Convergence of Per Capita Output Levels Across Regions of Bangladesh, 1982-97

Contributor Notes

Author’s E-Mail Address: ahossain@sti.imf.org.sg.

This paper has examined the phenomenon of convergence of per capita output levels across regions of Bangladesh during 1982–97. The main finding is that most of the regions of Bangladesh experienced strong convergence of per capita output levels during 1982–91. There are two other findings within the domain of convergence. First, a few poorer regions of the country did not demonstrate any output convergence for the full or part of the sample period. Second, no evidence has been found for regional convergence of per capita output levels during 1991–97 that coincided with opening up the economy.

Abstract

This paper has examined the phenomenon of convergence of per capita output levels across regions of Bangladesh during 1982–97. The main finding is that most of the regions of Bangladesh experienced strong convergence of per capita output levels during 1982–91. There are two other findings within the domain of convergence. First, a few poorer regions of the country did not demonstrate any output convergence for the full or part of the sample period. Second, no evidence has been found for regional convergence of per capita output levels during 1991–97 that coincided with opening up the economy.

I. Introduction

The existence and persistence of regional income disparity, caused by either structural or policy reasons or both, has economic as well as political implications for developing countries. There are a plethora of examples in Asia and Africa where the people of depressed economic regions have demonstrated their resentment and engaged in secessionist movements of one form or the other. Even if regional income disparity does not always lead to a secessionist movement, it can slow down national integration and retard the process of economic development. Therefore, it is widely acknowledged that regional development is as important as national development on the grounds of both equity and political sustainability2

When it gained independence in 1971, Bangladesh was hailed as a homogeneous economic and political entity. Although this was far from reality, its socioeconomic differences across regions remained somewhat hidden until the 1974 famine that caused a havoc to some of the poorer regions, such as Faridpur, Jessore, Mymensingh, Rangpur and Dinajpur. By then it became apparent that neither infrastructural nor socioeconomic development was uniform across regions of Bangladesh. In fact, in the absence of planned development programs, most modern economic activities were concentrated to a few metropolitan centers, such as Chittagong, Dhaka, Khulna and Rajshahi. There were of course a few urban satellites as well, but they were far from being called the growth poles. In short, the North-West regions were in particular economically depressed and socially and culturally underdeveloped, while the economic fortunes of most of the South and South-West deltaic regions were at the mercy of the mighty Padma and its tributaries. All these underdeveloped regions have indeed been the fertile ground of radical politics.3 A large number of underground political parties in fact flourished during the early 1970s and challenged the authority of the Dhaka-based political establishment. Therefore, regional development has become both an economic and a political issue since the early 1970s and also featured prominently in the early strategy of development within a socialist paradigm (Islam, 1977).

Although Bangladesh’s macroeconomic performance since then has not been impressive, the general contention is that its aid-financed development programs have given priority to both infrastructural and social development and that might have benefited the poorer regions more than the richer ones in a relative sense. Moreover, under various administrative decentralization and socioeconomic development programs, there have been regional catching up processes in economic and social affairs, including agricultural modernization, industrialization, urbanization, education and health care. Some sociocultural and political changes are also afoot. The younger generation that observed the political events of the early 1970s and also experienced the trauma of the 1974 famine has become desperate for upward mobility - social, economic and political. This has generated both social and political unrest that has been manifested in various forms. The rise in labor mobility has been one of them. The increased rural-to-urban migration has in particular contributed to the rapid urbanization of the country. It is indeed significant that many younger people have responded to overseas employment opportunities, especially in the Gulf countries4, from the mid-1970s and thus helped to bring changes in the economy through inflows of foreign capital, technology, and entrepreneurial skills and initiatives. As various studies have already indicated, all these factors have jointly and individually unleashed both economic and sociopolitical forces that contributed to a steady growth of the economy at the rate of around 5 per cent per annum from the mid-1970s (Hossain, 1996). This has, in turn, brought a structural change in both output and employment and caused some integration of regional economies.

This paper formally investigates the question of whether there has been any convergence of per capita output levels across regions of Bangladesh since the early 1980s. Insofar as I am aware of, there has been no study on this and other related issues for Bangladesh despite their wide range of policy implications. Importantly, as the Bangladesh economy has gradually been integrated into the global economy since the mid-1980s (Hossain, 2000), the issue that has gained significance is whether economic openness has provided growth impetus to different regions evenly given their factor endowments and structural characteristics. The rest of the paper is organized as follows. Section 2 outlines the phenomenon of convergence in an historical context. Section 3 defines three concepts of convergence that are used for empirical analysis. Section 4 reviews the rank and relative mobility of the regions of Bangladesh during the period 1982 to 1997. Section 5 tests for convergence of per capita output levels across regions of Bangladesh for different time intervals. Section 6 highlights some economic and non- economic developments that provide plausible explanations for aspects of output convergence across regions of Bangladesh. Section 7 summarizes the major findings and draws policy implications within a broader perspective. The paper has an appendix that estimates the speed of convergence of per capita output levels across regions of Bangladesh for different time intervals and also identifies some regions that can be considered outliers for the present purpose.

II. The Phenomenon of Convergence

The word convergence has been used by scholars to mean different things in different contexts (Abramovitz, 1986). However, as Baumol (1986:1075) points out the phenomenon of convergence in economic history has an unambiguous meaning: “that forces accelerating the growth of nations who were latecomers to industrialization and economic development give rise to a long-run tendency towards convergence of levels of per capita product or, alternatively of per worker product”. David Hume suggested technology transfers as an explanation for the tendency of poorer countries or regions to catch up the richer ones: “technology transfer creates a natural tendency for poor countries to converge toward rich ones, and rich countries benefit from the convergence since it creates new markets for their exports and keeps the spirit of emulation alive” (Elmslie, 1995:212). He made his case on the basis of experience of England that was developed by technology transfers from its neighbors and hoped that having lagged behind, “Scotland’s commercial relations with England would one day result in rapid growth of Scottish wealth” (Elmslie, 1995:213). Since then other prominent economists have been impressed with the idea of convergence of income levels across regions or nations. For example the nineteenth century economists like John Stuart Mill and Karl Marx and the twentieth century economic historians like Alexander Gerschenkron were optimistic about convergence of laggard developing nations (De Long, 1988).

The phenomenon of convergence has found a lease of life in modern growth economics that started with the Harrod-Domar or the Solow growth model (Domar, 1946; Harrod, 1939; Solow, 1956). According to the neoclassical growth model, poorer countries with lower capital-labor ratios would grow faster than richer countries with higher capital-labor ratios, especially when all these countries remain open to foreign trade and capital flows. When income differentials result from differences in technology, poorer countries would grow faster than richer countries as technological know-how would flow from technologically advanced countries to technologically lagging countries.5 The neoclassical growth model, however, does not predict an absolute convergence of per capita output levels unless the rate of saving and the growth rate of population, as well as technology, are identical across countries. That means, countries with faster growing populations and lower saving rates are expected to be poorer than countries with lower growing populations and higher saving rates. It is only in a world of fully integrated markets for goods, capital and ideas that one would expect to observe absolute convergence regardless of countries’ saving rates or demographics (Obstfeld and Rogoff, 1996).

Empirical evidence suggests that convergence of income levels is not a universal phenomenon. While some middle-income developing countries - the club of convergence - demonstrated convergence of per capita income levels in recent decades, most low-income developing countries showed little evidence of convergence (Zind, 1991). There are two kinds of explanations given for weak convergence of income levels between the poor and rich countries. The first explanation is that the forces of convergence work well only when political and economic institutions in poor countries are supportive of inward flows of foreign capital and technology.6 The second explanation relies on the new endogenous growth theories that suggest that an initial advantage of a rich country over a poor country in the level of human capital per-worker results in a permanent difference in incomes between them. The logic is that when externalities related to human capital are strong, richer countries achieve sufficiently higher output levels due to high human capital endowment per-worker and are able to maintain their lead indefinitely by generating enough new savings and investment compared with poorer countries (Romer, 1986,1990; Sachs and Larrain, 1993).

The factors that lead to convergence of per capita output levels across countries remain largely valid for convergence of per capita output levels across regions; that is, regions with lower per capita output levels would grow faster than regions with higher per capita output levels provided that all the regions have the same steady-state growth path per capita output. As convergence is essentially a long-term process, this feature becomes prominent at the maturing stage of development. Barro and Sala-i-Martin (1991,1992), for example, found convergence of per capita income levels across the United States during the period 1880 to 1988. When there are less cultural, ethnic, political and religious differences across regions, the speed of convergence is expected to be stronger.

In essence, the phenomenon of output convergence across regions can be justified analytically and also expected to be realized within a reasonably long period of time. Whether it happens in reality is, however, a different proposition. As Cardenas and Ponton (1995:10) point out in the context of Colombia, “convergence is not a linear process or it is not the only force at play behind the process of economic growth”.

III. Three Concepts of Convergence

There are three interrelated concepts of convergence in growth economics: (1) sigma (σ) convergence; (2) beta (β) convergence; and (3) conditional beta (βc) convergence.7 Sigma convergence concerns with cross-sectional dispersion of per capita income or productivity levels; that is, there exists a convergence if the cross-sectional dispersion of per capita income or productivity levels decreases over time.8 Thus the presence of sigma convergence suggests a tendency to equalization of per capita income or productivity levels across regions or economies. The presence of sigma convergence, however, does not necessarily imply the presence of beta convergence, which suggests that poorer countries or regions grow at faster rates than richer countries or regions given that they all have the same steady-state growth path for per capita output (elaborated below). For example, Barro and Sala-i-Martin (1991) point out that cross-sectional standard deviation in log yi,t (per capita output), denoted δt, falls (or rises) over time if it starts above (or below) its long run constant value δ. The value of σt at a particular point of time can be higher than its long-term value if, for example, there are shocks to output (economic and non-economic) that raise temporarily its cross-sectional standard deviation. Subsequently, the value of δt may fall and converge to its long-run value δ that is assumed unchanged. Thus, what does the falling value of sigma suggest is how the distribution of per capita output across regions converged in the past and is likely to behave in the future.

Whether the convergence of per capita output levels, measured by sigma convergence, is due to higher growth rates of poorer regions than the richer ones can be examined by testing for the presence of beta (or conditional beta) convergence. Beta convergence, as defined in the empirical literature, concerns with cross-section regression of the time averaged income growth rate on the initial per capita income level; that is, there exists a beta convergence if the coefficient of the initial per capita income level in a cross-section regression for per capita output growth bears a negative sign. This suggests that countries or regions with higher initial income levels grow slowly than countries or regions with lower initial income levels.

The concept of conditional beta convergence concerns with cross-section regression of the time averaged output growth rate on the initial per capita output level and a set of additional explanatory variables that define the steady-state growth path for per capita output (Barro and Sala-i-Martin, 1991). Within such augmented growth regression, there exists a conditional beta convergence if the coefficient of the initial per capita output level bears a negative sign.

IV. The Rank and Relative Mobility of Regions of Bangladesh

The early 1970s were highly unstable—politically and economically—for Bangladesh. It was only in the late 1970s that some economic and political stability returned to this country. As a prelude to testing for regional convergence of per capita output levels from the early 1980s, a brief review is made here of the socioeconomic condition of people across regions of Bangladesh in and around 1981.

A. Socioeconomic Indicators Across Regions of Bangladesh: 1981

Table 1 shows that there were significant differences in the values of various socioeconomic indicators across regions of Bangladesh in the early 1980s, such as the density and growth rate of population, the adult literacy rate, the degree of urbanization, the incidence of rural landlessness, the intensity of high-yielding-variety (HYV) technology and the level of per capita agricultural output of rural population.

Table 1.

Socioeconomic Indicators Across Regions of Bangladesh, 1981

article image
Sources: Author’s compilation/computation based on Boyce (1987); BBS, Statistical Yearbook of Bangladesh (various years); and BBS, Bangladesh Census of Agriculture and Livestock: 1983-84.Notes:

The inter-censal rural population growth rale is for the period 1971-1981.

The size of household is measured as the average number of persons in a household.

Farm households are expressed as a per cent of all rural households.

Figures for per-capita cultivated area, irrigated area, and bovine (callie and buffaloes) animals are for the year 1983-84.

Figures for HYV (high-yielding-variety) intensity are for AUS and Aman rice combined.

Per-capita agricultural output (Bangladesh = 100)

n.a. Not available

B. The Rank and Growth of Per Capita Output Levels Across Regions of Bangladesh: 1982–97

Table 2 reports the rank of each region by per capita gross regional product at 198–85 market prices for three years: 1982, 1993 and 1997. In 1982, Chittagong and Chittagong Hill Tracts were the richest regions of the country, and Khulna, Sylhet, Patuakhali and Dhaka followed them. During this time Pabna, Rajshahi, Rangpur, Noakhali, Jessore, Faridpur and Comilla were the poorest regions. In 1997, although Chittagong and Chittagong Hill Tracts remained the richest regions, there were at least two regions, Faridpur and Jessore, that made a significant improvement in their relative economic position. Also, there were two traditionally prosperous regions with an agricultural base, Tangail and Mymensingh, that experienced a sharp deterioration in their relative economic position during the period 1982 to 1997. Most other regions experienced some changes in their relative economic positions during the same period.

Table 2.

Rank and the Growth Rate of Per Capita Gross Regional Product at 1984-85 Market Prices, 1982-97

article image
Sources: Data for per capita gross regional product at 1984–85 market prices are taken from BBS (1993) and BBS, Statistical Yearbook of Bangladesh 1997.

Taka is Bangladesh’s currency unit.

The compound growth rate (g) of per capita output (y) for each region was calculated by estimating the following generic regression equation: In yt = α + βt + εt, where α = ln (yt−T),β = ln (1+g), t is time, such that g = eβ - 1, and c is the random error term.

Table 3 shows the distribution of regions by per capita output level relative to the country average for both 1982 and 1997. It is found that there was an upward mobility of some poorer regions during this period. For example the number of regions in the relative per capita output range: 0.90 or less decreased from 10 to 7 and the number of regions in the relative per capita output range from 0.91 to 1.00 increased from 3 to 8. This changed the shape of the distribution of regions from a heavily skewed to the right in 1982 toward a bell-shaped form in 1997 (Figure 1). The country map (Figure 2) reveals that most of the South and Western regions that were depressingly underdeveloped at the time of independence performed better than some of the North and Eastern regions over the sample period. The dotted regions in the map can indeed be roughly considered the growth poles. By contrast, the Northern regions with an agricultural base (shaded with parallel lines in the map) have now become economically depressed.

Figure 1a.
Figure 1a.

Distribution of Regions by Relative Per-capita Output, 1982

Citation: IMF Working Papers 2000, 121; 10.5089/9781451854374.001.A001

Figure 1b.
Figure 1b.

Distribution of Regions by Relative Per-capita Output, 1997

Citation: IMF Working Papers 2000, 121; 10.5089/9781451854374.001.A001

Figure 2.
Figure 2.

The Country Map Showing the Change in the Relative Economic Position of Regions: 1982-97

Citation: IMF Working Papers 2000, 121; 10.5089/9781451854374.001.A001

Source: Adapted from BBS (1986, Vol. II:104).
Table 3.

The Distribution of Regions by Per Capita Output Relative to the Country Average: 1982 and 1997

article image

V. Tests for Convergence

A. Sigma (σ) Convergence

In order to test for sigma convergence the standard deviation (σ) of (log of) per capita output across regions has been computed for each year for the period 1982 to 1997. The calculated sigma value has been found to show a negative trend from 0.38 in 1982 to 0.31 in 1997. This was, however, not a linear process. In fact the following non-linear trend equation has been found to fit the data best:

σt(tratio)=0.359(32.96)0.013(4.52)T+0.006(3.79)T2
Sample:19821997 R2=0.62 DW=1.33(T=1 for 1982,2 for 1983 and so on)

Figure 3 plots the mean value of (log of) per capita real gross regional product (LY-Hat) ± 2*SIGMA. It also reveals the presence of sigma convergence of per capita output levels, especially during the early years of the sample period.

Figure 3.
Figure 3.

LY - HAT + or - 2 * Sigma of (Log of) Per-capita Gross Regional Product

Citation: IMF Working Papers 2000, 121; 10.5089/9781451854374.001.A001

B. Beta (β) Convergence

As indicated earlier, sigma convergence does not necessarily imply a beta convergence. The latter phenomenon needs to be tested by running a growth regression for per capita output across regions. Baumol (1986) has estimated a growth equation of the following kind to test for β convergence in the historical data (1870–79) of per capita output for 16 industrialized countries:

ln(yi,t/yi,tT)=α+βln(yi,tT)+uit(1)

where yi, is per capita output in country i, ui is an error term, the subscript t indicates the end of the time interval and t-T is the beginning of the time interval. In this regression equation a significant negative sign of β suggests the presence of beta convergence in per capita output levels across countries: that is, countries with lower levels of initial per capita output grow faster than countries with higher levels of initial per capita output.

Although such a simple test for convergence has generated controversy (De Long, 1988; Quah, 1993), Baumol (1986:1076) argues that “a relationship such as that… for [16 industrialized countries] is no tautology, nor even a foregone conclusion”. What is the theoretical basis for such a statistical relationship? Following the tradition of David Hume and others, Baumol interprets it as an evidence of the technological catching up hypothesis. However, Baumol’s specification is also equivalent to the transitional dynamics of per capita output that can be derived from the neoclassical growth model for closed economies (Ramsey, 1928; Solow, 1956; Cass, 1965; Koopmans, 1965). If all countries have the same steady-state growth path for per capita output, then those countries with lower initial income levels would grow faster than countries with higher income levels. Barro (1991) also points out that the tendency of low-income countries to grow at higher rates than the rich countries is reinforced in extensions of the neoclassical models that allow for international mobility of capital and technology.

C. Empirical Results with Bangladeshi Data

Equation (1) has been estimated by OLS using the per capita output growth rate as the dependent variable for different time intervals. Table 4 reports two sets of beta estimates. The first is the set of estimates of beta in which 1982 is the beginning of the time interval while the end of the time interval was moved forward from 1990 by one year until the end of the sample period was reached. The estimated beta coefficient bears a negative sign and remains significant whichever time interval is used for calculation of the growth rate of per capita output. The second is the set of beta estimates in which 1991 is the beginning of the time interval while the end of the time interval was moved forward from 1993 by one year until the end of the sample period was reached. The year 1991 was significant for two reasons. First, this was the beginning of a democratic form of government after about 15 years of military or quasi-military rule. Second, although the economy was opened up in the late 1980s, the process got momentum once the civilian government was installed. Unlike the first set of estimates of beta, the second set suggests no evidence of convergence of per capita output levels across regions. This is indeed an unexpected finding that needs further investigation and reconciliation with theory and evidence.

Table 4.

Crosssectional Regressions of the Growth Rate of Per Capita Output

article image
Notes:

This represents the period for which the growth rate of per capita output for each region was calculated.

Significant at the one per cent level.

To begin, one fundamental assumption behind absolute beta convergence is that all the regions or countries have the same steady-state growth path for per capita output. Thus the poorer regions or countries show a tendency to grow faster than the richer regions or countries. However, if the assumption that all the regions or countries have the same steady-state growth path for per capita output is relaxed, then it is possible to provide an explanation of the possibility that the growth rates of the poorer and richer regions remain the same or even the richer regions or countries grow faster than the poorer ones. This follows the idea of relative

convergence; that is, the growth rate of a region or country does not depend on its absolute value of per capita output, but rather on the value measured relative to its own steady-state position.9

D. Conditional Beta (βc) Convergence

As indicated above, there is a shortcoming in the specification (1) for testing for convergence of per capita output levels across regions or countries. Within the framework of the neoclassical growth theory, it imposes the restriction that the steady-state growth path for per capita output is identical for all the regions or countries. This is indeed a restrictive assumption even when the presence of convergence is investigated across regions for an apparently homogeneous country like Bangladesh. There are at least three variables that can be used to define the steady-state growth path for per capita output: (1) the rate of saving/investment; (2) the growth rate of population; and (3) the rate of technological progress. If zero restrictions are imposed on these variables in the specification for the output growth rate, the estimate of beta could be biased and may lead to misleading conclusions. Considering this, Mauro and Podrecca (1994) have estimated an equation of the following form to test for convergence across regions of Italy:

ln(yi,t/i,tT)=α+βlnyi,tT+γlnsit+δlnnit(2)

where si is the share of saving/investment on gross domestic product for region i, ni is the growth rate of population for region i and α, β, γ and δ are parameters to be estimated.

Although the conditioning variables si and ni remain crucial in the specification (2), it is possible to include other region-specific variables to test for their impacts on the growth rate of output for different time periods. For example, Barro and Sala-i-Martin (1991) have used a number of additional variables in a non-linear beta specification (see Appendix) and tested for convergence of per capita output levels across the United States and countries.

Table 5 reports the results of a number of regression equations with variables that are expected to determine the steady-state growth path for per capita output. The variables used for experimental purposes included the growth rate of population, the net migration rate, rainfall, the share of agriculture in output, the intensity of high-yielding-variety technology and the literacy rate.10 For lack of data, the rate of saving/investment has not been used for experimentation. Both t and F-test results are reported to determine whether these variables, individually or jointly, are significant in the growth equation.

Table 5.

Conditional Beta Convergence Regressions of Per-Capita Output+

article image
Notes: Equation 1 = Conditional beta convergence regression equation with the population growth rate (Ln ni).Equation 2 = Conditional beta convergence regression equation with the population, growth rate (Ln ni) and the literacy rate (Ln li).Equation 3 = Conditional beta convergence regression equation with the population growth rate (Ln ni) and the net migration rate (Ln migi).Equation 4 = Conditional beta convergence regression equation with the population growth rate (Ln ni), the net migration rate (Ln migi), and the HYV technology (Ln hyvi).Equation 5 = Conditional beta convergence regression equation with the rainfall and the HYV technology (Ln hyvi).Equation 6 = Conditional beta convergence regression equation with the rainfall, the HYV technology (Ln hyvi) and the share of agriculture in output (Ln sai).+ I(II) represents the regression equation for the growth rate of per-capita output for the time interval 1982-1997 (1982-1993). The figures in parentheses below the the coefficients are absolute t-ratios.

A: Serial correlation; B = Functional form; C: Normality, and D: Heteroskedasticity.

Significant at the one percent level.

Significant at the five percent level.

As Bangladesh remains predominantly an agricultural country, rainfall, the HYV technology and the share of agriculture in output can be considered key variables in the growth equation. Of these variables, rainfall and the HYV technology have been found to have significant impact on the growth rate of output. Note that rainfall is generally used as a proxy for weather and weather related factors (such as floods) that affect the growth rate of output. However, it is difficult to disentangle the net effect of floods and related variables on the growth rate of output as they have both positive and negative effects on production. For the present purpose, the net effect of rainfall has been found to have a negative effect on the growth rate of output. The share of agriculture in output (a proxy for production structure) does not have a significant impact on output growth for the complete sample period.

Three other variables that have been used for experimentation are the net migration rate, the population growth rate and the literacy rate. The coefficient of the net migration rate bears a positive sign but is not significant at the conventional 5 per cent level. Note that the data used for this variable are for the 1970s as data for later years are not available at the regional level. The later discussion will suggest that this variable remains important in the specification for the growth rate of output as labor mobility has increased significantly in Bangladesh since the early 1980s.11 Contrary to the a priori expectations, neither the population growth rate nor the literacy rate has been found to have any significant impact on the growth rate of per capita output.

While all the conditioning variables in the growth equation provide some information on factors that explain variation in regional output growth, the variable of interest for the present purpose is the initial income level. The coefficient of this variable bears a negative sign and remains significant irrespective of which conditioning variables are used in the regression. This result is consistent with the view that there was conditional beta convergence of per capita output levels across regions of Bangladesh, especially during the 1980s.

VI. Factors Behind Convergence of Per Capita Output Levels Across Regions of Bangladesh

The phenomenon of convergence in growth economics remains controversial as there is no consensus on factors that lead to convergence of output levels across regions or countries. Economic theorists postulate that both economic and non-economic factors cause the poorer regions to grow faster than the richer ones. As indicated earlier, there are two broad analytical frameworks that can explain the phenomenon of output convergence across regions or countries. The first is the technological catching up hypothesis that emphasizes the spread of technology from the technologically-advanced regions to the technologically-backward regions as a factor causing convergence of per capita output levels. David Hume considered foreign trade as the major conduit of technology transfers and argued for freer trade and other laissez-faire policies with the hope that such policies would promote convergence of incomes across regions or countries. Within the modern context, a body of literature has grown around this hypothesis (Abramovitz, 1986,1990; Baumol, 1986; Baumol and Wolff, 1988; De Long, 1988; Dowrik and Nguyen, 1989; Maddison, 1987,1991; and Soete and Verspagen, 1993). The second hypothesis is derived from the transitional dynamics of the neoclassical growth model. This analytical approach has become prominent in the works of Mankiw, Romer and Weil (1992), Barro (1991) and Barro and Sala-i-Martin (1991,1992,1994). They show that if different regions or countries have different capital-labor ratios, their growth paths would eventually converge to a steady-state growth path because of diminishing returns to reproducible capital but would exhibit different growth rates during the transition phase. The process of such convergence relies on natural forces in markets and hence any failure of convergence could be attributed to distortions in both labor and capital markets. Along with these two analytical approaches, there is another explanation for convergence of per capita output levels across countries that economic growth is contagious (Baumol, 1994). The phenomenal growth of East Asian economies has given some credibility to the phenomenon of convergence based on contagion.

This section reviews major economic and non-economic developments in Bangladesh from the early 1970s with a view to identifying factors that may provide explanations for convergence of per capita output levels across regions, especially in the 1980s. It then highlights the rapid opening up the economy since the early 1990s that might have contributed to the slowing down of the process of convergence.

A. Diffusion of the HYV Technology

The introduction of the modern methods of agricultural production (or simply the HYV technology) has been the major agricultural development in Bangladesh since the mid-1970s. Until then, agricultural production was heavily dependent on monsoons (Faaland and Parkinson, 1976). The diffusion of the HYV technology was both supply and demand determined. The government provided subsidies on seeds, irrigation facilities and pesticides, and low interest bank credits to facilitate the diffusion of this technology (Hossain, 1988). At the same time, there were at least two reasons why farmers became receptive of this technology. First, the 1974 famine created an urgency for raising the quantum of food production to feed the rapid growth of population. The introduction of the HYV technology then became necessary for raising food production in a land-scarce country. Second, from both the private and public sources, farmers became aware that high rates of return could be gained from the introduction of the HYV technology in food production.

Although the HYV technology spread rapidly throughout the country from the mid-1970s, as of 1977 the rate of its adoption varied significantly across regions (Table 6). Some regions, such as Chittagong, Chittagong Hill Tracts, Comilla, Noakhali and Kushtia, already had high rates of adoption of this technology. By contrast, in most of the poorer regions, such as Faridpur, Jessore, Rajshahi and Rangpur, the rate of its adoption was low. There was thus a technological catching up process in operation in the poorer regions from the mid-1970s. As agriculture contributed more than 50 per cent of GDP in most of the poorer regions, the contribution of the HYV technology to the growth of output was higher in these regions than the richer ones.

Table 6.

Diffusion of the HYV Techonology in Agriculture Across Regions of Bangladesh, 1977–84

article image
Notes and Sources:

HYV intensity in 1977: Author’s computation based on BBS, Statistical Yearbook of Bangladesh, and BBS, Agricultural Yearbook of Bangladesh.

Percentage change figures for irrigated area to cultivated area: BBS (1986: Vol. 1).

B. Development of Economic and Social Infrastructure

Bangladesh has received about US$30 billion as foreign aid and loans since its independence. The bulk of this foreign aid and loans has been used for the development of economic and social infrastructure. There are two main reasons why infrastructural development got priority in the strategy of development. First, as there was heavy damage to infrastructure during the Independence War and later by frequent floods and other natural disasters, the government spent the major portion of its aid resources for infrastructural development. Second, the government spent a portion of its aid resources for public works programs that were designed to create non-farm employment opportunities for vulnerable groups in the rural community. The distribution of aid resources for public works programs, however, was not uniform across regions. Some traditionally underdeveloped regions in the Western part of the country received priority in infrastructural development. Most of the metropolitan areas and industrial growth centers also received large shares of resources for infrastructural development. While there were economic reasons for such a non-uniform distribution of developmental resources, it is to be emphasized that political factors were also behind various developmental or public works programs. Chittagong, for example, remains an economically and politically strategic region. It receives priority in infrastructural development irrespective of the color or form of government.

Table 7 shows that the growth of road length (a proxy for public works programs) varied significantly across regions during the period 1986 to 1994. Although it cannot be substantiated statistically with a high degree of confidence, it can be argued that variation in developmental activities led to differences in economic performance across regions. Despite such arbitrariness of developmental and public works programs, they in general contributed to integration of rural economies with the urban centers. For example, one visible economic progress over the past two decades has been the development of transport and cornmunication facilities. Most rural areas are now connected with the urban centers through semi-pucca roads that are used for plying of rickshaws and different types of motorized or wheeled transports. Similarly, as Bangladesh remains a riverine country, the introduction of motorized boats and other water transports has in some sense revolutionized the inland transport system. The impact of improvement in transport and communication facilities on economic activity has been substantial, especially in the backward areas.

Table 7.

Road Length and Primary Schools Across Regions of Bangladesh

article image
Source: BBS, Statistical Yearbook of Bangladesh (various issues).

In addition to aid-financed development activities, there has been an upsurge of construction activity in the private sector from the mid-1970s. This has been partly due to large-scale remittances of workers in the Gulf countries. Note that beginning from the late 1970s, about fifty thousand workers went mostly to the Gulf countries. The number of emigrating workers has increased to about 200 thousand per annum since the early 1990s. Along with about US$2 billion aid and loans, Bangladesh receives almost an equal amount of workers’ remittances each year - both officially and unofficially. One main item of expenditure from remittances has been the improvement and construction of housing in both the rural and urban areas. While such construction activities benefited the different regions differently, their marginal impact on economic growth has been high in the poorer regions.

Along with development of physical infrastructure, there has been considerable expansion of educational, health and other service facilities across regions since the early 1970s. Table 7 shows that the number of primary schools has increased by 40 per cent during the period 1986 to 1994. The numbers of secondary schools, colleges, and universities have also increased significantly. Two factors have particularly facilitated the expansion of educational institutions. First, with the democratization of politics, politicians of all colors, especially businessmen-turned-politicians, have helped to build educational institutions, health facilities and community centers as part of establishing their political credential. Second, there has been a proliferation of non-government educational institutions and health services throughout the country. It is to be noted that more than 8 thousand non-government organizations (local and foreign) are now active in various service-providing sectors. Although the effectiveness of these organizations in the provision of services vary significantly, it is widely accepted that these organizations have made a substantial contribution to basic literacy, health care, and other social development. The poorer regions that lacked educational facilities have particularly benefited the most from activities of the non-government organizations.

C. Labor Mobility

For reasons suggested above, one significant development in Bangladesh from the early 1970s has been the rise in labor mobility from both the rural to urban areas and the relatively underdeveloped to developed regions. The big cities have particularly attracted a large number of people from the countryside. Within the framework of the neoclassical growth theory, the emigration of labor from the poorer regions with low capital-labor ratios to the richer regions with high capital-labor ratios speed up convergence of per capita output levels, provided that migrants possess no human capital12 (Barro and Sala-i-Martin, 1991).

Table 8 reports data for the net migration of population across regions of Bangladesh during two census periods 1961-74 and 1974-81. It shows that during the 1960s and 1970s there were large-scale in-migrations to Sylhet, Chittagong Hill Tracts, Dhaka, Khulna and Dinajpur, while Noakhali, Faridpur, Tangail, Mymensingh, Patuakhali and Pabna experienced large-scale out-migrations. Data for the net-migrations during the 1970s have been used to examine their impact on convergence of per capita output levels across regions.

Table 8.

Net Migration of Population Across Regions Between 1974 and 1981

article image
Source: BBS, Statistical Yearbook of Bangladesh, 1985 and 1989.

Data for net migration across regions during the 1980s and 1990s are not available.

+/- indicates net in migrants/out-migrants. The sum of net migrants of all regions is not zero due to rounding and other statistical errors.

Mymensingh includes Jamalpur.

The regression results reported in Table 5 show that the coefficient of the net migration of population in the specification for conditional beta convergence bears a positive sign. It appears that the impact of the net in-migration of population on the growth rate of output in regions where they settle is positive. This finding is consistent with the idea that immigrants are generally young and possess above average human capital. While this could be a plausible explanation, it is to be recognized that the regions that grow faster also attract people from the slow-growing regions.

D. Economic Openness

While the factors mentioned above appear to have contributed to convergence of per capita output levels across regions during the 1980s, the regression results in Table 4 show that the process of convergence weakened, if not stopped, during the period 1991 to 1997. As indicated earlier, the Bangladesh economy has been opened up rapidly since the early 1990s and this could be a factor behind the slowing down of output convergence across regions.

Here follows an elaboration of this point. Note that economic openness generally promotes output growth and accelerates the speed of income convergence across countries (Ben-David, 1993; Levine and Renelt, 1992). This could be true for Bangladesh as well. However, it appears that the richer regions gained the most from economic openness due to both structural and policy reasons. For example, foreign investment in Bangladesh has generally been concentrated to developed regions, such as Dhaka and Chittagong. Modern technology transfers to these regions have also been considerable. Thus the factors that promoted output convergence across regions during the 1980s might have been neutralized to some extent following the opening up the economy. Recall that although the share of agriculture in output was not significant in the growth equation for the whole sample period, it was found to bear a negative sign and was significant in the growth regression for the period 1991 to 1997 (this result is not reported). This suggests the possibility that the opening up the economy has changed the steady-state growth path of per capita output levels for different regions as new factors such as knowledge-based production technology, export-oriented investment and infrastructural development have started to play a critical role in economic growth. To the extent that these factors have become important, it is possible to explain the weakening of the process of regional convergence. Note that the speed of convergence in a region depends on the gap between its actual and steady-state per capita output levels. As this gap might have narrowed over time for those regions whose steady-state per capita output levels did not rise, they have experienced lower rates of economic growth. By contrast, there are regions whose steady-state per capita output levels increased due to factors that have become prominent in the growth process after opening up the economy and this could have led to higher rates of growth in these regions despite their relatively higher initial output levels.

VII. Summary and Conclusion

This paper has examined the phenomenon of convergence of per capita output levels across regions of Bangladesh for the period 1982 to 1997. Three concepts of convergence, such as sigma (σ) convergence, beta (β) convergence and conditional beta (βc) convergence have been investigated. The results suggest that most of the regions of Bangladesh experienced strong convergence of per capita output levels for the period 1982 to 1991. The estimated speed of convergence for this period has been found to be around 3 per cent per annum (Appendix). However, along with this positive feature of development, the paper has identified two related developments within the domain of convergence that have important policy implications. First, a few poorer regions of the country did not demonstrate any output convergence for the full or part of the sample period. Second, no evidence has been found for convergence of per capita output levels across regions for the period 1991 to 1997 that coincided with opening up the economy.13 The paper has highlighted both economic and non-economic developments within the country since the early 1970s and thus provided plausible explanations for all these features of convergence of output levels across regions.

It is possible to draw policy implications from empirical findings within a broader perspective.14 At the early stages of development, convergence of national output performance vis-à-vis other countries is as important as convergence of output levels across regions. Historically, economic openness has been important for convergence of national outputs across countries. From this viewpoint, the opening up Bangladesh’s economy since the early 1990s has been a major step towards its goal of achieving rapid economic growth. At the same time it is necessary to remember that at the early stages of a country’s development (and international convergence), its modern economic activities remain concentrated around urban centers, which, by nature, offer the strongest incentives for investment- both domestic and foreign. This may weaken the process of regional convergence of output levels. It is even possible that international convergence is associated with regional divergence of output levels for a considerable period of time. However, over-time, rural-urban migration and intensifying capacity constraints are likely to lead to a progressive spillover of benefits to the relatively less urbanized areas. It is possible to draw two policy implications from such a pattern of development. First, opening markets for foreign products and investment tend to favor the country’s higher income households and industrialized regions, implying that scale economies and spillover effects play an important role in attracting foreign direct investment. Second, safety nets and redistribution policies may need to be put in place not just to share the growth dividends from opening up the economy but also to make the process politically sustainable. Insofar as Bangladesh, policymakers should take into account the structural features of poorer regions where the process of output convergence has slowed down, if not stopped, and then introduce remedial measures for expediting their integration with the developed regions. The government should particularly take measures to enhance human capital formation and to induce technology transfers to poorer regions so that they can reap the benefits of economic openness. Rapid infrastructural development is another bet for attracting investment in the poorer areas. This may not only promote regional convergence but also relieve increasing pressure on congested urban centers.

Convergence of Per Capita Output Levels Across Regions of Bangladesh, 1982-97
Author: Mr. Akhtar Akhtar Hossain
  • View in gallery

    Distribution of Regions by Relative Per-capita Output, 1982

  • View in gallery

    Distribution of Regions by Relative Per-capita Output, 1997

  • View in gallery

    The Country Map Showing the Change in the Relative Economic Position of Regions: 1982-97

  • View in gallery

    LY - HAT + or - 2 * Sigma of (Log of) Per-capita Gross Regional Product