Regional Economic Integration and Its Impact on Income Distribution and the Poverty Level: The Case of the WAEMU Zone

There are several theoretical approaches to how regionalisation impacts poverty and income distribution across countries. REI is expected to increase income inequality since economic integration helps create a larger labour market (Alderson, Nielson 2002; Beckfield 2008). Economic integration also increases the amount of trade and investment in the concerned economies. However, if the laws guiding such regional integration are not strong enough, this will lead to income inequality and poverty increase since workers in such economies will now have to compete with more highly skilled workers from other countries. Kuznets (1955) hypothesises that income inequality increases in the early stages of economic development until a certain point, when it starts diminishing, thereby reflecting an inverted U-shaped relationship with economic growth.

The Heckscher–Ohlin (HO) model postulates that international trade leads to an increase in national incomes for participating countries based on comparative advantages of factor endowments. A country produces goods and services in which they have factor endowments and adequate labour supply while importing goods for which they have scarce resources (Cornia 2011). As a result of the operation of the forces of trade, production that is based on capital-intensive techniques is relinquished in favour of labour-intensive approaches that favour exports. This has the benefit of increasing economic growth and reducing inequality in the countries. Stolper and Samuelson's theory also shows that trade openness increases income with abundant resources.

Adequate understanding of the regional integration and poverty reduction relationship can help policymakers to adopt appropriate and suitable economic policies. Bergh and Nillson (2011) examined the role of globalisation on poverty for a panel set for the period 1988–2007. The study employed the panel fixed-effects model and found that globalisation, information and trade restrictions significantly reduced poverty. However, Sharma (2013) found that globalisation increased unemployment, income inequality and human deprivation, and therefore does not contribute to poverty reduction in developing countries. Muriel and Guo (2016) analysed the nexus between regional integration and income inequality for Latin America for the 2000–2013 period. The study employed the fixed-effects model and showed that regional integration reduced inequality; the study also confirmed the existence of an inverted U-shaped relationship between trade and income in the community. Employing the fixed-effects estimation method, Bui et al. (2016) analysed the effect of economic integration on inequality and poverty in Vietnam for the period 2006–2010. The study showed that economic integration had a minimal effect on income distribution and poverty reduction in Vietnam. More recently, Huh and Park (2019) investigated the impact of regional integration and globalisation on economic growth and income inequality of 158 countries for the period 2006–2014. The study employed the fixed-effects model to show that globalisation and extra-regional integration increase economic growth while reducing income inequality.

Castilho et al. (2009) assessed the impact of trade liberalisation on poverty and inequality in Brazilian states for 1987–2005. The Generalised Method of Moments (GMM) results showed that trade increased both inequality and poverty level in the urban zone, while it had adverse effects in rural areas. Cornia (2011) investigated the role of economic integration in the evolution of income inequality and the growth rate in a comparative analysis of Latin America and European countries for the 2000–2008 period. The estimation results suggested that European countries’ GDP figures were higher than those of Latin American countries, although they both experienced a rise in inequality. Muriel et al. (2018) examined the nexus between regional integration and income inequality for the Economic Community of West African States (ECOWAS) for the 2004–2013 period. Using the Least Square Dummy Variable (LSDV), they found that political integration reduced income inequality while economic integration increased inequality in the zone. Park and Claveria (2018b) examined the impact of regional integration on economic growth, income inequality and poverty in over 156 countries for 2006–2016, using the System GMM methods. The results indicated that regional integration had been a significant driver of economic growth, poverty reduction and income distribution.

Javid et al. (2012) examined the impact of remittances on growth and poverty in Pakistan for 1973–2010, using the Auto-Regressive Distributed Lag (ARDL) method. The authors reported a significant and positive relationship between remittances and growth, as well as between remittances and poverty reduction. More recently, Kousar et al. (2019) probed the impact of remittances and financial development on poverty and inequality for Pakistan over the period 1980–2016, using the ARDL-Bounds testing method. The results suggested that foreign remittances increased poverty and inequality in both the long and short terms. Financial development, however, reduces income inequality and the poverty rate in the short term. The study also found support for the inverted U-curve relationship between income inequality and per-capita income in the short term.

Using the multinomial probit model, Adams and Cuecuecha (2013) sought to identify the impact of remittances on investment and poverty in Ghana for the 2005–2006 period and showed that remittances reduced poverty and increased investment. Onyekwena and Oloko (2016) analysed the effects of regional integration on inclusive development (growth, poverty, transport sector, etc.) in the case of the ECOWAS region for the period 1995–2014. The results of the descriptive analysis indicated that regional integration increased growth and extra-regional trade, in addition to having a significant impact on poverty reduction in the region.

Using the Computable General Equilibrium (CGE) model, Nguyen and Ezaki (2005) probed the impact of REI on growth, inequality and the poverty level for Vietnam for 2001. The results suggested that REI increased growth, improved income distribution and reduced poverty in Vietnam. Hartono et al. (2007) investigated the link between regional integration, growth, poverty and inequality in Indonesia for 2000. The CGE results showed that regional integration improved income distribution and poverty reduction. Ezaki and Nguyen (2008) analysed the impact of East Asian economic integration on economic growth, income distribution and the poverty level for 2001. The CGE model results revealed that the East Asian community increased economic growth and improved income distribution, thus reducing the region's poverty rate. Ean et al. (2020) employed the Fully Modified Ordinary Least Square (FMOLS) method to examine the effect of REI on inequality in the Association of Southeast Asian Nations (ASEAN) countries over the 2005–2018 period. The results showed that trade integration, as well as the agricultural and manufacturing sectors, reduced income disparity.

The influence of regional integration on income distribution and poverty has been a hotly debated topic among academics and policymakers alike. Although this subject of research has been extensively researched, previous research on this issue failed to control for common proxies of variables, that is, failed to use the appropriate methodology.

This study examines annual time-series data from 1994 to 2018 for eight WAEMU countries: Benin, Burkina Faso, Cote d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal, and Togo. Table 1 displays the definitions and data sources for selected variables. This study builds on the work of Muriel et al. (2018), and Huh and Park (2019) in the literature to create a weighted indicator for REI. REI is a multifaceted process that goes beyond trade liberalisation measures. Considering simply one factor, such as trade, may understate the impact of regional integration on poverty and inequality reduction. The weighted regional integration is calculated by applying Principal Component Analysis (PCA) to the trade liberalisation (% of GDP), foreign direct investment (% of GDP), migration flow (%) and inflation rate (%) figures for the eight WAEMU countries. The PCA results are shown in Table 11. Using this method helps in discovering useful relevant information about the variables while removing the possibility of multicollinearity in the regression model (Park, Claveria 2018a, b; Huh, Park 2019). For the poverty data, the poverty headcount index from the Foster–Greer–Thorbecke (FGT)

Foster et al. (1984)

This section introduces a theory that connects economic integration and income distribution to the level of poverty. The main theoretical nexus of REI, income inequality and poverty is captured in the HO model. According to this theory, trade increases unskilled workers’ income in developing countries and those of skilled workers in developed countries, thereby reducing income disparity and poverty in developing countries while increasing both in developed countries. According to Meschi and Vivarelli (2009), trade allows for the transfer of new and high technologies between countries using skill-intensive production techniques; as such, the HO model is appropriate for this study. As per Castilho et al. (2009), as well as Bukhari and Munir (2016), the study model is defined as follows: (1) Yi,t=β0EIi,t+∑iβiXi,t+μi+εi,t {Y_{i,t}} = {\beta _0}E{I_{i,t}} + \sum\limits_i {{\beta _i}{X_{i,t}} + {\mu _i} + {\varepsilon _{i,t}}} where Y is the level of income inequality/poverty and EI represents economic integration. X is the set of control variables used in this study. The subscripts i and t represent country-specific and time respectively, while μ_i is an unobserved individual effect and ɛ_i,t is the error term.

Based on the objectives of the studies, the two equations have been expressed separately as follows: (i)

REI and income distribution (2) GINIi,t=α0+α1GINIi,t−1+α2EIi,t+α3REMi,t+α4GDPi,t+α5CCi,t+α6POPi,t+μi+εi,t \matrix{ {GIN{I_{i,t}} = {\alpha _0} + {\alpha _1}GIN{I_{i,t - 1}} + {\alpha _2}E{I_{i,t}} + {\alpha _3}RE{M_{i,t}} + } \cr {{\alpha _4}GD{P_{i,t}} + {\alpha _5}C{C_{i,t}} + {\alpha _6}PO{P_{i,t}} + {\mu _i} + {\varepsilon _{i,t}}} \cr }

The study employed a dynamic analysis since past values of income distribution and poverty can affect present values of income distribution and poverty. The past values of GINI_i,t−1 and POV_i,t−1 are thus associated; moreover, Arellano and Bover's (1995) System of Generalised Method of Moments (SYS-GMM) can be employed to control the individual and time-specific effects, as well as to counteract the endogeneity bias of the variables induced by the lagged values. However, the method cannot be used here since it accounts for ‘short time and large panel members’. The alternative model is the Pooled Mean Group (PMG), introduced by Pesaran et al. (1999). It provides for short-run coefficients and a country-by-country variability in adjustment speed. Long-run coefficients, on the other hand, are homogeneous since they are similar across countries. Finally, the magnitude of T and N is important in helping to avoid bias in average estimators, in addition to helping to resolve the issue of heterogeneity. Both models are estimated through the PMG. (4) Δ(yi)t= αiECTi,t+∑j=1p−1γjiΔ(yi)i−j+∑j=1q−1δjiΔ(Xi)t−j+φi[(yi)t−1−{β0i+β1i(Xi)t−1}]+εi,t \matrix{ {\Delta {{\left( {{y_i}} \right)}_t} = \,{\alpha _i}EC{T_{i,t}} + \sum\limits_{j = 1}^{p - 1} {\gamma _j^i\Delta {{\left( {{y_i}} \right)}_{i - j}} + \sum\limits_{j = 1}^{q - 1} {\delta _j^i\Delta {{\left( {{X_i}} \right)}_{t - j}} + } } } \cr {{\varphi ^i}\left[ {{{\left( {{y_i}} \right)}_{t - 1}} - \left\{ {\beta _0^i + \beta _1^i{{\left( {{X_i}} \right)}_{t - 1}}} \right\}} \right] + {\varepsilon _{i,t}}} \cr } where y is employed for income inequality and the poverty headcount, X stands for the independent variables such as REI, control of corruption, remittances, economic growth and population growth, γ and Δ denotes the short-run coefficients of lagged variables, ECT means error-correction term, β represents the long-run coefficients and φ is the coefficient representing the speed of adjustment to the long-run equilibrium. The subscripts i and t represent country and time, respectively. All variables are in logarithm except for the income inequality index, regional integration index, corruption index and population growth.

Table 2 displays the findings of the descriptive statistics. The findings indicate that the average poverty rate is very high (53.61%) in the WAEMU countries and the average values (mean) GINI and REI are low (22.05 and 0.00, respectively). On average, the value of remittances is equal to 3.68% and economic growth per capita is 779.83, while the institutional quality value is −0.66 and population growth value is 2.84%. The maximum values of POV, GINI, REI, REM and GDP are, respectively, 92.84%, 49.9, 5.40, 10.71% and 1,692.54. The standard deviation values are large enough to explore the variations in the data. The results also show that only poverty and the income distribution index are adversely skewed, whereas others are positively skewed. The Kurtosis values reveal that poverty, income inequality, remittances, economic growth and control of corruption have platykurtic distributions, whereas other variables have leptokurtic distribution. According to the Jarque–Bera test statistics, only poverty, as well as control of corruption, fulfils the requirements of normal distribution.

The outcomes of the correlation matrix and the Variance Inflation Factor (VIF) coefficients tests are presented in Tables 3 and 4, respectively. The correlation analysis shows a positive link between poverty and population growth (0.21), while income disparity displays a positive link with REI, economic growth and control of corruption. Moreover, the association between economic growth and poverty is negative and has a moderate value, which means that they are perfectly correlated. REI, economic growth and population growth are positively correlated. In addition, financial integration regressors and economic growth are positively correlated. A negative correlation exists between remittances and control of corruption regressors.

The results in Table 4 show that the VIF coefficient is equal to 1.63, which means that there is no collinearity among the regressors in the study. Both tables indicate no multicollinearity among the variables.

Since countries in the study are interlinked via WAEMU, a CD test is necessary. The study thus employs the CD test of Pesaran (2004), as well as the Lagrange multiplier (LM) tests of Breusch Pagan (1980) and Pesaran, to check for dependence in the panel data. The null hypothesis is that there is no CD among the regressors.

Table 5 shows that only the results of the Pesaran CD test of poverty, income inequality, control of corruption and population are not significant, thus suggesting rejection of the null hypothesis. The overall results show that the variables are significant at a 1% level, and thus the null hypothesis of no CD is rejected.

CD is important in deciding which stationary test and cointegration tests are necessary for the study. Having shown that there is CD in the model, the second-generation unit root tests (CIPS and CADF)

Cross-section Augmented Im-Pesaran-Shin (CIPS) and Cross-section Augmented Dickey-Fuller (CADF).

Table 6 presents the results of the unit root test of the variables. The results suggest that poverty, income inequality, REI and population fulfilled the criteria of the CIPS test and are therefore stationary at level. To arrive at stationarity for all the remaining variables, the data are first differentiated. The results of the CADF test also show that poverty, income inequality and population growth are stationary at level, and to arrive at stationarity for all the remaining variables, the data are first differentiated. Therefore, the study concludes that although some of the variables are not stationary at levels, they all become stationary at first difference at a 1% level of significance. There is a presumption of cointegration among the regressors.

The problem of long-run cointegration is resolved via the Westerlund (2007) test, which employs four error-correction tests that are divided into two groups. The first group (Gt and Ga) assumes that the whole panel is cointegrated, while the second group (Pt and Pa) assumes that at least one cross-section is cointegrated (Zafar et al. 2019). The null hypothesis states that no cointegration exists between the variables. The results in Table 7 reveal that Gt and Pt are significant at 1%, implying that the null hypothesis is rejected; therefore, the variables are cointegrated in the long run.

The PMG models are estimated according to Eq. (4) and Table 8 summarises the findings of the models. As for the GINI model, the link between REI and income inequality is statistically significant and positive in the long run, with a 1-unit rise in economic integration increasing income disparity by 0.068. In other words, increased economic integration increases income disparity and reduces the equal distribution of wealth. Regional integration through migration and inflation increases population, thereby reducing job opportunities and increasing unemployment as well as income inequality. This correspondingly explains why population growth significantly increases income inequality in the region, with REI benefitting mostly highly skilled workers. Cornia (2011), Bui et al. (2016) and Muriel et al. (2018) reported similar results.

The coefficient of remittances with respect to the GINI index are negative and not significant in short and long runs. Keeping other things equal, a 1% increase in remittances decreases income inequality and improves income distribution. The findings indicate that remittances reduce income inequality and are a substantial contributor to income increase in the countries under investigation. This is explained by the fact that remittances to the poor help increase their income, thus enabling them to invest in education and get more employment opportunities so that income inequality can decrease. Economic growth and population growth increase income disparity both in short- and long-run estimates owing to the poor or negative growth experienced by member countries. The worsening state of income inequality might be justified here by the slow GDP growth rates of WAEMU member countries; moreover, the increasing population is also not helpful in regulating the distribution in the region. As a result, the findings have implications for both short-term economic growth and long-term population increase.

Controlling corruption increases income inequality in the short run while decreasing it in the long run.

As for the poverty model, the relationship between REI and poverty headcount is statistically significant and negative over time, with a 1-unit increase in economic integration significantly increasing poverty reduction by 0.076%. In other words, increased economic integration reduces poverty, possibly because it increases economic activities, thus leading to higher GDP per capita and resulting in poverty reduction. Greater participation in integration significantly contributes to poverty reduction. Regional integration through trade and foreign direct investment (FDI) increases domestic savings, productivity (through knowledge spillovers) and welfare, which also possibly creates job opportunities for the population and leads to a decrease in unemployment as well as income inequality and, hence, the poverty rate. Correspondingly, this explains why the income distribution index significantly reduces the poverty rate in these countries. Since the integration of the countries, important transformations have also been achieved that lead to improvements in economic activities, growth and income distribution. Nguyen and Ezaki (2005), Hartono et al. (2007), Ezaki and Nguyen (2008), as well as Park and Claveria (2018) reported similar results. The findings may support Agénor's (2004) J-curve hypothesis, which states that growing globalisation may raise poverty in the short run while lowering it in the long run. The coefficient on the lagged POV is statistically significant, implying that lagged poverty has an effect on the contemporaneous poverty headcount.

The coefficient of remittances with respect to poverty is significantly positive in the long run. Keeping everything else constant, a 1% increase in remittances increases poverty by 0.17%. These findings agree with those of Kousar et al. (2019) for Pakistan. This is explained by the negative effects of remittances and financial instability; however, it is contrary to the general belief that remittances lead to poverty reduction in poor households. Per-capita economic growth, control of corruption and population growth significantly increase poverty in long-run estimates, thus suggesting that poverty reduction in the WAEMU bloc depends on corruption and population control.

This work investigates time-series models for robustness using the DOLS approach, the findings of which are also provided in Table 9. It is also important to examine the elasticities of income distribution and poverty, as well as other variables, for each WAEMU member country. For this, the study employed the DOLS techniques designed by Saikkonen (1991) and Stock and Watson (1993). This method controls for endogeneity and autocorrelation problems by adding the leads and lags of first-difference regressors. Table 9 summarises the findings.

REI has a positive and significant influence on income disparity in Benin, Burkina Faso, Côte d’Ivoire and Togo, and a positive and significant impact on poverty in Côte d’Ivoire, Mali, Senegal and Togo, according to the results. Table 9 reveals that the coefficient estimates of income inequality are positively and statistically significant, with respect to poverty in Benin, Guinea-Bissau, Niger and Senegal. Remittances significantly reduce poverty in Côte d’Ivoire, Mali and Senegal and significantly improve income distribution in Burkina Faso, Cote d’Ivoire, Niger, Senegal and Togo. Economic growth per capita shows a negative and significant impact on income disparity for Benin, Mali and Niger and significantly reduces poverty in Benin, Burkina Faso, Guinea-Bissau, Mali, Niger and Senegal.

The results in Table 9 also show that control of corruption only has a negative and insignificant impact on income disparity for Burkina Faso, probably because income distribution insignificantly improves with a reduction of corruption. It however significantly reduces the poverty level in Benin, Mali, Niger and Senegal. Population growth significantly increases income inequality in Guinea-Bissau, Niger, Senegal and Togo and significantly increases the poverty rate in Benin, Burkina Faso, Guinea-Bissau, Mali, Niger, Senegal and Togo.

The study employed the Dumitrescu and Hurlin (2012) panel causality method to reveal the relationship between the variables, given the presence of cointegration between poverty/income distribution and its determinants. The results of this test are shown in Table 10. REI and poverty have a bidirectional causal relationship, implying that REI causes poverty and vice versa. The results also suggest that there is a unidirectional causal relationship between REI and income disparity, as well as between economic growth per capita, control of corruption, population growth and income inequality. The findings also point out a bidirectional causal relationship between REI and remittances, showing that REI causes remittances and vice versa. Table 10 shows that remittances, economic growth and population growth cause poverty, that is, poverty has a unidirectional causal relationship with remittances, economic growth per capita and population growth.