exports and econ growth in bd begum shamsuddin

Upload: anisur-rahman-faroque

Post on 30-May-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    1/27

    This article was downloaded by:[Hiroshima University]On: 20 June 2008Access Details: [subscription number 789277225]Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

    Journal of Development StudiesPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t713395137

    Exports and economic growth in BangladeshShamshad Begum a; Abul F. M. Shamsuddin ba Senior Lecturer, School of Accounting, Finance and Entreprneurship, VicePresident of Prime Bank, Dhaka, Bangladeshb University of New England, Australia

    Online Publication Date: 01 October 1998

    To cite this Article: Begum, Shamshad and Shamsuddin, Abul F. M. (1998)'Exports and economic growth in Bangladesh', Journal of Development Studies,35:1, 89 114

    To link to this article: DOI: 10.1080/00220389808422556URL: http://dx.doi.org/10.1080/00220389808422556

    PLEASE SCROLL DOWN FOR ARTICLE

    Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

    This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction,re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expresslyforbidden.

    The publisher does not give any warranty express or implied or make any representation that the contents will becomplete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should beindependently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,

    demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with orarising out of the use of this material.

    http://www.informaworld.com/smpp/title~content=t713395137http://dx.doi.org/10.1080/00220389808422556http://www.informaworld.com/terms-and-conditions-of-access.pdfhttp://www.informaworld.com/terms-and-conditions-of-access.pdfhttp://dx.doi.org/10.1080/00220389808422556http://www.informaworld.com/smpp/title~content=t713395137
  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    2/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    Exports and Economic Growthin Bangladesh

    SHAMSHAD BEGUM andABUL F. M. SHAMSUDDIN

    This study investigates the effect of exports on economic growth inBangladesh, based on a two-sector growth model. Using annualdata for the period 1961-92, the article estimates anAutoregressive Conditional Heteroscedastic model of economicgrowth, which is found to capture the volatility of the Bangladesheconomy. The results suggest that an increase in the share ofinvestment in GDP significantly increases the growth rate of GDPin normal years, but negligibly increases GD P growth in abnormalyears. Abnormalities in the economy arise from war, politicalturmoil and natural disasters. The key finding is that export growthhas significantly increased economic growth through its positiveimpac t on total factor produc tivity in the economy. Thecontribution of exports to economic growth was more pronouncedduring 1982-90 when the government pursued a policy of tradeliberalisation and structural reform, and political turmoil was no tpersistent. This finding is not sensitive to the choice of the mo delor the estimation technique.

    I. INTRODUCTIONThe recent literature on economic growth has focused considerableattention on determining the effect of foreign trade on economic growth.This article attem pts to analyse the effect of exports on econo m ic grow th inBangladesh by adopting the supply side approach.1 Our model assumes thatthere are two sectors in the economy, the non-exp ort and the export sectors,Shamshad Begum is Vice President of Prime Bank, Dhaka, Bangladesh. Abul Shamsuddin isSenior Lecturer, School of Accounting, Finance and Entreprneurship, University of NewEngland, Australia. The authors would like to thank R. Holmes, M. Khan, D. Schulze, H.Zafarullah and an anonym ous referee of this journal for helpful comm ents.The Journal of Development Studies, Vol.35, N o.1 , October 1998, pp.89-114PUBLISHED BY FRANK CASS, LONDON

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    3/27

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    4/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June2008

    EXPORTS AND ECONOM IC GROWTH IN BANGLADESH 91function [Romer, 1989]. Freer trade exposes domestic firms to foreigncompetition and forces them to rationalise their activities. It providesdomestic firms access to wider markets and enables them to gain technicalefficiency by exploiting both economies of scale and economies of scope.

    Rodrik [1992] provides an excellent discussion on the limitations oftrade policy reform in the context of developing countries. He draws themain reasons for trade reform in developing countries from the domain ofpolitical economy. First, a macroeconomic crisis encourages politicians toem brace trade policy reform. T he political costs of unde rtaking trade reformare perceived to be low during a national economic crisis. Second, foreigncreditors, particularly the IMF and World Bank, play an important role inpersuading cash-starved governments to undertake trade policy reform.However, the success of trade reform depends on some key factors:macroeconomic stability, the credibility of trade reform, and marketstructure.

    Many developing countries undertook trade policy reform in the 1980s,wh ich wa s also a decade of intense ma croecon om ic instability. Trade reformimproves economic welfare by reducing the distortions in relative pricesand b y chan nelling resources to sectors that can use them mo st efficiently.M acroec onom ic instability interferes w ith the effectiveness of trade reform.For example, during high and volatile inflation, economic agents may notbe able to disentangle relative price changes from changes in the generalprice level. Hen ce, they m ay not be able to respond p romp tly to trade reformwhich may slow economic activity and exacerbate unemployment at least inthe short term. A credible trade reform accompanied by a stablemacroeconomic environment provide unambiguous price signals toeconomic agents and enable them to take advantage of emergingopportunities from freer trade. However, the welfare consequences of atrade liberalisation in an economy with imperfect competition is ambiguousbecause freer trade increases welfare by improving the overall efficiency inresource allocation and consumption but it also decreases welfare ifunexploited scale economies exist in import-competing industries.Em pirical Evaluation of Trade Policy ReformSeveral studies in recent years have evaluated the impact of trade policyreform on economic performance in developing countries. Lopez [1991]provides evidence of the relationship between trade and macroeconomicpolicy and economic growth using cross-sectional data for 35 developingcountries. It is observed that sustainable policies (as in Korea, Taiwan,Singapore, Hong Kong, Thailand and Malaysia) promote growth throughrelatively stable real exchange rates that are either fully aligned orundervalued for prolonged periods o f tim e. Export-promo tion policies w ere

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    5/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    92 THE JOURNAL OF DEVELOPMENT STUDIESfound to be more effective in generating growth than policies that removeimport restrictions. However, contrary to conventional wisdom, Lopezfound that capital accumulation is stimulated by export restrictions and isnot directly sensitive to economic instability.

    Edwards [1992] uses alternative measures of trade orientation and tradedistortions to examine the role of trade policy in explaining cross-countrygrowth differentials. His study is based on a model where the country'scapacity to absorb new spillovers of world technology is negativelydependent on the level of trade distortions. In other words, countries thatliberalise trade tend to accumulate knowledge at higher rates and growfaster. In his basic empirical mode l, Edw ards ex presses the average growthrate of per capita real GDP as a function of the investment-GDP ratio, thegap between the world's and cou ntry's stock of knowledge and an index oftrade intervention or openness of the economy. The main results areobtained from cross-sectional data for 30 developing countries. The resultssuggest a strong and robust relationship between trade orientation andeconomic growth. That is, countries with more open and less distorted tradepolicies have experienced higher growth rates than those with moredistorted trade policies. This finding is robust to the estimation techniqueand the use of alternative indicators of trade orientation.Clarke and Kirkpatrick [1992] evaluate the effect of trade policy reformon economic performance in developing countries using a multivariateeconometric model. The performance indicators are: growth of GDP,growth of merchandise exports, the real exchange rate, growth of labourproductivity, balance of trade, growth of merchandise imports, and growthof manufacturing value added. Pooling data for 80 low- and middle-incomecountries for eight years (1981-88), they estimated the impact of tradepolicy reform on each of these performance indicators. This study suggests

    that lagged export growth leads to GDP growth and vice versa. With regardto the role of trade policy, they make the following remark: 'The overallconclusion of this study must be that there is no empirical evidence tosuggest that economic performance benefits from trade reform strategies'[Clarke and Kirkpatrick, 1992: 69].This result was attributed to some limitations of their methodology suchas omission of country-specific institutional characteristics and relativelevels of development. The former aspect is rightly emphasised in a recent

    book on East Asian Newly Industrialising Economies [Chowdhury andIslam, 1993: Ch.5]. Reviewing the debate on trade liberalisation andgrowth, they draw explanations for the spectacular success of trade policyreform in N IE s from the dom ain of political economy. The success of tradepolicy reform crucially depends on the ability and willingness of economicagents to take advantage of new opportunities created by the reform.

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    6/27

    DownloadedBy:[HiroshimaUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOM IC GROWTH IN BANGLAD ESH 93Governments in East Asia played a vital role in creating an institutionalframework conducive to export-oriented growth. As Chowdhury and Islam[1993: 86] state:

    [G ov ern m en ts in East Asia have in general been able to withstand thecontend ing pressures of distributional coalitions in order to im plem entcoherent economic policies that led to trade dependence. Oncechanges in domestic fortunes became closely linked to changes inworld market conditions, a virtuous circle set in leading to a processof simultaneous interaction between exports and rapid economicgrowth.An alternative to conventional econometric modelling is applied generalequilibrium analysis which allows interactions between sectors and includesmore general specifications of the behaviour of producers and consumers.Evaluating the literature on applied general equilibrium analysis of tradeliberalisation, Lai and Rajapatiram [1987] point out that under a generalequilibrium mo del with con stant returns to scale, net static gains from tradeliberalisation are small. However, once imperfect competition and scaleeconomies are taken into account, a multi-lateral reduction of all tariffsleads to a welfare gain of more than five per cent of GDP as was seen in a

    Canadian context [Harris, 1983]. Gunasekera and Tyers [1991] use anapplied general equilibrium model for South Korea under the assumptionsof imperfect competition and scale economies. Their simulation resultsindicate substantial gains from trade liberalisation.Cross-Country Studies on Exports an d G rowthCross-country emp irical studies use conventional econom etric techniques t oestimate the potential impact of exports on growth. In general, the earlyliterature supports the hypothesis of a positive correlation between exportsand economic growth. Michaely [1977] estimated the Spearman correlationcoefficient between the grow th rate of export share of GD P and the grow thrate of GDP using a sample of 41 less developed countries (LDCs) for theperiod 19 50 -7 3. H e found a statistically significant positive correlationbetween the two variables. However,

    [T]he positive association of the economy's growth rate with thegrow th of the export share appears to be particularly strong am ong themore developed countries, and not to exist at all among the leastdevelop ed . . . . This seem s to indicate that grow th is affected b y expo rtperformance only once countries achieve some minimum level ofdevelopment [Michaely, 1977: 52].

    Helleiner's [1986] study of low-income countries for the period 1960-80

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    7/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    94 THE JOURNAL OF DEVELOPMENT STUDIESprovides further support for Michaely's finding for least developedcountries. However, the sample used in Helleiner's study was over-represented by inward-oriented sub-Saharan countries. Thus, it was notsurprising to find no significant relationship between export growth andeconomic growth.

    Since a correlation coefficient does not provid e an estimate o f the pa rtialeffect of exports on GDP, a more sophisticated econometric analysis undera neoclassical production function framework was pursued by manyauthors. The essence of this approach was that exports raised total factorproductivity through their favourable effect on the efficiency of resourceallocation, economies of scale, capacity utilisation and technologicalchange. In his 1982 cross-sectional study, Feder employed an intuitivelyappealing model of GDP growth. In this model, export growth can affectGDP growth if factor productivity in the export sector differs from that ofthe non-export sector, and/or the export sector generates productionexternalities. The results support the hypothesis that export growth leads toGDP growth.5Lai and Rajapatiram [1987] also provide estimates of a GDP growthmodel using data from the 1986 World Development Report. The empiricalresults for a sample of 18 low-income countries indicate that the growth ofexports share of GDP over the period 1965-73 significantly increaseseconomic growth over the period 1973-84.

    Time Series Studies on Exports and GrowthMost of the studies mentioned above rely on cross-country data. Sinceinstitutional characteristics and the quality of inputs vary widely acrosscoun tries, no firm conc lusion can be drawn from studies wh ich do not allowthe structure of the growth model to vary across countries. A country-specific study based on time series data is not subject to this shortcoming.This view is supported by Edw ards [1993] in a comprehe nsive review of them odern emp irical literature on trade-orientation and grow th. With regard tocross-country empirical analysis, Edwards makes the following remark:'Ap plied econo m ists often ask too mu ch of their data sets, and try to ex tractinformation that simply is not there. In that sense, cross-country aggregatedata sets have little information regarding the relationship between tradepolicy and growth' [1993: 1390]. The country-specific analyses have beenuseful not only in enhancing academic literature about how different tradereform policies have affected growth in developing countries but also ininfluencing policy circles. How ever, most of these country-specific studiesare based on a traditional neo-classical framework where, in long-runequilibrium, the steady-state rate of grow th is independent of trade policies.As a modelling strategy, many applied researchers have invoked the non-

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    8/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOMIC GROWTH IN BANGLADESH 95steady state assumptions, such as the externality effect of exports andproductivity differential between the export and non-export sectors, in atraditional growth model [e.g., Feder, 1982]. Edwards argues that the newvintage of endogenous growth models provides a more promisingframework to establish a long-run equilibrium relationship between tradepolicies and economic growth.

    In the context of Bangladesh, the evaluation of the effect of exports oneconom ic grow th is an under-researched area. Our literature search failed toidentify any study that uses a theoretically sound model to examine theimpact of exports on growth in Bangladesh . Roy [1991] and Rahman [1993]provide some evidence of the impact of exports on economic growth. Royinvestigates Bangladesh's export performance and provides somepreliminary evidence of the relationship between exports and GNP. Usingannual data for 1975-88, Roy observes a large positive rank correlationcoefficient between GNP and the share of exports in GNP. In addition, hefound a positive correlation between the export and non-export sectorsoutput. The correlation result is indicative of a strong relationship; however,no firm conclusion can be drawn from Roy's study about the partial impactof exports on economic growth.

    Rahman [1993] evaluated the effect of exports on economic growthapplying an ad hoc regression model of GN P on a sample of only 14 annualobservations, covering the period 1972-73 to 1985-86. Regressing thecurrent level of GNP on the last period's levels of exports and investment,he ob serves that a unit (1 Taka) increase in the last period's exp orts leads toat least an 11 unit increase in current GNP. This surprisingly large effect ofexports on GNP seems to arise from the misspecification of the model, thenon-stationarity of the data and the small size of the sample. The model ismisspecified because it excludes an important input, labour, and expressesthe level of final output as a function of investment, rather than the stock ofcapital (K). The first difference of final output (AGNP) is dependent oninvestment, that is, the first difference of capital stock (AK). Furthermore,both GNP and exports in Bangladesh contain strong trend components.Hence, a regression in the levels of the variables produces spurious resultsand the co nventional t and F tests are biased towards the rejection of the nullhypothesis of no relationship. Rahman also provides evidence for Granger-causality between manufacturing GDP and manufacturing export based onannual data for the period 1972-73 to 1990-91. The results suggest thatman ufacturing exports cause an expansion of manufacturing GDP. This testis a heroic attempt because as a rule of thumb, the literature on time serieseconom etrics advo cates the use of a large sample (at least 40 observations)to cond uct the causality test. Alth ou gh th e empirical literature for export-ledgrowth is limited in the context of Bangladesh, a large number of studies

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    9/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June2008

    96 THE JOURNAL OF DEVELOPMENT STUDIEShave been undertaken by researchers for other countries after the mid-1970s.Jung and Marshall [1985] conducted the Granger causality test usingdata on 37 developing countries. They found support for the export-ledgrowth hypothesis for four countries: Indonesia, Egypt, Costa Rica, andEcuador. Surprisingly, they found no evidence for the export-led growthhypothesis for countries that are well known for export promotion policiessuch as South Korea, Taiwan and Brazil. Darratt [1986] also observes thatexports did not Granger-cause economic growth for any of the four newlyindustrialised economies of Asia (Hong Kong, Korea, Singapore andTaiwan) over the period 1960-82. Further evidence against the export-ledgrowth hypothesis comes from a recent Canadian study by Henriques andSadorsky [1996]. Using a VAR model, they found that export growth didnot Granger-cause GDP growth in Canada over the period 1870-1991.

    Applying a vector auto-regressive model to Taiwan, USA and Japan,Ghartey [1993] observes that export growth Granger-causes economicgrowth in Taiwan, economic growth Granger-causes export growth in theUSA, and a feedback casual relationship exists in the case of Japan. Thetime series literature casts some doubt on the effectiveness of an export-ledgrowth strategy.However, the results from the atheoretical Granger causality test cannotbe used to reject the claim that exports lead economic growth. The Grangercausality test is a test of precedenc e w hich exa mines the precedence of onevariable over another. A test for precedence does not necessarily imply atest for eco nom ic causality. Learne r's [1985] remark is worth m entioning inthis context: 'W e can all think of contexts in wh ich prece denc e is sugg estiveof causation and also contexts in which it is n o t . . . . It is altogether clear thatprecedence is not sufficient for causality. Weather forecasts regularly

    precede the weather, but few of us take this as evidence that the forecasts"cause" the weather ' [1985: 259, 283]. Thus further research usingalternative analytical tools might contribute to the ongoing debate onexport-oriented growth policy in LDCs like Bangladesh. Considering thelimitations of the Granger-causality test and the small size of our sample,this article employed a structural econometric model as the principalmethod of analysis. The Granger-causality test is used as a supplementarytool to verify the key finding from our structural model.

    III. STYLISED FACTSThis section describes the stylised facts of export performance andeconomic grow th in Bangladesh over the 196 1-92 period. The whole p eriodis divided into three identifiable policy regimes: pre-liberation regime

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    10/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOMIC GROWTH IN BANGLADESH 97(19 61 -71 ), post-liberation regime I (19 72-8 2) and p ost-liberation regime II(1983-92) .The main characteristics of the pre-liberation regime were a peggedforeign exchange rate, import-substituting industr ialisation andspecialisation in a few prim ary prod ucts. The policies of a pegged exchan gerate and import substitution were m aintained in the post-liberation regime I.The most distinguishing policy of this regime was the nationalisation ofmajor industries such as jute, cotton, textiles, sugar mills, banks andinsurance companies. Until 1975, the government pursued a publicownership strategy to attain a socialist economy. However, this goal wasabandoned in late 1975 in favour of a mixed economy. Finally, in the post-l iberation regime II , the government pursued the policies ofdenationalisation of the banking and industrial sectors, deregulation of thecapital market, structural reform of the tax system and trade liberalisation(under the sponsorship of the World Bank). This regime witnessed thereplacement of import-substituting industrialisation (ISI) by a policy ofexport-oriented industrialisation (EOI) and the introduction of a managedflexible exchange rate policy. As trade liberalisation allows the exportsector to grow along the line of comparative advantages, one may expect amo re pronounced contribution o f exports to econom ic growth in a regime offreer trade.Table 1 depicts the statistical properties of the data for each regime.Variable definitions and data sources are given in the Appendix (see TableA l ) . Real per capita income and the investment-GDP ratio were higher inthe post-liberation period than in the pre-liberation period. The averageannual growth rate of GDP increased from 4.2 per cent in 1961-71 to only4.4 per cent in 197 2-92 despite a large increase in the grow th rate of exports(2.7 to 12.9 per cent) over this period. It seems to indicate a weakrelationship between exports and economic growth.Furtherm ore, the average annual expo rt growth rate wa s 15.2 per cent inthe post-liberation reg ime I, and 10.1 per cent in the post-liberation regim eII, which does not seem to be consistent with the notion that an outwardorientation of the economy improves export performance. In general, thedescriptive statistics suggest that despite structural reform and fluctuationsin export growth, the GDP growth rate remained just above four per centover the period 1961-92.

    Further evidence on the relationship between exports and economicgrowth is depicted in Table 2. Since exports are a component of GDP, onewould normally expect a high correlation coefficient between GDP andexports. In order to determine a systematic economic relationship ratherthan an accounting relationship, we estimate alternative measures of thecorrelation coefficient. The second column of Table 2 presents correlation

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    11/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    98 THE JOURNAL OF DEVELOPMENT STUDIESTABLE 1DESCRIPTIVE STATISTICS OF THE VARIABLES8

    Variables

    Mean real per capitaGDP (in Taka)Mean export-GDP ratio

    Mean investment-GDPratioAverage annual growthrate of GDPAverage annual growthrate of PopulationAverage annual growthrate of exportAverage annual weightedgrowth rate of export

    Pre-Iiberationregime:1961-62 to1971-72

    3014.1(0.080)0.094(0.226)0.098(0.244)0.042(1.067)0.023(0.231)0.027b(2.750)0.03 l b(2.60)

    Me an share of manufacturing goods in exportsMean share of privateinvestment in grossdomestic investment

    Post-liberationregime:1972-73 to1991-92

    4086.2(0.091)0.094(0.336)0.116(0.273)0.044(0.873)0.022(0.127)0.129(1.620)0.013(1.297)0.399(0.161)0.558(0.191)

    Post-liberationregime I:1972-73 to1981-82

    3802.2(0.041)0.076(0.267)0.103(0.333)0.047(1.096)0.023(0.109)0.152(1.609)0.013(1.296)0.376(0.195)0.573(0.241)

    Post-liberationregime II:1982-83 to1991-92

    4433.3(0.053)0.117(0.243)0.131(0.154)0.041(0.323)0.020(0.079)0.101(1.637)0.014(1.367)0.407(0.109)0.540(0.093)

    Notes:a Figures in parentheses are the coefficients of variation.b These figures are calculated by excluding two years of political instability and war (1970-71and 1971-72) from the sample.

    coefficients for the full sample. The correlation between exports and non-export output is very high (0.83) which may be due to a strong trendcomponent in both variables or Feder's [1982] hypothesis that the exportsector generates positive prod uction externality. The correlation coefficientis 0.43 betwee n G DP an d the export share of GDP, and 0.33 be tween grow thrate of GD P and growth rate of exp orts. Thu s, the correlation results for thefull sample seem to provide tentative support for the export-led growthhypothesis. However, this support is weakened by the results for individualpolicy regimes. In particular, the correlation coefficients for the post-

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    12/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    E X P O R T S A N D E C O N O M I C G R O W T H IN B A N G L A D E S HTABLE 2CORRELATION RESULTS FOR EXPORTS AND GDP

    99

    Variables

    Export and non-export

    GDP and share of exportin GDPGrowth rate of GDP andgrowth rate of exportGrowth rate of GDP andweighted growth rateof export

    Fullsample:1961-62 to1991-92

    0.828(7.95)0.428(2.55)0.331(1.88)0.313

    Pre-liberationperiod:1961-62 to1971-72

    0.766

    0.868

    0.410

    0.418

    Post-liberationregime I:1972-73 to1981-82

    0.779

    0.732

    0.306

    0.348

    Post-liberationregime II:1982-83 to1991-92

    0.855

    0.859

    0.417

    0.455

    Notes: Figures in parentheses show compu ted t-statistics. The c ritical t-value at the per cent levelof significance is 2.04. t-statistics for the sub-periods are not reported due to the small sizeof the sample.liberation regime II, when an EOI policy was followed, are quite similar tothose of the pre-liberation regime when an ISI policy was followed. Thisfinding seems to suggest that the nature of the policy regime may not be acrucial factor in determining the effects of exports on economic growth.However, one should not be tempted to reject the export-led growthhyp othesis as the correlation an alysis does not take into accoun t the effectsof other factors on economic growth.

    IV. ANALYTICAL FRAMEWORKThis section presents the analytical framework of the study. For a smallopen economy like Bangladesh it seems realistic to assume that it isdomestic supply rather than foreign demand which imposes a bindingconstraint on the grow th of the export sector.6 H enc e, in mode lling the effectof exports on econom ic grow th we focus on the supply side of the economy.There are two channels through w hich exports affect econom ic growth.These are the externality effect and the productivity differential effect. Theexport sector operates in a highly competitive environment and to succeedin global competition it invests in advanced production techniques,workers' training and infrastructure. The non-export sector receives indirectbenefits from the export sector in two ways: the demonstration effect andthe public good characteristics of some export activities. The non-exportsector can replicate the management practices, improved production

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    13/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June2008

    100 THE JOURNAL OF DEVELOPMENT STUDIEStechniques and marketing strategies of the export sector and it may getready access to public infrastructure (transport and communicationfacilities) developed for the export sector. These indirect non-pricedbenefits to the non-export sector are referred to as the positive productionexternalities of the export sector. An expansion of a sector with positiveproduction externalities increases GDP, ceteris paribus.

    Factor productivity in the export sector may differ from that in the non-export sector for several reasons. First, export industries use improvedtechnology and skilled workers. Second, factors of production are notperfectly mobile between the two sectors due to institutional constraints aswell as differential skill requirements across sectors. In the absence ofperfect mobility of factors, the value of the marginal product of each inputin the export sector is likely to be higher than that of the non-export sector.The productivity gap would be larger if the skill composition of the workforce differed between sectors. Thus, export expansion may raise the totalfactor productivity in the economy through its favourable effect on theefficiency of resource allocation.

    It is assumed that the production function differs across sectors butremains constant within a sector over time. The allocation of resourcesbetween the export and non-export sectors is non-optimal due to thepresence of non-priced externalities and imperfect mobility of factors ofproduction.7 Adopting Feder 's [1982] theoretical framework, the supplyside of the econom y can be expressed as:

    Y = N + X (1)N = F ( K N , L N , X) (2)

    X = H ( K X , L X ) (3)Equation (1) states that the gross domestic product (Y) is the sum of thenon-exp ort sector output (N) and the export sector output (X ). % and K xare the sector-specific capital stock, and L N and L x are sector-specificlabour em ploym ent. Taking the total differential of equations (2) and (3), weget:

    dN = F K dK N + F L d L N + F x dX (4)dX = H K d K x + H L d L x (5)

    where, F ; and Hj (i =K, L) are the marginal productivities of input i in the

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    14/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOMIC GROWTH IN BANGLAD ESH 101non-export and export sectors respectively; F x represents the marginalexternality effect of X on N. Invoking the assumption that the productivityof each input in the export sector differs from that of the non-export sectorby a factor 8, i.e., H , = (1 + 8) F; and substituting equa tions (4) and (5) intothe identity dYs dN + dX, we obtain:

    dY = F K dK N + F L d L N + F x dX + (1+ 8)F K d K x + ( l + 8 ) F L d L x (6 )or

    dY = FL.L (f) +FKM -) + M - + F x ) ( ? ? )where, L = L X + L N , I = dK x + d K N , and ((8/1+8) + F x ) is the sum of theproductivity differential and the production externality effects. The term,X/Y(dX/X) can be interpreted as the weighted export growth, that is, theexport growth rate weighted by the share of exports in GDP. The modelstates that the annual growth rate of GDP is linearly dependent on theannual growth rate of the labour force, the annual investment-GDP ratio,and the weighted growth rate of exports. As noted earlier, F L and F Krepresent the marginal productivity of labour and capital respectively in thenon-export sector of the economy. The coefficient of the weighted exportgrowth rate can be interpreted as the social marginal productivity gapbetw een the export and the non-exp ort sector relative to the private m aiginalproductivity in the export sector.8 Marginal productivity of a factor for thewhole econ omy can be found by rew riting equation (6) as:

    dY = F KdK + F L dL + 8 F K d K x + 8 F L d L x + F x dX (8)Economy-wide marginal productivity (MP) of a factor depends on itsproductivity in the non-export sector, the extent of the productivitydifferential and the allocation of the factor between the two sectors. Wederive the following expressions for MPs from the above equation?

    (9)

    (10)where I x /I is the share of export sector investment in total investment andd L x /dL is the proportion of additional employment created in the exportsector. The marginal productivity of an input for the whole economy would

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    15/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    102 THE JOURNAL OF DEVELOPMENT STUDIESbe the same as that of the non-export sector if there was no productivitydifferential (8 = 0) between the export and the non-export sector or ifadditional inputs were employed only in the non-export sector In general,the marginal productivity of capital for the whole economy will be(8(Ix/I))l00 per cent higher than the marginal productivity of capital in thenon-export sector.Empirical IssuesA preliminary investigation of the data suggests that in abnormal years(periods of war, natural calam ity and political instability), G D P grow th rateswere negative or close to zero but the investment-GDP ratios were large. Inabnormal years, a large part of public investment, mainly financed byforeign donors, was used to rebuild the economic infrastructure andproduction plants destroyed by natural and man-made calamities, ratherthan to augment the pre-disaster stock of capital. In order to address thisempirical issue, equation (7) is augmented by including an interactionvariable D I/Y, which is the product of the dummy variable for theabnormal years (D) and the investm ent-GD P ratio.

    Furthermore, equation (7) implici t ly assumes that the state oftechnology, the skill composition of the labour force and the characteristicsof the socio-economic institutions remained unchanged over the sampleperiod, 196 1-92 . Strictly speaking, none of the above factors wa s constantduring this period. In Ban gladesh, the last three decad es have w itnessed thegradual adoption of the High Yielding Varieties (HYV) technology inagriculture; the emergence of the garment industry which uses industry-specific skills; a significant increase in the adult literacy rate and anexpansion of the underground economy through the institutionalisation ofcorruption. Reliable time series data on each of the above factors are notavailable for the full sample period.

    10As a second-best option, the empiricalmodel is modified by including a trend variable (T), which is expected topartially capture the composite effects of these factors. Thus, we adopt thefollowing empirical specification:

    Y = Po + Pii + P 2(-7-) + P 3 ^ .x ) + W.y + frT+e (11)

    where e represents a random error which is normally distributed with meanzero and variance a 2 ; the dot (') indicates the annual growth rate. Expectedsigns of the coefficients are: Pj > 0, P 2 > 0, P 3 > 0, P 4 < 0, The sign of P 5is a priori indeterminate.

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    16/27

    DownloadedBy:[HiroshimaUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOMIC GROWTH IN BANGLAD ESH 103V. EM PIR IC A L R ESULT SThe OLS ResultsInitially, we apply the OLS estimator to estimate both the basic model (7)and the augm ented equation (11). The OLS resu lts are presented in Table 3.The first row of the table shows that the basic model performs poorly. Allbut the weigh ted exp ort growth rate have statistically insignificant effectson the growth rate of GDP and the investment-GDP ratio has a negativecoefficient. The modification of our initial specification by including aninteraction dummy and a trend variable increases the R 2 from 0.16 to 0.36.A partial F-test also justifies the inclusion of these additional regressors inthe initial empirical specification. Both population growth and weightedexport growth have the expected significant positive effects on GDPgrowth. The coefficient of the interaction dummy is negative and highlysignificant and its absolute value is larger than that of the investment-GDPratio. This m eans that the marginal productivity of capital in the non -expo rtsector is positive in a normal year and negative in an abnormal yeanHowever, the former is not statistically different from zero.

    The OLS result on the coefficient of the investment-GDP ratio ispuz zling. This coefficient m easures the m arginal productivity of capital andis found to be insignificantly different from zero, which is unexpected in acapital-scarce econom y like Bangladesh." Hen ce, we conducted diagnostictests on the OL S residuals to verify wh ether they satisfy th e assum ptions ofthe classical linear regression model. The residual diagnostics indicate thatlarge and small residuals occur in clusters. This finding seem s to imply thatthe variance of the current error depends on the size of the last period'serror. In other wo rds, the conditional error varian ce, Var[e t/e t_{\ follows anauto-regressive process. This observation induces us to apply a formal testfor the Autoregressive Conditional Heteroscedastic (ARCH) process.12 ALagrange Multiplier (LM) test for the first-order ARCH process isconducted. The null hypothesis is that the error variance does not follow anautoregressive process, that is, the volatility of GDP growth in the currentperiod is indepen dent of the last perio d's volatility o f GDP. The LM statisticfollows a%2 distribution with 1 degree of freedom. The last column of Table3 shows that for both OLS equations, the computed y} is greater than thecritical x 2 = 3.84 at the 5 per cent level of significance. Hence, the nullhypothesis of no ARCH effect is rejected.The ARCH ResultsThe LM test results suggest that the classical linear regression model isincompatible with the data and the OLS is no longer the most efficientestimator. Hence, we adopt the estimation technique developed by Engle

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    17/27

    DownloadedBy:[HiroshimaUniversity]At:03:3920June20

    08

    104 THE JOURNAL OF DEVELOPMENT STUDIESTABLE 3

    THE OLS RESULTS: 1962-92(DEPENDENT VARIABLE: ANNUAL GROWTH RATE OF GDP)

    Equation

    Constant

    L

    I/Y

    (X/Y)X

    D.I/Y

    T

    R2 (F-value)

    DWLM

    Equation (7): no interactiondummy, no trend variable

    -O.025(-0.38)

    2.721(1.27)-0.010

    (-0.04)1.037**

    (2.17)-

    -

    0.16(1.67)2.398.68

    Equation (11): with interactiondummy and a trend variable

    -0.059(-0.97)

    4.148**(2.02)0.229

    (0.89)1.521**

    (3.22)-0.422*(3.22)-O.001*

    (-1.47)0.36

    (2.79)2.51

    10.96Notes:Numbers in parentheses are t-statistics.DW is the test statistic for first-order autocorrelation in the error term.LM is the test statistic for the first-order ARCH process in the error term.** The coefficient is statistically significant at the five per cent level.* The coefficient is statistically significant at the ten per cent level.

    [1982] which involves the maximum likelihood estimation (MLE) of themodel with an ARCH process. The results for the modified GDP growthequation (11) and the corresponding error variance equation are presentedbelow (t-statistics are in parentheses).The Growth Model:Y = -0.045 + 0.351(//T) + 2.589L + 0.957(X/Y. X) - 0.301(Z). I/Y)- 0.0017/ (12)

    (-0.99) (1.92) (1.69) (2.88) (-2.83) . (-1.75)Log-likelihood function= 66.51, DW= 2.52, Q-statistic= 9.37 at lag 8.

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    18/27

    DownloadedBy:[HiroshimaUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOMIC GROWTH IN BANGLA DESH 105Error Variance Equation:

    aj = 0.0004 + 0.598e?_! (13)(2.46) (1.69)where, aj is the variance conditional on past error, that is, a} =Var[e t Ie t_^\ and numbers in parentheses are t-statistics.14 The estimatedvariance equation satisfies the stationarity condition since the coefficient ofe^_l is less than unity. Bo th the Du rbin-Watson (DW ) and Ljung-B ox-Pierce(Q-statistic) statistics suggest the absence of an autocorrelation problem.15The ARCH model yields more efficient parameter estimates than the OLSestimates. For all explanatory variables, we obtain theoretically expectedcoefficients with smaller standard errors.

    The marginal productivity of capital in the non-export sector is 0.35 in anormal year and 0.05 (= 0.35-0.30) in an abnormal yean The populationgrow th rate variable obtains a coefficient of 2.59 w ith a t-statistic of 1.69.16With regard to the impact of exports, the result suggests that if weightedexports grow by one per cent, GDP increases by 0.96 per cent. Putdifferently, the sum of the productivity differential and externality effect is0.96. Our model cannot isolate the productivity differential effect (5/1+8)from the externality effect (F x ). For a plausible range of F x values, wecalculate the corresponding values of factor productivites for the exportsector as well as for the whole economy. Table 4 presents these estimates.For example, when the marginal externality effect is 0.2, the marginalproductivity of capital is 1.44 in the export sector and 0.90 in the wholeeconomy.

    The total contribution of exports to economic growth can be estimatedas the produ ct of weigh ted ex port grow th and its coefficient. F igure 1 show sthat, for most years, the contribution of exports to economic growth waspositive. In particular, we observe that the contribution of exports to thegrowth rate of GDP was more pronounced during 1982-90 when thegovernment pursued policies of trade liberalisation and deregulation of theeconomy. Although the process of economic reform continued in the early1990s, the contribution of exports to the growth rate of GDP was close tozero in 1990-91 and 1991-92 because of intense political turmoil in theseyears.

    Th e coefficient of the trend va riable is neg ative and statistically differentfrom zero, which seems to suggest that positive effects of technologicalimprovement and human capital accumulation on economic growth wereoutweighed by the negative effects of corruption, political turmoil and theadoption of inappropriate technology. The persistent dependence ofBangladesh on concessional foreign loans over the post-liberation period

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    19/27

    DownloadedBy:[HiroshimaUniversity]At:03:3920June20

    08

    1 0 6 T H E JOURNAL O F DEVELOPMENT STUDIESTABLE 4

    PREDICTED MARGINAL PRODUCTIVITY OF CAPITALFOR POSSIBLE VALUES OF THE EXTERNALITY EFFECTPossible Values

    for MarginalExternalityEffects(Fx)0.10.20.30.40.50.60.70.80.91

    Productivitygap between theExport andNon-Export Sector

    (5)i5.9933.1151.9161.2570.8420.5550.3460.1860.060-0.041

    Marginalproductivity inthe Non-ExportSector

    (FK)20.3510.3510.3510.3510.3510.3510.3510.3510.3510.351

    Marginalproductivity inthe ExportSector

    (H K)32.4551.4441.0230.7920.6460.5460.4720.4160.3720.337

    Marginalproductivity inthe WholeEconomy(M PK )4

    1.4030.8980.6870.5720.4990.4480.4120.3840.3620.344Notes:1. Com puted from the relation, ($3 = 5/1+5 + F x where, {J3 is the coefficient of weighted exportgrowth.2. F K is the coefficient of the investment-GDP ratio.3. H K = ( 1 + 5 ) F K4. M P K = F K (1 + 5(IX/T)) where, Ix/I = 0.094. It is assumed that the investment share of theexport sector is same as the output share of the export sector

    has led to adoption of inappropriate technology and investment projectswith low returns. By mid-1986, the outstanding medium- and long-termforeign debt of Bangladesh increased to US$6.14 billion which was eighttimes the value of merchandise exports in 1985-86. The terms of foreignindebtedness were very favourable, with an annual average interest rate ofonly 1.5 per cent, an average repayment period of 39 years and a graceperiod of nine years [Khan and Hossain, 1989: 25\. The availability offoreign loans on these generous terms induced the governm ent to und ertakeprojects which were incompatible with the factor price structure ofBangladesh. Furthermore, the diffusion of publicly subsidised HYVtechnology in agriculture was successful in increasing land productivity butnot labour productivity over the period 196 7-91 [Islam and Taslim, 1996].This type of development strategy constrained the scope of alternativetechnological innovations or adoptions in the agricultural sector In the non-agricultural sector, structural reform took place in a distorted way. State-owned enterprises were sold below their market value to private business.Nationalised financial institutions were encouraged to make loans onconcessional terms to individuals with high-level political connections but

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    20/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June2008

    E X P O R T S A N D E C O N O M I C G R O W T H IN B A N G L A D E S H 107FIGURE 1CONTRIBUTION OF EXPORT GROWTH TO GDP GROWTH

    sOO HQO

    odowo

    0.20 T

    Post-liberationregime II-0 .15

    YEAR

    no business experience. Despite deregulation of the economy, import-competing industries have not yet been able to achieve any significantimprovement in technical efficiency. Thus, the negative coefficient of thetrend variable is not inconsistent with these stylised facts in Bangladesh.We now compare our finding on exports and growth with that of otherstudies. Table 5 presents the coefficients of export growth found in somerepresentative studies. The first four studies use simple export growth(AX/X) as an independent variable in the GDP growth equation, while therest of the studies use weighted export growth ((X/Y).(AX/X)). Thecoefficient of the simple export growth rate (coefficients with superscript'a ') represents the partial elasticity of GD P with respect to export, and it liesbetween 0.10 and 0.57. On the other hand, the coefficient of the weightedexport growth rate (coefficients with superscript 'b') lies between 0.42 and1.97. O ur A RC H m odel provides a coefficient of 0.957, which seems to bea plausible estimate as it lies within the range of previous estimates.

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    21/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    108 THE JOURNAL OF DEVELOPMENT STUDIESTABLE 5ESTIMATED EFFECTS OF EXPORTS ON ECONOMIC GROWTH

    Study Nature of the Study Coefficient of ExportGrowth in Models ofGDP GrowthKavoussi [1984:247] A cross-sectional analysis of 73developing countries, 1960 -78Ram [1985: 419] A cross-sectional study of 73 LDCs,1970-77Balassa [1985: 33] A cross-sectional study of 43 developingcountries, 1973-78

    0.105 a

    0.148 a

    0.161 a

    Tyler [1981: 128] A cross-sectional study of 55 middle-incomeLDCs, 1960-77Feder [1982:128] A cross-sectional study of 31 LDCs, 1964-73Ram [1987: 64] A study of 88 developing countries, 1960-82Greenaway andSapsford [1994:161]

    A time series study on Pakistan, 1971-85

    O.57Oa

    0.422b1.552b1.971b

    Present study A time series study on Bangladesh, 1962-92 0.957D

    Notes: a The coefficient of the growth rate of exports, X The coefficient of the weighted growth rate of export, X/Y. X

    Specification TestsThe Hausman specification test was conducted to examine whether or notthe growth model is subject to simultaneity bias. The null hypothesis of theexogeneity of exports is tested by implementing the Hausman test in twosteps. The first step involves a regression of the export variable oninstrumental variables" (which determine the demand for exports), and theremaining exogenous variables of the GDP growth model. In the secondstep, the GDP growth equation (11) is augmented by adding estimatedresiduals from the first step as a regressor. This augmented equation isestimated and a t-test is cond ucted for the coefficient of the residual se ries.The relevant t-statistic (1.28) suggests that the null hypothesis of theexogeneity of exports cannot be rejected at the five per cent level ofsignificance. In other words, the Hausman test confirms that GDP growth

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    22/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    EXPORTS AND ECONOMIC GROWTH IN BANGLADESH 109has no contemporaneous influence on exports. Thus, our empirical findingon the impact of exports on economic growth is not subject to simultaneitybias.

    Although the Hausman test confirms that GDP growth has nocontem poraneo us effect on expo rts, it says nothing abou t the potential non -contemporaneous impact of GDP growth on exports. In order to examinethis latter possibility, the Granger-causality test is employed. To test thehypothesis that export growth (x) leads to (or Granger-causes) economicgrowth (y) we estimate the following equation:m N

    yt = 0 + X ^t-i + 2 aixt-i + ztand to test the hypothesis that economic growth Granger-causes exportgrowth we estimate

    x, = Xn

    where, e^ and r | t are random error terms . We use the likelihood ratio test thatcompares the values of the maximised likelihood functions under the null(H o) and alternative (Hj) hypotheses. The null hypothesis that exportgrowth does not Granger-cause GDP growth is rejected at the 2.5 per centlevel of significance.18 This finding complements our results from thestructural model. We do not find any statistically significant evidence insupport of reverse causality which indicates that GDP growth does not leadto export growth. This result strengthens our finding from the Hau sm an test.VI. SUMMARY AND CONCLUSIONA common problem facing policy-makers in LDCs like Bangladesh is howto efficiently allocate scarce resources to increase the productive cap acity o fthe economy. Under our analytical framework, the essence of the problemis how to determine an optimal allocation of resources between the exportand non-export sector of the economy. In order to address this importantpolicy issue, this study put forward two basic questions: (1) Is the factorproductivity in the export sector greater than that of the non-export sector?(2) Does the export sector generate positive production externalities for thenon-export sector?

    If the sum of the productivity differential and externality effects ispositive, then any marginal reallocation of resources from the non-export

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    23/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    110 THE JOURNAL OF DEVELOPMENT STUDIESsector to the expo rt sector will increase econ om ic grow th. This is obv iouslyan em pirical issue and this pap er is an attem pt to investigate this issue. Theuniqu e features of this study, in the context of Bang ladesh, are the use of awell-structured theoretical framework, application of the ARCH model andutilisation of a relatively large sample (1961-92) for analysis. The majorfindings of this article are: The weighted grow th rate of exports has a positive effect on econo m icgrowth; that is, the sum of the productivity differential and externalityeffects of the export sector is po sitive . Th is finding is not sensitive to thespecification of the model and the estimation technique. If weighted

    exports grow by one per cent, GDP increases by 0.957 per cent. Theregression analysis also indicates that the contribution of exports toeconomic growth was more pronounced during 1982-90, when thegovernment pursued policies for the liberalisation of the economy. Theerror variance equation associated with the ARCH model suggests thatthe volatility in GDP growth is finite. This means that any volatility inGDP growth, originating from natural disaster or random humanbehaviour, diminishes over time. The marginal productivity of capital in the non-e xport sector is higher ina normal year (0.35) than in an abnormal year (0.05). Under theassum ption that the expo rt sector generates a marginal externality effectof 0.2, we find that marginal productivity of capital is 1.44 in the exportsector and 0.90 in the whole economy. Th e pop ulation grow th rate (a prox y for labour force grow th) ha s atheoretically expected positive effect on econo mic grow th.

    Overall, our study seems to support the export-led growth hypothesis.This study has obvious implications for the allocation of resources betweenthe export and non-export sectors. As the sum of the productivitydifferential and the externality effects is close to unity, any marginalreallocation of resources from the non-export to the export sector willenhance the productive capacity of the economy. The recent change in thestructure of exports towards labour intensive non-traditional products (forexample, garments and fish) also indicates that export expansion willpositively affect employment. Further research should be directed toquantify the impacts of export expansion on employment and incomedistribution. Analysis needs to be conducted at the disaggregated level toidentify the effects of the com position of expo rts on GDP, em ploym ent an dincome distribution. If the data become available, other important

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    24/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June2008

    EXPORTS AND ECONOMIC GROWTH IN BANGLAD ESH 111determinants of econom ic grow th in Bang ladesh such as skills of the labourforce and the extent of corruption should be included in the growth model.

    final version received February 1998N O T E S

    1. T his approach is formally developed by Fe der [1982].2. Strictly speaking, the predicted effect on GDP is consistent with the theory of comparativeadvantage if the non-export sector includes importables and excludes non-tradables. In ourmodel specification the non-export sector includes both importables and non-tradables.However, it does not create any methodological problem because we do not impose any apriori restriction on the potential effect of export growth on GDP growth.3. Consequently, the export sector changed significantly both in size and structure during the1980s. The share of export in GD P increased to around 12 per cent in 1989 -90 and the shareof primary export in total merchandise export decreased from 43 per cent in 1972-73 to 23per cent in 1989-90. Newly industrialising countries in South-East Asia were losing thecompetitive edge in labour-intensive products due to rapidly growing labour costs, whichpartially contributed to the growth of labour-intensive exportables in Banglade sh such as thegarment industry.4. To our knowledge there exists no published study which investigates the impact of exportgrowth on GDP grow th in Bangladesh under a well-structured empirical framework.5. Further studies by Balassa [1985], Ram [1987] and Greenaway and Sapsford [1994]complem ent the m ajor findings of Feder [1982].6. We recognise the importance of export demand in determining the structure of exports.Protectionist measures of foreign countries and business cycle fluctuations have obviouseffects on foreign demand for particular commodities such as clothing, footwear, leatherproducts, etc. However, an investigation of the export structure is beyond the scope of thisstudy.7. There are reasonable assumptions in the context of Bangladesh. First, leading export-oriented industries such as garments, leather and jute manufacturing use sophisticatedtechnology relative to that of leading non-export industries, such as food grains, housing andnon-traded services. Bangladesh can be viewed as a typical Lewis-Fei-Rains 'du al ec onom y'where most industries in the non-export sector use traditional technology while majorindustries in the export sector use modern technology imported from abroad. Second, factorsof production are not perfectly m obile across sectors in Bangaldesh. Skill requirements in theexport sector differ from those in the non-export sector. More specifically, export-orientedindustries are largely based on skills which are acquired primarily through on-the-jobtraining rather than through general secondary o r tertiary education. The non-export sector,on the other hand, is primarily based on skills acquired either through intergenerationaltransmission of skills (for example, commercial activities in the urban informal sector andsubsistence activities in the rural sector) or through the completion of general college oruniversity education. Thus, skills used in one sector may not be readily transferable to theother sector. Furthermore, in a hierarchical society such as Bangladesh, individuals oftenforego higher wages to enjoy the non-pecuniary benefits of working in white-collaroccupations, for examp le, higher social status. Individuals may prefer to w ork in the familybusiness with a subsistence reward instead of working in blue-collar occupations as wageworkers. These preferences simply reflect a strong positive relation between occupationalstatus and social status. Hence, one cannot expect perfect mobility of labour across sectorsin the presence of this cultural con straint.

    8. To further clarify the meaning of the coefficient of weighted export growth we write:

    1 + 8 + Fx ~ H:

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    25/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June2008

    112 THE J OUR N AL OF DE VELOP M E N T STUD I ESwhere, SMPi = Hi + FxH i. The social marginal product of factor i (i = K, L) in the exportsector (SMPi) takes into acocunt both the direct effect of factor i on GDP through its effecton exports (Hi) and the indirect effect of factor i on G D P through its externality effect on thenon- export sector FxH i.9. D ividing both sides of equation (8) by dK and setting dL = dX = d L x = 0, we find theexpression for M P K . A similar procedure is used to find M P L.10. Based on various sources, we were able to construct a time series for the primary enrolmentratio (a proxy for human capital accumulation) for the full sample range. This variable wasfound to be statistically insignificant in explaining economic growth of Bangladesh and wasconsequently omitted from our final econometric model.

    11 . Possible economic explanations for this result are the misallocation of resources in the publicsector and public investment on infrastructure. The contemporaneous effect of th e latter kindof investment on growth may not be significant.

    12. Heteroscedasticity is often associated with the cross- sectional, rather than time series data.However, analysing time series data on inflation, Engle [1982]observed that large and smallprediction errors occur in clusters, i.e. error variance follows an autoregressive process whichviolates the assumption of homoscedasticity.

    13. Initially we estimated both versions equations (7) and (11) of the growth model under theARCH framework. After investigating the results, we accept the modified GDP growthequation (11) because a likelihood ratio test provides justification for including theinteraction dummy and the trend variables in the model.

    14. Critical values of the t- statistics are 1.7 at the five per cent level and 1.32 at the ten per centlevel.

    15 . The Ljung-Box- Pierce Q- statistic follows a 2 distribution with L degrees of freedom, whereL is the maximum order of autocorrelation (p) tested. In our case, L= 8 and the nullhypothesis is: p1 = p2 = . .. = p8 = 0. The computed value of the test statistic is 9.37, whileth e critical value of 2 at the 5 per cent significance level is 15.51. Thus, the null hypothesesof no autocorrelation can not be rejected.

    16. Recall from equation (7) that this coefficient is the ratio of marginal productivity of labourto average productivity of labour. Using the sample mean of per capita G DP (TK 3671) as aproxy for average productivity, we derive an annual marginal productivity of a labourer ofonly TK 9508. No t e that the labour force data for the whole sample period are not availablefor Bangladesh. Hence , we are faced with two alternatives: one is to generate labour forcedata from the available data pools using an arbitrary growth formula, and the other is to usepopulation growth rate as a proxy variable. The former yields a labour force growth serieswith little fluctuation over time and use of this artifically-generated variable may not bedesirable in a regression equation. Therefore, this study uses the population growth rate as aproxy for the labour force growth rate, following the tradition in the literature [Feder, 1982;Balassa, 1985; Ram, 1987; Greenaway and Sapsford, 1994].

    17. The instrumental varaiables employed here are the annual growth rate of world income andth e difference between foreign and domestic inflation rates. Da t a are collected from variousissues of the International Financial Statistics, IMF.

    18. An empirical investigation of th e lag- structure leads us to choose a lag order of 4 in all cases.Th e likelihood ratio test statistic is: LR = 2[L(H 1) - L (H 0) ] . The LR has an approximate 2distirbution with J degrees of freedom. In our case, J= 4. Unde r the null hypothesis, exportgrowth does not Granger-cause GDP growth, we obtain LR = 2[64.30 - 59.72] = 9.16. Forth e null hypothesis that G D P growth does not Granger- cause export growth, we obtain LR =2[87.11 - 83.80] = 6.62. No t e that right- tail critical values of 2 distribution with 4 degreesof freedom are 9.49 and 11.14 at five per cent and 2.5 per cent levels of significance.

    REFERENCESAlamgir, M. and L. Berlage, 1972, An Analysis of National Accounts of Bangladesh 1949/50 -

    1968/69, Research Report Series No.7, Dhaka: Bangladesh Institute of DevelopmentEconomics .

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    26/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    E X P O R T S A N D E C O N O M I C G R O W T H IN B A N G L A D E S H 1 1 3Balassa, B., 1985, 'Exports, Policy Choices and Economic Growth in Developing Countries afterthe 1973 Oil Shock', Journal of Development Economics, Vol.18, No .2, pp.2 3-35 .Bangladesh Bureau of Statistics, 1993, Twen ty Years of National Accounting of Bangladesh,Dhaka: Statistics Division, Ministry of Planning.Chowdhury, A. and I. Islam, 1993, The Newly Industrialising Economies of East Asia, London:

    Routledge.Clarke, R. and C. Kirkpatrick, 1992, 'Trade Policy Reform and Economic performance indeveloping countries: Assessing the Empirical Evidence', in R. Adhikari, C. Kirkpatrick andJ. Weiss (eds.), Industrial and Trade Policy Reform in Developing Countries, Manchester:Manchaster University Press.Darrat, A. F., 1986, 'Trade and Development: The Asian Expe rience', Cato Journal, Vol.6, No .2,pp.695-700.Dornbusch, R., 1991, 'Policies to Move from Stabilization to Growth', Proceedings of the WorldBank Annual Conference on Development Economics 1990, World Bank.Dornbusch, R., 1992, 'The Case for Trade Liberalization in Developing Countries', Journal ofEconomic Perspectives, Vol.6, No .1, pp.69-85 .Edwards, S., 1992, 'Trade Orientation, Distortions and Growth in Developing Countries',Journal of Development Economics, Vol.39, No. 1, pp.3 1-57 .Edwards, S., 1993, 'Openness, trade liberalization and growth in developing countries', Journalof Economic Literature, Vol.XXXI, pp.1358-93.Engle, R., 1982, 'Autoregressive Conditional Heteroscedasticity with Estimate of theVariance ofUnited Kingdom Inflation', Econometrica, Vol.50, pp.987-10 08.Esfahani, H ., 1991, 'Exp orts, Imports and Econom ic Growth in S emi-Industrialised Cou ntries',Journal of Development Economics, Vol.35, No.1 , pp93-116.Feder, G., 1982, 'On exports and economic growth', Journal of Development Economics, Vol.12,No.1/2, pp.59-74.Ghartey, E.E., 1993, 'Causal Relationship between Exports and Economic Growth: SomeEmpirical Evidence in Taiwan, Japan and the US', Applied-Economics, Vol.25, No.9,pp.1145-52.Greenaway, D. and D. Sapsford, 1994, 'W hat Doe s Liberalisation Do for Exports and G row th? ',Weltwirtschaftliches Archiv, Band 130, pp.152-73.Gunasekera, D. and R. Tyers, 1991, 'Imperfect Competition and Returns to Scale in a NewlyIndustrialising Econ om y ', Journal of Development Economics, Vol.34, pp.223-4 7.Harris, R., 1983, Trade, Industrial Policy and Canad ian Manufacturing, Toronto: OntarioEconomic Council.Helleiner, G., 1986, 'Outward Orientation, Import Instability and African Economic Growth: anEmpirical Investigation', in S. Lall and F. Stewart (eds.), Theory and Reality in

    Development, London: Macmillan.Henriques, I. and P. Sadorsky, 1996, 'Export-Led Growth or Growth-Driven Exports? TheCanadian C ase' , Canadian Journal of Economics, Vol.XXIX, No.3, pp.540-55.Islam , T. and M. Taslim, 1996, 'Demographic Pressure, Technological Innovation and Welfare:The Case of the Agriculture of Bangladesh ', Journal of Development Studies, Vol.32, No .5,pp.734-70.Jung, W. and P. Marshall, 1985, 'Exports, Growth and Causality in Developing Countries',Journal of Development Economics, Vol.18, No.2, pp .l- 12 .Kavoussi, R. M., 1984, 'Export Expansion and Economic Growth. Further Empirical Evidence',Journal of Development Economics, Vol.14, No.1/2, p p.241 -50.Khan, A.R. and M. Hossain, 1989, The Strategy of Development in Bangladesh, London:Macmillan.Lal, D. and S. Rajapatirana, 1987, 'Fore ign Trade Regions and Econom ic Growth in Develop ingCountries ' , World Bank Research O bserver, Vol.2, No.2 , pp.189 -217.Leamer, E. E., 1985, Vector Autoregressions for Causal Inference? Carnegie Rochester Series onPublic Policy 22, Amsterdam: North Holland.Lopez, R., 1991, 'How Trade and Macroeconomic Policies Affect Economic Growth andCapital Accumulation in Developing Countries', Working Paper Series No.625, CountryEconomics Department, Washington, DC: World Bank.

  • 8/14/2019 Exports and Econ Growth in Bd Begum Shamsuddin

    27/27

    DownloadedBy:[Hiroshim

    aUniversity]At:03:3920June20

    08

    114 T H E J O U R N A L O F D E V E L O P M E N T S T U D I E SMichaely, M., 1977, 'Exports and Growth: An Empirical Investigation', Journal of DevelopmentEconomics, Vol.4, No .1, pp.4 9-53 .Rahman, M.A., 1993, Export and Economic Development of Bangladesh, Dhaka: YoungEconomists' Association.Ram, R., 1985, 'Exports and Economic Growth: Some Additional Evidence', EconomicDevelopment and Cultural Change, Vol.33, No .2, pp.415 -25.Ram, R., 1987, 'Exports and Economic growth in Developing Countries: Evidence from TimeSeries and Cross-Sectional data', Economic Development and Cultural Change, Vol.36,No.1, pp.51-72.Rodrik, D., 1992, The Limits of Trade Policy Reform in Developing Countries, Journal ofEconomic Perspectives, Vol.6, N o.1, pp.87-105 .Romer, P., 1989, 'Capital Accumulation in the Theory of Long Run Growth', in R. Barro, (ed.),Modem Business Cycle Theory, Cam bridge, M A: Harvard University P ress.Roy, D.K., (1991), 'Determinants of Export Performance of Bangladesh', The BangladeshDevelopment Studies, Vol.19, No.4, p p.2 7-4 8.Tyler, W. G., 1981, 'Growth and Export Expansion in Developing Countries: Some EmpiricalEvidence', Journal of Development Economics, Vol.9, N o.l , p p.121-30.World Bank, World Tables, 198 4-85 , 1987 and 1 988 -89, Oxford: Oxford University Press.

    A P P E N D I XTABLE Al

    VARIABLE DEFINITIONS AND THE DATAVariable nam e DefinitionGD P (Y) Gross domestic product in Taka at 1984-85 constant pricesLabou r (L) Population in the mid-year is used as a proxy for labourInvestment (I) I is the sum of private and public investment, mea sured at 1984 -85

    constant price and exp ressed in TakaExport (X) Export includes receipts on account of mercha ndise (f.o.b.) and non-factorservices. The former comprises the market value of goods including non-monetary gold and the latter includes shipment, passenger and othertransport services and travel, as well as current account transactions notseparately reported. X is measured at 1984-85 constant price andexpressed in Taka.Sources:(i) Alamgir and Berlage (1972) for the period 1960-61 to 1968-69(ii) World Bank (various issues of World Tables) for the period 1969-70 to 1971-72(iii) BBS (1993) for the period 19 72-73 to 1991-92