The effects of economic liberalisation on income ... · The economic liberalisation process, undertaken on a global scale since the early 1980s, has induced support for a set of market-oriented

The effects of economic liberalisation on income distribution: a panel data analysis

By

Gerardo Angeles – Castro∗∗∗∗

October 2004

Abstract: From the economic prescription that has been promoted by leading globalizers since the early 1980s and which is often referred to as the Washington Consensus or first generation reforms, we find that lower inflation and the promotion of employment can reduce income inequality. However, the flow of FDI worsens inequality, while international trade, which is deemed the corner stone to provide distributional effects, exerts a weak benefit on income distribution. In addition, the expansion of exports and employment based on the primary sector does not form an appropriate basis for reducing inequality, whereas a strategy based on industrialisation can have better consequences for income distribution. These results undermine basic neoliberal postulates and suggest that there is room for contesting approaches explaining the effect of trade and investment on income distribution and the relationship between primary production, industrialisation and inequality. Widespread financial crises that hit developing economies over the late 1990s and increasing sites of resistance to globalisation have led to the emergence of a Post Washington Consensus which places special emphasis on a stronger governance dimension, in which the understanding of governance is the effective and efficient management of the state. We show that the socio-political norms enveloped in this new global governance agenda represent an improvement of the economic liberalisation process, because domestic efficiency, the re-empowerment of the state, and human capital formation, which is one of the main social policies advocated by this approach, can help to mitigate the adverse effects of economic liberalisation on income distribution. However, we also show that the role of the state is not enough to socialise the operation of market forces and therefore further supranational mechanisms are required.

1. Introduction The economic liberalisation process, undertaken on a global scale since the early 1980s, has induced support for a set of market-oriented policies that can be summarised as deregulation, privatisation, liberalisation of markets and macroeconomic discipline. This prescription creates the preconditions for the expansion of trade and the flow of investments across countries. The theoretical support for this development model is standard neoclassical theory (Jones 1988, 30-33; Corden, 1993), which argues that trade, investment, and in general the market mechanisms boost growth and facilitate development. This view also holds that an important factor affecting growth and the effectiveness of the market system is the efficient organization of the domestic economy itself (Gilpin 1987, 265-6). The implications of this model for income distribution are that high and sustained rates of growth and the expansion of exports foster employment, reduce poverty, and eventually provide additional resources that facilitate the distribution of income. Moreover, economic liberalisation facilitates the operation of market forces and the adjustment to world prices, which allow resources to be allocated more efficiently.

∗ PhD student sponsored by Conacyt (Mexico) and teaching assistant of the Economics Department, University of Kent.

2

The theoretical formulations explaining this effect are the orthodox principle of comparative advantages (Jones 1988, 34-35) and the Stolper-Samuelson theorem (FitzGerald 1996, 32; Litwin 1998, 3). The latter is a neoclassical two-factor model, in which liberalisation of foreign trade increases the use of the cheaper-abundant factor as exports and imports adjust according to the principle of comparative advantages, while the costly-scare factor is used less. This mechanism increases the income of the factor which is relatively most used in the export sector and which is also most abundant. For example, this factor is conventionally assumed to be unskilled labour in developing countries, by the same token income distribution is assumed to improve. The economic policies involved in this early stage of the economic liberalisation process are often referred to as the Washington Consensus1 or first generation reforms (Ortiz, 2003). These terms are applied especially in developing countries. The chapter is aimed at testing the effect of the general variables that are assumed to improve income distribution over this early process of economic liberalisation. These variables are mainly trade, investment, macroeconomic discipline, and employment. Over the last few years, leading globalizers and multilateral institutions have induced support for a set of socio-political norms and have also recognised the need for a stronger governance dimension. They have added institution building, civil society participation, social and human capital formation, safety nets, transparency and accountability, among others, to the original economic norms or first generation policies, enveloped in the Washington Consensus. This new global governance agenda is a response to the financial crises that have hit emerging markets since 1995, the increasing perception that liberalisation brings with it inequality, and other subsequent forms of resistance to globalisation (Higgott 2000, 131-140). This further stage in the economic liberalisation process is usually called Post Washington Consensus (PWC) and the set of socio-political norms embedded in this approach are often referred to as second generation reforms. The PWC is an attempt to socialise and humanize the operation of market forces and to legitimise global economic liberalisation, although there is also a genuine recognition of the importance of tackling issues of fairness and inequality (Edwards, 1999). In this development paradigm, the re-empowerment of the state plays a central role for addressing the socioeconomic dislocation that may be generated by global liberalisation. In this context, the understanding of governance is thus the effective and efficient management of the modern state. Moreover, from this perspective, domestic efficiency and sound and disciplined macroeconomic policies accentuate the benefits of globalisation. These policies are associated with sustainable economic growth and are assumed to improve equity over the long-run. In contrast, those countries which do not adopt sound policies and show evidence of pronounced macroeconomic disequilibria are likely to fall behind in relative terms (IMF 1997, 72; Camdessus 1998, xiv; Higgott and Phillips 2000, 363).

1 John Williamson (1990) is given credit for first labelling as Washington Consensus the package of policies that multilateral institutions endorsed in trade and loan negotiations during the 1980s. This package of policies insisted on unregulated markets and a reduced role of the governments in economic activity.

3

In this context, the chapter also tests the effects of the PWC approach on income distribution. For this purpose we use variables such as government expenditure to proxy the size of the state, and we also explore the effect of first generation policies under different scenarios of macroeconomic stability and governance. Finally, we test the effect of human capital formation, represented by secondary school enrolment, on income distribution. This variable is deemed as one of the key elements in the set of socio-political norms advocated by the PWC. We find that the variable on trade exerts a weak benefit on income distribution and the effect is statistically significant only on the samples which comprise countries associated with good governance or macroeconomic stability, while the effect is significant only on the latter sample when the variable on trade is represented by changes in trade volume. The flow of FDI increases inequality under any scenario although the effect is mitigated in those countries that exhibit domestic efficiency. Inflation worsens inequality in those countries with domestic inefficiency; and this variable adversely affects income distribution even in those countries which exhibit good governance, although in any case the effect is weak. Not surprisingly, inflation is not significant in those countries that have traditionally kept macroeconomic discipline. The export-led growth strategy is able to reduce inequality when the export sector is oriented toward manufactured production. On the other hand, the expansion of primary exports does not exert benefit on income distribution under any scenario. Employment benefits income distribution; however when we explore the effects of employment by sector we notice that the expansion of employment in industry reduces inequality, whereas employment in agriculture is not able to improve income distribution. Consequently, the analysis of exports and employment by sector suggests that emphasis on primary production does not form the basis for redistributional effects. Even those countries with some form of domestic inefficiency, which are associated with lower levels of development and also with comparative advantages supported on natural resources and unskilled labour, do not seem to improve income distribution with the expansion of primary exports. These results are in keeping with the theoretical foundations and expectations that have supported the global liberalisation process to the extent that low inflation, fiscal discipline, larger employment and domestic efficiency can benefit income distribution. On the other hand, the results undermine other aspects of these theoretical foundations and expectations, as the benefits of trade on income distribution are weak and FDI worsens inequality, besides the fact that the export-led growth strategy and the expansion of employment based on the primary sector do not improve income distribution. Consequently, the results suggest that there is room for contesting theories and hypotheses explaining the relationship between trade, investment and income distribution and the relationship between primary production, industrialisation and income distribution. As for the set of policies underlined in the second stage of the economic liberalisation process, we find that domestic efficiency can help to reduce the adverse effect of FDI on income distribution, while it can also help to obtain some benefits from trade liberalisation in income distribution. Moreover, a stronger state is important to decrease inequality, and the set of socio-political norms enclosed in the PWC

4

approach, and represented in this study by a proxy of human capital formation, improves income distribution. Accordingly, these findings suggest that the PWC represents an improvement to mitigate the adverse effects that the global process of economic liberalisation can exert on income distribution. Nevertheless, the empirical evidence in this study suggests that the role of the state is not enough to socialise the operation of trade and investment. In this context, even under conditions of macroeconomic stability and high governance, FDI does not benefit income distribution; on the contrary, it seems to hinder it. Moreover, the beneficial effect of trade on income distribution is weak in those countries with domestic efficiency. Hence, this fact suggests that further supranational mechanisms, beyond the scope of the state, are required to socialise the flow of trade and investment. The chapter is organised as follows: section two analyses the characteristics of the data sets on income distribution available in the literature and selects the appropriate option for this study. Section two also presents the features of the explanatory variables included in the model. Section three explains the econometric method applied in the analysis. Section four gives results. Finally, concluding remarks are provided. 2. The Data One of the features which characterise the available data sets on income inequality is that the coverage is sparse and varies widely across countries and over time. In the absence of adequate longitudinal data, some studies attempting to assess the trend of inequality worldwide over time draw general conclusions from cross sectional data so as to try to overcome this major drawback (Bourguignon, 1994; Milanovic, 1995; Jha, 1996). However, this type of data does not deal with intertemporal relationships. In addition, other studies restrict attention to a subset of the data such as five-year intervals (Forbes, 2000; De Gregorio and Lee; 2002) or group the data in five-year or ten-year averages (Deininger and Squire, 1998; Calderon and Chong, 2001), but in these cases there is a risk of bias in the selection of the subset or in the construction of the average respectively. Finally, in order to improve coverage across space and through time, other studies use data sets measuring a component of overall income inequality (Galbraith and Kum, 2002); however, these data sets are not representative samples covering all of the population. Accordingly, so as to provide an accurate assessment, the data set on income inequality must contain a substantial coverage across countries and over time. It must also be consistent and harmonised, and based on a representative measure covering all of the population. Selection of the data set on income inequality The competing options available in the literature have the following characteristics: The World Bank data set. It is a compilation of Gini coefficients reported in the literature and was assembled by Deininger and Squire (1996), hereafter D & S. When

5

one applies the high-quality selection criteria2 the database yields 681 observations and an average number of 6.3 observations per economy, across 108 countries. The coverage is concentrated between the 1960s and early 1980s, but it varies widely across regions and decades. The World Bank data set suffers from a substantial degree of heterogeneity, even after the high-quality filters are applied. In fact, this sub-sample includes measures of expenditure inequality and income inequality; it also comprises a mixture of household and individual as unit of observation, while it contains Gini coefficients based on income gross and income net of taxes. World Income Inequality Database (WIID). In an attempt to expand the coverage of the D & S database, the United Nations University/World Institute for Development Economics Research (UNU/WIDER) collects and stores information on income inequality from different sources such as D & S, World Bank (1996), Luxemburg Income Study, Central Statistical Offices, UNICEF, and research studies. D & S was its starting point and forms half of the current WIID. If we try to obtain a more homogeneous sub-data set from WIID, for example Gini coefficients based on gross income and household/family over a national sample, the sub-data set results in 603 observations, across 105 economies, which yields 5.7 observation per country on average, concentrated between the 1960s and 1990s. Despite the fact that WIID represents an improvement in terms of homogeneity, its coverage still varies widely across countries and over time. Luxembourg income study. The LIS data set overcomes many of the problems of making comparisons across countries and over time that plagued other data sources since it is based on micro-level data and primary data sets from household income surveys. Through the LIS approach, it is therefore possible to assemble a data set more harmonized and standardized than the World Bank high-quality database and even than any high-quality sub-sample from WIID. However, the LIS coverage is constrained to 29 countries and 132 observations over the period 1969 – 2000, which yields an average number of 4.6 observations per country. UTIP-UNIDO data set. The University of Texas Inequality Project has produced a set of measures of the dispersion of pay across industrial categories in the manufacturing sector, drawn from the industrial statistics database published annually by the United Nations Industrial Organization (UNIDO). This data set is therefore based on source data that are likely to be accurate, both through time and across countries. Moreover, it has approximately 3,200 observations over about 36 years (1963 – 1999), across 153 countries, which averages 20.9 observations per country. However, its major drawback is that industrial pay inequality is a component of overall income inequality and its importance will differ across countries and vary over time (Galbraith and Kum 2003, 3). Thus, by construction it is not an index of overall income inequality, since it is confined to industrial pay inequality.

2 D & S suggested three high-quality filters. The first one is household or individual as a unit of observation and it requires that the data be based on units drawn from household surveys. The second filter requires that the data be based on a representative sample covering all of the population. The most usual deviations from this principle are surveys that cover only economically active individuals or indicators that cover only rural or urban sectors. Finally, the third filter requires that the measures of inequality be based on comprehensive coverage of different income sources. In this case the most frequent deviations from this principle are surveys that cover only wage income and tend to exclude non monetary income.

6

The Estimated Household Income Inequality (EHII). This income inequality measure was constructed by Galbraith and Kum (2003) and aims to fill the gaps and correct what they consider errors in the D & S data set. This indicator takes advantage of the information in D & S and the information in the UTIP – UNIDO database. As a matter of fact, they replicate the coverage of the latter with estimated measures of household income inequality, taking into account the relationship between industrial pay inequality, household income inequality, and an additional set of variables. The EHII covers 153 economies over about 36 years (1963 – 1999). In average it contains 20 observations per country. Galbraith and Kum underline two advantages from their data set.

“First, the coverage basically matches that of the UTIP-UNIDO exercise, providing substantially annual estimates of household income inequality for most countries, including developing countries that are badly under-represented in D & S. Second, the data set borrows accuracy from the UTIP – UNIDO… with due adjustment for the different weight of manufacturing in different economies. At the same time, unexplained variations in the D & S income inequality measure are treated for what they probably are: as inexplicable. They are therefore disregarded in the calculations of the UTIP Gini coefficient” (Galbraith and Kum 2003, 9).

Data sets such as D & S, WIID and LIS, can vary in terms of homogeneity and standardisation. In any case these data sets have sparse coverage and are widely unbalanced. For this reason, it may be necessary to group the data in year averages or to restrict attention to certain subsets so as to achieve a more balanced and less sparse data set. However, it has been stated that these procedures can cause bias in the selection of the subset and in the construction of averages. On the other hand, the UTIIP-UNIDO data set achieves a substantial improvement in terms of coverage, but it is not a representative sample covering all of the population. The EHII data set provides consistency, representation of household income inequality, and large coverage. In addition, it is a new data set available in the literature which can provide further insights for the study of income distribution. As a consequence, we select it from the competing data sets as a source of information for this study. Explanatory Variables It has already been noted that the dependent variable on income inequality is the EHII indicator. The analysis also includes four different sets of explanatory variables. The first set comprises market liberalisation proxies, and it is aimed at testing the effects of the original Washington Consensus prescription on inequality. The second set contains macroeconomic stability proxies and it will explore the role of macroeconomic efficiency on income inequality. The third set includes governance proxies, and it is aimed at assessing the influence of the PWC themes on the distribution of income. Finally, the fourth set contains variables related to employment; these variables are included because employment is considered one of the main factors affecting inequality in the neoliberal model. The explanatory variables are described as follows. The market liberalization proxies are trade volume, which is the sum of exports and imports of goods and services measured as a share of GDP. Alternatively, we also use changes in trade volume as a measure of trade liberalisation. So as to test the effect of

7

exports by sector on income distribution we apply two variables representing the percentage of manufactures exports and the percentage of primary exports to merchandise exports. FDI inflow is measured as a percentage of GDP and represents investment liberalisation. The source for the variables on trade is the World Development Indicators (2002); the source for the variable on FDI is a compilation of data from the WDI 2002 and UNCTAD. The variables are presented on a yearly basis. Inflation is included as both fiscal discipline and macroeconomic stability proxy. It reflects the annual percentage of change in consumer prices and is taken from the WDI 2002 on a yearly basis. This analysis also uses standard deviation of inflation for the period 1980 – 1998, calculated from the same source to construct two sub-samples. The first sub-sample comprises countries with standard deviation lower than 8.5, whereas the second sub-sample groups the countries with standard deviation greater than 8.5. This indicator allows us to assess the effect of economic liberalisation on inequality in two different scenarios of macroeconomic stability. The sub-sample classification is carried out through a two-step cluster analysis. The variables on governance comprise human capital formation, which is one of the main elements included in the concept of governance, and is represented by gross secondary school enrolment. It is measured by the total enrolment in secondary education, regardless of age, expressed as a percentage of the eligible official school age population corresponding to the same level of education in a given school year. This indicator is a compilation of data from UNESCO and WDI 2002, and is presented on a yearly basis. The analysis also includes an aggregate governance indicator, for the year 1996, which is provided by the World Bank. It is the average of six indicators measuring the following dimensions of governance: voice and accountability, political stability and absence of violence, government effectiveness, regulatory quality, rule of law, and control of corruption. Its score lies between -3.0 and 3.0 with a higher score corresponding to better governance. The indicator is used to construct low governance and high governance sub-samples that involve countries with scores lower than zero and greater than zero respectively. From these two sub-samples it is possible to analyse the trend of inequality within two different performances of governance. It has been claimed that the understanding of governance also involves the re-empowerment of the state as a means to tackle issues of inequality. Bearing this in mind, we include Government expenditure as a ratio of GDP as a variable to represent the size of the state and to explore its effect on income distribution. Because of constraints on data availability, this exercise is carried out only with the overall sample. The source is WDI 2002. The employment proxies comprise an unemployment variable, which is measured by the share of the labour force that is without work in the total labour force. Furthermore, in order to assess the effect of employment by sector on income distribution, we consider variables on employment in two main sectors, industry and agriculture. They are the proportion of total employment recorded as working in the corresponding sector. As government expenditure these variables are only analysed with the overall sample because of constraints on data availability. The source is WDI 2002

8

The dimension of the panel The original equation includes trade, investment, inflation and education variables. After assembling these variables and their different sources of information, it is possible to construct an unbalanced panel consisting of 1,301 observations across 93 countries over the period 1980 – 1998. After including the government expenditure and employment variables, the number of observations and countries drops. Hence, when these variables are included in the equation we conduct the regression only for the overall sample. The analysis of primary and manufactured exports is conducted in a separate equation including these variables only. In this case the number of observations and countries also drops, but allows us to conduct regressions across the sub-samples. In every panel the number of time periods available may vary from country to country, but the number of variables included in each year is the same. 3. The Model Initially we explore a general regression model for income inequality as follows: EHII it = αi + β1TRAGDPit + β2FDIGDPit + β3INFL it + β4EDUSECit + uit (1) Where EHII is the estimated household income inequality indicator, TRAGDP is the rate of trade to GDP, FDIGDP is the inflow of FDI as a percentage of GDP, INFL shows the annual percentage change of consumer prices, and EDUSEC represents the gross secondary school enrolment as outlined earlier. Subscripts i and t indicate country and year respectively, the error term uit is assumed to satisfy white noise assumptions, αi lets the intercept vary for each country and captures country-specific differences; finally, β1 to β4 are parameters to be estimated. Standard Methods The process of estimation starts with the standard OLS method pooling or combining all the observations, and assuming that αi = α. As column 1 in table 1 illustrates, all the variables are individually statistically significant. However, the traditional OLS approach has two mayor drawbacks. It assumes that the intercept value of the countries is the same and it does not control for country-specific factors. So as to confirm whether these are implausible characteristics in the model, the Breusch and Pagan Lagrange Multiplied test (1980) is conducted. Based on the OLS residuals and under the null hypothesis: σ2

ε = 0, i.e., αi = α, the LM test is distributed as a χ2 with one degree of freedom (Greene, 2000, 572-3). In this case, the result of the test is to reject the null hypothesis3. Thus, it is possible to conclude that the classical regression model with a single constant term is inappropriate. We turn therefore to panel estimation methods that may take in to account the specific nature of the countries. The fixed effect model lets the intercept vary for each country by adding dummy variables that take into account country-specific effects. In the Random effect model,

3 We obtain a Lagrange Multiplier test statistic of 4,466, which far exceeds the 5 percent critical value of χ2 with one degree of freedom, 3.84. Since the null hypothesis is rejected, it is concluded that there are individual effects.

9

differences across countries are captured through a disturbance term ωit, which follows ωit = εi + uit, where εi is an unobservable term that represents the individual specific error component, and uit is the combined time series and cross-section error component. The REM assumes that εi is not correlated to any explanatory variable in the equation. In order to choose between the FEM and the REM, we apply the Hausman test for specification (1978). The null hypothesis underlying this test is that the regressors and the unobservable individual specific random error are uncorrelated. If the test statistic, based on an asymptotic χ2 distribution, rejects the null hypothesis, then the random effect estimators are biased and the fixed effect model is preferred. The result of the Hausman test from equation (1) suggests that the REM estimates are inconsistent and the FEM would be more appropriate4. Since the LM test suggests that there are individual effects, and the Hausman test emphasises that these effects are correlated with the other variables in the model, it is possible to conclude that of the alternatives previously considered, the FEM is the better choice. Column 2 in table 1 shows the results from this model. Before adopting the FEM as the final estimation procedure, it is important to conduct an additional test in equation 1. It has been already contended that the error term uit is assumed to satisfy white noise assumptions, i.e. zero mean, constant variance σ2, and serially uncorrelated, which is denoted as uit ~I.I.D.(0, σ2). By the same token an AR(1) test should be available. In the presence of autocorrelation both σ2 and the standard errors are likely to be underestimated and biased, which leads to misleading conclusions about the statistical significance of the estimated regression coefficients5. The test for first-order serial autocorrelation is not satisfied, as it rejects the null: no evidence of AR(1) or ρ = 0 6. The P value of the test is also presented in column 2, table 1. To deal with this problem it is essential to explore the possibility that autocorrelation may arise due to model mis-specification; to be precise, because of omitted lagged dependent variables. In this context, equation 1 is extended and transformed into a dynamic panel data model (DPDM) by adding a lagged dependent variable as follows: EHII it = αi + γEHII it-1 β1TRAGDPit + β2FDIGDPit + β3INFL it + β4EDUSECit + ηi + uit (2) However, the inclusion of a lagged dependent variable introduces a source of persistence over time: correlation between the right hand regressor yit-1 and the error term uit. Furthermore, DPDM are characterised by individual effects ηi caused by heterogeneity among the individuals7. As a consequence, it is necessary to adopt different estimation and testing procedures for this model.

4The value of the Hausman test statistic is 130.65 with a negligible P value; hence, the test rejects the null hypothesis. In this case, the key assumption of the REM “the unobservable individual specific error εi is not correlated to any explanatory variable” is violated; thus, the FEM is preferred. 5For an empirical application see James k. Galbraith and Hyunsub Kum “Inequality and Economic Growth: Data Comparison and Econometric Tests” UTIP Working Paper Number 21 (2002), p. 12. 6 The AR test statistic of order one is equal to 29.99 and the P value is negligible. Hence, the AR(1) test suggests evidence of first order autocorrelation. 7 For an elaboration in this point see Badi H. Baltagi, Econometric analysis of panel Data (Sussex: John Wiley & Sons, 2001) 2nd Ed., pp. 129-30.

10

Table 1. Estimation Methods

OLSOLS (1 Group DV (2 sys GMM (3

EHIIt-1 0.841 *TRAGDP -0.015 * -0.003 -0.027 *FDIGDP 0.474 * 0.218 * 0.594 *INFL 0.001 * 0.000 0.001 *EDUSEC -0.098 * 0.054 * -0.102 *Constant 44.702 * 36.438 * 45.959 *

Adjustment coef 0.159Obs 1302 1302 1209Countries 93 93 93

BP LM test [0.000]Hausman test [0.000]Sargan test: [0.247]AR(1) test: [0.000] [0.000] [0.000]AR(2) test: [0.074]Notes.

Dependent variable: EHII

P values in parenthesis

* Significant at 5%; **Significant at 10%

Column (3) are long-run parameters The sys-GMM method In order to estimate the model, we use a generalized method of moment estimation (GMM) for DPDMs, initially proposed by Arellano and Bover (1995). Firstly, the estimation method eliminates country-effects (ηi) by expressing (2) in first differences as follows: EHII it - EHIIit-1 = γ(EHII it-1 - EHIIit-2) + βk(X it – Xit-1) + (uit - uit-1) (3) Where X is the set of explanatory variables outlined earlier. On the basis of the following standard moment condition: E(EHIIi,t-s ∆ uit) = 0, for t = 3,….,N and s ≥ 2 That is, lagged levels of EHIIit are uncorrelated with the error term in first difference, the method uses lagged endogenous variables as instruments to control for likely endogeneity of the lagged dependent variable, reflected in the correlation between this variable and the error term in the transformed equation. The resulting GMM estimator is known as the difference estimator. Blundell and Bond (1998) contended that the GMM estimator obtained after first differencing has been found to have large finite sample bias and poor precision. They attribute the bias and poor precision of this estimator to the problem of weak instruments, as they assert that lagged levels of the series provide weak instruments for the first difference (Blundel and Bond 1998, 115-6). So as to improve the properties of the standard first-differenced GMM estimator, they justified the use of an extended GMM estimator, on the basis of the following moment condition:

11

E[∆EHII it-1 (ηi + uit)] = 0 That is, there is no correlation between lagged differences of EHIIit and the country-specific effect. The method therefore uses lagged differences of EHIIit (the endogenous variable) as instruments for equations in levels, in addition to lagged levels of yit as instruments for equations in first differences. The extended GMM encompasses a regression equation in both differences and levels, each one with its specific set of instrumental variables. This type of estimation, called system estimator, not only improves precision but also reduces finite sample bias8. The method assumes that the disturbances uit are not serially correlated. If this were the case, there should be evidence of first order serial correlation in differenced residuals (i.e. uit - uit-1) and no evidence of second order serial correlation in the differenced residuals (Doornik et al 2002, 5-8). It is an important assumption because the consistency of the GMM estimators hinges upon the fact that E[∆uit ∆uit-2] = 0. Accordingly, tests of auto correlation up to order 2 in the first-differenced residuals should be available. The results of the sysGMM regression are reported in column 3, table 1. The tests of serial correlation in the first differenced residuals are in both cases consistent with the maintained assumption of no serial correlation in uit. The AR(2) test fails to reject the null hypothesis that the first differenced error term is not second order serially correlated, whereas by construction, the AR(1) tests rejects the null that this process does not exhibit first-order serial correlation9. In order to assess the validity of the instruments, a Sargan test of overidentifying restrictions, proposed by Arellano and Bond (1991), is also reported. Under the null hypothesis that the instruments are not correlated with the error process, the Sargan test is asymptotically distributed as a chi-square with as many degrees of freedom as overidentifying restrictions. In this case, the test is unable to reject the validity of the instruments10. 4) Results Original equation The results are illustrated in table 2. Column 1 shows the results initially obtained by regressing equation 3 with the whole sample. Subsequently, in order to assess the effect of economic liberalisation on income distribution, under different scenarios of

8 A discussion of the method through an empirical application can be seen in Cesar Calderón and Albert Chong “External Sector and Income Inequality in Interdependent Economies Using a Dynamic Panel Data Approach” Economics Letters, 71(2001), pp. 226 - 7 9 The P value in the test statistic for second-order serial correlation based on residuals from the first-difference equation is equal to 0.074. In this case we can not reject the null hypothesis if we establish a 95% confidence interval. Even in these circumstances the sys-GMM model represents a substantial improvement in terms of AR compare with ordinary methods. The P value in the test statistic for first-order serial correlation is negligible; in this case the null is rejected at any conventional level of significance. 10 The Sargan test statistic is equal to 55.38 with a P value equal to 0.247. Thus, the test is unable to reject the hypothesis that the instruments are not correlated with the error process.

12

governance and macroeconomic stability, we conducted regressions with four different sub-samples. The first two sub-samples comprise countries with low and high governance and their results are illustrated in columns 2 and 3 respectively. The last two sub-samples contain countries with high and low standard deviation of inflation, and their results are shown in columns 4 and 5 respectively. All the equations are regressed using the sys GMM procedure. Bearing in mind that a positive sign in the corresponding coefficient indicates a worsening in the distribution of income, the results yield the following conclusions. Trade. The overall sample indicates that trade reduces income inequality. At first glance this finding is in keeping with the expectations supporting trade liberalisation and is also consistent with other studies (Calderón and Chong, 2001). However, if we look at the magnitude of the coefficients, we find that the impact is relatively low. A 37 units increase in the rate of trade leads to a long-run decline of 1 point in the income distribution indicator ((0.004/(1-0.841))*37)11. Trade liberalisation is deemed the corner stone supporting distributional effects in the process of economic liberalisation (Bulmer-Thomas 1996, 10). In this sense, we should expect a larger benefit from trade on income distribution but the weak evidence above is not supportive of this assumption. It should also be emphasised that trade liberalisation is able to exert a long-term positive impact on income distribution under conditions of macroeconomic balance and high governance. On the other hand, the rate of trade as a percentage of GDP is not significant in countries with macroeconomic distortions and low governance. Thus, the outcome indicates that in the presence of macroeconomic disequilibria and failing to notions of good governance, trade does not exert an adverse effect on income distribution, but it is unable to deliver the set of positive benefits that a country should expect. To some extent this finding in particular is in accordance with the neoliberal assertion that the overall efficiency within a country is an essential factor for the appropriate operation of trade policies. Investment. The outcome of the overall regression suggests that FDI has an adverse effect on income distribution. An upturn of 1.7 points on the rate of investment as a percentage of GDP raises the inequality indicator by 1 point over the long-term ((0.094/(1-0.841))*1.7). This result undermines the assumptions and expectations outlined in the neoliberal postulates. In addition, table 2 indicates that the effect of FDI is adverse under any scenario. However, if we compare the coefficients across columns 2 to 5, it is possible to observe that this effect is worse on countries with macroeconomic disequilibria and low governance. For example, a 1.7 point upturn on the rate of investment as a percentage of GDP raises the inequality indicator over the long-run by 0.75 points ( (0.14)/(1-0.684)*1.7) and by 0.61 points ((0.021)/(1-0.942)*1.7) in low and high

11 Dollar and Kraay (2004) stress that between the late 1970s and late 1990s the rate of trade to GDP in OECD countries rose 29 points, to say from 21% to 50%. While trade volume in the top one third of developing countries in terms of this rate expanded 17 points, to say from 16% to 33%. Accordingly, the effect of trade volume on income distribution in this study seems to be weak because the expansion of the rate of trade to GDP required to improve income distribution by one point far exceeds the average increase in trade volume over the last two decades. Thus, this statistical evidence suggests that the effect of trade on income distribution has not been as large as it was originally expected.

13

governance countries respectively, whereas the effects on countries with high inflation and low inflation standard deviation is 1.53 points ((0.203)/(1-0.775)*1.7) and 0.54 points ((0.032)/(1-0.899)*1.7) respectively. Consequently, inefficiency within countries seems to be a factor that accentuates the adverse effect of FDI on income distribution. Table 2. Scenarios

Whole Low High High LowSample (1 Governance (2 Governance (3 Infl SD (4 Infl SD (5

Short-run parametersEHIIt-1 0.841 * 0.684 * 0.942 * 0.775 * 0.899 *

TRAGDP -0.004 ** -0.003 -0.001 * -0.012 ** 0.000 *FDIGDP 0.094 ** 0.140 * 0.021 * 0.203 * 0.032 *INFL 0.000 * 0.000 * 0.000 * 0.000 * 0.001EDUSEC -0.016 * -0.020 * -0.002 * -0.021 * -0.007 *Constant 7.306 * 13.886 * 2.449 * 10.698 * 4.332 *

Long-run parametersTRAGDP -0.027 * -0.009 -0.017 * -0.055 -0.005 *FDIGDP 0.594 * 0.442 * 0.359 * 0.900 * 0.317 *INFL 0.001 * 0.001 * 0.003 * 0.001 * 0.012EDUSEC -0.102 * -0.065 * -0.039 * -0.094 * -0.072 *Constant 45.959 * 43.876 * 42.129 * 47.532 * 42.707 *

Adjustment coef 0.159 0.316 0.058 0.225 0.101Obs 1209 546 663 517 692Countries 93 47 46 44 49

Sargan test: [0.247] [0.617] [0.929] [0.923] [0.565]AR(1) test: [0.000] [0.007] [0.001] [0.007] [0.002]AR(2) test: [0.074] [0.070] [0.385] [0.124] [0.159]Notes.


Sargan and Serial Correlation Test are P values


All the equations use the sys GMM estimation method Inflation. Macroeconomic imbalance is represented by the inflation variable. Results from the overall regression reveal that this variable is a determinant of income inequality. This finding is in accordance with the orthodoxy supporting liberalisation and stabilisation policies, although the magnitude of the effect is low. An increase in inflation of 930 points, which could be considered an episode of hyperinflation12, is required to raise inequality by 1 point ((0.0002/(1-0.841)*930). Hence, inflation has a negative effect on the distribution of income, but this effect does not seem to be large. Table 2 reveals that inflation has adverse effects on countries with low governance and macroeconomic distortions. It is worth noting that the repercussions of inflation also affect the distribution of income in countries with high governance. Not

12 The threshold customarily used to define hyperinflation is an increase of prices averaging 50% a month or 600% a year (Havrylyshyn et al, 1994).

14

surprisingly, this is not significant in equation 5; this fact indicates that the inflation rate is a variable that does not affect income distribution in countries that have traditionally kept a stable economy. Education. Our results confirm that education, measured through gross secondary school enrolment, reduces income inequality. Similar conclusions have also been obtained in previous studies (Chong and Calderon, 2000; De Gregorio and Lee, 2002). As for the overall sample, the education coefficient implies that an increase of 10 points in gross secondary school enrolment is required to reduce the inequality indicator by 1 point over the long-run ((0.016/(1-0.841)*10). Moreover, education exerts a positive impact under any scenario. Hence, human capital formation can mitigate the negative effect that may be caused by economic liberalisation on income distribution and it is in keeping with the postulates of the PWC13. The adjustment coefficient. The outcome of the regression illustrates that the adjustment coefficient is larger for those countries that exhibit macroeconomic mismanagement or a low level of governance. This fact suggests that these countries are more vulnerable to the effect of the variables involved in the economic liberalisation process, because their distribution of income adjusts faster to the long-term level, or changes faster compared to those countries with more domestic efficiency. Applying changes in trade volume Some authors contend that trade volume is a variable that reflects countries’ geographical characteristics such as their proximity to major markets, their size, or whether they are landlocked. As a consequence, this variable may tell us little about the effect of trade on growth or income distribution (Dollar and Kraay, 2004). With the above in mind, changes in trade volume is a variable that may eliminate geography or any other unobserved country characteristic. On the other hand, trade volume is a variable applied frequently in the empirical literature, including those studies exploring the relationship between economic liberalisation and income distribution (Calderón and Chong, 2001). In addition, this variable can be more effective in panel data studies, because they also consider variations over time and not only variations across countries. Finally, trade volume improves its explanatory power when it is applied in first difference estimations, such as GMM methods, because in this way unobserved country characteristics are eliminated. So as to test if our results can vary depending on the applied variable on trade liberalisation, we regress equation 3, replacing trade volume with changes in trade volume. The outcome of this exercise, including both the overall sample and the four sub-samples, is illustrated in table 3. 13 As complementary information, we found that a lower rate of population growth leads towards less inequality, as the corresponding coefficient enters positively and significantly in the overall sample and the four sub-samples (regression not reported). In this context, Heerink (1994) shows that income inequality and population growth reinforce each other, because lower levels of fertility results in a more equal income distribution; he also underlines the negative relationship between improvements in nutrition, health and education on the one hand and the fertility levels on the other. Not surprisingly, our data base also reveals a negative relationship between education and the rate of population growth. These findings suggest that education can contribute directly to improve income distribution, but also contributes indirectly to this effect by reducing the rate of population growth.

15

It is worth nothing that the long-term coefficient of the variable changes in trade volume is not significant in the overall sample. This outcome suggests that there is no significant long-term relationship between inequality and changes in trade volume. This finding is consistent with the recent study by Dollar and Kraay (2004). Furthermore, changes in trade volume is not significant in the sub-samples except in column five, low inflation standard deviation, where the variable enters negatively and at significant levels. Consequently, this result illustrates that countries with macroeconomic stability can improve income distribution owing to the expansion of trade. A 36% increase in trade volume leads to a long-term decline of 1 point in income inequality ((0.003/1-0.90)*36). As for the remaining three variables, they keep the same sign as the results in table 2, across the five columns, while the magnitude of their coefficients does not differ substantially. Only in the case of high governance countries in table 3, column 3, there are two variations that deserve highlighting. Firstly, the secondary school enrolment variable is not significant. If we consider that high governance countries are associated with high levels of secondary school enrolment and educational attainment, it is likely that an upturn in this variable does not impact considerably on income distribution. As a matter of fact, table 2 shows that the coefficient of secondary school enrolment is significant in any scenario, but in the high governance sub-sample, it is the smallest compared with the corresponding coefficients in the other sub-samples. In this case, we suggest that higher levels of education, for instance tertiary education, can provide better benefits to income distribution. Secondly the coefficient of the ratio of FDI to GDP is not significant either. Thus, when we consider changes in trade volume, FDI does not seem to exert an adverse effect on income distribution in this kind of economies. In summary, the effect of trade on income distribution seems to be weak and not robust, because changes in trade volume do not have a significant relationship with income distribution in the overall sample and, of the four sub-samples, only those countries associated with macroeconomic stability can obtain benefits from the expansion of trade, but the effect is weak as table 3 illustrates. In addition, trade volume does not have a significant impact on countries that exhibit domestic inefficiency, and only has a small impact on the remaining scenarios in table 2. Nevertheless, there is some statistical evidence, emerging from the two variables on trade, which illustrates that countries with domestic efficiency are more likely to improve income distribution on account of an upturn in trade. The conclusions for all other variables practically do not change, except for the ratio of FDI to GDP and secondary school education, in high governance countries, as outlined earlier. The effects of exports by sector In the economic liberalisation process trade liberalisation plays a preponderant role, as it is assumed to give an unambiguous boost to the exportable sector and therefore to export-led growth. In this context, export growth may raise employment directly in the exportable sector and indirectly by permitting faster GDP growth. This process is expected to have positive implications for income distribution and longer term growth.

16

Table 3. Scenarios using changes in trade volume (CTRAGDP)


Short-run parametersEHIIt-1 0.845 * 0.677 * 0.956 * 0.754 * 0.900 *CTRAGDP -0.001 -0.003 -0.000 -0.003 -0.003 *FDIGDP 0.060 * 0.129 * 0.006 0.170 * 0.026 *INFL 0.000 * 0.000 * 0.000 * 0.000 * 0.003EDUSEC -0.016 * -0.021 * -0.001 -0.025 * -0.007 *Constant 6.889 * 14.081 * 1.747 * 11.060 * 4.215 *

Long-run parametersCTRAGDP -0.007 -0.010 -0.007 -0.013 -0.028 *FDIGDP 0.388 * 0.398 * 0.147 0.688 * 0.258 *INFL 0.001 * 0.001 * 0.004 * 0.001 * 0.035EDUSEC -0.101 * -0.066 * -0.019 -0.101 * -0.071 *Constant 44.467 * 43.532 * 40.015 * 44.908 * 42.247 *






All the equations use the sys GMM estimation method We have already contended that the basis for tracing the income-distribution effects of trade is the Stolper-Samuelson theorem. According to this neoclassical model, foreign trade increases the income of the factor which is relatively most used in the export sector and which on the principle of comparative advantage is most abundant, this factor being conventionally assumed to be unskilled labour in developing economies; so income distribution improves (FitzGerald 1996, 32). In this context, we may expect that primary exports, based on natural resources and unskilled labour, provide larger benefits to income distribution than manufactured exports in those countries which exhibit low governance and high inflation standard deviation, since these countries are associated with lower levels of development and therefore their comparative advantages are based on natural resources and unskilled labour.14 In order to test the effect of manufactured exports and primary exports on income distribution we conduct a regression including two variables representing the percentage of manufactures exports and primary exports to merchandise exports. Table 4 illustrates that the proxy of manufactures exports is negative and statistically

14 Kaufmann et al (1999) provide evidence of a strong positive relationship between governance and better development outcomes. In this sense, countries with low level of governance are associated with low level of development and therefore they can have substantial reliance on natural resources and unskilled labour.

17

significant in the overall sample and in the four sub-samples. An upturn of 8.9 points in the rate of manufactured exports to merchandise exports is linked with a long-run decline of 1 point in income inequality in the overall sample (0.013/(1-0.885)*8.9). On the other hand, the proxy of primary exports is not significant in any case, even in the low governance and high inflation variance sub-samples. To some extent, this outcome undermines theoretical pillars supporting the economic liberalisation process, such as the principle of comparative advantages and neoclassical assumptions. Table 4. Exports by sector to total exports


Short-run parametersEHIIt-i 0.885 * 0.736 * 0.925 * 0.860 * 0.879 *

MANEXP -0.013 * -0.011 * -0.006 * -0.012 ** -0.011 **PRIEXP 0.001 0.009 0.003 0.021 -0.010Constant 5.039 * 11.107 * 3.116 * 6.039 * 5.151 *

Long-run parametersMANEXP -0.112 * -0.041 ** -0.077 * -0.084 ** -0.091 *PRIEXP 0.005 0.033 0.039 0.151 -0.081Constant 43.850 * 42.114 * 41.693 * 43.154 * 42.717 *






All the equations use the sys GMM estimation method The role of employment and government size It has already been noticed that employment is deemed to be one of the main factors that can drive a better distribution of income over the economic liberalisation process. So as to test this assumption, we extend the original equation by adding a proxy for the level of employment. We include the unemployment rate variable. Results are shown in table 5, column 2. In addition, so as to test the role of employment by sector, we include variables of employment by industry and agriculture to total employment. Results are provided in table 5, columns 3 and 4 respectively. Finally, we undertake an additional exercise to explore the role of the state and its empowerment in the distribution of income. For this purpose, we use the government expenditure to GDP variable in order to represent the size of the state. Results are provided in table 5, column 5. In all the equations, the exercise is conducted only for the whole sample because of constraints on data availability. That is, after including the unemployment,

18

employment by sector and government expenditure variables, the number of observations drops. Firstly, we observe that unemployment enters positively and significantly in the equation, indicating that higher levels of employment are associated with less inequality. A 4.75 points reduction of unemployment drops the inequality indicator by 1 point over the long-run ((0.020)/(1-0.904)*4.75). What is striking is that trade volume does not remain significant. However, this result is not owing to the inclusion of unemployment in the equation but rather on account of the reduction of observations in the sample. As a matter of fact, when we conduct the regression (not reported) with the same sample comprising 781 observations and dropping the unemployment variable it is worth noting that trade volume is neither significant. When we explore the effect of employment by sector on income distribution it is interesting to note that the employment in industry to total employment variable is significant and negative. An increase of 2.4 points in this ratio is required to reduce inequality by one point over the long-term ((0.007/1-0.982)*2.4). It should also be added that the variables FDI to GDP and secondary school enrolment are no longer significant in this equation. However, as in the case of the unemployment variable this outcome is not due to the inclusion of an additional variable in the equation, in this case employment in industry to total employment. This outcome is rather caused by the reduction of the sample. On the contrary, when the regression is conducted (not reported) with the same sample (690 observations) but dropping the employment in industry to total employment variable both FDI to GDP and secondary school enrolment are not significant either. On the other hand, what is striking is that the employment in agriculture to total employment variable is not significant. In this sense, results in table 4 and 5 are supportive and complement each other. In table 4 we show statistical evidence suggesting that the growth of manufactured exports decreases inequality, while in table 5 we illustrate that employment in industry is associated with improvements in income distribution. In contrast, from the two previous tables there is not statistical evidence suggesting that primary exports and employment in agriculture can reduce inequality, even in those countries associated with lower levels of development and abundant unskilled labour such as countries with low governance and high inflation standard deviation. Consequently, results in table 4 and table 5 columns 3 and 4 undermine the theoretical foundations supporting the economic liberalisation process. As for government expenditure, it enters negatively and significantly, while the other variables remain significant. An upturn of 15 points on government expenditure as a ratio of GDP drops the inequality indicator by 1 point over the long-run ((0.0044/(1-0.932))*15). Hence, the re-empowerment of the state is a factor that can improve income distribution. This finding is in keeping with the prescription advocated by the PWC.

19

Table 5. Level of employment (UNEMP), employment by sector (EMPSEC) and government size (GOVEXPEN)

Whole Unemployment Employment in Employment in Governmentsample (1 rate (2 industry to total Agric. to total Expenditure (5

employment (3 employment (4Short-run parametersEHIIt-i 0.841 * 0.904 * 0.982 * 0.981 * 0.932 *TRAGDP -0.004 ** 0.000 -0.001 * -0.001 * -0.001 *FDIGDP 0.094 ** 0.023 * 0.006 0.009 0.024 *INFL 0.000 * 0.001 * 0.002 * 0.002 * 0.000 *EDUSEC -0.016 * -0.009 * 0.001 0.000 -0.004 *UNEMP 0.020 *EMPIND -0.007 *EMPAGR 0.000GOVEXPEN -0.004 *Constant 7.306 * 4.219 * 0.947 * 0.804 ** 3.174 *

Long-run parametersTRAGDP -0.027 * -0.002 -0.031 ** -0.032 ** -0.021 *FDIGDP 0.594 * 0.236 * 0.356 0.476 ** 0.347 *INFL 0.001 * 0.010 * 0.100 * 0.089 * 0.005 *EDUSEC -0.102 * -0.096 * 0.030 0.014 -0.064 *UNEMP 0.212 *EMPIND -0.417 *EMPAGR 0.000GOVEXPEN -0.064 *Constant 45.959 * 44.019 * 53.995 * 43.109 * 46.347 *






All the equations use the sys GMM estimation method 5) Concluding remarks From the statistical evidence presented above, we find that a strong state, associated with higher levels of government expenditure, is important to reduce inequality. We also find that countries with macroeconomic stability and high governance can mitigate the adverse effect of FDI on income distribution, while there is some evidence that they can obtain weak benefits from trade liberalisation, in terms of income distribution. In contrast, countries that exhibit domestic inefficiency do not benefit from trade and the effect of FDI on their distribution of income is worse. Inflation, which can be a proxy of fiscal discipline, worsens inequality, but this effect is weak and seems to exert a real impact on income distribution only in episodes of hyperinflation. Not surprisingly, this variable is not statistically significant in the sample comprising countries with a stable economy.

20

In addition, education socialises the operation of market forces, since it reduces inequality, although there is some evidence that this effect is weak on high governance countries. We claim that one of the reasons for this finding is that tertiary education can be more effective to reduce inequality in high governance countries, as they are associated with higher secondary school enrolment and educational attainment. In short, domestic efficiency, the empowerment of the state, and the set of second generation policies, represented in this study by a proxy of human capital formation, can mitigate the adverse effect of economic liberalisation on income distribution, and to some extent can reduce inequality. Consequently, the PWC and second generation reforms represent an improvement in the global process of economic liberalisation. It has been argued that exports increase employment in the exportable sector and this trend facilitates the distribution of income. In this study we confirm that higher employment reduces inequality. However, when we explore the effects of employment and exports by sector we find that emphasis on the primary sector is not a strategy that improves income distribution, even in those countries that to some extent can be associated with comparative advantages based on natural resources and unskilled labour. On the other hand, we find that the expansion of exports and employment through emphasis on the industrial sector can have better consequences for income distribution. Therefore, the effects of exports and employment by sector on income distribution illustrated in this study are in keeping with what might be called “The new Keynesian approach”. This approach stresses that those attempts to exploit comparative advantages based on natural resources and unskilled labour do not form an appropriate basis for sustained export growth, since successful export growth depends on technological innovation and the improvement of skills in the labour force; furthermore, from this perspective such an alternative strategy has better consequences for income distribution than the neoliberal model. In this concern FitzGerald points out that from the new Keynesian view industrialisation can raise productivity and hence labour incomes throughout the economy. Moreover, manufactured products require greater investment in human capital formation which also leads toward higher labour incomes. In addition, manufactured exports involve higher rates of both public and private investment. This set of events strengthens domestic markets and forms the basis for sustained economic growth. He also explains that from this critical viewpoint reliance on natural resources and unskilled labour does not solve the problems of depressed real wages and slow rate of growth, which leads to weak domestic markets, recessive fiscal retrenchment and structural unemployment, resulting in a worsening of income dispersion and absolute poverty. In contrast, with an appropriate industrialisation wage cuts are unnecessary, domestic markets are stronger, faster growth increases employment and welfare expenditure and eventually inequality declines (FitzGerald 1996, 34-35) FDI increase inequality under any scenario and the effect is worse on those countries that exhibit any form of domestic inefficiency; although there is some evidence that its effect is not significant in high governance countries when we replace trade

21

volume by changes in trade volume. Accordingly, the effect of FDI on income distribution tends to be adverse and it does not provide any distributional effect. This finding undermines the neoclassical postulate which holds that the international flow of investments increases efficiency in resource allocation between and within countries. This assertion faces opposite arguments in the literature that may be generally supportive of the results that emerge from this study in terms of FDI. Some of these arguments are summarised as follows The race to attract new inward investment and/or to retain multinational enterprises results in subsidy packages, downward pressure on corporation and income taxes, and in general in tax incentives and tax reductions. Such tendencies have three major adverse consequences that can be outlined with the following statements. Bailey, et al stress that fiscal policies designed specifically to serve transnationals’ interests lead to an evaporating tax base that constrains the scope for redistributive and social expenditure by governments (Bailey, et al 1998, 296). In addition, Held, et al notice that the perceived capacity of MNCs to shift production might reduce countries’ ability to tax capital, reduces their tax base and increasingly moves the burden of taxation on the less mobile factors such as labour (Held et al 1999, 277). Finally, Easterly argues that preferential tax treatment and other incentives to induce inward FDI may place domestic industry at a disadvantage and may also introduce a distortion affecting domestic investment. He adds that such distortion between the return to foreign and domestic capital could have a strongly negative effect on growth and employment (Easterly 1993). Thus, the corporate and capital tax competition between nations that has emerged on account of capital liberalisation and privatisation is likely to result in declining social expenditure because of a reduction in the tax base, or a rising tax burden on the less mobile factors. As a matter of fact, Held et al notice that corporate tax rates among countries have tended to fall since the early 1980s (loc. cit). This set of characteristics, in which FDI operates, can be a justification of our results. The increasing bargaining power of MNCs is another possible cause of inequality. Privatisation of state-owned firms and liberalisation of FDI encourage a surge of mergers and acquisitions across borders that tend to create dominant positions and oligopolistic markets. This likely pattern decreases the market power of small and medium sized enterprises (SMEs) and leads to a deterioration of the domestic industry and to capital concentration.15 In addition, the ability of MNCs to organise production transnationally and to shift production to reap the benefits of low wages increases corporate power relative to the power of labour, putting downward pressure on wages and working conditions. In this respect the globalisation of production may contribute to widening wage differentials between skilled and unskilled workers within and between countries16. Hence, the fact that the balance of power between multinational capital and social actors (the state, SMEs, and labour, among others), under conditions of liberalisation, may shift in favour of the former, represents an explanation of the likely adverse effect of FDI on income distribution.

15 For an elaboration of the expansion and challenges of cross-border mergers and acquisitions see UNCTD World Investment Report: Cross-Border Merger and Acquisition and Development (New York: United Nations Publication, 2000), pp. 15-28 16 For a discussion about the balance of power among labour and capital see David Held, et al, Global Transformations: Politics, Economics and Culture (Cambridge: Polity Press, 1999), pp. 278-280

22

The operation of MNCs impacts on the effectiveness of the traditional tools of macroeconomic management and government economic policy in several ways. Held, et al most especially underline two ways. The effectiveness of national monetary policy may be compromised since MNCs can borrow abroad when domestic interest rates are high, and conversely take advantages of low domestic interest rates to borrow to fund project overseas. In addition, MNCs may also play an important role in exchange rate markets. In this sense, although speculators may initiate an attack on a currency, it is when both MNCs and institutional investors shift out that currency, even as a precautionary measure, that pressure on the exchange rate can become irreversible (Ibid, 276-277). As a consequence, if the monetary and exchange rate policy are aimed at supporting an efficient allocation of resources and redistributive actions, the apparent erosion in the effectiveness of these policies adversely affect the distribution of income. Arguments such as tax competition between nations, the increasing bargaining power of MNCs, and the erosion of national macroeconomic policy instruments are neglected in the neoliberal thesis, and provide a different perspective to explain the adverse effect on income distribution that may arise owing to the flow of FDI. As for the variables on trade, we find that there is a weak benefit of the expansion of trade on income distribution when this policy is represented by trade volume, and this effect is significant only in those countries that exhibit domestic efficiency and in the overall sample. There is no significant relationship between trade and income distribution in the overall sample when the former variable is represented by changes in trade volume, and it is significant only in the sub-sample comprising countries with low inflation standard deviation. We have pointed out that international trade is traditionally considered the corner stone supporting distributional effect in the process of economic liberalisation (Bulmer-Thomas 1996, 10). Therefore, we should expect a larger benefit from this variable on income distribution but the weak evidence above is not supportive of this assumption. Consequently, the empirical evidence obtained from this study is in keeping with the assumptions and expectations supporting the set of first generation reforms to the extent that employment and low inflation can benefit income distribution. However, the results undermine these assumptions and expectations because the benefit of trade on income distribution is weak and FDI worsens inequality. In addition, an export-led growth strategy and the expansion of employment based on the primary sector do not improve income distribution Although second generation reforms represent an improvement in the economic liberalisation process, the empirical evidence in this study suggests that the role of the state is not enough to socialise the operation of trade and investment. In this context, even under conditions of macroeconomic stability and high governance FDI does not benefit income distribution; in contrast, it seems to be adverse. Moreover, the beneficial effect of trade on income distribution is weak in those countries with domestic efficiency. For these reasons, we argue that further supranational mechanisms, beyond the scope of the state, are required to obtain distributional effects from trade liberalisation and from the flow of FDI.

23

References � Arellano, Manuel & Bond, Stephen (1991) ‘Some Tests of Specification for Panel

Data: Monte Carlo Evidence and an Application to Employment Equations’, the Review of Economic Studies, 58:2, pp. 277 – 297.

� Arellano, Manuel and Bover, Olympia (1995) ‘Another Look at the Instrumental Variable Estimation of Error-Component Models’ Journal of Econometrics, 68, 29-51.

� Bailey, David; Harte, George and Sugden, Roger (1998) ‘Debate: Transnational Corporations. The Case for a Monitoring Policy Across Europe’ New Political Economy, 3:2, pp. 279-300.

� Baltagi, Badi H. (2001) Econometric analysis of Panel Data. Sussex: John Wiley & Sons, 2nd Ed.

� Blundell, Richard and Bond, Stephen (1998) ‘Initial Conditions and Moment Restrictions in Dynamic Panel Data Models’ Journal of Econometrics, 87, pp. 115-143.

� Bourguignon, F. (1994) ‘Growth, Distribution and Human Resources’. In G. Rains, eds., In Route to Modern Growth, Essays in Honour of Carlos Diaz-Alejandro. Johns Hopkins University Press: Washington, DC, pp. 43-69.

� Breusch, T and Pagan, A. (1980) ‘The LM Test and Its Applications to Model Specification in Econometrics’, Review of Economic Studies, 47, 99. 239-254

� Bulmer-Thomas, Victor (1996) ‘Introduction,’ in Victor Bulmer-Thomas, eds., The New Economic Model in Latin America and its Impact on Income Distribution and Poverty. London: Macmillan Press LTD, pp. 7-26

� Calderón, César and Chong, Albert (2001) ‘External Sector and Income Inequality in Interdependent Economies Using a Dynamic Panel Data Approach’ Economics Letters, 71, pp. 225 - 231

� Camdessus, Michel (1998) ‘Income Distribution and Sustainable Growth: the Perspective from the IMF at Fifty’, in Vito Tanzi and Ke-Young Chu, eds., Income Distribution and High-Quality Growth. Massachusetts: MIT Press, pp. xi-4

� Chong, A and C, Calderón (2000) ‘Institutional Quality and Income Distribution’ Economic Development and Cultural Change, 48:4, pp. 761-786

� Corden, W. M. (1993) Protection and Liberalisation: a Review of the Analytical issues, Washington , D. C. IMF Occasional Paper 54.

� De Gregorio, José and Lee, Jong-Wha (2002) ‘Education and Income Inequality: New Evidence from Cross-Country Data’, Review of Income and Wealth, 48:3, 395-416

� Deininger, Klaus and Squire, Lyn (1996) ‘A New Data Set Measuring Income Inequality’ The World Bank Economic Review, 10:3, pp. 565-591.

� Deininger, Klaus and Squire, Lyn (1998) “New Ways of Looking at Old Issues: Inequality and Growth” Journal of Development Economics, 57, pp. 259-287

� Dollar, David and Kraay, Aart (2004) ‘Trade, Growth, and Poverty’ The Economic Journal, 114, pp. 22-49

� Doornik, Jurgen A. et al (2002) ‘Panel Data Estimation Using DPD for OX’ Centro de Estudios Financieros y Monetarios. Available at http://www.cemfi.es/~arellano/

� Edwards, Michael (1999) Future Positive: International Cooperation in the 21st Century . London: Earthscan.

� Easterly, W. (1993) ‘How much do distortions affect growth’ Journal of Monetary Economics, 32, pp.187-212.

24

� Fitz Gerald E. V. K. (1996) ‘The New Trade Regime: Macroeconomic Behaviour and Income Distribution in Latin America’ in Victor Bulmer-Thomas, eds., The New Economic Model in Latin America and its Impact on Income Distribution and Poverty. London: Macmillan Press LTD, pp. 29-52.

� Forbes, K. (2000) ‘ A Reassessment of the Relationship Between Inequality and Growth’ American Economic Review, 90, pp. 869-887

� Galbraith, James K. and Kum, Hyunsub (2002) ‘Inequality and Economic Growth: Data Comparison and Econometric Tests’, UTIP Working Paper Number 21.

� Galbraith, James K. and Kum, Hyunsub (2003) ‘Estimating the Inequality of Household Incomes: Filling Gaps and Correcting Errors in Deininger & Squire’, University of Texas Inequality Project Working Paper No. 22

� Gilpin, Robert (1987) The Political Economy of International Relations. Princeton, Oxford: Princeton University Press.

� Greene, William H. (2000) Econometric Analysis. New Jersey: Prentice Hall, 4th Ed.

� Hausman, J. A. (1978) ‘Specification Test in Econometrics’, Econometrica, 46, pp. 1251 – 1271.

� Havrylyshyn, Oleh; Miller, Marcus and Perraudin William (1994) ‘Deficits, Inflation and the Political Economy of Ukraine’ Economic Policy, 19, pp. 353-401.

� Held David, et al (1999) Global Transformations: Politics, Economics and Culture. Cambridge: Polity Press

� Heerink, Nico (1994) Population Growth, Income Distribution, and Economic Development: Theory, Methodology, and Empirical Results. New York and Berlin: Spriner-Verlag.

� Higgott, Richard (2000) ‘Contested Globalisation: the Changing Context and Normative Challenges’ Review of International Studies, 26, 131-153

� Higgot, Richard and Phillips, Nicola (2000) ‘Challenging Triumphalism and Convergence: the Limits of Global Liberalisation in Asia and Latin America’, Review of International Studies, 26, pp. 359-379.

� International Monetary Found (1997) World Economic Outlook: Globalisation and the Opportunities for Developing Countries. May issue, Chapter 4

� Jacobsen, Peter W. F. and Giles, David E. A. (1998) ‘Income Distribution in the United Sates: Kuznets’ Inverted-U Hypothesis and Data non-Stationary’, The Journal of International Trade & Economic Development, 7:4, pp. 405-423.

� Jha, S. (1996) ‘The Kuznets Curve: a Reassessment’, World Development, 24:4, pp. 773-780.

� Jones, R. J. Barry (1988) ‘Liberal Political Economy’, in Jones, eds., The worlds of political Economy, pp. 27-56

� Kaufmann, Daniel; Kraay Art and Zoido-Lobaton, Pablo (1999) “Governance Matters” World Bank Policy Research Working Paper No. 2196. Available at http://www.worldbank.org/wbi/governance/pubs/govmatters.html

� Litwin, Carol (1998) ‘Trade and Income Distribution in Developing Countries’ Working Papers in Economics, Göteborg University, 9, pp. 1-42

� Luxembourg Income Study – LIS (2003) The LIS Database, Available at http://www.lisproject.org/

� Milanovic, B. (1995) ‘Poverty, Inequality and Social Policy in Transition Economies’ Transition Economics Division, Research Papers 9. Washington, DC, The World Bank.

25

� Ortiz, Guillermo (2003) ‘Latin America and the Washington Consensus: overcoming reform fatigue’, Finance and Development, September 2003, pp. 14-17.

� Williamson, John (1990) ‘What Washington Means by Policy Reform’, in John Williamson eds., Latin American Adjustment: How Much Has Happened? Washington DC: Institute for International Economics, pp. 7-20.

� UNESCO Institute for Statistics (2003) Statistical Tables . Available at http://portal.unesco.org/uis/ev.php?URL_ID=5275&URL_DO=DO_TOPIC&URL_SECTION=201

� United Nations Conference on Trade and Development (2000) World Investment Report: Cross-Border Mergers and Acquisitions and Development. New York: United Nations Publication.

� United Nations Conference on Trade and Development (2003) Foreign Direct Investment Database. Available at http://www.unctad.org/Templates/Page.asp?intItemID=1923&lang=1

� United Nations University/World Institute for Development Economics Research (2003) World Income Inequality Data Base. Available at http://www.wider.unu.edu/wiid/wiid.htm

� University of Texas Inequality Project (2003) ‘An estimated household income inequality data set for the global economy’. Available at http://utip.gov.utexas.edu

� University of Texas Inequality Project (2003) UTIP – UNIDO Database. Available at http://utip.gov.utexas.edu/

� World Bank (2003) Deininger and Squire Data Set. Available at http://www.worldbank.org/research/growth/dddeisqu.htm

� World Bank (2003) World Development Indicators, CD-ROM (2002). Washington, DC. Described in http://www.worldbank.org/data/wdi2002/cdrom/

26

Appendix List of countries (93) Algeria Greece NigeriaArgentina Guatemala NorwayAustralia Guinea PakistanAustria Haiti PanamaBangladesh Honduras PeruBarbados Hong Kong PhilippinesBelgium Hungary PolandBolivia Iceland PortugalBotswana India RussiaBrazil Indonesia SenegalBulgaria Iran SingaporeBurundi Ireland SloveniaCameroon Israel South AfricaCanada Italy SpainCentral African Rep Jamaica Sri LankaChile Japan SurinameColombia Jordan SwazilandCosta Rica Kenya SwedenCote d'Ivoire Korea Syrian Arab RepublicCyprus Lesotho ThailandDenmark Malawi TogoDominican Republic Malaysia Trinidad and TobagoEcuador Malta TunisiaEgypt Mauritius TurkeyEl Salvador Mexico UgandaEthiopia Morocco UkraineFiji Mozambique United KingdomFinland Nepal United StatesFrance Netherlands UruguayGabon New Zealand VenezuelaGhana Nicaragua Zimbabwe

Documents

The effects of economic liberalisation on income ... · The economic liberalisation process, undertaken on a global scale since the early 1980s, has induced support for a set of market-oriented