The Elusive Quest for Inclusive Growth
Growth, Poverty, and Inequality in Asia
Author: Mr. Ravi Balakrishnan, Mr. Chad Steinberg, and Mr. Murtaza H Syed

Contributor Notes

Author’s E-Mail Address: rbalakrishnan@imf.org; csteinberg@imf.org; msyed@imf.org
Publication Date:
26 Jun 2013
eISBN:
9781475531169
Language:
English
Keywords:
WP; Gini coefficient; labor market; financial market; venture capital; market clearing wage; export heartland; quintile income; poverty incidence
Download PDF (1.7 MB)

This paper assesses how pro-poor and inclusive Asia’s recent growth has been, and what factors have been driving these outcomes. It finds that while poverty has fallen across the region over the last two decades, inequality has increased, dampening the impact of growth on poverty reduction. As a result, relative to other emerging and developing regions and to Asia’s own past, the recent period of growth has been both less inclusive and less pro-poor. Our analysis suggests a number of policies that could help redress these trends and broaden the benefits of growth in Asia. These include fiscal policies to increase spending on health, education, and social safetynets; labor market reforms to boost the labor share of total income; and reforms to make financial systems more inclusive.

Abstract

This paper assesses how pro-poor and inclusive Asia’s recent growth has been, and what factors have been driving these outcomes. It finds that while poverty has fallen across the region over the last two decades, inequality has increased, dampening the impact of growth on poverty reduction. As a result, relative to other emerging and developing regions and to Asia’s own past, the recent period of growth has been both less inclusive and less pro-poor. Our analysis suggests a number of policies that could help redress these trends and broaden the benefits of growth in Asia. These include fiscal policies to increase spending on health, education, and social safetynets; labor market reforms to boost the labor share of total income; and reforms to make financial systems more inclusive.

I. Introduction

Income inequality has risen across the world over the last two decades. The academic literature attributes the rise mainly to three factors: globalization, skill-biased technical change, and the decreasing bargaining power of workers. The global crisis and recent social turmoil in different parts of the world have heightened awareness of the potential impact of rising inequality on economic and social stability and on the sustainability of growth. Such concerns have not bypassed Asia, with policymakers throughout the region looking for ways to arrest rising inequality and make growth more inclusive.

This paper examines how pro-poor and inclusive Asia’s recent growth has been compared to its own history and other emerging regions, what factors lie behind these outcomes, and which policies could be considered to help make growth more inclusive in the region. The main findings are that poverty has fallen in recent decades in Asia but inequality has increased, and that the rise in income inequality has dampened the impact of growth on poverty reduction. Relative to other regions and to Asia’s own past, the recent period of growth has been both less inclusive and less pro-poor. There is scope for policy measures to broaden the benefits of growth, notably enhanced spending on health and education, stronger social safety nets, labor market interventions, financial inclusion, and strengthened governance.

The rest of this paper is organized as follows. Section II motivates our research by comparing recent trends in poverty and inequality in Asia to those in other regions of the world. Section III proposes ways to quantify how pro-poor and inclusive growth is in any economy and uses a regression approach to assess Asia’s performance on these metrics. Section IV analyzes what factors contribute to making growth pro-poor and inclusive. On this basis, section V proposes potential policy interventions for broadening the benefits of growth. Section VI concludes.

II. How Does Asia Compare to Other Regions?

Over the last two decades, growth in most Asian economies has been robust and higher on average than in other emerging regions (Figure 1). In turn, this has translated into significant reductions in poverty; nevertheless, Asia is still home to the largest number of the world’s poor, with China and India together accounting for almost half (Table 1 and Box 1).

Table 1.

Number of People Living on Less Than $1.25 per Day

(At 2005 PPP prices)

article image
Source: World Bank, PovcalNet database.
Figure 1.
Figure 1.

Stylized Facts: Asia’s Growth Experience over the Last Two Decades

Citation: IMF Working Papers 2013, 152; 10.5089/9781475531169.001.A001

China and India: Why Does a Rising Tide Not Lift All Boats?

Both China and India have seen considerable poverty reduction since their economic take-offs.

In China, poverty fell fastest during the early 1980s and mid-1990s, spurred by rural economic reforms and low initial inequality. With a relatively equal allocation of land—through land use rights rather than ownership—agricultural growth unleashed by the rural economic reforms of the early 1980s translated into rapid poverty reduction. High access to health and education opportunities also ensured that the subsequent nonfarm growth was poverty reducing. When China’s reforms began, it was one of the poorest countries in the world. In 1981, 84 percent of the population lived on less than $1.25 a day, the fifth-largest poverty incidence in the world. By 2008, this proportion had fallen to 13 percent, well below the developing country average.

Since the mid-1990s, however, the nature of poverty in China has been changing. Growth in the agricultural sector has slowed and the benefits of agrarian reforms have started to dissipate. This has resulted in slower growth in rural employment and incomes, and an increased rural-urban income gap. At the same time, there has also been a rise in urban poverty, partly reflecting large-scale migration from rural areas.

India has also had success in reducing poverty, although at a somewhat slower rate than in China. In 1981, 60 percent of India’s population lived on less than $1.25 a day, lower than in China. By 2010, the share fell to 33 percent, but now two and a half times that in China. While reforms started about a decade later in India, the growth elasticity of poverty reduction has also been lower than in China (as discussed in section III).

High inequality in education and health may have contributed to this outcome. India’s schooling, health and gender inequalities were larger than those of China at the beginning of their reform periods, and, despite significant progress in the last decade, continue to be high. This could have impeded the poor from being able to contribute to, and benefit from, India’s high growth. These detrimental effects may have been especially pronounced since India’s trend growth has been higher in the modern services sector, which is largely urban-based and requires higher human capital.

However, inequality has increased.

The rise has been more dramatic in China. According to official estimates, China’s Gini increased from 37 percent in the mid 1990s to 49 percent in 2008. It has since ticked down to 47.4 in 2012 but remains higher than that in the United States and close to levels in parts of Africa and Latin America. India has also witnessed a rise in inequality, though much less pronounced than in China, with its Gini ticking up from 33 percent in 1993 to 37 percent in 2010 according to the ADB.

However, there are other important dimensions of inequality in India that are not evident in conventional inequality indices based on consumption or income. These are inequalities associated with identity, such as gender or caste, and inequalities in access to education and health. According to some estimates, while growth has benefitted almost every segment of society, poverty reduction has been slower in disadvantaged groups (notably Scheduled Castes and Scheduled Tribes) during the reform period (Thorat and Dubey, 2012). This is consistent with the econometric results presented in section III.B.

In both China and India, a significant part of the rise in inequality reflects a widening of disparities between rural and urban areas as well as between regions. Estimates suggest that spatial disparities account for between one-third and two-thirds of overall inequality in China and India (Asian Development Bank, 2012). In China, the rural-urban income gap has increased significantly since 1998, reaching a ratio of more than 3:1. Notwithstanding a slight decline since 2009, this gap remains high by international standards and is estimated to explain almost half of overall inequality in China. For most other Asian economies, the ratio falls between 1.3–1.8 (Eastwood and Lipton, 2004). At the same time, the historically slower pace of income growth in central and western regions, compared to the export heartlands on China’s eastern coast, has also opened up income gaps among regions. Similar patterns have been observed in India, where the ratio of urban to rural per capita consumption increased continuously from around 1.5 in 1987–8 to nearly two in 2009–10. At around three, rural-urban per capita income differentials are even larger (Sen and Himanshu, 2005) and regional disparities have also increased recently: for instance, the ratio of the per capita GDP of the richest major state (Punjab) to that of the poorest major state (Bihar) rose from 2.9 in 1980 to 4.1 in 2010.

The existing literature highlights a variety of potential causes of this rising inequality, including:

Health and education spending. Fiscal decentralization is much higher in China than in OECD and middle-income countries, particularly on the spending side. More than half of all expenditure takes place at the sub-provincial level, including social spending, but they lack own-revenue sources. The result has been that poor villages cannot afford to provide good services, and poor households cannot afford the high private costs of basic public services. Public spending per capita in the richest province is almost 50 times that in the poorest. Similar patterns are observed in India, putting an onus on central efforts to assure greater fiscal redistribution to poor regions from better-off ones.

Declining labor share of income. Across most of the OECD as well as Asia, the last two decades have seen a decline in the income share of labor and a rise in that of capital—in the case of China and India, the share of labor income to manufacturing value added fell from 50 percent in both countries during the early 1990s to around 40 and 25 percent, respectively, by the mid 2000s according to the Asian Development Bank (2012). This contributes to inequality, since capital income tends to be less evenly distributed than income from basic wage labor. It is partly the result of technological change that has raised the return to capital and lowered the employment elasticity of growth—between 1991 and 2011, this elasticity fell from 0.44 to 0.28 in China and from 0.53 to 0.41 in India. This has been exacerbated in the case of China by an artificially low cost of capital. In both countries, the pool of surplus labor in rural areas has also reduced the bargaining power of workers, contributing to holding down wages relative to productivity. In India, for instance, between 1990 and 2007, while labor productivity rose by nearly 7½ percent annually, real wages grew by only 2 percent per year (Kumar and Felipe, 2010).

Unbalanced regional development. The coastal regions, China’s export heartlands, have provided more opportunities for nonagricultural employment and income. In India, coastal states have also fared better than inland ones. In both cases, this was partly the result of geographical advantages but compounded by preferential policies as well as persistent disparities in human capital and infrastructure (Fan, Kanbur, and Zhang., 2009).

Skill premia and increasing returns to human capital. Between 1988 and 2003, wage returns to one additional year of schooling increased in China from 4 to 11 percent (Zhang and others, 2005) and disparities in educational attainment beyond primary school have emerged. In India, schooling inequalities are even larger and have inhibited pro-poor growth (Ravallion and Datt, 2002).

Financial exclusion. For both China and India, several empirical studies suggest that uneven access to financial services has contributed to inequality. For instance, Zhang and others, 2003 find that after controlling for other factors—such as provincial infrastructure, institutional transition in rural areas, and degree of international integration—differential financial development and urban biases in lending have contributed significantly to the rise in China’s urban-rural income disparity since the late 1980s. In the case of India, Ang (2008) finds that underdevelopment of financial systems hits the poor more than the rich, resulting in higher income inequality.

Moreover, inequality has increased across Asia—in sharp contrast to the previous three- decade record of fast and equitable growth in Japan, the Newly Industrialized economies (NIEs), and the ASEAN. While some decline in the impact of growth on poverty is to be expected as poverty rates fall, in Asia this has been exacerbated by the larger rise in inequality than in other emerging regions. Earlier work (IMF, 2006) attributes this rise in inequality to skill-biased technological change and the transition from agriculture to industry for lower-income Asian economies (consistent with the Kuznets hypothesis).2 At the same time, even as the size and purchasing power of Asia’s middle class has grown in the last two decades, their share of overall income has fallen while that of the richest quintile has increased. By contrast, in Latin America and the Middle East and North Africa, the share of the richest quintile has fallen.

More recently, poverty has generally continued to fall in Asia, but the global crisis has exacerbated the rise in inequality in several economies for which data are available (Figure 2). This trend has been particularly pronounced in rural China and Indonesia, but has also been observed for Japan and some NIEs.

Figure 2.
Figure 2.

Selected Asia: Change in Poverty and Inequality during Global Crisis

(In percentage points)

Citation: IMF Working Papers 2013, 152; 10.5089/9781475531169.001.A001

Sources: CEIC Data Company Ltd.; PovcalNet database; WIDER income inequality database; national authorities and IMF staff calculations.

III. The Links Between Growth, Poverty, and Inequality

Going beyond the stylized facts, regression analysis can be used to quantify how pro-poor and inclusive growth has been in Asia relative to other emerging regions.3

A. What is Pro-Poor and Inclusive Growth?

There are various ways to interpret what it means for growth to be inclusive and pro-poor. In this paper, we follow the Ravallion and Chen (2003) approach and define growth as pro-poor simply if it reduces poverty. Inclusive growth, on the other hand, is defined as growth which is not associated with an increase in inequality, following Rauniyar and Kanbur (2010). In particular, we define growth as inclusive when it is not associated with a reduction in the income share of the bottom quintile of the income distribution.

B. How Pro-Poor is Growth?

To examine the relationship between poverty reduction and growth, the following regression is estimated:

lnPi,t=γi+βi,d⁢lnyi,t+δ⁢ln⁢G⁢I⁢N⁢Ii,t+ρd+εi,t(1)

where Pi,t is the poverty headcount below the $2 line in country i at time t, γi is a country dummy, yi,t is per capita income in country i at time t, GINIi,t is the GINI coefficient in country i at time t, and ρd is a set of decade dummies. As the equation is in logs, β gives the impact of income growth on poverty reduction, and δ gives the impact of a change in the Gini coefficient. β is allowed to vary across country and decade.

The regression model follows the literature that argues while per capita income growth is a key factor, the same rate of growth can bring very different rates of poverty reduction, meaning that other factors matter. In particular, factors that change the income distribution (e.g., shocks to agricultural incomes, changes in tax regimes, etc). Thus following Ravallion and Chen (1997), we allow poverty to also depend on the gini coefficient, which proxies for the underlying factors causing a change in the distribution of income. One can think of growth in average income shifting the income distribution and changes in inequality modifying the shape of the distribution, both of which can affect the poverty headcount (the cumulative distribution below a line at a particular income level, in this case the 2 dollar line).

To estimate the fixed effects, we need to choose a set of benchmark countries. Since we are mainly interested in comparing Asia with Latin America, we include all other countries in other emerging/developing regions in the benchmark. We also use an instrumental variables approach to take account of endogeneity bias and potential measurement error in the income variable. In particular, we use lags of real per capita income as measured in the Penn World Tables (PWT) to instrument the household-survey-based average income variable.4 Specifically, the lagged variables help correct for endogeneity bias by identifying the component of income that is predetermined, and the PWT measure of income help corrects for measurement error by identifying the component of income as measured by the household survey that is also consistent with this secondary measure of income. As both endogeneity bias and measurement error are relevant, the direction of the bias in the estimates that are not instrumented is uncertain.

The regression analysis presented in Table 2 suggests that growth is in general pro-poor, leading to significant declines in poverty across all economies and time periods. Specifically, a 1 percent increase in real per capita income leads to about a 2 percent decline in the poverty headcount (column 1). However, a 1 percent increase in the Gini coefficient more or less directly offsets the beneficial impact on poverty reduction of the same increase in income. Moreover, inequality interacts with income, meaning that a higher level of inequality tends to reduce the impact of income growth on poverty reduction (column 2).5 An increase in the Gini coefficient of about 25 percent (for instance, as observed in urban China from 1995–2005) reduces the impact of a 1 percent increase in income to about a 1½ percent decline in the poverty headcount from 2 percent in the base case. The implication of this result is that past rises in inequality in Asia are likely to reduce the future impact of income growth on poverty, even if the level of inequality remains constant. In addition, the impact of growth on poverty reduction is found to be somewhat lower during the 1990s, possibly due to a changing nature of growth (column 3).

Table 2.

Pro-Poor Growth Regressions1

article image

Dependent variable is the log of poverty headcount below the $2 line. Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1

The relationship, however, varies across regions and economies (columns 4 and 5 and Figure 3). In particular, in East Asia and Latin America, income growth has a significantly lower impact on poverty than in the Middle East and North Africa, Eastern Europe and Central Asia, and sub-Saharan Africa, which make up our baseline economies. The impact is particularly weak in India and Indonesia, where it is significantly less than the impact of an equivalent reduction in the Gini coefficient.6

Figure 3.
Figure 3.

Income Elasticity of Poverty Reduction1

(Impact on poverty headcount, in percent, of 1-percent increase in per-capita income)

Citation: IMF Working Papers 2013, 152; 10.5089/9781475531169.001.A001

Sources: World Bank, PovcalNet, Penn World Tables; and staff calculations.1 The red bars represent countries for which the estimated income elasticity of poverty reduction is significantly different to that of the baseline countries.2 EAP includes Cambodia, Malaysia, Philippines, Thailand, and Vietnam.

C. How Inclusive is Growth?

As a second step, we follow Dollar and Kraay (2002) and look at the relationship between per capita income and a broader definition of “the poor”—the income of the bottom quintile of the income distribution. If the income of the poor tends to rise equiproportionately with average incomes—that is, income growth is not associated with a decrease in the income share of the bottom quintile—then growth would be considered inclusive. Specifically, we use the following panel regression model:

lny⁢p⁢1i,t=θi+λi,d⁢lnyi,t+ηd+εi,t(2)

where yp1i,t is per capita income of the bottom quintile of the income distribution in country i at time t, θi is a country dummy, yi,t is per capita income in country i at time t, and ηd is a set of decade dummies. λ—which is allowed to vary across country and decade—is the elasticity of growth in income of the bottom quintile with respect to growth in average income. This equation can be rewritten as:

lnQ⁢1i,t=θi+(λi,d−1)⁢lnyi,t+ηd+εi,t(3)

where Q1i,t is the bottom quintile share of the income distribution in country i at time t. As equation 3 shows, if λ is less than one, income growth is associated with a decrease in the income share of the bottom quintile: in other words, growth is not inclusive. Equation 3 is the model we estimate. Given that much of the ongoing debate on inclusiveness has not just focused on the poorest fifth of society being left behind, but the richest fifth doing particularly well, we also estimate a similar relationship for income in the top quintile. As with the pro-poor regressions, we use an instrumental variables approach to take account of endogeneity bias and potential measurement error in the income variable.

The results are shown in Table 3. If we simply pool all observations or just use country specific effects, then we get the familiar Dollar-Kraay result that average incomes of the poorest fifth of society rise proportionately with per capita income (column 1), something which also holds for the richest fifth at the 5 percent significance level (column 4). However, once we instrument for the income variable (columns 2 and 5), the results change: income of the bottom quintile rises significantly less than proportionately with average income, and income of the top quintile rises significantly more than proportionately with average income—an important departure from the Dollar-Kray stylized fact. These results also validate concerns that both measurement error and attenuation bias affected the estimates presented in Dollar-Kray.

Moreover, these elasticities vary significantly across regions and countries (columns 3 and 6). For the bottom quintile, the elasticity is significantly less than one for China, the NIEs, and South Asia (excluding India), whereas for Brazil, it is significantly greater than one (Figure 4).7 Turning to the top quintile, the results are the mirror image of those for the bottom quintile (Figure 5).8 The elasticity is significantly greater than one for China and South Asia (excluding India); and significantly less than one for Brazil. In sum, the results suggest that growth has generally not been inclusive in China, the NIEs, and South Asia (excluding India), whereas it has been inclusive in Brazil.9

Figure 4.
Figure 4.

Degree of Inclusiveness of Growth1

(Impact on income of the bottom quintile, in percent, of a 1-percent increase in per-capita income)

Citation: IMF Working Papers 2013, 152; 10.5089/9781475531169.001.A001

Sources: World Bank, PovcalNet, Penn World Tables; and staff calculations.1 The red bars represent countries for which the estimated degree of inclusiveness is significantly different from one.2 EAP includes Cambodia, Malaysia:, Philippines, Thailand, and Vietnam.
Figure 5.
Figure 5.

Degree of Inclusiveness of Growth1

(Impact on income of the top quintile, in percent, of a 1-percent increase in per-capita income)

Citation: IMF Working Papers 2013, 152; 10.5089/9781475531169.001.A001

Sources: World Bank, PovcalNet, Penn World Tables; and staff calculations.1 The red bars represent countries for which the estimated degree of inclusiveness is significantly different from one.2 EAP includes Cambodia, Malaysia, the Philippines, Thailand, and Vietnam.
Table 3.

Inclusive Growth Regression1

article image

Dependent variable is the log share of the income distribution of the bottom/top quintile. Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1

D. How Important is Growth for the Poor?

Using the regression estimates, Table 4 constructs measures of pro-poor and inclusive growth for Brazil, China, India, Indonesia, Russia, and Mexico over recent decades. The table highlights that although the income elasticities of poverty and income of the bottom quintile vary significantly across economies, per capita income growth remains a key driver of income of the poorest fifth of society. Some of the more specific results include:

  • Inequality has widened in China, in contrast to Brazil and Mexico. Yet China has still experienced the greater poverty reduction given its higher growth in average income.

  • The importance of average income growth is reinforced when looking at trends in Indonesia and Russia. For both economies in the 2000s relative to the 1990s, poverty reduction was much greater despite inequality worsening, as growth was much higher.

  • A similar story emerges when looking at measures of inclusive growth. For example, while growth has been half as inclusive in China as in Brazil, the income of the poorest fifth of society has increased by relatively more in China as average income growth has been much stronger.

Table 4.

Pro-Poor and Inclusive Growth Measures

article image
Sources: World Bank, Povcalnet, Penn World Tables; and IMF Staff Calculations.

Set equal to that of the baseline countries when the null of a significant difference.

As proxied by 100 times the change in the log over the corresponding period.

IV. What Determines How Pro-Poor and Inclusive Growth is?

In this section, we try to uncover which factors drive how pro-poor and inclusive growth is. To do this, we first compile a database of structural reform variables (see Appendix 2 for further details). This includes variables designed to capture some of the factors mentioned earlier in the paper (e.g., education, healthcare, labor share of income, share of employment in manufacturing, and openness).10 Then, we estimate regressions similar to equations 1–2, but instead of using fixed effects we allow variation of structural factors across countries to pick up country-specific effects, including on the impact of income on poverty and inclusiveness.

Specifically, we estimate the following equations:

lnPi,t=γx⁢Xi,t+β⁢lnyi,t+βx⁢Xi,t⁢lnyi,t+δ⁢lnG⁢I⁢NIi,t+εi,t(4)
lnQ1i,t=θxXi,t+(λ−1)lnyi,t+λxXi,tlnyi,t+εi,t(5)

where X is the set of structural variables. Our particular interest is in the impact of these factors on the income elasticity of poverty reduction and the degree of inclusiveness (the βxs and the λxs, respectively). As before, a version of equation 5 is estimated for the top quintile as well as the bottom one.

Tables 5–7 show the results of the estimations. In terms of what determines how pro-poor growth is, regressions that add each structural variable one by one suggest that years of schooling, educational spending, credit penetration, trade openness, the labor share and the share of employment in industry all significantly increase the impact of income on poverty, while financial openness reduces it. We do this first as the coverage of the structural variables is uneven.

Turning to the multivariate specifications, the significant positive impact of the share of employment in industry and the negative impact of financial openness survive. An important caveat when interpreting some of the multivariate regressions is the number of observations, which falls significantly given the need to have overlapping data for the structural variables.

Turning to the inclusive growth regressions, both the bottom and top quintile results are similar. Labor share, education spending, years of schooling, industry employment, and financial reform significantly increase the degree of inclusiveness (increase the impact of average income on bottom quintile income and reduce the impact of average income on top quintile income). And in the multivariate regressions, financial reform, education spending, and industry employment remain significant.

To sum up, education and industry employment seem to be important for increasing the impact of income on poverty and inequality. Interestingly, financial openness appears to reduce the effect of income on poverty,11 while financial reform increases the degree of inclusiveness. The robustness of industry employment may at first blush seem to be inconsistent with the Kuznets hypothesis, which is more suggestive of a positive correlation between industry (or manufacturing) employment and inequality at earlier stages of development. However, it is consistent with the idea that labor shifts from agriculture to industry raise the productivity of the agricultural sector where most of the poor are employed, while decreasing relative productivity in industry.

Table 5.

Structural Determinants of How Pro-Poor Growth Is1

article image

Dependent variable is the log of poverty headcount below the $2 line. Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1

Table 6.

Structural Determinants of How Inclusive Growth is (Bottom Quintile)1

article image

Dependent variable is the log share of the income distribution of the bottom quintile. Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1

Table 7.

Structural Determinants of How Inclusive Growth is (Top Quintile) 1/

article image

Dependent variable is the log share of the income distribution of the top quintile. Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1