Journal Issue
Share
Article

Inequality, Poverty, and Growth

Author(s):
Garbis Iradian
Published Date:
February 2005
Share
  • ShareShare
Show Summary Details

I. Introduction

Poverty alleviation and equity considerations are playing an increasingly important role in the work of the International Monetary Fund (IMF). This is because it is socially unacceptable to have poverty in the midst of prosperity and because equitable adjustment programs are more likely to be sustainable.2 In this regard, efforts are being made to increase the use of poverty and social-impact analysis techniques for the assessment of the impact of policy reforms on income distribution. These reforms are supported in the programs of the IMF, notably under the Poverty Reduction and Growth Facility (PRGF).

This paper analyzes the impact of income inequality on growth, in addition to testing the validity of the Kuznets curve, according to which income inequality rises with per capita income to a certain level and declines thereafter. It also examines the relationship among economic growth, income distribution, government spending, and poverty reduction. The paper attempts to respond to the following questions:

  • Is inequality harmful for growth?
  • Is inequality related to the level of per capita income (Kuznets curve)?
  • How responsive is poverty to economic growth and changes in inequality?
  • Would an increase in government expenditures reduce the incidence of poverty and improve the income distribution?
  • What is the minimum annual per capita real GDP growth needed for sub-Saharan African countries to reach their respective poverty targets, under the Millennium Development Goals, by 2015?

I use a new dataset on inequality and poverty. With that, I apply appropriate econometric techniques which address the potential biases induced by simultaneity, omitted variables, and unobserved country-specific effects—all of which have plagued previous empirical work on the links among growth, inequality, and poverty. A panel dataset for 82 countries for the period 1965–2003 has been assembled with the data averaged over periods of three to seven years, depending on the availability of inequality and poverty data. The minimum number of observations for each country is three and the maximum, seven. That is, only countries with observations for at least three consecutive periods are included. In the dataset, two household surveys for one country define what is called an interval of three to seven years in length. The entire sample includes 380 observations and 290 intervals.

Poverty in this paper is measured using the World Bank’s definition, that is, the percentage of the population living on less than $1 a day at 1993 prices, adjusted for purchasing power parity. For a few developing countries, the national household surveys’ definition of poverty is used. As to the measure of income distribution, I use the Gini coefficient, which is one of the most popular representations of income inequality. It is based on the Lorenz curve, which plots the share of population against the share of income received and has a minimum value of 0 (reflecting perfect equality) and a maximum value of 1 (reflecting total inequality).

The paper is structured as follows. Section II reviews analytical arguments and the related literature regarding the relationship among growth, inequality, poverty, and government spending. Section III presents data issues and suggests using more consistent data to reduce measurement error. Section IV analyzes and evaluates the panel regression results. Section V discusses the feasibility of the Millennium Development Goals of poverty reduction in light of the estimated growth elasticity of poverty. Section VI summarizes the empirical results. Appendix I describes the empirical methodology, and Appendix II reports the complete dataset used in this paper.

The empirical results in this paper challenge the belief that income inequality has a negative effect on growth and confirm the validity of the Kuznets curve. The paper identifies credit market imperfections in low- and medium-income countries as the likely reason for the positive link between inequality and growth over the short-to-medium term. The results also find evidence that higher government spending has a statistically significant impact on reducing inequality and poverty. Data quality, period length, and estimation technique may explain why the results in this paper are different from previous studies.

II. Theory and Evidence

Is income inequality harmful for growth? What are the factors which explain the differences in inequality across countries? Do higher government expenditures reduce inequality? To answer these questions, this section examines the related theory and the available cross-country evidence.

A. Impact of Inequality on Growth

There is as yet no consensus throughout the economics profession on the relationship between income inequality and growth. Early thinking on the effects of inequality on growth suggested that greater inequality might be good for growth, for example by redistributing income to the rich, who save, from the poor, who do not. This view implied a trade-off where more growth could be bought for the price of more inequality, with ambiguous effects on poor people. Figure 1 presents three different approaches of the channels through which income inequality affects growth:

  • ▪ The classical approach (Kaldor, 1957 and Bourguignon, 1981) suggests that the marginal propensity to save of the rich is higher than that of the poor, implying that a higher degree of initial inequality will yield higher aggregate savings, capital accumulation, and growth.
  • ▪ In contrast, the modern approaches emphasize the main four channels through which income inequality lowers growth: (a) the impact of inequality on encouraging rent-seeking activities that reduce the security of property rights; (b) unequal societies are more prone to difficulties in collective action—possibly reflected in political instability, a propensity for populist redistributive policies, or greater volatility in policies—all of which can lower growth; (c) the median voter in a more unequal society is relatively poorer and favors a higher (and thus more inefficient) tax burden; and (d) to the extent that inequality in income or assets coexists with imperfect credit markets, poorer people may be unable to invest in their human and physical capital, with adverse consequences for long-run growth.
  • Galor’s (2000) “unified model” provides an intertemporal reconciliation for the above two conflicting approaches (Box 1). He argues that the classical approach holds at low income levels but not at later stages of development. In the early stage of development, inequality would promote growth because physical capital is scarce at this stage and its accumulation requires saving. Inequality in income would then result in higher savings and rapid growth. In later stages of economic development, however, as the return to human capital increases owing to capital-skill complementarity, human capital becomes the main engine of growth. Credit constraints, however, become less binding as wages increase, and the adverse effect of income inequality on human capital accumulation subsides, and thus the effect of inequality on the growth process becomes insignificant.

Figure 1.Channels Through Which Inequality Can Affect Growth

Box 1.The Unified Model

The unified approach complements the research of Galor and Weil (1999, 2000) who developed unified models that encompasses the transition between three distinct regimes that have characterized the process of economic development: the Malthusian Regime, the Post-Malthusian Regime, and the Modern Growth Regime, focusing on the historical evolution of the relationship between population growth, technological change, and economic growth.

Galor and Moav (1999) argue that inequality has a positive effect on capital accumulation but negative effect on human capital accumulation in the presence of credit constraints. In the early stages of development physical capital is scarce, the rate of return to human capital is lower than the return on physical capital and the process of further development is driven mainly by capital accumulation.

In the early stages of development, the positive effect of inequality on aggregate saving more than offsets the negative effect on investment in human capital and, since the marginal propensity to save is an increasing function of the individual’s wealth, inequality increases aggregate savings and capital accumulation, enhancing the process of development. In the later stages of development, however, the positive effect of inequality on saving is offset by the negative effect on investment in human capital.

There is empirical evidence that growth depends on human capital, economic policies—such as openness to international trade, sound monetary and fiscal policies (reflected in small budget deficits and the absence of high inflation)—and a well-developed financial system. Other factors, such as geography, initial incomes and level of corruption, matter as well. Strong evidence suggests that growth is higher in countries with lower initial per capita income and in countries that have experienced a sharp fall in output (such as the transition economies in the early 1990s).3

In the past few years, inequality has been added as an additional independent variable to such cross-country growth regressions. But the literature has found mixed results using different samples and different econometric techniques. On the one hand, Alesina and Rodrik (1994), Clarke (1995), Perotti (1996), and Panizza (2002) found support for a negative impact of inequality on growth using cross-country growth regressions. Meantime, Deininger and Squire (1998) questioned the robustness and the validity of the negative association between inequality and growth.

On the other hand, Forbes (2000) found positive effects of income inequality on growth. She argued that country-specific effects and omitted variables are the cause of a significant negative bias in the estimations of the effects of inequality on growth. She also concluded that fixed-effect estimations yield the consistent result of a positive short- and medium-term correlation between inequality and growth. Smith (2001), examined empirically two hypotheses—subsistence consumption and credit market imperfections—of specific channels for inequality to affect private saving rates. He found that there is econometric evidence that especially at low per capita income levels, income inequality may be associated with higher aggregate savings.

B. Kuznets’s Law

Kuznets (1955) argued that the income distribution within a country was likely to vary over time with its progress from a poor agricultural society to a rich industrial society. The average per capita income of the rural population is usually lower than that of the urban population, whereas income distribution within the urban population is more unequal. In the urban population, savings are concentrated in the upper-income groups and the cumulative effects of such savings would be the concentration of an increasing proportion of income yielding assets in the upper-income groups. Thus, as the weight of the urban sector in the economy increases with industrialization, the country’s overall income distribution will tend to deteriorate until such time as the urban sector dominates. Thereafter, the income distribution will tend to stabilize because of three factors: (i) the slower growth in the population of the wealthier classes; (ii) the exploitation of the opportunities for wealth-creation offered by technology undertaken by those whose assets are not in established industries; and (iii) the shift of workers away from lower-income to higher-income industries.

The literature in the 1960 and 1970s in general supported the hypothesis that income inequality is related to the level of per capita income (see especially Ahluwalia 1976). According to Kuznets’ law, the relationship between income inequality and per capita income may be described by a curve in the shape of an inverted U, with an upward phase in which income inequality increases with rising per capita income, and a downward phase in which inequality declines with increases in per capita income. Some of the recent literature, however, challenged this hypothesis and several empirical studies found no significant relationship between inequality and per capita income, see Anand and Kanbur (1992). Li, Squire, and Zou (1998) argue that the Kuznets curve works better for a cross section of countries at a point in time than for the evolution of inequality over the time within countries.

During the past three decades, diverse patterns have emerged with respect to income distribution. On balance, Table 1 shows that more countries have experienced some worsening in inequality (see Appendix II for the full data set). Most South and East Asian economies grew at high per capita rates since the early 1970s while maintaining moderate levels of inequality, although increasing over time, in particular in China. In contrast, Latin American countries grew by less than half of the average growth rate in South and East Asia while maintaining high inequality. The differences in inequality at a given rate of growth could reflect a different combination of policies and institutions across countries and that these differences in policies matter for income distribution.4

Table 1.Inequality and Growth in Selected Countries
Household

surveys

based on 1/
Inequality

(as Measured by Gini Index)
Per Capita Annual Real GDP

Growth (in %)
19701980199020001970–791980–891990–2000
ArgentinaI‥‥4345520.8-2.83.0
BrazilI585863596.20.91.5
ChileI515356570.91.05.2
ColombiaI524851583.9-0.71.1
Dominican Rep.I‥‥455150‥‥0.81.0
MexicoI585155554.10.42.0
VenezuelaI494844483.0-3.00.1
Subtotal534952543.2-0.52.0
BangladeshE262728321.82.33.8
ChinaE‥‥3235454.68.78.9
IndiaE303231331.63.64.1
Korea, Rep. ofI353934326.96.05.4
MalaysiaI514946444.93.44.6
ThailandE424343435.66.33.2
Subtotal313231333.64.34.3
EgyptE‥‥3234344.43.12.2
MauritaniaE‥‥‥‥4039‥‥-0.42.1
MoroccoE‥‥3939403.22.71.2
PakistanE323231331.73.11.4
TunisiaE484640405.41.83.0
UgandaE‥‥‥‥3841‥‥-0.24.3
ZambiaE‥‥‥‥4853‥‥-2.1-2.3
Subtotal39401.11.7
CanadaI323131332.91.91.5
FinlandI323126272.83.21.6
GermanyI393736382.81.22.7
SwedenI211922251.61.81.3
United StatesI394043412.51.81.9
Subtotal333232332.52.01.8
Total average464345473.92.23.2
Sources: Derived from World Bank, OECD, and IMF reports and databases.

I denotes household surveys based on per capita income, and E, household surveys based on consumption.

Sources: Derived from World Bank, OECD, and IMF reports and databases.

I denotes household surveys based on per capita income, and E, household surveys based on consumption.

C. Growth, Inequality, Government Spending, and Poverty

The positive relationship between economic growth and poverty reduction is clear. However, there are significant differences across countries and over time in how much poverty reduction occurs at a given rate of economic growth. The extent of poverty reduction depends on how the distribution of income changes with growth and on initial inequalities in income and the sources or quality of growth. In theory at least, if income inequality increases, it is possible for a country to enjoy positive economic growth without significant benefit to its poorest segment of population—the rich get richer while the incomes of the poor stagnate. Therefore, establishing the relationship between economic growth and income distribution is critical for poverty reduction.

Fiscal policy is important both for reducing poverty and for improving social indicators through government expenditures programs. But increasing total government expenditures is not always the answer to improving the well-being of the poor. The composition of expenditures greatly influences the nature and outcome of government spending. Several developing countries were able to maintain social spending or even increase it as a share of GDP while total government expenditures were reduced. Chile, for example, managed to protect services to the poor during its fiscal adjustment in the 1980s and 1990s. Despite lower public spending on goods and services overall, basic health and child nutrition programs targeted to the poor expanded. This helped to sustain a continued improvement in social conditions in the 1980s and 1990s.5

Larger public spending on the social sectors (education, health, and housing) and on infrastructure is necessary to alleviate poverty and promote human development. The markets for education and health services are imperfect and governments in many countries have no other choice but to intervene on grounds of equity and efficiency. The link between social spending and income distribution is particularly strong, and public investment in human capital can be an efficient way to reduce income inequality over the long run. Investment in infrastructure could also be considered as poverty-reducing public expenditure. For example, building a road that eases access to a market for rural farmers enhances their income.

However, a larger government (as measured by the ratio of public expenditure to GDP) is also likely to harm growth prospects.6 This is particularly the case if the government maintains ineffective public programs and a bloated bureaucracy. In a retrenchment of the public sector, programs that benefit the poor might be cut. Also, if public employment plays a safety net role, then retrenchment may lead to increasing income inequality.

III. Data Issues

Data quality and measurement errors are major concerns in cross-country studies, particularly in the case of inequality and poverty data. In this paper, a concerted effort has been made to ensure that the statistics are comparable across countries and over time, using similar definitions of variables for each country and year. However, perfect comparability is not attainable, since the coverage of and questionnaires used in household surveys differ among countries and frequently also within countries over time. Whenever a trade-off arises, I decided to preserve comparability within a country over time rather than across countries.

While the quality of the World Bank data on poverty and inequality has recently improved, it is still far from being problem free. The data available at http://www.worldbank.org/research/povmonitor/ includes a data set on poverty and inequality for about 60 developing and transition countries. However, many of these countries had only one or two observations of three or more years apart. I have, therefore, expanded the existing data set by including comparable data on poverty and inequality from recent household surveys included in IMF staff reports and in Poverty Reduction Strategy Papers (PRSPs). I have also added to the sample data on inequality for the Organization for Economic Development (OECD) countries. All regions are well represented in the whole sample (16 countries from Latin America, 12 from sub-Saharan Africa, 12 from South and East Asia, 11 from the former Soviet Union, 6 from Central and Eastern Europe, 8 from the Middle East and North Africa, and 17 OECD countries)

The data set refers to an unbalanced panel of 82 countries observed from 1965 to 2003 (unequal country sizes or data are not available for all countries in the same period). The use of panel data allows us to control for time-specific effects, as well as country-specific effects. Also, the likely endogeneity of some explanatory variables can be accounted for using previous observations of the variables in the panel as instruments. The constraining factor is the scarcity of inequality and poverty data over time for many countries. Those included in the regressions have at least three observations. In the data set, two household surveys for one country define what is called an interval. In constructing the intervals the following criteria were used: intervals must be three or more years in length. They come from nationally representative surveys, and use either expenditures or income per person over time.

Data on poverty and inequality may not be comparable across countries as a result of differences in definitions and methodologies. There are some problems in comparing household surveys across countries. Different countries have different definitions of poverty, and consistent comparisons between countries based on the same definition can be difficult to obtain. The most widely used poverty indicator for developing and transition economies is the one used by the World Bank: the percent of the population living below $1 a day of consumption or income at 1993 prices, adjusted for purchasing power parity (PPP).

National household surveys are often the source for constructing consumption or income distributions and estimating poverty. But their design is not standardized across countries and over time, leading to significantly different estimates of average consumption or personal income. Some surveys only obtain information on income of households and others only on consumption. For developing and transition countries, slightly more than half of the observations are based on expenditures, and the remaining on income. Also, the Gini coefficient for some OECD countries is based on individual rather than households incomes. Household surveys based on expenditure data are usually regarded as more accurate than income data because they are likely to have fewer errors of underreporting. Also, data on expenditures yield a lower estimate of inequality than that based on income data, as a result of the higher saving rates of upper-income classes, the size of the informal economy, and private transfers.7 While there are significant methodological differences across surveys in different countries, these differences are likely to be less important in surveys conducted in a particular country across time. Although there is no perfect solution for this problem, using only one type of survey for each country and restricting cross-country comparisons to changes (as opposed to levels) of poverty and inequality, should go a long way toward addressing these problems.

IV. Econometric Results

A. Growth and Inequality

This section presents the panel regression results on the relationship between inequality and growth. Previous studies utilized Ordinary Least Squares (OLS) to estimate the cross-country growth regression. The resulting estimates of a negative coefficient on inequality suggested that countries with a more equal income distribution (that is a lower Gini index) tend to have higher levels of income. Due to the limited availability of comparable inequality statistics, sample selection is always a problem in estimates of the relationship between inequality and growth. This problem is magnified by using inappropriate econometric techniques of panel data. This paper specifically addresses these problems by using the fixed effect and the Generalized Method-of-Moments (GMM) econometric estimation techniques for panel data (for a brief description of these techniques see Appendix I).

Credit market imperfections could be a reason why inequality may increase growth. Galdor and Zeira (1993) have argued that when individuals cannot borrow against future income, the initial income inequality level affects physical and human capital accumulation and growth. Their models suggest that the poor are likely to be most affected by credit market imperfections.

One way to econometrically evaluate this hypothesis is to include the inequality variable and measures of credit market imperfections in a standard growth equation. Following King and Levine (1993), two variables from International Financial Statistics (IFS) were used to proxy financial market development and credit market imperfections. The first is the share of broad money (M2) in GDP, and the second is the share of credit to the economy in GDP. The specification estimated in Table 2 is as follows:

Table 2.Growth, Income Inequality, and Credit Market Imperfections
Estimation

Sample
(1)(2)(3)(4)(5)(6)
Full Sample 1/Financial Intermediation 1/Long-term Effect 2/
Low 3/High 4/
Inequality (Gini index)0.07

(3.2)
0.04

(1.9)
0.07

(2.7)
0.02

(0.4)
-0.05

(1.9)
-0.04

(1.8)
Log (per capita income)-1.36

(4.9)
-0.86

(4.1)
-1.63

(3.7)
-0.71

(3.6)
0.98

(2.6)
-0.98

(2.6)
Investment/GDP0.19

(6.5)
0.21

(7.0)
0.21

(5.8)
0.23

(7.5)
0.17

(5.3)
0.17

(5.3)
Inflation rate-0.03

(5.8)
-0.03

(7.2)
-0.03

(5.0)
-0.05

(2.5)
-0.02

(1.3)
-0.01

(0.8)
Inequality * HFI-0.07

(1.9)
0.08

(1.6)
HFI Dummy 5/3.67

(2.4)
-2.44

(1.5)
Countries828256266464
Number of observations2692691641058181
Source: Authors’ own calculations.Notes: Dependent variable: average annual per capita growth rate between two survey years of 3 to 6 years apart. See Appendix II for full set of data and definitions of variables. T-statistics in parenthesis are heteroskedasticity corrected.

Short- to medium-term effect. Each observation is derived from household surveys of 3 to 7 years in length.

Each observation is derived from household surveys of 10-15 years in length.

Algeria, Argentina, Armenia, Azerbaijan, Bangladesh, Brazil, Bulgaria, Cameroon, China, Colombia, Costa Rica, Dominican Rep., Ecuador, Egypt, El Salvador, Estonia, Ethiopia, Georgia, Ghana, Honduras, Hungary, India, Indonesia, Iran, Jamaica, Ivory Coast, Kazakhstan, Kyrgyz Rep., Latvia, Lesotho, Lithuania, Mali, Mauritania, Mexico, Morocco, Nepal, Nigeria, Pakistan, Paraguay, Peru, Philippines, Poland, Romania, Russia, Senegal, Sri Lanka, Tajikistan, Turkey, Uganda, Venezuela, Vietnam, and Zambia.

Austria, Belgium, Canada, Chile, China, Finland, France, Germany, Ireland, Italy, Japan, Jordan, South Korea, Malaysia, Netherlands, New Zealand, Norway, Panama, Portugal, Spain, Sweden, Thailand, Tunisia, the United Kingdom, Uruguay, and the United States.

HFI is a dummy variable to indicate countries with relatively high financial intermediation level. The shares of credit to the private sector and broad money in GDP are used as proxies to determine the level of financial intermediation.

Source: Authors’ own calculations.Notes: Dependent variable: average annual per capita growth rate between two survey years of 3 to 6 years apart. See Appendix II for full set of data and definitions of variables. T-statistics in parenthesis are heteroskedasticity corrected.

Short- to medium-term effect. Each observation is derived from household surveys of 3 to 7 years in length.

Each observation is derived from household surveys of 10-15 years in length.

Algeria, Argentina, Armenia, Azerbaijan, Bangladesh, Brazil, Bulgaria, Cameroon, China, Colombia, Costa Rica, Dominican Rep., Ecuador, Egypt, El Salvador, Estonia, Ethiopia, Georgia, Ghana, Honduras, Hungary, India, Indonesia, Iran, Jamaica, Ivory Coast, Kazakhstan, Kyrgyz Rep., Latvia, Lesotho, Lithuania, Mali, Mauritania, Mexico, Morocco, Nepal, Nigeria, Pakistan, Paraguay, Peru, Philippines, Poland, Romania, Russia, Senegal, Sri Lanka, Tajikistan, Turkey, Uganda, Venezuela, Vietnam, and Zambia.

Austria, Belgium, Canada, Chile, China, Finland, France, Germany, Ireland, Italy, Japan, Jordan, South Korea, Malaysia, Netherlands, New Zealand, Norway, Panama, Portugal, Spain, Sweden, Thailand, Tunisia, the United Kingdom, Uruguay, and the United States.

HFI is a dummy variable to indicate countries with relatively high financial intermediation level. The shares of credit to the private sector and broad money in GDP are used as proxies to determine the level of financial intermediation.

where GR is the average growth rate of per capita GDP at 1993 prices and PPP adjusted; GINIit-1 is the Gini index in the previous period; LogYit-1 is the natural logarithm at the beginning of the period of per capita GDP in dollars at 1993 prices and PPP adjusted; INVit is the share of gross capital formation in GDP; INFit is the average CPI inflation rate; HFI is a dummy variable equal to one for countries with a high level of financial intermediation, that is, above the sample median (as measured by the share of M2 and credit to the private sector in GDP); i =1, 2, …, n cross-sectional units (in this case, countries); µi is a country-specific unobservable effect, νt is a time-specific factor; and εit is the disturbance term.

Table 2 shows the panel regression estimates for the determinants of per capita real GDP growth. As to the impact of other explanatory variables (excluding inequality and financial intermediation level) on growth the main findings of the panel regressions are as follows:

  • ▪ There is a negative and significant correlation between growth and initial income per capita expressed in U.S. dollars. A poor country, other things being equal, tends to grow faster than a rich country.
  • ▪ There is a strong association between investment shares and GDP growth.
  • ▪ Macroeconomic instability (as measured by inflation) is negatively correlated with growth. The links appear to operate through a dampening of both investment and productivity.

The positive relationship between inequality and growth challenges previous empirical results. The estimated coefficients on inequality (GINIit-1) are positive in columns (1) to (4) (which test the short to medium-term effect of inequality on growth). Columns (3) and (4) in Table 2 split the full sample by low and high financial intermediation levels. The effect of inequality on growth differs between low and high financial intermediation sub-samples.

It is expected that β1>0, β5<0, and β6> meaning that the positive effect of inequality on growth is weaker in countries with high financial intermediation levels (or developed financial markets). The interaction term, GINIit-1* HFI, is strongly negative in column (1). Despite the relatively low t-statistics, an F-test shows that GINIit-1* HFI and the dummy variable HFI are jointly highly significant. Also the coefficient for HFI (β6) is positive and highly significant as expected. The insignificance of the inequality coefficient in column (4) is consistent with the argument that inequality has no explanatory power in countries with developed financial markets.

The long-term relationship between inequality and growth, however, is different than the short to medium-term relationship as shown in columns (5) and (6). Here the data are constructed as 10–20 year averages, and the estimated inequality coefficients are negative and statistically significant at the 10 percent level of confidence.

In conclusion, credit market imperfections may be a source of the positive link between inequality and growth. The results show that inequality stimulates growth in the short- to medium-term in countries with low levels of financial market development and credit available to the private sector. Over the long-term, however, inequality may have an adverse impact on growth.

B. Determinants of Income Inequality

Subsequently, I consider the factors influencing the variation of income inequality between countries and over time. In this paper, an effort has been made to compile an improved set of inequality statistics not only to reduce measurement error, but also to utilize panel estimation to control for time-invariant omitted variables.

Figure 2 shows a scatter of the inequality values against values of the log of per capita real GDP. A Kuznets’ curve would appear as an inverted-U relationship between the Gini value and log of GDP per capita (a proxy measure of economic development). The framework does not include country fixed effects, which would eliminate the cross-sectional information in the data. While the results should reflect both cross-sectional differences among countries as well as variations over time within countries, the main information comes from the cross-sectional dimension due to the fact that few observations are used for each country.

Figure 2.Inequality Versus Per Capita Income

Table 3 shows the panel regression results of the following equation:

Table 3.Determinants of Inequality
Independent VariableDependent Variable: Log (Gini Index)
Log (per capita GDP)1.12

(5.4)
1.25

(7.1)
0.98

(6.3)
Log (per capita GDP) squared-0.16

(6.1)
-0.19

(7.3)
-0.15

(6.7)
Log (government expenditure as % of GDP)-0.15

(3.8)
-0.17

(4.13)
Population growth0.02

(2.5)
0.02

(2.4)
Log (secondary school enrollment)-0.18

(4.6)
-0.15

(3.7)
Dummy for sub-Saharan Africa 1/0.08

(4.9)
Dummy for Latin America 2/0.14

(12.1)
Dummy for transition economies 3/-0.07

(6.6)
Dummy for income based inequality 4/0.04

(3.2)
Adjusted R-squared0.190.310.61
Number of countries909090
Number of observations378378378
Source: Authors’ own calculations.Notes: T-statistics in parenthesis. Estimation is by fixed effects.

The first dummy variable equals one if the country is in sub-Saharan Africa and zero otherwise.

The second dummy variable equals one if the country is in Latin America and zero otherwise.

The third dummy variable equals one for transition economies or socialist countries prior to 1994, that is, before significant progress was made to move to a market economy.

The fourth dummy equals one if the Gini index is based on income rather than consumption.

Source: Authors’ own calculations.Notes: T-statistics in parenthesis. Estimation is by fixed effects.

The first dummy variable equals one if the country is in sub-Saharan Africa and zero otherwise.

The second dummy variable equals one if the country is in Latin America and zero otherwise.

The third dummy variable equals one for transition economies or socialist countries prior to 1994, that is, before significant progress was made to move to a market economy.

The fourth dummy equals one if the Gini index is based on income rather than consumption.

where Log GINIit is the natural logarithm of the Gini index8; LogYit is the natural logarithm of income per capita and Log2Yit is included in equation (2) to test the hypothesis of a nonlinear conditional convergence; Log (EXP) is the natural logarithm of government expenditures in GDP (proxy for government expenditures on social sectors); EDUC is the secondary school enrolment rate (in percent of the total secondary school-aged population); POPGR is the percent change in population; i =1, 2, …, n cross-sectional units (here countries); µi is a country specific unobservable effect, νt is a time specific factor; and εit is the disturbance term. The following dummies are used:

D1: a dummy variable equal to one if the country is in Sub-Saharan Africa;

D2: a dummy variable equal to one if the country is in Latin America;

D3: a dummy variable equal to one if the country is in East Europe or the former Soviet Union for the period prior to 1995; and

D4: a dummy variable equal to one for the income-based measure of the Gini index.

The panel regression results presented in Table 3 show clear evidence of a non-monotonic relationship between inequality and the level of development as measured by per capita income. The estimated coefficients of LogYit and Log2Yit reported are highly significant and of the expected sign, implying a quadratic relationship between income per capita and inequality (Kuznets’ curve). The estimated coefficient is about 1 on the linear term and -0.15 on the squared term. The estimated coefficients were also found to be stable over time. The Gini coefficient value rises with per capita GDP to a certain level (estimated at about 4,000 in PPP dollars of 1993 prices) and declines thereafter. However, per capita income explains only about 20 percent of the variations in inequality across countries or over time (first column of Table 3).

These results are consistent with the findings in section A on the relationship between inequality and growth. They may have important implications for policies relating to income distribution in poor developing countries. Thus, if during a given period income inequality shows a tendency to increase modestly, than according to Kuznets’ law (which is confirmed in this paper) income inequality would have to be treated as an inevitable consequence of an increase of per capita incomes in the early stages of growth. It could then be argued that it is only in subsequent stages of their growth that there would be a “trickle down” of the effects of growth and a reduction of income inequalities.

There are other factors that also explain differences in inequality across countries. These may include the following: (i) school enrollment ratio; (ii) share of agriculture in GDP; (iii) real growth rate in agriculture; (iv) population growth; (v) governance; and (vi) other social and structural factors particular to countries or regions. However, historical data on governance and growth in agriculture is not available for most developing countries. Columns two and three of Table 3 include secondary school enrollment ratio, government spending as percent of GDP, and population growth in addition to per capita income.

Governments may be inefficient (more government means less growth) but appear to be benevolent since more government spending may reduce inequality. The estimated coefficient for the government size or government expenditures in terms of GDP has the right sign and is highly significant. Higher targeted government spending can be expected to improve the income distribution to the extent that rent seeking by privileged groups is avoided and government bureaucracies concentrate on enhancing the possibilities of the poor. While cutting the size of the government is likely to lead to faster growth,9 it could increase inequality.

The coefficient on the secondary school enrollment ratio (a proxy for human capital) was found to be negative and highly significant. This implies that improvement in education could reduce inequality. In contrast, high population growth would increase inequality.

Subsequently, I re-estimated the model using regional dummies. The estimated coefficients of the dummy variables for Latin America10 and sub-Saharan Africa are each positive and statistically highly significant. The results show that Latin America and the sub-Saharan region are 14 and 8 points, respectively, more unequal than the average for all countries. The dummy variable for transition countries is also significant but of negative sign as expected. The dummy variable for inequality based on income (D4) is also highly significant with a negative sign as expected. The estimated coefficient for D4 implies that inequality based on income is on average about four percentage points higher than inequality based on expenditures for the whole sample.

C. Role of Growth, Equity, and Government Expenditures in Reducing Poverty

Poverty is a multidimensional phenomenon, encompassing both monetary and non-monetary aspects. A common component of all poverty measurement and analysis is the setting of a poverty threshold, or a poverty line. People with welfare levels below the line are defined to be poor, and those above are not poor. Despite the limitations of such an approach, poverty measures of these sorts are useful in that they: (i) serve a monitoring role on the evolution of living standards, and (ii) can be an important means of focusing policy attention and public debates on the deprived groups.

The choice of the definition of poverty depends on the purpose of the analysis or the policy objective. There is no universally accepted concept of poverty that can be applied to every conceivable situation in every country. The one dollar-per-day poverty line, which is adopted in this paper, is the most widely used benchmark for developing and transition economies with a low per capita income.11

Rapid and sustainable economic growth is generally viewed as the primary vehicle for poverty reduction. The basic proposition is that if the economies of low-income countries grow rapidly enough and their income distributions are not unusually skewed against the poor, poverty reduction should occur. The experience of several countries has shown that poverty can increase not only because of a fall in output, but also because of increased inequality in the distribution of income. One important reason why inequality hinders poverty reduction is that the higher the level of inequality, the smaller are the absolute gains of the poor as the economy grows.

There is a significant difference across countries and over time in how much poverty reduction occurs at a given rate of economic growth. A key magnitude in assessing the impact of growth on poverty is the elasticity of poverty with respect to per capita real GDP growth. The correlation between per capita income and poverty incidence suggests some general tendencies related to income and poverty levels. First, lower poverty levels are associated with higher per capita income. Second, the incidence of poverty varies widely among countries with similar annual per capita incomes. The variation in poverty among countries with similar economic growth rates reflects the degree of income inequality of the countries. Also, as discussed earlier, in a retrenchment of the public sector, programs that benefit the poor might be cut. Larger government spending on social sectors (education, health, housing) and on infrastructure is necessary to alleviate poverty and to promote human development.

To capture the impact of growth, change in inequality, and government expenditure on poverty, the following equation is estimated:

where ΔPit is the headcount change in poverty of the total population (in percentage points) for country i between two household survey years; GRit is the per capita real GDP growth rate for country i between two survey years; ΔGINIit is the change in inequality (in percentage points) between two survey years; ΔEXPit is the change in government expenditures as percent of GDP for country i between two household survey years (a proxy for government expenditure on social sectors and infrastructure in the absence of such data for many developing and transition economies); and GINIMit-1 is the initial inequality level. Ideally, social spending should be used instead of total government expenditures. However, reliable and comparable breakdowns of government spending in many developing countries are not readily available. For this reason, equation (3) uses the change in total government expenditures as percent of GDP as a proxy for the change in social spending

The results of the panel regressions are reported in Table 4. The regression coefficients of growth in per capita GDP and change in inequality are statistically significant with the right signs. The estimated β1 coefficient shows how much poverty could be reduced in percentage points for a given growth in real GDP per capita. The coefficient for the change in the Gini index or income distribution (β2) is positive as expected and is also highly significant (robust). This finding suggests a positive and significant association between changes in inequality and the change in the level of poverty within a country. Growth will reduce poverty more if it is accompanied by a decrease in inequality, while poverty reduction will be tempered if growth is accompanied by an increase in inequality.

Table 4.Poverty, Growth, Inequality, and Government Size 1/
Sample
Whole SampleLow Income Cuntries1/
Per capita real GDP growth-0.30

(11.4)
-0.29

(10.5)
-0.30

(11.2)
-0.32

(8.5)
-0.28

(8.2)
Change in Inequality0.57

(6.7)
0.30

(2.9)
0.83

(5.7)
0.45

(3.2)
Change in government expenditure-1.01

(8.3)
-0.77

(5.7)
-1.01

(5.7)
Initial level of Inequality0.05

(3.2)
0.04

(3.3)
Adjusted R-squared0.580.620.680.610.71
Number of countries7272725151
Number of observations196196196142142
Sources: Authors’ own calculations.Notes: t-statistics in parenthesis. All estimated coefficients are significant at the 1% level.

Dependent variable is the poverty change in percentage points between two survey years.

2/ Developing and transition countries with per capita income of less than $3,000 calculated at PPP.
Sources: Authors’ own calculations.Notes: t-statistics in parenthesis. All estimated coefficients are significant at the 1% level.

Dependent variable is the poverty change in percentage points between two survey years.

2/ Developing and transition countries with per capita income of less than $3,000 calculated at PPP.

The regression results in Table 4 also show that the estimated coefficient for government expenditures (β3) is highly significant (robust) and has the expected negative sign. Controlling for the per capita GDP growth and income distribution level, an increase in government expenditures (particularly on social sectors and infrastructure) can be expected to reduce poverty. One percentage point reduction in government expenditures to GDP ratio would increase poverty by 0.7 percentage points.

It is likely that poverty increases if the adverse impact of an increase in inequality more than offsets the reduction in poverty associated with growth. The regression results show that the extent of poverty reduction also depends on the initial inequality level and not only on the change in inequality. That is, for the same growth in per capita income, poverty will be reduced more in countries with low initial equality than in countries with high initial inequality. Other things being equal, growth leads to less poverty reduction in unequal societies than in egalitarian ones.

Higher growth is associated with a lower poverty rate, but the response of poverty reduction to growth varies among regions. The estimated β1 coefficients by region could be used to calculate the elasticity for different regions, by dividing by their respective mean values of poverty rates. Such elasticity estimates generally measure the percentage change in the share of the population living below the poverty line following an increase of 1 percent in the average income or private consumption per capita of the population as a whole.

The calculated elasticity varies substantially, depending on the particular sample of countries chosen (Table 5). Transition countries, as a region, have the highest poverty elasticity of growth. A 10 percent decline in real per capita growth would lead to a 14 percent increase in poverty incidence. Such a sharp increase in the poverty rate in response to economic contraction suggests that the collapse of the centrally planned economic system in the early 1990s has affected the welfare of the population, not only through economic decline, but also through the deterioration of social conditions because of reductions in social expenditures by governments. For sub-Saharan Africa, the estimated elasticity of poverty reduction with respect to economic growth is -0.79 and the calculated inequality elasticity of poverty is 1.20. In general, if growth has a weak effect on poverty, it could be due to high inequality or a worsening income distribution, and thus poverty reduction policies should also focus on measures that could reduce inequality.

Table 5.Estimated Growth and Inequality Elasticities of Poverty, by Region
Growth

Elasticity

of

Poverty
Inequality

Elasticity

of

Poverty
Mean Values
Poverty

(% of pop.)
Inequality

(Gini

index)
Gov’t

Spending

(% of GDP)
Investment

(% of GDP)
Human

capital 1/
Whole sample 2/-1.081.402842232250
Latin America 3/-1. 312.021652222052
Sub-Saharan Africa 4/-0.791.204944221928
Middle East and North Africa 5/-1. 151.441740272347
East and South Asia 6/-0. 791.353038192653
Transition countries 7/-1.411.303131301971
Source: Author’s own calculations.

As measured by the secondary school enrollement rate.

The World Bank studies (including Ravallion, 1997) show much higher growth elasticity of poverty (-1.68) and inequality elasticity of poverty (1.90).

Brazil, Chile, Colombia, Costa Rica, Dominican Rep., Ecuador, El Salvador, Honduras, Jamaica, Mexico, Panama, Paraguay, Peru, Uruguay, and Venezuela.

Cameroon, Côte d’Ivoire, Ethiopia, Ghana, Lesotho, Madagascar, Mali, Mauritania, Nigeria, Senegal, Uganda, Zambia.

Algeria, Egypt, Pakistan, Iran, Jordan, Morocco, Tunisia, and Turkey. This group of countries have relatively low incidence of poverty (except for Pakistan) and income inequality as compared with other countries of similar per capita income due to the substantial remittances and the large size of the public sector in the economy. International migration to the Persian Gulf and Europe and the large public sector employment has helped boost the income of the poor in several Middle Eastern and North African countries (Adams and Page, 2003).

Bangladesh, China, India, Indonesia, Republic of Korea, Malaysia, Nepal, Pakistan, Philippines, Sri Lanka, Thailand, and Vietnam.

Armenia, Azerbaijan, Bulgaria, Georgia, Hungary, Kazakistan, Kyrgyz Republic, Romania, Tajikistan, Russia, and Ukraine.

Source: Author’s own calculations.

As measured by the secondary school enrollement rate.

The World Bank studies (including Ravallion, 1997) show much higher growth elasticity of poverty (-1.68) and inequality elasticity of poverty (1.90).

Brazil, Chile, Colombia, Costa Rica, Dominican Rep., Ecuador, El Salvador, Honduras, Jamaica, Mexico, Panama, Paraguay, Peru, Uruguay, and Venezuela.

Cameroon, Côte d’Ivoire, Ethiopia, Ghana, Lesotho, Madagascar, Mali, Mauritania, Nigeria, Senegal, Uganda, Zambia.

Algeria, Egypt, Pakistan, Iran, Jordan, Morocco, Tunisia, and Turkey. This group of countries have relatively low incidence of poverty (except for Pakistan) and income inequality as compared with other countries of similar per capita income due to the substantial remittances and the large size of the public sector in the economy. International migration to the Persian Gulf and Europe and the large public sector employment has helped boost the income of the poor in several Middle Eastern and North African countries (Adams and Page, 2003).

Bangladesh, China, India, Indonesia, Republic of Korea, Malaysia, Nepal, Pakistan, Philippines, Sri Lanka, Thailand, and Vietnam.

Armenia, Azerbaijan, Bulgaria, Georgia, Hungary, Kazakistan, Kyrgyz Republic, Romania, Tajikistan, Russia, and Ukraine.

V. The Millennium Development Goals of Poverty Reduction

The results in the previous section contribute to our knowledge of the relative importance of growth and equity in reducing poverty. The estimated growth and inequality elasticities of poverty could be used to determine the minimum economic growth required to achieve the Millennium Development goal of poverty reduction. The United Nations Millennium Declaration (General Assembly Resolution 55/2 of September 2000) endorsed the commitment to halve the proportion of people living in extreme poverty between 1990 and 2015. The poverty line in this case is based on one dollar per day in purchasing power parity.

Table 6 shows that poverty has declined significantly on a global level over the past two decades. However, most of this improvement was due to the sharp reduction in poverty in China and India, where the largest share of the world’s poor people live. The poverty rate in east Asia and the Pacific in 20 years dropped from about 58 percent of the population in 1981 to 15 percent in 2001, mainly because of the dramatic poverty reduction in China. In contrast, poverty rates in sub-Saharan Africa have increased, in the same period moving from 42 percent of the population in 1981 to 47 percent in 2001. The explanation for this is mainly the stagnant annual per capita growth in sub-Saharan Africa over the past two decades. The decline in poverty in recent years in several Asian economies was due to both improved income distribution and sustained rapid growth.

Table 6.Poverty and Growth Across the Globe
Regions/CountriesPercentage of Population Living

on Less Than $1.08/day at 1993 PPP
Per Capita Growth
198119902001Millennium

2015 goal
1981–901991–2001
East Asia and Pacific 1/57.729.614.914.85.75.9
 China63.833.016.616.58.18.6
South Asia 2/51.541.331.120.73.43.6
 India54.442.134.721.13.34.3
Sub-Saharan Africa 3/41.644.646.922.3-0.60
Latin America 4/15.716.313.58.2-0.11.7
Middle East & North Africa 5/14.213.910.27.00.31.5
Total40.327.921.314.02.44.1
Total (excluding China)31.626.122.813.11.42.9
Source: Poverty figures extracted from http://www.worldbank.org/research/povmonitor/ on August 4, 2004, except for the Middle East and North Africa, which is derived from the data reported in Appendix II. Per capita annual growth figures are derived from the IMF WEO database. Note: PPP denotes purchasing power parity.

China, Indonesia, the Republic of Korea, Malaysia, the Philippines, Thailand, and Vietnam.

Bangladesh, India, Nepal, Pakistan, and Sri Lanka.

Benin, Burkina Faso, Cameroon, Côte d’Ivoire, Ethiopia, the Gambia, Ghana, Lesotho, Madagascar, Malawi, Mali, Mauritania, Niger, Nigeria, Senegal, Tanzania, Uganda, Zambia, and Zimbabwe.

Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela.

Algeria, Egypt, Iran, Jordan, Morocco, Tunisia, and Turkey.

Source: Poverty figures extracted from http://www.worldbank.org/research/povmonitor/ on August 4, 2004, except for the Middle East and North Africa, which is derived from the data reported in Appendix II. Per capita annual growth figures are derived from the IMF WEO database. Note: PPP denotes purchasing power parity.

China, Indonesia, the Republic of Korea, Malaysia, the Philippines, Thailand, and Vietnam.

Bangladesh, India, Nepal, Pakistan, and Sri Lanka.

Benin, Burkina Faso, Cameroon, Côte d’Ivoire, Ethiopia, the Gambia, Ghana, Lesotho, Madagascar, Malawi, Mali, Mauritania, Niger, Nigeria, Senegal, Tanzania, Uganda, Zambia, and Zimbabwe.

Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela.

Algeria, Egypt, Iran, Jordan, Morocco, Tunisia, and Turkey.

A variation between countries has arisen to the extent to which poverty responds to growth. Initial inequality and per capita income are important factors. In this paper, the estimated growth elasticity of poverty is -1.08 and the inequality elasticity of poverty is 1.40 for the whole sample. The World Bank studies (including Ravallion, 1997) show much higher growth elasticity of poverty (-1.68) and inequality elasticity of poverty (1.90).

Based on the estimated growth elasticity of poverty in this paper, the per capita real GDP in sub-Saharan African countries needs to grow by at least 4.5 percent a year to reduce poverty from 47 percent in 2001 to 22 percent by 2015 (Table 7). Given an average population growth of at least 2 percent, this implies that the minimum needed economic growth should be about 6.5 percent a year. If we assume that inequality, as measured by the Gini index, improves by half a percentage point every year (that is average inequality in sub-Saharan Africa declines gradually from about 0.48 in 2001 to 0.40 in 2015) then real GDP needs to grow by 5.5 percent a year.

Table 7.Projection of Poverty in Sub-Saharan Africa(In percent of the population)
Per Capita Annual Growth Rates

(Inequality remains fixed = 0.47)
Per Capita Annual Growth Rates

(Inequality declines to 0.40 by 2015)
Gini

Index
2%3%4%5%2%3%4%5%
200246.145.745.345.045.745.344.944.50.463
200543.842.240.739.142.140.639.037.40.448
201039.936.432.929.436.132.629.125.60.423
201536.030.525.119.630.224.719.313.80.398
Source: Authors’ own calculations using estimated growth and inequality elasticities of poverty.
Source: Authors’ own calculations using estimated growth and inequality elasticities of poverty.

VI. Conclusion

The empirical results in this paper challenge the current belief that income inequality has a negative effect on economic growth. The panel regression results suggest that in the short-to-medium term, an increase in a country’s level of income inequality may have a positive relationship with subsequent economic growth. Credit market imperfections may be a source of the positive link between inequality and growth. In the long term, however, inequality would have an adverse impact on growth.

The results in this paper also confirm the validity of the Kuznets curve, according to which income inequality first increases and later decreases during the process of economic development. Inequality rises with per capita income to a certain level, estimated at about 4,000 in PPP dollars and 1993 prices, and declines thereafter. However, per capita income explains only about 20 percent of the variations in inequality across countries or over time.

Higher growth in per capita income is associated with higher rates of poverty reduction. The variation in poverty with similar economic growth rates reflects the degree of income inequality of countries. Poverty would increase if the adverse impact of an increase in inequality more than offsets the reduction in poverty associated with growth. For the same growth in per capita income, poverty will be reduced more in countries with low initial equality than in countries with high initial inequality. Other things being equal, growth leads to less poverty reduction in unequal societies than in egalitarian ones.

Sub-Saharan African countries will need to grow by at least 6.5 percent a year to reduce poverty from 47 percent of the population in 2001 to 22 percent by 2015 (assuming that the level of inequality remains constant).

APPENDIXES
APPENDIX I

I. Panel Estimation Techniques

This section describes briefly three panel data estimation econometric techniques. The technique of Ordinary Least Squares (OLS) may be plagued by problems of reverse causality. Baltagi (2001) proposes several other econometric techniques to estimate panel data which could avoid the problem of reverse causality including Fixed Effects and the Generalized Method-of-Moments (GMM) dynamic panel estimators.12 The optimal estimation technique is evaluated on the basis of the following two criteria: (i) the presence of unobserved time- and country-specific effects; and (ii) the likely endogeneity of some of the regressors. This chosen technique is necessary to control for unobserved time- and country-specific effects because these may be correlated with the right-hand side variables, and produce biased coefficients if omitted. The unobserved time-specific effects could be controlled for by using time-period dummies; this entails the elimination of information related to those variables that vary across time periods but not across countries.

The class of models that can be estimated using panel data can be written as:

where yit is the dependent variable, and αit and Xit are k-vectors on non-constant regressors and parameters for i =1, 2, …, n cross-sectional units (here countries); uit is a general disturbance, including a country specific unobservable effect, µi, a time specific factor νt, and an idiosyncratic disturbance εi,t. The fixed effects µi act as proxy for other determinants of a country’s steady state not included in Xit and the time specific factor νt controls for shocks common to all countries. Each country is observed for dated periods t =1, 2 … T.

Fixed-Effects Model

The fixed-effects estimator allows αi to differ across countries by estimating different constants for each country.13 The fixed effects model is equivalent to taking deviations from individual (country) means and then estimating an ordinary OLS regression using the transformed data:

The deviation from the mean purges the data of the fixed effects by removing means of these variables across countries. The OLS estimates of β in the fixed effects model are inconsistent, although as T→∞, the inconsistency disappears. But for finite, typically small T, the inconsistency remains, as is the case in this paper, with N=90 countries and T=7 (maximum periods for each country.

The coefficient covariance matrix estimates are given by the usual OLS covariance formula applied to the mean differenced model:

where X represents the mean differenced x, and

where uFE′uFE is the SRR from fixed effects model and (nT-n-k) is the correct number of degrees of freedom (nT being the number of observations, n the number of countries and k the number of parameters).

The fixed effects themselves are not estimated directly. They are computed from:

The fact that the fixed effects estimator can be interpreted as a simple OLS regression of means-differenced variables explains why this estimator is often called a within group estimator. That is, it uses only the variation within a country’s set of observations.

Generalized Method-of-Moments (GMM)

Using the GMM estimator that was developed for dynamic models of panel data can be written as follows:

where y is the natural logarithm of real per capita GDP, x is the set of explanatory variables (other than lagged per capita GDP), µ is an unobserved country-specific effect, ε is the error term.

Following the procedure of Anderson and Hsiao (1981) to account for unobserved country-specific effects, all variables in equation (7) are first-differenced. This eliminates not only the unobserved country-specific effects but also all variables for which only cross-sectional information is available. After first-differencing and rearranging the terms of the dependent variable, equation (7) becomes:

First-differencing, however, introduces a correlation between the error term (εi,t - εi,t-1) and the differenced lagged-dependent variable (yi,t-1- yi,t-2). An OLS estimation in this case would produce biased results, even when the set of explanatory variables X is strictly exogenous. The other econometric problem to be addressed is the likely endogeneity of some of the regressors.

To deal with these two estimation problems, we need to use certain instruments. Under the assumption that in (8) the error terms εi,t are serially uncorrelated (i.e., is E(εi,t - εi,s) = 0 for t≠s, and the explanatory variables, X, are weakly exogenous that is E(Xi,t εi,s) = 0 for s > t, then values of y and x, respectively, which lagged two periods or more, are valid instruments in the equations in first differences. These two assumptions imply a set of moment restrictions that can be used in the context of the GMM to generate consistent and efficient estimates of the parameters of interest. This methodology follows work by Arellano and Bond (1991) on dynamic panel data estimation.

Consistency of the GMM estimator, however, depends on the validity of the instruments. To address this issue, I present two specification test suggested by Arellano and Bover (1995). The first is the Sargen test of over-identifying restrictions, which tests the overall validity of the instruments by analyzing the sample analog of the moment conditions used in the estimation process. The second test examines the hypothesis that the error term in the differenced regression (εi,t - εi,t-1) is not serially correlated. One potential limitation of the GMM approach is that not much heterogeneity is allowed across countries. Heterogeneity is restricted to the intercept but is not permitted in the slope coefficients. Yet, if the slope coefficients vary across units lagged, values of serially correlated regressors cannot be used as valid instruments

APPENDIX II

II. Data Definition, Sources, and Dataset

Data Definition and Sources

  • Per capita real GDP growth rates are annual averages between two survey years and are derived from the IMF WEO and the International Financial Statistics (IFS) data bases.
  • The inequality data (Gini coefficient) are derived from World Bank data, OECD, and the IMF staff reports and Poverty Reduction Strategy Papers (PRSPs).
  • The secondary school enrollment as % of age group is at the beginning of the period and derived from the World Bank data base.
  • Population growth rates are from the World Bank Development Reports.
  • Data on the ratio of government expenditure and investment as shares of GDP are averages for the period between two survey years and come from the IFS.
  • Inflation rates, annual averages between two survey years, are calculated using the IFS’s CPI data.
  • GNP per capita at PPP are from the World Bank.
  • The poverty data is defined as the percentage of population living on less than $1 a day at 1993 prices and adjusted for purchasing power parity. The sources of the poverty data are the World Bank and recent IMF country reports and PRSPs.
  • Credit as % of GDP: Claims on the nonfinacial private sector/GDP, line 32d of the IFS.
  • M2 as % of GDP: Broad money/GDP, lines 34 plus 35 of the IFS.
Dataset
CountryHousehold

survey

year
Per capita

annual

growth
Inequality

(GINI

index)
Secondary

school

enroll. (%)
Pop.

growth

%
Gov’t

expend.

% of GDP
Invest.

as %

of GDP
Inflation

rate

%
GNP per

capita

PPP (US$)
Credit

as %

of GDP
M2

as %

of GDP
Poverty

as %

of pop.
Algeria198838.74,060407013.9
Algeria1995-2.035.3652.631.225214,358144815.1
Algeria19981.035.3692.230.025124,81384212.2
Argentina199044.79,4892220
Argentina19965.048.2731.415.419411,1722017
Argentina19987.049.5771.315.420112,1622228
Argentina2001-1.352.2771.216.918-111,5441928
Armenia198925.938.22,580302514.3
Armenia1998-6.037.986-1.025.617801,850101055.1
Armenia20017.036.0860.023.22012,680101250.0
Armenia200311.034.0870.024.02223,00091544.0
Australia198039.310,0162840
Australia19852.837.6561.438.926912,7903039
Australia19902.641.7551.333.925917,3145148
Australia19940.535.2511.235.721220,4076758
Austria197029.36,2004541
Austria19764.531.2640.042.02796,7225550
Austria19813.631.4620.050.323610,8297063
Austria19871.731.6600.252.020415,6488072
Austria19952.030.5580.452.522321,7029485
Azerbaijan198932.82,31033.6
Azerbaijan1995-9.045.0491.017.0241701,790131268.1
Azerbaijan20025.136.5501.026.02952,650111249.6
Bangladesh197425.926041672.1
Bangladesh19821.927.0162.517.0101142072062.1
Bangladesh19863.326.3192.516.0118570142560.0
Bangladesh19924.528.3232.415.0156845192858.8
Bangladesh19963.033.6272.013.31851,006202954.5
Bangladesh20003.831.8501.614.42251,252243249.8
Belarus198923.86,728
Belarus1995-2.528.969-0.242.4262204,9391228
Belarus20005.230.469-0.245.9261107,7121122
Belgium198028.310,517
Belgium19853.626.2980.140.919412,9082775
Belgium19903.026.6980.241.320218,3473589
Belgium19961.825.0970.340.620222,2513286
Brazil197057.61,5101819
Brazil19758.061.9402.322.0252002,2192618
Brazil19851.459.5462.025.4221803,989231315.8
Brazil19884.362.5541.826.8211205,672353018.6
Brazil1993-2.061.6601.629.8191505,831454418.8
Brazil19963.360.0621.429.020606,671403514.9
Brazil20011.459.0661.230.02067,5713031
Bulgaria198923.36,1262.0
Bulgaria1994-4.224.366-0.544.914445,48523708.0
Bulgaria2001-0.331.968-0.741.016126,625143312.8
Cameroon198449.02,215282440.0
Cameroon1996-2.847.7243.015.515131,843142053.3
Cameroon20012.544.6392.816.51722,22091740.2
Canada196531.64,3882237
Canada19703.032.3801.434.42355,6012438
Canada19753.031.6841.336.82297,1483442
Canada19802.631.0871.240.0241011,4344045
Canada19851.232.8901.147.022914,5294846
Canada19902.630.6921.146.822519,4074950
Canada19980.333.1931.044.020223,6436060
Chile197546.01,4391222
Chile19801.453.2491.728.0151502,6433226
Chile19871.056.4581.625.017213,999513610.2
Chile19906.756.1621.523.221204,81056408.3
Chile19946.554.8661.421.024136,74349384.2
Chile19976.757.5681.421.02678,44253413.0
Chile20001.757.1701.323.02359,09760401.0
China198132.0483304063.8
China19859.031.4541.420.52913797355541.0
China19908.834.6581.320.129141,332408033.0
China19949.040.0601.114.530102,2304510127.0
China19989.340.3631.113.93383,1974211019.0
China20016.344.7671.114.93614,0595013816.6
Colombia198853.14,53815124.6
Colombia19912.051.3522.113.018284,81814153.0
Colombia19962.457.1542.015.320236,074162011.0
Colombia19991.357.6561.919.119145,7272025
Costa Rica198634.44,130184412.5
Costa Rica19902.545.6462.525.621175,070164011.1
Costa Rica19932.746.3482.320.420166,100143510.3
Costa Rica19963.347.1522.223.318176,67018359.6
Costa Rica20003.846.5542.022.418118,4702238
Czech Republic198919.410,9800.4
Czech Republic1993-1.826.6840.037.3281510,42450681.3
Czech Republic19961.725.485-0.242.230912,81856740.0
Dominican Rep.198045.02,1361522
Dominican Rep.19850.243.3282.218.017352,555151911.0
Dominican Rep.19891.550.5282.114.522273,49218268.0
Dominican Rep.19970.849.7262.015.62184,67520274.0
Ecuador199252.33,732202224.5
Ecuador19952.054.8212.415.020324,689232328.9
Ecuador19980.556.2232.315.518283,3132822
Egypt198232.21,533237017.2
Egypt19902.934.0242.227.829172,416288625.0
Egypt19951.434.5282.033.719142,844328322.9
Egypt20003.034.4311.930.52153,519488116.7
El Salvador198949.02,8975525.5
El Salvador19952.751.3322.011.918124,0814625.2
El Salvador20001.853.2332.012.31734,5815622.0
Estonia198929.98,6781.9
Estonia1995-3.835.484-1.040.522356,92215278.9
Estonia20006.137.686-0.840.528119,7072638
Ethiopia198132.041262748.0
Ethiopia1995-1.340.0223.025.014846563945.5
Ethiopia20003.441.0252.930.0173557204244.2
Finland197031.84,7744234
Finland19755.427.0850.431.229155,8084537
Finland19802.030.9870.436.627139,2404638
Finland19853.030.8900.441.427712,2075545
Finland19903.826.2920.444.426617,6107763
Finland1995-1.225.6950.447.022219,6648065
Finland20004.526.9960.443.020225,7356553
France196547.04,171
France19703.444.0870.537.62455,3233943
France19754.043.0890.539.325106,4764250
France19802.434.9950.546.0241310,2817477
France19850.634.9970.552.2221212,8907871
France19951.732.7950.551.521321,4268663
Georgia198929.24,84413.0
Georgia1995-10.041.675-0.521.1133101,4224760.0
Georgia20025.138.971-0.318.015102,19081052.0
Germany197039.25,2696551
Germany19752.636.6530.344.02276,4106951
Germany19803.036.6600.248.321410,5497655
Germany1985-1.035.2700.147.521413,3308557
Germany19902.636.0650.145.022318,5318861
Germany19983.138.2630.246.023223,36010567
Ghana198735.41,23631447.7
Ghana19922.038.9473.020.014181,45641650.0
Ghana19981.836.0442.921.322321,78061739.5
Honduras199057.62,216243155.0
Honduras19951.056.1343.016.025172,417222948.3
Honduras1999-0.355.0342.916.526142,350334253.0
Hungary198923.39,5661.1
Hungary1993-4.527.978-0.556.020248,47133495.0
Hungary19981.324.484-0.449.0211810,07024463.0
India197833.1561203355.8
India19841.530.5302.113.6208848243949.8
India19883.031.2382.017.22191,298274446.3
India19943.731.9451.916.422101,709244542.3
India19975.733.7501.815.62392,010254740.0
India20004.032.5541.716.02262,370254834.7
Indonesia197030.740058.0
Indonisia19805.635.6202.019.5213183691729.0
Indonesia19873.732.0291.921.52471,489192517.0
Indonesia19935.831.7401.817.72792,400454314.8
Indonesia19965.736.5451.716.02893,029524515.7
Indonesia1999-2.031.0501.716.725232,736435727.1
Iran198647.03,523254227.3
Iran1990-6.843.4423.021.123213,219245026.0
Iran19947.343.0452.520.025223,665224421.3
Iran19982.343.0461.822.528253,691204020.9
Ireland197043.73,376
Ireland19753.638.7760.334.023173,3763966
Ireland19802.635.7780.438.325195,7404663
Ireland19851.235.2820.546.32287,9094953
Ireland19964.135.9870.441.019318,5547071
Italy197539.05,8067380
Italy19803.034.3630.141.924149,8135985
Italy19851.833.2650.150.8231212,5475170
Italy19902.832.7650.253.022617,9904965
Italy19981.536.0640.247.420322,4155260
Ivory coast198541.21,479392818.9
Ivory Coast1989-2.436.9392.826.0851,500363031.0
Ivory Coast19953.036.7372.625.31181,600282836.8
Ivory Coast19984.043.8342.420.71541,789162333.6
Jamaica199038.43,6382543
Jamaica19961.740.3671.016.428243,526224027.5
Jamaica1999-2.037.9690.918.62693,481264218.7
Japan197035.54,2567774
Japan19754.834.4951.025.335145,6958782
Japan19803.633.4960.832.63289,4598487
Japan19853.835.9980.632.730312,5329594
Japan19904.635.01000.430.529219,194120114
Jordan198634.93,44068963.0
Jordan1992-2.738.4503.529.62583,5606011514.4
Jordan19971.836.4524.032.42743,765679511.7
Jordan2003-0.836.5534.232.72324,106
Kazakistan198928.74,70115.5
Kazakistan1996-4.335.483-0.818.6171103,452111434.6
Kazakistan20012.831.284-0.623.418105,2251819
Korea Rep.196534.3710252341.4
Korea Rep.19708.235.3702.017.530151,032302723.2
Korea Rep.19757.838.0791.715.630211,435343020.0
Korea Rep.19808.238.6841.517.030242,573333314.5
Korea Rep.19855.634.5871.417.52984,155353514.2
Korea Rep.19905.833.6891.217.53267,522384010.5
Korea Rep.19956.232.0901.019.337511,67657457.4
Kyrgyz Rep.198928.72,01032.9
Kyrgyz Rep.1996-6.440.5720.833.417651,21781560.0
Kyrgyz Rep.19995.037.0730.833.918321,34951355.0
Kyrgyz Rep.20034.531.0740.829.019141,63961441.0
Latvia198927.48,7402.4
Latvia1994-5.627.682-1.041.015455,0401634
Latvia19984.332.481-0.843.321186,3501127
Lesotho198856.075630.3
Lesotho19933.657.9212.322.640151,341193443.1
Lesotho19952.060.0232.224.045101,5381732
Lithuania198927.39,1302.3
Lithuania1993-6.633.676-0.233.022456,1561945
Lithuania1996-0.732.477-0.234.223366,8001117
Lithuania20004.531.978-0.133.62128,6381224
Madagascar197946.9650212549.2
Madagascar1993-2.543.4242.919.21118759152260.2
Madagascar1997-0.642.0272.917.11320775102074.0
Madagascar20011.847.5292.919.01581,102102369.5
Malaysia197051.31,37149.3
Malaysia19804.649.1412.926.32442,318294632.5
Malaysia19854.046.8472.828.53463,167465720.7
Malaysia19901.845.7522.629.32824,562656717.1
Malaysia19956.645.6602.522.03847,23575779.3
Malaysia20002.644.3702.421.03638,884105994.0
Mali198936.5550142256.5
Mali19940.450.5162.724.9215609122372.3
Mali20013.249.4182.825.4242824162463.8
Mauritania199040.11,217301856.7
Mauritania19960.038.9252.827.12261,612321751.1
Mauritania20001.539.0272.629.32651,616291546.3
Mexico196555.51,411
Mexico19702.857.7473.315.020351,8883332
Mexico19753.457.9513.114.719402,5262931
Mexico19803.450.0532.825.021604,395203012.1
Mexico19840.852.5582.622.920704,9549289.0
Mexico1988-2.555.1612.423.219805,966182210.1
Mexico19951.153.7692.320.619217,042302817.9
Mexico20001.254.6702.119.520168,837202615.0
Morocco198539.22,040204418.0
Morocco19902.839.3372.028.82142,781285213.1
Morocco19991.239.5401.832.52253,384487319.0
Nepal198533.4619102838.3
Nepal19893.035.7402.618.32012826123040.0
Nepal19963.036.7522.516.62381,152273842.0
Netherlands198128.310,4286770
Netherlands19850.328.1950.659.722412,6326375
Netherlands19892.329.6970.656.820117,1478280
Netherlands19942.032.6990.654.920320,6829087
New Zealand197530.06,251
New Zealand1980-0.834.8851.033.025208,4901827
New Zealand19851.635.8871.133.7241211,4472132
New Zealand19902.640.2901.235.3221113,5867575
New Zealand19961.336.2931.236.719218,0309888
Nigera198538.7493163043.0
Nigeria19931.545.0183.017.41230812122434.1
Nigeria1997-0.850.6202.818.57357991221
Norway198212,7033457
Norway19863.822.6880.437.026814,5404460
Norway19901.323.7920.539.624618,1276563
Norway19952.625.8970.541.721323,9245556
Norway20003.725.8970.437.021329,9366954
Pakistan197031.5340254546.5
Pakistan19801.032.3262.819.7189669244230.7
Pakistan19853.633.4342.721.7177952274129.1
Pakistan19933.333.2492.622.41791,380274428.6
Pakistan19991.333.0472.522.61671,682284632.6
Panama198956.63,497503716.6
Panama19963.756.3682.027.72215,077645510.3
Panama20003.056.4711.827.72616,20586667.2
Paraguay199049.74,050502021.8
Paraguay1995-0.259.5572.716.223164,605602919.4
Paraguay19990.756.8612.618.02284,496803314.9
Peru198545.73,267922
Peru1994-1.648.6821.718.518803,9431426
Peru20003.549.4851.518.42384,6792632
Philippines198541.02,373302822.8
Philippines1988-2.340.7702.618.217103,110183018.3
Philippines19913.043.8722.519.421123,167213215.7
Philippines1994-1.342.9752.418.42383,332264318.4
Philippines19972.746.2782.318.32483,712476114.4
Philippines20000.946.1792.219.02163,8973459
Poland198925.55,7402041
Poland1993-1.028.0820.446.020826,187183513.8
Poland19995.531.6850.242.722278,90130403.0
Portugal198532.06,9488785
Portugal19905.031.0710.140.6271110,8785569
Portugal19982.831.6900.139.825415,0747873
Romania198923.35,73014.0
Romania1994-6.028.275-0.436.021715,144173021.5
Romania20001.030.376-0.334.818465,66182420.0
Russia198927.88,8175.0
Russia1994-6.843.683-0.229.2212527,038102030.9
Russia1996-6.348.184-0.125.219306,0451019
Russia2000-0.245.685-0.123.716427,2601322
Senegal199154.11,157292445.4
Senegal1995-1.041.3182.721.01531,594232457.9
Senegal20013.042.6192.620.61822,082192753.9
Slovenia198922.011,340
Slovenia1994-1.825.888-0.144.2221911,7002632
Slovenia19984.828.489-0.144.4211214,1803142
Spain197537.14,8027486
Spain19800.833.4760.832.925277,0746980
Spain19850.631.8780.642.122168,6376666
Spain19904.232.5860.442.022613,1297273
Sri Lanka196547.051037.0
Sri Lanka19792.343.5591.632.0207848182719.0
Sri Lanka19873.846.7631.431.324101,679203127.0
Sri Lanka19911.838.1661.329.622131,956193220.0
Sri Lanka19964.041.1681.227.725112,804183525.0
Sweden197521.36,7484260
Sweden19801.219.4970.361.621139,9754256
Sweden19852.020.5970.364.6201112,9994156
Sweden19902.421.9970.458.819617,7195051
Sweden19950.025.0990.460.017320,3054345
Sweden20003.025.0990.355.018324,9343943
Tajikistan198930.81,95451.2
Tajikistan1999-6.034.7401.816.01870707171365.4
Tajikistan20036.036.0431.817.520301,050191155.0
Thailand197542.9776243330.0
Thailand19817.247.3221.718.227131,572383825.7
Thailand19853.847.4261.518.42831,999424827.0
Thailand19884.747.4291.316.22933,104606221.0
Thailand19929.551.5351.215.04044,530827413.1
Thailand19966.843.4401.016.44056,47789783.0
Thailand2000-1.043.2470.819.02676,7779998
Tunisia197550.63,4503436
Tunisia19804.246.1432.434.029142,3383941
Tunisia19851.443.0482.336.528112,978484411.2
Tunisia19900.640.0532.234.62293,75554507.4
Tunisia19952.841.7672.132.82674,78054487.6
Tunisia20003.239.8752.032.02646,20552544.0
Turkey197056.01,436
Turkey19752.251.0422.116.719501,586
Turkey19871.843.6461.917.121403,93318261.5
Turkey19942.941.5521.723.322654,85715292.4
Turkey20002.740.0581.627.023606,18923481.0
Uganda198937.37104762.0
Uganda19924.339.2252.816.116408004756.0
Uganda19964.837.4263.018.21681,0705944.0
Uganda20004.340.5283.219.01931,23061435.0
Ukraine198923.57,2106.0
Ukaine1992-6.825.775-0.240.023896,31552711.0
Ukraine1998-4.832.576-0.238.421393,54781526.0
United Kingdom197727.05,9012838
United Kingdom19802.728.0880.244.919159,1753335
United Kingdom19851.029.0890.246.518811,7115039
United Kingdom19904.036.0900.340.118616,85710075
United Kingdom19950.836.0920.342.817420,44611070
Uruguay198942.212,5213853
Uruguay19952.042.7630.629.0144512,4572542
Uruguay19993.043.8650.632.0131512,521414411.0
Uruguay2002-3.044.6680.633.014612,118606612.0
USA197139.47,0445264
USA19751.839.7900.930.71988,1665865
USA19802.840.3900.933.7191113,0166063
USA19851.441.9921.036.719616,9035964
USA19902.842.8941.035.319423,4446467
USA19971.340.8961.033.319330,1135860
Venezuela198148.05,70029396.3
Venezuela1987-4.053.5242.624.020126,30025358.5
Venezuela19950.746.8352.420.019408,51014299.4
Venezuela19980.047.6442.320.618508,939111810.0
Vietnam199335.78,93952150.9
Vietnam19986.636.1442.119.72891,744183037.4
Vietnam20022.336.4552.120.82842,240415328.9
Zambia199048.3973103058.6
Zambia1993-3.546.2382.828.0121101,056102362.0
Zambia1996-5.649.8402.622.4294593461869.2
Zambia19982.552.6422.425.6322591571772.9
References

    AdamsRichard and JohnPage2003Poverty, Inequality and Growth in Selected Middle East and North Africa CountriesWorld DevelopmentVol. 31 pp. 202846.

    • Search Google Scholar
    • Export Citation

    AhluwaliaM.1976Income Distribution and DevelopmentAmerican Economic ReviewVol. 66No. 5 pp. 12835.

    AlesinaA. and D.Rodrik1994Distributive Politics and Economic GrowthQuarterly Journal of EconomicsVol. 109 pp 46590.

    • Search Google Scholar
    • Export Citation

    AnandS. and S. M.Kanbur1993The Kuznets Process and the Inequality-Development RelationshipJournal of Development EconomicsVol. 40 pp. 2572.

    • Search Google Scholar
    • Export Citation

    ArellanoM. and S.Bond1991Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment EquationsReview of Economic StudiesVol. 58 pp. 27797.

    • Search Google Scholar
    • Export Citation

    ArellanoM. and O.Bover1995Another Look at the Instrumental Variable Estimation of Error-Components ModelsJournal of EconometricsVol. 68 pp. 2951.

    • Search Google Scholar
    • Export Citation

    BaltagiB. H.2001Econometric Analysis of Panel Data2nd ed. (New York: John Wiley & Sons).

    BarroRobert2000Inequality and Growth in a Panel of CountriesJournal of Economic GrowthVol. 5 pp. 532.

    BourguignonF.1981Pareto Superiority of Unegalitarian Equilibrium in Stiglitz’s Model of Wealth Distribution with Convex Saving FunctionEconometricaVol. 49 pp. 146975.

    • Search Google Scholar
    • Export Citation

    ClarkeGeorge R.1995More Evidence on Income Distribution and GrowthJournal of Development EconomicsVol. 47 pp. 40327.

    • Search Google Scholar
    • Export Citation

    DeiningerKlaus and LynSquire1996A New Data Set Measuring Income InequalityWorld Bank Economic ReviewVol. 10 (September) pp. 56592.

    • Search Google Scholar
    • Export Citation

    de FerrantiDavidGuillermoPerryFranciscoFerreira and MichaelWalton2004Inequality in Latin America: Breaking with History? (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    FischerStanley1993The Role of Macroeconomic Factors in GrowthJournal of Monetary EconomicsVol. 32 pp. 485512.

    ForbesK.2000A Reassessment of the Relationship between Inequality and GrowthAmerican Economic ReviewVol. 90 pp. 86997.

    • Search Google Scholar
    • Export Citation

    GalorO.2000Income Distribution and the Process of DevelopmentEuropean Economic ReviewVol. 44 pp. 70612.

    GalorO. and O.Moav1999From Physical to Human Capital Accumulation: Inequality in the Process of DevelopmentCEPR Discussion Paper No. 2307 (London: Center for Economic Policy Research).

    • Search Google Scholar
    • Export Citation

    GalorO. and J.Zeira1993Income Distribution and MacroeconomicsReview of Economic StudiesVol. 60 pp. 3552.

    IradianG.2003Armenia: The Road to Sustained Rapid Growth, Cross-Country EvidenceIMF Working Paper 02/103 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    JohnstonJ. and J.DiNardo2000Econometric Methods (New York: McGraw-Hill).

    KaldorN.1957A Model of Economic GrowthEconomic JournalVol. 67 pp. 591624.

    KeeferP.KnackS.2000Polarization Politics and Property Rights: Links Between Inequality and Growth Policy Research Working Paper No. 2418August (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    KingR. and R.Levine1993Finance, Entrepreneurship, and Growth: Theory and EvidenceJournal of Monetary EconomicsVol. 32 pp. 51342.

    • Search Google Scholar
    • Export Citation

    KuznetsSimon1955Economic Growth and Income InequalityAmerican Economic ReviewVol. 45No. 1 (March) pp. 128.

    LiH.L.Squire and H.Zou1998Explaining International and Intertemporal Variations in Income InequalityEconomic JournalVol. 108 pp. 2643.

    • Search Google Scholar
    • Export Citation

    PanizzaUgo2002Income Inequality and Economic Growth: Evidence from American DataJournal of Economic GrowthVol. 7 pp. 2541.

    • Search Google Scholar
    • Export Citation

    PerottiRoberto1996Growth, Income Distribution, and Democracy: What the Data SayReview of Economic StudiesVol. 60 pp. 75576.

    • Search Google Scholar
    • Export Citation

    RavallionMartin1997Good and Bad Growth: The Human Development ReportsWorld DevelopmentVol. 25 pp. 63138.

    RavallionMartin and ShaohuaChen2003Measuring Pro-Poor GrowthEconomic LettersVol. 78 pp. 9399.

    SmithD.2001International Evidence on How Income Inequality and Credit Market Imperfections Affect Private Saving RatesJournal of Development EconomicsVol. 64 pp. 10327.

    • Search Google Scholar
    • Export Citation

    TanziVitoKe-youngChu and SanjeevGuptaeds.1999Economic Policy and Equity (Washington: International Monetary Fund).

1I would like to thank Jean Le Dem, Erik De Vrijer, Amor Tahari, Sena Eken, and Mohsin Khan for helpful comments and suggestions. All remaining errors are mine.
3Examples of this line of work include Fischer (1993), and Iradian (2003). Fischer, for example, shows that growth is negatively associated with inflation, large budget deficits, and distorted foreign exchange markets. Iradian demonstrates that higher levels of financial intermediation (as measured by broad money and private credit to GDP) are associated with higher growth prospects, and that growth is usually higher in countries following a sharp fall in output (such as the transition economies in the early 1990s)
7In many transition and developing economies, the poor, who are often engaged in the shadow economy, particularly in trade, housing construction, maintenance, and some traditional service sectors, have higher income than is recorded formally in the household budget survey.
8To address the problem of inconsistency resulting from the use of Gini coefficients based on expenditure and income, I included a dummy variable with a value of 1 for inequality observations that are based on income.
10Latin America’s inequality has deep historical roots and pervades contemporary institutions (De Ferranti and others, 2004).
11Poverty lines are cut-off points separating the poor from the non-poor. They can be monetary (e.g., a certain level of consumption) or non-monetary (e.g., a certain level of literacy). The use of multiple lines can help in distinguishing different levels of poverty. There are two main ways of setting poverty lines—in a relative or absolute way: (i) relative poverty lines are defined in relation to the overall distribution of income or consumption in a country; for example, the poverty line could be set at 50 percent of the country’s mean income or consumption; and (ii) absolute poverty lines are anchored in some absolute standard of what households should require in order to meet their basic needs. For monetary measures, these absolute poverty lines are often based on estimates of the cost of basic food needs (i.e., the cost of nutritional basket considered minimal for the healthy survival of a typical family), to which a provision is added for non-food needs.

Other Resources Citing This Publication