forman journal of economic studies vol 8

Forman Christian College (A Chartered University) was founded in 1864 by

Dr. Charles W. Forman, a Presbyterian missionary from USA. In 1972 the

college was nationalized by the government of Pakistan and it was returned to

the present owners of the college on March 19, 2003. In March 2004, the

government of Pakistan granted university status to Forman Christian

College.

For submission of articles for publication and purchase of Forman Journal of

Economic Studies:

Contact Editor Forman Journal of Economic Studies Department of Economics Forman Christian College (A Chartered University) Ferozepur Road, Lahore-54600, Pakistan E mail: [email protected] Ph: +92 42 99231581-8, Ext: 380 Fax: +92 42 99230703 www.fccollege.edu.pk Subscription Rate Inland

Students Rs.200 General Rs.300

Overseas US $ 40 ISSN: 1990-391X Abbreviated Key Title: Forman j. econ. stud. Recognized by: Higher Education Commission, Category-Y Internationally Indexed by: EconLit, EBSCOhostTM, IBSS & Ulrich’s Journal Website: www.fccollege.edu.pk/academics/departments/academic- departments/department-of-economics/research

i

Forman Journal of Economic Studies Patron

James A. Tebbe Editor Associate Editors Managing Editor Muhammad Aslam Chaudhary Tanvir Ahmed Ghulam Shabbir Muhammad Akbar National Advisory Board Asad Zaman International Islamic University, Islamabad Eatzaz Ahmad Quaid-i-Azam University, Islamabad Fazal Hussain PIDE, Islamabad Imran Sharif Chaudhry Bahauddin Zakariya University, Multan Khair-uz-Zaman Gomal University D. I. Khan Michael Murphy Forman Christian College University, Lahore Muhammad Aslam Lahore School of Economics, Lahore Muhammad Idrees Quaid-i-Azam University, Islamabad Mumtaz Anwar Ch. University of the Punjab, Lahore Mushtaq Ahmed Lahore University of Management Sciences, Lahore Naveed Ahmed Institute of Business Administration, Karachi Razaque H. Bhatti International Islamic University, Islamabad Shah Nawaz Malik Bahauddin Zakariya University, Multan International Advisory Board Chang Yee KWAN National University of Singapore, Singapore David Graham Institute of Defense Analysis, Alexandria, VA, USA Ismail Cole University of California, PA, USA James Fackler University of Kentucky, USA Kiyoshi Abe Hanazono, Hanamigaaku, Chiba City, Japan M. Arshad Chaudhary University of California, PA, USA Mack Ott Gravitas International LLC, USA Marwan M. El Nasser Fredonia University, USA Muhammad Ahsan Academic Research Consultant / Adviser, UK Nasim S. Sherazi Islamic Development Bank, Saudi Arabia Roger Kormendi University of Michigan, USA Sarkar Amin Uddin Fredonia University, USA Soma Ghosh Albright College, USA Stephen Ferris Carleton University, Canada Steve Margolis North Carolina State University, USA Suchandra Basu Rhode Island College, USA Toseef Azid Taibah University Madinah, Saudi Arabia Thomas Zorn University of Nebraska, USA

ii

Declaration

The findings, interpretations and conclusions expressed in this journal are

entirely those of the authors and should not be attributed in any manner to the

FCC or Editorial Board. The journal does not guarantee accuracy of the data

included in this publication and accepts no responsibility for any consequence

of their use.

iii

FORMAN JOURNAL OF ECONOMIC STUDIES

Volume: 8 2012 January-December An Analysis of International Income Inequality 1

Muhammad Idrees and Eatzaz Ahmad

Impact of Natural Disasters, Terrorism and Political News on 13 KSE-100 Index

Mian Ahmad Hanan, Saleem Noshina, Saqib Ali Siddiqui and Shahid Imran

Performance of Alternative Price Forecast for Pakistan 31 Yaser Javed and

Eatzaz Ahmad Impact of Trade Openness on Exports Growth, 63 Imports Growth and Trade Balance of Pakistan

M. Aslam Chaudhary and Baber Amin

Determinants of Youth Activities in Pakistan 83 Rizwan Ahmad and

Ijaz Hussain A Study of Implicit Tax in Pakistan’s Agriculture, with 107 Special Reference to the Case of Rice

Mohammad Aslam Determinants of Residential Electricity Expenditure in Pakistan: 127 Urban-Rural Comparison

Ijaz Hussain and Muhammad Asad

DEPARTMENT OF ECONOMICS FORMAN CHRISTIAN COLLEGE (A CHARTERED UNIVERSITY)

FEROZEPUR ROAD, LAHORE, PAKISTAN

Forman Journal of Economic Studies Vol. 8, 2012 (January–December) pp. 1-11

An Analysis of International Income Inequality Muhammad Idrees and Eatzaz Ahmad1

Abstract

This study measures and decomposes world income inequality between world’s geographic regions during the past two decades using Theil’s two measures of inequality. The study finds that the extent of income inequality has been decreasing over the years mainly because of increasing per capita income in China and to some extent India. Income inequality has been highest, but declining sharply over time in East Asia & Pacific. Sub-Saharan Africa and Middle East & North Africa show moderate income inequalities, while other regions of the world show low inequality. The study finds that the contribution of income inequality between regions has been substantially larger than the contribution of inequality within regions.

Keywords: Income inequality; Geographical blocks; Theil’s Entropy

JEL classification: D63, O18, P25

1. Introduction With the passage of time and the world’s economies emerging as a

global village, the issue of world income distribution has gained importance. A small number of countries are very rich, accounting for a significant proportion of the world GDP. According to World Development Indicators (WDI) data, based on the country-specific per capita incomes, the richest 20% of the world population (those living in the richest countries) are found to account for 80% of world income, while the share of poorest 20% of the population has remained less than 2% of the world income. Table 1 indicates that the income share of the middle 60% of world population has varied between 13.78 and 26.6, signifying a skewed distribution with long tail of poverty.

Quite a few studies have analysed income inequality across countries. As noted in Heshmati (2004), the earlier work can be divided into two categories. The first approach is to measure international inequality as economic disparities between countries considering per capita GDP of each 1 The authors are Assistant Professor and Professor/Dean at School of Economics, Quaid-i-Azam University, Islamabad, respectively.

Idrees and Ahmad

2

country as the unit of analysis [e.g. Andic and Peacock (1961), Rati R. (1979), Berry et al. (1983), Chotikapanich (1997), Deininger and Squire (1996), Park (2001), Podder (1993), Schultz (1998), Sala-i-Martin (2002), Firebaugh and Goesling (2004) and Theil (1979, 1996), Theil and Seale (1994)]. This approach is simple but ignores inequality within each country. The second approach, which is to measure global inequality as economic disparity between all individuals in the world, uses household income as the unit of analysis by utilizing national level surveys [e.g. Milanovic (2002, 2005, 2010)]. But practical application of this approach is limited, as national surveys of all countries in a given period are not easily available and the household income measurement practices can vary considerably across countries.

Table 1: Quintiles of Countries based on Per Capita GDP

Year Share of Top 20% Share of Bottom 20%

1960 84.24 1.29 1970 72.88 0.52 1980 76.08 0.67 1990 84.49 0.94 2000 85.20 1.02 2010 80.63 0.61

Note: Calculations are based on WDI. The present study provides time series (1990 to 2010) of international income inequality measures across countries of the world. In addition it also decomposes international income inequality between different geographical blocks of the world in order to observe the relative extent of inequality within and between various regions.

2. Analytical Framework and Data The study carries out two tasks; a) measurement of international income inequality over the period 1990 to 2010 based on per capita GDP at PPP adjusted constant 2000 US$ and b) decomposition of the international inequality into geographical blocks of the world.

Of all the measures of inequality, Gini index and Theil’s Entropy measures have attracted much attention in empirical literature because of their relative agreement with theoretically desirable properties of an inequality measure. Since the decomposition of Gini index between regions would

Analysis of International Income Inequality

3

include, besides between and within components, a term called trans-variation which has no straightforward interpretation [Dangum (1997)]; the study employs Theil’s two well-known measures that are neatly decomposable. Denoting per capita income of country i, per capita world income and the number of countries by iY , Y and n respectively, Theil’s measures are given by:

Theil’s First Measure: ∑=

⎟⎠⎞

⎜⎝⎛

⎟⎠⎞

⎜⎝⎛=

n

i YYi

YYi

nT

11

ln1 (1)

Theil’s Second Measure: ∑= ⎥

⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛=

n

i iYY

nT

12

ln1 (2)

In case of perfect equality both 1T and 2T take the values equal to zero, while in case of perfect inequality 1T and 2T take the values equal to )log( n and

nn)/1log( respectively. According to Shorrocks (1980), the two measures are decomposable as

follows.

4342143421

BW T

K

k

kk

T

K

k

kk Y

YsTsT ∑∑

== ⎟⎟⎠

⎞⎜⎜⎝

⎛+=

1111

ln. (3)

44 344 2143421

BW T

K

k Kk

T

K

k

kk Y

YpTpT ∑∑== ⎟

⎟⎠

⎞⎜⎜⎝

⎛+=

1122

ln. (4)

Where ks and , kp are respectively the income and population shares of group k, used as weights. The first term in each case measures weighted inequality within the k sub-groups and the second term explains inequality between the sub-groups. In order to calculate between groups inequality, Theil measures set mean income of each country within each group equal the respective group mean. The income inequality within groups is measured as the weighted sum of inequalities within various groups.

Income inequality between countries can be based on GDP as the unit of analysis, but in this case all countries are given equal weight irrespective of their populations. Per capita GDP is obviously a better unit of analysis, provided in the income inequality measures the income units (countries) are

Idrees and Ahmad

4

appropriately weighted by population or income shares as the case may be. Furthermore, all per capita incomes have to be converted to one currency (e.g. US$) and adjusted for PPP. The data for the present study are taken from latest issue of World Development Indicators ((WDI)-2012, an annual publication of World Bank. Since data for many countries are missing for the earlier years, the present study covers the period 1990 to 2010 for 170 countries.

3. Trends in International Inequality The results of international income inequality presented in Figure 1 show a declining trend throughout the period of analysis. This indicates that divergence between income and population share of different countries has been decreasing with the passage of time. Furthermore, as expected, Theil’s second measure that uses income shares rather than population shares as weights shows greater degree of inequality and faster rate of decrease over the years2.

Figure 1: Trends in International Income Inequalities

(….. Theil’s First Measure, ____ Theil’s Second Measure)

0.19

0.21

0.23

0.25

0.27

0.29

0.31

0.33

0.35

0.37

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

Our results are in harmony with those preseted in Firebaugh and Goesling (2004) that if countries are weighted according to population, the

2 The estimates of Theil’s indices without PPP adjustment of per capita GDP show no substantial difference in trend.


5

international inequality shows declining trend. Chotikapanich et.al (2009) also found that international income inequality declined between 1993 and 2000. They emphasized that decline in inequality was largely attributable to economic growth in China. Similar results were found in Sutcliffe (2004), Warner et al. (2011) for China and Wolf (2004) for China and India.

China accounts for more than 21% of the world population and had annual compound GDP growth rate at 9.5% during 1990-2000. India accounts for more than 18% of the world population and had annual GDP growth rate at 4.7% during the same period. In comparison to these two countries the growth rate of rest of the world had been just 2.13%. Therefore, it is worthwhile to determine how the trends in inequality are affected if one or both of these countries are excluded from the sample. The results are shown in figures 2 to 4.

As suspected the trends in international income inequality are almost reversed when China is excluded from the sample (Figure 2). Both the indices show an increasing trend in income inequality till 2000 and a mild decreasing trend thereafter. Thus, the substantial decrease in income inequality over the years may be attributed to inclusion of China, a relatively poor but fast growing country, in the sample. On the other hand, exclusion of India (Figure 3) has no substantial effect on the general trends of income inequality across countries, though the rate of decline in inequality is somewhat dampened.

Figure 2: Trends in International Income Inequalities (Excluding China) (….. Theil’s First Measure, ____ Theil’s Second Measure)

0.19

0.21

0.23

0.25

0.27

0.29

0.31

0.33

0.35

0.37

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

Idrees and Ahmad

6

Figure 3: Trends in International Income Inequalities (Excluding India) (….. Theil’s First Measure, ____ Theil’s Second Measure)

0.19

0.21

0.23

0.25

0.27

0.29

0.31

0.33

0.35

0.37

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

Figure 4: Trends in International Income Inequalities

(Excluding China and India) (….. Theil’s First Measure, ____ Theil’s Second Measure)

0.19

0.21

0.23

0.25

0.27

0.29

0.31

0.33

0.35

0.37

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

This is so because the growth rate in India has not been as phenomenal as in China. Both China and India account for close to one fifth of the world’s population, both have per capita income less than the world average and the GDP growth rate of China has been about twice as fast as that of India, which in turn has been more than twice as fast as the growth rate of the rest of the world. This explains why, as indicated in Figure 4, the presence of China and


7

India in the sample have suppressed the income inequality across countries in recent years to a reasonable extent.

4. Decomposition of International Income Inequality The decomposition is carried out with respect to the Wold Bank’s classification of countries in seven geographic regions. The decomposition statistics for all the years under consideration, not presented here, indicate smooth trends over the years with no sudden jumps. Therefore in order to preserve space, the statistics are presented with the gap of five years, that is, for the years 1990, 1995, 2000, 2005 and 2010. The decomposition results are reported in table 2.

The first two blocks of the table provide some indication of regional income disparity. For example, as of the year 2010 the share of North America in world income has been more than four times its share in world population while the income share of Sub-Saharan Africa has been only one fifth of its population share. This means that per capita income in North America has been about 20 times the per capita income in Sub-Saharan Africa. On the upper side of income distribution North America is followed by Europe & Central Asia with per capita income about half of the former. On the lower tail, Sub-Saharan Africa is closely followed by South Asia. The income share of Latin America & Caribbean has been slightly higher than its population share whereas the income share of East Asia & Pacific has been somewhat lower than its population share.

The next two blocks show income disparity between countries within each region. Both the measures show that the degree of inequality was highest in East Asia & Pacific, which declined drastically after every five years. Sub-Saharan Africa and Middle East & North Africa show moderate income inequalities, which remained quite stable over the years. The level of inequality in the other four regions has been quite low and stable.

Coming now to the last two blocks, it is noticeable that Theil’s second measure produces a larger contribution of inequality within regions than Theil’s first measure. The reason is that Theils second measure assigns largest weight (based on population) to the region of East Asia & Pacific where the income inequality is the highest whereas Theil’s first measure assigns largest weight (based on income) to the region of Europe & Central Asia and North America where the income inequality is relatively quite low. In any case the results show that the contribution of income inequality between regions has

Idrees and Ahmad

8

Table 2: Decomposition of World’s Per Capita Income Inequalities by Geographic Regions

Inequality statistics

Regions 1990 1995 2000 2005 2010

Percentage incom

e shares

East Asia & Pacific 18.78 21.94 22.18 24.12 27.91 South Asia 3.87 4.44 4.83 5.57 6.97 Europe & Central Asia 37.67 32.87 31.61 30.35 28.21 Sub-Saharan Africa 2.29 2.19 2.18 2.34 2.63 Latin America & Caribbean 8.59 9.14 8.88 8.50 8.85 North America 24.71 25.06 25.90 24.49 21.95 Middle East & North Africa 4.09 4.36 4.42 4.61 3.48

Percentage population shares


Theil’s first m

easure


Theil’s second m

easure


Decom

position of Theil's 1st m

easure (%)

Contribution of Inequality within East Asia & Pacific 23.79 22.09 18.63 16.95 15.06 Contribution of Inequality within South Asia 0.09 0.11 0.12 0.17 0.33 Contribution of Inequality within Europe & Central Asia 6.67 10.21 10.12 8.39 7.88 Contribution of Inequality within Sub-Saharan Africa 1.40 1.43 1.43 1.72 2.21 Contribution of Inequality within Latin America & Caribbean 0.53 0.61 0.75 0.76 0.89 Contribution of Inequality within North America 0.04 0.06 0.06 0.06 0.06 Contribution of Inequality within Middle East & North Africa 2.02 2.19 2.19 2.36 2.16 Total Contribution of Inequality within All Regions 34.54 36.71 33.29 30.40 28.59 Contribution of Inequality Between All Regions 65.46 63.29 66.71 69.60 71.41

Decom

position of Theil's 2nd m

easure (%)

Contribution of Inequality within East Asia & Pacific 30.55 24.51 19.91 16.39 12.88 Contribution of Inequality within South Asia 0.48 0.58 0.59 0.78 1.27 Contribution of Inequality within Europe & Central Asia 3.07 5.79 5.85 4.76 4.53 Contribution of Inequality within Sub-Saharan Africa 4.16 5.21 5.87 7.07 8.70 Contribution of Inequality within Latin America & Caribbean 0.49 0.69 0.85 0.95 1.13 Contribution of Inequality within North America 0.01 0.01 0.00 0.01 0.01 Contribution of Inequality within Middle East & North Africa 1.45 1.69 1.76 1.98 1.98 Total Contribution of Inequality within All Regions 40.21 38.48 34.84 31.95 30.51 Contribution of Inequality Between All Regions 59.79 61.52 65.15 68.05 69.49


9

been substantially larger than the contribution of inequality within regions. Furthermore, the degree of inequality between regions has been increasing almost steadily over the years. Another notable observation is that, as expected, income inequality within East Asia & pacific has been the main contributor of total inequality within regions.

5. Conclusions The study arrives at several interesting conclusions. It is shows that the degree of inequality in income between countries has been decreasing steadily over the years. However, this trend in income distribution does not mean that economic conditions in poor countries are improving in most of the countries. Far from it, if China alone is taken out of the picture, the trend is almost reversed, showing slight deterioration in the 1990s and mild improvement in the 2000s. Furthermore, if both China and India are taken out of the picture, the trend would show a no net improvement in income over the past two decades despite some improvement in recent years. Nevertheless, this need not be viewed pessimistically. After all China and India account for about 40% of world’s population and, therefore, improvement of economic conditions even in these two countries alone cannot be taken lightly. When it comes to standards of living, what matters is the proportion of world population, rather than the number of countries than show improvement.

Another useful finding of the study is that international inequality mainly comprises of inequality between geographic regions while the contribution of inequality within regions has been relatively small. Further, the contribution of inequality between regions has increased consistently. This pattern has serious implications for the way world economic cooperation contributes to reducing inter-country economic disparity through free trade and movement of production activities. Most formal efforts towards economic cooperation are confined to regional economic cooperation in the form of free-trade area, etc. Where there has been any cross regions (or cross continental) economic cooperation, the effects on reduced disparity are visible. This can be seen in the form of significant improvements in economic conditions in China and India. While both countries have benefitted from free trade and transfer of production facilities, China has reaped the maximum gains. In any case, the study clearly indicates that there is substantial scope foe bridging the gap between rich and poor countries of the world and highlights the importance of international economic cooperation beyond regional boundaries.

Idrees and Ahmad

10

References Andic, S., & Peacock, A.T. (1961). The international distribution of income,

1949 and 1957. Journal of the Royal Statistical Society, 124, 206-218.

Berry, A., Bourguignon, & Morrisson, C. (1983). Changes in world distribution of income between 1950 and 1977. Economic Journal, 93, 331-350.

Chotikapanich, D., Griffiths, W., Rao, P., & Vicar, V. (2009). Global income distribution and inequality: 1993 and 2000. Department of Economics, Working Papers Series 1062, The University of Melbourne.

Chotikapanich, D., Valenzuela, R., & Rao, D. S. (1997). Global and regional inequality in the distribution of income: Estimation with limited and incomplete data. Empirical Economics, 22, 533-46.

Dangum, (1997). A new approach to the decomposition of Gini income inequality ratio. Empirical Economics, 22, 515-531.

Deininger, & Squire, (1996). A new data set measuring income inequality. World Bank Economic Review, 10, 565-591.

Firebaugh, G., & Goesling, B. (2004). Accounting for the recent decline in global income inequality. American Journal of Sociology, 110, 283-312.

Fredrik, & Christian, (2005). Estimates of trends in global income and resource inequalities. Ecological Economics, 55, 351-364.

Heshmati, (2004). The world distribution of income and income inequality. Discussion Paper: 1267, IZA.

Milanovic, B. (2002). True world income distribution, 1988 and 1993: First calculation based on household surveys alone. Economic Journal, 112, 51-92.

Milanovic, B. (2005). Can we discern the effect of globalization on income distribution? Evidence from household surveys. World Bank Economic Review, 19, 21-44.

Milanovic, B. (2010). Global inequality recalculated and updated: The effect of new PPP estimates on global inequality and 2005 estimates. Journal of Economic Inequality, 10, 1-18.


11

Park, (2001). Recent trends in the global distribution of income. Journal of Political Modelling, 23, 497-501.

Podder, N. (1993). A profile of international inequality. Journal of Income Distribution, 3, 300-314.

Ram, R. (1979). International income inequality: 1970 and 1978. Economic Letters, 4, 187-90.

Sala-i-Martin, (2002). The distribution “Rise” of global income inequality, NBER working paper, 8933.

Schultz, (1998). Inequality in distribution of personal income in world: How it is changing and why? Journal of Population Economics, 11, 307-344.

Shorrocks, (1980). The class of additively decomposable inequality measures. Econometrica, 48, 613-625.

Sutcliffe, (2004). World inequality and globalization. Oxford Review of Economic Policy, 20(1) 15-37.

Theil, H., & Seale, J. (1994). The geographic distribution of world income, 1950-90. De Economist, 4, 387-419.

Theil, H. (1979). World income inequality. Economics Letters, 2, 99-102.

Theil, H. (1996). Studies in global econometrics. Kluwer Academic Publishers, Amsterdam.

Warner, D., Rao, P., Griffiths, W. E., & Chotikapanich, D. (2011). Global inequality: Levels and trends, 1993-2005. Discussion Papers Series, 436, School of Economics, University of Queensland, Australia.

Wolf, M. (2004). States are cure and disease. World Economy, Special Report.

World Development Indicators (2012). http://data.worldbank.org/data-catalog/world-development-indicators, World Bank.


IImmppaacctt ooff NNaattuurraall DDiissaasstteerrss,, TTeerrrroorriissmm aanndd PPoolliittiiccaall NNeewwss oonn KKSSEE--110000 IInnddeexx

MMiiaann AAhhmmaadd HHaannaann,, SSaalleeeemm NNoosshhiinnaa,,

SSaaqqiibb AAllii SSiiddddiiqquuii aanndd SShhaahhiidd IImmrraann11

Abstract

This paper discusses the impact of news related to natural disaster, terrorism and political on the Karachi Stock Exchange (KSE-100) index. This study is based on Event Study Methodology. It focuses on 21 different news events:10 political, 9 terrorism and 2 natural disasters. It attempts to determine the statistical relationship and effect of these events on the KSE -100 index by using an 11 day stock market window. This paper concludes that news events have a strong impact on the KSE-100 index and among political, terrorism and natural disaster events; news related to terrorism has the most profound influence on trend of KSE-100 index. It also shows that the bigger the news event the greater is the impact on KSE-100 index. It also supports the notion that good news has a positive and bad news has a negative impact on the KSE-100 index. This study also reveals that the Karachi Stock Exchange is “informationally efficient.”

Keywords: Good and bad news; Political, Terrorism and natural disasters news; KSE-100 index

JEL classification: G14, H80, H84

1. Introduction

Stock exchange is one of the key financial markets of every country and considered as one of the imperative sources for the corporations to raise capital. It has also made it easier for the investors including individual and

1 The authors are Professor/Chairperson at DDeeppaarrttmmeenntt ooff MMaassss CCoommmmuunniiccaattiioonn,, FFoorrmmaann CChhrriissttiiaann CCoolllleeggee ((AA CChhaarrtteerreedd UUnniivveerrssiittyy)), Lahore; Assistant Professor at IInnssttiittuuttee ooff CCoommmmuunniiccaattiioonn SSttuuddiieess,, UUnniivveerrssiittyy ooff tthhee PPuunnjjaabb,, LLaahhoorree aanndd RReesseeaarrcchh AAssssiissttaannttss,, rreessppeeccttiivveellyy.. Corresponding Author’s Email: [email protected]

Hanan, Noshina, Siddiqui and Imran

14

institutions to invest in the highly liquid securities traded in the stock exchange in contrast to invest in less liquid investments like real estate. Due to its high liquidity characteristics, it is easy for investors to buy and sell their securities quickly. It has become the key indicator of economic expansion and strength of a country. It is widely believed that the performance of stock market largely depends upon the arrival of information and news related to different events. Chen and Siems (2004) describe the effect on news in these words. “…global capital markets today are tightly intertwined; news spread rapidly (especially bad news), with quick spill over or contagion effects” (p. 363). Furthermore, Goonatilake and Herath (2007) investigated the impact of news on DJIA, NASDAQ and S&P 500 and found the association between the news and stock market fluctuations. In addition, the efficient-market hypothesis (EMH) asserts that financial markets are "informationally efficient". There are three main versions of this hypothesis: "weak", "semi-strong", and "strong". Weak EMH maintains that prices on traded assets (e.g., stocks, bonds, or property) already reflect all past publicly available information. Semi-strong EMH claims both that prices reflect all publicly available information and that prices instantly change to reflect new public information. Strong EMH argue that prices instantly reflect even hidden or "insider" information. There is evidence for and against the weak and semi-strong EMHs, while there is powerful evidence against strong EMH. Oncu and Aktas (2006) maintain:

An implication of EMH is that market prices reflect all available information and expectations, and that any new information is properly incorporated into prices without any delay. A stock market’s speediness to incorporate new information into prices is referred to informational efficiency. Market’s ability to reflect new information properly is referred to market rationality (p. 233).

Therefore, the news of different types of events happening in the country and around the world have a major impact on the performance of stock market no matter it is directly linked to it or not. If the country’s economic, political and law and order situation is unsatisfactory, it will also have a deteriorating impact on the performance of stock market and vice versa. Therefore, stock exchange is a key indicator of country progress and economic development. News regarding political events has both constructive and deteriorating impact on the stock market depending on its nature and consequences. When the political condition of the country is stable, then it


15

will have a positive impact on the stock market performance and attract more investors to invest their capital and vice versa. Sandler and Enders (2002) define terrorism as “the premeditated use, or threat of use, of extra normal violence to obtain a political objective through intimidation or fear directed at a large audience.” Barth et al (2006) conducted a research report on “Impacts of Economic Terrorism: From Munich to Bali” and stated that approximately 20,000 terrorist activities and incident had been occurred in the world during the past 35 years and concluded that “terrorism is associated with adverse effects on overall economic activity” of a country (p. 3).

The basic objective of this paper is to examine the impact of news related to politics, terrorism and natural disasters on the KSE-100 index. This study is based on Event Study Methodology. It focuses on 21 different news events including 10 political, 9 terrorism and 2 natural disasters and determine the statistical relationship and effect of these events on the KSE -100 index by analyzing the 11 days stock market trend. This paper concludes that news has the strong impact on the KSE-100 index and among political, terrorism and natural disaster news; news related to terrorism has profound influence on trend of KSE-100 index. It also proves that bigger the news in terms of its consequence, the more will be the impact on KSE-100 index. Finally, it also claims that (1) Good news has an optimistic impact on the KSE-100 index; (2) Bad news has a pessimistic impact on the KSE-100 index. In the end, this study also supports the EMH that KSE is nearly efficient and security prices fully reflect the available information and news.

2. Literature Review

2.1. Political News and its Impact on Stock Exchange

The political news always has a great impact on country’s economic, social and political activities. Khan et al (2009) conducted the study about the impact of news related to Pak-U.S. relations on the KSE-100 index by applying the event methodology and concluded that the association between political news and the KSE-100 index was highly significant. Moreover, Fornari et al (2002) examined the impact of schedule and unscheduled news on Italian financial market between the year 1994 and 1996 and found that unscheduled news produced more volatility in the Italian financial market than the schedule news (p. 611). Chan and Wei (1996) revealed that the favorable and unfavorable news were correlated with the positive and negative returns


16

of the Hang Seng Index. In the same way, Chan, Chui and Kwok (2001) analyzed the impact on political news on Hang Seng Index and concluded that impact of political news was more than the economic news. Zach (2003) investigated the impact of political events on one of the major Israel Stock Exchange namely Tel Aviv Stock Exchange Index from 1993 to 1997 by applying different statistical techniques on the sample including statistical test- location, statistical test- spread, regression analysis, cross-sectional analysis, and News-intensive periods and found that political events had more significant impact on the returns instead on the days when there were no significant political events occurred (p. 243). Niederhoffer (1971) found that the world events exert a discernible influence on the movement of the S&P 500. Returns following the world events tend to be larger in absolute value than returns on other days (p. 193). Franck and Krausz (2009) analyzed the impact of institutional reforms, political events and wars on the Israel Stock Market between 1945 and 1960. They found that likelihood of war did not have any effect on the stock market but it would increase the risk at the time of skirmishes. Furthermore, domestic political instability also increased the stock market risks (p. 141). In the Pakistan’s context, Khan and Ahmed (2009) examined the relationship between events happening from December 2007 to October 2008 and their impact on aggregate stock market trading volume and daily stock returns and maintained that political events noticeably fluctuate the stocks returns and trading volume of KSE-100 index (p. 373). Chan and Wei (1996), and Kim and Mei (2001) concluded that political news produced more stock volatility in the Hong Kong. Niederhoffer et al (1970), Peel and Pope (1983) and Gemmill (1992) found that stock market prices were significantly affected by the elections both government and congressional in different developed countries. They found changing in governmental administration caused by elections tends to effect financial policies or legislation, thereby significantly affecting stock prices (Chen, et al, 2005, p. 167). On the other hand, Aktas and Oncu (2006) tested the Efficient Market Hypothesis with the case study of the refusal of controversial bill allowing deployment of U.S troops in Turkey by Turkish Parliament and its impact on the Turkish Stock Exchange performance and maintained that in response to unfavorable political events, stock prices are expected to behave differently in the efficient market since the new information will have different economic impact on individuals firms (p. 78).

Chuang and Wang (2009) examined the impact of major political changes related to U.S, U.K, France and Japan from the year 1979 to 2001 on


17

the stock exchange indexes including Nikkei 225, SBF-250, FTSE 30 and Dow Jones 30. They concluded that political changes negatively affected all the countries indexes with the significance level of 5% and suggested that political changes basically provided the opportunity for progress according to the democracy but they had an inverse relationship with the stock return in the developed countries (p. 948). Kaminsky, and Schmukler (1999) argued that “in the chaotic financial environment of Asia in 1997–1998, daily changes in stock prices of about 10 percent became commonplace. They found that market movements were triggered by local and neighboring-country news, with news about agreements with international organizations and credit rating agencies having the most weight. In addition, the evidence suggests that investors over-react to bad news” (p. 537). In the context of political election, Pantzalis et al (2000) had used UIH to investigate the behavior of the stock market indexes across 33 countries around the election days during the sample period of 1974-1995 and found a positive impact of these elections on stock market indexes resulting positive abnormal returns (p. 1575). Jones and Banning (2009) analyzed the impact of U.S elections on the stock exchange performance and found little relationship between both of them. They concluded that elections and election cycle didn’t play much role in forecasting stock market returns (p.273). Similarly, Goriaev and Zabotkin (2006) study regarding the Russian Stock markets supported the argument that economic and political events always influence the Stock Market trends (p. 380).

2.2. Studies related to Terrorism Impact on Stock Exchange

Several studies support the argument that the terrorism incidents always influence the stock exchange performance. Chen and Siems (2004) applied the event methodology to measure the impact of 9/11 on global and U.S share prices as well as to compare it with political, economic and natural disasters impact and found that the 9/11 had less effect on financial markets as compare to former terrorist events (p.349). Elder and Melnick (2004) analyzed the impact of Palestinian terrorist attacks on the Israel’s Stock and Foreign Exchange markets by using daily times series data from 1990 till 2003 and establish the argument that the attacks had a permanent effect on both Stock and Foreign Exchange markets but location of terrorist attacks had no effect in either of the markets (p. 367; also cited in Sathye et al 2008, p. 6). In an other study, Eldor and Melnick (2004) investigated the impact of terrorism on stock and foreign exchange markets in Israel by using the data of 639 terrorist


18

attacks classified into location, type, target, and causalities from the year 1990 to 2003 and argued that terrorist events had a permanent pessimistic impact only on the stock markets. Johnston and Nedelescu (2006) revealed that International Monetary Fund Report (2001b) regarding impact of September 11 attacks on the Standard and Poor’s 500 and NASDAQ indexes maintained that the both indexes dropped to 11.6 percent and 16.1 percent respectively between 17th and 21st September. In addition, the 9/11 attack had a significant impact on the world major financial markets causing sharp declines and fall. European Stock Market also had a severe decline after 17th September. Correspondingly, Dow Jones STOXX index had dropped by 17.3 percent during 11th and 21st September after the 9/11 terrorist attack (p. 12). Arin, Ciferri and Spagnolo (2008) investigate impact of terrorists’ events on the behavior of six financial markets including Indonesia, Israel, Spain, Thailand, Turkey and U.K and concluded that the effects of terrorism not only on the stock market, but also on the stock market volatility. In addition they found that the magnitude of terrorist effects is larger in emerging markets (p. 164).

Barros et al (2009) studied the impact of Basque terrorism on the Basque stock market. They took the data from July 2001 to 15th November 2005. They investigated the level of violence so called Kale Borroka (street fighting) in this area, the police action and repressive policy measures by the government against this violence and impact of this violence on the Stock Market. They had found that they could reduce the violence by banning the radical party Herri Batasuna and the occurrence of violence had a negative effect on the stock market index. Ahmed and Farooq (2008) investigated the impact of 9/11 terrorist attacks on the volatility of KSE-100 index and found that this incident had a significant impact on the KSE. Their study also negated that volatility was not due to the implementation of regulatory reforms by the SECP as they had found the same qualitative results by dividing the 9/11 period into pre and post reforms period (p. 71). Furthermore, Chen and Wei (2005) investigated U.S capital market response to 7 major terrorist and 7 military attacks from 1915-2001 through event methodology. They applied their analysis to some other capital markets as well, but focus on the impact of only two events: the 9/11 terrorist attacks and Iraq’s invasion of Kuwait in 1990. They found that the U.S. capital markets rebound and stabilized quicker after these two events compared to other markets, and U.S. markets are more resilient now than in the past, which they explain by the strength of the banking and the financial sectors in the U.S. One of the main conclusions of their study is that the U.S. financial markets are efficient in


19

absorbing the shocks caused by the terrorist attacks and can continue to function in an effective way (p. 399).

2.3. Natural Disasters and Stock Exchange

To explain the impact of natural disaster incidents on stock markets, Worthington and Valadkhani (2004) investigated the effect of natural disasters on the Australian equity market by studying the impact of 42 natural disasters including storms, floods, earthquakes, cyclones etc on the All Ordinances Index from the December 31, 1982 to January 1, 2002 by applying ARMA model and concluded that bushfires, cyclones’, earthquakes had a major impact on market returns as compared to severe storms and floods (p. 2177). In another study, Worthington and Valadkhani (2005) compared the effects of natural, industrial and terrorist disasters on the Australian Capital Markets by applying Box and Tiao Intervention analysis on 10 market sectors: consumer discretionary, consumer staples, energy, financial, health care, industrial information technology, materials, telecommunication services and utilities and found that “the sectors most sensitive to disaster of any type are the consumer discretionary, financial services and material sectors. The most significant single event during the past eight years would appear to be the September 11 terrorist attacks” (p. 331). Shelor et al (1990) investigated the impact of California Earthquake on the firms’ value dealing in Real Estate industry. They found that this event had negative impact on the firm’s stock returns that were operating in the area hit by the earthquake. To sum up, various studies discuss in the literature related to different countries support the argument that news has a significant impact on the stock markets.

2.4. Conceptual Framework

The proposed conceptual model of this study is presented in Figure1. The independent variable is News and dependent variable is KSE-100 index. News is categorized into political, terrorism and natural disasters. One dimensional arrow is demonstrating that political, terrorism and natural disaster news have an impact on the KSE-100 index.

3. Methodology

Event Study Methodology is applied in this study to analyze the impact of each event on KSE-100 index. Purpose based sampling technique is


20

used for the selection of news events. Three different types of 21 news including 10 political, 9 terrorism and 2 natural disaster events (see Table 1) were identified and determined the impact of them on the KSE -100 index by analyzing the 11 days trend. The occurrences of these events have been identified from the headlines appearing in the newspapers “The News” and "the Dawn” respectively. These two newspapers are among the elite and most trusted newspapers of Pakistan and have vast circulation across the Pakistan. Statistical technique independent t-test is applied to analyze the impact of political, terrorism and natural disaster news on stock exchange after testing the equality of variances by Levene’s Test. If the variances turned out to be unequal, in independent sample t-test, where variances were assumed to be unequal had been used to compare the pre and post event mean values.

Figure 1

3.1. Event Windows

Event Windows are divided into three parts: The five days has been selected for the Pre-Event Window, one day for the Event, and five days for the Post Event (Figure 5). Therefore, the trend of total eleven days has been studied because of following reasons: (1) larger time span creates more noise in the trend that may affect the results (2) accurate results are not possible with shorter time span because impact is more dispersed. (3) The 11 days trend presents more accurate variation in KSE-100 index. In the pre and post


21

event windows, those days are not included when the stock market was closed either for a weekend or on account of public holiday.

Figure 2

3.2. Hypotheses

This study proposes following hypothesis to analyze the impact of political, terrorism and natural disaster news on KSE-100 index.

H1: News has a significant impact on the KSE 100 Index

H2: Political News has a significant impact on the KSE 100 Index.

H 3: Terrorism News has a significant impact on the KSE 100 Index.

H 4: Natural Disaster News has a significant impact on the KSE 100 Index.

H 5: International news has a significant impact on the KSE-100 index.

H 6: News related to assassination of prominent political personalities has a negative impact on KSE-100 index.

H 7: Good news has a positive impact on the KSE-100 index.

H 8: Bad news has negative impact on the KSE-100 index.


22

3.3. Research Questions

The study addresses three research questions: RQ 1: Which type of news including political, terrorism or natural disaster has more impact on the KSE 100 index? RQ 2: To what extent the bigger news in terms of its consequence has a profound impact on KSE-100 index? RQ 3: How far the news related to failed assassination attempts on prominent personalities affects the stock market trends? RQ 4: Which type of bad news, local or international, has more negative impact on KSE-100 index?

4. Findings and Results This section presents answers to the research questions as well as validating hypotheses on the basis of quantitative analysis.

4.1. Hypothesis 1 Table 1 determines that majority of news regarding politics; terrorism and natural disaster incidents have a significant impact on the KSE-100 index. For instance, political news 2, 3,4,5,6, and 9, terrorism related news 11, 13, 14,15,16,17 and 18, and natural disaster news 20 and 21 have a significant impact on KSE-100 index. This validates the hypothesis (H1) as well as supports the EMH. Therefore, these findings support that the KSE is informational efficient.

4.2. Hypothesis 2

Table 1 reveals that political news has a significant impact on the KSE-100 index. Political news namely 2, 3,4,5,6, and 9 support this hypothesis. For instance, news item 2-“17th Amendment passed” is statistically significant with respect to its impact on the KSE-100 index as t (8)= -11.455, p<0.05 and there is an upward trend in the KSE-100 index as Post Event KSE trend mean value rises sharply as compared to the mean value of Pre-Event KSE trend.

4.3. Hypothesis 3 The news related to terrorism incidents including 11, 13, 14,15,16,17 and 18 supports the hypothesis that terrorism news has significant impact on KSE-100 Index. For example, News item 11- “9/11 attacks” has a significant impact [t(8)= 4.377, p<0.05] on KSE-100 index and causes a radical change of - 94.73 points between Pre-Event and Post Event mean values. Similarly,


23

Table 1: News Impact on KSE-100 Index


24

news item 17- “Benazir Bhutto’s assassination,” and 8- “Suicide attack on Marriot Hotel Islamabad” have a negative impact on the KSE-100 index as t (8)=4.595, p< 0.05 and t (8)= 3.885, p<0.05 respectively ( see Table 1).

4.4. Hypothesis 4 The news related to natural disasters also has an impact on the KSE-100 index. News items 20-“Massive earthquake” and 21-“Flood” had an impact on the KSE-100 index as t (8) = -4.536, p<0.05 and t (8) = -4.324, p< 0.05 respectively (Table 1). However, both news incidents reflected that the KSE-100 index went up because of the two factors: (1) these events had provided an opportunity for the cement, manufacturing and other related sectors to rebuild the infrastructure again in the disaster areas; and (2) the KSE was not significantly affected because the earthquake (2005) and flood (2010) largely devastated the non-industrial areas of Pakistan.

4.5. Hypothesis 5 The news item 11 -“9/11 attacks” has a significant impact on the KSE-100 index as t (8) = 4.377, p<0.05. The September 11 terrorist attacks have brought a declining trend of (-94.73) points between the Pre-Event and Post-Event mean values. This international news likely brought fluctuations in KSE-100 index because of two reasons: (1) it destabilized U.S. Pakistan relations due to Pakistan’s support of Taliban regime at that time, and (2) U.S. declared war against terrorism altered the geo-strategic situation of the region. This shows that international news related to Pakistan has a significant impact on the KSE-100 index (Table 1).

4.6. Hypothesis 6 The news item 17- “Benazir Bhutto’s assassination” had a major impact on the KSE-100 index as t (8)=4.595, p< 0.05 that caused a decline in the Post Event KSE trend mean value to 13870.332 from 14663.622 because Benazir was a renowned political leader of Pakistan and her killing caused a law and order problem in the country. This declining trend is consistent with the belief that assassination of a prominent political personality has a negative impact on the stock market (Table 1).

4.7. Hypothesis 7 The good news including news items 2, 3, 6, 7, 8, 9, and 10 (see Table 1) have a positive impact on KSE-100 index because (1) they increased the investors’ confidence, and (2) stabilized the country politically that boosted


25

the economic activities in the Pakistan. The quantitative findings support the argument that good news has significant impact on the stock market.

4.8. Hypothesis 8 Table 1 depicts that news item 5 “Emergency rule declared by Musharraf” in Pakistan has a significant influence on the KSE-100 index as t (8) = 5.313, p<0.05. It caused a declining trend of -591.668 points difference between the Pre-Event mean value (14117.132) and Post Event mean value (13525.464). The rationale behind this decreasing trend was the result of President Musharraf‘s move to dissolve the parliament and introduction of amendments in constitution to strengthen his power in the country. Moreover, emergency rule destabilized the political and democratic institutions of Pakistan due to which investors’ confidence diminished that resulted in downwards trend. In addition, News item 11 “9/11 attacks” is statistically significant with respect to its impact on the KSE-100 index as t (8) = 4.377, p<0.05. There was a sharp declining trend of -94.728 in the KSE-100 index as the Post Event KSE trend mean value (1149.736) decreased as compared to the mean value (1244.464) of the Pre-Event KSE trend. This declining trend was the result of the emerging severe tensions between the U.S. and the Muslim countries, especially with Afghanistan and Iraq. U.S. declared the War on Terror and divided world in to friends and foe. These U.S. moves badly damaged the confidence of investors, business cycles and economies around the world. Therefore, all the international financial markets including the KSE demonstrated a sharp declining trend. Similarly, news items 17 “Benazir Bhutto’s assassination” and 18 “Suicide attack on Marriott Hotel Islamabad” had also caused a negative impact on KSE-100 Index (Table 1). Finally, news 1 “Musharraf wins presidential referendum” had no significant impact on KSE-100 index as t (8) =0.532, p>0.05. But, it has caused a decline in the KSE-100 index from 1875.546 to 1863.624 because this incident had strengthened dictatorial regime in Pakistan. Hence, all the bad news including this study support the argument that bad news has a negative impact on the KSE-100 index.

4.9. Research Question 1

On the basis of the higher mean values difference, three news items were selected from political, terrorism and natural disaster categories respectively resulted that news related terrorism has profound impact on KSE-100 Index as compare to political and natural disaster news. Among the political news, “Emergency rule declared by Musharraf” had created a


26

significant downward mean value difference of -591.666. On the other hand, the terrorism incident, “Benazir Bhutto’s assassination,” had a momentous impact on the KSE-100 index because the pre and post event mean difference was -793.29. Interestingly, “News item 20-Massive earthquake” in Pakistan had stimulated a positive trend in the KSE-100 (Table 1).

4.10. Research Question 2 Table 1 reveals that news item 5 -“Emergency rule declared by Musharraf” had caused the decline of -591.668. Similarly, news item 9 “Supreme Court acquits Nawaz from hijacking charges” had increased the KSE-100 index by difference of 144.48 from pre-event mean value 7644.108 to Post-Event Mean value of 7788.588 points. In addition, news related to terrorist incidents including news items, 11- “9/11 attacks,” and 17-“Benazir Bhutto’s assassination” have created a falling trend of -94.728 and -793.29 points respectively. These findings support the assumption that bigger news in terms of its consequence has a strong impact on KSE-100 index.

4.11. Research Question 3 Table 1 depicts that the news related to failed assassination attempts on prominent political personalities, news items, such as 12, 13, 15 and 16 have no negative impact on the KSE-100 index because KSE is an efficient market that absorb the shocks caused by these news incidents successfully. Therefore, it appears that news related unsuccessful assassination attempts on prominent personalities do not have a negative impact on KSE-100 index.

4.12. Research Question 4 The local news showed a greater negative impact on KSE-100 index than international news. For instance, a local bad news -Benazir Bhutto’s assassination (pre and post event mean value difference (-793.29) had produced more volatility in KSE -100 Index than the international news related to “9/11 attacks (-94.728) (see Table 1: news items 17 & 11).

5. Conclusion

On the whole, it is concluded that news has the strong impact on the KSE-100 index and news related to terrorism appears to have more influence on the KSE-index as compared to political and natural disaster news. This study has the following results. Firstly, political, terrorism and natural disaster news has a significant impact on the KSE 100 Index. Secondly, the bigger news in terms of its consequence has a profound impact on KSE-100 Index.


27

Thirdly, this study also validates the view that good news has a positive impact on the KSE-100 index and vice-versa. Finally, this study is consistent with the notion that the Karachi Stock Exchange is “informationally efficient.”


28

References

Ahmed, S., & Farooq, O. (2008). The Effect of 9/11 on the Stock Market Volatility Dynamics: Empirical Evidence from a Front Line State. International Research Journal of Finance and Economics, 16, 71-83.

Aktas, H., & Oncu, S. (2006). The stock market reaction to extreme events: The evidence from Turkey. International Research Journal of Finance and Economics, 6, 78-85.

Arin, K., Ciferri, D., & Spagnolo, N. (2008). The price of terror: The effects of terrorism on stock market returns and volatility. Economics Letters, 101, 164-167.

Barros, C., Caporale, G., & Gil-Alana, L. (2009). Basque terrorism: police action, political measures and the influence of violence on the stock market in the Basque Country. Defence and Peace Economics, 20, 287-301.

Barth, J., Li, T., Mccarthy, D., Phumiwasana, T., & Yago, G. (2006). Economic impacts of global terrorism: From Munich to Bali. Capital Series, 1-38.

Chan, Y., Chui, A., & Kwok, C. (2001). The impact of salient political and economic news on the trading activity. Pacific-Basin Finance Journal, 9, 195-217.

Chan, Y., & John WEI, K. (1996). Political risk and stock price volatility: The case of Hong Kong. Pacific-Basin Finance Journal, 4, 259-275.

Chen, A., & Siems, T. (2004). The effects of terrorism on global capital markets. European Journal of Political Economy, 20, 349-366.

Chen, C., & Wei, Y. (2005). Computational issues for quantile regression. Sankhy : The Indian Journal of Statistics, 399-417.

Chen, D., Bin, F., & Chen, C. (2005). The Impacts of Political Events on Foreign Institutional Investors and Stock Returns: Emerging Market Evidence from Taiwan. International Journal of Business, 10(2), 165-188.

Chuang, C., & Wang, Y. (2009). Developed stock market reaction to political change: a panel data analysis. Quality and Quantity, 43, 941-949.


29

Eldor, R., & Melnick, R. (2004). Financial markets and terrorism. European Journal of Political Economy, 20, 367-386.

Fornari, F., Monticelli, C., Pericoli, M., & Tivegna, M. (2002). The impact of news on the exchange rate of the lira and long-term interest rates. Economic Modeling, 19, 611-639.

Franck, R., & Krausz, M. (2009) Institutional changes, wars and stock market risk in an emerging economy: evidence from the Israeli stock exchange, 1945–1960. Cliometrica, 3, 141-164.

Gemmill, G. (1992). Political risk and market efficiency: tests based in British stock and options markets in the 1987 election. Journal of Banking & Finance, 16, 211-231.

Goonatilake, R., & Herath, S. (2007). The Volatility of the Stock Market and News. International Research Journal of Finance and Economics, 11-53.

Goriaev, A., & Zabotkin, A. (2006). Risks of investing in the Russian stock market: Lessons of the first decade. Emerging Markets Review, 7, 380-397.

International Monetary Fund. (2001b). World economic outlook -the global economy after 11 September: A survey by the staff of the international monetary fund, World Economic and Financial Surveys.

Jones, S. & Banning, K. (2009). US elections and monthly stock market returns. Journal of Economics and Finance, 33, 273-287.

Johnston, R., B. & Nedelesar, O. M. (2006). The Impact of Terrorism on Financial Markets. Journal of Financial Crime, 13(1), 7-25.

Kaminsky, G., & Schmukler, S. (1999). What triggers market jitters?: A chronicle of the Asian crisis. Journal of International Money and Finance, 18, 537-560.

Khan, A., & Ahmed, M. S. (2009a). Trading volume and Stock Returns: The Impact of Events in Pakistan on KSE 100 Indexes. International review of Business Research Papers, 5, 373-383.

Khan, M. A., Javed, S., Ahmad, S., Mehreen, & Shadzad, F. (2009b). Impact of Pak-U.S. Relationship News on KSE-100 Index. International review of Business Research Papers, 5, 273-288.


30

Kim, H., & Mei, J. (2001). What makes the stock market jump? An analysis of political risk on Hong Kong stock returns. Journal of International Money and Finance, 20, 1003-1016.

Niederhoffer, V. (1971). The analysis of world events and stock prices. Journal of Business, 193-219.

Niederhoffer, V., Gibbs, S., & Bullock, J. (1970). Presidential elections and the stock market. Financial Analysts Journal, 26, 111-113.

Pantzalis, C., Stangeland, D., & Turtle, H. (2000). Political elections and the resolution of uncertainty: the international evidence. Journal of Banking & Finance, 24, 1575-1604.

Peel, D., & Pope, P. (1983). General Election in the U.K. in the Post-1950 Period and the Behavior of the Stock Market. Investment Analysis, 67, 4-10.

Sandler, T., & Enders, W. (2004). An economic perspective on transnational terrorism. European Journal of Political Economy, 20, 301-316.

Sathye, M., Sharma, D., Liu, S., Chetty, G., & Shadabi, F. (2008). Impact of Terrorism on Australian Stock Market: An Empirical Investigation. Proceedings of the 3rd Biennial Conference of the Academy of World Business, Marketing and Management Development, Conference of the Academy of World Business, Marketing and Management Development, 14-17 July 2008, Rio De Janeiro, Brazil, , 1012-1026

Shelor, R., Anderson, D., & Cross, M. (1990). The impact of the California earthquake on real estate firms' stock value. Journal of Real Estate Research, 5, 335-340.

Valadkhani, A., & Worthington, A. (2005). Catastrophic Shocks and Capital markets: A Comparative Analysis by Disaster and Sector. Global Economic Review, 34(3), 331-344.

Worthington, A., & Valadkhani, A. (2004). Measuring the impact of natural disasters on capital markets: An empirical application using intervention analysis. Applied Economics, 36, 2177-2186.

Zach, T. (2003). Political Events and the Stock Market: Evidence from Israel. International Journal of Business, 8, 243-266.


PPeerrffoorrmmaannccee ooff AAlltteerrnnaattiivvee PPrriiccee FFoorreeccaasstt ffoorr PPaakkiissttaann Yaser Javed and Eatzaz Ahmad1

Abstract

To evaluate the price forecasts, we use two data frequencies i.e., annual and quarter with two most demanding techniques, i.e., ARIMA and VAR models to forecast the four index of inflation, named, Consumer Price Index (CPI), Wholesale Price Index (WPI), GNP Price Deflator (GNPPD), and Implicit Price Deflator of Total Domestic Absorption (DAPD).2 In order to test the performance of price forecast for Pakistan, we found Consumer Price Index (CPI) and Implicit Price Deflator of Total Domestic Absorption (DAPD) better than Wholesale Price Index (WPI) and GNP Price Deflator (GNPPD). In general more elaborate Vector Autoregressive (VAR) models outperform the simplistic Auto Regressive Integrated Moving Average (ARIMA) models in forecasting a price series. Another useful conclusion is that the quarterly data provide better forecasts than the annual data. All these results support the econometricians’ maintained hypotheses that, data observed at high frequency and statistically more elaborate use of a given data set provides better predictions than the data observed at low frequency and analyzed with simplistic statistical tools.

Keywords: ARIMA models; Cointegration; ECM; VAR Models

JEL classification: C12, C13, C15, C22, E30, E31, E37, E58

1. Introduction Uncertainty about future events influence our present decision, the main reason why expectations are made is that we want to incorporate that uncertainty in our present decision to minimize the risk. For example, a student does not know whether it will rain in the afternoon when he/she returns from university. The student has to decide now on the basis of his/her judgment or given knowledge about the pattern of climate whether carry umbrella or leave it at home. A good decision about afternoon, in the morning is important. 1 The authors are Ph.D. Student at Federal Urdu University of Arts, Science and Technology, Islamabad and Professor/Dean at School of Economics, Quaid-i-Azam University Islamabad, respectively. 2 Total domestic absorption price deflator is obtained from addition of imports and subtraction of exports from GNP. This deflator was used by Ahmad and Ram (1991).

Javed and Ahmad

32

Macroeconomic policy makers are interested to know the inflation rates for the coming years. If these figures are alarming then they suggests monetary authorities that to tighten their steps towards monetary policy right now, so that the remedy starts before occurring of the disease. Forecasting is an important exercise in the context of time series analysis according to Yin-Wong and Menzie (1997) a large industry is involved in the forecasting of key macroeconomic variables.

None of the variable can be predict with certainty; decisions are made on the basis of forecasts made by researches are individuals, but no forecast is ever perfect there must be some errors. Importance of correct forecast is obvious, from the observation of Blix et al. (2002) that a bad forecast can lead to loss of business opportunities, loss of investment or to misguide government macroeconomic policies; good forecast, on the other hand, can lead to the opposite. So it is important to test the performance of such forecasts.

The remaining portion of the study is organized as follows. In section 2, we review the existing literature on measuring performance of price forecast and in section 3, empirical findings of the pertinent studies. In section 4, we present data sources, estimation techniques and in section 5 we present different type of performance hypothesis. In section 6, we present the results of our performance tests. Finally section 7 concludes the study.

2. Review of Literature Analysis of time series started before the evolution of modern macroeconomics, according to Yule (1927) forecasting has an even longer history. Importance of time series analysis and forecast is obvious from the observation of Ruey (2000) that objectives of the two studies may differ in some situations, but forecasting is often the goal of a time series analysis.

Forecasting of economic time series is an important but difficult task; especially in case of developing countries due to the poor quality of data, there is also persistently destabilize economic and political environment. Economics outcomes are often influenced by unanticipated events and data may be inadequate, particularly in developing countries. According to Paula (1996) economic forecasting is an art, not a science.

Granger (1996) points out that it is easy to find criticisms of economic forecasts, both of their perceived quality and of the methods used in their construction. No forecast can be properly evaluated in isolation and so it is

Performance of Alternative Price Forecast for Pakistan

33

worth noting that famous book by Box and Jenkins (1976) on univariate models, has attracted substantial opponents in forecasting competitions. According to Granger (1989) it is not possible to give a definite answer to the question like ‘What is the best forecasting method?’. In any particular forecasting situation some methods may be excluded either because of insufficient data or because the cost is too high. If there are no such limitations, it is still not possible to give a simple answer.

Importance of correct forecast is obvious, as according to Blix et al. (2002) a bad forecast can lead to loss of business opportunities3, missed investment or misguide government macroeconomic policies; good forecast, on the other hand, can lead to the opposite. Accuracy of forecast is important to policymakers, as several studies evaluate the forecasts, such as Gavin and Mandal (2000), Oller and Barot (2000) and Batchelor (2001). As mentioned Nordhaus (1987), given the heightened importance of forecasts and expectations, it is natural to inquire into their accuracy and adequacy.

2.1. Consistent Forecast Generally a forecast having lower RMSE is considered better than the ones having a higher value of RMSE. As mentioned by Yin-Wong and Menzie (1997) when examining forecast accuracy researchers examine the mean, variance and serial correlation properties of the forecast errors. The issues of integration and cointegration are rarely addressed. These issues are very important as pointed out by Clement and Hendry (1993) and Armstrong and Fildes (1995) make a criticism on the RMSE, and mention that RMSE is not a good benchmark.

After the rejection of conventional tools of analyzing the forecast, the cointegration approach named ‘consistency’ was introduced, and this technique was also used by Liu and Maddala (1992) and Aggarwal et al. (1995) to assess the unbiasedness, integration and cointegration characteristics of macroeconomic data and their forecasts.

2.2. Efficient Forecast Efficiency norm is defined by different researchers, and in different ways. In a Congressional Budget Office Report (1999) efficiency indicates the extent to which a particular forecast could have been improved by using

3 The expectations of the businessmen and investors play a key role in the business cycles theories presented by Pigou (1927) and Keynes (1936).

Javed and Ahmad

34

additional information that was at the forecaster’s disposal when the forecast was made. Nordhaus (1987) define efficiency in two ways i.e., ‘weak’4 and ‘strong’ efficiency. This kind of efficiency states by Beach et al. (1999).

Bonham and Cohen (1995) criticize the methodology used by Keane and Runkle (1990) that directly tests conditional efficiency of forecast using an approach that based on incorrect integration accounting. Their integrating accounting errors result in trivial cointegration and improper distributional assumption and, therefore, incorrect inference. Bonham and Cohen (1995) claim that they correct the integration accounting errors and show that the efficiency hypothesis is still rejected.5

2.3. Rational Forecast Doctrine of rationality is defined by Lee (1991) as follows, expectations are said to be rational if they fully incorporate all of the information available to the agents at the time the forecast is made. There are many studies like Hafer and Hein (1985), McNees (1986), Pearce (1987) and Zarnowitz (1984 and 1985) that places great weight on minimum mean square error (MSE) but do not incorporate accuracy analysis convincingly in their tests of rationality. However, there are many studies like Holden et al. (1987), Ash (1990 and 1998), Artis (1996), Pons (1999, 2000 and 2001), Kreinin (2000), Oller and Barot (2000) and Batchelor (2001), shows that the IMF and OECD forecasts pass most of the tests of rationality.

Rather than simply compare forecast on the basis of RMSE, Bonham and Douglas (1991) include a test for conditional efficiency6 in the definition of strong rationality. In order to analyze the rationality of price forecast Bonham and Douglas (1991) define a hierarchy of rationality tests starting from ‘weak rationality’ to ‘strict rationality’. The level of rationality in hierarchy is defined as, weak, sufficient, strong and strict.

4 Another notion of efficiency proposed by Bakhshi et al. (2003) is that current forecast errors should be uncorrelated with past forecast. 5 In this study we are not much concern with the colliding debate of efficiency related to Bonham and Cohen (1995) and Keane and Runkle (1990, 1994 and 1995) due to some flaws with respect to comparative analysis between the forecasts obtained from ARIMA and VAR models. 6 Granger and Newbold (1973), describe conditional efficiency as a forecast for which the combination forecast does not produce a lower RMSE than its component forecast.


35

2.3.1. Weak Rationality Most of the applied work such as Evans and Gulmani (1984), Friedman (1980), Pearce (1987) and Zarnowitz (1984 and 1985) view rationality in term of the necessary conditions of unbiasedness and information efficiency.7 According to the notion of weak rationality defines by Bonham and Douglas (1991), the forecast must be unbiased and meet the tests of weak information efficiency.

Ruoss and Marcel (2002) state that unbiasedness is often tested using the Theil-Mincer-Zarnowitz equation. This is a regression of the actual values on a constant and the forecast values. The null hypothesis to be tested is that, the intercept is equal to zero and the slope is equal to one. Holden and Peel (1990) pointed out that this null hypothesis is merely sufficient but not necessary for unbiasedness. Clement and Hendry (1998) suggest, running a regression of the forecast error on the constant, if the parameter estimate deviates from zero, the hypothesis that the forecast is unbiased is rejected.

2.3.2. Sufficient Rationality The forecast must be weak rational and must pass a more demanding test of sufficient orthogonality, namely, that the forecast errors is uncorrelated with any variable in the information set available at the time of prediction.

Rational expectation hypothesis played a critical role in macroeconomic analysis and in the theory of economics decision-making. Rational expectation assumes that economic agents are rational optimizers, especially in making forecasts and in taking actions based on such forecasts. Rational expectations hypothesis by Muth (1961) holds that predictions of future inflation are formed in a manner that fully reflects relevant information currently available.

2.3.3. Strong Rationality The forecast must be sufficiently rational and pass tests of conditional efficiency. Conditional efficiency requires a comparison of forecasts.8 Consider a sufficiently rational forecast as a benchmark. Combine benchmark 7 The same kind of unbiasedness and efficiency notion was build by Eichenbaum et al. (1988) and Razzak (1997). 8 Started from the classic study of Bates and Granger (1969), a large literature on forecast combination summarized by Clemen (1989), Diebold and Jose (1996) and Timmermann (2005) has found evidence that combined forecasts tend to produce better forecast than individual forecasting models.

Javed and Ahmad

36

with some competing forecast. Conditional efficiency refers to Granger and Newbold (1973) that measures the reduction in RMSE, which occurs when a forecast is combined with one of its competitors. Against such kind of notion Granger (1989) suggest that combining often produces a forecast superior to both components. Same kind of notion is build by Timmermann (2006). If the combination produces an RMSE that is significantly smaller than the benchmark RMSE, the benchmark forecast fails the test for conditional efficiency because it has not efficiently utilize some information contained in the competing forecast. Stock and Mark (2001) report broad support for a simple combination of forecasts in a study of a large cross-section of macroeconomic and financial variables.

2.3.4. Strict Rationality According to Bonham and Douglas (1991) a statement about rationality should not depend on arbitrary selection of time periods. A forecast is strictly rational if it passes tests of strong rationality in a variety of sub-periods, stated in section 5.3.4.

3. Empirical Findings Yin-Wong and Menzie (1997) concludes that the (final) Treasury bill rate, housing starts, industrial production, inflation and most of their respective forecasts appear to be trend stationary. The corporate bond rate, GNP, the GNP deflator, unemployment and most of their respective forecasts appear to be difference stationary. About half of the unit root pairs are cointegrated. In only one of these cases the unitary elasticity restriction is rejected the 1-quarter ahead GNP deflator forecast. In the study of Yin-Wong and Menzie (1997) 30 out of 36 cases fulfill the requirement that forecast and actual series possess the same order of integration. Surprisingly, the linkage between forecasts and unrevised actual series is not unambiguously stronger. However, while there is more evidence of cointegration, there is also a greater rate of rejection of the unitary elasticity restriction.

The evidence from the study of Aggerwal et al. (1995) indicate that there are significant deviations from the rational expectations hypothesis for survey forecasts of a number of macroeconomics series. They find that survey forecasts for the consumer price index and personal income are stationary and consistent with the rational expectation hypothesis and that the surveys of housing starts, the unemployment rate and the trade balance are rational forecasts in the sense that the announced values and their survey


37

forecasts are cointegrated. Aggerwal et al. (1995) suggests, that the quality of forecast of industrial production and retail sales can be improved significantly by using past values. These results have important implications for decisions by many economic agents and for research based on these survey forecasts and also favoring the univaraite methodology.

Results of weak efficiency hypothesis stated by Nordhaus (1987) are that 50 of 51 tests, the forecast were found to be positively correlated. The degree of correlation appears to be highest for institutional forecasts (such as those made by international agencies) and lowest for professional forecasters using time-series techniques. Nordhaus (1987) describes two reasons for this kind of inefficiency. First, perhaps the true forecasts are indeed efficient, while the published forecasts are not. Second, surely the high degree of forecast inefficiency of international institutions must contain some element of bureaucratically based forecast inefficiency.

Empirical results regarding the rationality of forecasts was explained by Lee (1991) that forecast is fail to be rational in the strong sense even though they are not rejected by the conventional test of weak rationality. Ruoss and Marcel (2002) examine the forecast rationality of the Swiss economy says that GDP forecasts in our sample do not pass the most stringent test i.e., the test of strong informational efficiency, because, in some cases, forecasts errors correlate with the forecasts of the other institutes.

Same kind of results is shown by Bonham and Douglas (1991) that the most stringent criteria for testing rationality will not be useful for empirical work. On these criteria there might not be a rational forecast of inflation. Bonham and Douglas (1991) states that, rational forecast is getting by relaxing the criterion that defines strict rationality.

Razzak (1997) and Rich (1989) test the rationality of National Bank of New Zealand’s survey data of inflation expectation and SRC expected price change data respectively. Both studies end up with a same conclusion, that the results do not reject the null hypothesis of unbiasedness, efficiency and orthogonality for a sample from their particular survey data series.

4. Data Sources and Forecast Modeling

In order to test the performance of price forecast for Pakistan, we forecast four proxies of prices, namely, Consumer Price Index (CPI), Wholesale Price Index (WPI), GNP Price Deflator (GNPPD) and Implicit Price Deflator of Total Domestic Absorption (DAPD). Annual data is taken from

Javed and Ahmad

38

various issues of Economic Survey of the Ministry of Finance, Government of Pakistan, and Annual Reports of State Bank of Pakistan. Quarterly data is taken from the IMF’s International Financial Statistics (2005) and World Bank’s World Development Indicator (2006). Data of quarter GDP is taken from the research paper of Kemal and Arby (2001). Data is taken on annual and quarter basis for the period from 1972-73 to 2004-05 and 1972Q2 to 2005Q2, respectively.

For a better forecast, our estimation is based on univariate and multivariate techniques. For the univariate technique, we use the Box-Jenkins approach to modeling ARIMA models (Box and Jenkins, 1976). For the multivariate technique, we use VAR approach presented by Sims (1980). In the estimation of VAR we use price variable alternatively with the four other variables, real GDP, Broad Money (M2), interest rate and exchange rate.

After three stages of identification, estimation and diagnostic checking, we present the specification of ARIMA models in table 4.1. In table 4.2, we present the lag specification of VAR models.

Table 4.1: Specification of ARIMA Models

Annual Data Consumer Price Index ARIMA (1,1,1) Wholesale Price Index ARIMA (0,1,1) GNP Price Deflator ARIMA (0,1,1) Domestic Absorption PD ARIMA (0,1,1)

Quarterly Data Consumer Price Index ARIMA (0,1,0) Wholesale Price Index ARIMA (4,1,0) GNP Price Deflator ARIMA (4,1,4) Domestic Absorption PD ARIMA (4,1,4)

Note: ARIMA (p,d,q) stands for a model with autoregressive process of order p and moving average process of order q applied to data integrated of order d.

Table 4.2: Specification of VAR Models

Annual Data Consumer Price Index Wholesale Price Index

VAR (1) VAR (1)

GNP Price Deflator VAR (1) Domestic Absorption PD VAR (1)

Quarterly Data Consumer Price Index VAR (1,2) Wholesale Price Index VAR (1,4) GNP Price Deflator VAR (1,2,4) Domestic Absorption PD VAR (1,4)

Note: The number in brackets show the lag periods specified in the VAR models.


39

5. Performance Hypothesis After getting the forecasts we test the performance of price forecasts by applying the different type of hypothesis under the definition of consistency, efficiency and rationality.

5.1. Consistency Test of Forecast Consistent forecast states that the, observed price index and their relevant forecast series are integrated of same order and they are cointegrated. To test the existence of unit root we follow the spirit of Dickey and Fuller (1979, 1981). According to them if yt follows AR(p) process.

tptpttt yyyy εφφφ ++++= −−− .......2211 , a series yt is said to be

stationary, if the value of ∑=

p

ii

1

φ is less than unity. If the observed variable and

their forecast are of same level of integration, say I(1). Then the first condition for consistency is met. Concept of cointegration was first introduced by Granger (1981) and elaborates further by Engle and Granger (1987). The spirit of the cointegration in this study is that observed price index (Po) is cointegrated with their forecast (Pe). Both series posses same order of integration, say I(1), then the linear combination9 of these two must be I(0). We define it in following way.

tto

te PP ε+Φ+Φ= 21 )0(It ≈ε (1)

Where { }21,ΦΦ is the cointegrating vector producing a linear combination of{ }to

te PP , , which is stationary. This will complete the proposition of

cointegration. After that there is a need to test the stability of long run relationship through error correction models.

5.1.1. Error Correction Models For the Error correction we estimate the following equations.

t

m

iit

eitt

e uPP +∆++=∆ ∑=

−−1

121 δεαα (2)

t

n

iit

oitt

o vPP +∆++=∆ ∑=

−−1

121 γεββ (3)

9 We will apply Granger Causality test presented by Granger (1969), to determine dependent variable in the linear combination of observed price series with their forecast series.

Javed and Ahmad

40

The selection of m and n in equation 2 and 3 depends on the significance of lags under t-statistics. For a stable long run relationship between observed price index with forecast the following feedback effect must be less than zero, that is.

0222 <Φ− βα (4)

If the above condition holds, it implies that disequilibrium in previous period leads to adjustment in current time period, which counter balance the disequilibrium forces.

5.2. Efficiency Test of Forecast Nordhaus (1987) define efficiency in the two classifications; weak efficiency is the necessary condition for strong efficiency, but clearly not the sufficient condition.

5.2.1 Weak Efficiency

A forecast is weakly efficient if it minimizes ( ) }{ 2t tu JΕ , where Jt is

the set of all past forecasts. Where Ut2 is the square of forecast error at time t.

In order to test weak efficiency of forecasts obtained from both techniques, we estimate the following regression.

t

k

iit

eiot PU εαα ++= ∑

=

−

1

2 (5)

Selection of k depends upon the significance under t-statistics. Only significant lags of expected price forecasts are included. Under this kind of efficiency norm, a forecast is said to be weak efficient if we are unable to reject the null that all the coefficients are simultaneously equal to zero.

5.2.2. Strong Efficiency

A forecast is strongly efficient if ( ){ }2t tu IΕ is minimized, where It is

all information available at time t. Strong efficiency requires that the square of forecast error was not explained by the information set available at time t. The information set in Univariate analysis is the past values of the variable itself, so we regress the following equation, to test the strong efficiency for the forecasts obtained from ARIMA models.


41

t

n

jit

ojot PU εαα ++= ∑

=

−

1

2 (6)

Here Pot is the observed value of price variable at time t. A forecast fails to

pass the strong efficiency hypothesis if α0 and αj are significantly different from zero. In order to test the strong efficiency of forecasts obtained from VAR we estimate the following regression.

tttttto

ot ERRMRGDPPU εαααααα ++++++= −−−−− 15141312112 2 (7)

A strongly efficient forecast obtained from VAR fail to reject the null hypothesis that all the coefficients in equation 7 are simultaneously equal to zero.

5.3. Rationality Test of Forecast Bonham and Douglas (1991) define a hierarchy of rationality tests starts from ‘weak rationality’ to ‘strict rationality’ the level of hierarchy define as follows:

5.3.1 Hypothesis of Weak Rationality A forecast must be unbiased and meet tests of weak information efficiency. Condition of unbiasedness and weak informational efficiency is set after the estimation of following equation.

tte

oto PP εαα ++= 1 (11)

A forecast is said to be unbiased if it satisfies the following conditions.

1. In equation 11, εt is serially uncorrelated.

2. In equation 11, αo and α1 are insignificantly different from zero and one respectively.

Weak information efficiency means that the forecast errors to

te

t PPE −= are uncorrelated with the past values of the predicted variables. To test the weak efficiency hypothesis we estimate the following regression equation.

t

m

iit

oiot PE εαα ++= ∑

=

−

1 (12)

Javed and Ahmad

42

If we fail to reject the following joint null hypothesis it implies that forecast errors are systematically different from zero and/or past values of the observed price series help to explain the forecast errors.

0: == jooH αα For all j = 1……….. m (13)

Acceptance of such hypothesis represent that the forecast error at time t is independent to the past information contained by relevant observed price index.

5.3.2. Hypothesis Sufficient Rationality The sufficient rationality requires that the forecast errors are not correlated with any variable in the information set available at the time of forecast. If Zt is a variable or a vector of variables used to build our forecast model, then Zt is the exogenous variable in the following equation.

t

m

iitiot ZE εαα ++= ∑

=−

1 (14)

Forecasts of ARIMA models have included only the lags of observed series as the information set. For ARIMA forecasts two lags of associated price index are used as information set. While forecasts obtained from VAR models depend upon the lags of price variables, real GDP, M2, interest rate, and exchange rate, so their lags with relevant price series are used to test sufficient rationality. After estimating the equation 14 we test the following null hypothesis.

0: == jooH αα For all j = 1……….. m (15)

The rejection of above mentioned hypothesis states that the information contained in the past values of related price series, real GDP, M2, interest rate and exchange rate, has not been used efficiently in forming the forecast.

5.3.3. Hypothesis of Strong Rationality A forecast is said to be strongly rational if it passes the test of conditional efficiency introduced by Granger et al. (1973). Conditional efficiency requires a comparison of forecasts. Call some sufficiently rational forecast as benchmark; combine the benchmark with some competing forecast. Estimate the following regression.

[ ] tttt SSD εβα +−+= (16)


43

Where Dt and St are the difference and the sum of the benchmark and combination forecast errors, respectively, and tS is the mean of the sum. Under the null hypothesis of conditional efficiency (α=β=0) the combination does not produce a lower RMSE. F test is appropriate if β>0 and the mean errors of both forecasts have the same sign as α. If the mean errors of the two forecasts do not have the same sign, then α cannot be interpret as an indicator of the relative bias of the two forecasts.

5.3.4. Hypothesis of Strict Rationality A forecast is strictly rational if it passes tests of strong rationality in a variety of sub-periods. In this study only quarter forecasts of CPI can be treated for strong efficient criterion, annual data do not have sufficient number of observation to sub-divide in various sub-periods, so we estimate equation 16 in the sub-periods;1972-Q3 to1982-Q4, 1983-Q1 to 1994-Q2 and 1994-Q3 to 2005-Q2.

If a strongly rational forecast pass the same test based on equation 16 in sub-periods mentioned above then according to Bonham and Douglas (1991) that particular forecast is awarded as strict rational.

6. Results and Discussion We are not going to discuss the conventional tools for analyzing the performance of forecasts, as lower RMSE and the maximum value of covariance proportion etc. as Clement and Hendry (1993), Armstrong et al. (1995) make a criticism on the RMSE, and mention that RMSE is not a good benchmark. In general, we can say that forecast of Consumer Price Index (CPI) is the best (among the others proxies of price variables used in this study), while the forecasts of Wholesale Price Index (WPI) are not performing well with reference to consistency, efficiency and rationality tests, because forecasts from VAR models are not able to meet the tests of weak and strong efficiency except for the quarterly CPI forecast that significantly accept the weak efficiency hypothesis.

6.1. Results of Consistency Tests of Forecast As an initial condition of consistency, observed and expected price variables should be the same order of integration. The results of unit root tests of observed data series are given in table 6.1.1.

It is obvious from the results given in table 6.1 that the four price series included in this study have unit root at levels form. Other variables also

Javed and Ahmad

44

have unit root except the annual series of interest rate that is stationary at level. One variable i.e., WPI is stationary at 10% level of significant. But following the general practice of considering the level of significant at 5%, we conclude that all the annual and quarter observed data series except the annual interest rate series are I(1). Simply the four annual and four quarter price series are I(1), then in order to satisfies the conditions of consistency the forecasts series must be I(1). The results of unit root test for the ARIMA and VAR forecasts are shown in table 6.1.2 and 6.1.3 respectively.

The results in these tables shows that all the forecasts series obtained in this study are I(1). For consistency the second condition is that the observed price series must be cointegrated with their respective forecast. In this study find the evidence on cointegration between observed

Table 6.1.1: Unit Root Tests of Observed Variables

Variables t-values (Annual Data) t-values (Quarterly Data)

Level First Diff. Level First Diff. Consumer Price Index -1.38 -4.71*** -1.53 -10.28*** Wholesale Price Index -0.79 -4.93*** -2.87* -8.54*** GNP Price Deflator -1.46 -4.04*** -1.72 -10.07*** Domestic Absorption Price Deflator

-1.18 -4.27*** -1.39 -10.54***

Real GDP -0.80 -8.45*** -0.98 -21.02*** Interest Rate -3.66*** ------ -1.97 -10.13*** Exchange Rate 0.45 -5.26*** 0.98 -11.81*** M2 -0.39 -4.20*** -0.18 -18.23***

* Significant at 10% level of Significance and *** Significant at 1% level of significance.

Table 6.1.2: Unit Root Test of Forecasts from ARIMA Models

Variables t-values (Annual Data) t-values (Quarterly Data)

Level First Diff. Level First Diff. Consumer Price Index -0.35 -5.12*** -1.37 -10.35*** Wholesale Price Index -2.15 -7.64*** -1.72 -10.98*** GNP Price Deflator -0.89 -3.87*** -1.18 -10.05*** Domestic Absorption Price Deflator

-0.81 -4.34*** -0.90 -10.50***

*** Significant at 1% level of significance.

price series and their relevant forecasts, we first used Granger Causality test to determine dependent variable in the linear combination of forecast and actual


45

series. The series of linear combination are stationary at level form so we say them I(0), except the annual WPI forecast series obtained from ARIMA models as shown in the table 6.1.4.

One forecast out of sixteen is shown to be not consistent, that is the annual forecast of WPI obtained from the ARIMA models. Otherwise all remaining fifteen forecast series turned out to be consistent on the basis of cointegration. This means that there exists a long-run relationship between

Table 6.1.3: Unit Root Test of Forecasts from VAR Models

Variables t-values of the Root with Annual Data

t-values of the Root with Quarterly Data

Level First Diff. Level First Diff. Consumer Price Index -0.70 -5.20*** -1.36 -10.61*** Wholesale Price Index -1.27 -5.68*** -1.44 -11.82*** GNP Price Deflator -0.43 -5.83*** -0.95 -11.56*** Domestic Absorption Price Deflator

-0.36

-6.58***

-0.78

-12.94***


Table 6.1.4: Unit Root Test of Linear Combination of Observed

Variables with their Forecasts

Variables t-values (Annual Data) t-values (Quarterly Data) Level First Diff. Level First Diff.

Consumer Price Index -4.14*** -10.13*** 4.58*** -12.04*** Wholesale Price Index -2.74* -11.40*** -8.48*** -7.57*** GNP Price Deflator -7.48*** -10.57*** -10.04*** -12.84*** Domestic Absorption Price Deflator

-6.99***

-10.78***

-10.55***

-12.64***

* Significant at 10% level of Significance and *** Significant at 1% level of significance.

observed and forecasted price series. Now there is a need to check the stability of the long run relationship that is to determine whether or not this relationship is stable in the long run. For a stable long run relationship the feedback effects obtained from the error correction mechanism should be negative.

Table 6.1.5 shows that all the feedback effects are negative, implying that all the consistent relationships between observed and forecasted price series are stable in the long run. Thus disequilibrium between observed and

Javed and Ahmad

46

expected series in any period is eliminated in the subsequent period. In short, we can say that we found fifteen out of sixteen forecast series consistent and having a stable consistent long-run relationship with their relevant observed price series.

Table 6.1.5: Feedback Effects10 of Forecasts

ARIMA VAR

Annual Data Consumer Price Index -1.604 -1.458 Wholesale Price Index -2.279 -2.664 GNP Price Deflator -2.524 -3.857 Domestic Absorption PD -2.442 -3.936

Quarterly Data Consumer Price Index -1.816 -1.564 Wholesale Price Index -1.865 -1.732 GNP Price Deflator -1.659 -1.755 Domestic Absorption PD -1.626 -1.511

6.2. Results of Efficiency Tests of Forecast In the debate of efficiency we present the results of weak efficiency, the concept represents by Nordhaus (1987) as a necessary but not the sufficient condition for strong efficiency. Tables 6.2.1 and 6.2.2 represent the results of weak efficiency of annual forecasts obtained from

Table 6.2.1: Weak Efficiency of Annual Forecasts (ARIMA Models)

t

k

iit


=

−

1

2 Ho: All the coefficients are equal to zero

Equation α0 α1 α2 χ2 for Ho F-stat. for Ho CPI

Equation 9.39

(-0.98) -0.023 (-0.15)

----------- 3.002 (0.22)

1.501 (0.24)

WPI Equation

-2.11 (-0.56)

-2.00 (-3.71)***

2.35 (4.12)***

35.89 (0.00)***

11.966 (0.00)***

GNPPD Equation

-16.62 (-1.08)

-9.821 (-2.79)***

11.42 (3.07)***

24.72 (0.00)***

8.239 (0.00)***

DAPD Equation

-15.61 (-1.03)

-11.03 (-3.21)***

12.72 (3.48)***

25.398 (0.00)***

8.466 (0.00)***

Notes: t-statistics are in parentheses under the coefficients. Probabilities are in parentheses under the test statistics. *** Significant at 1% level of significance. 10 We calculate the feedback effects using Engle and Granger (1987), procedure are defined in section 5.1.1.


47

ARIMA models and VAR models respectively. Annual forecasts obtained from ARIMA are not good on the basis of weak efficiency test, except the forecast of CPI. Results reported in the table 7 shows that only the CPI forecast is weak efficient. Table 8 shows that the situation is worse for those annual forecasts we obtained from VAR, where not a single forecast series is able to pass the test of weak efficiency.

Table 6.2.2: Weak Efficiency of Annual Forecasts (VAR Models)

t

k

iit


=

−

1

2 Ho: All the coefficients are equal to zero.

Equation α0 α1 χ2 for Ho F-stat. for Ho

CPI Equation

3.04 (0.485)

0.09 (0.884)

7.18 (0.03)**

3.59 (0.04)**

WPI Equation

-0.69 (-0.613)

0.07 (3.549)***

29.38 (0.00)***

14.69 (0.00)***

GNPPD Equation

-18.43 (-2.135)**

0.65 (4.084)***

21.54 (0.00)***

10.77 (0.00)***

DAPD Equation

-15.78 (-2.220)**

0.57 (4.336)***

24.39 (0.00)***

12.2 (0.00)***

*** Significant at 1% level of significance. ** Significant at 5% level of significance.

Results presented in table 6.2.3 shows that the quarterly forecast of WPI obtained from ARIMA models is not a weak efficient forecast, while forecasts of CPI, GNPPD and DAPD accept the weak efficiency hypothesis.

Table 6.2.3: Weak Efficiency of Quarter Forecasts (ARIMA Models)

t

k

iit


=

−

1



CPI Equation

383.41 (1.06)

-0.14 (-0.33)

2.33 (0.31)

1.16 (0.32)

WPI Equation

-99.61 (-1.40)

0.41 (5.43)***

57.58 (0.00)***

28.79 (0.00)***

GNPPD Equation

0.24 (-0.77)

0.04 (1.82)*

4.72 (0.09)*

2.36 (0.10)*

DAPD Equation

-0.25 (-0.79)

0.04 (1.92)*

5.26 (0.07)*

2.63 (0.08)*

*** Significant at 1% level of significance. * Significant at 10% level of significance.

Javed and Ahmad

48

Table 6.2.4: Weak Efficiency of Quarter Forecasts (VAR Models)

t

k

iit


=

−

1



CPI Equation

361.73 (1.13)

-0.10 (-0.26)

3.11 (0.21)

1.55 (0.22)

WPI Equation

-116.39 (-1.69)*

0.46 (6.13)***

71.42 (0.00)***

35.71 (0.00)***

GNPPD Equation

-0.26 (-0.9)

0.042 (2.03)**

5.70 (0.06)*

2.85 (0.06)*

DAPD Equation

-0.27 (-0.88)

0.05 (2.10)**

6.29 (0.04)**

3.14 (0.05)**

*** Significant at 1% level of significance. ** Significant at 5% level of significance. * Significant at10% level of significance.

Quarter forecasts of GNPPD from both techniques are passing the test of weak efficiency. These results are seems to be coherent with the Nordhaus (1987), as they also find a few week efficient forecasts. After describing the results of weak efficiency, we now present the results of strong efficiency test.

Table 6.2.5: Strong Efficiency of Annual Forecasts (ARIMA Models)

t

k

iit

oiot PU εαα ++= ∑

=

−

1



CPI Equation

8.21 (0.91)

-0.005 (-0.04)

3.00 (0.22)

1.50 (0.24)

WPI Equation

-4.13 (-0.98)

0.21 (2.92)***

13.98 (0.00)***

6.99 (0.00)***

GNPPD Equation

-22.72 (-1.51)

0.91 (3.16)***

13.14 (0.00)***

6.57 (0.00)***

DAPD Equation

-19.66 (-1.26)

0.82 (2.78)***

10.32 (0.00)***

5.16 (0.01)***


Results of strong efficiency presented in table 6.2.5, indicate that from annual forecast computed by ARIMA models, no index passes the test of strong efficiency except the forecast of CPI. The results of strong efficiency reported in table 6.2.6 shows that quarter forecasts of CPI, GNPPD and DAPD all pass the test of strong efficiency whereas the WPI forecast does not


49

pass the test. We are not stating the results of strong efficiency of annual and quarter forecasts, obtained from VAR models. These results show that neither annual nor quarter forecast pass the test of strong efficiency.

Table 6.2.6: Strong Efficiency of Quarter Forecasts ARIMA Models

t

k

iit

oiot PU εαα ++= ∑

=

−

1


Equation α0 α1 χ2 for Ho F-stat. for Ho CPI

Equation 372.04 (1.40)

-0.13 (-0.31)

2.31 (0.31)

1.16 (0.32)

WPI Equation

-103.42 (-1.47)

0.42 (5.55)***

59.08 (0.00)***

29.54 (0.00)***

GNPPD Equation

-0.22 (-0.72)

0.04 (0.766)*

4.53 (0.10)*

2.27 (0.11)

DAPD Equation

-0.23 (-0.73)

0.04 (1.862)*

5.05 (0.08)*

2.52 (0.08)*

*** Significant at 1% level of significance. * Significant at10% level of significance.

6.3. Results of Rationality Tests of Forecast In this section we discuss the results of the rationality tests of forecast we get from ARIMA and VAR models. We estimate a hierarchy of rationality tests starting from ‘weak rationality’ to ‘strict rationality’ presented by Bonham and Douglas (1991).

In table 6.3.1 we present the results of weak rationality of ARIMA forecast, which was the combination of unbiasedness and weak informational efficiency present in the top panel. Where the first regression equation is the famous Theil-Mincer-Zarnowitz equation. This is a regression of the observed series on a constant and the forecast series, and their regression residuals must be serially uncorrelated to fulfill the condition of unbiasedness as well as fail to rejecting the null presented in front of first equation and the second equation represents the weak informational efficiency, if the null in front of that equation is accepted.

According to the results of weak rationality, the forecast of CPI and DAPD pass this test in both time frequencies. Quarter forecast of GNPPD also passes the test of weak efficiency, but annual forecast of GNPPD is found to be biased. Annual forecast of WPI is biased and the null hypothesis of weak informational efficiency is rejected. On the other hand the quarter forecast of WPI is found weak efficient as shown in table 6.3.1.

Javed and Ahmad

50

Here we find the same kind of evidence about the forecasts of CPI and DAPD as we result out in efficiency analysis, that these two series are better than WPI and GNPPD. Annual and quarter forecasts of CPI and DAPD are pass the two sets of tests for the rationality, therefore ARIMA models for the two indices produce rational forecasts. Annual forecast of WPI fails in both tests, while quarter forecast amazingly passes both the conditions for rationality. This is the major breakthrough of this research, because according to Bonham and Douglas (1991), Lee (1991) and Ruoss and Marcel (2002), many of the forecasts were not able to pass these tests of rationality. In table 6.3.2 we present the same set of test applied on those forecasts obtained from VAR model. There exist some similarities between results stated in table 6.3.1 and 6.3.2.

If we summarize the results of weak rationality tests of ARIMA forecasts, we find that quarter forecasts pass the both tests, annual forecasts of CPI and DAPD also passes both tests, while forecast of GNPPD is biased, but it passes the weak informational efficiency hypothesis. Quarter forecasts obtained from VAR models are found to be weakly rational, except the forecast of WPI that is biased forecast. Forecasts of CPI from annual and quarter data frequencies pass both the tests. Annual forecasts of WPI, GNPPD and DAPD are biased, but they pass the test of weak informational efficiency.

This is a hierarchy of rationality test, so we apply sufficient rationality test, to those forecast series that pass both weak informational efficiency and unbiasedness test, which are required for the weak rationality. The results of sufficient rationality test are shown in table 6.3.3 indicate that the annual forecast of CPI and DAPD obtained from ARIMA are able to pass the test of sufficient rationality. Quarter forecast of GNPPD and WPI obtained from ARIMA models passes this test of sufficient rationality. Annual and quarter forecasts of CPI obtained from VAR are sufficiently rational, while quarterly forecasts of GNPPD and DAPD do not pass the sufficient rationality test.

Strong efficiency depends on the concept presented by Granger and Newbold (1973), requiring that a forecast is combined with one of its competing forecast and the combination forecast does not produce a lower RMSE. If we look at the quarter forecasts the WPI forecast obtained from VAR is not found to be as weakly rational, GNPPD and DAPD forecasts do not posses the same signs of mean forecast error, only one forecast i.e.,


51

Table 6.3.1: Weak Rationality Tests of Forecasts (ARIMA Models)

tte

oto PP εαα ++= 1 (HoA: α0 = 0, α1 = 1)

t

m

iit


=

−

1

(oB: α0 = αj = 0)

Dependent Variable

Data Frequency

α0 α1 F-stat. (Ser. Corr.)

Null Hypothesis

χ2 for Ho

F-stat. (Ho)

CPI Annual 0.19 (0.19)

1.00 (61.02)***

1.31 (0.26)

HoA 0.22

(0.89) 0.11

(0.89) Forecast

Errors of CPI Annual -0.19

(-0.19) -0.00

(-0.05) Ho

B 0.22 (0.89)

0.11 (0.89)

CPI Quarter 0.42 (0.15)

1.00 (302.3)***

1.42 (0.24)

HoA 0.47

(0.79) 0.23

(0.79) Forecast

Errors of CPI Quarter -0.43

(-0.15) -7.2e-04 (-0.21)

HoB 0.47

(0.79) 0.23

(0.79) WPI Annual 0.81

(1.19) 0.97

(89.2)*** 9.29

(0.00)*** Ho

A 6.73 (0.03)**

3.36 (0.05)**

Forecast Errors of WPI

Annual -0.73 (-1.08)

0.02 (2.18)**

HoB 6.11

(0.05)** 3.06

(0.06)* WPI Quarter 2.88

(1.29) 0.99

(418.5)*** 0.07

(0.792) Ho

A 5.17 (0.07)*

2.58 (0.08)*


Quarter -2.85 (-1.28)

0.005 (2.12)**

HoB 5.13

(0.08)* 2.56

(0.08)* GNPPD Annual 0.42

(0.35) 0.99

(47.54)*** 4.49

(0.04)** Ho

A 0.48 (0.786)

0.24 (0.79)

Forecast Errors of GNPPD

Annual -0.32 (-0.26)

0.01 (0.533)

HoB 0.35

(0.83) 0.17

(0.84)

GNPP Quarter -0.03 (-0.47)

1.00 (209.1)***

0.34 (0.56)

HoA 1.35

(0.51) 0.67

(0.51) Forecast Errors of GNPPD

Quarter 0.032 (0.48)

-0.004 (-1.01)

HoB 1.35

(0.51) 0.67

(0.51)

DAPD Annual 0.03 (0.24)

0.997 (49.68)***

2.56 (0.12)

HoA 0.04

(0.98) 0.02

(0.98) Forecast Errors of DAPD

Annual 0.03 (0.02)

0.001 (0.08)

HoB 0.02

(0.99) 0.01

(0.99)

DAPD Quarter -0.04 (-0.55)

1.01 (203.67)***

0.13 (0.72)

HoA 1.65

(0.44) 0.82


Quarter 0.04 (0.54)

-0.005 (-1.13)

HoB 1.63

(0.44) 0.82

(0.44)

*** Significant at 1% level of significance. ** Significant at 5% level of significance. * Significant at 10% level of significance.

Javed and Ahmad

52

Table 6.3.2: Weak Rationality Tests of Forecasts Obtained from VAR

tte

oto PP εαα ++= 1 (HoA: α0 = 0, α1 = 1)

t

m

iit


=

−

1

(HoB: α0 = α1 = 0)

Dependent Variable

Data Frequency

α0 α1 F-stat. (Ser. Corr.)

Null Hypothesis

χ2 for Ho

F-stat. for Ho

CPI Annual -0.14 (-0.14)

1.00 (61.5)***

0.42 (0.52)

HoA 0.07

(0.97) 0.04

(0.97) Forecast

Errors of CPI Annual 0.30

(0.30) -0.003 (-0.19)

HoB 0.11

(0.95) 0.05

(0.95) CPI Quarter 0.55

(0.19) 1.00

(296.8)*** 0.61

(0.44) Ho

A 1.08 (0.59)

0.54 (0.59)

Forecast Errors of CPI

Quarter -0.39 (-0.14)

-0.001 (-0.43)

HoB 1.13

(0.57) 0.57

(0.57) WPI Annual 0.41

(0.81) 0.99

(122.4)*** 6.95

(0.01)***Ho

A 1.64 (0.44)

0.82 (0.45)


Annual -0.39 (-0.77)

0.010 (1.19)

HoB 1.56

(0.46) 0.78

(0.47) WPI Quarter 1.59

(0.68) 1.00

(400.6)*** 5.75

(0.00)***Ho

A 1.51 (0.47)

0.75 (0.47)


Quarter -1.42 (-0.61)

-0.0003 (-0.120)

HoB 1.52

(0.47) 0.76

(0.47) GNPPD Annual 1.12

(1.15) 0.97

(59.48)*** 25.43

(0.00)***Ho

A 2.80 (0.25)

1.40 (0.26)

Forecast Errors of GNPPD

Annual -1.09 (-1.09)

0.029 (1.55)

HoB 2.51

(0.28) 1.25

(0.30)

GNPPD Quarter 0.004 (0.05)

1.002 (208.5)***

2.60 (0.109)

HoA 0.73

(0.69) 0.37

(0.70) Forecast Errors of GNPPD

Quarter -5.e-04 (-0.007)

-0.002 (-0.52)

HoB 0.80

(0.67) 0.40

(0.67)

DAPD Annual 0.95 (1.04)

0.98 (64.0)***

30.29 (0.00)***

HoA 1.97

(0.37) 0.99


Annual -0.94 (-0.99)

0.0233 (1.33)

HoB 1.80

(0.41) 0.89

(0.42)

DAPD Quarter -0.005 (-0.069)

1.005 (201.2)***

2.03 (0.16)

HoA 2.72

(0.26) 1.36


Quarter 0.009 (0.13)

-0.005 (-1.12)

HoB 2.88

(0.24) 1.44

(0.24)



53

forecast of CPI satisfying all the conditions for strong rationality. While from annual forecast series obtained from VAR, only CPI passes the test of sufficient rationality, and the forecast series obtained from the ARIMA also passes this test, but the sign of mean forecast error of two series is not same.

Table 6.3.3: Sufficient Rationality Tests of Forecasts

Regress the forecast error to information set and set the null hypothesis that all the coefficients are simultaneously equal to zero.

Forecast error of Obtained from

Data Frequency

χ2 for Ho F-stat. for Ho

CPI ARIMA Annual 1.44 (0.69)

0.48 (0.69)

DAPD ARIMA Annual 3.50 (0.32)

1.16 (0.34)

CPI ARIMA Quarter 1.97 (0.58)

0.66 (0.58)

WPI ARIMA Quarter 5.10 (0.16)

1.70 (0.17)

GNPPD ARIMA Quarter 1.53 (0.67)

0.51 (0.67)

DAPD ARIMA Quarter 1.65 (0.64)

0.55 (0.65)

CPI VAR Annual 20.07 (0.00)***

3.34 (0.01)***

CPI VAR Quarter 8.82 (0.18)

1.47 (0.19)

GNPPD VAR Quarter 29.99 (0.00)***

5.00 (0.00)***

DAPD VAR Quarter 36.26 (0.00)***

6.04 (0.00)

*** Significant at 1% level of significance. We apply strong rationality test only to the quarter forecasts of CPI from both techniques. The results of strong rationality are shown in table 6.3.4. Postulate that both the series posses the negative sign of mean forecast error, when we take ARIMA forecast of CPI as benchmark and combined it with VAR forecast, it gives us biased results because that the sign of α is positive, as shown in panel A of table 6.3.4, so forecast series obtained from ARIMA models do not pass strong rationality test. In panel B of the table 6.3.4 we take VAR forecast as benchmark and combine it with ARIMA forecast. We found that the forecast series of CPI obtained from VAR, when

Javed and Ahmad

54

combined with ARIMA forecast does not produce lower RMSE. This result means that the forecast of CPI obtained from VAR can be claimed to be as strongly rational.

Table 6.3.4: Test of Strong Rationality

Benchmark Forecast When Combined With Panel A CPI from ARIMA CPI from VAR Sign Mean Error -ve -ve α 0.38 β -0.04 Prob. 0.72 Conclusion Bias Panel B CPI from VAR CPI from ARIMA Sign Mean Error -ve -ve α -0.38 β 0.04 Prob. 0.72 Conclusion Cannot Reject

Data Sample: 1972Q3-2005Q2

Conditions of Strict rationality simply state that strongly rational forecasts pass the same test of strong rationality with different sub-time periods. We break the whole sample in three parts, when we check the sign of mean forecast errors of both series while taking the sample from 1972Q3 to 1982Q4, the sign of mean forecast error is not the same, while from 1983Q1 to 1994Q2 and from 1994Q3 to 2005Q2, the sign of mean forecast error are negative of both series. In the first time span that is from 1983Q1 to 1994Q2, we are not able to find unbiased results, as the sign of α is positive. When we take sample from 1994Q3 to 2005Q2, we find the CPI forecast passes the conditional efficiency test that is the RMSE of combination is not lower than the benchmark forecast as shown in table 6.3.5. But the condition of strict rationality is not satisfied, because from the three sub-sample time periods, forecast of CPI passes the test only for one sub-sample time periods that is from 1994Q3 to 2005Q2. So we are not able to say that VAR produce a strictly rational forecast of CPI.


55

Table 6.3.5: Test of Strict Rationality

Panel A 1983Q1 1994Q2 Benchmark Forecast When Combined With

CPI VAR CPI ARIMA Sign Mean Error -ve -ve α 4.59 β 1.78 Prob. 0.00 Conclusion Bias Panel B

1994Q3 2005Q2 Benchmark Forecast When Combined With CPI VAR CPI ARIMA Sign Mean Error -ve -ve α -0.19 β 1.98 Prob. 0.12 Conclusion Cannot Reject

7. Conclusions and Policy Implications In this section we rank the alternative price indicators on the basis of performance test used in this study. Annual forecast of WPI obtained from ARIMA is not found to be consistent. On the other hand although the quarter forecast of WPI is not efficient but passes the tests of weak rationality and sufficient rationality, it is a surprising results, because in empirical analysis many forecasts are not able to pass these tests. Annual and quarter forecast of CPI from ARIMA passes all the tests of consistency, efficiency and test of weak and sufficient rationality.

Forecasts obtained from VAR shows same results about the forecast of WPI in the context if efficiency and rationality but it is consistent. Annual forecasts of WPI, GNPPD and DAPD are not pass the tests of weak rationality i.e., unbiasedness test and weak information efficiency test. We rank CPI is the best indicator of inflation from the forecasting point of view. Forecast of DAPD stays at the second number, to satisfying the tests of consistency, efficiency and rationality. Here we provide support to the observation of Ahmad and Ram (1991) that DAPD is a better indicator of inflation as compared to the other popular price indices. Forecasts of GNPPD are less reliable, but the forecast of WPI is least reliable according to the findings of this study.

Javed and Ahmad

56

So we can say that to get the best price forecast, the better specification is VAR models with quarterly data, and we suggest CPI and DAPD instead of GNPPD and WPI. For a VAR forecast, we rank WPI at number third, better than GNPPD, while form ARIMA forecasts WPI is least satisfactory price variable for forecasting point of view.

If we look at the construction procedure of the price indices like CPI and WPI in Pakistan, there are also some facts that support results of the study. For the construction of CPI, the price data are taken from the 71 markets of 35 cities of Pakistan. On the other hand, coverage of WPI is very low. The wholesale price data are collected from a single market of 18 cities each. The relatively poor forecasts of WPI compared with CPI suggest that efforts need to be made to make the WPI more representatives by improving the coverage in terms of markets, commodities and cities. There is also a need to improve the skills of price collecting staff, especially for those enumerators who collect the prices for the construction of WPI, so that the problem of low coverage may be covered. In this way the qualities of survey indicators can be improved with the improvement in the human capital that makes the survey data a clear picture of the economy.

Although econometric forecasting is not yet very common among policy makers and other agencies/institution, a movement in that direction is in the making. For example, the State Bank of Pakistan has gone through rigorous training programs on model building, econometrics and forecasting. If econometric forecasts are used for policy making, they should also be aware of limitations of the techniques. Our results show that in general more elaborate VAR models outperform the simplistic ARIMA models in forecasting a price series. Another useful conclusion is that the quarterly data provide better forecasts than the annual data. All these results support the econometricians’ maintained hypotheses that data observed at high frequency and statistically more elaborate use of a given data set provides better predictions than the data observed at low frequency and analyzed with simplistic statistical tools.

Another implication of our findings is that researchers and policy makers are likely to make better predictions and policy prescriptions if they base their analyses on the price indices that have broader coverage like the CPI as compared to WPI or the price deflator based on gross domestic absorption as compared to gross domestic product or gross national product.


57

References Aggarwal, R., Mohanty, S., & Song, F. (1995). Are survey forecasts of

macroeconomic variables rational? Journal of Business, 68(1), 99-119.

Ahmed, E., & Hari, R. (1991). Foreign price shocks and inflation in Pakistan. Pakistan Economic and Social Review, XXIX, 1-20.

Armstrong, J. S., & Fildes, R. (1995). On the selection of error measures for comparisons among forecasting methods. Journal of Forecasting, 14, 67-71.

Artis, M. J. (1996). How accurate are the IMF’s short-term forecasts? Another examination of the world economic outlook. International Monetary Fund, Working Paper No. 96/89.

Ash, J. C. K., Smyth, D. J., & Heravi, S. M. (1990). The accuracy of OECD forecasts of the international economy. International Journal of Forecasting, 6, 379-392.

Ash, J. C. K., Smyth, D. J., & Heravi, S. M. (1998). Are OECD forecasts rational and useful? A directional analysis. International Journal of Forecasting, 14, 381-391.

Bakhshi, H., George, K., & Anthony, Y. (2003). Rational expectations and fixed-event forecasts: An application to UK inflation. Bank of England, UK, Working Paper No. 176.

Batchelor, R. (2001). How useful are the forecasts of intergovernmental agencies? The OECD and IMF versus the consensus. Applied Economics, 33, 225-235.

Bates, J. M., & Granger, C. W. J. (1969). The combination of forecasts. Operations Research Quarterly, 20, 451-468.

Beach, W. W., Aaron, B. S., & Isabel, M. I. (1999). How Reliable are IMF Economic Forecasts?. A Report of the Heritage Center for Data Analysis, Washington, D.C.

Blix, M., Kent, F., & Fredrik, A. (2002). An evaluation of forecasts for the Swedish economy. Economic Review, 3, 39-74.

Bonham, C. S., & Douglas, C. D. (1991). In search of a “Strictly Rational” forecast. The Review of Economics and Statistics, 73(2), 245-253.

Javed and Ahmad

58

Bonham, C. S., & Cohen, R. (1995). Testing the rationality of price forecasts: Comment. The American Economic Review, 85, 284-289.

Box, G. E. P., & Jenkins, G. M. (1976). Time Series Analysis, Forecasting and Control. Holden-Day: San Francisco.

Congressional Budget Office, Congress of the United States, (1999). Evaluating CBO’s Record of Economic Forecasts.

Clemen, R. T. (1989). Combining forecasts: A review and annotated bibliography. International Journal of Forecasting, 5(4), 559-581.

Clement, M. P., & Hendry, D. F. (1993). On the limitation of comparing mean square forecast errors. Journal of Forecasting, 12, 617-637.

Clement, M. P., & Hendry, D. F. (1998). Forecasting Economic Time Series. Cambridge University Press, Cambridge.

Dickey, D. A. & Fuller, W. A. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 74, 427-431.

Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica, 49, 1057-1072.

Diebold, F. X., & Jose, A. L. (1996). Forecast Evaluation and Combination. Handbook of Statistics, Elsevier: Amsterdam.

Eichenbaum, M., Hansen, L. P., & Singleton, K. J. (1988). Testing restriction in non-linear rational expectations models. Quarterly Journal of Economics, CIII, 51-78.

Engle, R. F., & Granger, C. W. J. (1987). Co-integration and Error Correction: Representation, estimation and testing. Econometrica, 55, 251-276.

Evans, G. W., & Gulmani, R. (1984). Tests for rationality of the Carlson-Parkin inflation expectation data. Oxford Bulletin of Economics and Statistics, 46, 1-19.

Friedman, B. M. (1980). Survey evidence on the ‘Rationality’ of interest rate expectations. Journal of Monetary Economics, 6, 453-465.

Gavin, W. T., & Mandal, R. J. (2000). Forecast inflation and growth: do private forecasts match those of policymakers. Federal Reserve Bank of St. Louis, Working Paper No. 026A.


59

Granger, C. W. J. (1969). Investigating causal relations by econometric models and cross spectral methods. Econometrica, 35, 424-438.

Granger, C. W. J., & Newbold, P. (1973). Some comments on the evaluation of economic forecasts. Applied Economics, 5, 35-47.

Granger, C. W. J. (1981). Some properties of time series data and their use in econometric model specification. Journal of Econometrics, 16, 121-130.

Granger, C. W. J. (1989). Forecasting In Business and Economics. London: Academic Press.

Granger, C. W. J. (1996). Can we improve the perceived quality of economic forecast?. Journal of Applied Econometrics, 11(5), 455-473.

Government of Pakistan, Economic survey (various issues), Ministry of Finance, Islamabad.

Hafer, R. W., & Hein, S. E. (1985). On the accuracy of time series, interest rate, and survey forecast of inflation. Journal of Business, 5, 377-398.

Holden, K., Peel, D. A., & Sandhu, B. (1987). The accuracy of OECD forecasts. Empirical Economics, 12, 175-186.

Holden, K., & Peel, D. A. (1990). On testing for unbiasedness and efficiency of forecasts. Manchester School, 58, 120-127.

International Monetary Fund (2005). International Financial Statistics 2005, Washington, DC.

Keane, M. P., & Runkle, D. E. (1990). Testing the rationality of price forecasts: New evidence from panel data. American Economic Review, 80(4), 714-735.

Keane, M. P., & Runkle, D. E. (1994). Are economic forecast rational? Unpublished Manuscript, Federal Reserve Bank of Minneapolis.

Keane, M. P., & Runkle, D. E. (1995). Testing the rationality of price forecasts: Reply. American Economic Review, 85, 290.

Kemal, A. R., & Arby, M. F. (2004). Quarterisation of annual GDP of Pakistan. Statistical Paper Series No. 5, Pakistan Institute of Development Economics.

Keynes, J. M. (1936). The General Theory of Employment, Interest and Money. London: Macmillan.

Javed and Ahmad

60

Kreinin, M. E. (2000). Accuracy of OECD and IMF projection. Journal of Policy Modeling, 22, 61-79.

Lee, B. (1991). On the rationality of forecasts. The Review of Economics and Statistics, 73(2), 365-370.

Liu, P., & Maddala, G.S. (1992). Rationality of survey data and tests for market efficiency in the foreign exchange markets. Journal of International Money and Finance, 11, 366-381.

McNees, S. K. (1986). The accuracy of two Forecasting techniques: Some evidence and interpretations. New England Economics Review, April, 20-31.

Muth, J. F. (1961). Rational expectations and the theory of price movements. Econometrica, July, 313-335.

Nordhaus, W. D. (1987). Forecasting efficiency: Concepts and applications. The Review of Economics and Statistics, 69, 667-674.

Oller, L. E., & Barot, B. (2000). The accuracy of European growth and inflation forecasts. International Journal of Forecasting, 16, 293-315.

Paula, R. D. M. (1996). The difficult art of economic forecasting. Finance and Development, December, 29-31.

Pearce, D. K. (1987). Short-term inflation expectations: Evidence from a monthly survey. Journal of Money, Credit, and Banking, 19, 388-395.

Pigou, A. C. (1927). Industrial Fluctuation. London: Macmillan.

Pons, J. (1999). Evaluating the OECD’s forecasts for economic growth. Applied Economics, 31, 893-902.

Pons, J. (2000). The accuracy of IMF and OECD forecasts for G7 countries. Journal of Forecasting, 19, 56-63.

Pons, J. (2001). The rationality of price forecasts: A directional analysis. Applied Financial Economics, 11, 287-290.

Razzak, W. A. (1997). Testing the rationality of the National Bank of New Zealand’s survey data. National Bank of New Zealand, G97/5.

Rich, R. W. (1989). Testing the rationality of inflation from survey data: Another look at the SRC expected price change data. The Review of Economics and Statistics, 71(4), 682-686.


61

Ruey, S. T. (2000). Time series forecasting: Brief history and future research. Journal of the American Statistical Association, 95(450), 638-643.

Ruoss, E., & Marcel, S. (2002). How accurate are GDP forecast? An empirical study for Switzerland. Quarterly Bulletin, Swiss National Bank, Zurich, 3, 42-63.

Sims, C. A. (1980). Macroeconomics and reality. Econometrica, 48(1), 1-48.

State Bank of Pakistan, Annual Reports (various issues), Karachi, Pakistan.

Stock, J. H., & Mark, W. W. (2001). A comparison of linear and nonlinear univariate models for forecasting macroeconomic time series. Oxford University Press, Oxford, 1-44.

Timmermann, A. (2005). Forecast Combinations. Forthcoming in Handbook of Economic Forecasting, Amsterdam: North Holland.

Timmermann, A. (2006). An evaluation of the World Economic Outlook forecasts. International Monetary Fund, Working Paper No. 06/59.

World Bank (2006). World Development Indicators 2006, Washington, D.C.

Yin-Wong C., & Menzie, D. C. (1997). Are macroeconomic forecast informative? Cointegration evidence from the ASA-NBER surveys. National Bureau of Economic Research, Working Paper, 6926.

Yule, G. U. (1927). On a method of investigating periodicities in disturbed series with special reference to Wolfer’s Sunspot Numbers. Philosophical Transactions of the Royal Society London, Ser. A, 226, 267-298.

Zarnowitz, V. (1984). The accuracy of individual and group forecasts from Business Outlook Surveys. Journal of Forecasting, 3, 11-26.

Zarnowitz, V. (1985). Rational expectations and macroeconomic Forecasts. Journal of Business and Economic Statistics, 3, 293-311.


Impact of Trade Openness on Exports Growth, Imports Growth and Trade Balance of Pakistan

M. Aslam Chaudhary and Baber Amin 1

Abstract

This study mainly aims at analyzing the impact of trade openness on exports growth, imports growth and trade balance of Pakistan. Pakistan has undergone substantial trade openness measures during the last three decades. The main objective behind the openness and liberalization has been to reap the fruits of higher exports which contribute to higher economic growth. The study analyzes the data from 1980-2008. The OLS and Auto Regressive Distributive Lagged modeling approaches have been employed to find empirical support. The results of the study reveal that trade openness affected both exports growth and imports growth positively although the imports growth increased more than exports one, which worsened the trade balance. Nevertheless, trade openness played a limited role and remained constrained in promoting economic growth through exports expansion. Thus, there is a need to create a balance between exports and imports growth to reap the fruits of openness.

Keywords: Export growth; Import Growth; Trade Balance; Trade Openness; Pakistan

JEL classification: F41, F43, F15, F42

1. Introduction Trade liberalization as well as openness of economy is seen as driving force to accelerate economic growth. Of course, openness of borders for trade leads to reap the benefits of expanded demand for exports. For this reason, most of the countries, particularly the developing ones, introduced reforms to open up the foreign sector and also reformed the domestic economy too; since the last three decades.2 The international financial institutions such as

1 The authors are Professor and lecturer at Forman Christian College (A Charted University), Lahore and Lahore Leads University, Lahore, respectively. The paper is based on M. Phil thesis of Amin B. (2011) and Working paper of Chaudhary, M. A. (2010). They are thankful to Naeem Rashid for his valuable comments which helped to improve the paper. 2 The reforms process of opening up of the foreign sector started in late1980’s in Pakistan.

Chaudhary and Amin

64

WTO, World Bank and IMF also encouraged trade liberalization and openness. In addition to above, one of the main objectives behind the openness and liberalization has been to promote efficiency, competition and discourage distortions.3 The more barriers on trade we have, the lesser will be exports expansions. For a country like Pakistan, which introduced rapid economic reforms and ended up with expanding imports and meager impact on its exports expansion, the result is trade balance worsened.4 Thus, trade openness might have beneficial, as well as harmful, effects for a country. If trade openness leads towards higher exports and more efficient allocation of resources, it is beneficial and could potentially accelerate growth by ensuring needed foreign exchange and attracting foreign investment. Pakistan has not generated efficiency and competition at domestic level and relied heavily on imports which could turn out as worsening economic conditions. Pakistan is suffering from twin deficits i.e. trade deficit and domestic budget deficit.5 So, there is a need to analyze whether trade openness has really contributed to accelerate economic growth of Pakistan or not. Most of the researchers focused their research on the expansion of exports, due to openness; however little attention has been paid towards increasing growth rate of imports which ultimately could worsen and balance of trade. Furthermore, deficit in trade balance again reflects as foreign borrowing which further aggravates the problem of deficit. The important point to be noted here is that if trade liberalization increases the import growth more than export growth, as it happens in case of most of the developing countries, it might lead towards creating worse conditions for the country. It is a well known fact that most of the under developed countries are already suffering from foreign reserves shortages, deficit in trade and low foreign direct investment. In this environment, liberalization of foreign sector helps to improve economic conditions. There is a limited research on these issues, particularly in the case of Pakistan.

Given the above background, this study empirically analyzes the impact of trade liberalization on both export and import growth. Moreover, trade balance which was ignored is being analyzed to dig the roots of the problem.

3 The Market Friendly Approach also conveyed the process of market competition, international linkages which take place due to investment in human development. 4 See: Pakistan Economic Survey, 2011-12. 5 Pakistan’s budget deficit was as high as over 7% of GDP. It is even expected higher for the current year i.e. 2012-13. For further details see: Pakistan Economic Survey, 2011-12.

Impact of Trade Openness on Exports Growth, Imports Growth and Trade Balance

65

The rest of the study follows a certain pattern. Section II describes the performance of various variables in the post and pre-reform era regarding trade openness and liberalization6. Section III, presents literature review and the results of other important studies regarding the impact of trade openness and liberalization on export growth, import growth and trade balance. Theoretical background and model specification have been discussed in Section IV. Section V presents the results of empirical estimation. Conclusions and policy implications are provided in section VI.

2. Economic Performance of Major Variables in Pre and Post-Reform Era. Pakistan brought significant trade liberalization during the 1980s7. Table 1 shows the average growth rates of various important variables before liberalization i.e. pre-reform era 1980-1990 and post-reform era 1991-2008. The table clearly indicates poor performance of all the indicators in the post-reform era. The average GDP per capita growth rate was 3.2% in the pre-reform era while it reduced to 1.9% in the post-reform period. The average real GDP growth was 6.3% in the pre-reform period, while it reduced to 4.36% in the post-reform era. In line with the pattern of the above-described indicators, real exports slowed down in the post-reform period from 9.4% to 6.8% in the pre-reform period. In contrast to the above-discussed indicators, average import growth increased from 4.37% to 5.28%; after liberalization. The trade reforms thereof increased the average import growth while decreased the export growth of Pakistan’s economy. Besides, the average growth rate of trade deficit was minus 1.9% in the pre-reform period while it increased to 26.8% in the post-reform period which indeed is a significant increase in the trade deficit. The economic performance improved earlier during the first decade of 2000’s and again deteriorated thereafter. It may be noted that economic growth is still around 2.5%8. On the basis of above discussion therefore, it may be inferred that trade reforms affected the economic growth of Pakistan adversely.

6 In Pakistan, not only liberalization of the foreign sector took place but there were substantial reforms to improve domestic economy such as privatization of the financial market etc. 7 See: Chaudhary M. A. (2004); Globalization: WTO, Trade and Economic Liberalization in Pakistan. 8 See: for details, Pakistan Economic Survey, 2011-12.

Chaudhary and Amin

66

Table 1: Performance of Various Variables in Pre and Post-Reform Era

Variable Pre-liberalization Post-liberalization GDP per capita 3.19 1.89 Real GDP 6.26 4.40 Real Exports 9.38 6.85

Real Imports 4.37 5.28

Trade Deficit -1.85 26.8

Source: Pakistan Economic Survey, GOP (various Issues)

3. Review of Literature

After the Washington consensus and emergence of WTO, the world has been witnessing a continuous debate on the nexus between trade liberalization and economic growth. It still remains to be seen if trade liberalization and openness is growth promoting. If yes, we need to see the channels through which it affects economic progress? There is an ample literature available on the nexus between openness and export growth and export led growth but very little attention has been paid on impact of trade openness on import growth and trade balance. A brief literature review on the issue is presented below:

Sherazi and Abdul Manap (2004) tested the ongoing issue that exports growth enhanced economic growth. They also found that there is a feedback impact on imports. However they have not tested this feedback impact. Our study is focused to contribute to the literature in this context, which is neglected so far.

Faini et al. (1992) analyzed the effects of trade policy on demand of imports in developing countries. The study divided the imports into two categories: under the quantitative restrictions and those which are freely movable among countries. The results show that income elasticity was greater than one among developing countries while the relative prices were proved to be significant having elasticity less than one. The other important finding of the study has been that the shortage of foreign exchange or when we have restrictions on import flows then the estimated effects of income and price elasticity becomes less prominent as compared with liberalized or more open trade regime where this impact is prominent. The study suggests that while interpreting the income and price elasticities in import demand studies, the


67

type of trade regime should be given special attention. It is the nature of trade and goods which contribute to gains from the trade.

Santos-Paulino (2002) analyzed import demand function for twenty two developing countries with special reference to their trade policy reforms, particularly liberalization of trade. He utilized panel data, Fixed Effect (FE) and Generalized Method of Moments (GMM), to draw empirical evidences. The study presented estimates at both regional and panel level. The main objective of this study was to observe the impact of trade liberalization reforms on imports in developing countries. This study also used the “Heritage Foundation Index of Economic Freedom” to categorize the countries from very low to very high level of protection of traded goods. Heritage Foundation Index of Economic Freedom classified countries into five classes. In the estimation of fixed effect model, country specific dummy was also used which takes into account the country specific factors and environment. The dynamic panel data estimation is done through FE and GMM method while the time series cross section analysis is based upon Two Stages Least Square (2SLS) and Maximum Likelihood Method (MLM). The results of the fixed effect model showed that all the variables had the sign according to the prediction of economic theory and all the variables were significant except the relative price indicator. The results also revealed that short and long run price and income elasticities are same. However, the variables of trade policy regime and import duty are statistically significant. The study also showed that trade liberalization enhanced 100% in the imports volume. The fixed effect estimates support the Melo and Vogt (1984) hypothesis.9 Thus, based upon the study of Santos-Paulino (2002), it can be stated that the affects of import duties vary from one region to the other region while we do not have stable and consistent results for all the regions of the world. Similarly, Income and price elasticities also differ among regions. Due to 100% increase in imports after liberalization, the study suggested important policy measures regarding the export promotion and current account deficit problem of developing countries. However, there was not significant increase in exports. The study suggested that liberalization should be carried out along with export promotion strategy so that countries should not face the severe problem of balance of payment which may reduce the fruits of liberalization in terms of higher growth.

9 For more details on this hypothesis see: Melo, O., & Vogt, M. G. (1984) and Yanikhra, (2003).

Chaudhary and Amin

68

In spite of using appropriate techniques of panel data estimation, they missed important determinants of import growth in its function; like industrial growth, exchange rate regime, law and order situation, incentives for investors, institutional development and domestic environment etc., which may have affected his results. Moreover, the study has found different results for aggregate sample and regions. So, it is difficult for any country to fully adopt the same policies based on these results because individual country results might get different outcomes from the regional results. Thus a time series comprehensive study for individual countries is also needed to provide additional evidences for sound policy suggestions regarding liberalization and export growth.

Santos-Paulino and Thrilwall (2004) studied and utilized different measures for liberalization and openness. Their focus was on the impact of trade liberalization on exports, imports and balance of payments problems of developing countries. The study used the data set from twenty two developing countries which brought significant changes and introduced liberalization in the 1970s.They used two types of measures of liberalization which are: (a) import and export duties (b) the dummy variable used for the year of liberalization selected on the basis of world trade organization (WTO) and World Bank’s criteria. The study used the Fixed Effect (FE) and Generalized Method of Moments (GMM) for analyzing panel data for developing countries using time series /cross sectional study for different regions of the world.

The study analyzed and compared the impact of trade liberalization on exports and imports growth for major developing countries. Besides, the impact of liberalization on prices, income elasticity of demand for exports and imports had also been estimated. Further, the impact of liberalization on balance of payment and trade balance was highlighted.

The study supported the notion that trade liberalization enhanced both exports and imports but the increase in former was greater than that of the later, which worsened the problem of trade deficit. It is well known that most of the developing countries have already been facing the problem of shortage of foreign exchange reserves. Liberalization has therefore very important policy implications for these because it may lead to growth below the potential level. The results of the study also pointed out that import and export duties have negative impact on import and export growth. The study concluded that ten percentage point decrease in duties leads to 2% growth of


69

exports, while import growth increased between 2 to 4 %. Moreover, liberalization increased the elasticity of demand for imports more than exports. Thus the developing countries have to be careful in terms of liberalization while remaining ready to handle balance of payment problem. They need not become a victim of foreign exchange shortage. If it happens so, a country may end up pilling up collusive amount of foreign debt10. This trend has already been set for the Pakistan economy. Pakistan has not borrowed as much as, in the first fifty five years since its birth, which it has borrowed in the last five years.

Keeping in mind the outcomes of the above cited literature, our study aims at exploring the important features and impact of liberalization, particularly the impact of trade liberalization on exports, imports and balance of payment. For this purpose, a model has been developed to draw empirical evidences. The model is discussed below.

4. Theoretical Background and Model Specification

4.1. Trade Liberalization and Exports Growth The demand for exports depends mainly upon relative prices and world demand for exports (s). By keeping the price and income elasticity constant and following Santos-Paulino and Thrilwall (2004), the export demand function can be written as:

.

Where ‘X’ is the exports at time period t, pd/pf is the ratio of domestic to foreign price in the same currency units, ‘W’ denotes the world income The value of β2 indicates income elasticity of demand for exports, while β1 is the price elasticity of demand for exports. After taking the logs and differentiating with respect to time, the above equation may be written as:

Now by adding trade openness variable (top), the equation becomes:

10 For other bottlenecks and trade contributions see Kroger (1978).

Chaudhary and Amin

70

Finally, by introducing dummy (lib) variable to capture the effect of the year of liberalization as taken by Lopez (2003), Santos-Paulino and Thrilwall (2004), the above equation will be:

Here Xt is the export growth, pxt is the growth rate of relative price change, topt

/ is the growth rate of trade openness and libt is the liberalization dummy which considers the year 1991 as liberalizing year, as commonly utilized in the literature.11

4.2. Trade Liberalization and Imports Most of the literature has focused on trade liberalization and export-led growth12. However, there is limited body of literature which explored the nexus of liberalization - import -growth phenomenon. Trade liberalization may increase the growth of imports much more than growth of exports which could create a problem of balance of payment deficit as well as that of shortage of foreign exchange which may squeeze economic growth. Therefore, it is also equally important to analyze the impact of trade liberalization and openness on import growth.13 In order to analyze the impact of trade liberalization on imports and economic growth by following Santos-Paulino and Thrilwall (2004), the given equation is derived in the same way as for exports growth model stated above. So the above described equation becomes as following which is utilized to analyze the impact of trade liberalization on import growth;

.

After taking the log, differentiating it with respect to time, and augmenting the variable of trade openness, the above equation becomes as follow:

11 See: Santos-Paulino and Thrilwall (2004) 12 For details see: Balassa (1985), Ram (1987). 13 Santos-Paulino and Thrilwall (2004) studied the impact of trade liberalization on export, import and trade balance growth in developing countries and proved that trade liberalization increased the import growth more than export growth which created the balance of payment problem too.


71

Now, after adding the dummy variable, the above equation may be written as: Where we have Mt as the growth rate of imports, pmt the growth of import price relative to domestic substitutes; yt the growth of domestic income, topt the growth of trade openness, libt the dummy for the liberalization year i.e. 1991 in case of Pakistan.

4.3 Trade Liberalization and Current Account The current account provides a good picture of a country’s position regarding foreign exchange and foreign reserves. Thus, taking the difference between exports and imports, as trade balance provides performance of trade liberalization. To capture such impact this study will estimate the following equation which is taken from Santos-Paulino and Thrilwall (2004):

Where ‘W’ is the world income, Y is the domestic income, P is real exchange rate, TOP is the trade openness and TOT is the terms of trade.

5. Empirical Estimation and Interpretation of Results

5.1. Empirical Evidences: Trade Liberalization and Exports Growth Two models discussed in the previous section were estimated by using the OLS method. All the variables have been taken in growth rates and were found stationary at level form. The results are presented in the following table 2. The shows that growth rates of all the variables are I(0) , so OLS can be applied for empirical results. Table 3 shows the results of OLS regarding the impact of trade liberalization on export growth.

The results of the regression analysis (table 3) show that trade openness have significant and positive relationship with exports growth. The results also reveal that 1% increase in trade openness leads to 1.06% increase in exports growth while the world income growth and relative price change variables remain insignificant.

By adding dummy variable to the model for capturing the affects of liberalization, the following equation is estimated:

Chaudhary and Amin

72

Table 2: Results of the Unit Root Tests

Augmented Dickey Fuller Philips Perron Result Variable Level Level

Intercept Trend & Intercept

Intercept Trend & Intercept

GX -5.40708* -5.3058* -5.432890* -5.306255* I(0)

GRER -4.710111* -4.572675* 3.286711* -3.134843*** I(0)

GTOP -4.903515* -4.820744* -4.888834* -4.798468* I(0)

GW -3.173418* -3.444151 -3.076926* -3.385888 I(0)

GY -3.278274* -3.467158** -3.254017* -3.467158** I(0)

Note: *, ** and *** show level of significance at 1%, 5% and 10 %, respectively.

The results of the above equation are provided in appendix I. The results indicate that trade liberalization (openness) has significant and positive impact on export growth. The variable is significant at 1% level of significance. The above results reveal that one percent increase in trade openness led to 1.17% increase in export growth. The world income growth and liberalization dummy are also found significant. Interestingly, the sign of liberalization dummy is negative but it is logical since after introducing the trade reform policies and becoming liberalized, the openness squeezed exports growth. The results are consistent with Santos-Paulino and Thrill wall (2004).

5.2 Empirical Evidences: Trade Liberalization and Imports Growth In order to analyze the impact of trade liberalization (trade openness) on import growth, the given equation is estimated14.

The results of the unit root test reveal that all the variables are integrated

at I(0). So, OLS can be applied and the results are given Table 2.

The results of the OLS have been given in Table 4 which shows that the variable of trade openness is significant at 1% level of significance with positive sign. It suggests that one percent increase in trade openness could

14 See Chapter 4, Amin B. (2011).


73

Table 3: Impact of Trade Liberalization on Exports Growth

(Dependent Variable: Growth of Real Exports)

Variable Coefficient Std. Error t-Statistic Prob.

C 6.740863 7.358236 0.916098 0.3700

GTOP 1.058946* 0.267205 3.963043 0.0007

GW 0.344935 1.956751 0.176280 0.8618

GPX -0.134371 0.160071 -0.839446 0.4107

AR(1) 0.700776* 0.164761 4.253287 0.0004

MA(1) -0.997480* 0.107147 -9.309462 0.0000

R-squared 0.557946

Adjusted R-squared 0.452695

F-statistic 5.301109

Prob.(F-statistic) 0.002646

Durbin-Watson stat 2.313301

Note: * indicates significant at 1% level of significance

lead to almost 1.2% increase in import growth. The import growth is positively related to trade liberalization.

Now by adding the dummy variable for capturing the affects of the year of liberalization, the following equation is estimated.

Where libtt represents liberalization dummy. The results of the above regression are given in appendix II. The results of the regression analysis show that the variable of trade openness is still highly significant with positive sign along with the co-efficient almost equal to one. The value of adjusted R2 is 0.81, while the value of DW is 1.8. The liberalization dummy is also found significant with positive sign. Both the variables of trade openness and liberalization dummy are significant at 1% level of significance, respectively.

Chaudhary and Amin

74

Thus trade liberalization is positively and significantly contributing to imports growth. The liberalization dummy has 3.84 co-efficient which shows that 1% increase in trade openness leads to 3.84% increase of imports. However, it may be noted that positive association of trade liberalization and import growth may not be very healthy for the economy. The increasing imports and squeezing exports potentially create a serious problem of trade deficit which Pakistan is being faced by Pakistan.

Table 4: Impact of Trade Liberalization on Imports Growth

(Dependent Variable: Growth of Real Imports)

Variable Coefficient Std. Error t-Statistic Prob. C -2.538340 3.285175 -0.772665 0.4476 Y 1.028560** 0.503367 2.043360 0.0526 PM 0.103221 0.114265 0.903345 0.3757

TOP 1.199407* 0.161098 7.445208 0.0000

MA(1) -0.574645* 0.187534 -3.064215 0.0055 F-statistic 15.15 Prob. (F-statistic) 0.01 D.W. stat. 1.78 R-squared Adjusted 0.73

Note: *, ** indicate significant at 1% and 5%, respectively.

5.3. Impact of Trade Liberalization on Trade Balance The following equation has been estimated in order to analyze the impact of trade liberalization on trade balance following Santos-Paulino and Thrilwall (2004).

First unit root tests have been conducted in order to determine the order of integration of the variables. It helps to decide about the technique of estimation.

The results of Table 5 indicate that some variables are I (0) while the others are I(1) . In these circumstances econometric theory suggests that Bounds Procedure and ARDL approach seem appropriate for determining long and short run dynamics as described by Pesaran and Shin (1996, 1999


75

and 2001). The empirical estimations of long run and short run dynamics are analyzed in the following section.

5.3.1. Estimation of ARDL Model After analyzing the order of integration of the variables, the following error correction version of the ARDL model has been used in order to determine the short run and long run dynamics of the relationship among the variables.

tt

t

ttttit

n

iit

n

i

it

n

iit

n

iit

n

iit

n

i

GDPTBLW

LRERLTOTLTOPLYLYLW

RERLTOTLTOPGDPTB

GDPTB

ξββ

ββββαα

ααααα

+⎟⎠⎞

⎜⎝⎛++

++++∆∑+∆∑+

∆∑+∆∑+∆∑+⎟⎠⎞

⎜⎝⎛∆∑+=⎟

⎠⎞

⎜⎝⎛∆

−−

−−−−==

====

1615

14131211_60

_50

_40

_30

_20_

11

0

Where the parameters of ‘’α” show short run while of ‘’β “show long run co efficient in the above equation.

Table 5: Results of the Unit Root Tests Augmented Dickey Fuller Test Philips Perron Test Level 1st Difference Level 1st Difference Variable Trend Trend &

intercept Trend Trend &

Intercept Trend Trend &

intercept Trend Trend &

intercept Result

X 0.7843 2.1526 6.0841* 5.9954* 0.7843 2.1988 6.0846* 5.9954* I(1) M 1.0017 2.8549 5.4257* 5.3353* 0.8943 2.9114 5.6673* 5.6542** I(1) TOP 3.75** 4.000** 5.1769* 5.0452* 3.09** 2.8427 5.2750* 5.1265* I(0) TOT 1.4210 0.3873 4.7451* 4.9103* 1.4033 0.3502 4.7350* 5.0892* I(1) TB/GDP 4.586* 4.4383 .1511 .8247 9.863* 10.2900* 4.5865* 4.4383* I(0) RER 1.7119 0.6268 4.9808* 6.1885* 1.6933 0.6709 4.9863* 7.5889* I(1) W 1.4574 2.98879 3.306** 3.25*** 0.9576 1.9568 5.4444* 4.2314** I(1) Y 2.3383 2.0533 3.227** 3.46*** 1.9726 2.2736 3.2479* 3.4635** I(1)

Note: *, ** and *** indicate significant at 1%, 5% and 10%, respectively.

5.3.2. Estimation of the F-Statistics The joint significance test is applied to determine the existence of the long run relationship among variables and then compared with the critical bound values.15 The results show that calculated F- Stat is 5.18, which is greater than the critical Bound values which are (3.23- 4.35) for 5% level of 15 For details see Pesaran & Shin (1996, 1999 and 2001).

Chaudhary and Amin

76

significance. So, we reject the null hypothesis of no co-integration at 5% level of significance for the above model. The results of the estimated long run elasticities are reported in table 6, given below:

Table 6: Estimated Long Run Elasticities Using the ARDL Approach

[ARDL (1, 0, 0, 1, 0, 0) Selected Based On Schwarz Bayesian Criterion]

Dependent variable is TB/GDP Regressors Coefficient Standard Error T-Ratio Probability

C 32.6324** 14.8950 2.1908 .038 LTOT .094994 .14383 .66045 .515 LTOP -.59178* .17621 -3.3584 .003 LRER -.31419*** .17772 -1.7679 .090

LW 3.0325* 1.1545 2.6267 .015 LY .66488* .21938 3.0306 .006

Note: *, **, *** indicate significant at 1%, 5% and 10%, respectively

5.3.3. Error Correction Representation for Selected ARDL Table 7 shows the results of error correction representation (ARDL) model of the impact of trade liberalization on trade balance.

The results of both the short run and long run elasticities of ARDL (table 6 and table 7) show that trade openness is significantly and negatively related to trade balance. The variable is significant at 1% level of significance. The results reveal that trade openness leads to the worsening of the trade balance which means an increase in trade deficit. It may be noted that the finding is in line with the previous findings that trade liberalization increased the import growth more than export growth implying that it has negative impact on trade balance. However, the variable of real exchange rate remains insignificant while the world income growth has significantly positive impacts on trade balance because it positively and significantly affects the exports of Pakistan. The country’s income growth is negatively related to trade balance. The variable is significant at 1% level of significance. It is quite logical to have negative sign with it because we found in our previous analysis that the domestic income growth leads to increased imports growth.

The results of the long run analysis show that 1% increase in trade openness leads to 0.59% reduction in trade balance. The results of the error correction representation show that the adjustment parameter is highly


77

significant at 1% level of significance with negative sign which is according to theory. The co-efficient of the error correction term suggests that 67% of the error will be adjusted in the first time period. It shows relatively fast speed of adjustment. It also means that 67 % of the disequilibrium caused by the previous period shocks will converge back to the equilibrium.

Table 7: Error Correction Representation for the Selected ARDL Model [ARDL (1,0,0,1,0,0) Selected Based On Schwarz Bayesian Criterion]

Dependent variable is ∆TBGDP

Regressors Coefficient Standard Error T-Ratio Prob. ∆C 22.1739** 9.7532 2.2735 0.032

∆LTOT 0.064549 0.096428 .66940 0.509 ∆LTOP -0.40212* 0.098787 -4.0705 0.000 ∆LRER -0.0093804 0.090250 -0.10394 0.918 ∆LW 2.0606* 0.76359 -2.6986 0.012 ∆LY -0.45179* 0.16608 -2.7203 0.012

Ecm (-1) -0.67951* 0.13226 -5.1378 0.000 Note: *, ** indicate significant at 1% and 5%, respectively.

6. Conclusions and Policy Implications The main objective of the study was to analyze the impact of trade openness on export and imports growth. Moreover, trade balance has also been analyzed by highlighting its determinants. As per our knowledge, such analysis has been ignored in the previous literature. In other words, the issue of deterioration in trade balance was the ultimate prime focus of this study. For this purpose, the study analyzed the data form 1980 to 2008. The OLS and ARDL approaches were applied to draw empirical investigations.

Most of the previous studies analyzed the impact of trade liberalization on the performance of economic growth, exports, inequality and income distribution etc. Hardly any study has analyzed the above-cited issue. The impact of trade liberalization on trade balance and imports growth is very important for a developing country like Pakistan. The general notion of liberalization of trade, accelerating exports and bringing improvement to trade balance may not be true for all. The liberalization may increase greater growth of imports than exports and ultimately it might have serious effects on country’s balance of payments. It may increase deficit of trade balance, which

Chaudhary and Amin

78

will affect foreign exchange reserves, foreign exchange rate and ultimately economic growth unless the balance between imports and exports is maintained. As a result, ultimately the economic growth is hampered.

The results of the study suggest that increase in trade openness and liberalization has significant positive impact on the growth of imports and exports where this influence on imports is greater. The results of the analysis also show that exports growth is greater in the pre- reform era than the post- reform era while the situation was vice versa for the imports growth. The results of the study also revealed that trade openness and liberalization worsened the trade balance.

The above cited findings have important bearings for policy formulation. There is a need to review trade liberalization policy since it has worsened the balance of payments. The increasing imports, more than exports, could create further serious bottleneck for the economy. The trade deficit is already on the verge of increase and it can pose a serious problem, if appropriate measures are not taken. Pakistan must improve its exports and also cut on imports to improve trade balance. There is also a need to review trade openness policy and take additional necessary steps to reap the benefits of trade liberalization.


79

References

Amin, B. (2011). Trade liberalization, competitiveness & economic growth: Empirical evidence from Pakistan. M. Phil thesis, Department of Economics F C College (A Chartered University) Lahore, Pakistan.

Balassa, B. (1965). Trade liberalization and revealed comparative advantage. The Manchester School, 33( 2), 327-345.

Balassa, B. (1985). Exports, policy choices, and economic growth in developing countries after the 1973 oil shock. Journal of Development Economics, 18, 23-35.

Chaudhary, M. A. (2004). WTO and economic reforms in: Globalization, WTO, trade and economic liberalization in Pakistan. Lahore, Ferozsons.

Chaudhary, M. A. (2010). Trade liberalization, reforms and their failures: A case study of Pakistan. Working Paper, Department of Economics, F C College (A Chartered University) Lahore.

Dickey, D. A., & Fuller, W. A. (1979). Distributions of the estimators for autoregressive time-series with a unit root." Journal of the American Statistical Association, 75, 427-831.

Faini, R., Lant, P., & Fernando, C. (1992). Import demand in developing countries. In M. G. Dagenis and P.-A. Juet (Eds), International trade modeling, international studies in economic modeling , 11, 279-297.

Govt. of Pakistan. Pakistan economic survey. Various issues. Economic Adviser’s Wing, Finance Division, Islamabad.

Kemal A. R. (2000). A plan to foster regional trade cooperation in South Asia. A study for SANI-Project. Pakistan Institute of Development Economics, Islamabad.

Kroger, A. O. (1978). Foreign trade regime and economic development: Liberalization attempts and consequences. Cambridge, MA Ballinger Pub. Co. NBER.

Melo, O., & Vogt, M. G. (1984). Determinants of the demand for imports of Venezuela. Journal of Development Economics, 14(3), 351-358.

Chaudhary and Amin

80

Pacheco-Lopez, P. (2005). The impact of trade liberalization on exports, imports, the balance of payments and growth: The case of Mexico. Journal of Post Keynesian Economics , 595-619.

Pakistan Economic Survey. Govt. of Pakistan. Various Issues. Economic Adviser’s Wing, Finance Division, Islamabad.

Perish, R. (1950). The economic development of Latin America and its principal problems. United Nations Commission for Latin America, New York.

Pesaran, M. H., Shin, Y. & Smith, J. R. (1996). Testing for the existence of a long-run relationship. DAE Working Papers 9622, Department of Applied Economics, University of Cambridge.

Pesaran, M. H., Shin, Y., & Smith, J. R. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289-326.

Ram, R. (1985). Exports and economic growth: Some additional evidences. Journal of Economic development and Cultural Change, 36, 415-25.

Santos-Paulino, A. U. (2002). The effects of trade liberalization on imports in selected developing countries. World Development, 30(6), 959-974.

Santos-Paulino, A., & Thrilwall, A., P. (2004). The impact of trade liberalization on exports, imports and the balance of payments of developing countries. The Economic Journal, 114(493) , F50-F72.

Shirazi, N. S., & Abdul Manap, T. A. (2004). Exports and economic growth nexus: The case of Pakistan. The Pakistan Development Review, 43(4), 563-581.

Yanikhya, H. (2003). Exports and economic growth: A cross country empirical evidence. Journal of Development Economics.


81

Appendix I

Impact of trade liberalization on export growth

Dependent Variable: Growth of Real Exports (GRX) Variable Coefficient Std. Error t-Statistic Prob. C 4.433517 8.317090 0.533061 0.5993 TOP 1.167103* 0.268145 4.352509 0.0003 W 3.378534*** 1.863828 1.812685 0.0836 PX 0.705932 0.427948 1.649573 0.1132 LIBD -8.306201** 3.264085 -2.544726 0.0185 MA(2) -0.942631* 0.034193 -27.56768 0.0000 R-squared 0.582761 Adjusted R-squared 0.487934 F-statistic 6.145515 Prob(F-statistic) 0.001045 Durbin-Watson stat 1.872891


Appendix II

Impact of trade liberalization on import growth

Dependent Variable: Growth of Real Imports Variable Coefficient Std. Error t-Statistic Prob. C -9.023977** 4.249876 -2.123351 0.0452 GTOP 0.984337* 0.179387 5.487229 0.0000 GY 1.863871*** 0.617836* 3.016772 0.0063 GPM 0.091508 0.074030 1.236087 0.2295 LIBD 3.843301** 1.383668** 2.777617 0.0110 MA(1) -0.997458* 0.111083* -8.979418 0.0000 R-squared 0.817887 Adjusted R-squared 0.776498 F-statistic 19.76088 Prob(F-statistic) 0.000000 Durbin-Watson stat 1.758968



Determinants of Youth Activities in Pakistan Rizwan Ahmad and Ijaz Hussain1

Abstract

This paper analyzes the youth labour market activities in Pakistan. Based on micro data of Labour Force Survey (2006-07), the strength of analysis presented in the paper is twofold. First, it highlights some issues of youth in labour market, their attitude towards work and education in Pakistan and second, the econometric analysis investigates the supply side determinants of youth activities in Pakistan. Our descriptive analysis shows that a substantial percentage of youth is neither in labour force nor enrolled as student which shows the wastage of human resources in the society. Moreover, higher unemployment among educated youth, poor level of education and skills, predominance of informal economy are some of the major issues of youth labour market in Pakistan. Results of multinomial logit model show that being a female reduces the chances of full time work and full time students in Pakistan. In general, young people with educated parents are more likely to enroll in education, while those whose parents are working in agriculture or informal sector are more likely to be full time workers. Similarly, responsibilities within household as a head or having more number of siblings also increase the economic participation of youth.

Keywords: Labor market; Youth activities; Education; Pakistan

JEL classification: J16, J21, J28

1. Introduction Since 1950s, the world’s population has gone through some major

changes. Some countries are facing the problem of ageing population while some are having a larger share of youth2population. Pakistan is also one of the countries which will have larger share of youth population in future. According to United Nations Population Projections; by the year 2050, there will be 50 million young people in Pakistan3. These young people can be

1 The authors are Assistant Professors at Department of Economics, Forman Christian College (A Chartered University), Lahore and Department of Economics, Gomal University, D. I. Khan, respectively. 2 In Pakistan, Youth constitutes a group of 15-24 years of age people. 3 World Population Prospects: The 2006 Revision.

Ahmad and Hussain

84

considering as an asset to produce demographic dividend4 for the society if proper education and economic opportunities are provided to them. However, unfortunately in Pakistan, these young people are facing number of challenges and difficulties in their way to education and work (Nayab, 2008; Ahmad & Azim, 2010). Early start of career, higher rate of unemployment, lack of educational and vocational opportunities are some of the issues of youth in Pakistan. On one hand, many young people start their career early which can adversely affect their earnings later in life5 while on other hand, a substantial percentage of youth especially females are not engaged in any economic activity6. Usual status7 unemployment rate (22%) for youth in Pakistan is almost three times higher than their official unemployment rate (7.5%). Moreover, a substantial percentage (30.5 %) of youth in Pakistan is neither in school nor in labour force. Only 27 percent of young people (22.1 percent female and 31.7 percent of male youth) are currently enrolled as a student in the age when they are supposed to complete their education. About 14.2 percent of employed youth work for less than 35 hours a week. Majority of them (about 94.4 percent) are not available for additional work and those who are available for additional work do not take serious measures to find work. Only 2.6 percent of them are actively seeking for additional or alternative work. Some of these figures are summarized in table 1.

All these facts and figures show that a substantial percentage of our population will consist of young people in future and majority of them are either sitting idle or facing difficulties in labour market. The opportunity cost of sitting idle of these people can be very high and there is a need to analyze what factors determine the outcomes of youth activities in Pakistan. This paper is an attempt to analyze the supply side determinants of activities of youth in Pakistan. For this purpose, we divide the activities of youth in four categories8, i.e., those who are full time student, full time worker, combine work with school and neither working nor enrolled as student. These 4 Different researchers have already claimed that countries with larger share of youth and working age population may experience a boost in economic growth which is termed as ‘demographic dividend’. See for example, studies by Bloom et al., 2001; Lee et al., 2006. 5 Emerson and Souza, 2006; Faizunnisa, 2005. 6 Durrant, 2000. 7Official unemployment rate is based on one week reference period while usual status approach uses last twelve months status as reference period. For details about usual status approach, see Ahmad (2010). 8 We follow the methodology used by Burki and Tazeen (1999) to divide youth activities in four mutually exclusive categories.

Determinants of Youth Activities in Pakistan

85

outcomes are then regressed on their personal, household and regional characteristics to find out the factors affecting each outcome. Complete descriptions of dependent and independent variables are given in table 4 while remaining paper is organized as follows.

Table 1: Activities of youth (15-24 years)

Full time student 21.8% Full time work 36.1% Combine work with school 1.2% Out of school and out of labor force 30.5% Labour force participation rate (LFPR) 47.6% Unemployment rate a) Official b) Usual status

7.5% 21.7%

Vulnerable employment9 (as percentage of total employment)

52.1%

Source: Calculated from LFS (2006-07)

Section two briefly presents some issues faced by youth in labour market, section three gives a brief overview of literature survey and section four discusses the data source and methodology for empirical analysis. Results of empirical analysis are presented in section five while conclusions and recommendations are discussed in section six.

2. Issues of Youth Labour Market in Pakistan Labour market in Pakistan is confronted with number of challenges. Some of critical issues related to the youth labour are discussed below.

2.1. Poor Level of Education and Skills

Education plays an important role in the development of a country. It raises the productivity and efficiency of individuals in the labour market. Unfortunately, Pakistan is lagging behind in the field of education and skills. About one-third of the youth population, a total of 10.4 million (3.7 million males and 6.7 million females) are illiterate (GOP, 2008). Statistics of educational attainment of youth also do not present a good picture of situation. In the year 2006, more than half of the youth labour force (62.2 per cent) had

9 Unpaid family helpers and own account workers are considered as vulnerable in labour market, for detail see Pakistan Employment Trends for Youth ,2008.

either lesimagine teducation

Figu

Sour

Technicaprocess oinvestmemore moone yearincome VocationAccordinTechnicastudents institutiodevelopeare also n2005, the

10 Cited in

ss than one ythe skills ann.===

ure 1: Educa

rce: Calculated

al and Vocaof employm

ent in educatobile and flexr increase inof individua

nal Educationg to Econoal and Vocat

after matrns as comp

ed countries1

not very gooe performanc

UNESCO, 200

A

year or just pd productivi

ational Atta

d from LFS, 20

ational Instiment generatition and trainxible. A studn technical eals in Paki

onal Institutiomic Surveytional Instituriculation apared to 8 p

0. Moreoverod. Accordince of 28 per

06

Ahmad and Huss

86

primary eduity of the lab

ainment of t

06-07

itutions alsoion especialning are verdy by Nasir aeducation restan. Howeions is not y of Pakistautions in Pakare enrolledpercent in dr, the qualityng to Asian Drcent of Voc

ain

ucation (Figubour force w

he Youth L

o play an illy for younry high for yand Nazli (2esulted in 2.ver, the stavery satisfa

an (2008-09)kistan (Tabled in Techndeveloping ay and structuDevelopmencational and

ure 1). It is nwith such a lo

Labour Forc

important rong people. Ryouth becaus2005) has sho.4 percent inate of Techactory in the), there are e 2). Only 1

nical and Vand 18 percure of these innt Bank Surv

Technical In

not hard to ow level of

ce (%)

ole in the Returns on se they are own that a ncrease in

hnical and e country. just 1522

1.6 percent Vocational ent in the nstitutions

vey in year nstitutions


87

in Pakistan was poor, 60 percent was fair and only 12 percent of institutions’ performance was ranked as good11.

Table 2: Trends in Technical and Vocational Education in Pakistan

Indicators 1995-96 2006-07 Number of Institutions 577 1522 Total Enrollment 86000 314188 Percentage enrollment (Technical and Vocational education)

0.56 % 1.66 %

Source: GOP, 2008a. Economic Survey of Pakistan, 2007-08 and HDR, 2007

2.2. Higher Unemployment12 among Educated Youth in Pakistan There is an incidence of higher unemployment among educated people in Pakistan which shows the mismatch between type of education and opportunities available in labour market. It is clear from figure 2 that unemployment rate among those with higher level of education is much greater than those with lower level of education. It does not mean that education is not good for labour market success in Pakistan. The main reason is that in general, tertiary education in Pakistan is not providing required skill for jobs in the labour market. Students generally do not have any practical

Figure 3: Unemployment on the basis of Education Level in Pakistan

(Youth, 2006-07)

Source: Based on GOP, 2008. Pakistan Employment Trends

11 Cited in HDR, 2007 12 Figures presented in this section are based on weekly status approach and are official figures.

05

101520

Une

mpl

oym

ent R

ate

(%)

Education Level

Ahmad and Hussain

88

knowledge and skills during their education in schools. After education, they usually demand comparatively higher wages as compared to people with low level of education which results in the higher level of unemployment for them.

2.3. Predominance of Informal Economy

In Pakistan, informal economy is formulated in terms of household enterprises owned and operated by own-account workers (LFS 2006-07). Share of informal economy in total GDP is 37 percent in Pakistan which is much higher than the average share of informal sector in South Asian economies (26 percent of GDP). It is considered as the primary source of job generator after agriculture sector and provides more than half of the total employment in urban areas of Pakistan13. The main reason of this may be its biasness towards unskilled labour. Youth in early stage of their careers, get involved in informal economic activities which may result in low wage and productivity in future.

2.4. Gender Gap in the Labour Market Outcomes Last column of Table 3 highlights the gender gap in different labour

market outcomes for youth in Pakistan. All indicators show the biasness against female youth in the labour market. Their LFPR is almost 51 percentage points lower than that of male LFPR. Similarly, literacy rate of female youth is 19.7 percentage points less than that of male youth. Most of employed females are working as unpaid family helpers, which show the lack of proper work opportunities for them as compared to their male counterparts.

3. Literature Review

There has been a debate among researchers over the effects of early start of career. Their focus is on the impact of early start of career on educational achievements, human capital accumulation, productivity and finally on earnings later in life. In USA, a study by Michael and Nancy (1984) has shown that early work experience of youth should not be ignored as it does impact on labour market experience later in life.

For example, researchers like Elahi et al., (2005); Emerson and Andre (2006) found that boys who enter labour market early earn less and more likely to be in lowest income quintile later in life. Similarly, in Pakistan,

13 ILO, 2005.


89

Faizunnisa (2005) found that, early start of career is often a phenomenon that exists in poor families which adversely affect their life time earnings.

Table 3: Gender Gap in Labour Market Indicators (2006-07)

Indicators Male Female Gender Gap14 Labour Force Participation Rate 69.2 18.4 50.8 Unemployment Rate 7.1 8.9 -1.8 Employment-to-population Ratio (EPR)15 64.0 17.0 47.0 Literacy Rate 77.2 57.5 19.7 Share of Employment in Formal economy as percentage of total employment

14.0 9.0 5.0

Share of Employment in informal economy as percentage of total employment

53.0 32.0 21.0

Share of unpaid family helpers in total employment

35.0 56.5 -21.5

Source: Calculated from LFS, 2006-07

Some researcher analyzed the factors that can affect the decision of schooling and work of young people in market. For example, Rosati and Rossi (2003) analyzed the decision of household regarding the school attendance or labour supply (hours worked) by young people in Pakistan. Using Household Survey data, they applied Tobit model for the dependent variable of hours worked per week and Probit model on the decision to school enrolment of children. Independent variables include age, age squared (as a proxy variable for experience), household income, household size, number of children in household, and dummy variables for being female, and residence of rural areas. Their results showed that household size and number of children present in the household reduce the probability of school enrollment. Similarly, children living in rural areas are also less likely to be enrolled. The model of labour supply (hours worked per week) by children showed that increase in the income of household reduces the number of hours worked by children. Female children with larger household size worked fewer hours in market, this may be due to the fact that they spend more time in household work which increases in case of large household size. 14 Gender Gap is calculated by deducting the respective indicator of female youth from that of male youth. For example, gender gap in literacy rate = male literacy rate – female literacy rate. 15 EPR is taken from GOP (2008) “Pakistan Employment Trends for youth”.

Ahmad and Hussain

90

Female labour force participation in some developing countries like Pakistan is very low. Majority of the women are engaged in household works which are mostly unpaid and hidden. A study by Durrant (2000) showed that 45 percent of females aged 10–19 are apparently not engaged in any economic activities in Pakistan. Similarly, Sathar (2005) also investigated women work at home and found that at every age from 15-24, women work more hours than men but their work is largely unpaid and hidden.

Lloyd and Monica (2004) used Adolescent and Youth Survey of Pakistan (2001-02) and developed a model to analyze the determinants of youth activities in Pakistan. They divided the youth activities into three categories, i.e. household work, schooling, and paid work. Their study concluded that the presence of children, elderly and young people in household is associated with increase in the time of non-economic household work by young females. Having literate parents decrease the time spend on household work by young females especially in urban areas. Their study also highlights that the availability of school, technical institution, and opportunity for job (presence of factory in the area) are strongly associated with time use pattern of young males and females in Pakistan. Availability of schools within one kilometer of area reduces the chances of paid work among young females, while presence of factory in the area increases the time spent by young males and females on paid work.

Fafchamps and Wahba (2006) used labour force survey (1998-1999) in Nepal and found that Children residing near or in urban areas attend school more and work less. Moreover, higher education of parents reduces the probability of child work. Kingdon and Soderbon (2008) by using Pakistan Integrated Household Surveys (PIHS) (1998-99, 2001-02) found that along with increase in education, the likelihood of involving in agricultural production reduces for young men rather they prefer to quite labour force.

4. Data Source and Methodology of the Study This article is based upon micro data from Labour Force Survey of

Pakistan (2006-07), the survey provides information about 32,000 households containing information of 224,000 individuals. From this data set, we selected a sample of 44,902 individuals whose age was between 14 to 24 years, after dropping the 4682 individuals with missing values we left with a sample of 40,220 to use for our empirical analysis. Table 4 describes the description of dependent and independent variables of the study which is self explanatory.


91

Table 4: Description of Variables

Dependent Variable Description

Youth Activity = 1 if full-time student = 2 if combine work with school = 3 if full-time worker = 4 if neither enrolled nor economically active (reference category)

Independent Variables Covariates Sub-groups/Description Youth characteristics Age Age in completed years Age squared Square of age (to capture the experience) Gender

= 1 if female = 0 if male (reference category)

Married = 1 if married = 0 if not married at present (reference category)

Training = 1 if have some technical training and skills = 0 if do not have technical training (reference category)

Migration =1 if migrated from one district to another = 0 if did not migrate from one district to another (reference category)

Head = 1 if head of the household = 0 if not head of the household (reference category)

Education Level16 = 0 if education level is below primary17 (reference category) = 1 if education level is primary but below middle and 0 otherwise =1 if education is middle but below matric and 0 otherwise =1 if education level is matric but below inter and 0 otherwise =1 if education level is inter but below degree and 0 otherwise =1 if education level is degree or above and 0

16 Different softwares require different methods to construct variables, we use STATA 9 which requires to generate variables as described in table 4 17 This category includes illiterate as well as those whose education level is below primary.

Ahmad and Hussain

92

otherwise

Regional Factors Location = 1 if location is rural

= 0 if location is urban (reference category) Province = 0 if province is Punjab (reference category)

= 1 if province is Sind and 0 otherwise = 1 if province is KPK and 0 otherwise = 1 if province is Baluchistan and 0 otherwise

Household Characteristics Female head = 1 if head is female

= 0 if head is male (reference category) Household size Numbers of persons in household Siblings Number of children under the age of 15 years in

household Head education = 0 if education level is below primary (reference

category) = 1 if education level is primary but below middle and 0 otherwise = 1 if education is middle but below matric and 0 otherwise = 1 if education level is matric but below inter and 0 otherwise = 1 if education level is inter but below degree and 0 otherwise = 1 if education level is degree or above and 0 otherwise

Head activity = 0 if head is unemployed or out of labour force (reference category) = 1 if head is working in formal sector and 0 otherwise = 1 if head is working in informal sector and 0 otherwise = 1 if head is working in agricultural sector and 0 otherwise

4.1. The Model As our dependent variable has more than two categories, we estimate multinomial logit model with maximum likelihood estimation procedure on a set of explanatory variables to model the determinants of youth activities in Pakistan


93

Probabilities in the multinomial model18 are given by

| ∑

, 0,2, … , 0 ………(1)

While J log-odds ratios are define as: ln ́ ́ if k = 0 ………………………(2)

We assume that the odds ratio, does not depend upon other choices. As described by Green (2008), the log-likelihood can derived by

defining for each individual, = 1 if alternative j is chosen by individual i, and 0 if not, for the j-1 possible outcomes, then for each i, one and only one of the ’s is 1. The log-likelihood is given by: ln ∑ ∑ ln ……………………………….(3) To interpret the effect of independent variables on the probabilities of each choice we also calculate marginal effects of each outcome. By differentiating equation (1) we find the marginal effects of the characteristics on the probabilities are

∑ ………………………(4)

Likelihood Ratio (LR) Chi-Square test is used to test the null hypothesis that all the slope coefficients in the model are zero.

4.2. Issues and Hypotheses

4.2.1. Age of Youth Economic theory states that along with increase in age, people start

taking part in economic activities and enter in labor market. In the beginning of career, a young person may experience unemployment due to lack of experience and skills but as a young person gets experience, he or she becomes less likely to be economically inactive. In our society, there is a great emphasis on early start of career especially, in rural areas where children start working with their families in fields, so chances to enroll as a student may also decrease along with increase in age.

18 Multinomial Logit model described here is drawn from Green (2008).

Ahmad and Hussain

94

4.2.2. Gender In a male dominant society, females are less likely to participate in economic activities or enroll as a student. They are expected to engage in household work.

4.2.3. Marital Status It is assume that marriage brings some responsibility and a married person is more likely to be engage in economic activities. However, this relationship cannot be expected for females, rather it is assume that married females are more likely to engage in household work and taking care of siblings instead of enroll in educational institution or doing some job.

4.2.4. Education Level Education level of a young person can also affect his/her activities in two ways depending upon how we argue it. One can assume that investment in human capital increases the chances of getting employment in the labor market as educated people are more skillful and can better search for a job as compare to those with low level of education. On the contrary, one can also argue that young people with higher level of education usually have higher expectations about pay and jobs. They become more status conscious and prefer to wait for the time to get better and suitable employment instead of being involved in low paid or informal economic activities.

4.2.5. Head of Household It is assume that being the head of household increases the responsibility of youth, they might start working earlier in their life, it is also expected that there will be more chances of employment at early stage of life if the young person is also having the responsibility of being the head of household.

4.2.6. Location It is assume that youth living in rural areas are more likely to engage in economic activities instead of getting education as compare to their urban counterparts.

4.2.7. Province

Due to diverse culture, cast system, and traditions, young people in each province are expected to have different opportunities and attitude


95

towards work and education. It is assume that more and better employment and educational opportunities are available in Punjab and Sind as compared to Baluchistan and KPK. So, youth living in Baluchistan and KPK are more likely to be inactive as compare to youth living in Punjab and Sind.

4.2.8. Household Size and Number of Siblings Household size may also affect the attitude of youth towards economic activities. Generally, large families increase the burden on young persons to engage in economic activities. This may have a positive impact on labor force participation of youth in the household. One can also expect a negative impact of household size and number of siblings present in house on female labor force participation. As more children in household will require young females to stay at home and take care of young siblings instead of going to work.

4.2.9. Status of Household Head In our society, head is usually responsible to fulfill the financial requirements of household. Therefore, status of household head may greatly affect the activities of young persons in labor market. If head is unemployed then other members of household especially young people will have to take the responsibility to finance the household expenditures. Moreover, it is also expected that if head is working in formal sector he would be better able to finance his household and young persons may get their education instead of participating in economic activities.

4.2.10. Gender of Household Head Head are generally male in our society; it will be interesting to find out whether youth living in female-headed household are more or less likely to engage in economic activities. Youth living in female-headed household may feel responsibility to manage their household and may start their career early.

4.2.11. Education Level of Household Head Increase in the level of education of the household head is expected to reduce the chances of youth to start their career early. A highly educated person is expected to earn enough money that is sufficient to support their families. Therefore, it is expected that higher the level of education of the household’s head more will be the chances that youth will engage in educational activities instead of economic activities.

Ahmad and Hussain

96

5. Determinants of Youth Activities in Pakistan Table 5 presents the results of multinomial logit estimates for youth

activities in Pakistan. For this purpose, we divide activities of youth in four mutually exclusive categories, i.e. full-time students, those who combine work with school, full-time workers and those who neither work nor go to school. Using fourth category (neither work nor school) as our reference category we estimate multinomial logistic coefficients with maximum likelihood estimation. Results suggest that age has an important impact on the decision about schooling and employment for youth in Pakistan. For example, in case of full-time student, the estimated parameters of age and age squared show that the probability of being a full-time student decreases at an increasing rate and reached at its minimum point at the age of 26.59 years. Probability derivative of age also indicates that a one year increase in age decreases the probability of being a full-time student by 6.6 percentage points. Similarly, the probability of combining work with school also decreases along with increase in age while the probability of being a full-time worker increases by 11 percentage points. The main reason of this may be the increase in the cost of education and opportunity cost of staying at school which rises with age.

Similar kind of results is reported by different researchers in Pakistan. For example, studies by Naqvi and Shehnaz (2000) and Arif et al., (2002) found that participation in economic activities increase with age for both male and female youth in Pakistan. However, In Kuwait, Aly and Quisi (1996) found that age is inversely related to women economic participation. The results on the probabilities of female youth show that females are 0.2 percent less likely to be full-time student, 0.9 percent less likely to combine work with school and 71 percent less likely to be full-time workers than their male counterparts. These results depict a traditional bias of society towards females which are mainly considered to do household work instead of going to work or school. Moreover, the probabilities of being a full-time student or full-time worker also decrease if the young person is married. This may be due to high rate of inactivity among female youth which are not expected to work or get education after marriage. These results also confirm the results of earlier studies of Durrant (2000) and Sathar (2005) which show that mostly females in Pakistan are not economically active and their work is largely unpaid and hidden.

As expected, migration, training and being the head of household have positive impact on the probabilities of being a full-time worker. For example,


97

probability of full-time work increases by 17 percentage points if the respondent is the head of household, by 9 percentage points if have some technical training and by 37 percentage points if migrates to earn his or her living. Generally, in society like Pakistan, head of the household is considered to take the responsibility of financial matters of family. Therefore, one can expect an increase in economic participation and chances of full-time work by young people as a head of household. Similarly, a person who gets some training or migrates to earn living may also be expected to fully participate in economic activities in order to maximize benefits of migration or technical training.

To find the impact of education on youth activities we divide it in different categories and take ‘below primary’ as our reference category. Coefficients of this variable show some interesting results, along with increase in the level of education, the probabilities of being full-time student or combine work with school increase while that of full-time work decreases. This may be so as youth with low level of education start their career early (due to limited availability of options) and with higher level of education prefer to get higher education instead of getting involved in low paid economic activities. While on the other hand, all those who have education level of primary or above are more likely to enroll for further education. The highest enrolment rate is among those who have intermediate or degree level education.

Household size and number of siblings present in the household do not affect the first two outcomes (full-time student and combine work with school). One can expect the household size to reduce the school enrolment rate especially for female as concluded by Rosati and Rossi (2003). However, in our results, only number of siblings presents in the household increases the probability of full-time work by 2 percentage points.

Activity of head by sector of employment does not have much impact on the decision of schooling or combining work with school but it does affect the probability of being a full-time worker significantly. Our results suggest that the probability of being full-time work increases by 17 percentage points if the head is working in agricultural sector and by 6 percentage point if the head is working in informal sector. It may be due to the fact that informal sector in Pakistan is considered as the major source of employment in the economy. It consists of households enterprises owned and operated by own-account workers or an enterprise owned and operated by an employer with

Ahmad and Hussain

98

less than ten persons involved in the business. Therefore, our results are not surprising in the sense that youth living with the head who is either working in informal or agriculture sector may be more likely to get involve in work with their families in fields or in household enterprises.

In countries like Pakistan, one can expect that young people in female-headed households may start their career early. However, this variable does not seem to have any impact on youth activities. It may be due to the limited number of data points of this variable as only 0.1 percent of the households are headed by female in our data set.

As expected, the education level of the head of household has strong impact on youth activities in Pakistan. Our results confirm the hypothesis that along with increase in the level of education of head of the household, the probability of being full-time student increases and that of full-time work decreases. A young person with the head’s qualification of degree or above is about 7 percentage points more likely to be full-time student and 22 percentage points less likely to work as compared to a young man who lives in house where head is illiterate or below primary. As far as regional variables are concerned, our results show that young people living in rural areas are 2.6 percentage points more likely to work full-time; however, this variable does not have much impact on other two outcomes (being full-time student or combine work with school). Earlier study by Rosati and Rossi (2003) has also shown that youth living in rural areas are less likely to be enrolled and more likely to work. Provincial difference does not have much impact on the probabilities of full-time student or combine work with school. However, Punjab is the province where young people are more likely to work full-time as compared to the youth in other three provinces.

6. Conclusions and Recommendations

Based on micro data, this paper in descriptive terms shows that a substantial number of our youth is neither in school nor working, moreover, their attitude towards work and level of education also shows the areas need to address. Empirically we investigated the supply side determinants of youth activities in Pakistan. Our results show that being a female, reduces the chances of full time work and full time students in our society. Results also show that being a head of household increases the chances of full time work substantially. The results of this paper show an overall pattern of youth activities and the factors affecting them. In general, the young people with educated parents are more likely to enroll in education, while on the other


99

hand those whose parents are working in agriculture or informal sector are more likely to be full time worker. Similarly, responsibilities within household as a head or having more number of siblings also increase the economic participation of youth. In general, we can conclude that region of residence, personal and parent’s level of education, their employment status, and status within household determine the outcomes of youth activities.

Being a labour abundant country, it would be fair to say that well-being of Pakistan, in future, will heavily depend upon the willingness of its people to work. Unfortunately, the study highlights that a substantial percentage of young people are inactive, neither have they worked nor they study. One can expect high rate of inactivity among females due to household responsibilities but higher inactivity rate among male youth shows the wastage of human resources in the society. Current labour force survey provides very little information about the activities of youth who are neither in school nor in labour force. It is recommended that FBS (Federal Bureau of Statistics) should set a questionnaire that evaluates what young people do in their spare time. How much time they spend in family work, in schooling, loafing and so forth? For this purpose, a time-use survey of youth can also be initiated. The survey must provide information in much more comprehensive way about youth time usage and activities instead of asking just few basic questions. It would also be helpful to differentiate between those who are discouraged workers from those who do not want to work or show any commitment in finding work.

To reduce gender difference in labour market, a motivational campaign is required to educate the society to change their attitude about women work. Providing equal opportunities to young women in education and labour market should be the focus of this campaign. Income generating projects like handicrafts and other home based activities need to be identified for young females in the informal sector. For this purpose, training and educational programs should be launched. Government should also provide a minimum social protection package to vulnerable youth especially for young females in rural areas. The study also highlights the fact that more than half of the youth labour force (62.2 per cent) has either less than one year or just primary education. Moreover, those who are educated face higher unemployment as compared to those with low level or no education. It shows the need to address the issues of relevance and practical application of education in Pakistan. In order to identify the market requirements and needs, link between

Ahmad and Hussain

100

educational institutions and industry should be developed. We also observe that parent’s education significantly affects the activities of youth. Having educated parents improves the chances of youth to get higher education. Any motivational campaign to educate the parents regarding the education and work of their children could improve their economic participation and enrollment in educational institutions.


101

References Ahmad, R. (2011). Youth in Labour Market: An Econometric Analysis of

Micro Data in Pakistan. Unpublished Doctoral Dissertation. Government College University Lahore, Pakistan.

Ahmad, R., & Azim, P. (2010). Youth population and labour market of Pakistan: A micro level study. Pakistan Economic and Social Review, 48 (2), 183-208.

Aly, Y. H., & Quisi, I. A. (1996). Determinants of women labor force participation in Kuwait: A logit analyses. The Middle East Business and Economic Review, 8(2), 25-55.

Arif, G. M., Muhammad, F. K., & Khalid, H. S. (2002). Labour market dynamics in Pakistan: Evidence from the longitudinal data. The Pakistan Development Review, 41(4), 701-720.

Bloom, D. E., David, C., & Jaypee, S. (2001). Economic Growth and Demographic Transition. NBER Working Paper (8685), National Bureau of Economic Research.

Burki, A. A., & Tazeen, F. (1999). Households’ non-leisure time allocation for children and determinants of child labour in Punjab, Pakistan. Paper presented at fourteenth annual general meeting of Pakistan Society of Development Economists, January 28-31, 1999. Pakistan Institute of Development Economics, Islamabad.

Durrant, V. L. (2000). Adolescent girls and boys in Pakistan: Opportunities and constraints in the transition to adulthood. Research Report No. 12, Population Council, Islamabad.

Emerson, P. M., & Andre, P. S. (2006). Is Child Labour Harmful? The Impact of Working Earlier in Life on Adult Earnings. University of Colorado at Denver and Cornell University. Denver, C.O. and Cornel, N.Y.

Faizunnisa, A. (2005). The poverty trap: Leveling the playing field for young people. Brief based on Adolescents and Youth in Pakistan 2001–02: A Nationally Representative Survey. Islamabad: Population Council.

Fafchamps, M., & Wahba, J. (2004). Child labour, urban proximity and household composition. Economics Series Working Papers 213, University of Oxford, Department of Economics.

Ahmad and Hussain

102

GOP. (2008). Pakistan Employment Trends. Ministry of Labour, Manpower and Overseas Pakistanis, Labour Market Information and Analysis Unit. Islamabad, Pakistan.

Green, W. H. (2008). Econometric Analysis. Pearson Education, Inc.

HDR. (2007). Human Development in South Asia: A Ten-year Perspective. Mahbub ul Haq Human Development Centre, Oxford University Press. Karachi.

Ellahi, N., Peter F, O., & Guilherme, S. (2005). How does working as a child effect wages, income and poverty as an adult? Social Protection Discussion Series No.0514, World Bank, Washington, D.C.

ILO. (2005). World Employment Report 2004-05: Employment, Productivity and Poverty Reduction. ILO, Geneva.

Kingdon, G., & Soderbon, M. (2008). Education, skills and labour market outcomes: Evidence from Pakistan. Education Working Paper Series Number 11, World Bank. Available at www.worldbank.org/education

Lee, R., Sang-Hyop, L., & Andrew, M. (2006). Charting the Economic Life Cycle. NBER Working Papers (12379), National Bureau of Economic Research.

Lloyd, C. B., & Monica J. G. (2004). Growing up in Pakistan: The separate experiences of males and females. Policy Research Division, Working Paper no. 188. New York: Population Council.

Michael., & Nancy, B. T. (1984). Youth employment: Does life begin at 16? Journal of Labour Economics, 2(4), 464-476.

Nasir, Z. M., & H. Nazli. (2005). Education and earnings in Pakistan. The Pakistan Development Review. Available at http://www. pide.org.pk/Research/Report177.pdf

Naqvi, Z. F., & Shahnaz, L. (2002). How do women decide to work in Pakistan? The Pakistan Development Review, 41(4), 495-513.

Nayab, D. (2008). Demographic dividend or demographic threat in Pakistan. The Pakistan Development Review, 47(1), 1-27.

Rosati, F. C., & Rossi, M. (2003). Children's working hours and school enrollment: Evidence from Pakistan and Nicaragua. World Bank Economic Review, 17, 283-295.


103

Sathar, A. Z. (2005). Documenting the gender gap in opportunities among Pakistani youth. A brief based on Adolescents and Youth in Pakistan 2001-02: A Nationally Representative Survey. Islamabad: Population Council.

UNESCO. (2006). EFA Global Monitoring Report 2007: Education for All: Strong Foundations, Early Childhood Care and Education. UNESCO, Paris:

United Nations. (2006). World Population Prospects: The 2006 Revision, Dataset on CD-ROM. New York: United Nations.


Appendix

Table A-5: Multinomial Logit Estimates of Youth Activities in Pakistan

Covariates Subgroups Full time Student Combine work Work only

Coefficients Odds Ratios

Marginal Effects


Marginal Effects


Marginal Effects

Personal Characteristics

Age -2.638* 0.07 -0.066 -1.469* 0.23 -0.015 0.271* 1.31 0.111

Age square 0.050* 1.05 0.001 0.028* 1.03 0.000 -0.004** 1.00 -0.002

Gender Male (ref) Female

--- -2.214*

--- 0.11

--- -0.002

--- -3.473*

--- 0.03

--- -0.009

--- -3.830*

--- 0.02

--- -0.714

Married No (ref) Yes

--- -2.618*

--- 0.07

--- -0.034

--- -1.646*

--- 0.19

--- -0.009

--- -0.575*

--- 0.56

--- -0.114

Migrated No (ref) Yes

--- 0.942

--- 2.56

--- -0.018

--- 2.928*

--- 18.69

--- 0.008

--- 2.817*

--- 16.72

--- 0.374

Training No (ref) Yes

--- -0.281

--- 0.75

--- -0.010

--- 0.505

--- 1.66

--- 0.003

--- 0.404*

--- 1.50

--- 0.095

Head of the household

No (ref) Yes

--- -0.007

--- 0.99

--- -0.010

--- 0.566

--- 1.76

--- 0.001

--- 0.778*

--- 2.18

--- 0.171

Educational level

Below Primary(ref) Primary

--- 4.692*

--- 109.07

--- 0.530

--- 0.917*

--- 2.50

--- 0.001

--- -0.019

--- 0.98

--- -0.314

Middle 7.058* 1161.64 0.870 2.875* 17.72 0.006 -0.074 0.93 -0.519

Matirc 7.493* 1795.08 0.916 3.144* 23.19 0.004 -0.187* 0.83 -0.549

Inter 9.413* 12245.21 0.976 4.850* 127.77 -0.001 -0.099 0.91 -0.573

Degree or above 9.315* 11100.68 0.972 4.769* 117.75 -0.003 0.226* 1.25 -0.565

Household Characteristics

Household Size -0.003 1.00 0.001 -0.034 0.97 0.000 -0.048* 0.95 -0.012


105

Covariates Subgroups Full time Student Combine Work Work only


Marginal Effects


Marginal Effects


Marginal Effects

No. of siblings -0.001 1.00 -0.001 0.064** 1.07 0.000 0.086* 1.09 0.021

Head activity

Unemployed (ref) Formal

--- 0.113

--- 1.12

--- 0.001

--- 0.385*

--- 1.47

--- 0.003

--- 0.142*

--- 1.15

--- 0.031

Agricultural 0.186* 0.91 -0.007 1.285* 1.79 0.009 0.262* 1.30 0.174

Informal -0.089 1.20 -0.006 0.582* 3.62 0.004 0.784* 2.19 0.062

Head education

Below Primary (ref) Primary

--- 0.014*

--- 1.01

--- 0.004

--- 0.006

--- 1.01

--- 0.002

--- -0.289*

--- 0.75

--- -0.072

Middle 0.296* 1.34 0.013 -0.068 0.93 0.001 -0.307* 0.74 -0.080

Matric 0.336* 1.40 0.017 -0.083 0.92 0.001 -0.448* 0.64 -0.116

Inter 0.823* 2.28 0.045 -0.234 0.79 0.000 -0.615* 0.54 -0.168

Degree or above 1.105* 3.02 0.069 -0.250 0.78 0.001 -0.782* 0.46 -0.215

Regional Characteristics

Region Urban (ref) Rural

--- -0.176*

--- 0.84

--- -0.006

--- 0.262*

--- 1.30

--- 0.002

--- 0.104*

--- 1.11

--- 0.027

Province

Punjab (ref) Sind

--- -0.434*

--- 0.65

--- -0.005

--- -0.282*

--- 0.75

--- -0.001

--- -0.385*

--- 0.68

--- -0.088

NWFP -0.085 0.92 0.012 -0.026 0.97 0.005 -1.053* 0.35 -0.256

Baluchistan -0.654* 0.52 -0.010 -0.060 0.94 0.001 -0.184* 0.83 -0.038

Constant 25.904* 13.326* -1.034

Log Likelihood LR Chi2 Pseudo R2 Observations

-23012.35 45198.5 0.4955 40220

Note: * indicates significant at five percent level and ** indicates significant at ten percent level. Omitted category is neither work nor school.


A Study of Implicit Tax in Pakistan’s Agriculture, with

Special Reference to the Case of Rice

Mohammad Aslam1

Abstract

The study examined ‘implicit tax’ argument of the agriculturists’ lobby to oppose imposition of an agricultural income tax. The paper discovered a widening gap between procurement and export prices of both Basmati and IRRI. The gap between procurement and consumer prices of the two varieties also widened significantly. Thus while both producers and consumers remained on the losing end, first government and then after the policy reforms the exporters and other intermediaries, were the substantial gainers. Since RECP has been disbanded and the Government has opted out of purchase and export of rice, the margin now goes to the exporters instead of the Government. Under the changed rice policy, the ‘implicit’ tax argument has therefore lost much of weight and relevance.

Keywords: Agricultural prices policy; Basmati; IRRI; acreage; yield; procurement price; consumer price; export price; implicit tax

JEL classification: Q11, Q17, Q18

1. Introduction

The Agricultural Prices Policy in Pakistan has traditionally covered both important inputs and outputs. The Input Price Policy is implemented through provision of subsidized inputs to farmers. The Output Price Policy, on the other hand, is implemented through fixation of procurement and support prices of important food and cash crops such as wheat, rice, sugarcane, potatoes etc. Although both tiers of the Agricultural Prices Policy are important and interdependent too, the present study is limited to an analysis of the output prices policy only.

The Output Prices Policy has been used by economic decision makers in Pakistan since the early 1960s, as an incentive for growers and to expand production frontiers of different crops. During the early years after independence, the government did not use this policy due the widely held

1 The author is Professor of Economics at Lahore School of Economics (LSE), Lahore.

Aslam M.

108

view that subsistence farmers in developing countries are not responsive to price incentive. It was a general perception that they produce only for self consumption and are not influenced by prevailing market prices. The later studies however showed that in Pakistan farmers do respond positively to changes in prices of important food and cash crops and adjust their acreage decision accordingly. This prompted the government to use the policy in the 1960s for expanding production.

1.1. The Transition of the Policy and Problem Statement The policy has undergone a change over a period of time and more radically in the recent past and there is also reorganization of the institutional framework. This is particularly true of the rice output price policy. Rice is a crucial crop and occupies important place in the export economy of Pakistan besides being a food supplement to wheat. The rice policy since late 1970s and early 1980s has undergone many changes. Firstly, the compulsory procurement policy was replaced with voluntary procurement policy. Secondly, the ban on inter-district movement of rice was discarded. Thirdly, the Rice Export Corporation of Pakistan (RECP) which was created in 1974 and was assigned the responsibility of procurement and export of rice in the public sector was disbanded in 2000 and merged with the Trading Corporation of Pakistan. The important thing however was that the government opted out of rice export business and decided to discontinue fixation of procurement price of rice. This was done to allow market forces to prevail in the area of rice production and export. The Agricultural Prices Commission (APCOM) responsible for recommending procurement and support prices was also recast and renamed as the Agricultural Prices Institute (API).

Originally, the support price program covered crops like wheat, rice, sugarcane, cotton, potatoes, onions, grams, and non-traditional oil seeds such as sunflower, soybean, canola and safflower. In May 2001, on recommendation of the MINFAL, the Economic Committee of the Cabinet (ECC) reduced the coverage to wheat, rice, sugarcane and cotton crops. In September 2002, the ECC decided to further limit it to wheat, rice and cotton at the federal level while price of sugarcane was to be determined by the provinces. The government opted out of the export of both rice and cotton and specialized institutions created for the purpose i.e. Rice Export Corporation (RECP) and Cotton Export Corporation of Pakistan (CECP) were disbanded and merged with the Trading Corporation of Pakistan (TCP).

A study of Implicit Tax in Pakistan’s Agriculture

109

Since then the thinking on fixation of prices has undergone a major change with acceptance of an enhanced role for the markets. The coverage thus was restricted to wheat and cotton only. At present support price system stands discarded in case of almost all the crops. The prices fixed are only ‘indicative’ in character and provide growers a base level for negotiating better prices for themselves.

The Rice Exporters Association of Pakistan (REAP) has taken the place of RECP as regards procurement and export of rice. The TCP facilitates fulfillment of orders in consultation with the REAP. The government has also established a Quality Review Committee (QRC) that certifies the quality of rice before shipment.

The agricultural sector has traditionally been exempted from levy of a tax on agricultural incomes. The levy of the tax was opposed by the farming community on grounds of paying an ‘implicit tax’ to the government. The Government procured rice at prices arbitrarily fixed by her and then sold it internationally at prices many times higher than prices paid to the farmers. The margin accruing to the government was referred to as the so-called ‘implicit tax’.

1.2. The Study Objective The study examines ‘implicit tax’ plea advanced by farmers for avoidance of an agricultural income tax, particularly in the background of important institutional and policy changes referred to above.

2. The Literature Review There are many studies on the subject and its related matters. Aslam, M. (1982) studied the rice economy of Punjab with a particular focus on consumption aspect. The basic purpose of the study was to explore the prospects of promoting rice consumption with a view of releasing pressure on wheat. The study also analyzed the issue of the ‘implicit tax’. The important finding of study pertaining to the ‘implicit tax’ issue was that the gap between procurement and export price of rice had been widening over time and that government was the real beneficiary and earned increasing revenue due to this gap. The time series data of the three sets of prices for the period 1964-65 to 1979-80 was used for purpose.

Roberto Eliseu and Pastore Affonso (1978) studied the problem of import substitution and implicit taxation of agriculture in Brazil. According to them, industrialization in Brazil prior to the World War 2 had taken place at

Aslam M.

110

the cost of agriculture through a shift in resources to the industrial sector. In the post World War 2 periods the same thing had happened through import substitution industrialization. This was ensured through providing protective devices, subsidized of credit and stable wages.

Chaudhry M. G. and Kayani N. N. (1991) discussed the issue of implicit taxation of Pakistan’s agriculture. They compared import and export parity prices of major agricultural commodities with their domestic procurement prices and discovered that implicit tax argument was not without substance. The implicit tax rate for some of the years under study 1970-71 to 1989-90 was as high as 75% in the case certain commodities.

Chaudhry, M. G. (2001) discussed the current tax policy in Pakistan’s agriculture in the backdrop of the theory of optimal taxation. He quantified total amount of implicit tax on agriculture that declined from Rs. 82 billion in 1989-90 to Rs. 65 billion in 1999-2000. Despite reduction, implicit tax, calculated on the basis of parity and support prices, constituted 7-8.5% of value added by agriculture.

Noor, P. K. (2002) reviewed implications of government intervention in Pakistan’s wheat and cotton sectors. The study revealed overall transfers from wheat and cotton producers to society. The study also showed that WTO trade liberalization in wheat and cotton would have no significant impact on wheat and cotton production.

Ronge, Eric; Wanjala Bernadette and others (2005) studied implicit taxation of the agricultural sector in Kenya. They had concluded that agriculture was being taxed implicitly through changes in macroeconomic policies. They recommended that the government must ensure that this should not have an adverse impact on Kenyan agriculture.

Lin, Justin Yifo and Liu, Mingsing (2007) examined the historical evolution of China’s rural taxation system. The period under review was from pre-reform period to the late 1990s.The study discovered excessive local informal taxation on farmers. This necessitated a policy review that resulted in a change in the traditional approach of implicit taxation.

Salam, A. (2010) studied recent trends in distortions in incentives for production of major crops in Pakistan. The study compared domestic producer prices between 1991 and 2008 with the corresponding international prices with a view to measure nominal protection coefficients (NPCs). The study


111

revealed that in the case of rice, average implicit tax per ton of Basmati paddy was around $ 21.38.

3. Methodology To monitor relationships between procurement and export price, on the one hand, and between procurement and consumer price, on the other, first ordinary or actual curves were drawn. Then least square straight lines or trend lines were estimated to examine the overall long term trend.

The actual curves generally exhibit wide fluctuations from one year to the other and may not reveal much at first sight. That necessitated estimation of the trend lines. The straight line equations and coefficients of variation were also estimated for the three sets of prices.

3.1. Data Collection Secondary data was used for the study. This was collected mainly from government of Pakistan publications such as Foreign Trade Statistics of Pakistan, annual Economic surveys, Foreign Trade of Pakistan (an EPB/TDAP publication), Agricultural Statistics of Pakistan and Pakistan Statistical Yearbook. The data was also gleaned through publications and studies of the international Rice Research Institute, Manila, Food and Agricultural Organization (Rome) and Rice Research institute in Kala Shah Kaku in District Sheikhpura. The time-series data used pertained to the period 1990 to 2008. This covered procurement, consumer and export prices of Basmati and IRRI.

4. Results and Interpretation 4.1. Basmati Rice 4.1. 2. Actual Lines for Procurement and Export Prices of Basmati The figure 1 shows actual lines for both procurement and export prices of Basmati rice. The actual curve of export price of basmati shows more severe fluctuations compared to the actual curve of procurement price of basmati. The last two years of the period particularly show unusual and rapid upward trend in the export price of basmati. There was phenomenal food inflation at the world level and rice was no exception. The actual curve is almost flat and is shown increasing only gingerly. 4.1.3. Trend Lines for Procurement and Export Prices of Basmati The figure 2 shows trend lines of the procurement and export prices of basmati.

Aslam M.

112

Figure 1:Actual Lines for Procurement and Export Prices of Basmati

0

500

1000

1500

2000

2500

3000

3500

4000

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008 Years

Pric

e (p

er 4

0 kg

)

ProcurementExport

Figure 2:Trend Lines of Procurement and Export Prices of Basmati

0

500

1000

1500

2000

2500

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

Years

Pric

e (p

er 4

0 kg

)

Procurement Export

The trend lines of procurement and export prices of basmati are shown strongly drifting apart during the period under review. The closing years of the period show even greater rapid divergence of the two curves. Under assumption that overhead cost of exporters in terms of storage and transportation charges did not increase abnormally, rapidly widening gap


113

shows increasing profit margin for exporters. This also implies that while exporters reaped huge profits, producers were the real losers.

The trend line linear equations for procurement and export prices of basmati were estimated as under.

pY 105.965 20.1193t= + (Procurement Price)

eY 14.8070 110.407t= − + (Export Price)

The trend linear equation of procurement price shows an average increase of Rs.20 per year while trend linear equation of export price shows an average increase of Rs.110.4 per year.

4.1.4. The Impact on Consumer

In order to gauge impact on consumers in this process of production, consumption and export of basmati, combined actual and trend graphs for the

three sets of prices were also constructed as in figure 3 and figure 4.

Figure 3: Actual Lines for Procurement, export and consumer prices of basmati

0

500

1000

1500

2000

2500

3000

3500

4000

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008 Years

Pric

e (p

er 4

0 kg

) ProcurementConsumerExport

Aslam M.

114

Actual lines for three sets of prices were combined in a single line chart, the relative position of consumers became clear. The lines representing export and consumer prices remained glued to each other over whole length of the period and even submerged at times particularly starting early twenties.

When trend lines of the three sets of prices were jointly drawn in one graph, its graph looked as in figure 4.The trend line representing consumer goods is keeping pace with the export price trend line at a small distance. The gap between trends lines of export and consumer prices on the one hand and procurement price on the other is shown continuously widening over time. This means both producers and consumers remained at a disadvantage compared to the exporters of basmati.

Figure 4: Trend Lines for Procurement, Consumer and Export Prices of Basmati

0

500

1000

1500

2000

2500

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008 Years

Pric

e (p

er 4

0 kg

)

ProcurementConsumerExport

The trend linear equation for consumer prices was estimated as under.

cY 90.5088 110.361t= − + (Consumer Prices)

This showed an average increase of Rs.110.4 per year in consumer price of basmati over the period. Earlier average increase per year of export price of basmati had also approximated to the same figure.


115

4.2. IRRI Rice

4.2.1. Actual and Trend Lines for the Prices of IRRI

The figure 5 shows actual lines for both procurement and export prices of IRRI rice. The actual curve of export price of IRRI shows more severe fluctuations compared to the actual curve of procurement price of IRRI. The last two years of the period particularly show rapid upward trend in the export price of IRRI. There was severe food inflation at the international level and rice was no exception. The actual curve is found increasing only modestly.

Figure 5: Actual Lines for Procurement and Export Prices of IRRI

0

200

400

600

800

1000

1200

1400

1600

1800

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008Years

Pric

e (p

er 4

0 kg

)

ProcurementExport

The trend lines of procurement and export prices of IRRI are shown strongly drifting apart during the period under review. The closing years of the period show even greater rapid divergence of the two curves. Under assumption that overhead cost of exporters in terms of storage and transportation charges did not increase abnormally, rapidly widening gap shows increasing profit margin for exporters. This also implies that while exporters reaped huge profits, producers were the real losers. The trend line linear equations for procurement and export prices of IRRI were estimated as under.

pY 29.3860 14.0561t= +

Aslam M.

116

eY 22.7018 49.1965t= − +

The trend linear equation of procurement price shows an average increase of Rs.14 per year while trend linear equation of export price shows an average increase of Rs.49.2 per year.

Figure 6: Trend Lines of Procurement and Export Prices of IRRI

0

100

200

300

400

500

600

700

800

900

1000

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008 Years

Pric

e (p

er 4

0 kg

) ProcurementExport

4.2.2. Actual and Trend Lines for Procurement, Consumer and Export Prices of IRRI

When actual lines for three sets of prices of IRRI were combined in a single line chart, the relative position of consumers became clear. The actual lines representing export and consumer prices are seen rising in close proximity with one another but overtaking each other alternately during certain intervals.

4.2.3. The Impact on Consumer

In order to gauge impact on consumers in this process of production, consumption and export of IRRI, combined trend line graph for the three sets of prices were also constructed as in and figure 8.

When trend lines of the three sets of prices were jointly drawn in one graph, its graph looked as in figure 8. The trend line representing consumer


117

goods is rising very close to the export price trend line and during certain interval the two lines are seen coinciding with one another.

The gap between trends lines of export and consumer prices on the one hand and procurement price on the other is seen continuously widening over time. This means both producers and consumers remained at a disadvantage compared to the exporters of IRRI.

Figure 7: Actual Lines for Procuremnt, Consumer and Export Pricess of IRRI

0200400600800

10001200140016001800

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008 Years

Pric

e (p

er 4

0 kg

) Procurement

Consumer

Export

Figure 8: Trend Lines for Procurement, Consumer and Export Pricess of IRRI

0

100

200

300

400

500

600

700

800

900

1000

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

Years

Pric

e (p

er 4

0 kg

)

ProcurementConsumerExport

The trend linear equation for consumer prices of IRRI was estimated as under.

Aslam M.

118

cY 34.5263 44.8579t= +

This showed an average increase of Rs.45 per year in consumer price of IRRI over the period. Earlier average increase per year of export price of IRRI was approximated to Rs.49 per year over the same period.

5. Conclusions and Policy Implications The study pertained to the three sets of prices and an examination of the hidden tax argument. The important findings were as under.

1. During the period under study, the spread between procurement/indicative price and export price of basmati kept on widening.

2. This was also true of the spread between procurement and export prices of IRRI although the spread was more pronounced in the case of basmati due to its being premium quality rice.

3. The basmati and IRRI rice farmers receive prices that are many times below the world prices. Thus on the face of it their contention of an ‘implicit tax’ being paid by them sounds logical.

4. This conclusion will not be significantly altered even after milling, storage and transportation charges are duly accounted for and adjustment made.

5. The consumers, on the other hand, pay quite high prices and in the case of IRRI, consumer price even overtakes the export price. By implication, it may be stated that exports of basmati and IRRI and particularly the latter, do adversely impact upon domestic supply and domestic prices of the two rice varieties.

6. Presently there are no exports in the Public sector. The Rice Export Corporation of Pakistan was disbanded in 2000. The government now only facilitates exports and exporters through Trade Development Authority of Pakistan (former Export Promotion Bureau). The residual is thus appropriated by the intermediaries including rice exporters.

7. Thus under changed circumstances, the ‘implicit tax’ argument is no longer tenable. The government of late has opted out and does not fix procurement prices in order to allow market forces to play their due role.

8. After reversion to the market system, farmers are better advised to form their own rice export associations in order to reduce the role of intermediaries.


119

References Abbas, S. A. (1967). Supply and demand of selected agricultural products in

Pakistan-1961-75. Oxford University Press, Lahore.

Aslam, M. (1982). Rice economy in the Punjab with a particular focus on consumption aspect. Punjab Economic Research Institute (PERI), Lahore.

Chaudhry, M. G., & Kayani, N. N. (1991). Implicit taxation of Pakistan’s agriculture: An analysis of the commodity and input prices. The Pakistan Development Review, 30.

Chaudhry, M. G. (1999). The theory and practice of agricultural income tax in Pakistan and a viable solution. The Pakistan Development Review, 22.

Chaudhry, M. G. (1995). Recent input-output price policy in Pakistan’s agriculture: Effects on producers and consumers. The Pakistan Development Review.

Cummings, J. T. (1975). Cultivator market responsiveness in Pakistan–cereal and cash crops. The Pakistan Development Review, XIV(3).

Eric, R., Wanjala, B. et al. (2005). Implicit taxation of the agricultural sector in Kenya. Kenya Institute for Public Policy Research and Analysis (KIPPRA), Discussion Paper No. 52.

Eddie, S. M. (1971). Farmers’ response to price in large estate agriculture in Hungary 1870–1913. The Economic History Review, 24(4).

Government of Pakistan, Export Promotion Bureau,: Foreign Trade Statistics of Pakistan (various Years).

Government of Pakistan, Federal Bureau of Statistics, Statistics Division, Pakistan Statistical Yearbook (various Years).

Government of Pakistan, Federal Bureau of Statistics, Statistics Division, Foreign Trade of Pakistan (various Issues).

Falcon, W. P. (1962). Farmer response to price in an underdeveloped area: A case of West Pakistan. Harvard University Press, Harvard, USA.

Falcon, W. P. (1964). Farmer response to price in a subsistence economy: The case of West Pakistan. The American Economic Review, 54(3).

Fulginiti, L. E., & Perrin, R. K. (1993). Prices and productivity in agriculture. The Review of Economics and Statistics, 75(3).

Aslam M.

120

Hossein, A., & Cummings, J. T. (1977). Estimating agricultural supply response with the Nerlove Model: A survey. International Economic Review, 18(2).

Koc, A. A. (1999). Acreage allocation model estimation and policy evaluation for major crops in Turkey. Centre for Agricultural and Rural Development Working Paper 99, Iowa State University.

Khan, N. P. (2002). Government intervention in Pakistan’s wheat and cotton sectors: Concepts, policies and implications. Asian Journal of Plant Sciences, 1(4).

Krishna, R. (1963). Farm supply response in India-Pakistan: A case study of the Punjab region. The Economic Journal, 73(291).

Mythili, G. (2006). Supply response of Indian farmers: Pre and post reforms. Indira Gandhi Institute of Development Research, Mumbai, India.

QRC, Rice Inspection Cell, Karachi: comparative Variety-Wise statement of Exports of Rice (various issues)

Rezitis A. N. (2008). Greek beef supply response and price volatility under CAP reforms. Working Papers Series, University of Ioannina, Greece.

Roberto, E., & Pastore, A. (1978). Import substitution and implicit taxation of agriculture in Brazil. American Journal of Agricultural Economics, 60(5).

Salam, A. (2010). Distortions in incentives for production of major crops in Pakistan: Recent trends and emerging challenges. IATRC Policy Brief No PB2010-01.

Stern, R. M. (1962). The price responsiveness of primary producers. The Review of Economics and Statistics, 44(2).

Shafiq, M. (2005). Supply response of major crops in different agro-ecological zones in Punjab. PhD Thesis, Faculty of Agricultural Economics, University of Agriculture, Faisalabad.

Turvey, R., & Cook, E. (1976). Government procurement of price support of agricultural commodities: A case study of Pakistan. Oxford Economic Papers, 28(1).


121

Umar, F., Trevor, Y. et al. (2001). The supply response of Basmati rice growers in Punjab, Pakistan: Price and non-Price determinants. Journal of International Development, 13(2).

William, L., Westcott, P. C. et al. (2000). Supply response under the 1996 Act and Implications for the US field crops sector. USDA Economic Research Service, Technical Bulletin No.1888.

Aslam M.

122

Appendices

Appendix Table 1: Actual and trend prices (per 40 kg) of Basmati (1990-2008)

Years Procurement Price

Consumer Price

Export Price

Trend (procurement)

Trend (consumer, linear)

Trend (export, linear)

Trend (consumer , Quad)

Trend (export, Quad)

1990 143 322 423 126.084 19.85 95.6 479.74 558.48 1991 150 366 411 146.204 130.21 206.01 436.81 514.59 1992 155 411 448 166.323 240.58 316.41 411.91 488.86 1993 175 454 498 186.442 350.94 426.82 405.04 481.28 1994 185 467 502 206.561 461.3 537.23 416.21 491.85 1995 211 452 558 226.681 571.66 647.64 445.42 520.57 1996 222 630 701 246.8 682.02 758.04 492.66 567.44 1997 255 648 793 266.919 792.38 868.45 557.93 632.47 1998 310 778 966 287.039 902.74 978.86 641.24 715.65 1999 330 905 1056 307.158 1013.11 1089.26 742.58 816.98 2000 350 913 1002 327.277 1123.47 1199.67 861.96 936.46 2001 385 990 1152 347.396 1233.83 1310.08 999.38 1074.1 2002 385 1187 1176 367.516 1344.19 1420.48 1154.82 1229.89 2003 385 1205 1190 387.635 1454.55 1530.89 1328.31 1403.83 2004 400 1300 1266 407.754 1564.91 1641.3 1519.83 1595.92 2005 415 1350 1343 427.874 1675.27 1751.71 1729.38 1806.16 2006 460 1509 1486 447.993 1785.64 1862.11 1956.97 2034.56 2007 460 2255 2361 468.112 1896 1972.52 2202.59 2281.11 2008 460 3107 3364 488.232 2006.36 2082.93 2466.25 2545.81


123

Appendix Table 2: Actual and trend prices (per 40 kg) of IRRI (1990-2008)

Years Procurement Price

Consumer Price

ExportPrice

Trend (procure)

Trend (consumer ,linear)

Trend (export ,linear)

Trend (consumer, quad)

Trend (export ,quad)

1990 66 166 156 43.442 79.384 26.495 261.13 295.66 1991 73 192 193 57.498 124.242 75.691 245.41 255.13 1992 78 214 215 71.554 169.1 124.888 236.81 225.17 1993 85 239 207 85.611 213.958 174.084 235.34 205.75 1994 90 231 239 99.667 258.816 223.281 241 196.89 1995 103 300 320 113.723 303.674 272.477 253.78 198.59 1996 112 429 315 127.779 348.532 321.674 273.69 210.84 1997 129 388 347 141.835 393.389 370.87 300.73 233.65 1998 153 433 418 155.891 438.247 420.067 334.9 267.01 1999 175 601 396 169.947 483.105 469.263 376.19 310.93 2000 185 423 347 184.004 527.963 518.46 424.62 365.4 2001 205 401 412 198.06 572.821 567.656 480.17 430.43 2002 205 453 412 212.116 617.679 616.853 542.84 506.02 2003 205 465 487 226.172 662.537 666.049 612.65 592.16 2004 215 549 524 240.228 707.395 715.246 689.58 688.86 2005 230 619 561 254.284 752.253 764.442 773.63 796.11 2006 300 623 612 268.34 797.111 813.639 864.82 913.92 2007 310 959 1151 282.396 841.968 862.835 963.13 1042.28 2008 310 1494 1604 296.453 886.826 912.032 1068.57 1181.2

Aslam M.

124

Table 1: Procurement, Consumer and Export Prices of Basmati (1990-2008)

Years Procurement Consumer Export Price Price Price 1990 143 322 423 1991 150 366 411 1992 155 411 448 1993 175 454 498 1994 185 467 502 1995 211 452 558 1996 222 630 701 1997 255 648 793 1998 310 778 966 1999 330 905 1056 2000 350 913 1002 2001 385 990 1152 2002 385 1187 1176 2003 385 1205 1190 2004 400 1300 1266 2005 415 1350 1343 2006 460 1509 1486 2007 460 2255 2361 2008 460 3107 3364

Sources: 1. Federal Bureau of Statistics, Statistics Division, Government of Pakistan GOP): “Foreign Trade Statistics of Pakistan” (various years) 2. Export Promotion Bureau, Government of Pakistan: “Foreign Trade of Pakistan (Various years) 3. Economic Advisor’s Wing, Finance Division, Government of Pakistan: “Pakistan Economic Survey (various years). 4. Economic Wing, Ministry of Food, Agriculture and Livestock, Government of Pakistan (GOP): “Agricultural Statistics of Pakistan” (various years).


125

Table 2: Procurement, Consumer and Export Prices of IRRI (1990-2008)

Years Procurement Consumer Export Price Price Price 1990 66 166 156 1991 73 192 193 1992 78 214 215 1993 85 239 207 1994 90 231 239 1995 103 300 320 1996 112 429 315 1997 129 388 347 1998 153 433 418 1999 175 601 396 2000 185 423 347 2001 205 401 412 2002 205 453 412 2003 205 465 487 2004 215 549 524 2005 230 619 561 2006 300 623 612 2007 310 959 1151 2008 310 1494 1604

Source: 1. Federal Bureau of Statistics, Statistics Division, Government of Pakistan (GOP): “Foreign Trade Statistics of Pakistan” (various issues). 2. Export Promotion Bureau, Government of Pakistan (GOP): “Foreign Trade of Pakistan (various issues). 3. Economic Advisor’s Wing, Finance Division, Government of Pakistan (GOP): “Pakistan Economic Survey (various issues). 4. Economic Wing, Ministry of Food, Agriculture and Livestock, Government of Pakistan (GOP):“Agricultural Statistics of Pakistan” (various issues).


Determinants of Residential Electricity Expenditure in Pakistan: Urban-Rural Comparison Ijaz Hussain and Muhammad Asad1

Abstract

In this study the authors attempted to find out the determinants of the consumption expenditure on electricity by households. Explanatory variables are income of household, family size, number of rooms in the house, region, province and electricity consuming appliances like AC, fridge, freezer, computer, washing machine and air cooler. The authors found out that expenditure on electricity is income inelastic, increase in family size and number of rooms increases the expenditure on electricity. Households living in urban areas have more expenditure on electricity as compared to the rural households. Households in urban and rural areas of Punjab have more electricity expenditure as compared to the rest of the provinces. The acquisition of electric appliances contributed positively towards the electricity expenditure. A.C. and Freezer are the two most powerful contributors.

Keywords: Households; Electric expenditure; Electric appliances; Pakistan

JEL classification: Q4, Q41, Q43

1. Introduction At present, Pakistan is facing a power shortage ranging between 4000-

5000 megawatts (MW), because supply of electricity is increasing much slowly as compared with its demand. On overage demand for electricity has increased at a rate of 9.5% per annum during last four years due to urbanization, industrialization and electrification of the rural areas.. It is projected to grow by 8.7% per annum.2

If we look at sectoral consumption of electricity by economic groups, we find that domestic group is the largest consumer of electricity with average annual share of 45%. In the last four years (2003-04 to 2006-07), on average, consumption in domestic sector has increased by 8.9% annually. Number of electricity consumers in March 2008 was 17.73 million, out of which 15.02 million were the domestic consumers. In 1997-98 domestic consumers were 1 The authors are Assistant Professor at Department of Economics, Gomal University, D I Khan and Officer, State Bank of Pakistan, Lahore, respectively. 2 Source: Pakistan Economic Survey (PES) 2007-08.

Determinants of Residential Electricity Expenditure

128

8.4 million. Thus, the number of electricity consumers has doubled within 10 years as shown in table 1.2. If we look at the supply side we find that projected supply is 2000-3000 MW lesser than demand.

Since the supply falls short of the demand and there is continuous increase in the electricity consumption, it is highly desirable to conduct a demand side analysis regarding domestic consumers of electricity, as they constitute the largest group of electricity users (see table and figure, 1.1). Instead of considering the supply side of electricity, the alternative option is to study the demand side approach in electricity through demand management.

Table 1.1: The Share of Consumption of Electricity by End-Users (in %)

Year Households Commercial Industrial Agriculture 1997-98 42.2 5.2 27.6 15.5 1998-99 44.8 5.5 27.7 12.9 1999-00 46.9 5.5 29 9.9 2000-01 46.9 5.8 29.4 10.1 2001-02 45.8 5.9 29.8 11.1 2002-03 45 6.1 30.8 11.4 2003-04 44.9 6.4 30.3 11.7 2004-05 45 6.7 30.3 11.4 2005-06 45.4 7 29.3 11.7 2006-07 45.8 7.4 29 11.3 Average 45.27 6.15 29.32 11.7 July-March 2006-07 45 7.3 29.7 11.5 2007-08 45.6 7.4 28.4 11.8

Source: Pakistan Economic Survey: 2007-08

For this reason, detailed analysis of consumers’ electricity consumption is necessary and is the focal point of this paper. In this paper we intent to find out the determinants of domestic electricity expenditure (per month) using micro-data. This is desirable because a household level study can incorporate household characteristics and shed some light on the nature of consumer responses [See Filippini & Pachauri (2004)]. Moreover, by including different geographical factors we can see consumer behavior in different sub-groups. Thus, use of micro-data provides more detail and depth as compared to the aggregate level study. Unfortunately, all the studies which

Hussain and Asad

129

have taken up this topic in Pakistan have used aggregate level data. Therefore there is a need for micro level study on electricity demand in Pakistan using the micro data. Micro-data study is also important because it can suggest something about the demand management policy. A demand management policy could be a better solution, in the short run because changing the supply of electricity will require a longer timeframe. And even after increase in

Table 1.2: Consumers by Economic Groups (Thousands)

Year Households Commercial Industrial Agriculture Other Total1997-98 8455 1397 187 171 8 102181998-99 8912 1517 190 173 8 108001999-00 9554 1654 195 175 8 115862000-01 10045 1737 196 180 8 121662001-02 10483 1803 200 184 8 126782002-03 11044 1867 206 192 9 133182003-04 11737 1935 210 199 10 140912004-05 12490 1983 212 201 10 148962005-06 13390 2068 222 220 10 159102006-07 14354 2152 233 236 11 16986July-March 2006-07 14069 2132 230 233 11 166752007-08 15026 2214 240 243 11 17734Source: PES 2007-08

Figure 1.1: Available Capacity and Computed Demand (in MW)


130

supply the demand management policy will ensure against a power crisis as is faced by the country today.

2. Brief Literature Review Understanding the demand and supply forces and their determinants in electricity sector is important because today our lives are directly affected by it as we have become dependent on the use of appliances run by electricity. In the light of current electricity crisis the topic of demand side management has gained special significance. Despite its significance there has been no considerable work regarding electricity demand, on household level data, in Pakistan. Perhaps because the demand for electricity was considered as “given” or predetermined. Whatever the reason may be, the demand side of electricity is still waiting to be explored in Pakistan. We still have to develop insight about the dynamics of electricity demand in our country. In this paper our goal is to see; what are the major determinants of household expenditure on electricity (demand) using household level data.

There are a host of studies that have taken up the topic of electricity demand regarding domestic, industrial and commercial users. Some of the studies focused on residential demand for electricity are mentioned here. Houthakker (1951) has studied domestic demand for electricity in UK using cross sectional data on 42 provincial towns for a period from 1937-1938. He used OLS technique to estimate double log models which included variables like; average annual electricity consumption of each household with a decreasing two part tariff, average income , marginal price of electricity, marginal price of gas, and average holding of electricity consuming appliances per household.

Fisher and Kaysen (1962) have focused on both residential and industrial demand for electricity in US. by using a dataset having observations for 47 states for the period 1946 to 1957. They used OLS and analysis of covariance techniques. The model they estimated was in log form and included ex post average price and per capita income, both of them in real terms. They explicitly differentiated between the short run and long run domestic electricity demand, for the first time.

Houthakker and Taylor (1970) have analyzed the residential demand for electricity using annual time series data on personal consumption expenditure for the period 1947-1964. They used state adjustment model to make an equation for personal consumption expenditure on electricity. They

Hussain and Asad

131

estimated both short run and long run elasticities. Other studies on US residential demand for electricity include; Wilson (1971), Mount et al (1973) and Anderson (1973) among others.

Researchers discussed in detail the issues involved in modeling the demand for electricity including Houthakkar (1962), Fisher & Kaysen (1962), Houthakkar & Taylor (1970), Wilson (1971), Cargil & Meyer (1971), Mount et al (1973), Anderson (1973), Anderson (1971), Lyman (1973) and Houthakkar et al (1973), Taylor (1975).

Moreover, Reiss and White (2001) have studied US household electricity demand in the short run and have taken care of problems like non-linearity of electricity prices, data aggregation and heterogeneity in household’s price sensitivity. They used data of a representative sample of 1307 California households for year 1997. Estimation is done using Generalized Method of Moments (GMM) technique.

Filippini and Pachauri (2004) have studied residential demand for electricity for all urban areas of India. They have used cross section data containing 30,000 households for the year 1993-94. They estimated three demand functions in log form using monthly data for the summer, winter and monsoon seasons. The variables they included were average price of electricity, price of kerosene, price of LPG, total household expenditure, covered area of the house, size of town, size of household and age of head of the household. They did not include the information about the appliance held by the households. Their results show that the residential electricity demand is income and price inelastic in all three seasons whereas geographical, household and demographic variables included, show significant impact on electricity demand.

Other micro-data studies which have taken up this topic are; Halvorsen (1975) for USA, Parti & Parti (1980) for San Diego, Barnes et al. (1981) for USA, Murthy (2001) for India and Dubin & McFadden (1984) for USA. Studies which have taken up this topic on the aggregate level and studied it in the time series settings include; Holtedahl and Frederick (2004) for Taiwan, Akmal & Stern (2001) for Australia, Zachariadis & Pashourtidou (2006) for Cyprus, Halicioglu (2007) for Turkey, Dergiades and Lefteris (2008) for USA and Hondroyiannis (2004) for Greece, among others.

In this study our goal is to conduct a detailed analysis regarding the determinants of residential electricity demand in Pakistan by including the


132

relevant demographic and economic variables. In this respect, we have used cross-section data discussed in detail in section 5.

3. Residential Demand for Electricity in Pakistan Electricity is a commodity which is not directly consumed by the

households. Households get utility from the use of electricity consuming appliances, so the demand for electricity is a derived demand, originating from the demand for services provided by electricity consuming appliances. Use of the appliance may depend on the habits and preferences of the consumers, which are different hence leading to heterogeneity. In our analysis following literature [e.g. Taylor (1975)] we identify short run as a period in which the appliance stock of a household is assumed to be constant, hence the changes in electricity consumption occur due to changes in the utilization rate of the existing appliances. In long run the appliance holding can change3.

In the short run the residential demand for electricity is mainly determined by the price of electricity and the alternative forms of energy, income of the household, family size, number of rooms in the house, demographic factors like rural or urban area, temperature and seasonal factors and the appliance holding of the household.

In Pakistan we have increasing block pricing, this makes modeling demand difficult, hence in the our analysis we will drop the price variable. In our analysis we will not address the complex issue of multistep block pricing; there are two reasons for it. One is the data about the marginal price faced by the consumers is not readily available. Second is the unit prices faced by consumers are uniform thus, this variable lacks the required variability. In our study we have not included the seasonal variable, because of unavailability of data. Our study is thus prone to specification bias because of unavailability of data.

This paper is arranged as follows: section 4 is about the methodology used. Section 5 focuses on the data sources and sample details. Empirical results are summarized in section 6. Analysis of the results is in section 7. Section 8 gives the conclusion.

3 Long run analysis is skipped in this paper due to data availability constraint, since we have only cross sectional data for the short run (SR).

Hussain and Asad

133

4. Methodology In our analysis we will see how the monthly expenditure by

households on electricity is related with a set of given variables, using the OLS technique on cross section data of about 9,500 households. We will estimate the following general form;

( , , , , , )i i i i ik i iQ f Y N F N R D AP D RG D PR= (4.1)

Where

Q = consumption expenditure by household on electricity (Rs/month)

Y = monthly income of the household.

NF = Number of family members.

NR = the number of rooms in the house

DR = dummy showing the region. 1 for Urban, and 0 for rural.

DAP = shows the presence of a particular appliances. Appliances selected are freezer (fzr), fridge (frg), air conditioner (ac), air cooler (aclor), washing machine (wm) and computer (comp). Value of each category is 1 for the presence of the particular appliance, and is 0 otherwise.

DPR = dummy showing the province, i.e. Punjab, Sindh, NWFP and Baluchistan. 1 if the household belongs to the specific province, 0 otherwise.

Following literature we estimate equation in double log form, because in that case the coefficients of the variables will provide the respective elasticities and semi-elasticities. We estimate the following equation:

1 2 3 4 5 6 7 8ln lni i i i i i iQ Y NF NR DR Dsndh Dblch Dnwfpα α α α α α α α= + + + + + + +

9 10 11 12 13 14i i i i i iDfrz Dfrg Dac Dacolr Dwm Dcompα α α α α α µ+ + + + + + + (4.2)

The income elasticity of electricity demand 2α is expected to be positive, because as the income of the household increases their consumption of electricity also increases by consuming more appliances. The semi-elasticities 3α and 4α are expected to be positive, because as the number of family members and rooms in a house increases its electricity consumption is also expected to increase. The coefficients 5α through 14α cannot be interpreted as semi-elasticities. The percentage effects of the dummy variables


134

on the electricity expenditure can be derived by exponential transformation of the coefficients.

The electricity demand of a household depends on the demographic factors. The households living in urban areas are expected to consume more than those in rural areas. Similarly, there is expected to be province wise differences in electricity consumption, to capture these differences we are using dummy variables for each province by using Punjab as base category.

In the initial analysis we take a large sample which includes both urban and rural households. Then we conduct separate analysis for rural and urban regions to see the difference in response of electricity expenditure to the selected set of explanatory variables. It is expected that there will be strong heterocedasticity in the data because of its cross sectional nature. To counter this problem we took the log of the consumption expenditure of electricity and income. Other problems could be the presence of specification bias because of the missing data about the season in which the households were surveyed.

5. Data All the data used are taken from Pakistan Social and Living Standard Measurement Survey (PSLM) Round-1 (2004-05). This survey is conducted by the Federal Bureau of Statistics. The survey following Core Welfare Indicators Questionnaire (CWIQ) approach was conducted with the aim to provide data for use by the government in formulating the poverty reduction strategy as well as development plans at district level and rapid assessment of programs.4

This is the first time that Federal Bureau of Statistics (FBS) has conducted. The field work was carried out between September, 2004 and March, 2005. Simultaneously FBS conducted Household Integrated Economic Survey (HIES) by contacting more than 12000 households for the purpose of collecting detailed information about the income and consumption expenditure of the households. Hence, we have used the same households. But after accounting for missing values and outliers we were left with 9,238 household observations, which include households from all four provinces and from both rural and urban areas of Pakistan. Use of monthly data reduces the possibility of aggregation bias over time.

4 A sample survey covering approximately 76,520 households to provide district level indicators in the sectors such as Education, Health, Water Supply & Sanitation and Household Economic Situation & Satisfaction by facilities and services use.

Hussain and Asad

135

The combined sample (Rural and Urban) has 4,898 households from rural area and 4,340 households from urban areas. Province wise distribution of households included in the combined sample is given in Table 5.1.

Table 5.1: Province wise distribution of households (combined).

Province No. of Observations. Percentage

Punjab 4075 44.1

Sindh 2215 23.9

Baluchistan 1151 12.5

NWFP 1797 19.4

The separate sample used for urban area includes 4,409 households’ observations. The province wise distribution of households included in this sample is shown in table 5.2.

Table 5.2: Province wise distribution of households (Urban).


Punjab 1917 43.5

Sindh 1162 26.3


NWFP 744 13.3

The separate sample used for rural areas include 4,997 household observations. The province wise distribution of households included in the sample is shown in table 5.3.

Table 5.3: Province wise distribution of households (Rural).


Punjab 2760 44.8

Sindh 1094 21.9


NWFP 1094 21.9


136

6. Empirical Results The results of estimation of equation (4.2) for both the rural and urban combined and separate samples are given in table 6.1 below.

6.1. Analysis When we are looking at a cross section data of 9,238 households, it is

obvious that the appliance holding will be having different from one household to the other. Thus our estimated equation for consumption expenditure on electricity will be encompassing the effects of variations in the utilization rate and also the effect of intra-household change in appliance stock. Keeping this in mind our estimated elasticities suggest something both for short run and long run.5

Table 6.1: Estimated results of equation (4.2)

Variable Coefficients Overall Urban Rural

Ln Y 0.153* 0.167* 0.135* NF 3.52* 3.25* 3.86* NR 2.52* 3.76* 1.30*** DR 0.14*

Dsndh -0.02*** -0.01 -0.02 Dblch -0.22* -0.29* -0.14* Dnwfp -0.23* -0.25* -0.22* Dfrz 0.42* 0.47* 0.32* Dfrg 0.35* 0.31* 0.40*

Dacolr 0.06* 0.05*** 0.09** Dwm 0.16* 0.14* 0.19* Dac 0.56* 0.52* 0.84*

Dcomp 0.20* 0.22* C 3.98* 4.03* 4.11*

_2R = 0.38

_2R = 0.426

_2R = 0.211

Note: *,**,*** represent significance at 1%, 5% and10% respectively.

In case of the combined sample, we see that income elasticity is about 0.15, which means that expenditure on electricity consumption is inelastic to the income of the household. 100% increase in income of the household will on average lead to only 15 % increase in the expenditure on electricity. The coefficient associated with the number of family members give the semi- 5 See: Thomas (1987)

Hussain and Asad

137

elasticity. Its value in the case of combined sample is 3.52, which means if on average family size increases by 1 unit i.e. member, the household expenditure on electricity will increase by 3.52%. Similarly, the coefficient associated with the number of rooms in the house, represent semi-elasticity.

Its value is 2.52, which means that a unit increase in number of rooms i.e. one more room, will on average increase the electricity expenditure by 2.52%, this is because of increased expenses on lighting and air circulation.

The rest of the coefficients in our model are the dummy variables, and because our dependent variable is in log form, we cannot interpret the coefficients of these dummy variables as semi-elasticities. To find out the percentage effect of the dummy variables on the dependent variable we have to perform the exponential transformation of the coefficients of these dummy variables. Nevertheless, the sign of the coefficients also explain the effect of the dummy variables. The results show that the electricity expenditure is significantly higher in the urban areas as compared to the rural areas. This is probably because of more chances of electricity theft in rural areas as compared to the urban areas. Another reason could be the greater hours of load shedding in rural areas as compared to the urban areas. Also, there is less trend of using electricity consuming appliances and because of lower income in rural areas the appliance stock the households have is also limited.

Similarly, electricity expenditure is lesser in other provinces as compared to Punjab. The coefficients for Sindh, Baluchistan and NWFP are -0.02, -0.22 and -0.23 respectively. Though we cannot tell about the percentage changes in the electricity expenditure due to change in province but the magnitude of the coefficients is comparable. For example we can see that in Sindh electricity expenditure is slightly lower than Punjab, whereas electricity expenditure of Baluchistan and NWFP is much lower than Punjab. We can also see that the difference in electricity expenditure between Sindh and Punjab is very small, and that too is significant on 10% level significance. The results thus suggest that highest electricity expenditure is in Punjab, then comes Sindh, then Baluchistan and lowest electricity expenditure is in NWFP. This may be because of the non-payments of electricity dues in NWFP which is a common practice in some areas of NWFP and Baluchistan. The expenditure on electricity may also be lower because of more hours of load shedding in those provinces. Also, because these provinces are less developed as compared to Punjab and poverty is higher in those areas, appliance stock of households would be lesser than that of the Punjab.


138

The dummies for the appliances included in the model show that presence of an appliance always contributes positively towards the electricity expenditure. The highest contributor towards the electricity expenditure is air conditioner, followed by freezer, fridge, computer, washing machine and air cooler, respectively.

If we compare the results of urban and rural areas we find that income elasticity of expenditure is higher in urban areas as compared to the rural areas. The income elasticity is about 0.17 in urban and 0.135 in rural areas. It means a unit increase in the income of household living in urban area will increase their expenditure on electricity consumption by 17% whereas by 13.5% in rural areas. This is because appliance stock is expected to be lesser in rural households and trend of electricity consumption is comparatively lesser in rural areas. So, an increase in income will only lead to increase in the utilization rate of the existing lesser stock of appliances, thus showing lesser income elasticity.

The semi-elasticity associated with the number of household members is 3.25 in urban areas and 3.86 in rural areas. This results is different from expected, and cannot be rationalized. The semi-elasticity associated with the number of rooms in the house is 3.76 in urban areas and 1.30 in rural areas. This result is according to the expectations. In rural areas the construction and the degree of electrification of houses is different than those of the urban areas. In urban areas the lighting and air circulation equipment is more frequent and extensive in the rooms as compared to the rural areas. Thus an increase in the number of rooms in an urban household leads to increase in electricity expenditure of about 3.76% as compared to 1.30% in the rural households.

If we look at the province wise differences we see that in case of urban sub sample the expenditure on electricity is not significantly different in Punjab and Sindh, but is lower in NWFP and much lower in the Baluchistan.

Thus, electricity expenditure is lowest Baluchistan in case of urban sub sample. If we look at the province wise distribution of the household electricity expenditure in the rural areas we find that there is no significant difference in Punjab and Sindh, but electricity expenditure is lower in Baluchistan and lowest in NWFP. Thus, in case of rural sample, NWFP shows the lowest electricity expenditure, even lower than the Baluchistan, this may be due to high electricity theft in rural areas of NWFP or due to lack of electrification or load shedding.

Hussain and Asad

139

The appliance dummies we included in our model show that the presence of an appliances always contributes positively towards the electricity expenditure. The highest contributor in case of urban sample is AC, after it come freezer, fridge, computer, washing machine and air cooler, respectively. In case of rural areas the sequence of contribution in order of highest to lowest is; AC, fridge, freezer, washing machine and air cooler respectively. Computer was excluded in the case of rural sample because it appeared insignificant, and only less than 1% of rural households had a computer.

7. Conclusions and Policy Implication In the current study it is attempted to explore the determinants of the

consumption expenditure on electricity by households on entire country as well as on urban-rural basis. For this purpose, we have included the variables including income of household, family size, number of rooms in the house, region, province and electricity consuming appliances like air-conditioner (AC), refrigerator, freezer, computer, washing machine and air cooler. It was found that expenditure on electricity is income inelastic, increase in family size and the number of rooms raises the expenditure on electricity on household level. Households living in urban areas have more expenditure on electricity as compared to the rural households. Households in urban and rural areas of Punjab have more electricity expenditure as compared to those in other provinces. Since the presence of electricity-consuming appliances always contributes positively towards the electricity expenditure. The same evidence is empirically proved here. Air-conditioner and Freezer are the two most powerful contributors. Thus, to control or reduce the demand for electricity, use of air conditioner and freezer must be reduced.


140

References

Akmal, M., & Stern, D. (2001). "Residential energy demand in Australia: An application of dynamic OLS." The Australian National University, October.

Anderson, K. P. (1973). Residential energy use: An econometric analysis. The Rand Corporation (R-1297-NSF), October.

Barnes, R., Robert, G., & Robert, H. (1981). The short-run residential demand for electricity. The Review of Economics and Statistics, 63(4), 541-552.

Dergiades, T., & Lefteris, T. (2008). Estimating residential demand for electricity in the United States: 1965–2006. Energy Economics, 30, 2722–2730.

Filippini, M., & Pachauri, S. (2004). Elasticities of electricity demand in urban Indian households. Energy Policy, 32, 429-436.

Fisher, F. M., & Kaysen, C. (1962). A Study in Econometrics: The Demand for Electricity in the United States. Amsterdam: North Holland Publishishing Co.

Halicioglu, F. (2007). Residential electricity demand dynamics in Turkey. Energy Economics, 29, 199–210.

Halvorsen, R. (1975) Residential demand for electric energy. The Review of Economics and Statistics, 57(1) 12-18.

Holtedahl, P., & Frederick, L. (2004). Residential electricity demand in Taiwan. Energy Economics, 26, 201–224.

Hondroyiannis, G. (2004). Estimating residential demand for electricity in Greece. Energy Economics, 26, 319– 334.

Houthakkar, H. S. (1962). Electricity tariffs in theory and practice. In Electricity in the United States. Amsterdam: North Holland Publishing Co.

Houthakkar, H. S., & Taylor, L. D. (1970). Consumer Demand in the United States. 2nd ed. Cambridge: Harvard Univ. Press.

Hussain and Asad

141

Mount, T. D., Chapman, L. D., & Tyrrell, T. (1973). Electricity demand in the United States: An econometric analysis. Oak Ridge, Tenn.: Oak Ridge National Laboratory (ORNL-NSF-49), June.

Pakistan Social and Living Standard Measurement Survey (PSLM) Round-1 (2004-05). Federal Bureau of Statistic, Statistics Division.

Pakistan, Government of. (2008). Pakistan Economic Survey (2007-08). Finance Division, Islamabad

Parti, M., & Cynthia, Parti. (1980). The total and appliance-specific conditional demand for electricity in the household sector. The Bell Journal of Economics, 11(1) 309-321.

Reiss P. C., & White, M. W. (2001). Household electricity demand, revisited. Stanford University, October.

Taylor, Lester D. (1975). The demand for electricity: A survey. Bell Journal of Economics, 6, 74-110.

Thomas, R. L. (1987). Applied demand analysis. New York: Longman Publishers.

Wilson, J. W. (1971). Residential demand for electricity. Quarterly Review of Economics and Business, 11(1), 7-22.

Zachariadis, T., & Nicoletta, P. (2006). An empirical analysis of electricity consumption in Cyprus. Economics Research Centre, University of Cyprus, Discussion Paper 2006-04, April.

Forman Journal of Economic Studies SUBSCRIPTION FORM

1. I would like to subscribe to Forman Journal of Economic Studies for

………….year(s). 2. Tick one:

Name:………………………………………………………………………… Profession……………………Institution/ Department………………............ Address:……………………………………………………………………… ……………………………………………………………………… Ph. No:………………………… Cell…….……………………… E mail:…………………………….. 3. Delivery by (Tick one): A crossed cheque /demand draft in the name of Forman Christian College (A Chartered University), Ferozepur Road, Lahore-54600, Pakistan; for the sum of Pak Rs. / US $ ……… is enclosed to cover the above subscription. Signature……………………….. Date………………….. 4. Subscription Rate:

Inland Students: Rs. 200 General: Rs. 300 Overseas: US$ 40

Please address your order / mail to: Managing Editor Forman Journal of Economic Studies Department of Economics Forman Christian College (A Chartered University) Ferozepur Road, Lahore-54600, Pakistan. E mail: [email protected] Ph: +92 42 9923 1581-8, Ext: 380 www.fccollege.edu.pk

Institution Individual Student

Airmail Surface Mail

Department of Economics Forman Christian College (A Chartered University)

The Department of Economics

The Department of Economics was established in 1915. The

Department has eighteen faculty members; among them nine hold

PhD degrees. The department has over four hundred students. The

class rooms are fully equipped with all major academic needs. It

provides education in an American style in line with the ever

changing environment of the market.

Degree Programs

• B. S. (Honors) in Economics: 4-year Program • MS in Economics: One Year Program • M. Phil in Applied Economics: 2-year Program

Specialization

Development Economics

Environmental Economics

Applied Econometrics

Commitment to Excellence

“By Love Serve One Another”

Forman

Journal of Economic Studies Manuscript Submission guidelines

1. Manuscripts of Articles and book reviews shall be sent through

soft copy at email: [email protected] 2. All manuscripts should be double spaced by allowing a margin of

1” on the right and 1” the left. (Except footnotes and inset quotations).

3. Manuscripts are accepted on the assumption that they are original and have not been published anywhere else.

4. The articles so far as possible may be organized into the following sections: (a) a brief abstract, (b) introduction, (c) literature review, (d) theoretical framework / methodology, (e) empirical findings, (f) limitations (I) conclusions / summary and policy implications.

5. The first page of the article should contain the title of the paper, name of the author and a footnote showing the affiliation of the author and acknowledgements if any.

6. For the convenience and benefit of referees, detailed calculations and derivations of the model, estimation results reported in the study may also be submitted along with the article.

7. References at the end are to be written in alphabetical order but not numbered and should confirm to the APA Manual of style.

8. References given in the footnotes are numbered consecutively and their listing should also confirm to internal standardize practice.

9. The acknowledgement of an article/Paper does not automatically qualify it for publication. The acceptance of an article is generally communicated after receipt of comments from the referees.

10. The contributors are solely accountable for originality of the article; Intellectual integrity may not be compromised under any circumstances.

forman journal of economic studies vol 8

Documents

azam university

university status

zaman gomal university

nasser fredonia university

usa marwan

usa iideclaration

usa kiyoshi abe hanazono

lahore muhammad idrees