content i-netam.pdf · 2013. 4. 15. · content editorial 3 policy paper the evolution of tax and...

Content

Editorial 3

Policy paperThe evolution of tax and benefit policy in Latvia: what has been the place 5of distributional considerations? AnnaZasova,MarijaKrūmiņa,OlgaRastrigina

ArticlesCompetitiveness of Latvia’s Exporters 17 KonstantinsBenkovskis

What managers think of capital structure and how they act: Evidence from 47Central and Eastern Europe PéterHernádi,MihályOrmos

Student debt levels and income of University of Latvia graduates: 73 Prospects for income-contingent loan repayment by the field of studies and gender AliAitSiMhamed,RitaKaša,ZaneCunska

FDI in the post-EU accession Baltic Sea Region: A Global or a regional concern? 89 RichardNakamura,MikaelOlsson,MikaelLonnborg

How integrated are the exchange markets of the Baltic Sea Region? 109An examination of market pressure and its contagion ScottW.Hegerty

PhD news 123

3

Editorial

With this issue we come to the end of the twelfth volume of the Journal. Looking back on 2012 we see another year with a growing number of submissions. Another indicator of the development of the Journal is the geographical dispersion of authors. As seen from the cur-rent issue the number of authors with an affiliation outside the Baltic countries has increased.

Since this is the last issue of the current volume, we would like to take the opportunity to thank our referees for their contributions during the year – contributions that have been pivotal to the development of the Journal.

Altogether this bodes well for the thirteenth volume and we take the opportunity to re-iterate our invitation for submissions that, to quote the scope of the Journal, are “theoretical or empirical in their emphasis and with relevance to the Baltic countries”.

ZaneCunska AndersPaalzow AlfVanags

5

The evolution of tax and benefit policy in Latvia: what has been the place of distributional considerations?

Anna Zasova1,MarijaKrūmiņa2,OlgaRastrigina3

IntroductionLatvia has one of the highest levels of income inequality and poverty in the EU. During the boom, income inequality was rising and in 2007 was the highest in the EU. During the 2008-2009 recession, when Latvia experienced the world’s largest GDP contraction and a mass in-crease in unemployment, poverty, as measured by the at-risk-of poverty rate, declined as the incomes of the top quintiles fell more strongly than those of the relatively poor. The Latvian government has sometimes been applauded for this ‘achievement’ (e.g. Krastiņš, 2012). Yet the reduction in poverty is a result of the way in which this poverty indicator is measured: the at–risk-of-poverty rate shows the share of population receiving income below a certain pro-portion of the median income. Thus, the poverty rate can decline even with declining incomes if at the same time income distribution becomes more even, and vice versa – if incomes are growing unevenly, the poverty rate can increase. A more concrete indicator to assess the change in population welfare in the periods of rapidly changing average incomes is the share of population classified as severely materially deprived, which in Latvia declined from 38.9% in 2004 to 19.0% in 2007 and then increased to 30.9% in 20104 (compared to 18.5% in Lithu-ania, 8.7% in Estonia and 8.8% for the EU average in 2010).

In Latvia income inequality and poverty have become topics of active public debate after the publication of investigations undertaken by the Baltic Center for Investigative Journalism Re:Baltica –“The Hidden Side of Latvia’s ‘Success’ Story” (Sprinģe, 2012) and “The Invis-ible Side of Latvia’s ‘Success’ Story: Life with ‘God’s Mercy and the Goodness of Others’” (Rizga, 2012). As argued by Spriņģe (2012), not only is Latvia generally characterised by a high level of income inequality, but what should be a main virtue of the benefit system is distorted in Latvia. Thus, in contrast to other developed countries “where social security means assistance for people in difficult circumstances and support for the needy”, the Latvian benefit system allows for large benefits to wealthy households but often provides only meagre benefits to those who are really in need. Spriņģe identifies cases where the size of monthly child-related benefits can be measured in thousands of euros – in one particular case exceed-ing 16 thsd EUR per month. Moreover, this benefit was paid out in 2011, when the govern-ment had already introduced a “sliding” ceiling on the size of benefits, which stipulated that above a certain threshold only 50% of the benefit, to which a person is notionally entitled, is paid out. 1 Research fellow at Baltic International Centre for Economic Policy Studies, [email protected] 2 Research fellow at Baltic International Centre for Economic Policy Studies, [email protected] 3 Senior research officer at the Institute for Social & Economic Research (University of Essex), research associate at Baltic International Centre for Economic Policy Studies, [email protected] 4 Severe material deprivation is defined as inability to afford at least four of the following items: to pay rent, mortgage or utility bills; to keep the home adequately warm; to face unexpected expenses; to eat meat or proteins regularly; to go on holiday; a television set; a washing machine; a car; a telephone (Eurostat, Glossary) http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Glossary:Material_deprivation_rate

TheevolutionoftaxandbenefitpolicyinLatvia:whathasbeentheplaceofdistributionalconsiderations?

6

BalticJournalofEconomics12(2)(2012)5-16

The Baltic International Centre for Economic Policy Studies (BICEPS) participated in the Re:Baltica project by estimating the distributional impact of some of recent tax and ben-efit reforms. In what follows we provide some of the results of these calculations that were performed using the Latvian part of EUROMOD tax-benefit microsimulation model5. In ad-dition, here we present the results of an estimate of the distributional impact of austerity measures that were implemented in 2008-2012 and show that the impact was progressive, in the sense that higher income groups contributed relatively more to budget consolidation. At the same time, we also show that the distributional impact of the most recent reforms in personal income tax, which were designed and implemented after the end of the programme with international lenders, is likely to be regressive, despite the fact that the government had earlier committed to different reforms which would have had a more progressive impact. We argue that distributional considerations, especially assistance for those really in need, have had a low priority in designing tax and benefit policy in Latvia.

Increase in income inequality during boom years“Fairness in relation to taxation has a long history in the economics literature. One of the basic principles of taxation since the publication of Adam Smith’s WealthofNationsin 1776, has been the individual’s ability to pay, i.e. taxes raised to finance state expenditures should be related to each individual’s ability to pay. Taxation according to the ability to pay has since gained more or less unanimous acceptance as an equity norm for tax design. The concept is however too general for concrete policy purposes since application of the ability to pay principle requires, …, specifying exactly how much each individual should be asked to con-tribute.” (Vanags, 2010).

Both income inequality and poverty are influenced to a large extent by political decisions of the government mainly through treatment of tax policy using its redistributive function: coun-tries with more progressive tax systems that redistribute more income have lower poverty rates (for example Denmark and Austria). Countries with lower redistribution have higher numbers of people within the low-income group (U.S., Latvia).

“Despite the accelerated public increase in public spending, prior to the crisis in 2007, Latvia had one of the lowest reductions in inequality due to government taxes and spending when compared to high-income OECD countries. The Gini coefficient for household income was 0.44 before and 0.35 after government taxes and transfers” (World Bank, 2010). In OECD on average the reduction in Gini is nearly 0.15 points. Therefore, disadvantages of highly redis-tributive tax-benefit systems (low work incentives, high administrative costs) are not likely to outweigh the potential gain from a reduction in inequality.

Income inequality in Latvia was increasing in the boom years. During that time, the top end of society and the middle class benefited more than the poor, as there are no statutory index-ation rules for social transfers except for pensions and the growth of wages outpaced that of benefits. In 2007 inequality in Latvia was the highest among EU member states. The only year when income inequality slightly declined was 2006 (see Figure 1).

5 EUROMOD is a tax-benefit microsimulation model for the European Union (EU) developed by the core developer team based mainly at ISER, University of Essex, and financially supported by the European Commission DG-EMPL. For more information on the EUROMOD, see: https://www.iser.essex.ac.uk/euromod

7

Figure 1. S80/S20 income quintile share ratio in Latvia in 2004-2010

2010200920082007200620052004

5.0

8.0

9.0

7.0

6.0

4.0

3.0

0

1.0

2.0

Source: Eurostat

Tax experts in Latvia have noted that the country has had very conservative tax policies and prior to 2010 the established system was beneficial for wealthy individuals and businesses but imposed a relatively large burden on employment income. Furthermore, Eurostat data show that the greatest share of Latvian government revenue was collected through indirect taxes, primarily sales taxes (value added tax (VAT)) that are added to the cost of food, clothing, or medicines – items people with low incomes spend most of their money on (Eurostat, 2012). In countries with more progressive tax systems, governments collect more from people with higher incomes through direct taxes on things like real estate, capital gains, dividends and interest income. In Latvia before the crisis, capital gains and other sources of capital income were not subject to personal income tax. Attempts to introduce a progressive tax system in Latvia have so far failed and suggestions to reduce the tax burden on labour and to increase the real estate tax that appeared as early as 2000 as well as proposals to tax dividends were ignored by politicians in the pre-crisis period.

The personal income tax rate in Latvia is flat and has remained unchanged at 25% for many years. Some progressivity of personal income tax is ensured by a non-taxable income al-lowance, which, however, is quite low: in 2007 it was 50 LVL (71 EUR) per month and the allowance for a dependent was 35 LVL (50 EUR) per month. Until 2010, dividends (as well as interest income and income from capital gains) were not taxed, which created strong in-centives to receive compensation in the form of tax-free dividends rather than salary. For ex-ample, as shown by Spriņģe (2012), in the last ten years Latvian oligarch “Lembergs earned 14.8 million LVL from dividends and [interest on] bank deposits. If the dividend tax which was introduced during the crisis had been introduced ten years earlier, Lembergs would have had to pay nearly 1.5 million lats (2.1 million euros) in tax”. It was also widespread to take advantage of the real estate business as the land tax was rather low and the law allowed ways of avoiding a range of other taxes (Spriņģe, 2012).

The pre-crisis period was characterized by an overheated labour market that stimulated strong growth of wages in both public and private sectors, while pensions and other benefits were growing more slowly. However, 2006 was a year of parliamentary elections which brought increases in social transfers (the main examples being introduction of additional payment to


8

the childbirth allowance, permission for those in receipt of child-care benefit to work, a rise in the amount of the state social security benefit, which correspondingly increased the mini-mum guaranteed level of the old age pension) and as a result there was a slight reduction in inequality (see Figure 2).

Figure 2. Growth of household disposable income main components (per household mem-ber), % year-on-year

2010

Disposable incomeIncome from employmentSocial transfers

20092008200720062005

20

50

60

40

30

10

0

-30

-20

-10

Source: Central Statistical Bureau

The increase in child related benefits was a major development in the field of family policy during the pre-crisis years. Until 2005 child-care benefit was independent of individual social security payments and was established at 90% of the minimum wage for persons with chil-dren below 1.5 years, and at 70% for those with children between 1.5 and 3 years.

As of 2005, this benefit was made earnings related: it was equal to 70% of the individual’s average salary on which social insurance payments had been paid, but there were limits on the size of the benefit – it could not be less than 56 LVL and could not exceed 392 LVL (payable to either one of the parents). For socially uninsured persons child-related benefit amounted to 50 LVL per month. Initially child-care benefit was not compatible with employment, i.e., a person could only receive the benefit if s(he) was not working. From March, 2006, this restriction was removed. As of 2008, child care benefit for socially insured persons was re-placed by an earnings-related parental benefit, which had a minimum level, but the ceiling on benefit size was removed. Thus, the benefit was very much larger for families with higher incomes than for poorer households.

Other child related benefits were also increased in the pre-crisis period: before 2006, parents received a one-off childbirth allowance of 296 LVL. Starting from 2006 additional payments of 100 LVL for the first child, 150 LVL for the second child and 200 LVL for the third and each additional child were introduced.

There is also a state family benefit which is not means tested: a flat monthly amount of 8 LVL is provided for the first child, and until the crisis the amounts for subsequent children were larger - 9.60 LVL for the second, 12.80 LVL for the third child, and 14.40 LVL for the fourth and each subsequent child.


9

As argued by Spriņģe (2012) “(t)he Latvian benefits system overall works largely in favor of the wealthy, allowing them to receive huge benefits, while people on small incomes receive the minimum benefit”.

The crisis years: a reduction in inequalityEarly in 2007, the government adopted an anti-inflation plan, which had as its main short-term goals: balancing fiscal policy, reducing speculative activities in real estate market and limiting expansion of consumer credit. Adoption of the plan and moderation of credit growth by the main banks triggered a gradual slowdown in the economy in 2007. By mid-2008, GDP year-on-year growth approached 0% (from growth in excess of 10% at the beginning of 2007), private consumption was down by 3.2%, imports plummeted by 7.9%. The fiscal balance quickly deteriorated as a result of modest tax revenues, and in the 3rd quarter of 2008 the situation was further exacerbated by the government’s decision to nationalise the second largest commercial bank, Parex bank, which had faced a deposit run and had been unable to finance its syndicated loans. The deterioration in public finances is illustrated in Figure 3 where it can be seen that by mid-2008 revenue growth had declined to less than 10% (down from almost 30% a year earlier), while expenditure growth remained as high as 20%.

Figure 3. Year-on-year growth of general government budget total revenues, tax revenues and expenditures, %; seasonally adjusted budget balance, % of GDP

Budget balance ExpendituresTotal revenues Tax revenues

2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4

20.0

30.0

10.0

0.0

-30.0

-40.0

-20.0

-10.0

Source: Eurostat, authors’ calculations

Faced with a rapidly deteriorating fiscal position and the situation in the global financial markets, the Latvian government sought financial assistance from international lenders. After negotiations in the autumn of 2008 Latvia was provided a 7.5 billion euro (about 1/3 of GDP) bailout facility from the IMF, the European Commission, the World Bank and the Nordic countries. The Latvian authorities took a firm position to proceed with the adjustment under a fixed exchange rate, despite estimates by lenders which suggested that this might result in a more protracted recession than in a scenario of widening exchange rate bands to the 15% range permitted by ERM2 (IMF, 2009a). As a consequence, the government had to accept especially strict and wide-ranging budget consolidation measures. The total size of consolidation implemented during the programme was impressive: accord-ing to the Ministry of Finance (Ministry of Finance, 2012) the total size of budget consolida-tion measures implemented in 2008-2011 amounted to 16.6% of GDP, of which most (10% of GDP) consisted of expenditure cuts (see Figure 4). Under pressure from international lenders,


10

consolidation was front-loaded, in 2009 reaching almost 10% of GDP. Although lenders were closely involved in controlling the structural reforms implemented, the exact mix of measures was proposed by the Latvian authorities.

Figure 4: Size of implemented consolidation measures and budget deficit outturn, % of GDP

Revenue measures Expenditure measures Budget deficit2008 2009 2010 2011

10

12

8

6

0

2

4

Source: Ministry of Finance, Eurostat

The 2009 consolidation mainly consisted of expenditure measures, of which a major share consisted of cuts in the government wage bill, including both a reduction in the number of civil servants and wage cuts. On the revenue side, the government stuck to the goal of shifting the tax burden from labour to consumption, thus a major measure was an increase in the VAT rate, while at the same time the personal income tax rate for employees was reduced from 25% to 23% (for the self-employed it remained unchanged at 15%) and the non-taxable mini-mum and exemptions for dependents were raised. Another revenue-side measure introduced in 2009 was removal of ceilings on social security contributions. This measure was aimed at easing pressure on the social budget. However, it opened possibilities for abuse of contribu-tory benefits, and, as shown by Spriņģe (2012) these possibilities were indeed exploited to obtain very high benefits.

However, when designing the 2009 supplementary budget, it became obvious that the eco-nomic forecasts underlying the 2009 budget were overly optimistic (GDP growth forecast 2.0%, unemployment 7.7% (Ministry of Finance (2008)), while in practice in the second quarter of 2009 GDP was down 17.9% year-on-year and the LFS-based unemployment rate reached 16.7%, with the implication that the reduction in labour taxes could not be sustained. As a result, the non-taxable allowance was more than halved (from 90LVL to 35LVL per month). As an additional measure to reduce budget outlays, the government attempted to cut pension expenditures and in June 2009 passed amendments to the Law on State Pensions, which stipulated that old-age pensions should be cut by 10% and pensions to working pen-sioners by 70%. This decision was strongly opposed by the public and on December 21, 2009, the Constitutional Court ruled that the government’s decision was unconstitutional arguing that the state must guarantee peoples’ right to social security. Moreover, from July 2009 and to the end of 2013, old-age pensions are not to be indexed.

In the 2010 budget, the personal income tax rate reduction was reversed and the rate was raised even above the pre-crisis level (to 26% for both employees and the self-employed,


11

while for the latter the rate was previously 15%). Other major revenue measures included broadening the personal income tax base to include capital gains, dividends and interest and an increase in both the rate and the tax base of real estate tax. Although these reforms had been called for much earlier by international organisations (IMF, 2006) they had not been implemented by the government until pushed by severe revenue shortfalls. On the expendi-ture side, major cuts in the public wage bill continued. In addition, a number of benefit restric-tions were introduced in 2010. First, a “sliding” ceiling was applied to a number of contribu-tory benefits, including unemployment benefit, parental, maternity and paternity and sickness benefits. The ceiling stipulated that above a threshold of 11.51LVL per day, only 50% of the benefit, which a person is notionally entitled to, would be paid out. Yet, combined with aboli-tion of the ceiling on social security contributions implemented earlier, this still permitted very large benefits for high earners. Another measure with respect to parental benefit was that working parents were no longer able to receive benefit. Second, supplementary payments to state family benefit and child birth benefit were abolished. Third, the amounts of maternity and paternity benefits were cut from 100% to 80% of the average contribution wage.

Finally, the 2011 consolidation mainly comprised revenue measures, the most important of which was a further increase in the VAT rate and an increase in employee social security contributions from 9% to 11%.

In what follows we present estimates of the distributional impact of selected budget austerity measures as described above. The estimates for Latvia along with other 8 EU member states have been done using the EUROMOD micro-simulation tax-benefit model (Avram et al, 2013). The measures that are taken into account in the estimation include the old-age pension freeze, the increase in social security contributions and personal income tax rate, broadening of the personal income tax base and removal of ceilings on social security contributions, the reduction in tax exemptions, the cuts and ceilings on child-related benefits, the removal of eligibility for employed parents for parental benefits, and public pay cuts. The distributional impact is estimated by comparing two scenarios: (i) the baseline scenario - simulation of the 2012 tax-benefit policy system (with austerity measures implemented), and (ii) the counter-factual scenario – simulation of the tax-benefit policy system that would have emerged in 2012 in the absence of austerity measures.

There are some limitations to the analysis. First, the EUROMOD input data used in this es-timation exercise is based on European Union Statistics on Income and Living Conditions 2008 (with income data referring to 2007). The data are adjusted up to 2012 using updating factors based on the aggregate evolution of respective income categories according to na-tional statistics. Second, only the direct impact of the measures is modelled, not taking into account possible secondary effects such as the behavioural response of people to reforms implemented. Figure 5 shows the estimated impact on household income by household type and by decile group.

The results suggest that the impact of the austerity measures was progressive in the sense that relatively well-off households experienced a proportionately larger reduction of household income. It should be kept in mind though that the input data come from household survey data which do not normally fully capture individuals who are at the extreme tails of income distri-bution. The progressivity of the impact is likely to be a consequence of ceilings introduced on


12

contributory benefits and cuts in public wages. The effect of cuts in contributory child-related benefits is also reflected in the fact that for households with children the progressivity of the impact is stronger than on average for the whole population.

Figure 5: Estimated impact of austerity measures on household income by household type and by decile group, %

1 2 3 4 5 6 7 8 9 10

-2

0

-4

-6

-12

-14

-16

-18

-10

-8

TotalHouseholds with childrenHouseholds with elderly

Source: Avram et al (2013)

The results presented above do not include the effect of the increase in VAT rates imple-mented as part of the crisis measures. This reform is likely to have an opposite distributional impact, i.e., poorer population groups, who spend a relatively large share of their income on consumption, are expected to be more sensitive to these changes. The effect of a change in VAT is hard to estimate, firstly because it requires information about the pass-through of a VAT change into prices. What we present below is an estimation assuming a 100% pass-through, which might be a reasonable assumption given that the change in VAT in Latvia was broad-based, in which case the pass-through tends to be higher (Institute for Fiscal Studies et al, 2011). Moreover, since our main purpose is to evaluate the distributional impact of the reform, the exact extent of the pass-through is not crucial: what is important is that the degree of the pass-through is equal across different groups of goods and services. We base our calcu-lations on data about composition and structure of consumption expenditure by quintile from CSB published aggregate statistics from Household Budget Survey. We compare two VAT regimes – the 2011 rate and the 2008 rate, thus our estimation covers two VAT increases and we assume an unchanged consumption structure. We do not model the impact of the reduction in the VAT rate implemented in mid-2012 since we are looking only at budget consolidation measures implemented in response to the crisis. Figure 6 illustrates the calculated increase in the share of disposable income spent on VAT by income quintiles.

Our results suggest that the impact of the rise in VAT was, as might be expected, regressive: the share of income spent on VAT in lower quintiles increased more strongly than in the higher quintiles, thus increasing income inequality.


13

Figure 6: Estimated increase in the share of disposable income spent on VAT due to an in-crease in VAT rates in 2011 vs. 2008 by income quintiles, percentage points

5

4

3

2

6

7

0

1

1 2 3 4 5

Source: authors’ calculations

The dynamics of the S80/S20 income quintile ratio suggests that overall income inequality in Latvia declined in the crisis years. According to Eurostat data, the S80/S20 income quintile ratio declined from 7.3 in 2007 (the highest ratio in the EU at that time) to 6.6 in 2010. In part, the reduction was due to the overall impact of the crisis. As shown by Bičevska (2012), higher wages fell more strongly than lower wages. Additionally, the employed population was more affected than pensioners as a result of the fall in wages (while old-age pensions were frozen) and of the increase in unemployment. However, the poverty rate in absolute terms, as mea-sured by the share of severely materially deprived households, remains among the highest in Europe and increased strongly during the crisis from 19.0% in 2007 to 30.9% in 2010 (compared with 18.5% in Lithuania, 8.7% in Estonia and 8.8% in the EU on average in 2010). The mix of austerity measures implemented in response to the crisis was designed under considerable scrutiny and was generally supported by the international lenders. However, today some of the most recent reforms implemented by the government are leading in quite the opposite direction from the path supported by the lenders.

End of the programme with international lenders: distributional impact of newly pro-posed reforms

In December 2011, Latvia completed the programme with the international lenders, but con-tinues a close policy dialogue with the IMF under post-programme monitoring. While posi-tively assessing the progress achieved by Latvia in recent years and the return to economic growth, the IMF emphasizes that “(t)he 2012 budget, which aims for a deficit below 2.5 percent of GDP, demonstrates the authorities’ commitment to fiscal discipline and to meeting the Maastricht criteria. However, alternative measures such as higher real estate tax, progres-sive personal income tax, and improved targeting of social benefits, might have proved less distortionary and facilitated a more sustainable adjustment. Continued high unemployment and poverty rates make it important to maintain a strong social safety net.” (IMF, 2011).

In 2009, the IMF expressed concerns about the reduction implemented in the non-taxable income allowance, arguing that the impact of this reform is regressive (IMF, 2009b). The IMF generally supports “the goal of cutting labour taxes to stimulate employment but, given high


14

marginal tax rates for low-income workers and to promote inclusion, its long-standing recom-mendations include: (i) raising the tax-free threshold rather than cutting the headline rate; (ii) introducing a progressive personal income tax; and (iii) introducing tax credits for net new hires, especially the long-term unemployed” (IMF, 2012). The tax rate for low wage earners in Latvia is one of the highest in the EU: the tax wedge for a single worker at two-thirds of average earnings in Latvia in 2010 was 41.5%, vs. 36.0% on average in the EU-27, 38.6% in Estonia and 38.9% in Lithuania (Eurostat, 2012). Nevertheless, in May 2012 the parlia-ment urgently passed amendments to the law on personal income tax, which foresee that the personal income tax rate will be reduced by 1 percentage point as of 2013 and then gradually reduced to 20% by 2015. The tax allowance for dependents will be raised from the current level of 70 LVL per month to 80LVL per month as of July 2013. While the reduction in the tax rate as such is welcome, it has precluded the government from raising the non-taxable mini-mum and from introducing a more generous increase in allowances for dependents, despite the previous intention to raise the non-taxable allowance from the current 45LVL to 90LVL and to raise the allowance for dependents to 100LVL.

Figure 7 illustrates the estimated distributional impact of the currently implemented decrease in the personal income tax rate vs. the previously considered plans to raise the non-taxable minimum to 90LVL and to raise the allowance for dependents to 100LVL. The results suggest that the latter reform would have had a much more progressive impact.

Figure 7: 2012 personal income tax reforms - impact on average household net income vs. baseline by quintiles, %

5.0

4.0

3.0

2.0

6.0

0.0

1.0

1 2 3 4 5

Alternative1/Baseline



Alternative1:Increaseinnon-taxableminimumandallowancesfordependentsAlternative2:Reductioninpersonalincometaxratefrom25%to24%(effectiveJan1,2013)Alternative3:Reductioninpersonalincometaxrateto20%(effectiveJan1,2015)Source: authors’ calculations using EUROMOD

Another major reform which will come into force in 2013 and which raises concerns about the impact on socially least protected groups in Latvia are changes in the amount of, and the provisions for, the Guaranteed Minimum Income (GMI). GMI is the basic social assistance benefit and the reform will reduce the minimum level of GMI that local governments have to provide from 45LVL a month to 35LVL. The government argues that this will increase the incentive for job search. Given the already tiny level of guaranteed minimum income an improvement in work incentives through cuts in the tax burden for low paid workers or via


15

provision of tax credits for new hires would be a more desirable option from the point of view of protecting the poorest population groups. Moreover, the other feature of the reform – the transfer of funding of GMI benefit from central government to the municipalities – is likely to cause “perverse incentives at the local level, and disparities in provision of these mandatory benefits across wealthy and poor municipalities. For this reason few EU and OECD countries retain decentralised financing of mandatory targeted benefits” (World Bank, 2010)6.

Concluding remarksDespite a period of reduction in the crisis years, income inequality in Latvia remains one of the highest in the EU. The poverty rate as measured by the share of severely materially deprived persons has increased strongly and is currently the second highest in the EU, much above the level observed in Estonia and Lithuania. This suggests that policies aimed at pro-tecting the most vulnerable population groups and reintegrating them into the labour market should remain a genuine priority.

The reforms implemented during the recession did result in higher income households con-tributing more to budget consolidation. However, many of the reforms implemented in re-sponse to the crisis had been proposed much earlier (e.g., making capital gains subject to income tax, an increase in the rate and the tax base of real estate tax, avoiding too rapid wage bill growth in the public sector), thus in a way these anti-crisis measures were simply correct-ing previous policy failures.

Although not directly involved in designing the exact mix of anti-crisis policies, the interna-tional lenders played a major role in defining the direction of the reforms and ensuring protec-tion of the poorest population groups. In some cases their role was decisive: for example, one of the reforms proposed by the government in 2009 but rejected by the lenders was to cut the non-taxable income allowance to 0 LVL (Eglītis, 2012). Some major reforms implemented by the government since completion of the programme with international lenders will have a regressive impact, despite earlier government commitments and lenders’ recommendations. To conclude, we argue that there is a general lack of concern about income distribution in de-signing tax and benefit policy in Latvia and that protection of those in need has a low priority.

6 World Bank (2010), “Latvia: From Exuberance to Prudence. A Public Expenditure Review of Government Adminis-tration and the Social Sectors,” September 27, 2010.


16

ReferencesAvram, Figari, Leventi, Levy, Navicke, Matsaganis, Militaru, Paulus, Rastrigina, Sutherland

(2013), “The distributional effects of fiscal consolidation in nine EU countries,” European Commission, Employment, Social Affairs and Inclusion, Research note 01/2012, Forth-coming 2013.

Bičevska (2012), “Kā krīze ietekmēja dažādu iedzīvotāju grupu pirktspēju,” http://www.makroekonomika.lv/ka-krize-ietekmeja-dazadu-iedzivotaju-grupu-pirktspeju

Eglītis (2012), “Nebija lētu triku”, Interview with Gatis Eglītis by Ir.lv, April 2012, http://www.ir.lv/2012/4/19/nebija-letu-triku

Eurostat (2012), Taxationtrendsin theEuropeanUnion.Datafor theEUMemberStates,IcelandandNorway, 2012 edition.

Eurostat. Glossary. Available at http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Glossary:Material_deprivation_rate

IMF (2012), IMF Country Report No. 12/171, International Monetary Fund, July 2012.IMF (2011), IMF Concludes Fifth and Final Review Under Stand-By Arrangement with Lat-

via, Press Release No. 11/481, December 21, 2011.IMF (2009a), IMF Country Report No. 09/3, International Monetary Fund, January 2009.IMF (2009b), IMF Country Report No. 09/297, International Monetary Fund, October 2009.IMF (2006), IMF Country Report No. 06/353, International Monetary Fund, October 2006Institute for Fiscal Studies et al (2011), “A retrospective evaluation of elements of the EU

VAT system”, Final report, London, 1 December 2011. Krastiņš (2012), “Ekonomists: Dombrovska laikā samazinājusies ienākumu nevienlīdzība un

nabadzība,” http://puaro.lv/lv/puaro/izradas-dombrovska-laika-samazinajusies-ienaku-mu-nevienlidziba-un-nabadziba

Ministry of Finance (2012), “Veiktā budžeta konsolidācija laika posmā no 2008. līdz 2011. gadam (faktu lapa),” http://www.fm.gov.lv/lv/aktualitates/jaunumi/27969-veikta-budzeta-konsolidacija-laika-posma-no-2008-2011-gadam-faktu-lapa

Ministry of Finance (2008), „Likumprojekta “Par valsts budžetu 2009. gadam” Paskaid-rojumi,” http://helios-web.saeima.lv/Likumdosana/Budzets_09/2009Budz_paskaidr/FMPask_A_081008.htm

Rizga (2012), “The Invisible Side of Latvia’s ‘Success’ Story: Life with ‘God’s Mercy and the Goodness of Others’”, http://www.rebaltica.lv/en/investigations/the_other_side_of_latvias_success_story_/a/826/the_invisible_side_of_latvia%E2%80%99s_%E2%80%98success%E2%80%99_story_life_with_%E2%80%98god%E2%80%99s_mercy_and_the_goodness_of_others%E2%80%99.html, Re:Baltica article, 17 October 2012.

Spriņģe (2012), “The Hidden Side of Latvia’s “Success” Story”, http://www.rebaltica.lv/en/investigations/the_other_side_of_latvias_success_story_/a/799/the_hidden_side_of_latvia%E2%80%99s_%E2%80%98success%E2%80%99_story.html, Re:Baltica article, 16 October 2012

Vanags (2010), “Tax reform in Latvia: Could it be fair?” SSE Riga/BICEPS Occasional paper No. 8

World Bank (2010), Latvia From Exuberance to Prudence: A Public Expenditure Review of Government Administration and the Social Sectors – Analytical Report, World Bank. https://openknowledge.worldbank.org/handle/10986/3009


17

Competitiveness of Latvia’s exporters

KonstantīnsBeņkovskis1, 2

AbstractThis paper evaluates the competitiveness of Latvia’s exporters from various aspects by using detailed trade data from UN Comtrade. Competitiveness represented by the market share of Latvia’s products in world trade was on a rising trend, growing almost two times between 1999 and 2010. This dynamic improvement was mainly accounted for by intensive margin, as Latvia’s exporters increased their presence on traditional markets. Moreover, the contribution of extensive margin was also positive due to geographical expansion. Analysis of non-price competitiveness signals that although Latvia’s export unit values were increasing faster than those of its main competitors, relative quality and taste for Latvia’s products were rising even faster, and, overall, the competitiveness of Latvia’s exporters improved.

Keywords: C43, F12, F14, L15JEL classification: exports, extensive margin, intensive margin, non-price competitiveness, Latvia

1. IntroductionDespite the scope of discussion and empirical work on the topic, the concept of competitive-ness is still elusive. The definition of competitiveness is so broad that it includes an extremely large set of macroeconomic and microeconomic issues: per capita income levels, perfor-mance of institutions, levels of productivity, comparative costs, and many others not men-tioned here. As a result, the number of ways a researcher can evaluate the competitiveness of a country is vast. This paper is restricted to only a few approaches, which can be applied to highly disaggregated trade data. Thus we are narrowing the definition of competitiveness to the one given by the OECD: “Competitiveness is a measure of a country’s advantage or disadvantage in selling its products in international markets”3, and concentrate on the perfor-mance of Latvia’s exporters.

The motivation for focusing on export activities is obvious: Latvia is a very open and ex-tremely small economy, where exports are the main source of economic growth in the long run. Our research is by no means a unique attempt to discuss the competitiveness of Latvia’s exports. However, some empirical papers are already outdated and observe years before ac-cession to the EU (e.g. Dulleck et al., 2005, or Fabrizio et al., 2007), some relate competitive-ness issues mostly to effective exchange rates (e.g. Purfield and Rosenberg, 2010), some do not cover all Latvia’s exports (Benkovskis and Wörz, 2012). Therefore, the need arises to update the assessment of competitiveness and to broaden the set of available indicators.

1 Monetary Policy Department, Bank of Latvia, Kr.Valdemara iela 2A, Riga, LV-1050, Latvia, e-mail to: [email protected] The views expressed in this publication are those of the author, an employee of the Bank of Latvia Monetary Policy Department. The author assumes responsibility for any errors or omissions.3 See OECD Glossary of Statistical Terms: http://stats.oecd.org/glossary/detail.asp?ID=399

CompetitivenessofLatvia’sexporters

18 BalticJournalofEconomics12(2)(2012)17-45

Are Latvia’s products gaining export market shares? Are exporters doing it by expanding in new markets or are they intensifying their presence in traditional markets? Who are the main competitors for Latvia’s enterprises in external markets? Is the real effective exchange rate a complete measure of competitiveness? Can we assess non-price competitiveness of Latvia’s enterprises? Our paper tries to address these questions. Detailed trade data from UN Com-trade allow us not to restrict analysis to some specific geographical area or subset of products, while disaggregation enables tracking the performance of separate sectors and to take into account structural differences. The important contribution to existing empirical literature is the decomposition of changes in export market shares into intensive and extensive margins. This paper modifies the methodology of Hummels and Klenow (2005) for dynamic analysis. We also evaluate non-price competitiveness of Latvia’s total exports using UN Comtrade data. This analysis is performed using methodology recently developed by Benkovskis and Wörz (2012).

The next section illustrates the data, which are extracted from UN Comtrade. Section 3 then focuses on value data, decomposing market share changes into extensive and intensive mar-gins as well as presenting a geographical breakdown of Latvia’s main competitors. Section 4 uses information on trade volumes and prices, briefly describes the methodology behind the evaluation of price and non-price competitiveness at a highly disaggregated level, and pres-ents the empirical results. The last section concludes.

2. Description of databaseFor empirical analysis we use trade data from UN Comtrade. The main reason for this choice of data source is its almost full country coverage. Although the data in UN Comtrade have a lower level of disaggregation and a longer publication lag in comparison with Eurostat Comext, the world-wide coverage of the UN database is a significant advantage, for a view on Latvia’s exports would not be complete without such important trade partners as Russia or Belarus. Moreover, despite the current low shares of such countries as China, India and Brazil in Latvia’s exports, these markets are huge, dynamically growing, and have significant potential for Latvia’s products. Comext contains detailed data on Latvia’s exports outside the EU, but only UN Comtrade can give information on the product and partner structure of non-EU markets.

UN Comtrade provides a reasonably good disaggregation of export and import flows, and we are using the most detailed available, i.e. at the six-digit level of the HS (Harmonised System, 1996), which includes 5 132 different products. As mentioned above, this level of disaggre-gation is lower than provided by Eurostat (more than 10 000 products) but is still reasonably high to calculate unit values.

Notwithstanding our final goal to evaluate the competitiveness of Latvia’s exports, this pa-per achieves this by using import data of partner countries in several cases. The reason for focusing on imports from Latvia rather than on Latvia’s exports is driven by the theoretical framework underlining evaluation of price and non-price competitiveness. The methodology used in section 4 is based on the consumer’s utility maximisation problem. Import data are clearly preferable in this case, as imports are reported in CIF (cost, insurance, freight) prices and include transportation costs to the importer’s border; therefore, import data provide a bet-ter comparison of prices from the consumer’s point of view. On the other hand, use of import

19

data implies some drawbacks. Obviously, data on imports from Latvia do not fully coincide with Latvia’s export data due to differences in valuation, timing, sources of information, and incentives to report. The problem can be more severe for intra-EU trade, as measurement of trade in goods within the EU relies on VAT reports. This creates a greater incentive for report-ing export activities, which are subject to VAT returns.4 For that reason we still use Latvia’s export data where possible, namely when calculating extensive and intensive margins as well as determining the structure of Latvia’s exports for computations of competitors’ double-weights and aggregated adjusted relative export price index.

The import dataset contains annual data on imports of 75 countries at the six-digit HS level between 1999 and 2010.5 The list of reporters (importers) can be found in the Appendix, Table A1. By collecting data on imports of the abovementioned 75 countries we are covering more than 96% of world imports in 2010. Several importer countries (the United Arab Emir-ates, Vietnam, Egypt and Kazakhstan) were not included in the dataset due to lack of detailed data or missing information for 2010. To avoid calculation burdens, we restrict the list of part-ners (exporters) to 75 countries as well. The list of exporters can also be found in Table A1 (note that the list of exporters does not fully coincide with the list of importers). These 75 most important exporter countries cover around 93% of world imports in 2010; therefore, our database is a representative reflection of world trade flows.

We use unit value indices (dollars per kg) as a proxy for import prices and trade volumes (in kg) as a proxy for imported quantities. If data for either values or volumes are missing or data on volumes are not observed directly and are estimated by statistical authorities, no unit value index can be calculated. Unfortunately, the possibility to estimate unit values is relatively scarce for many reporting countries. Even the import database of the US, the major world importer, allows for calculating unit values only for approximately 70% of imports in 2010 (in value terms). The situation is much better for EU countries, China and Japan, but there are countries (e.g. Canada, Mexico and Australia) where the coverage is around 50% or even less. In addition, coverage is usually worse for the first half of the sample period. This problem makes analysis of non-price competitiveness more challenging, and the results of this study should be treated with a pinch of salt. However, the low coverage of available unit values in several countries is rather homogenous across different products and we can argue that this problem should not bias our results significantly. Another adjustment to the database relates to structural changes within categories of goods. Although we use the most detailed classification available, it is still possible that sometimes we are comparing apples and or-anges within one particular category. One indication of this problem is the large price level differences within a product code. Consequently, all observations with outlying unit value indices were excluded from the database.6

4 An extreme case of this problem is a VAT missing trader intra-Community fraud, which was not captured in import data and significantly overstated the UK trade balance in 2001-2002 (see Ruffles et al., 2003).5 For some countries data are not available for several years at the beginning or middle of the sample period: import data for South Africa, the Philippines, Oman and Tunisia are not available for 1999, Ukraine and Ethiopia – for 1999-2000, Malaysia, Bahrain and the Dominican Republic – for 1999-2001, Pakistan and Bosnia Herzegovina – for 1999-2002, Serbia – for 1999-2004, Sri Lanka – for 2000, Panama – for 2004, Nigeria – for 2004-2005.6 An observation is treated as an outlier if the absolute difference between the unit value and the median unit value of the product category in the particular year exceeds four median absolute deviations. The exclusion of outliers does not significantly reduce the coverage of the database. In the majority of cases only less than 2% of total import value was treated as an outlier.



As to Latvia’s export dataset, which is mainly used for analysis of extensive and intensive margins, this contains annual data on exports to the abovementioned 75 importer countries (actually 74, as obviously there are no data on Latvia’s exports to Latvia) at the six-digit HS level between 1999 and 2010. In the case of the export dataset we restrict ourselves only to value data, which is enough to calculate market shares and export structure.

Overall, the import and export database gives information about 379 768 potential markets for Latvia’s products (5 132 products times 74 importing countries), which can be used in a detailed analysis of Latvia’s competitiveness. At the beginning, analysis is restricted to value data for calculating extensive and intensive margins as well as for a description of Latvia’s main competitors, while later we will use also volume and unit value data.

3. Market shares and competitors3.1. Extensive and intensive margins of tradeTrade theories suggest that there are different ways by which a country can increase its ex-ports and market share in world trade. Models that follow Armington (1969) and assume an unchanged set of export products and destinations stress the intensive margin or exported quantity on a single market. The only way to increase exports in this model is to increase the average exported quantity in each market without altering the set of markets. On the other hand, monopolistic competition models (like the one developed by Krugman, 1979, 1980), allow for changes in the number of exported varieties. These models put emphasis on the role of extensive margin and state that exports can be enlarged by accessing new markets (in a geographical or product variety sense).

There is a considerable debate in empirical economic literature about the relative role of extensive and intensive margins in trade. Some authors state that the extensive margin is prevailing (e.g. Hummels and Klenow, 2005, who report that the extensive margin accounts for 62% of export increases in larger economies) while others find the intensive margin con-tributing more (e.g. Amiti and Freund, 2010, who conclude that China’s export growth was mainly accounted for by a notable growth in exports of existing products). These debates are important both from theoretical and practical points of view, as the dominance of one margin dictates the choice of modelling framework, underpins divergent predictions about the terms-of-trade effect of export expansion, and alters conclusions about consumer welfare gains. In this paper, however, it enables an answer to the question whether growth of export market shares and competitiveness was mainly driven by increasing diversification of export prod-ucts and/or destinations or whether producers were able to gain competitiveness in traditional markets.

One of the most popular ways to measure the extensive margin is by counting the number of products that a country exports (e.g. see Dennis and Shepherd, 2007). This measure is simple, intuitive and consistent with theoretical concepts. In a similar way, one can compute the number of markets (a specific product exported to a specific country) and the average number of countries to which one product is exported. Table 1 reports these calculations for Latvia’s exports.

21

Table 1. Number of markets, products and importers per product1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Markets 8959 9550 10055 11035 11686 13412 18968 20472 20827 21033 22593 24905

Products 2638 2674 2747 2817 2854 3065 3377 3490 3416 3462 3562 3610

Importers per product 3.4 3.6 3.7 3.9 4.1 4.4 5.6 5.9 6.1 6.1 6.3 6.9

Source: UNComtrade, author’s calculations.

The number of markets where Latvian enterprises are present increased almost three times between 1999 and 2010. This leads to the conclusion that the extensive margin was an impor-tant factor behind Latvia’s export growth. The growing number of Latvia’s export markets was partly due to an increasing set of exported goods, while the main driver was significant enlargement of geographical diversification: in 2010 one product was on average exported to 6.9 countries in comparison with only 3.4 countries in 1999.

Although the measures presented in Table 1 are informative and simple, they do not shed light on the role of the intensive margin and do not allow comparison with the contribution of the extensive margin. Several papers propose ways to decompose growth in trade (see e.g. Fel-bermayr and Kohler, 2006, or Besedes and Prusa, 2011). Our goal, however, is export com-petitiveness which is usually associated with market share. Therefore, we need to decompose the export market share, which is a more complicated task. Hummels and Klenow (2005) proposed methodology to decompose relative exports (and the export market share) into ex-tensive and intensive margins. However, their methodology is developed to compare differ-ent exporters at one point in time,7 while we are interested in a dynamic analysis of Latvia’s competitiveness.

This paper proposes disaggregation of changes in export market share (MSt) into three parts instead of two. Besides variations in the intensive (IMt) and extensive (EMt) margins, a shift in demand structure may also affect changes in market share. The reasoning for this decom-position is twofold. First, as changes in market share depend also on changes in world im-ports, we need to include a demand factor in the analysis. Second, our decomposition gives an opportunity to distinguish between the endogenous and exogenous components of market share changes. While the extensive and intensive margins are affected by behaviour of ex-porting firms, shifts in the demand structure are exogenous with respect to exporters at least in the medium term. Changes in market share can be expressed as

(1)

where Xig,t is Latvia’s nominal exports of good g to country i at time t, Mig,t is total nominal imports of good g by country i at period t, I is the set of importing countries, G is the set of products in world trade.7 Dynamic analysis of margins evaluated by the methodology of Hummels and Klenow (2005) will lead to incorrect conclusions. As the intensive margin is evaluated using a set of non-zero export categories in the current period, com-parison of intensive margins at different points in time will also include shifts in the product set, thus also accounting in part for changes in the extensive margin.



A crucial point of the analysis is the decision on distinction between intensive and extensive margins. The analysis can be done at the product level (as in Amiti and Freund, 2010), coun-try level (as in Felbermayr and Kohler, 2006) or country-product level (as in Besedes and Prusa, 2011). We follow the latter approach and define distinctions at the product-country level, which means that exporting an existing product to a new destination or a new product to an existing destination is also qualified as the extensive margin. This, together with the de-tailed 6-digit HS classification, obviously leads to a higher contribution of extensive margin to exports in comparison with alternative definitions.

Another important issue is the time dimension in definition of intensive and extensive mar-gins (see discussion in Besedes and Prusa, 2011). Here we follow the mainstream and exam-ine year-to-year survival of an exporter in a particular market. Exports to a new market are clearly classified as an extensive margin during the first year of appearance; however, if it survives further, it is reclassified in the intensive margin. In other words, the definition of ex-tensive margin is restricted to those markets in which no exports are observed either in period t–1 or in period t; all cases where Latvia’s exports are present in both periods are classified as an intensive margin. This definition will clearly decrease the contribution of the extensive margin, which should be kept in mind when interpreting the results.

Thus, changes in the intensive margin are calculated as follows:

(2)

where Gi,t,t-1 is the set of products exported by Latvia to country i in both periods. It is pos-sible that Latvia has no exports to some countries in several periods; in these cases Gi,t,t-1 is an empty set. Simply speaking, equation (2) calculates the changes in market shares in “old” or “traditional” markets. Following Amiti and Freund (2010), we define the contribution of extensive margin as follows:

(3)

This is similar to Feenstra’s (1994) index accounting for changes in import variety. Equation (3) compares the share of traditional markets in Latvia’s total exports in periods t–1 and t. If this share decreases over time, it means that the share of disappeared export markets was smaller than the share of new export markets, and the contribution of the extensive margin to changes in the export market share is positive. However, as mentioned by Amiti and Freund (2010), it should be kept in mind that Feenstra’s (1994) index reports the balance between new and disappearing markets and could somewhat understate the importance of new markets.

23

In order to fully decompose movements of the export market share, we need the following term, interpreted as changes in demand structure:

(4)

Equation (4) represents changes in the share of Latvia’s traditional markets in world trade. An increase in this share improves the total market share of Latvia’s exports, although it is problematic to qualify this effect as either an extensive or an intensive margin.

The extensive margin of exports in equation (3) can increase for two reasons: either producers start to export a new product or an existing export product is sold to a new country. To distin-guish between these two effects we further decompose the extensive margin into the product and geographical/importer dimensions (EMt

prod and EMtimp).

(5)

;

where Gt,t-1 is the set of products exported by Latvia in periods t–1 and t. The product dimen-sion of the extensive margin is again evaluated by Feenstra’s (1994) index, although now it focuses on the share of “traditional” products in total exports. If this share diminishes, the extensive margin improves due to the appearance of new exported products. The remaining part of the extensive margin is attributed to the importer dimension and includes establishing new geographical links by exporting existing goods to new countries.

Finally, similar decomposition is carried out for the demand structure, which makes it close in spirit to the constant market share analysis (see e.g. Richardson, 1971). The share of Latvia’s traditional markets in world imports can shift either due to changing demand for products or due to shifts in importer’s relative demand.

(6)

;



where DStprod is the product dimension and DSt

imp is the geographical/importer dimension of the demand structure effect.

Hummels and Klenow (2005) proposed decomposing the intensive margin further into price and volume effect, i.e. determining whether the share of exporters in traditional markets is growing due to more rapid price increases or due to larger physical volumes of exported pro-duction. Even though this information is potentially useful, analysis of price and volume data is left for the next section. This is determined by the abovementioned problem concerning availability of unit values and volume data, which will make the results incomparable with those for the total intensive margin.

Figure 1 presents the decomposition of Latvia’s export market share dynamics between 1999 and 2010.8 It shows that competitiveness, as indicated by the total world market share of Lat-via’s exporters was rapidly enhancing during the observation period. Except for two periods of marginal decrease (in 2000 and 2006), changes in market share were always positive and competitiveness almost doubled in ten years. As to the contribution of margins, growth in competitiveness was largely determined by the increasing intensive margin, although we also observe growing extensive margin of Latvia’s exports. At the same time, the results point to negative changes in the demand structure. Overall, we can conclude that Latvia’s producers are increasing their presence in old markets, while the falling share of Latvia’s traditional markets in world trade is compensated by the expansion of Latvia’s exporters into new mar-kets.

Figure 1. Extensive and intensive margin of Latvia’s exports

2005 20061999Market shareExtensive margin

Intensive marginDemand structure

2000 2001 20032002 2004 2007 2008 2009 2010

140

200

180

160

120

100

80

Source: UN Comtrade, author’s calculations.Notes: Calculated using equations (1)-(4); 1999=100.

Now we explore the way an increase in diversification of exports is achieved (see Figure 2a) and discover why the effect of the demand structure was negative (see Figure 2b). Estimates show that the major part of growing diversification is due to Latvia’s producers selling exist-8 The list of countries for which data are not available for several years has been given above. Fortunately, all these countries (except Ukraine) play a non-significant role in Latvia’s trade, so that the effect of missing years on the results is negligible.

25

ing export products to a new partner country. Thus, the geographical dimension of the exten-sive margin is prevailing. Exports of new products are also observed, although the intensity of this process is modest. Moreover, it was driven by one-off effect in 2005, which could be explained by EU accession and, to some extent, by changes in statistical methodology.9

Figure 2. Product and importer dimension a. extensive margin

2005

2006

1999

Product dimensionImporter dimensionExtensive margin

2000

2001

2003

2002

2004

2007

2008

2009

2010

110

125

130

120

115

105

100

95

90

110

70

80

75

105

100

95

85

2005

2006

1999

2000

2001

2003

2002

2004

2007

2008

2009

2010

90

Product dimensionImporter dimensionDemand structure

Source: UN Comtrade, author’s calculations.Notes: Calculated using equations (3)-(6); 1999=100.

9 Before May 1, 2004, foreign trade data were collected from customs declarations. Afterwards, data on trade with EU countries were collected by INTRASTAT monthly surveys. Therefore, changes between 2003 and 2005 may be driven by this change in the source of information.


b. demand structure


The small role of new products in increasing competitiveness contradicts the results presented in Table 1 and differs from the conclusions of Funke and Ruhwedel (2005), and Benkovskis and Rimgailaite (2011) who report a significant increase in product variety of Latvia’s ex-ports. In the case of Funke and Ruhwedel (2005), this is most likely determined by a different sample period (between 1993 and 2000 when the process of expanding the set of products should have been more intensive), and a different benchmark as well (product variety of Lat-via’s exports compared with US exports). Benkovskis and Rimgailaite (2011) in their turn use a different approach for assessing the extensive margin in the EU market, where variety was calculated relative to German exports while the importance of new products in total exports was not taken into account. A comparison with the results in Figure 2a may indicate that the share of products Latvia started to export recently in total exports is not very significant. Another possible explanation is the relatively lower disaggregation level of UN Comtrade, which leads to an underestimate of product set expansion.

Table 2. Market shares of Latvia’s exports by main product sector1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Wood and articles of wood 100.0 104.8 107.5 114.9 135.3 132.7 129.8 124.2 151.6 139.0 142.3 170.1

Base metals and articles of base metal 100.0 110.1 111.1 132.7 133.5 175.1 167.8 158.7 170.6 206.5 187.7 197.9

Machinery and mechanical appliances 100.0 104.3 133.6 159.0 193.1 254.3 316.4 334.5 451.2 562.0 605.0 573.4

Prepared foodstuffs 100.0 101.4 172.9 221.4 191.1 279.4 324.7 358.3 428.2 449.4 386.2 413.2

Chemical products 100.0 95.5 104.9 95.5 103.1 120.0 127.7 159.2 199.8 233.6 209.5 199.2

Vehicles and other transport equipment 100.0 111.5 150.9 165.8 201.4 342.2 550.8 841.0 1117.2 1324.0 1316.1 1168.2

Source: UNComtrade, author’s calculations.Notes: Calculated using equation (1); the six largest product sectors are chosen using 2010 export data for Latvia (cover 64.9% of Latvia’ exports in our database); 1999=100.

The geographical dimension is also prevailing over the demand structure effect. While the share of traditional products exported by Latvia in world imports remained roughly un-changed, the share of traditional geographical destinations of Latvia’s products decreased. On the one hand, this could be explained by Latvia’s geographical location. Although the closest neighbours Estonia and Lithuania experienced a rapid growth in imports, other important partners like Germany, Sweden and the UK did not increase their imports as rapidly as the developing countries of Asia. On the other hand, most of the effect is observed in 2000, while demand structure is almost unchanged afterwards.

Extensive and intensive margins can be calculated for separate product sectors; this is done in Table 2. A disaggregated view of export market shares and margins uncovers some interest-ing details. During the period observed, market shares increased for all major product sectors. The market share of vehicles improved more than ten times, machinery and mechanical ap-pliances grew more than five times, for food products the increase exceeded four times; we also observe positive and dynamic changes in market shares of wood, metals and chemicals. Analysis of extensive and intensive margins by sector of production (see Table A2 in Appen-dix) confirms dominance of the intensive margin in development of Latvia’s competitiveness, with all main export sectors showing strongly growing shares in traditional markets. How-ever, several sectors significantly expanded their export activities to new markets as well.

27

Overall, the story of Latvia’s exports is heterogeneous, and we can divide the main sectors into two broad groups. Exports of machinery, vehicles and food products showed the most impressive improvement in competitiveness, with both intensive and extensive margins being important. Latvia’s producers of machinery, vehicles and food were able to increase diversi-fication of their sales (mainly expanding the geographical dimension without losing product diversification, although exporters of vehicles were also able to increase their set of products by almost 15%) and at the same time to enhance their presence on traditional markets. A simi-lar development, although not as rapid, was observed for base metals. A different strategy was used by wood and chemical exporters. The wood sector is the only important export sector with almost unchanged diversification over the last 12 years. A lack of geographical and prod-uct expansion was compensated by a more intensive presence of Latvia in traditional markets for wood products. The same strategy was used by exporters of chemical products: changes in the extensive margin were small (albeit positive), while competitiveness was improved by growing presence in traditional markets.

3.2. Main competitors of Latvia’s exportersThe previous subsection gives some preliminary information about the performance of Lat-via’s exporters in external markets and evaluates competitiveness using the extensive and intensive margins. However, when we speak about competitiveness and competition, it is also useful to know the competitors so this section specifies countries from which the most important competitors for Latvia’s producers originate. Identifying the countries of origin of the main competitors is not a trivial task. If one wants to take into account both bilateral trade links and third-market competition, it is necessary to use a system of double weighting (see Durand, 1986). The method must take into account the relative importance of all competitors in each market, and the importance of each market for an exporter. Ideally, one also needs to have information on domestic producers in every market. This, however, is not possible due to data restrictions. In order to capture the importance of competitors from different countries for Latvia’s exporters, we define the double weights (Wc

comp):

(7)

where WMigc represents the share of imports from country c in total imports of good g by coun-

try i, while WXig shows the share of exports of good g to country i in Latvia’s total exports.

Therefore, double weights are calculated as the share of competitors in all 379 786 markets and weighted by the importance of those markets in Latvia’s exports.

Table 3 reports the top 15 countries whose firms were the most significant competitors for Latvia’s producers in 2010. It also shows how the weights of competitors evolved over time. According to our calculations for 2010, Latvia’s exporters face the most severe competition from Germany. This is a rather expected outcome, as Germany is the third largest world ex-porter and the largest exporter in Europe. Exporters coming from the biggest world exporter, China, form the second largest group of competitors for Latvia’s producers, but we can ex-pect more competition from this region in the future taking into account the rapid increase of China’s weight in comparison with 1999. The growing importance of China is naturally



explained by its rocketing export performance during the last ten years, while higher competi-tion with German firms is driven by the expansion of Latvia’s exports of machinery, vehicles, and chemical products. The third and fourth largest competitor groups for Latvia come from Poland and Russia, which can primarily be explained by geographical closeness and, to a lesser extent, by some similarities in export structure. Overall, the top 15 list of exporters is dominated by European countries, especially those from Northern Europe, again mainly on account of the geographical factor. A significant decline in importance is observed for competitors from Sweden (Latvia’s largest competitor back in 1999) mainly due to the dimin-ishing share of wood products in Latvia’s exports and the decreasing presence of Sweden’s producers in the wood products market.

Table 3. Double weights of Latvia’s competitors in 1999, 2004 and 20101999 2004 2010

Germany 7.1 9.6 11.6

China 2.1 3.9 5.7

Poland 3.4 3.9 5.4

Russia 5.0 6.0 4.8

Sweden 8.7 6.1 4.5

France 2.9 3.5 4.3

Finland 7.4 5.4 3.8

Netherlands 2.3 2.9 3.6

Italy 3.5 3.4 3.6

UK 2.7 2.2 3.6

US 3.2 2.3 2.6

Estonia 3.4 2.9 2.5

Lithuania 2.2 2.1 2.4

Belgium 1.7 1.9 2.1

Denmark 2.2 2.2 2.1

Source: UN Comtrade, author’s calculations.Notes: Calculated using equation (7); %.

Double weights of competitors in individual product sectors are evaluated in Table 4. These results corroborate our previous conclusion that German producers are Latvia’s main com-petitors in machinery, vehicles, and chemical products. Therefore, the importance of German competitors is increasing as Latvia is getting more similar to Germany in terms of export structure.

It should be noted that competition from German firms is also significant in Latvia’s other major export areas. Competition from China mainly focuses on machinery and mechani-cal appliances (not to forget textile products, where the share of competitors from China is almost 25%). The share of firms from Russia in machinery is negligible, while the presence of Russian competitors is very significant in wood products. Apart from Russia, significant competition in the wood product sector comes from Latvia’s northern neighbours Sweden, Finland and Estonia. Finally, in the food products market Latvia’s producers are competing with firms from France, the UK and, to a lesser extent, also Poland.

29

Table 4. Double weights of Latvia’s competitors by main product sectors in 2010Wood and articles of

wood

Base metals and articles

of base metal

Machinery and mechanical appliances

Prepared foodstuffs Chemical products

Vehicles and other transport

equipment

Share in Latvia’s exports 18.3 13.5 12.6 7.4 7.1 6.0

Germany 6.5 12.9 12.9 6.0 16.9 22.4

China 2.6 3.6 13.3 0.7 2.5 2.1

Poland 4.5 5.1 5.0 6.5 6.4 3.8

Russia 11.8 4.1 1.1 1.9 1.7 1.2

Sweden 9.3 3.0 3.8 3.1 2.5 3.3

France 1.3 3.6 2.6 12.8 7.8 8.0

Finland 6.9 2.2 4.5 2.3 2.7 2.0

Netherlands 1.5 2.8 3.5 4.1 5.1 2.6

Italy 0.9 5.2 4.5 5.3 3.7 4.8

UK 1.6 3.8 3.1 10.6 3.3 5.3

US 1.9 2.2 2.8 2.1 4.5 4.9

Estonia 5.6 1.0 0.8 1.5 1.6 1.4

Lithuania 2.3 2.5 0.9 2.9 1.3 0.7

Belgium 1.9 2.2 0.9 1.3 5.4 3.4

Denmark 1.5 1.3 1.5 1.8 1.9 3.4

Source: UN Comtrade, author’s calculations.Notes: The six largest product sectors are chosen using 2010 export data for Latvia (cover 64.9% of Latvia’s exports in our database); calculated using equation (7); %.

Last, but not least, we should remember the absence of information on domestic producers in every market, which definitely leads to biased estimations of weights. The results here some-what underestimate the competition coming from Lithuania, Estonia, Russia, Germany (to a smaller extent Sweden and Poland), as these are the main importers of Latvia’s products, and domestic producers obviously have strong positions in these markets.

4. Price and non-price competitiveness4.1. Traditional real effective exchange rate indicesThe real effective exchange rate is one of the most widely used tools in analysis of a coun-try’s competitiveness. It proxies relative changes in prices of a country’s exports by changes in nominal exchange rates and inflation differentials, which can be captured in various ways, leading in turn to different real exchange rate measures. The most popular indicator is based on inflation differentials as measured by the CPI due to data availability and comparabil-ity. Other popular definitions are PPI-based and ULC-based real effective exchange rates. Figure 3 reports CPI-based and ULC-based real effective exchange rates for Latvia. Both indicators show similar pictures with moderate changes in real effective exchange rate before 2005, a sharp increase in relative prices during the boom years of 2006-2008, and regaining of competitiveness after the financial crisis. At the end of the period observed, the real effec-tive exchange rate is 25-35% higher vis-à-vis 1999, which might be interpreted as a loss of price competitiveness. Such a simple interpretation of these indices, however, can be quite misleading for various reasons.



Figure 3. Real effective exchange rates for Latvia

CPI-based ULC-based

2005

2006

1999

2000

2001

2003

2002

2004

2007

2008

2009

2010

2011

140

170

180

160

150

130

120

110

100

90

80

Source: EurostatNote: 36 trading partners; 1999=100.

Traditional real effective exchange rates have several drawbacks related to approximation of export prices. The CPI-based index captures the dynamics of relative consumer prices. Domestic and export prices face different demand and supply conditions and can therefore differ greatly. Further, the CPI-based index includes changes in indirect taxes, which do not affect export activities directly. Although the PPI-based index is closer to the production side of the economy, it still includes production for the domestic market (data on export-oriented PPI are usually very scarce).

The ULC-based index has a similar drawback. Moreover, it usually refers to the total econo-my, also including the services sector. In addition, the ULC refers only to a part of production costs and ignores such important factors as profit margins. A solution to these shortcomings is to use the relative export price index, i.e. an indicator that is often used in macroeconomic models when explaining the dynamics of real exports. However, an aggregate export deflator still ignores one serious problem: the structure of exports differs across countries. Therefore, the need arises to conduct the analysis at the most disaggregated level to ensure that similar export products are compared for different countries.

In addition, real effective exchange rate indices measure only price competitiveness while ignoring non-price factors that affect the performance of exports. One such non-price fac-tor, emphasised by Flam and Helpman (1987), is related to vertical differentiation or quality of exported products. Another non-price factor is changes in consumer tastes, which can be driven by such subjective factors as image or branding. Finally, as emphasised particularly in recent empirical trade literature, consumers gain additional utility from increased product variety through international trade. Therefore, changes in the set of competitors can affect the competitiveness of exporters (larger numbers of competitors exporting the same product

31

to one particular market means increasing variety for consumers). Although several price measures (CPI and PPI) are adjusted for changes in product quality, they do not ensure any possibility to incorporate changes in consumer tastes or product variety.

4.2. Disaggregated approach to measure price and non-price competitivenessIn this section, we will apply the disaggregated approach proposed by Benkovskis and Wörz (2012) to measure price and non-price competitiveness of Latvia’s exports. This ap-proach is based on the methodology developed by Feenstra (1994) and Broda and Wein-stein (2006), while evaluation of the unobserved quality or taste parameter is based on work by Hummels and Klenow (2005).

The main idea is that consumers are not focused just on physical quantities but they also value variety (a set of exporters as we are sticking to Armington’s, 1969, assumption). Moreover, consumer utility also depends on the quality and taste parameter of a product. By solving the consumer maximisation problem, it is possible to introduce the abovementioned non-price factors into the relative export price measure.

4.2.1. Import price indexWe define a nested constant elasticity of substitution (CES) utility function of a representative household in country i which consists of three nests. At the upper level, a composite import good and a domestic good are consumed:

(8)

where Di,t is the domestic good, Mi,t is composite imports, and κi is elasticity of substitution between the domestic and the foreign good. At the second level of utility function, the com-posite imported good consists of individual imported products:

(9)

where Mig,t is subutility from consumption of imported good g, γi is elasticity of substitution between different import goods, while G denotes the set of imported goods. The third level utility function is the place where variety and quality are introduced into the model. Each imported good consists of various varieties (is imported from different countries of origin, therefore product variety indicates the set of competitors on the particular market). The taste and quality parameter denotes the subjective or objective quality that consumers attach to the product. Mig,t is defined by a non-symmetric CES function:

(10)



where migc,t denotes quantity of imports of good g from country c, C is the set of all partner countries, digc,t is the taste and quality parameter, and σig is elasticity of substitution among varieties of good g.

After solving the utility maximization problem subject to the budget constraint, the minimum unit cost function of import good g is represented by:

(11)

where denotes the minimum unit cost of import good g, pigc,t is the price of good g im-ported from country c.The price indices for good g could be defined as a ratio of minimum unit-costs in the current period to minimum unit costs in the previous period ( ). The conventional assumption is that quality and taste parameters are constant over time for all imported vari-eties and products, ( ) and the price index is calculated over the set of product varieties available in both periods t and t–1, where is the subset of all varieties of goods consumed in period t. Sato (1976) and Vartia (1976) proved that for a CES function the exact price index will be given by the log-change price index

(12)

whereby weights wigc,t are computed using cost shares sigc,t in the two periods as follows:

;

The import price index in equation (12) ignores possible changes in quality and variety (set of partner countries). The underlying assumption that variety is constant was relaxed by Broda

and Weinstein (2006). According to them, if for , , then the exact price index for good g is given by:

(13)

where

and

.

33

Therefore, the price index derived in equation (12) is multiplied by an additional term, which captures the role of new and disappearing variety.

Broda and Weinstein (2006) assume that taste and quality parameters are unchanged for all varieties of all goods ( ), i.e. vertical product differentiation is ignored. Benkovs-kis and Wörz (2011) introduced an import price index that also allows for changes in taste and quality:

(14)

Equation (14) can be taken as a modified version of equation (13) where the additional term captures changes in the quality and taste parameter.

4.2.2. Relative export price indexEquation (14) gives us a formula for a variety- and quality-adjusted import price index. How-ever, we can easily interpret migc,t, which is imports by country i of product g originating from country c, as exports from country c of product g to country i. Another problem arises from the need to compare the performance of one particular country relative to its competitors, while equation (14) gives the aggregate import price from all suppliers. According to Ben-kovskis and Wörz (2012), changes in the relative export price of good g exported by Latvia to country i could be defined in the following way:

(15)

where denotes the minimum unit cost of good g when exported by (imported from) Lat-via, while is the minimum unit cost of good g when exported by (imported from) all countries, except Latvia. After combining (14) and (15) we obtain:

(16)

where Cig-LV is the set of product varieties available in both periods, excluding varieties com-

ing from Latvia, wigc,t-LV and λig,t

-LV are calculated similar to wigc,t and λig,t, again excluding Latvia from the set of exporters (varieties).

The index of adjusted relative export price in equation (16) can be divided into three parts. The first term gives the traditional definition of changes in relative export prices, which are driven by changes in relative export unit values weighted by the importance of competitors in a given market (represented bywigc,t

-LV). An increase in relative export unit values is inter-preted as a loss of price competitiveness.



The second term represents Feenstra’s (1994) ratio capturing changes in varieties (i.e. the set of exporters of this product in our case). This term is calculated with exports coming from Latvia excluded. It can be interpreted as the effect from a changing set of competitors: more competitors for the same product give higher utility and lower minimum unit costs for con-sumers while at the same time lowering the market power of Latvia’s producers. Therefore, more competitors imply a positive contribution to the adjusted relative export price index and are associated with a loss in non-price competitiveness.

The third term is simply the change in relative quality and taste of exports. If the quality and taste of Latvia’s exports is rising faster than that of its competitors, the contribution to the adjusted relative export price index is negative, thus signalling improvements in non-price competitiveness. Although relative quality and taste are unobservable, it is possible to evalu-ate them using information on relative unit values and real market shares (see section 4.2.3).Finally, one needs to design an aggregate relative export price, as the index in equation (16) describes relative export prices only for one specific product g, which is exported to one particular market i. We calculate the aggregated adjusted relative export price index (RXPt) as a weighted average of market-specific indices. Weighting is done on the basis of Latvia’s export data, as this source of information is preferable for determination of a country’s export structure. If we denote the export price and volume of product g exported by Latvia to coun-try i as pxigLV,t and xigLV,t accordingly, the aggregate adjusted relative export price index can be defined as

(17)

where

;

.

Equation (17) shows that the aggregated index is just another Sato (1976) and Vartia (1976) log-change index, with its weights computed using the share of product g exports to country i out of Latvia’s total exports.

4.2.3. Evaluation of relative quality and tasteCalculation of the adjusted relative export price index in equation (16) is a challenging task due to the fact that relative quality and taste are unobservable. As in Hummels and Kle-now (2005), we evaluate unobserved quality and taste from the utility optimisation problem in the following way: after taking first order conditions and transformation into log-ratios, we can express relative quality and taste in terms of relative prices, volumes and elasticity of substitution between varieties:

(18)

where k denotes a benchmark country (any country can be chosen).

35

4.2.4. Estimation of elasticitiesTo derive elasticity of substitution, one needs to specify demand and supply equations. The demand equation is defined by re-arranging the minimum unit cost function in terms of mar-ket shares, taking first differences and ratios to a reference country:

(19)

where , therefore we assume that the log of quality and taste is a random walk process. The export supply equation relative to country k is given by:

(20)

where ωig≥0 is inverse supply elasticity assumed to be the same across partner countries. The unpleasant feature of the system of equations (19) and (20) is the absence of exogenous vari-ables which would be needed to identify and estimate elasticities. To get these estimates one needs to transform the system of two equations into a single equation by exploiting Leam-er’s (1981) insight and the independence of errors εigc,t and δigc,t. This is done by multiplying both sides of the equations. After these transformations, the following equation is obtained:

(21)

where

;

;

Broda and Weinstein (2006) argue that one needs to define a set of moment conditions for each good g, by using the independence of unobserved demand and supply disturbances for each country over time:

where represents the vector of estimated elasticities. For each good g im-ported by country i the following GMM estimator is obtained:

(22)

where is the sample analog of and B is the set of economically feasible values of β (σig>1 and ωig≥0). W is a positive definite weighting matrix, which weights the data such that variance depends more on large shipments and becomes less sensitive to measurement error.Elasticity of substitution between varieties is estimated using equation (22) for all products



where data on at least 3 countries of origin were available. Table A3 in the Appendix displays the main characteristics of estimated elasticities of substitution between varieties. For easier interpretation one can calculate the median mark-up, which equals σig/(σig–1).

4.3. Results of disaggregated approach for Latvia’s exportsNow we can calculate the relative export price index for Latvia, which will take into account non-price factors like quality, taste and changes in the set of competitors. This is done us-ing equations (16) and (17), while unobserved relative quality is evaluated by equation (18). Figure 4 shows three different relative export price indices for every country. The first is the traditional or conventional relative export price index (RXP), which does not take into account changes in quality and set of competitors and is calculated using the first term in equation (16). This index can serve as a benchmark denoting pure price competitiveness of Latvia’s exports. The second index also takes into account changes in the composition of competitors in the market. This is calculated using the first two terms in equation (16). A comparison with the conventional index indicates the contribution of changes in the set of countries to competitiveness. Finally, the relative export price index adjusted to non-price factors is calculated using all three terms of equation (16). This index includes all non-price competitiveness factors analysed in this paper. By comparing it with the conventional RXP, we can highlight the role of non-price factors in Latvia’s export competitiveness.

Figure 4. Latvia’s relative export prices

Conventional RXP RXP adjusted by changes in the set of competitorsRXP adjusted by non-price factors

2005

2006

1999

2000

2001

2003

2002

2004

2007

2008

2009

2010

100

115

120

110

105

95

90

85

Source: UN Comtrade, author’s calculations.Notes: Relative export prices are calculated by cumulating RXP changes from equations (16)-(18); 1999 = 100.

Before analysing the role of non-price factors for export competitiveness, we shall contrast the relative export price index based on trade data to the more frequently used real effective exchange rates reported in Figure 3. As both real effective exchange rates mostly describe price competitiveness, we must compare them with the conventional relative export price index. Although all indicators signal overall losses of price competitiveness between 1999

37

and 2010 for Latvia’s exporters, the magnitude of losses and the dynamics over years differ.Both real effective exchange rates calculated from aggregate price indices show a more pro-nounced real appreciation. At the peak, they point to around 70% appreciation (ULC-based) and around 35% appreciation (CPI-based) in comparison with the 1999 level. Price competi-tiveness improved significantly during and after the crisis; however, the real exchange rate level is still significantly higher than in 1999 (by around 35% for ULC-based and around 25% for CPI-based rates). By contrast, the relative export price index calculated on the basis of highly disaggregated trade data shows a much more moderate loss of price competitiveness of Latvia’s exporters, with the highest point observed in 2008 (losses of almost 15% compared with 1999). Second, there is a difference in time pattern for changes in price competitiveness. All indices show the weakest point of competitiveness in 2008-2009 (for the CPI-based index the late peak is due to an increase in VAT and excise tax rates in Latvia), although in the case of aggregated indices, price competitiveness is rather stable until 2006, while the disaggregat-ed index shows a gradual loss of price competitiveness until 2008. These differences could be driven by various causes, including differences between the CPI, ULC and export prices (unit values). In contrast to the ULC, export prices include profit margins, which declined during the boom years, thus partly compensating rapid growth in labour costs. After the crisis, how-ever, profit margins gradually returned to their initial level. Another crucial factor is structural differences between Latvia and its competitors, which are not captured by aggregated indices. A slower increase of disaggregated relative export price might show that losses of price com-petitiveness were much less pronounced in the main exporting sectors of Latvia.

Table 5. Cumulated contribution of non-price factors to competitiveness of Latvia’s exports by main sector and market

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Sectors

Wood and articles of wood 100.0 101.3 103.3 106.8 113.9 113.3 112.5 111.8 126.8 124.7 116.6 124.9

Base metals and articles of base metal 100.0 97.5 96.1 97.5 99.0 104.6 113.0 113.3 110.2 115.3 110.2 112.1

Machinery and mechanical appliances 100.0 100.4 99.0 100.0 102.1 128.3 134.7 129.4 127.2 137.9 149.2 156.7

Prepared foodstuffs 100.0 106.5 142.7 166.0 173.2 180.5 186.1 195.4 201.1 196.5 205.6 206.9

Chemical products 100.0 100.5 87.1 92.0 82.2 89.4 97.5 105.4 117.3 129.1 137.7 148.4

Vehicles and other transport equipment 100.0 92.4 94.4 87.4 88.0 85.4 85.6 85.2 86.1 89.0 99.0 103.6

Importers

Lithuania 100.0 99.4 99.6 98.9 96.0 112.0 115.8 114.6 111.5 116.2 123.4 126.3

Estonia 100.0 96.7 85.6 87.9 85.2 91.0 97.1 99.2 102.7 110.6 114.6 120.8

Russia 100.0 116.2 136.6 157.4 186.4 196.3 200.2 204.6 200.5 202.3 200.7 206.9

Germany 100.0 99.7 102.4 101.9 102.2 101.1 104.5 102.7 107.7 111.3 112.0 113.1

Sweden 100.0 94.6 94.8 94.7 99.1 97.5 96.4 97.2 105.5 103.2 105.7 123.1

Poland 100.0 102.8 100.7 97.7 95.0 102.2 99.2 83.5 93.8 91.7 88.2 85.8

Source: UNComtrade, author’s calculations.Notes: The six largest product sectors and importers are chosen using 2010 export data for Latvia (the six largest sectors cover 64.9 % of Latvia’ exports, the six largest importers – 62.8%); calculated using equations (16)-(18); 1999=100.



Comparison of RXP adjusted to changes in the set of competitors with the conventional RXP shows no material effect from changes in the set of competitors. In other words, a rising or falling number of competitors is not an important driver of Latvia’s export competitiveness. However, when we look at the RXP adjusted by non-price factors, we observe a rather strong impact of changes in quality and taste on Latvia’s export competitiveness. Figure 4 shows that this index decreases, indicating that Latvia was gaining non-price competitiveness. Al-though Latvia’s export unit values were increasing relative to those of the main competitors, the relative quality of Latvia’s exports (or taste for Latvian products) was rising even faster, compensating the price effect and leading to improvement in overall competitiveness. Unfor-tunately, our methodology does not allow for disentangling tangible and intangible compo-nents of non-price competitiveness, therefore we cannot calculate the contribution of changes in physical quality of exports. Most probably Latvia managed to improve both physical qual-ity of products and their image, branding and market placement.

This finding is mostly corroborated by earlier literature on quality performance in Central and Eastern European countries (CEEC). Dulleck et al. (2005) find overall evidence for qual-ity increases in CEEC exports between 1995 and 2000, although they report serious cross-country differences. For instance, the authors conclude that quality was, to some extent, a concern for the Baltic States. Also Fabrizio et al. (2007) state that the gains in market shares of CEEC, despite the pronounced appreciation trend of their currencies, can be ascribed to a shift in the quality of their exports. The performance of Latvia in terms of quality was posi-tive between 1994 and 2004, albeit worse compared with several Central European coun-tries. Some divergence in the results might be explained by different periods for analysis, as Figure 4 suggests a pronounced improvement in non-price competitiveness starting only from 2002. Finally, Benkovskis and Wörz (2012) use the same methodology and evaluate non-price competitiveness of ten CEEC countries, including Latvia, in the EU market (based on data from Comext). The main conclusions are similar: although relative export prices in-creased more strongly in Latvia in comparison with its competitors, the average quality and taste for Latvia’s goods increased even faster, thus fully compensating for the rise in prices.

Analysis by product sector shows significant improvements in non-price competitiveness for all major export goods (see Table 5). The most rapid improvement in quality or shift in con-sumer tastes is observed for food products, machinery and chemicals. The role of non-price factors for wood and base metal products is positive, although less significant, while quality or taste for Latvia’s exports of vehicles remained unchanged. Analysis of non-price competi-tiveness at different geographical destinations states that the highest contribution of non-price factors to Latvia’s competitiveness is observed in Russia (the most important destination outside the EU). Non-price competitiveness in Lithuania, Estonia, Sweden and Germany is improving, although at a lower speed in comparison with Russia.

ConclusionsThis paper attempts to assess Latvia’s competitiveness in external markets. Acknowledg-ing that the topic of competitiveness is far too broad for one research project, we restrict ourselves to only a few approaches which can be applied to highly disaggregated trade data. Thus, the analysis in this paper still remains in the macro area, albeit at a detailed level. For empirical analysis we use trade data from UN Comtrade at the six-digit level of the HS. The dataset contains annual data on imports of 75 reporter countries from 75 partner countries as

39

well as annual data on Latvia’s exports to 75 countries between 1999 and 2010.

One of the questions the paper addresses is about Latvia’s competitors. From which countries are the main competitor producers coming? According to our results, Latvian exporters face the most severe competition from German producers. Enterprises from China are the second largest competitor group for Latvian producers, but we can expect more competition from this region in the future taking into account a rapid increase in China’s weight. The third and fourth largest group of competitors are from Poland and Russia. As to sectoral composition of competitors, the most significant competition in the wood products markets is staged by Russia, Sweden, Finland and Estonia. Germany and China are by far the two main competi-tors for Latvia’s machinery exporters. The presence of German firms is also very significant in the vehicles, chemical products and base metals sectors.

A very quick and intuitive way to assess the competitiveness of a country is to calculate its export market share. Analysis at a very detailed level allows for extracting contributions of extensive and intensive margins, thus more information is obtained about drivers of competi-tiveness. Overall, competitiveness represented by the total market share of Latvia’s products in the world market was rapidly enhancing during the period observed. The upward trend in competitiveness is driven by the increasing presence of Latvia’s producers in old markets, while the diminishing share of Latvia’s traditional markets in world trade is compensated by the expansion of Latvia’s exporters into new markets. The growing extensive margin is dominated by the geographical dimension, as producers start to export existing products to new destination countries. These results are not uniform across product sectors, however. Some, like vehicles, machinery and food, performed well both in new geographical destina-tion countries and in traditional markets; producers of other articles like wood and chemicals focused on a more intensive presence in traditional markets.

The real effective exchange rate is by far the most popular way of measuring cost competi-tiveness. However, these traditional aggregate indicators have a rather long list of drawbacks, including poor proxying for export activities, ignoring structural differences of competitors, and focusing solely on price competitiveness. Indeed, real effective exchange rates are based on price dynamics and almost ignore changes in product volumes. The abovementioned drawbacks can be resolved, at least partly, by using price and volume trade data on a disag-gregated level. Therefore, we use the relative export price index developed by Benkovskis and Wörz (2012), which takes into account structural differences and allows for disentangling the impact of changes in relative quality and taste from changes in price competitiveness. The results show that Latvia experienced a loss of pure price competitiveness over the sample period, although our index signals that losses of price competitiveness were much smaller than suggested by traditional REER measures. This could be driven by various factors, in-cluding changes in indirect tax rates, counter-cyclical behaviour of profit margins, differences in export structures, and more rapid productivity improvements in export-oriented sectors of Latvia.

When looking at the relative export price adjusted by non-price factors, we observe a rather strong impact of changes in quality and taste on Latvia’s export competitiveness. Although Latvia’s export unit values were increasing relative to those of its main competitors, the rela-tive quality of Latvia’s exports (or taste for Latvia’s products) was rising even faster, fully



compensating for the price effect and improving overall competitiveness. Analysis by prod-uct sector shows significant gains in non-price competitiveness for all major export goods. Analysis of non-price competitiveness in the main geographical destinations shows that the highest contribution of non-price factors to Latvia’s competitiveness was observed in the Russian market (the most important destination outside the EU). Contributions of non-price competitiveness in the EU market are positive as well.

Finally, it should be stressed that this paper can by no means fully describe the issue of Lat-via’s competitiveness and cannot even be regarded as a complete analysis of the subject from the international trade perspective. There is a clear need for further research on microeco-nomic and institutional determinants of Latvia’s competitiveness.

BibliographyAmiti, M., & Freund, C. (2010). An Anatomy of China’s Export Growth. In R. C. Feenstra,

S. J. Wei (Ed.), China’s Growing Role in World Trade (pp 35–56). The University of Chicago Press.

Armington, P. S. (1969). A Theory of Demand for Products Distinguished by Place of Produc-tion. International Monetary Fund Staff Papers, 16(1), 159–178.

Benkovskis, K., & Rimgailaite, R. (2011). The Quality and Variety of Exports from the New EU Member States: Evidence from Very Disaggregated Data. Economics of Transi-tion, 19(4), 723–747.

Benkovskis, K., & Wörz, J. (2012). Evaluation of Non-Price Competitiveness of Exports from Central and Eastern European Countries. Bank of Latvia Working Paper, 1-2012.

Benkovskis, K., & Wörz, J. (2011). How Does Quality Impact on Import Prices? Austrian National Bank Working Paper No. 175.

Besedes, T., & Prusa, T. J. (2011). The Role of Extensive and Intensive Margins and Export Growth. Journal of Development Economics, 96(2), 371–379.

Broda, C., & Weinstein, D. E. (2006). Globalization and the Gains from Variety. Quarterly Journal of Economics, 121(2), 541–585.

Dennis, A., & Shepherd, B. (2007). Trade Costs, Barriers to Entry, and Export Diversification in Developing Countries. World Bank Policy Research Working Paper, No. 4368.

Dulleck, U., Foster, N., Stehrer, R., Wörz, J. (2005). Dimensions of Quality Upgrading Evi-dence from CEECs. Economics of Transition, 13(1), 51–76.

Durand, M. (1986). Method of Calculating Effective Exchange Rates and Indicators of Com-petitiveness. OECD Economic Department Working Papers, No. 29.

Fabrizio, S., Igan, D., Mody, A. (2007). The Dynamics of Product Quality and International Competitiveness. IMF Working Paper WP/07/97.

Flam, H., & Helpman, E. (1987). Vertical Product Differentiation and North-South Trade. American Economic Review, 77(5), 810–822.

Feenstra, R. C. (1994). New Product Varieties and the Measurement of International Prices. American Economic Review, 84(1), 157–177.

Felbermayr, G. J., & Kohler, W. (2006). Exploring the Intensive and Extensive Margins of World Trade. Review of World Economics 142(4), 642-674.

Funke, M., & Ruhwedel, R. (2005). Export Variety and Economic Growth in East European Transition Economies. Economics of Transition, 13(1), 25–50.

41

Hummels, D., & Klenow, P. J. (2005). The Variety and Quality of a Nation’s Exports. Ameri-can Economic Review, 95(3), 704–723.

Krugman, P. R. (1979). Increasing Returns, Monopolistic Competition, and International Trade. Journal of International Economics, 9(4), 469–479.

Krugman, P. R. (1980). Scale Economies, Product Differentiation, and the Pattern of Trade. American Economic Review, 70(5), 950–959.

Purfield, C., & Rosenberg, C. B. (2010). Adjustment under a Currency Peg: Estonia, Latvia and Lithuania during the Global Financial Crisis 2008-09. IMF Working Paper WP/10/213.

Richardson, D. J. (1971). Constant Market Share Analysis of Export Growth. Journal of In-ternational Economics, 1(2), 227–239.

Ruffles, D., Tily, G., Caplan, D., Tudor, S. (2003). VAT missing trader intra-Community fraud: the effect on Balance of Payments statistics and UK National Accounts. Economic Trends, 597, 58–70.

Sato, K. (1976). The Ideal Log-Change Index Number. Review of Economics and Statis-tics, 58(2), 223–228.

Vartia, Y. O. (1976). Ideal Log-Change Index Numbers. Scandinavian Journal of Statis-tics, 3(3), 121–126.



Importers(reporters)

Share in World

imports, %

Exporters(partners)

Share in World im-ports, %

United States 13.51 China 12.71China 9.59 United States 8.18Germany 7.33 Germany 8.03Japan 4.76 Japan 5.15France 4.12 France 3.56United King-dom

3.86 Korea 2.98

Italy 3.35 Netherlands 2.88Hong Kong 3.03 Italy 2.87Netherlands 3.02 Russia 2.69Korea 2.92 Canada 2.64Canada 2.69 United Kingdom 2.63Belgium 2.68 Mexico 2.15India 2.40 Belgium 2.07Spain 2.17 Malaysia 1.70Singapore 2.14 Switzerland 1.62Mexico 2.07 Spain 1.61Russia 1.71 Saudi Arabia 1.57Australia 1.30 India 1.47Turkey 1.27 Brazil 1.41Thailand 1.25 Singapore 1.41Brazil 1.24 Australia 1.39Switzerland 1.21 Thailand 1.34Poland 1.20 Indonesia 1.16Malaysia 1.13 Ireland 1.06Austria 1.03 United Arab

Emirates1.06

Sweden 1.02 Sweden 1.02Indonesia 0.93 Poland 0.98Czech Republic 0.86 Austria 0.96Saudi Arabia 0.73 Norway 0.92Hungary 0.60 Czech Republic 0.82Denmark 0.58 Turkey 0.70South Africa 0.55 South Africa 0.64Norway 0.53 Denmark 0.60Portugal 0.52 Hungary 0.60Finland 0.47 Nigeria 0.55Slovakia 0.44 Vietnam 0.51Greece 0.44 Finland 0.49Romania 0.43 Philippines 0.48

Ukraine 0.42 Chile 0.47Ireland 0.42 Hong Kong 0.46Israel 0.41 Argentina 0.45Philippines 0.40 Qatar 0.45Argentina 0.39 Venezuela 0.42Chile 0.39 Kuwait 0.42Nigeria 0.30 Algeria 0.40Algeria 0.28 Slovakia 0.40Colombia 0.28 Israel 0.38Pakistan 0.26 Ukraine 0.37Morocco 0.24 Kazakhstan 0.33Belarus 0.24 Romania 0.32Venezuela 0.22 Portugal 0.30New Zeeland 0.21 Colombia 0.28Peru 0.21 Peru 0.22Slovenia 0.18 Oman 0.21Bulgaria 0.17 New Zeeland 0.20Lithuania 0.16 Costa Rica 0.18Tunisia 0.15 Egypt 0.17Ecuador 0.14 Slovenia 0.16Luxembourg 0.14 Greece 0.15Croatia 0.14 Azerbaijan 0.15Oman 0.14 Pakistan 0.14Lebanon 0.12 Belarus 0.13Panama 0.11 Ecuador 0.13Serbia 0.11 Bulgaria 0.13Jordan 0.10 Morocco 0.13Dominican 0.10 Luxembourg 0.12Costa Rica 0.10 Lithuania 0.11Guatemala 0.10 Tunisia 0.11Estonia 0.09 Trinidad and

Tobago0.10

Sri Lanka 0.08 Sudan 0.07Kenya 0.08 Estonia 0.07Latvia 0.08 Croatia 0.07Bahrain 0.07 Cote d’Ivoire 0.06Bosnia Herze-govina

0.06 Latvia 0.06

Ethiopia 0.06 Panama 0.05Total 96.25 Total 93.01

Appendix

Table A1. Share of 75 exporters and 75 importers from our database in World imports in 2010Importers(reporters)

Share in World

imports, %

Exporters(partners)

Share in World

imports, %

Source: UN Comtrade, author’s calculations.Notes: Share of exporters and share of importers are calculated relative to total World imports.

43

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

Inte

nsiv

e m

argi

nW

ood

and

artic

les o

f woo

d10

0.0

110.

011

3.1

121.

314

3.9

138.

813

5.0

128.

615

8.6

145.

614

7.4

171.

7B

ase

met

als a

nd a

rticl

es o

f bas

e m

etal

100.

011

0.9

100.

712

4.4

133.

218

1.8

159.

515

6.2

158.

718

1.0

150.

415

4.4

Mac

hine

ry a

nd m

echa

nica

l app

lianc

es10

0.0

100.

813

9.4

166.

520

3.5

240.

128

2.7

305.

040

3.7

510.

552

5.9

503.

6Pr

epar

ed fo

odst

uffs

100.

010

1.9

168.

520

4.8

169.

822

6.3

249.

926

5.4

317.

133

9.8

294.

132

6.8

Che

mic

al p

rodu

cts

100.

098

.988

.586

.091

.610

0.1

101.

711

9.4

153.

817

3.9

157.

415

6.2

Vehi

cles

and

oth

er tr

ansp

ort e

quip

men

t10

0.0

111.

311

0.3

125.

213

2.4

177.

324

5.5

357.

246

3.8

549.

655

4.6

576.

1Ex

tens

ive

mar

gin

– to

tal

Woo

d an

d ar

ticle

s of w

ood

100.

099

.999

.499

.599

.710

0.1

100.

310

0.5

101.

110

1.8

102.

310

2.0

Bas

e m

etal

s and

arti

cles

of b

ase

met

al10

0.0

106.

511

3.8

111.

710

8.2

109.

112

0.1

110.

911

1.6

118.

012

3.0

127.

0M

achi

nery

and

mec

hani

cal a

pplia

nces

100.

010

6.4

99.6

101.

310

0.3

111.

411

7.1

114.

910

9.5

110.

611

5.4

113.

5Pr

epar

ed fo

odst

uffs

100.

099

.510

0.0

101.

910

5.9

116.

012

0.5

124.

312

3.6

124.

312

6.0

124.

9C

hem

ical

pro

duct

s10

0.0

101.

810

4.6

96.2

98.7

103.

110

5.3

106.

210

5.9

112.

910

5.9

107.

8Ve

hicl

es a

nd o

ther

tran

spor

t equ

ipm

ent

100.

010

9.5

145.

413

9.8

152.

418

7.5

213.

521

6.9

224.

422

6.7

238.

622

5.4

Exte

nsiv

e m

argi

n –

prod

uct d

imen

sion

Woo

d an

d ar

ticle

s of w

ood

100.

010

0.0

100.

010

0.0

100.

010

0.1

100.

110

0.0

99.9

99.9

99.9

99.9

Bas

e m

etal

s and

arti

cles

of b

ase

met

al10

0.0

99.9

101.

599

.699

.599

.499

.699

.599

.799

.910

0.2

103.

7M

achi

nery

and

mec

hani

cal a

pplia

nces

100.

010

0.5

100.

099

.910

0.3

102.

810

2.1

101.

295

.795

.595

.795

.1Pr

epar

ed fo

odst

uffs

100.

010

0.9

100.

610

0.3

100.

510

6.3

106.

310

6.4

106.

410

6.4

106.

510

5.8

Che

mic

al p

rodu

cts

100.

099

.599

.599

.910

1.1

102.

910

3.4

103.

410

4.2

105.

110

5.2

105.

4Ve

hicl

es a

nd o

ther

tran

spor

t equ

ipm

ent

100.

010

4.1

108.

110

8.7

115.

711

2.9

113.

811

5.5

116.

711

4.5

114.

011

3.6

Exte

nsiv

e m

argi

n –

impo

rter d

imen

sion

Woo

d an

d ar

ticle

s of w

ood

100.

099

.999

.499

.599

.710

0.0

100.

110

0.5

101.

210

1.9

102.

310

2.1

Bas

e m

etal

s and

arti

cle

of b

ase

met

al10

0.0

106.

611

2.1

112.

210

8.8

109.

812

0.6

111.

511

1.9

118.

112

2.7

122.

5M

achi

nery

and

mec

hani

cal a

pplia

nces

100.

010

5.9

99.6

101.

410

0.0

108.

311

4.8

113.

611

4.4

115.

812

0.6

119.

3Pr

epar

ed fo

odst

uffs

100.

098

.699

.410

1.6

105.

410

9.1

113.

311

6.9

116.

211

6.8

118.

411

8.1

Che

mic

al p

rodu

cts

100.

010

2.2

105.

196

.497

.710

0.2

101.

910

2.7

101.

610

7.3

100.

610

2.3

Vehi

cles

and

oth

er tr

ansp

ort e

quip

men

t10

0.0

105.

213

4.5

128.

513

1.7

166.

118

7.6

187.

819

2.3

197.

920

9.3

198.

4D

eman

d st

ruct

ure

Woo

d an

d ar

ticle

s of w

ood

100.

095

.595

.695

.194

.395

.595

.996

.194

.693

.894

.597

.1B

ase

met

als a

nd a

rticl

es o

f bas

e m

etal

100.

093

.296

.995

.592

.788

.287

.691

.696

.396

.710

1.5

100.

9M

achi

nery

and

mec

hani

cal a

pplia

nces

100.

097

.396

.394

.394

.595

.095

.695

.410

2.0

99.5

99.7

100.

3Pr

epar

ed fo

odst

uffs

100.

010

0.0

102.

610

6.1

106.

310

6.5

107.

910

8.6

109.

210

6.4

104.

210

1.2

Che

mic

al p

rodu

cts

100.

094

.911

3.3

115.

411

4.0

116.

311

9.2

125.

512

2.7

119.

012

5.8

118.

3Ve

hicl

es a

nd o

ther

tran

spor

t equ

ipm

ent

100.

091

.594

.194

.799

.910

2.9

105.

110

8.6

107.

310

6.3

99.5

90.0

Sour

ce: UNComtrade,

aut

hor’s

cal

cula

tions

.N

otes

: The

six

larg

est p

rodu

ct se

ctor

s are

cho

sen

usin

g 20

10 e

xpor

t dat

a fo

r Lat

via

(cov

er 6

4.9%

of L

atvi

a’s e

xpor

ts in

our

dat

abas

e); c

alcu

late

d us

ing

equa

tions

(2)-

(6);

1999

=100

.

Tabl

e A2.

Ext

ensi

ve a

nd in

tens

ive

mar

gin

of L

atvi

a’s e

xpor

ts b

y m

ain

prod

uct s

ecto

rCompetitivenessofLatvia’sexporters


Table A3. Elasticities of substitution between varietiesElasticities estimated Mean Standard

Deviation Maximum Minimum Median Median mark-up

Algeria 3261 22.0 125.2 6492.2 1.05 5.36 23.0Argentina 2920 20.6 69.1 2076.8 1.03 5.49 22.3Australia 2833 79.3 480.4 14517.1 1.02 5.83 20.7Austria 4501 23.8 84.8 4011.7 1.07 5.89 20.5Bahrain 2328 19.9 44.1 992.5 1.05 5.01 24.9Belarus 3326 22.7 71.7 2023.7 1.10 5.21 23.7Belgium 4856 18.4 44.2 905.8 1.05 5.35 23.0Bosnia Herzegovina 3282 22.5 61.7 1453.2 1.05 5.67 21.4Brazil 3946 21.3 82.5 3745.5 1.03 5.52 22.1Bulgaria 3893 19.8 49.1 1096.7 1.07 4.89 25.7Canada 3568 42.1 252.9 8201.7 1.03 8.26 13.8Chile 3525 43.5 210.2 6564.6 1.01 5.44 22.5China 4151 45.4 234.9 7385.5 1.01 6.71 17.5Colombia 3718 19.5 64.3 2305.4 1.06 5.02 24.9Costa Rica 3142 21.9 45.3 931.7 1.02 5.69 21.3Croatia 4029 17.7 40.8 979.8 1.04 4.58 27.9Czech Republic 4672 18.1 36.0 673.2 1.10 5.50 22.2Denmark 4440 19.1 52.2 2541.8 1.09 5.90 20.4Dominican 1053 75.8 482.7 12091 1.01 10.07 11.0Ecuador 3064 20.2 50.8 1368.1 1.05 4.92 25.5Estonia 3464 18.6 39.2 816.2 1.03 5.21 23.8Ethiopia 1778 18.5 43.2 1079.1 1.02 5.68 21.4Finland 4209 20.4 78.7 3478.7 1.04 4.99 25.1France 4963 24.2 150.0 10020.8 1.05 5.54 22.0Germany 4732 21.0 49.6 1695.9 1.02 5.62 21.6Greece 4291 18.1 48.7 1112.0 1.03 4.51 28.5Guatemala 2904 22.1 75.4 2474.5 1.02 5.28 23.4Hong Kong 3555 69.0 917 52025.5 1.01 6.11 19.6Hungary 4125 23.8 53.4 1012.6 1.05 5.56 21.9India 3835 63.6 421.5 15872.1 1.01 6.51 18.1Indonesia 4286 19.5 70.1 3613.6 1.07 5.58 21.8Ireland 4171 27.5 234.2 13318.6 1.02 5.59 21.8Israel 1418 137.2 1090.9 37958.5 1.02 9.03 12.5Italy 4913 19.2 43.5 893.9 1.02 5.05 24.7Japan 4349 22.9 90.5 4472.8 1.02 4.35 29.8Jordan 2145 19.7 47.6 714.1 1.05 4.73 26.8Kenya 2426 28.2 88.5 2177.7 1.05 5.45 22.5Korea 4499 18.3 52.3 2650.8 1.01 5.32 23.2Latvia 3451 21.0 51.4 1089.1 1.02 5.13 24.2Lebanon 3010 21.7 58.8 1469.7 1.03 4.90 25.6Lithuania 3673 18.5 45.6 1177.7 1.04 5.13 24.2Luxembourg 3598 27.5 112.6 5751.3 1.01 6.05 19.8Malaysia 3969 86.9 541.2 14903.0 1.01 4.59 27.8Mexico 3548 29.0 92.7 3528.0 1.01 5.60 21.7Morocco 3412 21.0 59.3 1857.2 1.02 4.87 25.9Netherlands 4193 55.6 329.8 12309.7 1.01 4.67 27.2New Zeeland 3949 19.7 49.4 1058.0 1.05 5.30 23.3Nigeria 1559 29.6 123.8 4373.9 1.03 5.18 23.9Norway 4321 17.3 49.9 1200.1 1.01 4.50 28.6Oman 2325 22.6 58.4 1185.7 1.03 5.12 24.3Pakistan 2387 56.4 404.5 12883.5 1.01 9.95 11.2

45

Panama 2503 21.5 59.5 1661.0 1.00 5.38 22.8Peru 3393 17.9 63.7 2902.9 1.02 5.03 24.8Philippines 3592 24.0 82.6 2832.5 1.03 4.74 26.7Poland 4566 18.6 72.5 4112.3 1.08 5.34 23.0Portugal 4338 19.9 51.1 970.9 1.02 4.86 25.9Romania 4238 20.5 59.4 2517.7 1.01 5.56 21.9Russia 4285 20.0 65.9 3443.2 1.08 6.35 18.7Saudi Arabia 3937 19.2 43.2 1270.7 1.01 5.12 24.3Serbia 3318 21.7 57.5 1222.5 1.01 5.81 20.8Singapore 3068 76.4 438.7 8874.8 1.00 5.79 20.9Slovakia 4130 21.0 76.5 3997.3 1.07 5.80 20.9Slovenia 4241 19.2 60.0 2002.4 1.06 5.27 23.4Southern Africa 4122 39.5 192.4 6241.9 1.01 6.49 18.2Spain 4872 17.9 43.8 1142.0 1.04 5.21 23.8Sri Lanka 2336 37.8 147.9 2872.4 1.02 5.75 21.0Sweden 3986 24.5 56.0 1452.2 1.03 6.21 19.2Switzerland 4684 20.0 46.3 1089.3 1.03 5.33 23.1Thailand 3754 31.5 207.5 6240.8 1.02 5.65 21.5Tunisia 3380 20.6 59.4 2001.7 1.03 5.02 24.9Turkey 4206 17.4 98.9 5958.3 1.04 5.05 24.7UK 4871 18.0 47.1 1381.1 1.05 4.37 29.7Ukraine 3721 20.9 57.2 2206.4 1.08 6.36 18.7US 3956 68.2 526.5 23647.6 1.01 4.98 25.1Venezuela 3520 23.6 80.6 2825.9 1.04 5.37 22.9

Source: UN Comtrade, author’s calculations.Notes: Elasticities of substitutions are estimated using equation (22) for all products where data on at least 3 countries of origin are available.


46 BalticJournalofEconomics12(2)(2012)

47

What managers think of capital structure and how they act: Evidence from Central and Eastern Europe

PéterHernádi1andMihályOrmos2

AbstractThis paper analyzes the capital structure and the choice of financing alternatives across a broad sample of Central and Eastern European companies. Our investigation is built on two methods: the first concentrates on capital structure decisions through quantitative informa-tion applying panel regression for the period 2005-2008 to allow a closer look at the strength of the pecking order and static tradeoff theories; and the second extends the analysis with a qualitative questionnaire on the explicit and latent preferences behind financing policy. The same set of randomly selected 498 firms that fairly represent size classes and countries by the weight of their economic performance are investigated. The CFOs’ answers reflect a pecking order driven behavior, with a limited role for the target leverage ratio; this is confirmed by the estimated coefficients of the panel regression.

JEL classification: G32Keywords: Capital structure; Emerging European countries; Static tradeoff theory; Pecking order theory;

AcknowledgementWe have received helpful suggestions from Matjaz Crnigoj and Aaro Hazak and useful com-ments from participants at the 3rd ECEE Conference in Tallinn (2011). We are grateful to the editor and an anonymous reviewer for valuable comments.

1. IntroductionWe analyze the capital structure and the choice of financing alternatives across a broad sample of small and medium (SME) and large-sized Central and Eastern European (CEE) companies from Poland, Hungary, the Czech Republic, Romania, Slovakia, Bulgaria, Slovenia, Lithu-ania, Latvia and Croatia. We contribute to the existing literature with a deeper understanding of the firms’ motivations in selecting their financing practices by building our investigation on two methods: the first examines regional firms’ capital structure decisions through quantita-tive information, while the second analysis extends the quantitative analysis with qualitative information about the explicit and latent preferences behind financing policies and practices.

1 Péter Hernádi: American Appraisal Hungary Ltd., Váci út 18., 1132 Budapest, Hungary, Phone: +36 1 388 9903, email: [email protected] and Department of Finance Budapest University of Technology and Eco-nomics2 Department of Finance Budapest University of Technology and Economics, Magyar tudósok krt. 2., 1117 Budapest, Hungary, Phone: +36 1 463 4220; email: [email protected]

Whatmanagersthinkofcapitalstructureandhowtheyact:EvidencefromCentralandEasternEurope


For the quantitative analyses, we collect a firm-level database from the BvD Amadeus; and we run panel regressions using well-known accounting proxies (see Rajan and Zingales, 1995; Booth et al., 2001; Delcoure, 2007) to test the strength of both the pecking order theory and the static tradeoff theory.

The drawbacks of a quantitative analysis may be partly resolved by using a more selective questionnaire technique that allows us to analyze not only the accounting track of financial decisions but also the same firms’ motivational background and the specific logic followed in the course of the decision-making process. Brounen et al. (2006) argue that the validity of any capital structure theory can be tested more reliably by minimizing potential distortions aris-ing from use of proxy accounting variables. We compiled a questionnaire concentrating on financial decision-making that was administered to a random sample of 498 firms; these firms fairly represent size classes and countries by weight of their economic performance. With the surveys completed by their executives’ answers, we have the opportunity to create a parallel investigation that allows us to test the implementation of theoretical aspects as suggested by the CFOs’ preferences.

The items in our questionnaire can be divided into two sets: the first set of questions addresses general information about the company, such as managerial ownership in the firm, the domi-nant management culture, the firm’s most important goals, and the shareholders’ most impor-tant goals. The remaining questions focus on financing practices and underlying preferences governing the firm’s decision-making process. Use of this detailed questionnaire allows us to directly test the relevance of capital structure theories, in particular the static tradeoff theory (STT) and the pecking order theory (POT). Questions such as “How important are…” and “What would you do if…” are particularly appropriate to capture preferences within financ-ing policy and provide an opportunity to contrast the executives’ declaration with their firms’ observable track record of financing decisions.

By linking these two distinct methods for the same sample of CEE companies we have the unique opportunity to test whether (i) CFOs’ preferences reflect the theoretical implications of the aforementioned theories (which can be measured through the survey); (ii) their finan-cial decisions (which can be measured through accounting data) in fact coincide with their preferences. Conventional one-stage analyses do not allow testing specifically either the rela-tionship between theory and motivation, or motivation and decision-making.

Regression results confirm that there are no strong tradeoff considerations in financing deci-sions for the broad sample of firms; this coincides well with the CFOs’ opinions revealed in our survey. Using our questionnaire, we find that the most important factor affecting the level of borrowing is the series of cash flows generated by the asset being financed, while corporate tax, non-debt tax shields and the potential costs of bankruptcy are only moderately important. At first sight, these answers indicate that a strong pecking order behavior drives leverage decisions. We find that country-specific results exhibit some diversity in the correlation be-tween tradeoff considerations and choice of debt. The results also show that the majority of firms (73%) do not set a target leverage ratio; but CFOs’ preferences show a high deviation among countries, with the highest proportion of firms having target leverage in Poland and the lowest proportion in the Czech Republic. Executives clearly express that, when financ-ing new assets, they attribute the highest importance to internally generated funds. A total of

49

73% of firms report internally generated funds as “important or very important”, ahead of asset restructuring, ordinary debt, new external common equity and convertible debt. These figures provide strong support for the POT. Using the survey answers, we divide the sample into firm classes based on different attributes. First, we examine whether the CFOs reporting a predefined leverage target give significantly different answers to questions when compared to those who do not have such a target. We find that a larger proportion of firms with target leverage are managed by non-owner managers, while firms without a target seem to be more often managed by owner-managers. We do not find any evidence that the origins of the domi-nant management culture have a similar influence on the presence of target leverage. Our results show that managers with a mixed shareholder-manager objective function are not only less inclined to keep the firm’s capital structure fixed but also consider the target a less strict objective, as firms with low managerial ownership more strictly respect the existing target. In addition, a substantially higher number of firms operated by owner-managers were active on the debt market than were firms in the other sub-sample. We find that non-owner-managers have a higher willingness to forgo an attractive investment opportunity than do insider man-agers if the investment cannot be undertaken without restructuring the firm’s balance sheet.

2. Summary of recent empirical findingsSince the pioneering papers of Miller and Modigliani (MM, 1958, 1963), a large number of studies have addressed the question of corporate financing over the last fifty years. The theo-rems of MM constitute the basis of modern thinking on capital structure by proving that, in the absence of all market imperfections, no optimal capital structure exists for a given firm. The irrelevance proposition has been challenged by MM admitting that asymmetrical tax systems, information and transaction costs likely lead firms to favor some specific financ-ing structure over others. Researchers realized that a firm’s risk is perceived differently by the market depending on its profitability (e.g., Jensen and Meckling, 1976), size (Rajan and Zingales, 1995), activity (Fama and French, 2002), age, market exposure, growth options (Myers, 1977) or the structure of its assets (Titman and Wessels, 1988). Moreover, neither firms nor international investors are subject to a uniform tax system as conceived originally by MM; so different tax systems, together with heterogeneous firm profiles, could ultimately harm the irrelevance of financing (Miller, 1977).

In the 70s and 80s, two competing capital structure theories were developed on the basis of the MM theorems and the empirical evidence about financing decisions. The first theory is the static tradeoff theory (STT), which aims to explain the leverage ratio by focusing on the cost-benefit function of borrowing; the function is supposed to be mostly dependent on observable firm attributes. The other family of theories called preference theories, such as the pecking order theory (POT) or agency theory, emphasize the role of transaction costs in choice of financing. These costs are associated with factors such as transaction timing, informational asymmetry among stakeholders, or prevailing market expectations. Although these theories differ in how they explain a firm’s capital structure, they all suggest an optimal financing strategy to follow under some conditions.

Most empirical analyses about capital structure have concentrated on developed markets (for instance Rajan and Zingales (1995) and Hovakimian et al. (2001)) so that developing coun-tries were not taken into consideration for a long time. Analyzing a sample of Asian, African and Latin American firms, Booth et al. (2001) conclude that knowing the firm’s national-



ity is less helpful when explaining leverage decisions than knowing it together with some firm-specific attributes. These attributes are mostly those that prove to be relevant in devel-oped countries (profitability, asset tangibility, growth). However, these variables alone fail to reliably predict leverage, indicating that one should expect much stronger country effects in less developed countries. These effects are not clear in this highly diverse international sample, nor are they clear in a more geographically proximate environment like Central and Eastern Europe (Nivorozhkin, 2004, 2005) among countries with common institutional and legal roots. Nevertheless, aside from these differences, many firm-specific factors are inter-nationally relevant. Booth et al. find a consistently positive relationship between tangibility and long-term debt as suggested by Rajan and Zingales (2001), emphasizing the increased importance of the collateral value of fixed assets when financial markets are under-developed. The negative correlation between profitability and leverage also seems to be experienced internationally; in underdeveloped financial markets, the cost of external financing, including both transactional and asymmetrical information costs, is higher. This implies that firms’ reli-ance on internal funds tends to be even stronger (Rajan and Zingales, 1998; Demirgüç-Kunt and Maksimovic, 1998).

In recent years, while the aforementioned papers provided answers to several fundamental questions about firms’ financing preferences, they also raised new questions to be answered. Nivorozhkin (2002) shows that Hungarian listed firms used excessively low debt compared to Western standards in the early 90s. Later, Nivorozhkin (2005) and Delcoure (2007) both find that, despite the considerable progress CEE countries have made in their financial markets and institutions from 2000 to the present following regime change, debt is still relatively un-derused. Moreover, following Chen (2004), Delcoure (2007) shows that external debt comes only as a last resort after equity in the hierarchy of financing because managers prefer not to constrain themselves with fixed debt service at the expense of ‘costless’ equity. Interestingly, Avarmaa et al. (2011) find in the Baltic States that firms with a foreign (multinational) back-ground have lower leverage than their local peers in spite of their lower profitability and low-er perceived credit constraints. Crnigoj and Mramor (2009) approach capital structure issue by modifying the conventional shareholder value maximization concept and putting owner-managers and employees (insider shareholders) in control of capital structure decisions. They find that corporate governance is strongly conservative, with a negative correlation between leverage and the extent to which firms are characterized by employee-governed behavior. The authors show that, despite the tendency for a lower leverage ratio in these firms, debt is still preferred to equity when external finance is desirable; therefore, these firms likely follow the POT. Crnigoj (2010) later argues that the probability of using debt in general drops in CEE firms when the largest shareholders are managers (employees). Capital structure choice, in particular the use of debt in the mixed corporate governance framework dominant in develop-ing countries with weak law enforcement such as Eastern European countries (with German/French legal roots) is an issue to be further analyzed. Recent studies present mixed evidence on the extent to which smaller, less-developed firms follow the POT. Based on Myers (1984) and Myers and Majluf (1984), one would expect SMEs to rely heavily on the least controlled funds, retained earnings, because these firms likely have the highest degree of informational asymmetry. Moreover, SMEs are facing higher relative transaction costs than bigger com-panies in their issuance decisions. Numerous recent studies (see Berger and Udell, 1998, Berggren et al., 2000, Chittenden et al., 1996, Michaelas et al., 1999 and Hernadi and Ormos, 2012) found the POT to be followed among SMEs. On the other hand, investigating this

51

field by directly surveying financial executives in the US, Graham and Harvey (GH, 2001) perceive a pecking order-driven behavior among firms but conclude that this is not due to conventional POT factors such as size or growth, both of which should influence the implied informational asymmetry. Brounen et al. (2006) also survey CFOs with a questionnaire in four developed European markets and find pecking order consistent behavior but, similarly to GH, also find that this behavior is not driven by asymmetrical information costs. In addition, corporate taxation also seems to raise new questions. GH show that US firms tend toward a predefined target level of leverage, however, CFOs report that tax concerns are only moder-ately important in their debt policy. More specifically, larger firms tend to be more interested in tax issues than smaller ones. On average, European firms are somewhat less concerned by corporate tax issues than their US counterparts, but Brounen et al. (2006) find relatively strong support for the existence of a target ratio among European firms. While the corporate tax system is, in fact, one of the critical policy tools currently among CEE countries compet-ing regionally, we know little about whether the diverse and country-specific corporate tax systems influence firms’ everyday financing practices.

3. DataSimilarly to Desai et al. (2003), and Hutchinson and Xavier (2006), we use Bureau van Dijk’s Amadeus firm-level database. Amadeus proves to be a valuable tool in the constitution of representative samples (country, size). The geographic scope initially covered ten countries that had either recently joined the EU or had been candidates (Poland, Hungary, the Czech Republic, Romania, Slovakia, Bulgaria, Slovenia, Lithuania, Latvia and Croatia). The num-ber of firms selected in the sample reflects each country’s economic weight within the total output of the block, estimated by their nominal GDP (2006). By considering each country’s relative weight, we can formulate general implications regarding capital structure choice that are representative of the region as a whole, but cannot draw specific conclusions for the many smaller countries that have lower populations and less economic power. Based on this meth-od, Poland represents the biggest share of the sample with 180 firms, followed by Hungary (73), the Czech Republic (72) and Romania (68). Slovakia, Bulgaria, Croatia, Slovenia and Latvia count 30, 25, 22, 16 and 12 firms, respectively. Due to unanticipated difficulty in the course of surveying Lithuanian firms, we have been constrained to exclude Lithuania from the analysis. Other than limiting the sample to exclude the smallest companies, each firm in Amadeus had an equal chance of being selected for the preliminary country samples (strati-fied sampling). These samples were created by randomly selecting a large number of firms, classified into five categories based on the number of employees (25-50, 51-250, 251-375, 376-650 and over 650). Altogether we have surveyed a random 498 firms with the aforemen-tioned weight by country and size and later we collected the necessary accounting data of exactly the same set of firms from the Amadeus database.

Data include the complete balance sheet and the income statement, as well as supplemental information on the industry, ownership structure, and creditworthiness. The analyzed ques-tionnaire is composed of 17 questions. Some of these questions address general company information (such as managerial ownership in the firm, the dominant management culture, the firm’s most important goals, as well as the shareholders’ most important goals and con-sideration of stakeholders) while others focus on everyday financing practice and underlying preferences governing a firm’s decision-making processes. Most questions address directly the personal opinion of the subject and most of the answers must be given by ranking mul-



tiple proposed alternatives. Ranking by importance/probability has to be done on a scale from 1 (not important, less likely) up to 4 (very important, most likely). The questionnaire was translated into all the official languages represented in the sample, and to minimize any data bias arising from language problems, executives were interviewed on the phone in their native language.

4. Quantitative Analyses of Capital StructureWe analyze quantitative aspects of capital structure through an unbalanced panel dataset cov-ering the sample firms for the period from 2005 to 2008. The panel consists of firms that have available balance sheet data for each year in the period investigated. Firms with an incomplete set of regressors have been automatically removed from the regression; however, a missing variable for a given period does not guarantee the firm’s ultimate rejection from the sample. Our sampling method resulted in a panel composed of 918 observations for the regression of the total leverage ratio (with 299 cross-sections, unbalanced).

We build our analysis on recent studies that identify the most relevant and widely accepted capital structure determinants. Our tests rely on an asymmetric panel structure (a relatively large number of cross-sections with few periods), the joint significance of the period effects has been rejected by an F-test in both estimations. Hence we apply only cross-section effects for which the Hausman test result shows that the consistency of the estimator with random effects could be rejected at the 5 percent significance level, but not at the 1 percent. If the Hausman test does not reject the null hypothesis, then both fixed and random effect estima-tors are consistent, but the latter is efficient. Therefore, we estimate the following model with both cross-section fixed and random effects (FE and RE, respectively) included:

(1)where Levi,t is the book leverage ratio for firm i in year t, Xi,t is the vector of explanatory variables, δi are cross-section specific effects (fixed or random), Di are time-invariant coun-try, industry and public company dummies, is the overall constant, β and γ (only in RE estimation) are the respective coefficient vectors and εi,t is the error term. Unequal variance in residuals is tested in both dimensions. Reported t-statistics are robust to heteroskedastic standard errors, as White diagonal standard errors are applied in FE estimation and White cross-section standard errors are reported for RE estimation.

As dependent variable we use the leverage ratio calculated as the sum of long-term liabilities plus current loans divided by the sum of book equity, long-term liabilities and current loans. We refer to this ratio as the total leverage ratio.

We incorporate tax effects in our model (TAX) as income tax effectively paid in the previ-ous financial year divided by the accounting profit before taxation. We consider this measure to be the best proxy for taxation for two reasons. First, even if the annual effective tax rate could be biased (Booth et al., 2001), averaging annual effective tax rates would be mislead-ing given that the statutory rates in the CEE countries have shown a regressive trend during the last decade (KPMG Tax Rate Survey, 2009). Second, historical income statements show that CEE firms rarely pay as much tax as they should if they were solely subject to statutory Corporate Tax rules.

53

We measure the size variable (SIZE) as the natural logarithm of annual net sales, converted into Euros. Using net sales as a size proxy, instead of assets, limits biases resulting from any differences between firms’ investment policies, and limits the risk of underestimating firms operating with an older asset structure.

Profitability (ROA) is measured as earnings before interest and taxes (EBIT) to total assets. We include cash in the model along ROA because de Haan and Hinloopen (2003) suggest that ample liquidity and high profitability are together indicators of financial health. Follow-ing Gaud et al. (2007), we measure the availability of internal funds (past accumulation of financial slack) by the ratio of cash and equivalents to total assets (CASH).

Volatile earnings are often considered to be good indicators of business risk. A financially distressed situation is the likely outcome of volatile profitability and has its own costs (e.g., renegotiating a supplier or debt contract and restructuring the organization), which influences the firm’s borrowing decisions. In controlling for risk (RISK), we measure the four-year aver-age variability of ROA, as introduced by Booth et al. (2001).

Numerous studies (e.g., Long and Malitz, 1985, and Friend and Lang, 1988) identify asset structure as a determining factor of leverage. Empirical evidence shows that the proportion of fixed to total assets (tangibility of assets) positively influences leverage, while the intangibil-ity of assets has the opposite effect. We follow Titman and Wessels (1988) by using two asset structure variables (TANG and INTAN) as proxies for the collateral value of assets. TANG is calculated as the sum of net tangible fixed assets plus inventories to total assets, while INTAN is calculated as the ratio of intangible assets to total assets.

Because non-financial cost items might fulfill the same favorable role in corporate taxation as interest payments, a firm incurring higher depreciation costs may be less motivated to increase borrowing. Following Fama and French (2002) and de Haan and Hinloopen (2003), we apply the ratio of annual depreciation and amortization to total assets (DEPR) as a proxy for non-debt tax shields. We expect that testing the effects of non-debt tax shields has a par-ticular secondary concern in the case of smaller private companies. These firms’ managers have many more options for optimizing tax liability (flexible cost structure, perquisite assets) because they are not necessarily interested in impressing the market.

Present and future growth opportunities are considered key determinants of leverage in all capital structure theories. With regard to current growth and based on the assumption that present growth is to hold in the near future (Fama and French, 2002), we apply the D_SIZE variable, a proxy for the growth rate of the firm calculated as the difference of log net sales in period t and in period t-1. Along with the above-described financial proxies, we include in the regression a public company dummy and dummies that capture country and industry effects (manufacturing, utilities, construction, wholesale, transportation and services).



Table 1. Regression of the total leverage ratio

Table 1 reports results for the panel regression of the total leverage ratio with cross-section fixed and random effects for years 2005 to 2008. The estimation is controlled for public company, country and industry dummies; these results are available upon request. Coefficient estimates are subject to a two-sided statistical test. *, **, *** indicate significant coefficients at 10%, 5% and 1%, respectively.

Table 1 summarizes the results of the panel regression of the total leverage ratio. At the le-verage optimum, the marginal benefit of an incremental portion of debt should equalize the marginal cost of increased financial distress and sub-optimal investments. To verify the rel-evance of the selected proxies to leverage, we directly compare the results obtained from the FE and RE estimators. Assuming that the independence of the explanatory variables and the unmodeled individual effects does not hold (only FE is consistent), we find that ROA, CASH and DEPR are negatively correlated with leverage, while INTAN and TANG are estimated to impact leverage positively.

All parameter estimates are significant at 1% except for CASH. As the independence of the omitted individual effects and the explanatory variables could not be rejected at the 1% level (Hausman test), we repeated the regression assuming cross-section random effects by includ-ing proxies such as RISK and industry and country dummies. This regression has a very simi-lar result; all the estimated parameters have the same impact on leverage and are significant. Neither RISK, SIZE nor D_SIZE became statistically significant in either of the estimations. RE regression also shows that most country effects, when separated from the unmodeled effects, had an economically significant impact on leverage. Unlike the country dummies, results generally eliminated the importance of industry dummies, such as manufacturing, trade or construction in the explanation of leverage differences among firms, with the sole exception of utilities.

Fixed Effect Random Effect

Variable Mean t-stat Mean t-stat

C 0.453 1.02 0.376 2.00 **

TAX -0.002 -0.17 -0.2 -0.54

ROA -0.212 -2.47 ** -0.252 -4.72 ***

RISK -0.13 -0.04

SIZE -0.017 -0.64 -0.008 -0.83

D_SIZE 0.023 1.03 0.029 1.58

TANG 0.326 3.94 *** 0.239 2.94 ***

CASH -0.146 -1.71 * -0.238 -3.79 ***

DEPR -1.414 -3.36 *** -1.070 -8.10 ***

INTAN 0.871 2.65 *** 0.9323 7.91 ***

Mean Leverage 0.270 0.270

Adj. R-square (within) 83.0%

Adj. R-square (overall) 12.3%

Number of observations 918 918

55

5. Qualitative Analyses of Capital StructureWhen interpreting the questionnaire, we emphasize uncovering the motivations of the CFOs that govern their capital structure decisions in a STT, a POT-driven or a mixed framework. Our questionnaire has separate questions on corporate governance and agency (stakeholder) issues. Additional information is used to divide the sample along stakeholder- and ownership-related attributes; we use these aspects to gain particular insights into the specific financing pattern of different firm classes. The identification of specific classes could help us to under-stand the diversity of previous empirical results and could answer why one specific theory cannot hold for a broadly diverse sample of firms.

From MM’s extended framework, we know that an optimal capital structure can create share-holder value in the presence of market imperfections. Therefore, one of the most analyzed questions in modern corporate finance is whether a firm makes financing decisions to keep a fixed leverage ratio where the cost of capital can be minimized or, alternatively, financing decisions are rather the outcome of customized cost-benefit analyses that lead to the observed capital structure without any preconception about how this structure should look.

Tables 2, 3 and Appendices 1, 2 exhibit the survey questions and answers for the full sample of firms and country sub-samples, respectively. They present CFOs’ answers to the question “Whatfactorsaffecthowyoudeterminetheappropriateamountofdebtforyourfirm?”by indicating the absolute importance of each given alternative. Our results show that the most important factor of indebtedness (mean is 2.98) is the level of cash flow generated by the asset to be financed. Given that 73.7% of firms considered this factor to be important, it is a much more decisive factor in borrowing than any other listed factor. Corporate tax rate, for instance, is only moderately important, as only 50% of CFOs assigned it a score 3 or higher (mean is 2.44). The potential costs of bankruptcy/financial distress have a similar outcome, with a slightly higher mean (2.55) and a 55% share of answers as “important.” As long as the STT holds and firms look to optimally balance the distress-costs and tax-benefits of incre-mental borrowing, these two factors are expected to be highly and equally important.

CFOs’ choice of other alternatives further contests the relevancy of the STT. As deprecia-tion, amortization and other non-debt-related costs can also decrease firms’ tax base, so they should impact the proper amount of debt. The CFOs say, however, that this aspect is just as important as either the corporate tax rate or the potential costs of bankruptcy; that is, none of them play a decisive role in debt policy. GH find a similar pattern for the relative importance of debt-related factors: US CFOs consider the most important goal to keep financial flex-ibility. They report that nearly 60% of firms considered financial flexibility as important or very important, suggesting that POT prevails in leverage decisions. Tax advantage is only the fourth most important factor, while volatility of cash flow and potential cost of bankruptcy, separately, rank third and seventh. Examining the four biggest CEE countries for which our sample allows a reliable analysis (the Czech Republic, Hungary, Poland and Romania), an-swers depict a fairly homogenous picture of debt-related factors (see Appendices 1 and 2). Hungarian CFOs attribute the lowest importance to the factors concerned, indicating the level of non-debt tax shields (2.28) and potential costs of financial distress (2.12) as the second and third most important issues. Romanian firms more rigorously follow the theory of corporate financing by attributing relatively high importance to both the potential costs of financial distress (3.15) and the corporate tax rate (2.82). Consequently, Romania is the only country



where the executives’ priorities are explicitly in line with the conventional implications of the STT. Between these two extremes, Czech and Polish firms report a similar average relevancy for the factors concerned.

Table 2 Summary of survey results for the total sample of CEE countries (part 1)What factors affect how you determine the appropriate amount of debt for your firm?

ImportanceMean

Low High

Projected cash-flow from the assets to be financed 26.3 73.7 2.98

The debt levels of other firms in the industry 57.5 42.5 2.27

The potential costs of bankruptcy or financial distress 44.9 55.1 2.55

The corporate tax rate 49.9 50.1 2.44

The level of depreciation and other non-debt tax shields 44.0 56.0 2.57

Does your firm have a target value for the leverage ratio?no yes

73.3 26.7

If „Yes”

We usually/permanently depart from it (flexible) 16.4

We occasionally dep. from it (somewhat tight) 44.0

We strictly respect it 39.7

What is this target value (book debt/total assets)? 37.8

Which of the following sources of long-term funds are / would be important for financing new investments?

ImportanceMean

Low High

Retained earnings 26.9 73.1 2.97

Restructuring assets 43.1 56.9 2.56

Straight debt 50.0 50.0 2.51

Convertible bond 86.8 13.2 1.48

External common equity 66.3 33.7 2.04

During the last three years, did your firm apply for new loans or capital leases?no yes

30.8 69.2

If „Yes”, these applications were

always approved 94.1

always denied 1.2

sometimes approved and sometimes denied 4.7

Table 2 presents the survey questions and answers for the full sample (country sub-samples are presented in Appendix 1). Figures show the mean score of importance/likelihood for each question and the distribution of answers ‘1’ and ‘2’, classified as ‘low’, and ‘3’ and ‘4’, classified as ‘high’.

The answers given to the question “Doesyourfirmhaveatargetvaluefortheleveragera-tio?” emphasize that the majority of CEE firms (73%) do not set a target leverage ratio. Out of the firms that report a definite target leverage (about 27%), only 16% respond that they treat this ratio as flexible, while 44% consider the target as a tighter goal that the firm can only occasionally depart from, and 40% indicate that they strictly respect it. Looking at countries, there is a larger share of firms with fixed target in Poland (38%) than in Hungary (28%) while the share of these companies is smaller in Romania (18%) and the Czech Republic (14%). Differences in fixing target leverage ratio are more than surprising when CFOs’ preferences

57

in tradeoff-related debt factors are taken into account. Managers in Hungary have an average commitment to keeping the firm’s capital structure fixed. However, when they are asked for their underlying motivations, they do not attribute high importance to any of the conventional factors of the STT. On the contrary, Romanian firms treat all these factors as important, but many of them do not have a predefined target leverage. The results in all countries strongly suggest that firms with fixed leverage try to hold the predefined leverage strictly, and only few report flexibility in their targets.

Answers regarding debt policy reveal a fairly high conservatism in the use of debt in each country. The main motivation for borrowing seems to relate directly to a class of assets rather than to firm-level optimization of costs and benefits. We see further evidence of conservatism by analyzing the question “Which of the following sources of long-term funds are/wouldbeimportantforfinancingnewinvestments?” With this question, we indirectly investigate whether managers are concerned with asymmetrical information problems. If they are, we expect that their preferences will be consistent with POT. In other words, CFOs find external funds less desirable than internal funds, and among external funds they prefer ordinary debt to convertible debt and convertible debt to new common equity. Our results show clearly that, when financing new assets, the highest importance is attributed to internally generated funds, with 73% of firms reporting it as important or very important. The mean of 2.97 for in-ternally generated funds is significantly higher than the second most important “new funds”, restructuring of assets (2.56), which, in turn, is ranked ahead of new external debt (2.51), new external common equity (2.04) and, as a last resort, convertible debt. Convertible debt has a low mean, and only 13% of CFOs consider it important in financing. These figures provide very strong support for the POT, which almost directly aligns with this conventional order of alternatives; as the sole exception we find evidence of a minimal role for convertible bonds in CEE countries. With regard to earlier studies, the importance of financing sources is unusual and raises important implications. Ang (1991) proposes a modified POT for SMEs, in which new external equity raised from existing owners ranks in second place behind retained equity and ahead of debt. Ang argues that it is difficult to draw a sharp distinction between “real” internal funds and the latent contribution of owners when the roles of owner and manager are not separated. Similarly, Chen (2004), Nivorozhkin (2005) and Delcoure (2007) find that, due to lack of long-term borrowing instruments caused by a combination of a missing corpo-rate bond market and an immature banking system, managers in developing countries prefer external equity to debt because it is not obligatory and because share capital appears to be a long-term and “free” source compared to debt.

By questioning CFOs’ motivation in their debt decisions, GH find that the most important of all aspects is keeping financial flexibility that is likely driven by managers’ desire for financial comfort rather than by the goal of optimizing present and future financing costs proposed by Fama and French (2002). Similarly to GH Brounen et al. (2006) find that, in Europe, despite POT-consistent behavior that is widely observable in everyday financing practice, this behav-ior does not seem to be caused by pure asymmetrical information concerns as first proposed by Myers (1984).

Our findings partially contradict previous studies on emerging markets as CFOs explicitly rank funds in a way that is strongly in line with POT. The analysis of the underlying firm attri-butes contributes to a better understanding of why some firms follow the theoretical scheme



more closely than others. Contrary to earlier findings, at the end of the 2000s CFOs firmly declare that long-term borrowing has become an accessible and viable option, and hence the revealed hierarchy matches that of the most developed and mature economies. Table 2 shows that nearly 70% of firms answered the question “Duringthelastthreeyears,didyourfirmapplyfornewloansorcapitalleases?” positively, a remarkably high ratio even if we take into account that a substantial number of firms target short-term loans. It seems that the CEE business environment has substantially changed since the era of Nivorozhkin (2002, 2004, 2005), a period when even the largest listed companies suffered from severe credit rationing problems that restrained their development. The presence of a more mature banking system and the availability of external credit are even more explicative considering that, based on our survey, 94% of all debt and capital lease applications were approved between 2005 and 2008. Contrary to Nivorozhkin’s findings suggesting that, even with serious asymmetrical informa-tion problems between insiders, creditors and shareholders, firms do not necessarily follow POT; we find that this reversion in POT is no longer noticeable among CEE firms.

Table 3 Summary of survey results for the total sample of CEE countries (part 2)What percent of equity is owned by the top three officers?

<5% 67.8

5-10% 4.2

10-20% 5.8

>20% 22.2

Management culture dominanceForeign Local

41.6 58.4

Does your firm take advantage of operating lease?no yes

53.3 46.7

In periods when the firm’s investment is low, does your firm retain a part of its free cash-flows?

no yes

57.3 42.7

Given an investment that could not be taken without modifying the actual bal-ance sheet structure, what action would you take?

LikelihoodMean

Low High

Deviate from the actual capital structure 51.2 48.8 2.45

Cut the dividends 44.5 55.5 2.59

Restructure assets 42.9 57.1 2.56

Forgo the investment opportunity 66.7 33.3 2.11

If you forwent the investment opportunity, why?

in order to hold independence 42.6 57.4 2.69

in order to keep moderate level of leverage 32.1 67.9 2.87

in order to keep the senior shareholders’ value 54.5 45.5 2.35

external equity financing is unavailable 50.0 50.0 2.41

borrowing is impossible 58.1 41.9 2.25

debt service is not expected to be satisfied 50.6 49.4 2.42

Table 3 presents the survey questions and answers for the full sample (country sub-samples are presented in Appendix 2). Figures show the mean score of importance/likelihood for each question and the distribution of answers ‘1’ and ‘2’, classified as ‘low’, and ‘3’ and ‘4’, classified as ‘high’.

59

Furthermore we find another common specificity in the region. Unexpectedly high impor-tance is attributed to asset restructuring (mean of 2.56), which we define as any kind of trans-action targeting the pool of assets to free potential internal cash reserves. The answers suggest that executives are clearly aware that their firms could temporarily operate with endogenous inefficiencies and use this sub-optimal investment option as an opportunity to solve financing problems. While this type of asset restructuring might not even exist in an efficiently man-aged firm, the answers imply that executives do consider this option as equally important to external debt financing (2.51).

One of the crucial elements in POT is management’s desire to keep the maximum amount of liquidity under control. Retained equity is cheaper than debt because there is no interest on it, and it is more easily accessible than external equity because even senior shareholders have to be persuaded to inject fresh equity into new projects. For non-owner managers, equity may look like the cheapest fund, but only if it is available internally. Table 3 presents that 43% of CFOs responded “yes” to the question: “Inperiodswhenthefirm’sinvestmentislow,doesyourfirmretainapartofitsfreecash-flows?” This result is significant because managers of cash-rich firms may be motivated to waste some part of their excess cash on perquisite assets and value-deteriorating investments. However, the ratio of these reserve-keepers does change significantly between countries. Only 36% of Polish managers build internal cash reserves, while Romanian and Hungarian firms report ratios of 57% and 53%. Interestingly, although retained equity represents a primary financing source in all major countries, Polish CFOs assess the importance of internal equity (3.18) as the highest relative to other alternatives. On the contrary, Hungarian and Romanian CFOs are more willing (and probably more able) to retain internal equity, but the results show much lower lead in its importance over debt. In Hungary and Romania funds potentially available through restructuring are ranked only in third place, while in Poland and the Czech Republic restructuring is preferable to debt. The ratio of loan applications supports these country-specific results, with Romanian (87% with a 97% approval rate) and Hungarian firms (72% with a 92% approval rate) significantly ahead of Czech and Polish firms. This outcome suggests that sufficiently tight and sustainable lending conditions, in addition to a mature financial system, are apparently a prerequisite to follow the POT.

As the next step we ask “whatactionwouldthey[the CFOs]takeifagiveninvestmentcouldnotbetakenwithoutmodifyingtheactualbalancesheetstructureofthecompany.” We offer them options on both the assets side and on the liabilities side, as well as the choice to forego this attractive investment opportunity for any reason.

Overall, executives are most prepared to cut dividends, as 57% of them consider that building cash reserves internally with the consent of senior shareholders is the easiest and probably the cheapest way to fund a promising future investment. The mean answer of 2.59 is consis-tent with CFOs’ expressed preference toward a POT-like hierarchy for financing funds. The second likeliest action, restructuring of assets, is also consistent, with a mean of 2.56. These results again support POT regarding the dominance of asset restructuring and retained cash and demonstrate a salutary consistency in CFOs’ answers. While CFOs show themselves to be quite flexible in adjusting dividends and fixing under-optimal balance sheet problems, they are less inclined to deviate from the firm’s actual capital structure when they are questioned about it directly. Even though dividend adjustments similarly influence the firm’s debt/equity



ratio, their impact is indirect, unlike direct debt and equity decisions. The mean of 2.45 indi-cates that such direct transactions are less preferred than internal action; nevertheless, CFOs still consider them a viable option. Consistent with favorably loose lending conditions, we find that, when managers see an attractive opportunity, they do not usually forgo it (2.11), instead trying to find ways to finance it by either restructuring assets or extending the balance sheet.

Among the countries examined we find substantial differences in the hierarchy of actions. In all countries but Poland it is less likely that CFOs would forego an attractive investment and more likely that they would deviate from the actual capital structure. Discarding the oppor-tunity ranks third in Poland, and the mean of 2.32 is well above the mean of other countries. Polish executives’ preferences are in line with their Czech counterparts by seeing an impor-tant opportunity in shifting asset composition and, hence, creating free liquidity. In contrast, Hungarian and Romanian managers that show a greater willingness to retain free cash flows and to borrow from external creditors when necessary will more likely deviate from the actual capital structure and cut back dividends if necessary to supply new funds. In both Hungary and Romania, the likelihood of foregoing the given opportunity is very low compared to Po-land and the Czech Republic.

5.1. Analyses of Sub-SamplesTo understand how much of the discussed heterogeneity is derived from country-specific fac-tors and how much of it might be explained by non-quantitative firm attributes, we divide the full sample into sub-samples along qualitative attributes by which pure country effects can be eliminated (the results are presented in Appendices 3 and 4).

The first attribute we use to differentiate firms is whether they have a target leverage or not. We find that out of the 116 firms being operated with a fixed target, the top 3 executives own less than 5 percent of the shares in 75% of the companies, and in only 16% of the firms the top 3 executives together hold more than 20% of shares. Firms without a defined target ratio are more commonly managed by owner-managers: we find 26% where the first 3 executives jointly hold more than 20% of shares. We do not find any evidence that the origin of a domi-nant management culture is correlated with the existence of a leverage target. If we assume that larger firms are more likely to belong to international owners, then given that the theory of optimal capital structure mainly gained ground in the most developed countries, we would expect to find relatively more firms with foreign management culture among leverage-target-ing firms than among those not having such a target. Contradicting this proposed relationship, in the two complementary samples the proportion of firms dominated by foreign culture is equally around 40-42%.

CFOs express an unexpected homogeneity in other policy issues, not just management ori-gins. We find no evidence that firms with fixed target leverage apply for new loans/capital leases more frequently than firms without a fixed target. Likewise, only a moderate gap exists in the proportion of firms building cash reserves in low-investment periods, with a 43-47% share in both samples. There is a larger difference in the number of firms taking advantage of operating leases, as firms with a target leverage turn more frequently to this tool. This outcome is consistent with our expectations, as an operating lease provides an important ad-vantage for a firm that tries to strictly keep a pre-defined capital structure.

61

There are both similarities and discrepancies between samples regarding borrowing factors. On the one hand, projected cash flow from the asset to be financed is the most important fac-tor in both sub-samples, while the least important is the debt level of rival firms. On the other hand, leverage-targeting companies assess conventional tradeoff aspects as more important in their debt decisions than firms without target leverage. These CFOs give more weight to corporate tax and non-debt tax shield considerations, which is in line with the basic principle of the STT. While we see a shift toward a more tradeoff-driven debt policy among leverage-targeting firms, interestingly our results do not confirm that these firms are less concerned by pecking order factors. Executives in both samples treat retained equity as the most important source of financing, ahead of asset restructuring and debt. In addition, external equity is com-monly considered one of the least preferred alternatives. This preference implies that despite the fact that the STT does not explicitly differentiate between internal and external equity in setting the optimal capital structure, even leverage-targeting firms are not indifferent to asym-metrical information problems.

Creating two sub-samples based on the degree of managerial ownership attribute, there are 120 firms where the top 3 executives hold more than 10% of shares (these firms are managed by “owner-managers” or “insiders,”); and there are 308 firms in which the top 3 executives together own less than 10% of shares. As expected, we find a higher proportion of firms hav-ing local management culture among the firms managed by owner-managers (67%) than in the complementary sample (56%). However, in absolute terms, a relatively high local man-agement influence had been developed in all firm classes by the end of the 2000’s.

We find that managers with a mixed (employee-owner) interest not only feel less constrained to keep the firm’s capital structure fixed than non-owner-managers (19% versus 30%) but also consider this target more flexible. Unlike firms with low managerial ownership, where the existing target is strictly respected by at least half of the managers, only 20% of insider CFOs report a strictly predefined ratio.

Despite the fact that owner-managers do not “like” high and pre-set leverage, they enter the external debt market quite frequently. A substantially higher proportion of firms operated by insiders applied for a new loan/capital lease in the preceding 3 years (86%) than firms in the other sub-sample (62%).

There is a notable discrepancy in how CFOs treat external borrowing when their manage-rial and shareholder interests are aligned as against when they are not. As to the importance of financing funds, CFOs of firms managed by insiders firmly declare that ordinary debt and retained earnings are about equally important. This outcome contrasts strongly with the other sub-sample, where non-owner managers have a clear preference for retained earnings over asset restructuring and straight debt. Just as the frequency and importance of using debt alters, so do the factors that determine the appropriate amount of borrowing. In firms man-aged by insiders, CFOs assign particularly high importance to cash flows from the asset to be financed, as opposed to other firms where this average is much lower. In addition, there is also a noticeable gap regarding the potential distress costs in favor of firms managed by owner-managers, meeting our prior expectation that financing decisions made under a mixed interest more vigorously represent shareholder-specific interests.



Overall, our results show that non-owner managers operate a firm in a more disciplined way that better adheres to theory than do insider managers; the latter can and do more likely devi-ate from the theoretical rules in their decisions. There is a higher probability that executives with low ownership are required to keep a fixed leverage ratio, and it is also more likely that they have to provide a reasonable argument before turning to the external debt market. In addition, we find that non-owner managers have a higher willingness to forgo an attractive investment opportunity than do insider managers if this investment cannot be undertaken without restructuring the firm’s balance sheet. The non-owner-managers are less likely to deviate from the actual capital structure and cut the dividends, where insiders evidently have more flexibility.

Based on the attribute of “the origin of management culture,” two samples have been cre-ated with 202 (foreign) and 284 (local) firms. We find that knowing which culture prevails in the firm’s management adds relatively little to our knowledge about the relevance of capital structure theories. There is no significant gap in the proportion of firms with fixed target leverage, even there is no gap in the proportion of firms opting to retain free liquidity when investment demand is low. However, there are some differences in the relative number of loan applications for the benefit of firms with local management culture, and in line with this, firms with foreign culture are more likely to take advantage of operating leases.

6. DiscussionIn the quantitative analyses (Table 1), our overall results confirm that conventional tradeoff aspects are not strongly present in financing decisions of the broad sample of firms. This outcome coincides well with CFOs’ verbal preferences. The TAX variable, which should positively relate to the leverage ratio under the STT and is denoted by half of the executives as not important, is not correlated with the total leverage ratio. The RISK, which is a time-invariant proxy representing volatility of earnings, does not play a determinant role in firms’ financing decisions. Similarly to TAX and according to our survey (Table 2), slightly more than half of CFOs report that the costs of getting into a distressed situation are a decisive fac-tor in borrowing. This is another result that does not support the STT, taking into account that a firm with more volatile operating profitability will target a lower debt/equity ratio to avoid potential problems in debt service. The SIZE (Table 1) also fails to reveal a positive relation-ship with the leverage ratio. Hence, coefficient estimates support that the majority of firms rejecting a target ratio (73%) in fact do not optimize their capital structure by analyzing the tax savings and distress costs of debt.

After rerunning the panel regression only on the leverage-targeting firms (see Table 4), we find that, although TAX becomes slightly positive, it remains insignificant similarly to RISK. The irrelevance of these factors on leverage is further confirmed by each sample-specific regression based on managerial ownership and culture attributes.

The strong irrelevance of RISK and the unexpected sign of SIZE have important implications that go beyond the relevance of the STT as both variables are explicitly considered to be important factors in the widely reported supply constraints of debt. For any credit applica-tion, one can expect that firm size correlates positively with transparency and negatively with implied informational asymmetry, while business risk correlates negatively with estimated future debt service. The measured lack of correlation with leverage may imply three possible

63

explanations: (1) Using log sales as size and the standard deviation of ROA as risk proxy is not appropriate in CEE countries; this possibility is less probably taken into account in numerous previous studies (e.g., Booth et al., 2001). (2) Lenders, regardless of their identity, do not strictly follow the conventional rules in risk assessment that most capital structure theories consider to be fundamental. This reasoning can be supported by the idea that local companies can use numerous alternative forms of debt, and these forms of intercompany or personal lending often bypass the conventional system. When managerial ownership is high and/or there is a substantial foreign economic interest in the firm (all CEE economies are relatively small, except Poland, and open), there is more likely a shift between the party who contracts the external debt and, therefore, is subject to external risk assessment and the party who finally uses the funds. (3) This explanation derives from the economic cyclicality that largely influences the banking sector’s lending policy. In an economic period when histori-cally high lending activity was further fuelled by impressive economic growth region-wide, the growing risk appetite eased the well-known constraints of bank financing (Avarmaa et al., 2011). Our survey confirms the belief that most firms did not encounter difficulties in the course of their loan applications during the period investigated.

Table 4 Regression results of the attribute-based sub-samplesTarget leverage Management Ownership Management Culture

Yes No High Low Foreign DomesticObs. 116 318 120 308 202 284

Coeff. t stat Coeff. t stat Coeff. t stat Coeff. t stat Coeff. t stat Coeff. t statC 0.39 1.80 0.49 2.89 0.13 0.31 0.46 3.63 1.18 3.02 0.06 0.20TAX 0.00 0.62 0.00 -0.78 -0.03 -1.32 0.00 0.55 -0.04 -1.17 0.00 0.47ROA 0.05 0.59 -0.28 -2.62 -0.06 -0.52 -0.44 -8.16 -0.27 -2.66 -0.21 -5.70RISK -0.09 -0.16 -0.16 -1.09 -0.38 -1.22 0.28 0.97 -0.25 -0.72 0.08 0.21SIZE 0.00 -0.43 -0.02 -2.17 0.01 0.33 -0.01 -1.83 -0.05 -2.24 0.01 0.87D_SIZE -0.02 -1.24 0.06 2.08 0.04 1.02 0.03 1.11 0.07 2.66 0.00 -0.12TANG 0.18 6.75 0.29 2.85 0.38 7.12 0.16 1.45 0.12 1.40 0.31 4.63CASH -0.38 -1.79 -0.21 -3.26 -0.14 -0.69 -0.32 -6.68 -0.29 -2.93 -0.22 -4.61DEPR -1.81 -3.11 -0.91 -3.39 -0.46 -0.99 -1.17 -5.66 -1.45 -8.06 -0.89 -3.19INTAN 1.10 2.12 1.01 5.88 -1.32 -2.48 1.00 6.92 0.01 0.02 1.14 3.39Adj. R-square 0.06 0.14 0.13 0.12 0.10 0.17

Table 4 reports the results of the fixed effect panel regression of the total leverage ratio for attribute specific sub-samples.

Unlike TAX and RISK, asset tangibility shows a strong positive relationship with leverage (Table 1 and 4), emphasizing the fact that long-term tangible assets could still serve as pri-mary collateral and that lending decisions are closely related to the nature of assets. This consideration is also highly dominant among CFOs, as out of the factors influencing the ap-propriate amount of borrowing, the highest importance is attributed to cash flow generated from assets to be financed (Table 2 and Appendix 1 and 3). Although TANG is undoubtedly a very strong positive determinant of leverage, its overall impact can vary notably between firm classes: TANG has a more significant impact on leverage in firms managed by insiders than in firms managed by non-owner-managers and it plays a more crucial role in firms operated with a local management culture.

The role of DEPR in borrowing decisions is ambiguous. Regression results (in Table 1) show that it is negatively correlated with leverage, which we apriori expected, assuming that the STT holds. Yet, little more than 50% of CFOs denote this factor to be important when we ex-



plicitly ask them about this issue (Table 2). Aside from DEPR, whose impact is only moder-ately supported by the survey, TANG is the sole conventional factor consistent with the STT, as neither profitability nor firm’s growth prove to be in accordance with the theory. Although profitability is significantly correlated with leverage, the estimated relationship is negative, which fails to prove that the main driver behind borrowing decisions is the goal of minimiz-ing corporate tax. Firms with higher profitability are more likely to use equity to finance their operations, which directly supports POT and is consistently backed by survey responses. The negative coefficient of CASH also underpins the basic principle of POT in line with CFOs’ explicit preferences for retained earnings.

Moderate support for the STT is exhibited in the attribute-specific regressions (Table 4), as the coefficient of ROA becomes non-negative only for firms with target leverage, although the estimation is not significant at 10%. Still, this result contrasts sharply with all the other samples for which we find a strongly significant negative ROA coefficient; moreover, the es-timated CASH and DEPR show a considerably weaker and a rather stronger negative effect, respectively, for leverage-targeting firms.

While the coefficients of ROA, TAX and RISK do not support the STT in the broad sample (Table 1), this does not imply that tax optimization is of secondary importance among firms’ objectives. Private companies are often given more flexibility in adjusting their cost struc-ture and optimizing taxes than are listed companies, whose actions are under permanent and thorough market control. Unlike listed firms, private firms can find several alternative ways to minimize their tax burden without increasing leverage, though some methods may threaten management’s position if the business takes a turn for the worse. This conservatism in bor-rowing decisions is clearly represented by insider CFOs.

The growth rate is positively related to leverage; this relationship explains the fact that credi-tors were willing to finance corporate sector growth during the period analyzed, consistent with banking sector statistics and with CFOs’ answers indicating that the majority of loan applications were approved. The estimated D_SIZE provides support for POT in the broad sample (Table 1), showing that dynamically expanding firms are willing and able to borrow externally if they are not profitable enough to build cash reserves.

INTAN’s positive coefficient raises another interesting issue. Titman and Wessels (1988) consider intangible assets as a negative tangibility that behaves like an inverse proxy for the collateral value of assets so negative correlation could be expected with leverage. Our results (Table 1 and 4) do not confirm this idea. Similarly, the positive correlation does not support the expectation that intangibility is generally a good indicator of a firm’s R&D activity (Long and Malitz, 1985), as one could expect that riskier assets, such as future growth opportunities incorporated in the value of present R&D expenses, are difficult to borrow against.

7. Concluding remarksWe test capital structure theories with two distinct methods in ten CEE countries. The first method we apply is a panel regression for the period from 2005 to 2008 for total financial leverage. The second method is a questionnaire-based test where the CFOs of the same 498 firms are questioned on their motivation in financing decisions. Furthermore, we create sub-samples from the overall sample to determine whether the results are robust in groups of

65

firms with specific attributes. We find that CFOs present rather strong pecking order-driven behavior, with a limited role for target leverage. Tradeoff-consistent behavior is more per-ceivable among firms with fixed target leverage. We do not find that the two rival theories are mutually exclusive. Among firm-specific attributes, the extent of managerial ownership has a strong impact on executives’ preferences and leads toward conservatism, while results do not confirm that a foreign influence in firms’ management has a strong impact on their financing decisions.

One of the main conclusions we draw is that the two independent methodologies give similar results; therefore, a robust picture can be drawn of the underlying motivations and preferenc-es of managerial behavior, and the regression equation gives a reliable prediction for firms’ financial behavior in CEE countries. Further investigation of the capital structure choices of CEE firms through creation of samples based on other firm-level attributes may be a promis-ing direction for future research.

The survey results show that a more qualitative approach could deliver useful information for a better understanding of capital structure decisions. From this analysis, we learn that the impact of country dummies might cover much more than the simple fact of belonging to a given country. Romanian firms, for example, tend to prefer lower leverage compared to other CEE countries, but this is probably not because they are from Romania. Rather, it is likely because a higher proportion of them are governed by owner-managers that follow a more conservative financing policy.



ReferencesAng, J. S. (1991). ‘Small business uniqueness and the theory of financial management’, Jour-

nal of Small Business Finance, 1(1), pp. 1-13. In Storey D. J. (ed.) Small business: critical perspectives on business and management edited, Routledge, 2000, pp. 574-586.

Avarmaa, M., Hazak, A., Mannasoo, K. (2011). ‘Capital structure formation in multinational and local companies in the Baltic States’, Baltic Journal of Economics, 11, pp. 125-145.

Berger, A. N., Udell, G. F. (1998). ‘The economics of small business finance: The role of private equity and debt markets in the finance growth cycle’, Journal of Banking and Finance, 22, pp. 613-673.

Berggren, B., Olofsson, C., Silver, L. (2000). ‘Control aversion and the search for external financing in Swedish SMEs’, Small Business Economics, 15, pp. 233-242.

Booth, L., Aivazian, V., Demirguc-Kunt, A, Maksimovic, V. (2001). ‘Capital Structures in Developing Countries’, Journal of Finance, 56, pp. 87-130.

Brounen, D., de Jong, A., Koedijk, K. (2006). ‘Capital structure policies in Europe: Survey evidence’, Journal of Banking and Finance, 30, pp. 1409-1442.

Chen, J. J. (2004). ‘Determinants of capital structure of Chinese-listed companies’, Journal of Business Research, 57, pp. 1341-1351.

Chittenden, F., Hall, G., Hutchinson, P. (1996). ‘Small firm growth, access to capital markets and financial structure: Review of issues and an empirical investigation’, Small Business Economics, 8, pp. 59-67.

Crnigoj, M., Mramor, D. (2009). ‘Determinants of Capital Structure in Emerging European Economies: Evidence from Slovenian Firms’, Emerging Markets Finance and Trade, 45, pp. 72-89.

Crnigoj, M. (2010). ‘Risk Averse Insiders with Specific Objective Function and Capital Struc-ture Choice in European Emerging Economies’, (August 11, 2010). Available at SSRN: http://ssrn.com/abstract=1100534

Delcoure, N. (2007). ‘The determinants of capital structure in transitional economies’, Inter-national Review of Economics and Finance, 16, pp. 400-415.

Demirgüç-Kunt, A., Maksimovic, V. (1998). ‘Law, finance, and firm growth’. Journal of Fi-nance, 53, pp. 2107–2137.

De Haan, L., Hinloopen, J. (2003). ‘Preference hierarchies for internal finance, bank loans, bond and share issues: evidence for Dutch firms’, Journal of Empirical Finance, 10, pp. 661-681.

Desai, M. A., Gompers, P. A., Lerner, J. (2003). ‘Institutions, Capital Constraints and Entre-preneurial Firm Dynamics: Evidence from Europe’, NBER Working Paper No. W10165.

Fama, E. F., French, K. R. (2002). ‘Testing tradeoff and pecking order predictions about divi-dends and debt’, Review of Financial Studies 15, pp. 1-33.

Friend, I., Lang, L. H. P. (1988). ‘An empirical test of the impact of managerial self-interest on corporate capital structure’, Journal of Finance, 43, pp. 271-281.

Gaud, P., Hoesli, M., Bender, A. (2007). ‘Debt-equity choice in Europe’, International Re-view of Financial Analysis, 16, pp. 201-222.

Graham, J. R., Harvey, C. R. (2001). ‘The Theory and Practice of Corporate Finance: Evi-dence from the Field’, Journal of Financial Economics, 60, pp. 187-243.

Hovakimian, A., Opler, T., Titman, S. (2001). ‘The Debt-Equity Choice’, Journal of Financial and Quantitative Analysis, 36, pp. 1-24.

67

Hutchinson, J., Xavier, A. (2006). ‘Comparing the Impact of Credit Constraints on the Growth of SMEs in a Transition Country with an Established Market Economy’, Small Business Economics, 27, pp. 169-179.

Hernadi, P., Ormos, M. (2012). ‘Capital Structure and its Choice in Central and Eastern Eu-rope.’, Acta Oeconomica, 62, pp. 229–263.

Jensen, M. C., Meckling, W. H. (1976). ‘Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure’, Journal of Financial Economics, 3, pp. 305-360.

KPMG Corporate and Indirect Tax Rate Survey (2009). Published annually by KPMG Inter-national’s Tax Division, October 2009, 11-14.

Long, M. S., Malitz, I. B. (1985). ‘The investment-financing nexus: Some empirical evi-dence’, Midland Corporate Finance Journal, 3, pp. 53-59.

Michaelas, N., Chittenden, F., Poutziouris, P. (1999). ‘Financial policy and capital structure choice in UK SMEs: Empirical evidence from company panel data’, Small Business Eco-nomics, 12, pp. 113-130.

Miller, M. H. (1977). ‘Debt and Taxes’, Journal of Finance, 32, pp. 261-75.Modigliani, F., Miller, M. H. (1958). ‘The Cost of Capital, Corporate Finance and the Theory

of Investment’, American Economic Review, 48, pp. 261-97.Modigliani, F., Miller, M. H. (1963). ‘Corporate Income Taxes and the Cost of Capital: A

Correction’, American Economic Review, 53, pp. 433-43.Myers, S. C., (1977). ‘Determinants of Corporate Borrowing’, Journal of Financial Econom-

ics, 5, pp. 147-175.Myers, S. C. (1984). ‘The capital structure puzzle’, Journal of Finance, 39, pp. 575-592.Myers, S. C., Majluf, N. S. (1984). ‘Corporate financing and investment decisions when

firms have information that investors do not have’, Journal of Financial Economics, 13, pp. 187-221.

Nivorozhkin, E. (2002). ‘Capital structures in emerging stock markets: the case of Hungary’, The Developing Economies, 40, pp. 166-187.

Nivorozhkin, E. (2004). ‘The Dynamics of Capital Structure in Transition Economies’, Eco-nomics of Planning, 37, pp. 25-45.

Nivorozhkin, E. (2005). ‘Financing choices of firms in EU accession countries’, Emerging Markets Review, 6, pp. 138-169.

Rajan, R. G., Zingales, L. (1995). ‘What Do We Know about Capital Structure? Some Evi-dence from International Data’, Journal of Finance, 50, pp. 1421-1460.

Rajan, R., Zingales, L., (1998). ‘Financial dependence and growth’, American Economic Review 88, pp. 559–573.

Rajan, R., Zingales, L., (2001). ‘Financial systems, industrial structure and growth’, Oxford Review of Economic Policy 17.

Titman, S., Wessels, R. (1988). ‘The Determinants of Capital Structure Choice’, Journal of Finance, 43, pp. 1-19.



App

endi

x 1

Sum

mar

y of

surv

ey re

sults

for t

he c

ount

ry su

b-sa

mpl

es (p

art 1

)

Cze

ch R

epub

licH

unga

ryPo

land

Rom

ania

Num

ber o

f obs

erva

tions

7273

180

68W

hat f

acto

rs a

ffect

how

you

det

erm

ine

the

appr

opria

te a

mou

nt o

f de

bt fo

r you

r firm

?Im

porta

nce

Mea

nIm

porta

nce

Mea

nIm

porta

nce

Mea

nIm

porta

nce

Mea

nLo

wH

igh

Low

Hig

hLo

wH

igh

Low

Hig

h

Proj

ecte

d ca

sh-fl

ow fr

om th

e as

sets

to b

e fin

ance

d30

.869

.22.

8644

.655

.42.

5725

.075

.02.

9422

.177

.93.

28

The

debt

leve

ls o

f oth

er fi

rms i

n th

e in

dust

ry39

.360

.72.

5476

.923

.11.

6864

.935

.12.

1850

.050

.02.

52

The

pote

ntia

l cos

ts o

f ban

krup

tcy

or fi

nanc

ial d

istre

ss50

.849

.22.

4661

.838

.22.

1246

.253

.82.

4720

.979

.13.

15

The

corp

orat

e ta

x ra

te61

.738

.32.

2076

.123

.91.

7951

.648

.42.

4630

.969

.12.

82

The

leve

l of d

epre

ciat

ion

and

othe

r non

-deb

t tax

shie

lds

59.7

40.3

2.16

55.

244

.82.

28 3

4.6

65.4

2.78

41.

858

.22.

63

Doe

s you

r firm

hav

e a

targ

et v

alue

for t

he le

vera

ge ra

tio?

noye

sno

yes

noye

sno

yes

85.9

14.1

72.3

27.7

61.9

38.1

81.8

18.2

If „

Yes”

W

e us

ually

/per

man

ently

dep

art f

rom

it (fl

exib

le)

28.6

29.4

6.6

30.0

W

e oc

casi

onal

ly d

ep. f

rom

it (s

omew

hat t

ight

)28

.611

.845

.950

.0

We

stric

tly re

spec

t it

42.9

58.8

47.5

20.0

Wha

t is t

his t

arge

t val

ue (b

ook

debt

/tota

l ass

ets)

? 44

.4

1

8.3

45.6

16

.9

Whi

ch o

f the

follo

win

g so

urce

s of l

ong-

term

fund

s are

/ w

ould

be

impo

rtant

for fi

nanc

ing

new

inve

stm

ents

?Im

porta

nce

Mea

nIm

porta

nce

Mea

nIm

porta

nce

Mea

nIm

porta

nce

Mea

nLo

wH

igh

Low

Hig

hLo

wH

igh

Low

Hig

h

Ret

aine

d ea

rnin

gs47

.852

.22.

5534

.365

.72.

7919

.680

.43.

1822

.177

.93.

01

Res

truct

urin

g as

sets

53.0

47.0

2.39

59.2

40.8

2.15

41.9

58.1

2.65

40.0

60.0

2.55

St

raig

ht d

ebt

73.5

26.5

1.91

35.7

64.3

2.94

57.2

42.8

2.30

35.3

64.7

2.87

C

onve

rtibl

e bo

nd95

.54.

51.

2198

.61.

41.

1685

.914

.11.

4879

.420

.61.

75

Exte

rnal

com

mon

equ

ity 7

8.3

21.7

1.71

77.

522

.51.

75 6

3.5

36.5

2.17

61.

838

.22.

10D

urin

g th

e la

st th

ree

year

s, di

d yo

ur fi

rm a

pply

for n

ew lo

ans o

r ca

pita

l lea

ses?

noye

sno

yes

noye

sno

yes

85.9

14.1

85.9

14.1

85.9

14.1

85.9

14.1

If „

Yes”

, the

se a

pplic

atio

ns w

ere

al

way

s app

rove

d90

.792

.393

.096

.6

alw

ays d

enie

d0.

03.

80.

91.

7

som

etim

es a

ppro

ved

and

som

etim

es d

enie

d 9

.33.

86.

11.

7

App

endi

x 1

pres

ents

the

surv

ey q

uest

ions

and

ans

wer

s for

the

coun

try su

b-sa

mpl

es (r

esul

ts o

f the

full

sam

ple

are

pres

ente

d in

Tab

le 2

). Fi

gure

s sho

w th

e m

ean

scor

e of

impo

rtanc

e/lik

elih

ood

for e

ach

ques

tion

and

the

dist

ribut

ion

of a

nsw

ers ‘

1’ a

nd ‘2

’, cl

assi

fied

as ‘l

ow’,

and

‘3’ a

nd ‘4

’, cl

assi

fied

as ‘h

igh’

.

69A

ppen

dix

2 Su

mm

ary

of su

rvey

res

ults

for

the

coun

try

sub-

sam

ples

(par

t 2)

C

zech

Rep

ublic

Hun

gary

Pola

ndR

oman

iaN

umbe

r of o

bser

vatio

ns72

7318

0 68

Wha

t per

cent

of e

quity

is o

wne

d by

the

top

thre

e of

ficer

s?

<5

%77

.171

.287

.323

.9

5-10

%2.

90.

04.

91.

5

10-2

0%1.

43.

01.

414

.9

>20%

18.6

25.8

6.3

59.7

Man

agem

ent c

ultu

re d

omin

ance

Loca

lFo

reig

nL

FL

FL

F56

.443

.746

.653

.464

.435

.661

.238

.8

Doe

s you

r firm

take

adv

anta

ge o

f ope

ratin

g le

ase?

no

yes

noye

sno

yes

noye

s49

.350

.765

.834

.245

.254

.875

.025

.0In

per

iods

whe

n th

e fir

m’s

inve

stm

ent i

s low

, doe

s you

r firm

re

tain

a p

art o

f its

free

cas

h-flo

ws?

noye

sno

yes

noye

sno

yes

53.5

46.5

47.1

52.9

63.5

36.5

42.6

57.4

Giv

en a

n in

vest

men

t tha

t cou

ld n

ot b

e ta

ken

with

out m

odify

ing

the

actu

al b

alan

ce sh

eet s

truct

ure,

wha

t act

ion

wou

ld y

ou ta

ke?

Like

lihoo

dM

ean

Like

lihoo

dM

ean

Like

lihoo

dM

ean

Like

lihoo

dM

ean

Low

Hig

hLo

wH

igh

Low

Hig

hLo

wH

igh

D

evia

te fr

om th

e ac

tual

cap

ital s

truct

ure

53.0

47.0

2.36

31.9

68.1

2.88

66.7

33.3

2.09

29.2

70.8

2.97

C

ut th

e di

vide

nds

68.7

31.3

1.91

52.2

47.8

2.45

38.8

61.2

2.78

28.8

71.2

2.92

R

estru

ctur

e as

sets

39.7

60.3

2.54

47.1

52.9

2.37

44.3

55.7

2.56

43.9

56.1

2.59

Fo

rgo

the

inve

stm

ent o

ppor

tuni

ty62

.737

.32.

1075

.424

.61.

8759

.240

.82.

3284

.815

.21.

86If

you

forw

ent t

he in

vest

men

t opp

ortu

nity

, why

?

in o

rder

to h

old

inde

pend

ence

62.5

37.5

2.08

52.9

47.1

2.35

39.2

60.8

2.81

40.0

60.0

2.60

in

ord

er to

kee

p m

oder

ate

leve

l of l

ever

age

28.0

72.0

2.92

16.7

83.3

3.06

36.0

64.0

2.79

10.0

90.0

3.00

in

ord

er to

kee

p th

e se

nior

shar

ehol

ders

’ val

ue

50.0

50.0

2.38

68.8

31.3

1.81

55.9

44.1

2.40

50.0

50.0

2.60

ex

tern

al e

quity

fina

ncin

g is

una

vaila

ble

56.5

43.5

2.30

70.6

29.4

2.00

50.0

50.0

2.46

10.0

90.0

3.20

bo

rrow

ing

is im

poss

ible

72.0

28.0

1.92

93.8

6.3

1.44

52.1

47.9

2.41

10.0

90.0

3.20

de

bt se

rvic

e is

not

exp

ecte

d to

be

satis

fied

53.8

46.2

2.35

52.9

47.1

2.12

55.7

44.3

2.39

20.0

80.0

3.10

App

endi

x 2

pres

ents

the

surv

ey q

uest

ions

and

ans

wer

s for

the

coun

try su

b-sa

mpl

es (r

esul

ts o

f the

full

sam

ple

are

pres

ente

d in

Tab

le 3

). Fi

gure

s sho

w th

e m

ean

scor

e of

impo

rtanc

e/lik

elih

ood

for e

ach

ques

tion

and

the

dist

ribut

ion

of a

nsw

ers ‘

1’ a

nd ‘2

’, cl

assi

fied

as ‘l

ow’,

and

‘3’ a

nd ‘4

’, cl

assi

fied

as ‘h

igh’

.



App

endi

x 3

Sum

mar

y of

surv

ey r

esul

ts fo

r th

e at

trib

ute-

base

d su

b-sa

mpl

es (p

art 1

)

Ta

rget

leve

rage

Man

agem

ent O

wne

rshi

pM

anag

emen

t Cul

ture

Yes

No

Hig

h Lo

wFo

reig

nD

omes

ticN

umbe

r of o

bser

vatio

ns11

631

812

030

820

228

4W

hat f

acto

rs a

ffect

how

you

det

erm

ine

the

appr

opria

te a

mou

nt o

f deb

t fo

r you

r firm

?Im

porta

nce

Low

Hig

hL

HL

HL

HL

HL

HPr

ojec

ted

cash

-flow

from

the

asse

ts to

be

finan

ced

25.7

74.3

27.1

72.9

18.0

82.0

27.9

72.1

28.5

71.5

25.4

74.6

The

debt

leve

ls o

f oth

er fi

rms i

n th

e in

dust

ry59

.640

.456

.343

.855

.944

.158

.641

.460

.139

.956

.443

.6Th

e po

tent

ial c

osts

of b

ankr

uptc

y or

fina

ncia

l dis

tress

41.8

58.2

48.2

51.8

33.9

66.1

48.3

51.7

49.7

50.3

41.0

59.0

The

corp

orat

e ta

x ra

te44

.555

.553

.946

.147

.752

.352

.247

.850

.649

.449

.051

.0 T

he le

vel o

f dep

reci

atio

n an

d no

n-de

bt ta

x sh

ield

s34

.965

.148

.951

.148

.251

.843

.456

.644

.455

.643

.856

.3D

oes y

our fi

rm h

ave

a ta

rget

val

ue fo

r the

leve

rage

ratio

?no

yes

noye

sno

yes

noye

sno

yes

noye

s0

100

100

081

.118

.970

.030

.073

.626

.472

.227

.8If

„Ye

s” W

e us

ually

/per

man

ently

dep

art f

rom

it (fl

exib

le)

15.9

35.0

12.5

19.5

15.1

If „

Yes”

We

occa

sion

ally

dep

. fro

m it

(som

ewha

t tig

ht)

43.9

45.0

38.8

43.9

43.8

If „

Yes”

We

stric

tly re

spec

t it

40.2

20.0

48.8

36.6

41.1

Wha

t is t

his t

arge

t val

ue (b

ook

debt

/tota

l ass

ets)

?

37.8

22

.9

39.6

34

.5

39.9

Whi

ch o

f the

follo

win

g so

urce

s of l

ong-

term

fund

s wou

ld b

e im

porta

nt

for fi

nanc

ing

new

inve

stm

ents

?Im

porta

nce

Low

Hig

hL

HL

HL

HL

HL

HR

etai

ned

earn

ings

23.5

76.5

30.1

69.9

22.6

77.4

28.9

71.1

28.6

71.4

25.4

74.6

Res

truct

urin

g as

sets

40.7

59.3

46.2

53.8

41.6

58.4

44.0

56.0

43.2

56.8

43.3

56.7

Stra

ight

deb

t50

.449

.650

.349

.731

.368

.758

.141

.953

.446

.648

.052

.0C

onve

rtibl

e bo

nd84

.016

.087

.912

.183

.017

.089

.310

.789

.210

.885

.814

.2 E

xter

nal c

omm

on e

quity

63.0

37.0

68.0

32.0

70.1

29.9

65.4

34.6

67.9

32.1

66.2

33.8

Dur

ing

the

last

thre

e ye

ars,

did

your

firm

app

ly fo

r new

loan

s or c

apita

l le

ases

? no

yes

noye

sno

yes

noye

sno

yes

noye

s

29.3

70.7

31.4

68.6

13.7

86.3

37.9

62.1

34.8

65.2

28.5

71.5

App

endi

x 3

pres

ents

the

surv

ey q

uest

ions

and

ans

wer

s fo

r the

attr

ibut

e-ba

sed

sub-

sam

ples

(res

ults

of t

he fu

ll sa

mpl

e ar

e pr

esen

ted

in T

able

2).

Figu

res

show

the

mea

n sc

ore

of

impo

rtanc

e/lik

elih

ood

for e

ach

ques

tion

and

the

dist

ribut

ion

of a

nsw

ers ‘

1’ a

nd ‘2

’, cl

assi

fied

as ‘l

ow’,

and

‘3’ a

nd ‘4

’, cl

assi

fied

as ‘h

igh’

.

71A

ppen

dix

4 Su

mm

ary

of su

rvey

res

ults

for

the

attr

ibut

e-ba

sed

sub-

sam

ples

(par

t 2)

Targ

et le

vera

geM

anag

emen

t Ow

ners

hip

Man

agem

ent C

ultu

reYe

sN

oH

igh

Low

Fore

ign

Dom

estic

Num

ber o

f obs

erva

tions

116

318

120

308

202

284

Wha

t per

cent

of e

quity

is o

wne

d by

the

top

thre

e of

ficer

s?

<5%

75.2

63.5

0.0

94.2

73.8

64.5

5-

10%

4.0

4.0

0.0

5.8

4.1

4.0

10

-20%

5.0

6.9

20.8

0.0

5.8

5.2

>2

0%15

.825

.679

.20.

016

.326

.2

Man

agem

ent c

ultu

re d

omin

ance

Fore

ign

Loca

lF

LF

LF

LF

LF

L

40.0

60.0

41.7

58.3

32.8

67.2

44.1

55.9

100

00

100

Doe

s you

r firm

take

adv

anta

ge o

f ope

ratin

g le

ase?

noye

sno

yes

noye

sno

yes

noye

sno

yes

47.0

53.0

54.6

45.4

55.9

44.1

53.8

46.2

50.5

49.5

56.4

43.6

In p

erio

ds w

hen

the

firm

’s in

vest

men

t is l

ow, d

oes y

our fi

rm re

tain

a p

art o

f its

free

cas

h-flo

ws?

noye

sno

yes

noye

sno

yes

noye

sno

yes

52.2

47.8

57.1

42.9

48.7

51.3

58.1

41.9

57.9

42.1

57.0

43.0

Giv

en a

n in

vest

men

t tha

t cou

ld n

ot b

e ta

ken

with

out m

odify

ing

the

actu

al

BS

wha

t act

ion

wou

ld y

ou ta

ke?

Like

lihoo

dLo

wH

igh

LH

LH

LH

LH

LH

D

evia

te fr

om th

e ac

tual

cap

ital s

truct

ure

56.4

43.6

49.3

50.7

36.3

63.7

55.7

44.3

50.8

49.2

51.7

48.3

C

ut th

e di

vide

nds

39.8

60.2

46.2

53.8

31.9

68.1

49.3

50.7

43.4

56.6

46.0

54.0

R

estru

ctur

e as

sets

41.3

58.7

43.1

56.9

40.0

60.0

42.6

57.4

45.6

54.4

40.8

59.2

Fo

rgo

the

inve

stm

ent o

ppor

tuni

ty66

.733

.367

.932

.178

.321

.762

.637

.469

.930

.164

.935

.1If

you

forw

ent t

he in

vest

men

t opp

ortu

nity

, why

?

in

ord

er to

hol

d in

depe

nden

ce43

.656

.444

.056

.032

.068

.047

.852

.250

.050

.039

.860

.2

in

ord

er to

kee

p m

oder

ate

leve

l of l

ever

age

25.6

74.4

34.7

65.3

16.0

84.0

36.6

63.4

35.6

64.4

31.6

68.4

in

ord

er to

kee

p th

e se

nior

shar

ehol

ders

’ val

ue

55.3

44.7

52.7

47.3

50.0

50.0

57.4

42.6

58.6

41.4

53.8

46.2

ex

tern

al e

quity

fina

ncin

g is

una

vaila

ble

55.3

44.7

49.0

51.0

30.8

69.2

50.9

49.1

56.9

43.1

45.7

54.3

bo

rrow

ing

is im

poss

ible

46.2

53.8

63.2

36.8

34.6

65.4

61.1

38.9

64.4

35.6

53.8

46.2

de

bt se

rvic

e is

not

exp

ecte

d to

be

satis

fied

57.9

42.1

49.5

50.5

34.6

65.4

51.4

48.6

55.9

44.1

46.8

53.2

App

endi

x 4

pres

ents

the

surv

ey q

uest

ions

and

ans

wer

s fo

r the

attr

ibut

e-ba

sed

sub-

sam

ples

(res

ults

of t

he fu

ll sa

mpl

e ar

e pr

esen

ted

in T

able

3).

Figu

res

show

the

mea

n sc

ore

of

impo

rtanc

e/lik

elih

ood

for e

ach

ques

tion

and

the

dist

ribut

ion

of a

nsw

ers ‘

1’ a

nd ‘2

’, cl

assi

fied

as ‘l

ow’,

and

‘3’ a

nd ‘4

’, cl

assi

fied

as ‘h

igh’

.


73

Student debt among University of Latvia graduates: Repayment prospects under income-contingent student

loan scheme

AliAitSiMhamed1,RitaKaša2,ZaneCunska3

AbstractThis study examines the relationship between student debt and income by gender and field of study among University of Latvia graduates of 2009. Data analysis in the paper is also framed by suggestions for an income-contingent student loan scheme in Latvia, modeled after the Australian example of the Higher Education Contribution Scheme. This research models a possible student loan payment pattern in Latvia under an income-contingent student loan plan. Data analysis shows that in the current mortgage-type student loan system, graduates of both genders in all fields of study and income groups have similar student debt levels, on average. At the same time, the largest portion of student debt is held by graduates in lower in-come groups. In the context of an income-contingent student loan payment scheme, analysis reveals that women and graduates in the field of humanities, pedagogy and psychology would be likely to belong to the low payment group. Consistent with findings in other studies, this research finds that female higher education graduates have lower income than men, on aver-age, while graduates in the field of economics and business management tend to have higher income than graduates in other fields of study.

Key words: Higher education graduates, gender, student loans, student debt, income-contin-gent student loan, LatviaJEL classification: 122

1. IntroductionAny public policy should be assessed for its outcomes in order to understand how implemen-tation of the given policy impacts its target group. This also applies in the case of student financial assistance policies and more specifically in the case of governmentally supported student loan programs. One question that is frequently studied in this respect in the context of different countries addresses the issues of equity and higher education access (e.g., Texeira, Johnstone, Rosa, & Vossensteyn, 2006). Another set of related literature is grouped around the question about indebtedness of graduates and student loan repayment (e.g., Andrew, 2010; Cataldi et al., 2011; Chapman, Lounkaew, Polsiri, Sarachitti, & Sitthipongpanich, 2010).

Due to its focus on University of Latvia graduates in terms of their indebtedness and pros-pects for loan repayment, the current paper belongs to the second body of literature. Using 1 Dr. Ali Ait Si Mhamed is an Assistant Professor at the Department of Adolescent Education, Canisius College, 2001 Main Street, Buffalo, New York 14228 USA. Phone number: 716-888-3732. Email: [email protected] 2 Dr. Rita Kaša is a Research Fellow at Stockholm School of Economics in Riga, Strelnieku 4a, Riga, LV 1010 Latvia. Phone number: 37167015840. Email: [email protected] Dr. Zane Cunska is a Research Fellow at Baltic International Center for Economic Studies, Strelnieku 4a, Riga, LV 1010 Latvia. Email: [email protected]

StudentdebtamongUniversityofLatviagraduates:Repaymentprospectsunderincome-contingentstudentloanscheme


University of Latvia graduate survey data of 2009, this study examines the amount of student debt in relation to income, gender, and field of study among University of Latvia graduates. It also models how graduates by gender and field of study would be distributed by the amount of student debt payments under an income-contingent student loan repayment scheme – which graduates would be likely to make higher payments and which graduates would belong to the low payments group based on their income.

There has never been a publicly funded income-contingent student loan plan in Latvia which would link the amount of loan payments to the income of graduates. The governmentally sub-sidized student loan program, which is the largest student loan program in the country, has al-ways provided so-called mortgage-type loans with a fixed repayment schedule (Kasa, 2006). However, there is an ongoing discussion about the need to reform the financing of higher education in Latvia. One of the options under discussion is the introduction of an income-contingent student loan system modeled after the Higher Education Contribution Scheme (HECS) in Australia (Dombrovsky, 2011). Under this scheme, all students would be required to pay tuition. Students who do not pay tuition up-front and borrow from the government’s student loan program to cover their higher education expenses would have to repay the loan to the state after completing their studies. The repayment would be linked to the graduate’s in-come. That is, payments would have to be made only when a certain income level is reached and would be of varying size, depending on the graduate’s earnings.

This would be a very different approach from the current model of student funding in Latvia in which 64% of students pay tuition while 36% primarily at public institutions of higher education study free of charge in so called ‘budget places’ entirely funded by the government (Ministry of Education and Science, 2011). Introduction of an encompassing income-contin-gent student loan program would be likely to increase the number of graduates with student loans if the current allocation of governmentally funded study places to public institutions of higher education were abolished. In 2011, governmentally guaranteed student loans for covering tuition were issued to 11% of all students in Latvia or to 18% of students admitted to tuition funded study places (Studiju un zinātnes administrācija, 2011). In the case of introduc-ing a governmentally supported income-contingent student loan scheme modeled after the Australian example as described in Dombrovsky (2011), all students who have their higher education expenses covered by the government would accumulate student debt which they would have to repay as graduates based on their income.

While there are few studies examining the population of graduates with student loans in Latvia (Ait Si Mhamed & Kaša, 2010), internationally this group of higher education recipi-ents is frequently studied in order to inform public policy debates. The current study aims to contribute to knowledge about the characteristics of higher education graduates with student loans in Latvia by using data from a survey of University of Latvia graduates in 2009. This paper addresses the following research question: in the context of potential implementation of an income-contingent student loan repayment scheme, what are the differences between student debt level and income of University of Latvia graduates by field of study and gender? This research question is framed by studies which find differences in the income level of higher education graduates by gender (Baum & O’Malley, 2003; Krūmiņš et al., 2007; Zepa et al., 2006) as well as varying income expectations by field of study (Krūmiņš et al., 2007; Minicozzi, 2005). Given that an income-contingent student loan scheme is in principle a

75

progressive taxing of income, where those who have higher income make higher student debt repayments, this paper will examine how the variables of gender and field of study will impact the income-contingent loan repayment pattern in Latvia.

2. Review of Literature Variations of student loan schemesAbout 75 various student loan schemes are implemented world-wide (Shen & Ziderman, 2009). Generally three types of student loan schemes can be distinguished by type of repay-ment. One type is fixed-schedule, also called conventional or mortgage-type loans. In this scheme, the monthly schedule of repayments, the interest rate, and the repayment period are all fixed in the loan repayment agreement (Johnstone, 2009). The second type is income-contingent loans, where graduate tax can be viewed as a variation of the income-contingent approach to student funding. In the income-contingent loan program student debt is cleared once the principal amount of the student loan and the interest rate attached are repaid. Under the graduate-tax scheme, graduates generally become obliged to income surtax for the rest of their earning lifetime (Johnstone, 2004). The third type of student loan scheme is the hybrid, or fixed schedule-income contingent loan (Johnstone, 2004, 2009), also called the soft in-come contingent plan (Usher, 2005). In this scheme, the borrower is required to meet a fixed schedule of payments “unless the monthly or annual repayments exceed some maximum percentage of monthly or annual earnings” (Johnstone, 2009, p. 191).

Multiple rationales frame implementation of any given student loan program with govern-mental involvement, ranging from facilitating the accessibility of higher education to in-creasing the degree of cost recovery for the government. Additionally, various provisions of funding for student loans exist. Some student loan programs are funded entirely by the gov-ernment, that is, by taxpayers. In the second kind of governmentally supported student loan programs funding comes from private resources such as commercial banks or capital markets while the government or taxpayers still participate with some degree of subsidy.

In the case of Latvia, both funding approaches have been implemented in the scope of student loan programs. In 1997, when the government of Latvia launched the first generally available student loan program, these were direct mortgage-type student loans, paid from the national budget (Kasa, 2008). Following the reform in 2001, the government became a secondary guarantor in the student loan program while the principal amount of student loans was pro-vided by commercial banks. This scheme remained a mortgage-type student loan program where graduates who have borrowed are expected to repay their loan over a fixed period in the form of fixed monthly payments. In 2011, representatives of the political elite initiated a debate about reforming the higher education funding system in Latvia and introducing an income-contingent student loan system similar to HECS in Australia as discussed in Dom-brovsky (2011). It should be mentioned that Australia was the first country to establish a universal income-contingent loan system for students in higher education in 1989 (Berlinger, 2009; Chapman, 2006; ICHEFAP, 2006). Many more countries have implemented similar student loan programs since then.

One of the most recent countries in Central Eastern Europe to implement such a scheme is Hungary, where an income-contingent student loan scheme was introduced in 2001 (Ber-



linger, 2009). As Berlinger (2009) describes it, Hungary has a student loan company, set up by the government, which provides non-subsidized governmentally guaranteed student loans that are paid directly to student bank accounts, financed through resort to the capital market. Student loan repayments are fixed as a percentage of individual graduate income. Borrowers have to repay six percent of the minimum wage or own income, whichever is higher. Borrow-ers have to repay in every instance, even if their income is less than the minimal wage. This approach allows constant contact with the borrower, helps to increase recovery of student loans, while the burden of repayment on the borrower is predictable and low in the case of low income. In contrast to this approach is the one implemented in the United States. In the US all student loan repayment schemes submit to the principle of protected income. This is an acknowledgement that some borrowers are too poor to pay, at least temporarily. How-ever, some researchers argue that the policy regarding the level and circumstances in which the principle of protected income is applied is not clear because of inconsistent deferment eligibility rules (Shireman et al., 2006). Different levels of protected income are established based on the borrower’s situation in terms of earnings and family size. In order to have loan repayments adjusted respectively, borrowers need to apply for “economic hardship” relief.

In a pure income-contingent student loan scheme, borrower hardship is addressed in the very design of the scheme. Advocates of income-contingent loans claim that these loans are more equitable than conventional mortgage-type student loans because they take into account graduates’ possible future poverty and thus provide them with a better safety net in case their income is too low to meet student loan payments (Barr, 2001, pp. 184 - 186). All income-contingent student loans have a provision for forgiving the remaining debts of some of the lowest earning borrowers who reach some maximum repayment period or some maximum age with a debt still outstanding (Johnstone, 2006, p. 89). Research using evidence from the UK finds that under the income-contingent student loan scheme, graduates have lower repay-ment burdens and higher taxpayer subsidies than in a mortgage-type loan system (Migali, 2010). Profiles of higher education graduates with student debt in relation to gender, field of study, and income are discussed in the following section of this paper. Profiles of graduates with student debt Studies that examine the socioeconomic effects of student loan policies focus on the rela-tionship between borrowing and student enrollment behaviors as well as their employment, income, and life decisions after graduation in relation to student debt repayment. These relationships are examined by such characteristics of graduates as income or socioeconomic status of the borrower, gender, ethnicity, and field of study.

Looking at student debt and future income, research on US graduates suggests that initial wages are higher for those who borrowed more while expecting higher wages after gradua-tion (Minicozzi, 2005). According to a study by Rosenblatt and Andrilla (2005), medical stu-dents in the US reported that higher levels of debt influenced their future career choices. Total student debt was associated with a lower likelihood of choosing a primary care career, which is more modestly compensated as compared to other medical specializations in America. At the same time, it is important to note that in the same study demographic factors such as gen-der and race appeared to provide a better explanation for choice of specialization in a medical career than accumulated student debt.

77

Variations in monetary returns to higher education were also found by Kelly, O’Connell and Smyth (2010) in their study on Irish graduates. This study found that relative to graduates of arts and humanities, graduates in medicine and veterinary, education, engineering and ar-chitecture, science and IT received higher returns (Kelly, O’Connell & Smyth, 2010). When examining the earnings of Canadian graduates, Finnie and Frenette (2003) observed low earnings among graduates in the fields of arts and humanities, agricultural and biological sci-ences, and social sciences, except for education and economics where earnings were higher. The highest earning graduates in Canada were in the fields of health, engineering and com-puter science, commerce, as well as mathematics and physics. Research on graduate income by field of study in Latvia showed that the net income of graduates in pedagogy, agriculture, health and social care and humanities belongs to the lower range of income distribution of up to about 305 LVL per month (Krūmiņš et al., 2007, p. 103).4 The highest average net monthly income reported, 469 LVL, was for graduates in architecture and construction. This evidence confirms that variations in income level by field of study are relevant when designing student loan repayment schemes.

Similarly, the gender variable is found relevant in relation to loan repayment possibilities. In their study on the relationship between income, gender and student debt, Baum and O’Malley (2003) found that gender has a strong effect on the income level of higher education gradu-ates in the US. These authors report that in 1997 median income for full-time women work-ers 25 years and older with bachelor’s degrees was slightly more than 28,000 USD. For men holding a comparable educational degree, median income was more than 41,000 USD. Similar findings that women graduates of higher education have lower income than men were also reported for Great Britain (O’Leary & Sloane, 2005), Thailand (Chapman & Lounkaew, 2009) and Canada (Finnie & Wannell, 2004). As for ability to repay student debt, in the case of Canada it was also found that twice as many women than men requested student loan interest relief due to high debt to income ratios (Human Resources and Skills Development Canada, 2004). Lower income among higher education female graduates compared to men was also found in Latvia (Krūmiņš et al., 2007).

At the same time, a study on US graduates found that gender bears no impact on the amount of student debt incurred (Price, 2004a). According to Price, the effect of being a female on the educational debt burden was positive but statistically non-significant compared to males. However, race and ethnicity in both genders as related to educational debt showed statisti-cally significant results (Price, 2004b). Price (2004b) reported that black students are more likely than white students to have an excessive educational debt burden. An ethnicity variable in relation to student borrowing in Latvia was examined by Ait Si Mhamed and Kaša (2010). They found that student loans are mostly held by ethnic Latvians compared to other ethnic groups in the country (Ait Si Mhamed & Kaša, 2010). Thus, the ethnicity variable is not in-cluded in the analysis in the current paper.

As for student loan repayment and borrower income, from the perspective of income-contin-gent student loan providers the most profitable borrowers to the system are those who belong to a fairly modest income category and achieve full repayment just at the point of retirement (Berlinger, 2009, p. 264). These conclusions are reported for Hungary. At the same time, a study on the US student borrower population finds that for graduates with low-income 4 LVL 1 equals EUR 1.42



student debt repayments cause more hardship than anticipated (Baum & O’Malley, 2003). A study by Migali (2010) involving higher education graduates in the UK shows that graduates with more uncertain incomes prefer an income-contingent student loan repayment system over a mortgage-type loan repayment. This question is yet to be examined in the context of Latvia. While the current study does not deal with the issue of which type of student loan is preferred by higher education debt holders in Latvia, it examines other variables found rel-evant in the context of student loan repayment by the literature such as income and student debt level by field of study and gender should an income-contingent student loan scheme be introduced in Latvia.

3. MethodologySampleData in this paper come from a University of Latvia graduate on-line survey conducted in the spring of 2009, in which 2,141 graduates took part. All respondents had graduated in the spring of 2009. For the analysis in this paper only those respondents who answered a ques-tion on accumulated student loan were included: 369 respondents in total. For the variable on field of study, however, the number of the respondents in the analysis is smaller. This is because in 37 cases it was difficult to identify accurate titles of the study programs for respon-dents included in this study. Those cases were excluded from the analysis, reducing the total number of observations for the field of study variable in the analysis to 332 respondents. The cases excluded represented various fields of study. More information about the representation of variables in the total survey sample and subsample used for the analysis in this paper is presented in Table 1. Table 1: Comparison of the variables in the samples of the study

Total sample (Column I)

Respondents who gave an answer on the amount of student debt

(Column II)

A comparison of sample proportions in columns II and III (two-tailed z-test)

N (%) N (%)

Field of study

Economics, business and management sciences

642 (30%) 124 (37%) p=.01

Natural sciences, IT, medical sciences

557 (26%) 72 (22%) p=.12

Humanities, pedagogy and psychology

492 (23%) 80 (24%) p=.69

Social sciences and law 450 (21%) 56 (17%) p=.09

Total 2141 (100%) 332 (100%)

Gender

Male 450 (21 %) 70 (19%) p=.38

Female 1991 (79%) 299 (81%) p=.21

Total 2141 (100%) 369 (100%)

Source: Table composed by the authors.

79

The field of study variable was re-coded for the purposes of analysis in this paper. There were multiple study programs with few graduates. In order to have a sufficient number of observations for statistical analysis, study programs were grouped in four categories without distinguishing the level of study program: (1) economics, business and management sciences; (2) natural sciences, IT, medical sciences; (3) humanities, pedagogy and psychology; (4) and social sciences and law.

The comparison of the sample proportions in the total sample (Column I) and in the sample of respondents (Column II) used for analysis in this paper shows that proportions in the samples are significantly different only for respondents in the field of economics, business and man-agement. The proportion of respondents in the field of economics, business and management sciences is larger in the sample used for the analysis in this paper than in the sample overall, p=.01. This discrepancy indicates a larger proportion of student borrowers among graduates in the field of economics, business and management sciences.

Variables for analysisThe main characteristic of an income-contingent loan scheme is that it links the amount of repayment to the income of the student debt holder. Thus, graduates with an income level below a certain threshold are required to make minimal payments or no payment at all to the loan scheme (Johnstone, 2006). As the income level of the graduate rises, so does the amount of debt repayment. Therefore, the income variable is central to the analysis in this paper.

A limitation to the dataset used in the analysis is that it asks graduates only about their net income within the last month before taking the survey. At this point of time – April to May 2009 – the graduates were still enrolled as students. However, it is important to note that it is a common practice in Latvia to combine full-time studies and full-time work especially at the level of master degree studies. Due to this particular situation, the majority of the respondents were working and receiving income, while 34 respondents said they had no income. Although the data on income in this paper are limited, they are valid for analysis on the relationship between graduate income and other variables in the study. We validate our analysis when dis-cussing our findings in relation to income; we compare our results to findings in other similar studies in order to provide a benchmark for interpreting our results.

The second variable used in the analysis is gender. This choice of variable is substantiated by a research finding that women, on average, have lower income than men (Baum & O’Malley, 2003), also in Latvia (Krūmiņš et al., 2007; Zepa et al., 2006). The current study will examine if this holds true among higher education graduates with student loans in Latvia. If so, and if an income-contingent student loan scheme were introduced, women would be among student debt holders making lower repayments to the student loan scheme.

The third variable examined for the relationship between income and student debt level is the field of study. Research shows that students entering various fields of study hold varying expectations for their income (Minicozzi, 2005). Research on different countries yields vary-ing results on highest and lowest earnings by field of study. This also holds true for Latvia. Krūmiņš et al. (2007) found that monthly net income for graduates in pedagogy, humanities, health care and social care was in the lower range of income distribution. Higher income was reported by graduates in social sciences and law, natural sciences, IT, engineering, com-



mercial and management sciences and services. Krūmiņš et al. (2007) base their results on a representative national sample of graduates while the data in the current paper represent graduates from the University of Latvia. When discussing our results, we will compare them to the findings in the earlier study by Krūmiņš et al. (2007).

The final variable included in the analysis in the paper is the reported student debt level. We will test if statistically significant differences exist between student debt accumulated by in-come, gender, and the field of study of graduates with student loans. This analysis is relevant as it will inform which groups of graduates are likely to hold higher student debt levels. The ethnicity variable will not be considered in our analysis because earlier research on higher education graduates with student debt in Latvia has shown that only an insignificant number of non-Latvians hold student debt (Ait Si Mhamed & Kaša, 2010).

Methods applied in data analysisIn order to examine the relationship between gender, field of study, income and student debt level among University of Latvia graduates, we applied several statistical techniques. We obtained chi-square values for categorical dependent variables. It should be noted that the in-come variable in the dataset used in this paper appears as a categorical variable. However, the student debt level variable was recorded as an ordinal variable allowing for ANOVA analysis of variance in order to determine variance in the student debt level among graduates.

In order to reduce the number of categories and to increase the N per cell, the field of study variable was recoded into four categories. Each of these categories includes bachelor and master degree study programs.

The results of the analysis are presented in the next section of the paper.

4. ResultsStudent debt and income by genderFrom the sample, 369 respondents indicated the amount of student loan incurred. The mini-mum amount of student debt stated was 50 LVL and the maximum – 25000 LVL, with the average amount M = 2636, SD =2790.73. Median student debt reported was 2000 LVL.

Reports on their student debt level were received from 299 women and 70 men. One- way ANOVA variance analysis showed no statistically significant differences in the amount of student debt accumulated among men and women, F(1, 367) = 0.648, p = 0.421. This result is consistent with the finding in the literature that gender has no impact on the amount of student debt incurred (Price, 2004a).

However, statistically significant differences were found in the amount of income of student debt holders by gender. On average, among respondents who reported the amount of student debt, men had higher income than women, X2(1, N = 369) = 19.76, p = .011. This is also consistent with observations in the literature about income discrepancies between men and women and allows the inference that women in Latvia have a higher student debt to income ratio compared to men.

81

Figure 1: Income levels by gender among respondents with student loans

Gender

FemaleMale

Perc

ent

20.0%

10.0%

0%

30.0%

No incom

e

< 100 LVL

100-299 LVL

300-499 LVL

> 1100 LVL

900-1099 LVL

700-899 LVL

500-699 LVL

Did not reveal

Source: Created by the authors.

Student debt and income by study programThe majority of student loan holders had acquired their degrees in full-time studies (N=341). The average debt reported by these respondents was 2714 LVL. Fewer respondents (N =26) had acquired their degree in part-time studies and their average debt was about 1440 LVL. On average, however, no statistically significant variation appeared in student debt levels for graduates in full-time study programs and in part time study programs, F(2, 364) = 2.341, p = .098. Analysis also showed that graduates in various fields of studies have similar student debt levels, F(3, 1386) = 1.533, p = .204.

At the same time, statistically significant differences were reported in income by field of study, X2(1, N = 332) = 37.31, p = .041. The largest proportion of high-income respondents, 29% with more than 500 LVL net income during the last month, were graduates in business, economics and management programs (N=124). The second largest proportion in this income group was graduates in social sciences and law (18%, N=56), followed by graduates in hu-manities, pedagogy and psychology (15%, N=80), and graduates in natural sciences, medical sciences and technology (12.5%, N=72).



Figure 2: Income levels of graduates with student loans by field of study

Degree in socialsciences

Degree in naturalsciences andtechnology

Field of study

Degree in economics,business andmanagement

Net income withinthe last month

No income< 100 LVL100-299 LVL300-499 LVL500-699 LVL700-899 LVL900-1099 LVL> 1100 LVLdon't want to reveal

Degree in humanities,pedagogy andphyschology

0

10

Num

ber o

f res

pond

ents

20

30

40


Of all respondents, 9% (N=30) said they had no income. Most of these respondents (N=12) had just graduated from a study program in the field of humanities, pedagogy and psychology. Income of up to 100 LVL was reported by 5% (N=16) respondents, the majority of whom were fresh recipients of a degree in humanities, pedagogy and psychology (N=6) and social sciences and law (N=4). In the income group of up to 300 LVL were 95 respondents (29%), most of whom, 10% (N=32) held a degree in natural sciences, medical sciences and technol-ogy. In the income group of 300 to 500 LVL were 27% (N=89) of respondents. The majority of these respondents, 11% (N=36) had just graduated in the field of economics, business and management.

For more robust conclusions on University of Latvia graduate income by field of study, it would be desirable to have a larger number of observations as well as information about part-time and full-time employment of respondents, which is not available in this dataset. Ideally, too, these data should be longitudinal. At the same time, results in the current study correspond to conclusions in an earlier study by Krūmiņš et al. (2007) on the population of higher education graduates in Latvia. The study by Krūmiņš et al. (2007) found that graduates from pedagogy, humanities, health care and social care study programs have lower income compared to graduates in social sciences, natural sciences, IT, engineering, commercial and management sciences and services. These trends are also identified by analysis in the current study on University of Latvia graduates in bachelor and master degree study programs.

83

Student debt and income among University of Latvia graduatesOne-way ANOVA analysis of variance showed that graduates of all income groups, on aver-age, hold a similar amount of student debt, F(8, 360) =.353, p = .944.

Table 2: Student debt of graduates by income as reported for spring 2009Net income within the last month before the survey

Number of respondents

Mean amount of student debt in LVL

Minimum amount ofstudent debt reported

Maximum amount ofstudent debt reported

No income 34 2770.59 500 11500

< 100 LVL 16 2498.75 500 5000

100-299 LVL 102 2651.49 75 12500

300-499 LVL 103 2666.99 200 22000

500-699 LVL 56 2951.88 300 25000

700-899 LVL 12 2558.33 500 12450

900-1099 LVL 3 1300.00 400 3000

> 1100 LVL 4 1650.00 700 2500

Don’t want to reveal 39 2232.28 50 8000

Total 369 2636.57 50 25000


At the same time, examination of descriptive statistics shows that graduates with a higher in-come level had accumulated a smaller amount of student loan, on average, compared to those in the lower income groups in the dataset. This suggests that the income level of students is related to their ability to pay upfront tuition and cover student living costs. This is also likely to hold true in the case of an income-contingent student loan scheme as students from lower income backgrounds will continue to use the deferred tuition option rather than make tuition payments upfront and also they will be in need of student loans to cover living costs.

Overall, the data in this study showed that for 43% of graduates with student loans net income was below 300 LVL; 27% said their net income was between 300 and 500 LVL; 15% said they earned between 500 and 700 LVL, and 5% reported earnings of more than 700 LVL in the month preceding the survey. Some 10% of graduates with student debt did not want to reveal their income. Thus, we can conclude that the majority of University of Latvia gradu-ates who make student loan payments belong to the modest income group.

5. DiscussionThe purpose of this paper was to examine the relationship between student debt and income by gender and field of study in order to model how graduates with student debt would be dis-tributed by these characteristics in an income-contingent student loan repayment scheme, if such a scheme is introduced in Latvia. The data source in this paper is a University of Latvia 2009 graduate survey.

Analysis of the student debt level among University of Latvia graduates in 2009 showed that, on average, graduates of both genders and all fields of study had similar student debt levels. There were no statistically significant differences in the student debt level by income of respondents either. At the same time, it was also true that graduates who reported their



income to be above 900 LVL had reported a smaller student debt amount. Due to the small number of respondents in this study, it is difficult to argue about how widespread this pattern is. If this finding holds in future studies, it will show that graduates in lower income groups hold the greatest portion of student debt in Latvia. Under an income-contingent student loan scheme this would mean that the majority of student loan repayments would also be made by graduates with low income, much as has been observed in the case of Hungary (Berlinger, 2009). On the other hand, this might not be a problem from the borrowers’ point of view since research suggests that graduates with more modest income levels prefer income-contingent student loan repayment over mortgage type student loans (Migali, 2010). At the same time, this assumption stemming from the literature should be tested among higher education gradu-ates in Latvia.

Additionally, it would be advisable for research findings regarding graduate income to be corroborated by other studies due to the limitations of the dataset used for the analysis in this paper. Data on income in the current analysis are limited to the month from April to May 2009. During the period to which the question on income referred to, respondents were either still enrolled as students or had recently graduated. Thus not only does the income variable not provide a longitudinal perspective but it also holds no information about the income of graduates after they had graduated. As to the second point, this shortcoming might be miti-gated by the fact that is it typical among students in Latvia to combine full-time studies and full-time work. Therefore, it is reasonable to expect a similar income level after graduation for those employed during studies. As for those with no employment, it is unlikely that fresh graduates who had no income during their studies would enter the job market with high com-pensation. Thus, it is reasonable to conclude that the largest portion of student debt among University of Latvia graduates rests on graduates with low and moderate income, as analysis in this paper shows.

Examining graduate income in relation to gender and field of study in the light of the plan for an income-contingent student loan scheme in Latvia we found statistically significant differ-ences in both cases.

For gender, the results of the analysis were consistent with findings in other studies (Baum & O’Malley, 2003; Krūmiņš, et al., 2007; Zepa et al., 2006) indicating that women would be likely to belong to the group of low student loan payments as their income, on average, was lower than for men.

Analysis of graduate income by field of study showed that graduates in the field of economics and business management tend to have higher incomes than graduates in other fields. Thus, this group of graduates would be likely to be among the high repayment group in an income-contingent student loan scheme in Latvia. Most of the low earning graduates, on the other hand, belonged to the study field of humanities, pedagogy and psychology.

Although information on level of income in the current study should be treated with caution, the median income for the majority of University of Latvia graduates of all fields of studies in 2009 was in the range of 100 to 500 LVL. For comparison, average net income in 2009 for people employed in Latvia was 342 LVL a month (CSB, 2012). As for higher education graduates specifically, a study by Krūmiņš et al. (2007) reported average monthly net income

85

of 322 LVL for bachelor degree holders and 424 LVL for master degree holders. A study by Ait Si Mhamed and Kaša (2010) using the same dataset as in Krūmiņš et al. (2007) found no differences in the income of graduates with and without student loans in Latvia. All this evi-dence combined allows the inference that, should an income-contingent student loan scheme be introduced in Latvia, the majority of student debt payments would belong to the income group in the vicinity of 300 to 400 LVL net.

6. ConclusionsThe purpose of this paper was to examine the relationship between student debt level and income by gender and field of study among University of Latvia graduates of 2009. Analysis in this paper was framed by a suggestion to introduce an income-contingent student loan scheme in Latvia modeled after HECS n Australia, as discussed by Dombrovsky (2011).

For student debt levels under the current mortgage-type student loan scheme, no difference was found in debt levels by gender, field of study or income. At the same time, analysis in this paper indicated that the largest share of student debt is held by graduates who have lower income. The relationship between student debt amount and graduate income should be exam-ined more closely in future studies as the sample used in this study was limited to University of Latvia graduates only as well as containing limited information on income level over time. As for the relationship between income, gender and field of study, the study found that should an income-contingent student loan scheme be introduced and should the overall situation remain as it is – with no changes in student enrollment and borrowing behavior – women and graduates in the field of humanities, pedagogy and psychology would belong to the low-est income group in terms of student loan payments. This conclusion based on a sample of University of Latvia graduates is consistent with findings in the literature. At the same time, a more extensive study would be advised to replicate the finding for sound public policy deci-sions regarding student loan scheme reforms in Latvia. Additionally, an inquiry into the views of higher education graduates on student loan payment mechanisms in Latvia would be sug-gested as international evidence shows that graduate preferences for different student loan re-payment modes are framed by socioeconomic characteristics such as income and education.



ReferencesAit Si Mhamed, A. & Kaša, R. (2010). Comparing tertiary graduates with and without student

loans. BalticJournalofEconomics,10(2), 49 – 62.Andrew, M. (2010). The changing route to owner occupation: The impact of student debt.

HousingStudies,25(1), 39 – 62.Barr, N. (2001). Thewelfarestateaspiggybank:Information,risk,uncertainty,andtherole

ofthestate. New York, NY: Oxford University Press.Baum, S., & O’Malley, M. (2003). College on credit: How borrowers perceive their educa-

tion debt. NASFAAJournalofStudentFinancialAid,33(3), 7 – 19. Berlinger, E. (2009). An efficient student loan system: Case study of Hungary. HigherEduca-

tioninEurope,34(2), 257 – 267.Chapman, B. (2006). Income related student loans. In P. Teixeira, B. Johnstone, M. Rosa, &

H. Vossensteyn (eds), Cost-sharingandaccessibilityinhighereducation:Afairerdeal?(pp. 79 - 105). Dordrecht: Springer.

Chapman, B., & Lounkaew, K. (2009). The effects for different loan schemes for higher education tuition: An analysis of rates of return and tuition revenue in Thailand. HigherEducationinEurope,34(2), 211 – 225.

Chapman, B., Lounkaew, K., Polsiri, P., Sarachitti, R., & Sitthipongpanich, T. (2010). Thai-land’s Student Loans Fund: Interest rate subsidies and repayment burdens. EconomicsofEducationReview, 29(5), 685-694.

Cataldi, E., Green, C., Henke, R., Lew, T., Woo, J., Shepherd, B., & Siegel, P. (2011). 2008–09BaccalaureateandBeyondLongitudinalStudy(BB:08/09):FirstLook(NCES 2011-236). U.S. Department of Education. Washington, DC: National Center for Education Statistics. Retrieved on August 17, 2012 from http://nces.ed.gov/pubsearch

CSB (2012). StrādājošomēnešavidējādarbasamaksaLatvijasstatistiskajosreģionos [The Average Wage of Employees in Latvia’s Statistical Regions]. Riga, Latvia: Central Statis-tics Bureau. Retrieved on May 17, 2012 from http://data.csb.gov.lv/DATABASE/Iedzsoc/Ikgad%C4%93jie%20statistikas%20dati/Darba%20samaksa/Darba%20samaksa.asp

Dombrovsky, V. (2011). Financinghighereducation:Amodelforreform. Riga, Latvia: SAKI. Retrieved on May 5, 2012 from http://newsaki.awtech.lv/wp-content/uploads/2011/12/fi-nancing_higher_educationa_model_for_reform_v.dombrovskis.pdf

Finnie, R., & Frenette, M. (2003). Earning differences by major field of study: evidence from three cohorts of recent Canadian graduates. EconomicsofEducationReview, 22, 179–192.

Finnie, R., & Wannell, T. (2004). Evolution of the gender earnings gap among Canadian uni-versity graduates. AppliedEconomics,36, 1967–1978.

Human Resources and Skills Development Canada (2004). Canadastudentloansprogram:Annual report, 2002-2003. Gatineau, QC: Human Resources and Skills Development Canada.

ICHEFAP (2006). Highereducationfinanceandcost-sharinginAustralia. SUNY, Buffalo: The International Comparative Higher Education Finance Project. Retrieved on June 7, 2012 from http://gse.buffalo.edu/org/inthigheredfinance/files/Country_Profiles/Australia/Australia.pdf

Johnstone, B. (2004). Cost-sharing and equity in higher education: Implications of income-contingent loans. In P. Teixeira, B. Jongbloed, D. Dill and A. Amaral (eds),Marketsinhighereducation (pp. 27 - 60). Dordrecht: Kluwer Academic Publishers.

87

Johnstone, B. (2006). Higher educational accessibility and financial viability: The role of student loans. In GUNISeriesonthesocialcommitmentofuniversitiesvolume1.Highereducationintheworld2006:Thefinancingofuniversities(pp. 84 – 101). GUNI & New York, NY: Palgrave Macmillan.

Johnstone, B. (2009). Conventional fixed-schedule versus income contingent repayment ob-ligations: Is there a best loan scheme? HigherEducationinEurope,34(2), 189 – 199.

Kasa, R. (2006). DevolutionofstudentfinancialassistanceinLatvia. (Unpublished doctoral dissertation). The State University of New York at Buffalo, Buffalo, New York.

Kasa, R. (2008). Aspects of fiscal federalism in higher education cost-sharing in Latvia. Pea-bodyJournalofEducation,83(1), 86 – 100.

Kelly, K., O’Connell, P., & Smyth, E. (2010). The economic returns to field of study and com-petencies among higher education graduates in Ireland. EconomicsofEducationReview,29, 650–657.

Krūmiņš, J., Goša, Z., Bāliņa, S., Rungule, R., Trapenciere, I., Seņkāne, S., Koroļeva, I., … Kristaps, G. (2007).Augstākounprofesionālomācību iestāžuabsolventuprofesionālādarbībapēcmācībubeigšanas [Professional Activity of Graduates of Higher and Profes-sional Educational Institutions]. Riga, Latvia: University of Latvia.

Migali, G. (2010). Funding higher education and wage uncertainty: Income contingent loan versus mortgage loan. Economics of Education Review, doi:10.1016/j.econe-durev.2012.06.001

Minicozzi, A. (2005). The short term effect of educational debt on job decisions. EconomicsofEducationReview,24, 417 – 430.

Ministry of Education and Science (2011). PārskatsparLatvijasaugstāko izglītību2011.gadā:skaitļi,fakti,tendences [Annual Report on Latvia’s Higher Education 2011: Num-bers, Facts, Tendencies]. Riga, Latvia: Ministry of Education and Science.

O’Leary, C., & Sloane, J. (2005). The returns to a university education in Great Britain. Na-tionalInstituteEconomicReview,193(1), 75 – 89.

Price, D. (2004a). Educational debt burden among student borrowers: An analysis of the Baccalaureate and Beyond Panel, 1997 Follow-Up. ResearchinHigherEducation,45(7), 701-737.

Price, D. (2004b). Borrowinginequality:race,class,andstudentloans. Colorado, CO: Lynne Rienner Publishers, Inc.

Rosenblatt, R., & Andrilla, C. (2005). The impact of U.S. medical students’ debt on their choice of primary care careers: An analysis of data from the 2002 medical school gradua-tion questionnaire. AcademicMedicine, 80(9), 815 - 819.

Shen, H., & Ziderman, A. (2009). Student loans repayment and recovery: International com-parisons. HigherEducation,57, 315 – 333.

Shireman, R., Asher, L., Talwalker, A., Li, S., Irons, E., & Cota, R. (2006). Whitepaper.Addressing student loan repayment burdens: Strengths andweaknesses of the currentsystem. Berkeley, CA: Project on Student Debt. Retrieved on June 2006 from http://pro-jectonstudentdebt.org/index.php

Studiju un zinātnes administrācija (2011). 2011.gadapubliskaispārskats [Report on 2011]. Riga, Latvia: SZA.

Texeira, P., Johnstone, B., Rosa, M., & Vossensteyn, H. (2006). Cost-SharingandAccessibil-ityinHigherEducation:AFairerDeal? Dordrecht: Springer.

Usher, A. (2005). Muchadoaboutaverysmallidea. Toronto, ON: Educational Policy Insti-tute.



Zepa, B., Pārsla, E., Beitnere, D., Paniotto, V., Hazans, M., & Pavlina, I. (2006). Dzimumulīdztiesībasaspektidarbatirgū [Aspects of Gender Equality in the Labour Market in Lat-via]. Riga, Latvia: Baltic Institute of Social Sciences.

89

FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?

H.RichardNakamura,MikaelOlssonandMikaelLonnborg1

AbstractThis paper investigates the dynamics of FDIs in the Baltic Sea Region (BSR) by applying the Poisson Pseudo-Maximum Likelihood estimation method on a gravity model. In particu-lar, we analyze the influence of macro and spatial factors on investment stock changes and discuss whether the origin of these investments and the 2004 EU enlargement have had any effects on BSR FDIs.

Our results suggest that EU enlargement has been significant for FDI activity in the region, and that FDI is basically a regional issue as it tends to be bilateral within the region. However, the same results also suggest that geographic distance is not a significant factor. We conclude that while being traditional in nature, the BSR FDI pattern is undergoing changes towards a lesser degree of geographic bias.

Keywords: Baltic Sea Region, FDI, Trade, Gravity Model, Poisson Pseudo-Maximum Like-lihood method.JEL classification: C21, C23, F21, F23, R12

1. IntroductionSince the EU accession of Estonia, Latvia, Lithuania and Poland, political rhetoric has envi-sioned a bright future for the overall development and integration of the Baltic Sea Region (BSR). However, the integration process had been under way long before the Baltic states and Poland became full EU members, and the integration of these countries into the European common market has recently been impeded by the global recession, where the Latvian expe-rience embodies the vulnerability of small open economies in times of international financial crises.

Earlier research (e.g., Petri, 1994; UNCTAD, 1996) has shown that trade and foreign direct investment (FDI) intensity are closely associated. However, on the global level, investment distribution is uneven and the BSR is no exception. As seen in Figure 1, investment flows in the BSR are up until 2008 primarily directed from the top industrialized countries on the western shores of the Baltic Sea to the eastern BSR countries, which regionally have cost ad-vantages in all or some input resources for industrial or service production, or being emerging consumer markets close to larger EU countries with a population gaining increasing purchas-ing power.

1 Centre for Baltic and East European Studies (CBEES), Södertörn University, Södertörn University, SE-141 89 Huddinge, Sweden. Tel: +46-8-608 5008, Fax: +46-8-608 4170. Main author, H. Richard Nakamura, e-mail: [email protected]. The present study and article was prepared, analyzed and written by the main author.

FDIinthepost-EUaccessionBalticSeaRegion:Aglobaloraregionalconcern?


Figure 1. Annual inward FDI in the BSR between 1992-2008; % of total world net inflows.

20

18

16

14BSR-W

BSR-E12

10

8

6

2

0

4

0,8

3,1

07,89,3

3,65,8

9,111,9

17,9

12,7

5,77,6

0,4

7,3 6,95,1 4,4

1,31,51,71,52,11,11,21,21,31,20,90,9 0,8 0,8 0,8 0,90,5

1992

1993

1994

1995

1996

1997

1998

1999

2001

2000

2002

2003

2004

2005

2006

2007

2008

Note: BSR-W: Germany, Denmark, Sweden and Finland. BSR-E: Estonia, Latvia, Lithuania and Poland. Source: UNCTAD and own calculations.

The statistics between 1992 and 2008, as summarized in Figure 1, indicate a development from an initial boom to a more modest flow of inward investment, suggesting that the Baltic states are losing their comparative advantage in production factor costs. However, this is not the only explanation. Concurring with increased loss of production cost advantages, the purchasing power of host country consumers has increased substantially, which means that the investment motives of the multinational corporations (MNCs) in the Baltic states changed over that period. This in turn might lead to market-oriented FDIs, which, by requiring less investment funds in absolute terms, affects total net FDI inflows downwards. At the very end of the period under analysis in this paper, the global financial crisis occurred, which also af-fected international FDI flows negatively.

Figure 2. Cumulative inward FDI per capita between 1992-2008 (USD in current prices).

35 000

30 000

25 000

20 000

15 000

32 058

18 138 17 811

13 70711 632 10 689

7 3425 454 4 257 3 573 3 297

10 000

5 000

0

Swed

en

Den

mar

k

BSR

-W /a

vg./

Finl

and

BSR

-8 /a

vg./

Esto

nia

Ger

man

y

BSR

-E /a

vg./

Latv

ia

Pola

nd

Lith

uani

a

Note: The averages reported for the region as such (BSR-8) and the two constituent parts (BSR-W and BSR-E) are unweighted averages. Source: UNCTAD and own calculations.

91

Decomposition of FDI statistics into country level data will reveal that the per capita value of foreign investment in the BSR area (in cumulative terms) is spread unevenly between mature economies and former transition economies. From Figure 2, it is possible to conclude that the largest value of investment stocks tends to be found in the established mature economies of the western BSR (Germany scoring relatively less in the Figure 2 ranking since the popula-tion size is substantially larger than the rest of the BSR countries). Another illustration of the size of FDI in the region is our compilation of all cross-border mergers and acquisitions (M&As) over USD 1 million in the BSR between 1998 and 2010 (See Appendix Tables A1 to A6), which gives an impression made by a few “pure” intra-BSR cross-border M&As (for example, 35% of all cross-border M&As taking place in the BSR between 1998-2010 were between BSR firms [source: Zephyr M&A database]) to the list of major international M&A deals in the region. This diverse picture is indicative of the unclear knowledge we have on contemporary FDI activity in the Baltic Sea Region.

The general purpose of this paper is to investigate and analyze the dynamics of FDIs in the BSR. In particular, this paper verifies the origin and target country directions of FDIs in the region. We will do so by an econometric analysis of the influence of gravity factors on contemporary investment flows, proxied by the inward FDI stocks of the 8 BSR countries of the region, and will discuss whether the geographic origin of these investments and the EU accession of the Eastern Baltic countries are of any importance when discussing regional integration and investment promotion in the BSR. Since EU accession is a central event for this analysis, we limit our study to the period between 2000 and 2008, that is, four years prior to and four years after the accession year of Estonia, Latvia, Lithuania and Poland. Thus, the data obtained cover the immediate pre-accession period to the start of the global financial crisis of 2008-2009.

2. Theoretical framework and earlier researchTransaction costs in international trade have been a factor traditionally dealt with more or less pragmatically in order to lift them off the basic models of international trade (e.g., Dixit and Norman, 1980; Jovanović, 2006; Eicher et al., 2009). This has led to trade models with strong assumptions, where the purpose has been to theoretically argue for the basic benefits of inter-national trade rather than to let the models describe a factual reality. If transaction costs and other institutionally related costs are introduced into the theoretical models, the geographic bias observed in trade statistics can also be better explained theoretically, provided the as-sumption holds that the longer the geographical distance, the more expensive the transport costs. If we presume the mainstream argument that trade and investment follow each other closely (see, for example, Petri, 1994; UNCTAD, 1996), a regional bias similar to those found in trade patterns should also be expected for investment flows.

Earlier studies (e.g., Petri, 1994; see also Blomström and Kokko, 2003) suggest that trade flows are more volatile than FDI, which is less surprising since trade is more sensitive to short-term institutional changes in the world trade environment such as exchange rates, war, trade conflicts and trade partners’ domestic political factors, as compared to FDI, which tends to be a more medium- or long-term commitment. Then what are the determinants of FDI? This question can be considered a classic area in the FDI literature, and in their early contri-butions, Penrose (1959) and Hymer (1960 [1976]) argue that opportunity-seeking behavior predominates internationalizing firms’ investment decisions, forwarding the idea that firms


92

make their first international expansion when the home market has become mature and the competition has driven profits toward zero – in other words, a situation well accommodated within the framework of neo-classic theory. More recent literature on FDI determinants (e.g., Dunning, 1992; Shapiro and Globerman, 2001) focuses on features connected to host country market attraction and production cost advantages, such as the size of the market, strong pur-chasing power, education level, infrastructure standards, trade policies, exchange rates and political and macroeconomic stability.

In general, increased global competition, liberalization of FDI regimes, technological and logistic advances have changed the conditions for firms of any origin in doing FDI (Sauvant et al., 2009). As a natural consequence of this development, the outward investment behav-ior of emerging market MNCs has increased dramatically, possibly making the validity of the mainstream assumptions of FDI determinants questionable. For example, Buckley et al. (2007) argue that capital market imperfections, diversification and strategic asset-seeking in institutional environments resembling home country conditions are strong influential factors for Chinese firms’ foreign investment decisions. Similar outward FDI determinants have also been found for Indian firms (Pradhan, 2004).

Even if we are discussing primarily macro aspects of FDIs in this paper, some words on mo-tives influencing FDI decisions are necessary for our theoretical discussion. Dunning (1992) has made important contributions to this literature, and has spelt out four different motives for outward FDIs: (Natural) Resource-seeking, Market-seeking, Efficiency-seeking and Stra-tegic Asset-seeking. The first FDI group aims to acquire all forms of input – physical as well as intangible – needed for production. The second FDI group aims for sale and supply of goods and services to other markets. This type of FDI motive is typically triggered as a result of strategic decisions by investing firms to deepen the commitment in the markets they have served earlier primarily by exports (see, for example, Johansson and Vahlne, 1977), by, e.g., starting host country production. The third type of FDI motive, that is, Efficiency-seeking, is denoted by Dunning (1992) as a desire among established firms to utilize economies of scale and scope to, for example, concentrate production in locations associated with internationally competitive production costs, factor endowments, institutional arrangements, and the like. Finally, the fourth FDI motive aims to sustain the investing firm’s long-term strategic objec-tives of being globally competitive (see Barney’s [1991] discussion of “sustained competitive advantage” of firms), through measures such as acquisition of brands and technology (includ-ing intellectual property rights).

If we assume the beneficial effects from trade firstly in terms of improved terms-of-trade and secondly in terms of improvement of real exchange rates which increases the purchasing power of the importing country, the benefits to the importing country as a market-seekingFDI destination are obvious. The situation for efficiency-seeking FDI might be the opposite. All other things being equal, the initial advantages of locating production in countries with low production costs can over time, as the overall wealth of the host country increases and spillover effects materialize, switch to a situation where production turns disadvantageous (as it has done in, e.g., Estonia). On the other hand, if upgrading of host country production factors occurs, a production location in the host country might still be advantageous from the efficiency viewpoint, due to a shift to skill-intensive production (as observed by, e.g., Görg and Strobl, 2003). Furthermore (and closely connected to market-seeking type FDI),


93

the lifespan of efficiency-seeking FDI is also affected by which end markets the products produced are sold in, i.e., whether these products are sold domestically in the host country or exported (or both). Thus, the sustainability of efficiency-seeking FDI is heavily dependent on how the foreign investor manages reinvestment in existing FDI stocks in the host country. For resource-seeking FDI, the relationship might be similar to efficiency-seeking FDI that aims for exports, as the investor home country typically imports resources extracted in the FDI host country and in this way increases trade between host and home countries. Finally, strategicasset-seeking FDI implies a home country (long-term) claim on the host country, thus contributing to an improved balance-of-payment position of the importing home country vs. the exporting host country.

The theoretical literature exemplified by, e.g., Lipsey’s (2006) discussion, argues that facili-tated access to a country’s main markets and resulting increase in trade flows should lead to increased growth and wealth, which in turn makes the country an attractive host for FDIs. This discussion should also be contrasted to the discussion by, e.g., Hummels (1999), who argues that geography is still an important factor for today’s trade and investment. Taken together, there are strong theoretical reasons to believe that economic integration of neighbor-ing regions might foster increased FDI activity.

As stated above, a gravity model is employed in this paper for the empirical analysis. Origi-nally suggested in the 1960s by, e.g., Tinbergen (1962), the gravity model has since been used in International Trade studies for analyses of influences of various impeding factors on inter-national trade flows (e.g., Bergstrand, 1985; Broadman, 2005; Flam and Nordström, 2010; Paas, 2000; Paas and Tafenau, 2005; Santos Silva and Tenreyro, 2003). Inspired by these developments in the trade analysis field, a number of researchers have also proposed and attempted to transfer the same analysis method to studies on FDI flows and stocks. The geo-graphic focus of these gravity analyses on FDI flows and stocks has, e.g., been studies on FDI and trade flows between Europe and Latin America (e.g., Africano and Magalhães, 2005), post-cold war Eastern European economic integration (e.g., Christie, 2003), or FDI flows in China (e.g., Marchant and Peng, 2004) and Southeast Asia under the ASEAN regime (e.g., Stone and Jeon, 1999). However, to the best of the authors’ knowledge, no similar studies have been made so far for FDIs in the BSR region. By emphasizing the overall importance of distance in bilateral economic relationships between countries, these attempts to emulate use of trade gravity models onto FDI analyses have in general confirmed theoretical discussion by, e.g., Krugman (1991) of the so-called new economic geography school. In other words, the FDI gravity model estimation results of these studies have suggested the significance of “resisting” factors influencing overall investment between two countries vis-à-vis the total FDI sample. In the following sections, we will explain our adaptation of the traditional grav-ity model to an FDI context.

3. ModelFollowing the traditional application of the gravity model (e.g., Paas and Tafenau, 2005; Santos Silva and Tenreyro, 2003, 2006), an FDI gravity model that included GDP, GDP per capita, trade and distance as independent variables was adopted. Furthermore, in order to ad-dress the time trend inherited in the cross-section time series, a time variable is also included in the regression model.



Formalizing the main gravity model used in this study, we obtain lnFDIijt = β0 + β1 lnGDPit + β2 lnGDPjt + β3 lnGDPCit + β4 lnGDPCjt + β5 lnTradeijt + β6 lnDistanceijt + β7 Time + β8 DEU + β9 DBSR + lnεijt (1)

where lnFDI = Total bilateral FDI stocks between home country i and host country j trans-formed into natural logarithms lnGDP = Gross Domestic Product transformed into natural logarithms lnGDPC = GDP per capita transformed into natural logarithms lnTrade = Total bilateral trade between home country i and host country j transformed into natural logarithms lnDistance = Distance between home country i and host country j transformed into natural logarithms Time = Time trend variable (2000 = 1, …, 2008 = 9) DEU = EU accession dummy K transformed into natural logarithms (K = 0 if 2000-2003 period, K = 1 if 2004-2008 period) DBSR = BSR country dummy K transformed into natural logarithms (K = 0 if non-BSR country, K = 1 if BSR country) ln ε = Error term i = Home country i j = Host country j t = time period t.As seen, the variables chosen for our FDI gravity model (1) follow the standard pattern for most analysis applications in trade and FDI research (e.g., Africano and Magalhães, 2005; Ledyaeva and Linden, 2006; Paas, 2000; Paas and Tafenau, 2005; Petri, 1994; Santos Silva and Tenreyro, 2003; Stone and Jeon, 1999). As the dependent variable, total FDI stocks are chosen as a measure for FDI due to better data availability for all countries and years in the sample. The independent variables are the gross domestic product (GDP) of countries i and j, GDP per capita (GDPC) of countries i and j, geographic distance between countries i and j (Distance) and total value of trade flows between countries i and j (Trade).

Since we are using cross-section time series data, there are obvious risks for various factors inherited in the data that might result in spurious estimations, making regression results un-reliable. In order to detect the presence of unit roots and non-stationarity in the data, Dick-ey-Fuller tests were executed on the estimated residuals and all dependent and independent variables (except for Distance). The test results2 allowed us to reject the null hypothesis of the existence of a unit root as observed values exceeded critical DF values at the 5% level. Thus, no tests for variable cointegration were performed due to the indicated absence of unit roots in Dickey-Fuller tests (e.g., Studenmund 2011).

In addition to this set of basic regressors, dummy variables were introduced. The purpose of these dummy variables was to estimate 1) the statistical significance of changes in the FDI pattern from the 2004 accession of the four Eastern European countries Estonia, Latvia, Lithuania and Poland to the EU, and 2) the propensity for FDI in the BSR to be a “regional” issue (i.e., that BSR FDI occurs between countries located around the Baltic Sea). Some notes on the interpretation of the dummy parameters have to be made here, since we have trans-formed the dummy variable K into natural logarithms in order to maintain the consistency 2 DF tests are not reported here due to space constraints.

95

of the estimation model. Obviously, this makes a traditional interpretation of the dummy parameters misleading and the parameter estimation has to be antilogged in order to obtain an estimated magnitude of the influence of the dummy parameters on the dependent variable (Giles, 2011a, 2011b). Therefore, the estimations for the dummy variables were antilogged in the final analysis in order to obtain correct interpretations.

Table 1. Summary of the variables of the main gravity model.Variable Variable type Description Expected sign Data sourceFDIij Dependent Total bilateral FDI stocks of home coun-

try i and host country j transformed into natural logarithms

OECD

WIIWEurostatHome GDPi Independent GDP of the home country i transformed

into natural logarithms+ OECD

EurostatHost GDPj Independent GDP of the host country j transformed

into natural logarithms+ OECD

EurostatHome GDPCi Independent GDP per capita of the home country i

transformed into natural logarithms+ OECD

EurostatHost GDPCj Independent GDP per capita of the host country j

transformed into natural logarithms+ OECD

EurostatTradeij Independent Total bilateral trade between home

country i and host country j transformed into natural logarithms

+ IMF DOTS

EcoWinDistanceij Independent Total distance (in km) between home

country i and host country j transformed into natural logarithms

- Google Earth

Time Independent Time trend variable (2000 = 1, …, 2008 = 9

DEU Dummy EU accession dummy K transformed into natural logarithms (K = 0 if 2000-2003 period, K = 1 if 2004-2008 period)

DBSR Dummy Home and host BSR country dummy K transformed into natural logarithms (K = 0 if non-BSR country, K = 1 if BSR country)

Note: For currency conversion purposes between WIIW, Eurostat, IMF DOTS and OECD data sets, annual average exchange rates for EUR, GBP and USD were obtained from the

Furthermore, some important points regarding estimation problems with the gravity model require discussion. In the past, mostly ordinary least square (OLS) estimations have been used to estimate gravity models. Recent contributions (e.g., Santos Silva and Tenreyro, 2006; Flam and Nordström, 2010) point to bias problems in the presence of heteroskedasticity. This makes estimations less reliable or, in the worst case, even misleading due to the differ-ent weights given to the observations in a sample. Santos Silva and Tenreyro (2006) point to the fact that “the expected value of the logarithm of a random variable is different from



the logarithm of its expected value” (Santos Silva and Tenreyro, 2006: 641), which has been ignored in past gravity analyses in economics. Instead, the authors suggest employing the Poisson Pseudo-Maximum Likelihood (PPML) estimation method, which gives equal weight to the observations in a sample. Through Monte Carlo simulations, Santos Silva and Ten-reyro (2006) show that estimations yielded from the PPML method were robust to different patterns of heteroskedasticity. Being aware of the bias risk from heteroskedasticity in OLS estimations of gravity models, we will follow Santos Silva and Tenreyro (2006) and use the PPML method for this study. All estimations and test statistics were obtained with a Stata 11 statistical package.

4. Data sources and descriptive statisticsWe define the BSR as the EU member countries surrounding the Baltic Sea, i.e., Denmark, Estonia, Finland, Germany, Latvia, Lithuania, Poland and Sweden. Furthermore, in order to control for major FDI countries outside this region, the main world FDI countries that are OECD members (i.e., US, UK, France and Japan) are also included in the sample. In total, this sample makes 12 countries, covering FDI (defined as the FDI stock of each sample coun-try), GDP, GDP per capita and trade between 2000 and 2008 (i.e., 4 years before and 4 years after EU accession by the three Baltic states and Poland). Finally, distance, defined as the geographic distance between the capitals of each bilateral relationship, was used as a proxy for the physical distance between the sample countries.

Table 2. Descriptive statistics (after conversion of variables to natural logarithms except for the time trend variable)Variable Obs Mean Std. Dev. Min Maxyear 1080 2004 2.583185 2000 2008country 1080 6.1 3.351676 1 12lnFDI_stock 1041 7.142286 2.713552 -2.67 12.8lnhome_GDP 1080 26.409601 2.204863 22.45947 30.27672lnhost_GDP 1080 26.409601 2.204863 22.45947 30.27672lnhome_GDPC 1080 9.887204 0.8487946 8.09 11.04lnhost_GDPC 1080 9.879565 0.8550506 8.09 11.04lntrade 1080 7.782435 1.846522 2.56 12.36lndistance 1080 7.281833 1.143067 4.43 9.1EU_acc_dummy 1080 0.7222222 0.4481107 0 1BSR_dummy 1080 0.5333333 0.4991188 0 1time 1080 5 2.583185 1 9

Since the quality and availability of macro data for the sample countries (especially for the three Baltic states) vary, the data used in this study were obtained and compiled from the following sources: OECD, The Vienna Institute for International Economic Studies (WIIW), IMF Direction of Trade Statistics (IMF DOTS), Eurostat, and EcoWin. For currency conver-sion purposes when transforming EcoWin and WIIW data, denoted in Euro, into US Dollars (current prices) in order to make all data comparable, annual average exchange rates for Euro, British Pounds and US Dollars were obtained from the ECB. Distances between sample country capitals were obtained through the Google Earth distance measurement tool.

97

Table 2 presents the descriptive statistics for the sample, which contains 1080 observations. Of these, 39 observations for FDI stocks were negative, and deleted during conversion to natural logarithms. This left us a total of 1041 observations. Since the PPML method only al-lows for zero or positive values of natural logarithms, another 14 observations were omitted, which left us with a total of 1027 observations to be estimated through the PPML method.

5. ResultsTable 3 reports the cross-section estimation results of the basic version of the main gravity model (i.e., without dummies for EU accession and home-host country location) for each of the years 2000 to 2008. As stated earlier, due to bias problems in using OLS in gravity model estimations (see Santos Silva and Tenreyro, 2006) and detection of error term heteroskedas-ticity in our OLS estimation3, we used the PPML method for this analysis.

The first item to note is the overall significance of the estimated models. Secondly, all vari-ables except for the GDP variable have the theoretically expected signs. For all individual years, trade tends to be the most influential variable when estimating FDI dynamics. This is also confirmed in a correlation analysis, where the partial correlation between the Trade vari-able and the dependent variable was 0.796 at the 1% level4. This follows the FDI theory out-lined and discussed in UNCTAD (1996), Blomström and Kokko (2003) and Lipsey (2006), and the findings of, e.g., Petri (1994). What is seemingly more theoretically puzzling is the negative sign of the GDP variable. Mainstream theory would suggest such a variable would yield a positive sign. We interpret this unexpected sign by the very nature of the BSR FDI pattern; when compared to the largest foreign direct investor countries in the world, the main share of FDI activity in the BSR is made by firms from small open economies investing in other small open economies of the BSR. Therefore, as one of two control variables for size ef-fects (i.e., GDP and GDPC) included in the classic form of gravity models, the negative sign of GDP has to be considered together with GDPC estimations. Turning our attention to the size of the home and host economies measured as GDP per capita also gives some indication of the importance of this factor for the build-up of BSR FDI stocks. This result is rather intui-tive, since the increasing wealth of a country, all other things being equal, tends to increase the average income level of the population, which in turn makes host markets interesting for foreign investors. However, we should not forget that FDI is global in scope, directed by business and profit opportunities identified by investing firms no matter where the location, which means that what we have observed here is not unique for the BSR; we will return to this observation later in our continued discussion.

Finally, regarding geographic distance as a theoretically impeding factor for trade and FDI location (e.g., Hummels, 1999; Krugman, 1991), this study supports the findings of, e.g., Stone and Jeon (1999) that distance is a statistically insignificant factor when determining the location propensity of FDI. This is also a highly interesting result, suggesting that distance is a secondary factor to FDIs in the BSR.

3 Tests are not reported here due to space constraints.4 Partial correlations are not reported here due to space constraints.



Table 3. PPML estimation results of basic FDI gravity equations from 2000 through 2008.2000 2001 2002 2003 2004 2005

Constant 0.643 0.952 1.000 0.225 -0.560 -1.056(0. 917) (0.908) (0.900) (0.785) (0.602) (0.729)

ln(Home GDP) -0.045 -0.061** -0.066** -0.020 -0.006 0.000(0.030) (0.031) (0.031) (0.027) (0.018) (0.020)

ln(Host GDP) -0.046 -0.062** -0.065** -0.019 -0.003 0.002(0.030) (0.030) (0.030) (0.024) (0.019) (0.021)

ln(Home GDPC) 0.095*** 0.110*** 0.119*** 0.075*** 0.097*** 0.114***(0.031) (0.034) (0.034) (0.023) (0.025) (0.032)

ln(Host GDPC) 0.098*** 0.113*** 0.119*** 0.074*** 0.080*** 0.105***(0.032) (0.036) (0.036) (0.024) (0.025) (0.031)

ln(Trade) 0.201 0.223*** 0.229 0.174*** 0.140*** 0.126***(0.035) (0.033) (0.035) (0.030) (0.020) (0.020)

ln(Distance) -0.028 -0.035 -0.0367 -0.012 -0.025 -0.035(0.041) (0.040) (0.039) (0.038) (0.026) (0.030)

Pseudo R2 0.181 0.196 0.187 0.160 0.144 0.138Log Pseudolikelihood -219.750 -233.501 -230.822 -230.087 -228.301 -233.606No. of observations (= n) 110 116 114 114 114 116Coefficients with robust standard errors within parenthesis. *, ** and *** represents statistical significance at the 10%, 5% and 1% levels respectively.

Table 3. Continued.2006 2007 2008

Constant -0.594 -0.261 -0.098(0.552) (0.434) (0.406)

ln(Home GDP) 0.001 -0.009 -0.021(0.017) (0.015) (0.013)

ln(Host GDP) 0.003 -0.007 -0.019(0.018) (0.017) (0.014)

ln(Home GDPC)

0.084*** 0.084*** 0.092***

(0.028) (0.028) (0.025)ln(Host GDPC) 0.078*** 0.080*** 0.081***

(0.029) (0.029) (0.024)ln(Trade) 0.131*** 0.138*** 0.161***

(0.017) (0.014) (0.015)ln(Distance) -0.031 -0.014 0.008

(0.026) (0.024) (0.021)Pseudo R2 0.126 0.115 0.107Log Pseudo-likelihood

-233.182 -234.044 -222.386

No. of observa-tions (= n)

116 116 111

Coefficients with standard errors within parenthesis. *, ** and *** represents statistical significance at the 10%, 5% and 1% levels respectively.

Turning our attention to the cross-section time series version of the same data set, we can make more interesting observations. In Table 4, we can see that besides the fact that OLS es-timators systematically overestimate the coefficients for the independent variables compared

99

to PPML estimators5, the PPML parameters of all independent variables are also statistically significant except for the Distance variable. The signs of the parameters follow the pattern from the cross-section estimations reported in Table 3.

Table 4. OLS and PPML estimation results of the FDI gravity model for the entire period, controlled for EU accession and geographic proximity.

Model 1: OLS estimation of the main gravity model for the period 2000-2008

Model 2: PPML estimation of the main gravity model for the period 2000-2008

Model 3: PPML Gravity model control-ling for 2004 EU accession

Model 4: PPML Gravity model controlling for home and host country being a BSR country

Model 5: PPML Gravity model control-ling for 2004 EU accession and home and host country being a BSR country

Constant -6.674*** -0.063 -0.211 -0.235 -0.317(1.423) (0.221) (0.233) (0.241) (0.246)

ln(Home GDP) -0.172*** -0.023*** -0.021*** -0.020*** -0.019**(0.047) (0.007) (0.007) (0.008) (0.008)

ln(Host GDP) -0.156*** -0.021*** -0.022*** -0.018** -0.020***(0.046) (0.007) (0.007) (0.008) (0.008)

ln(Home GDPC) 0.653*** 0.100*** 0.110*** 0.100*** 0.109***(0.068) (0.010) (0.011) (0.010) (0.011)

ln(Host GDPC) 0.597*** 0.094*** 0.101*** 0.094*** 0.100***(0.068) (0.010) (0.010) (0.010) (0.010)

ln(Trade) 1.301*** 0.166*** 0.167*** 0.164*** 0.166***(0.051) (0.008) (0.008) (0.008) (0.008)

ln(Distance) 0.023 -0.003 -0.000 0.003 0.004(0.075) (0.010) (0.010) (0.010) (0.010)

Time -0.062*** -0.009*** -0.015*** -0.010*** -0.015***(0.016) (0.002) (0.002) (0.002) (0.003)

DEU -0.061*** -0.057***(0.016) (0.016)

DBSR -0.033*** -0.022*(0.013) (0.013)

Adj R2 0.828Pseudo R2 0.156 0.156 0.156 0.156Log Pseudolike-lihood

-2070.025 -2068.620 -2069.649 -2068.457

No. of observa-tions (= n)

1041 1027 1027 1027 1027

Coefficients with standard errors within parenthesis. *, ** and *** represent statistical significance at the 10%, 5% and 1% levels respectively.

However, the estimation results of the two dummy variables DEU and DBSR tell us an interest-ing story. As expected, EU accession persedid have a statistically significant influence on changes in FDI stocks. In the PPML estimation models 3 and 5 (see Table 4), the EU acces-sion dummy yielded statistically significant results suggesting that, holding all other variables constant, post-accession FDI stock levels were on average 6% higher than prior to EU acces-sion. The location dummy for BSR countries’ FDI stocks (models 4 and 5 of Table 4) yield a

5 Probably due to the presence of heteroskedastic errors in the sample; the magnitude of this bias can be easily con-trolled for by comparing the elasticities (i.e., the coefficients) of the OLS and PPML estimations, respectively. The formula to compute the percentage effects is (eβi – 1) x 100, where βi is the coefficient estimated (Santos Silva & Tenreyro, 2006: 651).



statistically significant coefficient, indicating that, holding all other variables constant, intra-BSR FDI stocks are on average between 2% and 3% higher than FDI stocks held by firms from the major industrialized countries outside the BSR.

All in all, the PPML estimations of the FDI gravity model by and large confirm the “common wisdom” of the theoretical and empirical literature on FDI cited in this paper, with two in-teresting exceptions: the negative sign of the GDP variable, and the statistically insignificant parameter estimation for the Distance variable in combination with the statistically signifi-cant BSR location dummy variable. We will now move on to interpret the PPML estimation results in the next section.

6. Discussion and conclusionsIn this study, we used a gravity model to analyze recent FDI trends in the BSR measured by changes in FDI stocks in bilateral relationships between countries in and outside the BSR. First of all, we would like to forward the methodological contribution of this paper to the FDI literature, namely the econometric analysis on macro data using the Poisson Pseudo-Maximum Likelihood estimation method, which has been demonstrated by Santos Silva and Tenreyro (2006) to be unbiased and robust to different patterns of heteroskedasticity. This would imply the appropriateness of using PPML estimations for our sample here, by being the most consistent estimation method for gravity analyses of the type presented in this ar-ticle. This has, to the best of the authors’ knowledge, not yet been done on BSR FDI data, and leads us to forward our next contribution to the literature on BSR FDI, which is to show that the close connection between trade and FDI discussed in the theoretical literature also has its validity for the BSR setting. Earlier gravity model analyses have stopped at studying the relationship between trade and distance, and done so by using OLS for their estimation (e.g., Paas, 2000; Paas and Tafenau, 2005). Here, we present a variant of the gravity model that also includes FDI. The findings of our study are expected in line with Petri’s (1994) and Hum-mels’ (1999) discussions and earlier, similar, studies on FDI and trade (e.g., Africano and Magalhães, 2005; Bergstrand, 1985; Broadman, 2005; Christie, 2003; Ledyaeva and Linden, 2006; Santos Silva and Tenreyro, 2003; Santos Silva and Tenreyro, 2006), i.e., the estimation results suggest tradevolume,thesizeofhomeandhosteconomies and thelocationofinvest-ingcountriesas important factors for understanding foreign direct investment activity in the Baltic Sea Region.

The assumption of investing firms’ general preference for countries and cultures close to their home countries – at least in the early stages of an internationalization process – is something that we recognize from mainstream theory, such as the early theories on FDI and firm inter-nationalization (see, for example, Johanson and Vahlne, 1977; Dunning, 1977, 1979). This “basic paradigm” of early FDI theories has since been challenged by the current development pattern of global FDI, where we can observe firms such as the so-called “born globals” and where physical distance seems to play a lesser role in investment decisions (e.g., UNCTAD, 2011) for the benefit of other investment incentives such as market size, costs of obtaining resources and low production costs that might be decisive for investment decisions. This could be the explanatory factor behind the statistically insignificant results for the Distance parameters in combination with the other, statistically significant, parameters. Large physical distances perse might have become less of an effective deterrent for foreign direct investors.

101

In the midst of all this, our results also suggest that we should not abandon or forget that the “basic paradigm” is still valid, and that the “traditional” FDI pattern lingers on at least in the BSR in the sense that foreign direct investments in the BSR tend to be between countries within the same region. FDIs follow the trade pattern very closely in the BSR, which is as-sociated with the location of the principal trade partners of BSR countries. The answer to the title of this paper is therefore: “FortheBSR,FDIisindeedaregionalissueratherthanaglobalconcern” – with the addition “butthispatternisundergoingchanges”.

Thus, we can conclude that the FDIs taking place in the BSR are still characterized as being an “internal affair” of the region, which is also correlated to the overall trade pattern of the area. The imports of eastern BSR countries are dominated by products and services produced in western BSR countries or produced by firms owned by western BSR MNCs, and that one-third of all exports of the eastern BSR go to western BSR countries (Olsson et al., 2010).

It is also interesting, and highly relevant, to reflect our results with what is said in EU politi-cal circles in connection with the EU accession of the eastern BSR countries. The political rhetoric of hope and intentions prior to 2004, and manifested in the so-called EU Baltic Sea Region strategy in 2009 (Baltic Development Forum, 2009), envisioned significant effects and dynamism in economic integration, trade and FDI from the EU enlargement project. This is also what we would expect from the FDI literature (e.g., Lipsey, 2006). We might well use our results to discuss BSR economic integration also from a somewhat different viewpoint than those held by, e.g., policymakers in the BSR, who were hailing the EU accession of the Baltic states and Poland as historic for BSR economic development. We are not arguing against this fact as such. However, we would like to point to the statistically insignificant re-sult of the Distance variable, which normally yields a negative parameter in gravity models. In combination with the statistically significant dummy parameter result for EU accession, we would suggest that FDI decisions are more dependent on, e.g., FDI regimes, trade regimes and production cost advantages rather than geographical closeness (i.e., the distance) of mar-kets and production locations.

Blomström and Kokko (2003) argue in their overview of the FDI literature that the existence of other basic economic foundations concurring with regulatory reforms and political rheto-ric are crucial to make firms carry out FDI and host economies benefit from positive FDI spillovers. The eastern BSR countries are still in their “catching-up” transition process, albeit growing at a slower pace now than during the 1990s, and their market potential, defined as the overall purchasing power of the consumers in these markets, continues to attract FDIs into the region. Furthermore, the economic growth and development of the eastern BSR countries also spills over to the flow of outward FDIs of eastern BSR firms, which have increased steadily during the last 15 years.

In conclusion, policymakers in countries struggling to attract inward FDIs should be aware of the non-perpetual and “liquid” nature of FDI, especially in the manufacturing sector. As inward FDIs in the BSR are dominated by investments in the manufacturing sector, this par-ticular nature of FDI in the region should encourage governments to formulate realistic FDI policies attractive for foreign investors, aiming for sustained long-term retention of foreign investment in, e.g., non-manufacturing sectors. Contemporary global investor groups are not necessarily of EU origin, and BSR policymakers staying comfortable with inward FDI pro-



motion programs based on freedom of labor and capital within the EU might miss out factors which foreign investors consider more important than geographic home country closeness and free factor movements.

SuggestionsforFutureResearchThis paper is intended as a study of longitudinal dynamics in the FDI pattern of the BSR, explained by selected macro and spatial factors and estimated by using the Poisson Pseudo-Maximum Likelihood method. By doing this study, we have opened up a number of pos-sible developments for this research. One is, as suggested by, e.g., Nakamura (2005), that cross-border investment in the form of M&A tends to follow the overall M&A trend in the particular economy, that is, a tendency for behavioral isomorphism in the organizational field that the companies belong to (i.e., that firms are exhibiting a “follow-the-herd” mentality). The question is whether such behavior is also observable in the BSR. Another phenomenon that should be investigated in a BSR setting is institutional reasons for foreign companies to invest in a particular BSR economy that is not geographically close to the home country of the investor. Furthermore, in this study we only observed a small temporal “window” in the history of BSR FDI; to study the influence of firm-specific historical, path dependent, factors is probably still an alternative way to study historical FDI trends in the BSR.

103

ReferencesAfricano, Ana Paula and Magalhães, Manuela (2005). “FDI and trade in Portugal: A gravity

analysis”, FEP Working Papers no. 174, Universidade do Porto, Oporto.Barney, Jay B. (1991). “Firm Resources and Sustained Competitive Advantage”, Journal of

Management 17(1), 99-120.Baltic Development Forum (2009). State of the Region Report 2009: Boosting the Top of

Europe. Copenhagen/Helsinki: Baltic Development Forum.Bergstrand, Jeffrey H. (1989). “The gravity equation in international trade: Some micro-

economic foundations and empirical evidence”, Review of Economics and Statistics 67, 474-481.

Blomström, Magnus and Kokko, Ari (2003). ”The economics of foreign direct investment incentives”, NBER Working Paper 9489.

Broadman, Harry G. (2005). From disintegration to integration: Eastern Europe and the for-mer Soviet Union in international trade. Washington DC: The World Bank.

Buckley, P. J., Clegg, L. J., Cross, A. R., Liu, X., Voss, H. and Zheng, P. (2007). “The Deter-minants of Chinese Outward Foreign Direct Investment”, Journal of International Busi-ness Studies, 38(4), 499-518.

Christie, Edward (2003). “Foreign direct investment in Southeast Europe”, WIIW Working Papers no. 24.

Dixit, Avinash K. and Norman, Victor (1980). Theory of International Trade. Cambridge: Cambridge University Press.

Dunning, John H. (1977). “Trade, Location of Economic Activity and the MNE: A Search for an Eclectic Approach”, in Bertil Ohlin, Per-Ove Hesselborn and Per Magnus Wijkman (eds), The International Allocation of Economic Activity: Proceedings of a Nobel Sympo-sium. London: Macmillan, pp. 395-418.

Dunning, John H. (1979). “Explaining Changing Patterns of International Production: In De-fense of the Eclectic Theory”, Oxford Bulletin of Economics and Statistics 41(4), 269-295.

Dunning, John H. (1992). Multinational Enterprises and the Global Economy. Wokingham: Addison-Wesley.

Eicher, Teo S., Mutti, John H. and Turnovsky, Michelle H. (2009). International Economics, 7th ed. Oxon: Routledge.

Flam, Harry and Nordström, Håkan (2010). “Gravity estimation of the intensive and ex-tensive margin: An alternative procedure and alternative data”, Manuscript, Conference paper at the SNEE 12th annual conference on European Economic Integration in Swedish Research, Mölle, Sweden.

Giles, David E. (2011a). “Interpreting dummy variables in semi-logarithmic regression mod-els: exact distributional results”, Econometrics Working Paper EWP1101, Department of Economics, University of Victoria.

Giles, David E. (2011b). “On the inconsistency of instrumental variables estimators for the coefficients of certain dummy variables”, Econometrics Working Paper EWP1106, De-partment of Economics, University of Victoria.

Görg, Holger and Strobl, Eric (2003). “Multinational Companies, Technology Spillovers and Plant Survival”, Scandinavian Journal of Economics 105(4), 581-595.

Hummels, David (1999). “Have International Transportation Costs Declined?”, Manuscript, Graduate School of Business, University of Chicago.



Hymer, Stephen H. (1960 [1976]). The International Operations of National Firms: A Study of Direct Investment. Cambridge, MA: MIT Press.

Johanson, Jan and Vahlne, Jan-Erik (1977). “The internationalization process of the firm – A model of knowledge development and increasing foreign market commitments”, Journal of International Business Studies 8(1), 23-32.

Jovanović, Miroslav N. (2006). The Economics of International Integration. Cheltenham: Edward Elgar.

Krugman, Paul (1991). Geography and Trade. Cambridge, MA: The MIT Press.Lipsey, Robert E. (2006). “Measuring the Impacts of FDI in Central and Eastern Europe”,

NBER Working Paper 12808.Marchant, Mary A. and Peng, Xuehua (2004). “Prospects for China’s agricultural FDI in-

flows: A gravity model approach”, Manuscrpt, Conference paper at the Southern Agricul-tural Economics Association Annual Meeting, Oklahoma, USA.

Nakamura, H. Richard (2005). Motives, Partner Selection and Productivity Effects of M&As: The Pattern of Japanese Mergers and Acquisitions. Ph.D. Thesis. Stockholm: IIB.

Olsson, Mikael, Nakamura, H. Richard and Lönnborg, Mikael (2010). “Is there a Baltic Sea Region? A review of two decades of economic developments in trade and investment flows in the light of the European Union Strategy for the Baltic Sea Region”, Manuscript, Conference paper at the SNEE 12th annual conference on European Economic Integra-tion in Swedish Research, Mölle, Sweden.

Orbis (2011). Orbis European company database. Brussels: Bureau van Dijk.Paas, Tiiu (2000). “The gravity approach for modelling international trade patterns for econo-

mies in transition”, International Advances in Economic Research 6(4), 633-648.Paas, Tiiu and Tafenau, Elge (2005). “Regional trade clusters in promoting eastward enlarge-

ment of European Union”, Transition Studies Review 12(1), 77-90.Penrose, Edith T. (1959). The Theory of the Growth of the Firm. Oxford: Basil Blackwell.Petri, Peter A. (1994). “The regional clustering of foreign direct investment and trade”, Trans-

national Corporations 3(3), 1-24.Pradhan, Jaya Prakash (2004). “The Determinants of Outward Foreign Direct Investment: A

Firm-level analysis of Indian manufacturing”, Oxford Development Studies 32(4), 619-639.

Santos Silva, João M. C. and Tenreyro, Silvana (2003). “Gravity-defying trade”, Federal Reserve Bank of Boston Working Paper No. 03-1.

Santos Silva, João M. C. and Tenreyro, Silvana (2006). “The log of gravity”, The Review of Economics and Statistics 88(4), 641-658.

Sauvant, Karl P., Maschek, Wolfgang A. and McAllister, Geraldine (2009). “Foreign Direct Investment by Emerging Market Multinational Enterprises, the Impact of the Financial Crisis and Recession and Challenges Ahead”, Global Forum on International Investment VIII, 7-9 December 2009, OECD.

Shapiro, D. and S. Globerman (2001). “National Infrastructure and Foreign Direct Invest-ment”, Manuscript, Simon Fraser University.

Stone, Susan F., and Jeon, Bang Nam (1999). “Gravity-model specification for foreign direct investment: A case of the Asia-Pacific economies”, Journal of Business and Economic Studies 5(1), 33-42.

Tinbergen, Jan (1962). The World Economy: Suggestions for an International Economic Pol-icy. New York: Twentieth Century Fund.

105

UNCTAD (1996). World Investment Report: Investment, Trade and International Policy Ar-rangements. New York and Geneva: UNCTAD.

UNCTAD (2011). World Investment Report: Non-Equity Modes of International Production and Development. New York and Geneva: UNCTAD.

Zephyr (2011). Zephyr Global M&A database. Brussels: Bureau van Dijk.



Tabl

e A1.

Cro

ss-b

orde

r M&

A d

eals

in D

enm

ark

wor

th o

ver U

SD 1

bill

ion

1998

-201

0a, b.

Sour

ce: O

rbit/

Zeph

yr M

&A

dat

abas

e. *

Ast

eris

ks d

enot

e dea

l val

ue es

timat

ions

onl

y. a In

itial

Pub

lic O

ffers

(IPO

s) an

d Jo

int V

entu

res (

JVs)

not

incl

uded

. b Cou

ntry

code

s: A

T=A

ustri

a,

BH

=Bah

rain

, DK

=Den

mar

k, F

R=F

ranc

e, G

B=G

reat

Brit

ain,

IS=I

cela

nd, L

U=L

uxem

burg

, NL=

Net

herla

nds,

NO

=Nor

way

, SE=

Swed

en, U

S=U

SA.

Tabl

e A2.

Cro

ss-b

orde

r M&

A d

eals

in E

ston

ia w

orth

ove

r USD

1 b

illio

n 19

98-2

010a,

b.

Sour

ce: O

rbit/

Zeph

yr M

&A

dat

abas

e. *

Ast

eris

ks d

enot

e de

al v

alue

est

imat

ions

onl

y. a IP

Os a

nd JV

s not

incl

uded

. b Cou

ntry

cod

es: E

E=Es

toni

a, F

I=Fi

nlan

d, S

E=Sw

eden

.

107FDIinthepost-EUaccessionBalticSeaRegion:Aglobaloraregionalconcern?Ta

ble A

3. C

ross

-bor

der M

&A

dea

ls in

Fin

land

wor

th o

ver U

SD 1

bill

ion

1998

-201

0a, b

.

Sour

ce: O

rbit/

Zeph

yr M

&A

dat

abas

e. *

Ast

eris

ks d

enot

e de

al v

alue

est

imat

ions

onl

y. a IP

Os

and

JVs

not i

nclu

ded.

b Cou

ntry

cod

es: D

K=D

enm

ark,

FI=

Finl

and,

GB

=Gre

at B

ritai

n,

SE=S

wed

en, U

S=U

SA, Z

A=S

outh

Afr

ica.

Tabl

e A4.

Cro

ss-b

orde

r M&

A d

eals

in L

ithua

nia

wor

th o

ver U

SD 1

bill

ion

1998

-201

0a, b.

Sour

ce: O

rbit/

Zeph

yr M

&A

dat

abas

e. *

Ast

eris

ks d

enot

e de

al v

alue

est

imat

ions

onl

y. a IP

Os a

nd JV

s not

incl

uded

. b Cou

ntry

cod

es: L

T=Li

thua

nia,

PL=

Pola

nd.

Tabl

e A5.

Cro

ss-b

orde

r M&

A d

eals

in P

olan

d w

orth

ove

r USD

1 b

illio

n 19

98-2

010a,

b.

Sour

ce: O

rbit/

Zeph

yr M

&A

dat

abas

e. *

Ast

eris

ks d

enot

e dea

l val

ue es

timat

ions

onl

y. a IP

Os a

nd JV

s not

incl

uded

. bC

ount

ry co

des:

AT=

Aus

tria,

DE=

Ger

man

y, F

R=F

ranc

e, G

B=G

reat

B

ritai

n, IT

=Ita

ly, P

L=Po

land

.


Tabl

e A6.

Cro

ss-b

orde

r M&

A d

eals

in S

wed

en w

orth

ove

r USD

1 b

illio

n 19

98-2

010a,

b.

Sour

ce:

Orb

it/Ze

phyr

M&

A d

atab

ase.

*A

ster

isks

den

ote

deal

val

ue e

stim

atio

ns o

nly.

a IPO

s an

d JV

s no

t in

clud

ed.

b Cou

ntry

cod

es: A

E=U

nite

d A

rab

Emira

tes,

AT=A

ustri

a,

BH

=Bah

rain

, CN

=Chi

na, D

E=G

erm

any,

DK

=Den

mar

k, F

I=Fi

nlan

d, F

R=F

ranc

e, G

B=G

reat

Brit

ain,

IS=I

cela

nd, I

T=Ita

ly, N

L=N

ethe

rland

s, N

O=N

orw

ay, S

E=Sw

eden

, US=

USA

.

109HowintegratedaretheexchangemarketsoftheBalticSeaRegion?Anexaminationofmarketpressureanditscontagion

How integrated are the exchange markets of the Baltic Sea Region?

An examination of market pressure and its contagion

ScottW.Hegerty1

AbstractIn the two decades since independence, the Baltic nations’ re-integration with Western Europe has resulted in close linkages to the Euro. While only Estonia has as of yet joined the com-mon currency, Latvia and Lithuania maintain currency pegs in preparation for membership. But although the Eurozone’s pull is unmistakable, it is possible that the Nordic countries on the Baltic Sea—with which the Baltics have enjoyed economic ties for centuries—might also have an important influence on the region’s exchange markets. In particular, currency crises might more easily spread through the Baltic Sea region than to and from the Eurozone. This study investigates this relationship by generating indices of Exchange Market Pressure (EMP) for the three Baltic countries, Denmark, Sweden, and the Eurozone, before testing for contagion using Vector Autoregressive (VAR) methods. Granger causality tests and impulse-response functions show that pressure on the Scandinavian currency markets, as well as stock price declines, lead to increased Baltic EMP more than do events in the Eurozone.

JEL Classification: F31, F41Keywords: Exchange Market Pressure, Contagion, Vector Autoregression, Baltics, Scandi-navia

1. IntroductionAs they continue their path toward economic integration with Western Europe that resumed after the Soviet Union collapsed in 1991, the three Baltic nations of Estonia, Latvia, and Lithuania either have joined or plan to join the Euro. Firm pegs or currency boards form a strong link until accession has completed. Rather than sever these economic connections, the global financial crisis of 2008 seems only to have strengthened policymakers’ resolve to join the common currency, as Latvia’s refusal to devalue — and subsequent painful macroeco-nomic adjustment — attests.

But the German- and French-led currency union is not the only economic entity that enjoys close ties with the Baltic countries. As far back as the 13th century, Scandinavia and the Baltic Sea region have formed an important trading bloc. Even today, Sweden and Denmark serve as key sources of investment capital, banking centers, and export and import partners. In ad-dition, Finland has enjoyed a particularly close relationship with Estonia, which joined its northern neighbor as a fellow Euro member in 2010.

1 Department of Economics, Northeastern Illinois University, 5500 N. Saint Louis Avenue, Chicago, IL 60625, [email protected]


The other four countries in this study, however, still enjoy their own national currencies, even if the degree of flexibility is limited. Denmark, as a member of the European Exchange Rate Mechanism (ERM-II), is pegged to the Euro with a 2.25 percent band. Latvia and Lithu-ania joined this system in 2004, and the latter country has implemented a Euro-backed cur-rency board. As a result, the real and financial linkages among these countries might lead to interconnections among their currency markets. In particular, since maintaining a fixed exchange-rate regime introduces stress on a country’s currency following a crisis or a shock, it is possible that pressure in one might be transmitted to other countries in the region. This study examines transmissions of Exchange Market Pressure (EMP) among the Baltic and Scandinavian countries, generating indices of this pressure for five individual countries and for the Euro area. In addition, spillovers from Nordic stock prices to the region’s currencies helps uncover any linkages among the foreign exchange and asset markets.

Measuring EMP and its contagion in transition economiesExplained in detail below, Exchange Market Pressure (EMP) captures a weighted combina-tion of both currency depreciations and central bank action to avoid them. In short, an outright loss of value indicates an increase in market pressure, but so does a loss of foreign exchange reserves or an interest-rate hike. This index can be represented as a quarterly or monthly time series, which can then be analyzed via standard empirical methods. In particular, connec-tions among different countries might be examined to find evidence of “contagion” and the transmission of currency crises. This is particularly important in light of the Euro’s ongoing turmoil; it is possible that this can travel outside the common currency. Or, in the specific case of the Baltic countries, events outside the Eurozone might be relatively more significant. So far, a fairly limited number of studies have examined exchange market pressure in the countries of the former Soviet bloc. Often, studies use an alternative measure for “crises” in currency markets, and in addition, many of these focus primarily on the domestic determi-nants of these crises rather than their transmission of crises. For example, Gelos and Sahay (2001) examine interlinkages among the stock markets and currency markets of a set of tran-sition economies, as well as their macroeconomic linkages, up to 1998. Gibson and Tsakalo-tos (2004) study whether capital inflows increase the probability of a successful speculative attack in the ten new EU members that joined that year. Using panel methods, they find that the 1998 Russian crisis (proxied as a dummy variable) increased this probability.

Studies that generate a pressure index include Van Poeck etal. (2007), who use regression analysis to conclude that credit growth and current account deficits increase this pressure in a sample of eight transition economies. Hegerty (2009) finds that a withdrawal of portfolio capital increases EMP in the fixed-rate regimes of Bulgaria and the Baltics, but again, does not examine any cross-country transmission. Stavárek (2011) includes domestic determinants (such as domestic credit and the money multiplier) in his study of EMP in eight EU members, as well as foreign disturbances (such as Eurozone money supply). These external factors are shown to be influential, suggesting that further study of international transmission with the region, and particularly with the West, is warranted.

As Dornbusch et al. (2000) notes, and Hegerty (2011) explains, contagion can be caused by real, financial, or psychological factors. First, trade linkages are clearly affected by currency movements, particularly if one country’s exports are harmed by a neighbor’s depreciating currency. The importer’s resulting economic contraction can put pressure on its own currency

111

to depreciate. Second, one country’s currency depreciation, whether expected or merely an-ticipated, can lead to a flow of capital out of that economy and into another as investors seek to maximize their rates of return. This could lower EMP in the second country as pressure in the first increases. On the other hand, if investors expect neighboring countries to share re-gional characteristics, they might pull their capital out of additional nations and increase EMP throughout the region. Finally, psychological factors, such as “panic” selling or the tendency for uninformed investors to alter their investment behavior after an event in an unrelated country can in fact create a connection among countries where fundamentals are lacking.

Studies that do address the issue of contagion use a number of time-series methods. The most common of these include measuring correlations (as in Forbes and Rigobon, 2002) and Vec-tor Autoregressive (VAR) methods. While it is important to note that uncovering statistical interrelationships often follows outside the narrow definition of “contagion” (such as that of Forbes and Rigobon, 2002), these interrelationships can in fact uncover important inter-national connections. VAR analysis provides an important tool to examine the responses to shocks.

Notably, Boschi (2005) estimates a vector that includes six emerging markets’ (including Russia’s) exchange rates, stock returns, and sovereign debt spreads, but does not find much evidence of contagion. This method was extended by Hegerty (2011), who directly places seven transition economies’ EMP series into a VAR. Granger causality tests and impulse-response functions show that Russia has unexpectedly little influence on the region, while Hungary and Latvia have a stronger effect on the region’s currency markets. This study does not include any outside currencies or any measure of external shocks, however. In a study of Latin American EMP, Hegerty (2012) includes a measure of Brazilian stock prices and finds that stock-price declines increase EMP throughout the region.

As a result, we conclude that a study of EMP in the Baltic region must include some measure of global or regional market activity to better isolate events among the exchange markets in question. To that end, we measure contagion among the three Baltic countries (two if we exclude Estonia after it joined the Euro in 2011), two Scandinavian countries, the Eurozone, and a measure of Nordic stock prices. We find that the three Baltic countries respond more strongly to shocks to Scandinavian EMP or asset prices than they do to EMP in the Eurozone.This paper proceeds as follows. Section II outlines the EMP index used in the empirical study. Section III describes the results. Section IV concludes. 2. MethodologyIn the face of excess demand for foreign exchange, a central bank can either do nothing, or take action. Our EMP measure captures both options as a weighted average. As mentioned above, exchange market pressure indices capture currency depreciations, which are usually measured as percentage changes in a country’s bilateral exchange rate. Since central banks can intervene to thwart these movements, an EMP index captures the loss of reserves, and often interest-rate increases, as well. While a number of older methods (such as that of Girton and Roper, 1977) have modeled this pressure, one of the most commonly used approaches is that of Eichengreen etal. (1996):

HowintegratedaretheexchangemarketsoftheBalticSeaRegion?Anexaminationofmarketpressureanditscontagion


(1)

Exchange-rate depreciations result in an increase in units per dollar, and thus have a positive sign in the index. Since reserves are depleted to reduce pressure on a currency, this compo-nent has a negative sign. Likewise, interest-rate increases are also used to thwart deprecia-tions, so the third component has a positive sign.

In this model, the three components that comprise EMP are weighted by their own standard deviations so that the most volatile component does not dominate the series. The exchange rate and interest-rate changes are vis-à-vis the U.S., even though many countries peg to the Euro, because an EMP index is also created for the Eurozone. This allows for a consistent ref-erence point for all countries. Bilateral rates are used (instead of effective ones, for example) not only because this is standard in the literature, but also because changes in bilateral rates capture only events in the exchange market of the country in question, once those in the refer-ence country—which are common to all—are factored out. The currencies and weights used in constructing effective rates differ from country to country. Changes in reserves are scaled by the previous month’s monetary base. Monthly data from 2001m:11-2012m06 are taken from the International Financial Statistics of the International Monetary Fund. The variables are: e: nominal exchange rate (per US dollar); RES: foreign reserve stock, converted into domestic currency; MB = Monetary base, in domestic currency, deseasonalized with the Census-X12 procedure2; and r: short-term interest rate (money market rate).

While Weymark (1998) used model-dependent weights for the EMP components, the standard deviation has proven to be more empirically tractable. Nonetheless, the weighting scheme and other attributes of this EMP measure have begun to be questioned in the literature. Stud-ies include Pentecost et al. (2001), who use principal components to generate an alternative measure of EMP; Pontines and Siregar (2008), who try a number of methods but do not find a clear alternative; Bertoli etal.(2010), who question the weighting scheme of the EMP index; and Klaassen and Jager (2011), who also construct a new index. Because no suitable alterna-tive has yet been found, we choose the standard measure for our empirical analysis.Once these EMP series are generated, we examine their statistical properties by entering them in a VAR:

(2).

PS is a measure of Nordic stock process that is calculated as the first principal component of (log differenced) stock prices in Denmark, Sweden, and Finland, which were drawn from the IFS. The precise method is described below.

After choosing the appropriate VAR order by minimizing the Schwarz Information Criterion, connections among the variables are analyzed in two ways. First, Granger causality (block exogeneity) tests evaluate whether the addition of a variable has a significant impact on a country’s EMP. Should such an effect be shown, we can say that EMP in one country leads to pressure in another, and that a crisis originates that can spread to its neighbor. We next gener-ate impulse-response functions (IRFs) using the generalized VAR methodology of Pesaran 2 M1 is used for Estonia and the Eurozone.

113

and Shin (1998). This approach is invariant to the ordering of variables, which is useful for a model where there is no obvious ordering. This leads us to select this approach over the orthogonalized VAR technique of Sims (1980).

Using these two time-series methods, we can answer two key questions. First, we can see which countries are most vulnerable to currency shocks that are generated abroad, and which countries are most likely to cause a contagious crisis. It is possible that Scandinavia and the Eurozone, as two competing financial poles, might have differing influence on the region. Secondly, we can examine the role of asset-price movements and external events on the re-gion’s currency markets. These results are detailed below.

3. ResultsUsing monthly data from the International Financial Statistics of the International Monetary Fund, we generate EMP series according to Equation (1) for three Baltic and two Scandina-vian countries, as well as the Euro area. These series begin in 2002m1 and are depicted in Figure 1.3 As expected, there are “spikes” in the series that correspond to the 2008 crisis and its aftermath; low or even negative EMP often immediately follows these increases, as central banks respond to this macroeconomic turmoil. As was noted by Hegerty (2009, 2011), the mid- to late 2000s was also a period of below-zero EMP values. During this period, large capital inflows and corresponding current account deficits helped lead to reserve accumula-tion under the Baltics’ fixed exchange-rate regimes. Denmark, which is also fixed to the Euro under ERM-II, also registers similar patterns during 2008 and beforehand. Sweden, with its floating kronor, also sees its EMP spike in 2008 and exhibits fluctuations throughout the period. And the Euro, floating against the dollar, appears to experience similar events in its EMP series. This study hopes to uncover the relationships among these movements in the countries’ currency markets.

While most of the countries in this study currently peg their currencies to the Euro (although Lithuania switched from a dollar peg in 2004), the fact that dollar exchange rates are used in this study will introduce some fluctuation to the EMP series.4 The use of variance-smoothing weights should eliminate excess variability, however. Table 1 shows the weight (standard deviation) used for each of the three components in each country’s EMP series. We find that reserve changes have the largest variance in the series, particularly in Denmark. Latvia’s rate spike in 2009 might contribute to the large variance that is reflected in this table.

In addition to generating Exchange Market Pressure series, we also create a single measure of Nordic stock prices. The purpose of including this series is twofold: First, we capture external events in world markets (such as the 2008 financial crisis). Second, we can assess whether in-vestor behavior causes asset-price movements to spill over to the region’s currency markets.5 While it is possible to include Baltic stock prices as well, we choose to use an index that lies completely outside the Baltics. We thus proxy these prices as the first principal component of log changes to Danish, Swedish, and Finnish stock prices. The results of our principal com-ponents analysis are provided in Table 2.3 Note that the series vary slightly from Hegerty (2009, 2011) due to variations in the weighting scheme and choice of interest rate, and from Stavárek (2011) due to a different modeling strategy.4 Hegerty (2009) avoids this problem by assuming exchange-rate movements to be zero.5 Maneschiöld (2006) uses time-series methods to examine linkages between Baltic stock markets and their foreign counterparts.



We see that the first principal component explains 78 percent of the series’ variance, and that the other two components have eigenvalues below one. This confirms the choice of a single index using this method. We also find that the three countries make a roughly equal contribu-tion to the series. Finland and Denmark are the most closely correlated to one another, and have the highest factor loadings in the principal components.

The combined stock index is shown in Figure 2. Clear drops occur, as expected, during the 2008 financial crisis and the Euro crisis that followed a few years later. The VAR analysis will tell us whether these stock declines are related to increases in EMP in the region.

Table 3 gives the results of Phillips-Perron stationary tests on the EMP indices and (log dif-ferenced) stock price series. It also provides summary statistics for each variable used in our analysis. Since these variables are shown to be stationary, we enter them into our VARs without any further differencing. Our VAR analysis continues below.

Granger causality test results are provided in Table 4. We include two variations of our model. The first (Panel A) includes Estonia—even though the kroon no longer exists—to show a historical relationship that might still be of economic value. This estimation must, by necessity, end with Estonia’s independent currency in 2011. The second analysis (Panel B) excludes Estonia and runs to 2012. Both panels show similar results, indicating that the relationships are robust to the inclusion of Estonia or the shortening of the sample period.

Our most important finding is that the Nordic countries carry a stronger influence on the Baltic currency markets than do events in the Eurozone. Stock price changes Granger-cause EMP in Estonia and Lithuania, as well as in Denmark and Sweden. Danish EMP has an influ-ence on Estonia and (to a lesser extent) in Lithuania, while Swedish EMP has a strong effect on pressure in the Latvian foreign exchange market. EMP in the Eurozone Granger-causes EMP in Denmark and Sweden, and seems also to affect Nordic stock prices. Two other find-ings are significant: First, the inclusion of external forces in the model eliminates many of the bilateral relationships between Baltic EMP series that were uncovered by Hegerty (2011). Secondly, there is some evidence that market pressure in the Baltics leads to increased EMP in the Eurozone—perhaps due to psychological factors such as confidence, as investors be-come concerned with wider crisis— but these results are not consistent across specifications.These findings are confirmed via the Generalized Impulse-Response Functions (GIRFs) that are depicted in Figure 3.

We see that Estonian EMP registers a lasting increase after a one-standard-deviation shock to Danish EMP, and increases after a Nordic stock decline. Responses to increases in Eurozone EMP are much weaker and occur only after the passage of months. Latvia, likewise, sees its EMP index increase after a shock to the Swedish currency market, as was foreseen by the Granger causality tests. Again, Latvia is less sensitive to stock declines than its Baltic neigh-bors. Lithuania responds strongly to the Danish krone and to Nordic asset depreciations; here, perhaps surprisingly, we see Eurozone EMP cause a significant response to Lithuania’s pres-sure series. Again, since the Baltics were at the center of the European and global financial crises, it is no surprise that attention might be drawn to the Euro as a result.

115

4. ConclusionAs it struggles with maintaining its common currency, the Eurozone continues to serve as the epicenter of a potential crisis that could spread around the world. But perhaps no region is as susceptible as the new EU members on its periphery, which not only enjoys strong trade ties with its wealthy neighbors, but also includes countries that have chosen to peg to the Euro as a precursor to joining the common currency. It is thought that as pressure on the Euro spikes, these neighbors could fall victim to a “contagious” currency crisis. But, if linkages with other partners prove to be stronger, these effects will be relatively muted. This study examines this proposition for the Baltic nations, observing their strong econom-ic connections with the Scandinavian countries of the larger Baltic Sea region. Creating a monthly measure of exchange market pressure (EMP) for Latvia, Lithuania, Denmark, and Sweden, as well as Estonia (before its Euro accession) and the Eurozone itself, we use time-series methods to examine the relationship among these currency markets. Also incorporat-ing a single proxy for stock-price movements in Denmark, Sweden, and Finland—to proxy global events as well as to look at spillovers between two different asset markets—we con-duct Granger causality tests and generate impulse-response functions to arrive at a few key conclusions.

First, Scandinavia appears to have a stronger influence on the Baltics than does the Eurozone. This is shown by significant responses by Baltic EMP to shocks in stock prices and foreign EMP. Second, Eurozone EMP affects EMP in Denmark and Sweden, as well as stock prices. Finally, there is also evidence of causality running the other direction, originating in the Bal-tic and spreading to the Euro itself. It is possible that such small countries might affect their larger neighbor due to their effects on expectations, leading investors to adjust their behavior regarding the common currency as they witness events on the periphery. But the strength of Nordic influence on the Baltics, relative to that of the Eurozone, is the main finding of this study.

These results carry important implications for regional integration. It shows the Baltic Sea re-gion to be a highly integrated economic space, and confirms that the Baltic countries maintain strong linkages within it as well as with the currency to which their currencies are pegged. Even when Latvia and Lithuania join the Euro, these linkages are expected to persist. In particular, asset price movements will continue to have strong effects throughout the region. Policymakers should be wary of any assumption that the Baltic nations are solely tied to the Euro.



References:Bertoli, S., G.M. Gallo, and G. Ricchiuti 2010), “Exchange Market Pressure: Some Caveats

in Empirical Applications,” AppliedEconomics 42, 2435–2448.Boschi, M. (2005), “International Financial Contagion: Evidence From the Argentine Crisis

of 2001-2002,” AppliedFinancialEconomics 15, 153-163. Dornbusch, R., Y.C.Park, and S. Claessens (2000), “Contagion: Understanding How It

Spreads.” WorldBankResearchObserver 15(2), 177-197.Eichengreen, B., A. Rose, and C. Wyplosz (1996), “Contagious Currency Crises: First Tests,”

ScandinavianJournalofEconomics 98(4), 463-484.Forbes, K. and R. Rigobon (2002), “No Contagion, Only Interdependence: Measuring Stock

Market Co-Movements,” JournalofFinance57(5), 2223-2261.Gelos, R.G. and R.Sahay (2001), “Financial Market Spillovers in Transition Economies,”

EconomicsofTransition 9(1), 53-86.Gibson, H.D. and E. Tsakalotos (2004), “Capital Flows and Speculative Attacks in Prospec-

tive EU Member States,” EconomicsofTransition 12(3), 559-86.Girton, L. and D. Roper (1977), “A Monetary Model of Exchange Market Pressure Applied to

the Postwar Canadian Experience,” AmericanEconomicReview67, 537-548.Hegerty, S.W. (2009), “Capital Inflows, Exchange Market Pressure, and Credit Growth in

Four Transition Economies With Fixed Exchange Rates,” EconomicSystems 33(2), 155-167.

Hegerty, S.W. (2011), “Is Exchange-Market Pressure Contagious Among Transition Econo-mies?,” Applied FinancialEconomics 21(10), 707-716.

Hegerty, S.W. (2012), “Exchange Market Pressure, Commodity Prices, and Contagion in Latin America” Journalof InternationalTrade&EconomicDevelopment, in press, doi: 10.1080/09638199.2012.679292

Klaassen, F. and H. Jager (2011), “Definition-Consistent Measurement of Exchange Market Pressure,” JournalofInternationalMoneyandFinance 30, 74-95.

Maneschiöld, P.-O. (2006), “Integration Between the Baltic and International Stock Mar-kets,” EmergingMarketsFinanceandTrade 42(6), 25–45.

Pentecost, E.J., C. Van Hooydonk, and A. Van Poeck (2001), “Measuring and Estimating Exchange Market Pressure in the EU,” JournalofInternationalMoneyandFinance 20, 401–418.

Pesaran, M.H., and Y. Shin (1998), “Generalised Impulse Response Analysis in Linear Multi-variate Models,” EconomicsLetters58, 17-29.

Pontines, V. and R. Siregar (2008), “Fundamental Pitfalls of Exchange Market Pressure Based Approaches to Identification of Currency Crises,” InternationalReviewof EconomicsandFinance, 17(3), 345-365.

Sims, C.A. (1980) “Macroeconomics and Reality” Econometrica 48, 1-48.Sojli, E. (2007), “Contagion in Emerging Markets: the Russian Crisis,” AppliedFinancial

Economics 17(3), 197-213.Stavárek, D. (2011), “Comparison of Exchange Market Pressure Across the New Part of the

European Union,” EmergingMarketsFinance&Trade 47(Supplement 3), 21–39.Van Poeck, A., J. Vanneste, and M. Veiner (2007), “Exchange Rate Regimes and Exchange

Market Pressure in the New EU Member States,” JournalofCommonMarketStudies 45(2), 459-85.

Weymark, D.N. (1998), “A General Approach to Measuring Exchange Market Pressure,” OxfordEconomicPapers 50(1), 106-121.

117

Figure 1. Exchange-Market Pressure Indices, 2002-2012 Denmark

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

10.000

8.000

6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000

-8.000

Estonia

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

8.000

6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000

Eurozone

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

8.000

-8.000

6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000



Latvia

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

10.000

5.000

0.000

-5.000

-10.000

-15.000

Lithuania

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

6.000

4.000

-2.000

2.000

0.000

-4.000

-6.000

-8.000

Sweden

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

6.000

4.000

12.000

10.000

8.000

2.000

0.000

-4.000

-6.000

-2.000

119

Table 1. Standard Deviations of EMP Components.E RES R

Denmark 0.025 0.202 0.002Estonia 0.026 0.041 0.004Eurozone 0.026 0.002 0.002Latvia 0.024 0.099 0.025Lithuania 0.025 0.104 0.006Sweden 0.029 0.093 0.003

Table 2. Principal Components Analysis of Scandinavian Stock Prices.Eigenvalues Factor Loadings Correlations

Number Value Difference Proportion Index Component # 1 Denmark Finland Sweden1 2.339 1.883 0.780 Denmark 0.585 Denmark 12 0.456 0.251 0.152 Finland 0.604 Finland 0.787 13 0.205 --- 0.068 Sweden 0.540 Sweden 0.571 0.645 1

Figure 2. Percent Changes in “Scandinavian” Stock Prices

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

5.000

0.000

20.000

15.000

10.000

-5.000

-10.000

-20.000

-25.000

-30.000

-15.000

Calculated as the first principal component of percentage changes in Danish, Finnish, and Swedish stock prices. Source: IFS

Table 3. Phillips-Perron Stationarity Test Results and Summary Statistics.Country Statistic (p-value) Mean Std. Dev. Min Max NEstonia -10.370 (0.000) -0.25 1.72 -3.91 6.71 108Latvia -11.992 (0.000) -0.21 1.92 -13.64 6.34 126Lithuania -10.532 (0.000) -0.25 1.56 -5.66 4.77 126Denmark -8.158 (0.000) -0.34 1.83 -6.86 8.95 126Sweden -9.102 (0.000) -0.20 1.50 -4.24 10.06 126Eurozone -8.596 (0.000) -0.09 1.23 -6.18 5.64 126Stock Prices -7.399 (0.000) 0.00 1.54 -7.05 4.16 126

Bandwith = 5 for Newey-West standard errors in all cases.



Table 4. Granger Causality (Block Exogeneity) Test Results.PanelA:IncludingEstonia(2002-2010)

Country Estonia Latvia Lithuania Denmark Sweden Eurozone Stock Prices

(Excluded) Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.) Chi-sq (Prob.)

Estonia 0.859 (0.354)

3.323 (0.068)

5.825 (0.016)

0.797 (0.372)

7.703 (0.006) 2.174 (0.140)

Latvia 0.743 (0.389)

0.230 (0.631)

0.429 (0.512)

1.969 (0.161)

4.246 (0.039) 1.696 (0.193)

Lithuania 0.898 (0.343)

1.588 (0.208)

0.128 (0.721)

0.027 (0.869)

2.825 (0.093) 2.053 (0.152)

Denmark 3.021 (0.082)

0.335 (0.563)

3.098 (0.078)

7.101 (0.008)

0.625 (0.429) 0.000 (0.999)

Sweden 2.364 (0.124)

4.834 (0.028)

0.015 (0.902)

0.060 (0.806)

1.021 (0.312) 2.603 (0.107)

Eurozone 2.580 (0.108)

0.374 (0.541)

0.833 (0.362)

3.314 (0.069)

5.729 (0.017) 7.694 (0.006)

Stock Prices

7.758 (0.005)

0.905 (0.342)

2.288 (0.130)

3.639 (0.057)

7.688 (0.006)

0.694 (0.405)

All 35.165 (0.000)

12.821 (0.046)

14.526 (0.024)

13.658 (0.034)

21.563 (0.002)

15.818 (0.015) 15.939 (0.014)

PanelB:ExcludingEstonia(2002-2012)Latvia Lithuania Denmark Sweden Eurozone Stock Prices

Excluded Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.)

Chi-sq (Prob.) Chi-sq (Prob.)

Latvia 0.123 (0.726)

1.460 (0.227)

2.184 (0.140)

2.148 (0.143) 1.935 (0.164)

Lithuania 0.709 (0.400)

1.269 (0.260)

0.253 (0.615)

4.300 (0.038) 2.130 (0.145)

Denmark 0.494 (0.482)

1.972 (0.160)

6.883 (0.009)

0.119 (0.730) 0.989 (0.320)

Sweden 4.917 (0.027)

0.010 (0.921)

0.114 (0.736)

1.156 (0.282) 4.116 (0.043)

Eurozone 0.909 (0.340)

0.628 (0.428)

3.801 (0.051)

5.193 (0.023) 8.680 (0.003)

Stock Prices

1.228 (0.268)

2.994 (0.084)

3.412 (0.065)

10.632 (0.001)

0.557 (0.456)

All 12.378 (0.030)

12.670 (0.027)

10.442 (0.064)

22.903 (0.000)

7.725 (0.172) 16.270 (0.006)

Panel A: N = 107; Lag = 1, AIC = 24.371 (Akaike Information Criterion)Panel B: N = 125; Lag = 1, AIC = 20.653

121

Figure 3. Generalized Impulse-Response Functions to a One-Standard Deviation Shock, With ±2 Standard Error Bands. Estonia: Response to Latvia Lithuania

0.40.30.2

0.70.60.5

0.10

0 1 2 3 4 5 6 0 1 2 3 4 5 6

-0.2-0.3-0.4

-0.1

0.40.30.2

0.70.60.5

0.10

-0.2

0.8

-0.1

Denmark Sweden

0 1 2 3 4 5 6

0.4

0

-0.2

1

0.6

0.2

0.8

0 1 2 3 4 5 6

0.6

0.4

0.2

1.2

1

0.8

0

-0.2

Eurozone Stock Prices

-0.1-0.2-0.3-0.4-0.5-0.6-0.7-0.8-0.9

0.10.2

0

0 1 2 3 4 5 6 0 1 2 3 4 5 6-0.4

-0.2

0.2

0.4

0.6

0.8

0

Latvia: Response to Estonia Lithuania

5 6

-0.2

-0.4

-0.6

0.2

0.4

0.6

0.8

0

0 1 2 3 4 5 60 1 2 3 4

0

-0.2

-0.4

0.4

0.6

0.8

1

0.2

Denmark Sweden

5 60 1 2 3 4

-0.2

-0.4

1

0.2

0.4

0.60.8

0

5 60 1 2 3 4

-0.2

-0.4

1

0.2

0.4

0.60.8

0


5 60 1 2 3 4

-0.2

-0.4

1

0.2

0.4

0.6

0.8

0

5 60 1 2 3 4

-0.2-0.3

-0.5-0.6-0.7

-0.1

-0.4

0.10.2

0



Lithuania: Response to Estonia Latvia

5 60 1 2 3 4 5 60 1 2 3 4

-0.2

-0.4

0.4

0.6

0.8

0.2

0

-0.2-0.1

0.4

0.6

0.8

0.20.1

0.3

0.5

0.7

0

Denmark Sweden

5 60 1 2 3 4

0

0.8

0.6

0.4

0.2

1

-0.25 60 1 2 3 4

0

0.8

0.6

0.4

0.2

0.7

0.5

0.3

0.1

0.9

-0.1


5 60 1 2 3 4 5 60 1 2 3 4

0

0.6

0.4

0.2

0.7

0.5

0.3

0.1

-0.1

0

0.20.3

0.1

-0.1-0.2-0.3-0.4-0.5-0.6

Responses to Stock Price changes Denmark Sweden

5 60 1 2 3 45 60 1 2 3 4

0

0.20.1

-0.1-0.2-0.3-0.4-0.5-0.6-0.7

0

0.30.20.1

-0.1-0.2-0.3-0.4-0.5-0.6-0.7-0.8

123

The Impact of the Generosity of Unemployment Benefits on Estonian Labour Market Outcomes in a Period of Crisis1

AnneLauringson

This study analyses the effects of unemployment benefits on unemployment duration and post-unemployment employment in Estonia during the recent global economic crisis. The results indicate that receipt of unemployment benefits increases unemployment duration sig-nificantly even during a very severe recession. However, unemployment benefits also support job search so that the unemployed can find a job that suits them better. Hence, unemployment duration might not lengthen only because of lower job search activity, but also because the unemployed may have the opportunity to choose a job that for them is of higher quality (e.g. the tasks match the skills better).

In short, the analyses presented in this thesis suggest that there might be welfare effects if the system of unemployment benefits in Estonia were made more generous, particularly in terms of the potential benefit period, which is analysed more thoroughly. However, from previous studies conducted in other countries, it can be concluded that it might be reasonable to in-crease monitoring and sanctioning in the Estonian unemployment benefit system too, so that instead of the current situation where it is hard to obtain benefits in the first place but easy to stay on benefits, the system should be changed so that it would be easier to obtain benefits but harder to stay on benefits.

1 The publication of this dissertation is granted by the Doctoral School of Economics and Innovation created under the auspices of European Social Fund.

PhDnews

125PhDnews

Performance Appraisal and Remuneration Aspects of Performance Management on the Example of Estonian General Educational Schools1

ReelikaIrs

This dissertation aims to provide proposals for developing a teacher performance appraisal and remuneration aspects of performance management using the example of Estonian gen-eral educational schools. The development of a performance management tool is assumed in order to design and implement teacher performance appraisal and performance-related pay systems. As a result of this dissertation, critical aspects in designing both performance appraisal and performance-related pay are pointed out and recommendations are made for se-lecting appropriate criteria. Intotal,2,165 teachers and298 headmasters of Estonian general educational schools participated in this study. Additional casestudieswereconductedinthreegeneraleducationalschools.

The present study provided evidence that in order to employ new management practices more smoothly, aspects of school management should be taken into consideration. For ex-ample, the study showed that well-organised strategic management, resource management and organisational culture are important in performance appraisal and performance-related pay design, as it helps develop awareness, positive opinions and intention to adopt. Teacher performance management should be aimed at the balanced development of the school, and therefore, teacher performance appraisal and performance-related pay should combine crite-ria related to the learning process, the learning environment and school management. Teacher performance-related pay helps develop and improve teachers’ performance both as individu-als and in teams and guarantees teacher development in accordance with school objectives.

1 The publication of this dissertation is granted by the Doctoral School of Economics and Innovation created under the auspices of European Social Fund.

127

The relationship between governance and performance in water services provision in Estonian municipalities

PeeterPeda

Performance improvement in public services is recognised as a goal for many governments, and has resulted in a variety of public governance reform initiatives. Nevertheless, current knowledge about how governance arrangements influence the performance of public services is scarce and fragmented. The present dissertation aims to provide an in-depth understand-ing of how local governments set up and use governance mechanisms for public services provision and how the different governance patterns determine financial and non-financial performance in public services. This study seeks to contribute to the scientific debate on gov-ernance-performance relationships through research conducted in the Estonian water sector.

The first part of the empirical research comprises a quantitative analysis of the influence of ownership on the technical efficiency of Estonian water companies. The analysis reveals that none of the given ownership forms (public, private, mixed public-private) can be associated with greater or lower efficiency levels. The second part of the empirical research - a com-parative case study research - shows that the observed ownership structures had a different influence on the financial and non-financial performance of the water companies, which de-pending on the governance mechanisms applied could be either negative or positive. Private ownership of a water company could lead to relatively greater performance if the rights (e.g. tariffs) and obligations (e.g. quality requirements, accountability) are clearly fixed by strict regulatory contracts. Water services provision by a privately owned company without a clear ex ante written set of rules is likely to cause performance eroding agency problems between the local government and the company. Moreover, public ownership, which is often consid-ered to be old-fashioned and ineffective, could lead to good performance even without strong contractual arrangements when the water company is tied with the local government on a professional basis via the supervisory board.

The empirical findings of this study suggest that mere privatisation itself is not the key to solving performance problems in the water sector. Policymakers, regulators and managers of water companies involved in water services provision should consider the setup and use of regulatory contracts and corporate boards thoroughly in order to be effective in achieving targeted financial and non-financial performance objectives.

PhDnews

content i-netam.pdf · 2013. 4. 15. · content editorial 3 policy paper the evolution of tax and...

Documents