Economic Growth

Kuznets (1955) predicts that inequality increases with growth in the early stages of development and falls with growth in the later stages. As people move from the traditionally dominant agricultural sector to the modern industrial sector, inequality and polarization increase initially, but decline as the majority of people find employment in the high-income sector. Income disparity shows a similar pattern in Lewis (1954). In the modern sector (human) capital accumulation boosts incomes, whereas in the traditional sector incomes move little owing to surplus labor. Only after the pool of surplus labor is exhausted do incomes converge. Often the traditional and modern sectors are in distinct localities, giving rise to spatial inequality and polarization (Box 5.1).

This inverted U-relationship between income inequality and growth is confirmed for a subset of Asian economies. The Gini coefficient is regressed on (log) GDP per capita (expressed in purchasing power parity or ppp) and GDP per capita squared, using country fixed effects for an unbalanced sample of 11 Asian countries.⁴⁸ The estimated coefficients suggest an inverted U-curve relationship and are significant at the 5 percent level. The curve’s turning point occurs around $4,000 of GDP per capita, suggesting that a number of Asian countries may be approaching a turning point beyond which we may see a decline in inequality.⁴⁹ A similar analysis of wage dispersion also shows a rise in the early stages of development and a fall thereafter. Drawing on Freeman and Oostendorp (2005), wage dispersion is regressed on (logged) GDP per capita (ppp) and GDP per capita squared, using country fixed effects.⁵⁰ The coefficients are highly significant and again describe an inverted U-curve, with a turning point around $6,000 of GDP per capita. However, GDP per capita explains little of the variation in wage dispersion over time and across countries as evidenced by the low R-square.

Panel Regressions for the Gini Index with Country Fixed Effects¹

Source: IMF staff estimates.

¹Standard errors in parentheses. One, two, and three stars indicate significant at the 10, 5, and 1 percent level, respectively.

²No data available for Cambodia.

Panel Regressions for the Gini Index with Country Fixed Effects¹

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Constant	−457.71***	−436.29***	−275.06**	−290.80**	−320.56***	−253.13**	−252.12**
	(112.92)	(103.15)	(119.86)	(119.68)	(93.86)	(114.28)	(114.76)
Log (per capita GDP)	116.08***	111.84***	69.21**	73.27**	86.53***	90.73***	90.30***
	(25.47)	(23.26)	(28.43)	(28.41)	(22.45)	(27.59)	(27.77)
Log (per capita GDP)2	−6.64***	−6.82***	−4.13**	−4.37**	−4.86***	−6.99***	−6.96***
	(1.43)	(1.30)	(1.67)	(1.67)	(1.31)	(1.74)	(1.75)
Trade/GDP		0.288***	0.79***	0.80***	0.33*	0.48**	0.49**
		(0.05)	(0.20)	(0.20)	(0.18)	(0.21)	(0.21)
Trade*GDP per capita			−0.05**	−0.05***	−0.02	−0.03	−0.03
			(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
FDI/GDP				0.34
				(0.21)
Higher educaton share2					−497.33
					(313.98)
Trend						1.96***	1.95***
						(0.48)	(0.48)
							−1.00
							(5.24)
Number of observations	170	170	170	170	161	170	170
Number of countries	13	13	13	13	12	13	13
R-square	0.06	0.05	0.05	0.06	0.21	0.30	0.30

Source: IMF staff estimates.

¹Standard errors in parentheses. One, two, and three stars indicate significant at the 10, 5, and 1 percent level, respectively.

²No data available for Cambodia.

View Table

Panel Regressions for the Gini Index with Country Fixed Effects¹

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Constant	−457.71***	−436.29***	−275.06**	−290.80**	−320.56***	−253.13**	−252.12**
	(112.92)	(103.15)	(119.86)	(119.68)	(93.86)	(114.28)	(114.76)
Log (per capita GDP)	116.08***	111.84***	69.21**	73.27**	86.53***	90.73***	90.30***
	(25.47)	(23.26)	(28.43)	(28.41)	(22.45)	(27.59)	(27.77)
Log (per capita GDP)2	−6.64***	−6.82***	−4.13**	−4.37**	−4.86***	−6.99***	−6.96***
	(1.43)	(1.30)	(1.67)	(1.67)	(1.31)	(1.74)	(1.75)
Trade/GDP		0.288***	0.79***	0.80***	0.33*	0.48**	0.49**
		(0.05)	(0.20)	(0.20)	(0.18)	(0.21)	(0.21)
Trade*GDP per capita			−0.05**	−0.05***	−0.02	−0.03	−0.03
			(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
FDI/GDP				0.34
				(0.21)
Higher educaton share2					−497.33
					(313.98)
Trend						1.96***	1.95***
						(0.48)	(0.48)
							−1.00
							(5.24)
Number of observations	170	170	170	170	161	170	170
Number of countries	13	13	13	13	12	13	13
R-square	0.06	0.05	0.05	0.06	0.21	0.30	0.30

Source: IMF staff estimates.

¹Standard errors in parentheses. One, two, and three stars indicate significant at the 10, 5, and 1 percent level, respectively.

²No data available for Cambodia.

Panel Regressions for the Gini Index with Country Fixed Effects¹

Source: IMF staff estimates.

¹Standard errors in parentheses. One, two, and three stars indicate significant at the 10, 5, and 1 percent level, respectively.

Panel Regressions for the Gini Index with Country Fixed Effects¹

	(1)	(2)	(3)
Constant	−146.83*	−152.72*	−167.40
	(74.78)	(76.54)	(74.96)
Log (per capita GDP)	45.04**	46.93**	50.63**
	(18.93)	(19.52)	(19.03)
Log (per capita GDP)2	−2.72**	−2.88**	−3.09**
	(1.19)	(1.25)	(1.20)
Trade		0.01
		(0.23)
Asia crisis dummy			2.01
			(1.37)
Number of observations	50	50	50
Number of countries	11	11	11
R-square	0.36	0.38	0.34

Source: IMF staff estimates.

¹Standard errors in parentheses. One, two, and three stars indicate significant at the 10, 5, and 1 percent level, respectively.

View Table

Panel Regressions for the Gini Index with Country Fixed Effects¹

	(1)	(2)	(3)
Constant	−146.83*	−152.72*	−167.40
	(74.78)	(76.54)	(74.96)
Log (per capita GDP)	45.04**	46.93**	50.63**
	(18.93)	(19.52)	(19.03)
Log (per capita GDP)2	−2.72**	−2.88**	−3.09**
	(1.19)	(1.25)	(1.20)
Trade		0.01
		(0.23)
Asia crisis dummy			2.01
			(1.37)
Number of observations	50	50	50
Number of countries	11	11	11
R-square	0.36	0.38	0.34

Source: IMF staff estimates.

¹Standard errors in parentheses. One, two, and three stars indicate significant at the 10, 5, and 1 percent level, respectively.

View Full Size

GDP Per Capita of Sample Countries, 2005

(Purchasing Power Parity)

Source: World Bank, World Development Indicators.

• Download Figure
• Download figure as PowerPoint slide

Box 6.1.

Spatial Inequality in China and India¹

In both China and India, spatial inequality has increased markedly in recent years and is contributing to increased polarization. Amidst rapid economic growth, there are concerns that select regions are benefiting disproportionately while others are being left behind. In China, after a prolonged period of stable regional inequality, gaps in regional incomes now stand at their highest level since the early 1950s. In India, there is some evidence that the introduction of reforms in the early 1990s may have increased disparities in state incomes (Jha, 2002), or at least slowed the process of convergence (Purfield, 2006).

Trends in Spatial Inequality

View Full Size

China: Trends in Inequality

Source: Kanbur and Zhang (1999).

• Download Figure
• Download figure as PowerPoint slide

Growing regional inequalities have emerged as a thorn in China’s otherwise favorable reform experience. It is feared that China is fast becoming a polarized country along two dimensions—rural-urban and coastal-inland.² Although the urban-rural gap is much higher, it has remained relatively constant since the start of reforms in the late 1970s (Kanbur and Zhang, 2001). By contrast, coastal-inland polarization has increased dramatically, and explains most of the increase in total inequality observed during this period.³ Indeed, if current trends continue, the coastal-inland gap will eventually surpass the traditional rural-urban divide.

The gap between rich and poor states in India has also widened, as relatively rich states have on average grown two to three times as fast as poorer ones over the last three decades. In addition, the growth elasticity of poverty is also estimated to have been about 50 percent higher in richer states. There is also a high correlation between poverty and geographic location in India. While the poorest states—Bihar, Uttar Pradesh, Madhya Pradesh and Rajasthan—are primarily located in the central and northern regions of India (where the incidence of poverty is around 30 percent), middle- and high-income states such as Maharashtra, Gujarat and Tamil Nadu tend to be located along the coast. These regional differences have become particularly pronounced over the last 10 to 15 years, during which some poorer states, such as Orissa, have witnessed virtually no growth in per capita income and very little reduction in poverty (Deaton and Dreze, 2002).

Coastal-Inland Disparities in China¹

Source: Zhang and Zhang, 2003.

¹Ratio of costal to inland values.

²Sum of exports and imports to GDP.

Coastal-Inland Disparities in China¹

	GDP	Capital	Trade2
1985	1.12	1.20	4.68
1990	1.16	1.27	5.93
1995	1.39	1.39	4.86
1998	1.45	1.52	5.90
Percent increase	29	27	26

Source: Zhang and Zhang, 2003.

¹Ratio of costal to inland values.

²Sum of exports and imports to GDP.

View Table

Coastal-Inland Disparities in China¹

	GDP	Capital	Trade2
1985	1.12	1.20	4.68
1990	1.16	1.27	5.93
1995	1.39	1.39	4.86
1998	1.45	1.52	5.90
Percent increase	29	27	26

Source: Zhang and Zhang, 2003.

¹Ratio of costal to inland values.

²Sum of exports and imports to GDP.

What explains Spatial Inequality?

To some extent, it has been argued that spatial inequality may be a natural consequence of growth, with some regions enjoying a natural advantage by virtue of real geographic endowments, such as climate, natural resources or proximity to rivers, coasts, ports, and borders. This may be subsequently reinforced by the increasing returns that can arise out of interactions between economic agents in densely packed areas, creating virtuous cycles of development.⁴

At the same time, public policy can also influence spatial disparities. While the spatial distribution of natural features and agglomeration forces may give a region an initial advantage, geographical biases in government policies—especially investment, infrastructure, fiscal transfers and public service provision—and institutional barriers in product and factor markets can entrench spatial inequality.⁵ Some recent empirical work lends support to this hypothesis (see, for example, Jian and others, (1996) and Demurger and others, (2002) for China, and Cashin and Sahay (1996) and Ravallion (2005) for India). Other government policies also matter. In India, there is evidence that states that have sought to liberalize factor markets and promote good institutions usually perform better than others (Besley and Burgess, 2000), and improvements in education have been found to reduce income divergence between states (Aiyar, 2001). In addition, greater private sector investment and smaller governments tend to be positively associated with regional growth performance (Kochhar and others, 2006). These results suggest that policy choices could help stem—and potentially reverse—the recent increase in regional inequality.

Consequences and Policy Responses

Spatial inequality can have severe consequences. Spatial inequality can contribute significantly to changes in overall inequality and may retard growth, by reducing returns to job creation in new locations. It can also lead to more extreme outcomes—where geographical regions align with socioeconomic divisions based on ethnic, political, linguistic or religious affiliations, it can result in conflict and war. Indeed, some recent work suggests that spatial inequality amidst caste and ethnicity divisions is the major cause of the civil war in Nepal, which is at is most intense in the least developed mid and far western parts of the country (Murshed and Gates, 2005). Related to this is the Chinese leadership’s concern that rising income disparities between the coastal provinces and less dynamic North eastern parts of the country could affect the country’s political unity.

In both China and India, addressing regional disparities is high on the government’s policy agenda. During the Annual National People’s Congress in March 2006, the Chinese government stressed the need to provide more support to rural areas and less-developed regions. There are several components to the government’s strategy, including investment in transportation and communication infrastructure, raising education levels, improving the corporate governance of SOEs, and increasing the share of trade and FDI in less developed regions. Similarly in India, soon after being elected in 2004, the Congress government announced an agenda of “reforms with a human face”—described in the Common Minimum Program (CMP)—aimed at reducing poverty, increasing infrastructure and social spending, promoting FDI and improving access to credit for agriculture and small industry.

¹ The main author of this box is Murtaza Syed. ² The coastal region is typically defined as the following provinces: Beijing, Liaoning, Tianjin, Hebei, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian, Guangdong and Guangxi, with the remaining provinces classified as inland. ³ Kanbur and Zhang (2001) estimate that inland-coastal polarization nearly tripled between 1983 and 1995, while rural-urban polarization fell by around one-third. According to the OECD, differences in income between provinces accounted for 40 percent of total income inequality in 2000 (OECD, 2001). ⁴ For example, location externalities can result from technological spillovers, thick labor markets, or input-output linkages between firms. ⁵ Such as the household registration system (hukou) in China that constrains labor mobility, undermining the potential for income equalization across provinces through migration.

The Kuznets hypothesis may help explain the less favorable growth-equity tradeoff faced by the newly emerging economies compared with Japan, the NIEs, and the ASEAN-4. The latter countries’ initial record of high growth and low inequality may have been helped by relatively favorable conditions at the outset of their respective growth spurts. All the relatively advanced economies—with the exception of Japan—started their growth episodes from higher levels of per capita income, and with the transition from agrarian to industrial society much more advanced. For example, in 1967—the take-off date for NIEs—¾ of their population already lived and worked in urban areas, compared with a population share of ¼ for India at its growth take-off in 1982.⁵¹ Moreover, factor endowments were more equally distributed in the relatively advanced economies. Eight percent of the Japanese population had tertiary education in 1955, more than double the share in newly emerging economies at the time of their growth take-offs. Similarly, physical capital and land distributions were relatively equitable as a result of major postwar reforms (Japan, Korea, and Taiwan Province of China).

Initial Conditions of Fast-Growing Economies^{1 2}

Sources: World Bank, World Development Indicators; and IMF staff calculations.

¹The time of the growth take-off is 1955 for Japan; 1967 for the NIEs; 1973 for the ASEAN-4; 1979 for China; 1982 for India, and 1990 for Bangladesh and Vietnam.

²Simple averages for country groupings.

Initial Conditions of Fast-Growing Economies^{1 2}

		GDP per capita (current US$)	Urbanization (percent of population)	Agriculture (percent of GDP)	Human Capital (Percent of population with tertiary education)
Japan		259	39	…	8.0
NIEs		454	74	11	5.2
ASEAN-4		350	28	31	4.2
Newly emerging economies		232	21	34	3.4
	China	276	19	31	4.7
	India	285	24	36	3.4
	Bangladesh	269	20	30	2.2
	Vietnam	98	20	39	3.2

Sources: World Bank, World Development Indicators; and IMF staff calculations.

¹The time of the growth take-off is 1955 for Japan; 1967 for the NIEs; 1973 for the ASEAN-4; 1979 for China; 1982 for India, and 1990 for Bangladesh and Vietnam.

²Simple averages for country groupings.

View Table

Initial Conditions of Fast-Growing Economies^{1 2}

		GDP per capita (current US$)	Urbanization (percent of population)	Agriculture (percent of GDP)	Human Capital (Percent of population with tertiary education)
Japan		259	39	…	8.0
NIEs		454	74	11	5.2
ASEAN-4		350	28	31	4.2
Newly emerging economies		232	21	34	3.4
	China	276	19	31	4.7
	India	285	24	36	3.4
	Bangladesh	269	20	30	2.2
	Vietnam	98	20	39	3.2

Sources: World Bank, World Development Indicators; and IMF staff calculations.

¹The time of the growth take-off is 1955 for Japan; 1967 for the NIEs; 1973 for the ASEAN-4; 1979 for China; 1982 for India, and 1990 for Bangladesh and Vietnam.

²Simple averages for country groupings.

Openness and Technology

Standard trade theory predicts rising inequality in industrialized countries and declining inequality in developing countries in response to trade liberalization. According to the standard Heckscher-Ohlin model, the pattern of international specialization will reflect countries’ relative factor endowments after opening up to trade. Since the ratio of skilled to unskilled labor is higher in industrialized countries than in developing countries, they will specialize in skill-intensive products. Demand for skilled labor should rise and, therefore, inequality should rise after trade liberalization. The opposite pattern of specialization should emerge in developing countries, reducing inequality.

However, the prediction is not borne out by the data. There is still some controversy about the impact of trade openness on income inequality, with some studies suggesting a positive relationship (see e.g., Barro, 2000), and others finding no effect (see e.g., Dollar and Kraay, 2002). On the other hand, a broad consensus has emerged that trade liberalization increases wage dispersion, and more so in developing countries (see e.g., Milanovic and Squire, 2005).

In Asia, the impact of trade on inequality is modest at best. Trade openness, measured as the sum of exports and imports in percent of GDP, is added as an explanatory variable to the Gini coefficient regression. The coefficient is insignificant, suggesting that trade openness has had no impact on income inequality. The same explanatory variable is then added to the wage dispersion regression, where it is found to be significant and positively correlated with wage dispersion. However, the magnitude of this effect is small; a 10 percent increase in trade relative to GDP raises wage dispersion by 3 percentage points. Next, trade is interacted with GDP per capita (ppp) and added as another explanatory variable. This allows the effect of trade on wage dispersion to vary with the level of development, a proxy for differences in factor endowment. The coefficients for trade and trade interacted with GDP per capita—both significant—confirm that wage dispersion increases with trade liberalization, and more so in less developed countries. While the findings for Asia are in line with earlier empirical research and in contrast to the Heckscher-Ohlin model, the effects are relatively modest.

Why might increased openness be associated with rising wage dispersion in developing countries? There are a number of possible explanations. For one, relatively unskilled workers by industrial country standards may be skilled workers by developing country standards. Hence, specialization in manufacturing exports in the developing world may still benefit the better-off (see Wood, 1994). Alternatively, FDI and trade in goods produced by unskilled labor in developing countries may need to be supported by high-skilled labor, such as managerial staff. Given the scarcity of high-skilled labor in developing countries, skill premia would react particularly sharply to trade and capital account liberalization (see Feenstra and Hanson, 1997). The latter effect is explored by including the proportion of people with higher education and foreign direct investment (in terms of GDP) as additional regressors in the equation for wage dispersion. However, both coefficients turn out to be insignificant.

Skill-biased technological change has emerged as the main explanation for rising inequality and polarization in advanced economies (see Box 6.2 for a discussion focused on Korea’s experience). Most studies draw this conclusion by rejecting trade, the only alternative hypothesis for advanced economies, as a major determinant since prices of less skill-intensive products have not fallen and there has been little reallocation of labor from low-skill to high-skill sectors as expected under the standard Heckscher-Ohlin model (see e.g., Aghion and others, 1999). Skill-biased technological change represents a shift in the production technology that favors skilled over unskilled labor, by increasing the former’s relative productivity and, therefore, its relative demand. This in turn leads to a rise in earnings inequality.⁵² A few studies try to measure skill-biased technological change directly and find a close association with the skill-bias of employment (see e.g., Spitz-Oener, 2006). A time trend is added to the wage dispersion regression, as a proxy for skill-biased technological change. It is highly significant and leads to a large improvement in R-square.

Box 6.2.

Income Inequality and Social Polarization in Korea¹

Income inequality is rising fast in Korea, faster than in other industrialized countries. After falling for much of the 1980s and rising modestly in the early 1990s, income inequality surged during the financial crisis of 1997-98, as lower-wage employees were more severely affected by corporate layoffs. And the impressive post-crisis economic recovery has not brought inequality down; on the contrary, the increase in the Gini coefficient in Korea since 1997 is not matched by that in other industrialized countries, including the United States.² To boot, real incomes at the bottom of the distribution have fallen in absolute terms by close to 10 percent since the crisis.

View Full Size

Gini Coefficients in Korea and Comparators

Sources: U.S. Bureau of Labor Statistics; Statistics Canada; and Professor Kyungsoo Choi, Korea Development Institute.

• Download Figure
• Download figure as PowerPoint slide

Like elsewhere, skill-biased technological change and pressures from international trade are likely to have contributed to income inequality in Korea. Korea is exposed to competition from China, mostly affecting labor-intensive, small and medium enterprises in traditional manufacturing. At the same time, computer/IT penetration in the corporate sector is high. These two factors have lowered the relative demand for low-skill workers, and research does indeed point to a substantial rise in the college premium in Korea, with large effects on wage dispersion.

Yet if Korea is faring worse than other advanced countries, it is partly because job precariousness is greater. The economy has proven unable to generate permanent salaried employment, with a staggering 37 percent of salaried employees now under fixed-term contracts–ten percentage points higher than four years ago, and two and half times the OECD average. Some 30 percent of all new jobs created in the service sector in the past ten years have been accounted for by the self-employed, a much higher share than in most other OECD countries.³ Studies have repeatedly shown that fixed-term employees earn substantially lower salaries than permanent employees, even after accounting for all relevant work and personal characteristics. Beyond income considerations, the increased precariousness of work exacerbates other social inequalities, as fixed-term employees are substantially less likely to be covered by any worksite-based social insurance, face substantially higher probability of a return to unemployment, and together with the self-employed suffer from substantially higher income uncertainty. More flexible regulations on permanent employment, together with deregulation of the service sector to boost its productivity and growth, are key to creating permanent jobs in Korea.

View Full Size

Self-Employed

(In percent of labor force)

Source: OECD.

• Download Figure
• Download figure as PowerPoint slide

¹ The main author of this box is Jacques Miniane. ² The Gini coefficients come from different national sources, and hence caution is needed when comparing their levels across countries. However, it is safer to compare changes and trends. ³ There is evidence that much of this self-employment is involuntary, caused by a lack of salaried employment.

Did the Asian Crisis Increase Inequality?

Economic crises might be expected to increase inequality and polarization. The poor tend to have less flexibility to protect themselves against adverse shocks—they typically lack assets, such as savings and land, and have limited access to credit and insurance, preventing them from borrowing to smooth their impact. It has also been argued that lack of education and skills makes the poor less mobile across regions and economic sectors, undermining their ability to switch jobs and relocate in response to shifting demand conditions (Agenor, 2001). At the same time, crises may necessitate cuts in public expenditure that directly affect the poor, notably development spending and transfers.

There is little evidence, however, that the Asian crisis led to a widespread increase in inequality. In 3 out of the 5 countries most affected by the crisis—Indonesia, Korea, Malaysia, the Philippines, and Thailand—changes in the Gini index suggest that, if anything, income distribution improved in the immediate aftermath of the crisis (see also World Bank, 2000). Only in Korea did the crisis have a clear worsening effect.

View Full Size

Change in Income-based Gini Index During Asian Crisis¹

(Gini points)

Source: Zin (2005).¹ Compares Gini index in 1998 to that in 1996 for all countries, except Malaysia where 1999 is compared to 1995 and Philippines where 1998 is compared to 1997.

• Download Figure
• Download figure as PowerPoint slide

This is also confirmed by the insignificant coefficients on a crisis dummy—taking on the value 1 for the five countries in 1997 and 1998—that was added to both the Gini coefficient and wage dispersion equations. Finally, in 3 of these countries, the trend of rising inequality predates the Asian crisis.

Trends in Gini Index Before the Asian Crisis

(Gini points)

Sources: WIDER Database; World Bank, World Development Indicators; and Zin (2005).

Trends in Gini Index Before the Asian Crisis

(Gini points)

Indonesia	1987	33.1
	1993	34.4
	1996	36.5
Korea	1989	30.4
	1992	28.4
	1996	29.1
Malaysia	1989	46.2
	1992	47.7
	1995	48.5
Philippines	1988	40.6
	1991	43.8
	1994	42.9
Thailand	1988	43.8
	1992	46.2
	1996	43.4

Sources: WIDER Database; World Bank, World Development Indicators; and Zin (2005).

View Table

Trends in Gini Index Before the Asian Crisis

(Gini points)

Indonesia	1987	33.1
	1993	34.4
	1996	36.5
Korea	1989	30.4
	1992	28.4
	1996	29.1
Malaysia	1989	46.2
	1992	47.7
	1995	48.5
Philippines	1988	40.6
	1991	43.8
	1994	42.9
Thailand	1988	43.8
	1992	46.2
	1996	43.4

Sources: WIDER Database; World Bank, World Development Indicators; and Zin (2005).

Several factors are likely to have contributed to this relatively benign distributional impact, which is in stark contrast to the experience of Latin America during the 1980s.⁵³

• Initially, the crisis was predominantly an urban phenomenon, with direct effects concentrated on the relatively affluent. The crisis started as an exchange rate-induced financial downturn, directly affecting urban households and firms through declines in asset values, deteriorating balance sheets, and a squeeze on credit for consumption and investment. Wage declines and employment reductions were concentrated in the urban formal sector,

• Subsequently, poorer segments of the population also suffered, but the effect was dampened by a number of factors. First, in Malaysia, Thailand and the Philippines, the shift in relative prices associated with the currency devaluations favored agricultural products and other tradeables, preventing rural incomes from falling as sharply as in urban areas (FAO, 2001). Second, in some countries such as Thailand, the urban poor moved back to rural areas where they were relatively easily absorbed by flexible labor markets (Knowles and others, 2001). Third, welfare losses were also cushioned to some extent by dissaving and informal safety nets, particularly in Indonesia and Thailand (Gupta and others, 1998). which in these countries typically consists of workers with relatively high incomes. In Indonesia, for example, 14 of the 20 hardest hit areas were urban, while 13 of the 20 least affected areas were rural (Poppele and others, 1998).

Employment and Wages During Asian Crisis (1997-98)

(Percent change)

Source: World Bank (2000).

Employment and Wages During Asian Crisis (1997-98)

(Percent change)

		Agriculture	Manufacturing	Construction
Indonesia
	Employment	13.3	−9.8	−15.9
	Real wage	−35.0	−44.0	−42.0
Malaysia
	Employment	−5.3	−2.9	−13.4
	Real wage	…	−2.4	…
Korea
	Employment	0.0	−13.1	−26.4
	Real wage	…	−10.6	−14.7
Thailand
	Employment	−1.8	−1.9	−33.6
	Real wage	−8.9	−4.5	−2.2

Source: World Bank (2000).

View Table

Employment and Wages During Asian Crisis (1997-98)

(Percent change)

		Agriculture	Manufacturing	Construction
Indonesia
	Employment	13.3	−9.8	−15.9
	Real wage	−35.0	−44.0	−42.0
Malaysia
	Employment	−5.3	−2.9	−13.4
	Real wage	…	−2.4	…
Korea
	Employment	0.0	−13.1	−26.4
	Real wage	…	−10.6	−14.7
Thailand
	Employment	−1.8	−1.9	−33.6
	Real wage	−8.9	−4.5	−2.2

Source: World Bank (2000).

Conclusion

This study has looked at both inequality and polarization measures for Asia over the last decade. It finds that inequality and polarization have risen significantly for a broad and diverse set of countries, and more so in less developed countries. The study also begins to explore some explanations for this phenomenon. It finds a positive association between growth and inequality at low levels of development, and a negative one for more advanced countries. The impact of trade on inequality appears small. Nor does the Asian crisis account for much of the recent surge. Skill-biased technological progress appears to be behind the developments in advanced countries, whereas the transition from agriculture to industry is a likely driving force in developing countries.⁵⁴

While the precise causes of increased inequality remain subject to debate, many governments are interested in pursuing policies to stem this trend and to ensure opportunities for people to move up the income ladder. While specific polices will depend on individual country circumstances, a number of policy directions have been suggested. In addition to helping reduce inequality and polarization, these policies are also consistent with best practices for ensuring sustainable growth and sound macroeconomic management, and include:

• Spending on education. In both developing and advanced economies, a more equal distribution of human capital through greater—and more effective—spending on education is almost certain to reduce income inequality. In this context, the ASEAN-4 spent 3¼ percent of GDP on education in the decade following their growth take-off, compared with an average of 2½ percent of GDP for the newly emerging economies of China, India, Bangladesh and Vietnam. Whereas less developed countries should first aim for universal primary education, more advanced economies need to address skill-biased technological change by upgrading to the next highest level.

• Spending on infrastructure. Investment in transport and communications is a key to overcoming spatial inequality. It reduces transaction costs and, hence, the economies of agglomeration. Networks facilitate the flow of production factors, speed up the equalization of marginal returns, and spread prosperity from the centers of growth to the periphery. They also help to ensure that interior regions can benefit from integration into the global economy.

• Labor market policies. Labor market policies can have a significant impact on earnings, and hence inequality. Indeed, a number of countries have introduced specific labor market policies in an effort to influence income distribution. In some parts of Asia, lowering employment costs to facilitate new hiring could help to improve labor market efficiency and reduce inequality. In Korea, for example, a growing dualism in the labor market is an important source of increased inequality and the growing use of non-regular workers needs to be curbed, including through reduced employment protection for regular workers. Dismantling obstacles to internal migration is also important. Complementing public investments that aim to bring jobs to lagging regions, this would allow the poor to move to areas with greater potential and could be achieved by providing relocation assistance in the form of transport, housing, and training allowances. Such freer migration might, however, need to be phased in, to minimize problems of urban congestion.

• Access to financial markets. Inequality in access to financial markets can reduce economic efficiency and entrench disparities by preventing the poor from investing in themselves, or in their businesses. Addressing market failures, such as underdeveloped insurance or credit markets that restrict the ability of certain groups (such as farmers and small and medium-sized enterprises) to raise funds for investments or manage risk, is also likely to be an effective tool for redressing inequality.

• Investment climate. The cost of doing business remains prohibitively high in many parts of Asia, and extends to small and medium-sized enterprises. Streamlining regulatory regimes can help ensure that firms have an incentive to participate in the formal economy, with potentially wide-ranging effects on poverty and inequality. Key steps include encouraging competition by deregulating product and factor markets, maintaining non-distortionary tax and subsidy regimes, improving linkages with major markets, and ensuring access to information and finance to underpin investment.

At the same time, other policies are likely to be less effective. Studies generally find that the popular approach of providing fiscal incentives—in the form of tax advantages, risk-sharing arrangements, and subsidies—to attract industry and investment in specific regions is costly and ineffective (World Bank (2006a)). Ensuing competition between regions to attract industry and investment can also be highly distortionary, through its adverse effects on local tax bases and public services. Moreover, such interventions are unlikely to be successful at reducing inequality if underlying elements of the regional investment climate—such as the quality of institutions, the availability of skilled labor, infrastructure, and the efficiency of land and capital markets—remain weak.

Data Appendix

The following dates and welfare measures, where available, were used in the discussion and depiction of the changes in the Gini index, Polarization index, decline mean ratio and size of the middle class: Australia (Disposable income, 1995-2002); Bangladesh (Consumption, 1991-2000); China, rural (Consumption, 1990-2001); China, urban (Consumption, 1991-2001); Hong Kong SAR (Gross income, 1991-1996); India (Consumption, 1990-1999); Indonesia (Consumption, 1993-2002); Japan (Disposable income, 1994-2000); Korea (Gross earnings, 1995-2005); Lao PDR (Consumption, 1992-2002); Malaysia (Income, 1989-1997); Nepal (Consumption, 1995-2003); New Zealand (Disposable income, 1991-2001); Philippines (Consumption, 1988-2000); Singapore (Gross earnings, 1990-2000); Sri Lanka (Consumption, 1990-2000); Taiwan Province of China (Disposable income 1993-2003); Thailand (Consumption, 1992-2002); Vietnam (Consumption, 1993-2004).