ANALYSIS OF THE EFFECT OF REMITTANCES ON ECONOMIC GROWTH USING PATH ANALYSIS Violeta Diaz University of Texas-Pan American 20 W. University Dr. Edinburg, TX 78539, USA. vdiazzz@utpa.edu Tel: +-956-38-3383. ABSTRACT In this paper I propose two different channels, a direct channel and an indirect investment channel, for remittances to influence economic growth. By using data from Giuliano and Ruiz- Arranz (2006) I employ path analysis to test these channels. This is the first academic paper, that I know of, that analyzes a causal relation between workers remittances and economic growth. In addition I am adding to the current literature, the first remittance paper to employ path analysis. The results, from the estimated models, indicate that remittances only influence growth indirectly through investment. This path is statistically significant, supporting a mediating effect of investment between workers remittances and economic growth. INTRODUCTION Workers remittances are becoming an important discussion topic in the world agenda. The importance of this monetary flow has increased considerably in recent years. According to the World Bank, estimated official remittances were US$67 billion for developing countries in 2005. However, the relationship between remittances and growth has not been frequently studied. Foreign direct investment and Financial Aid, on the other hand, are topics commonly related to economic growth. The main reason for this oversight is that remittances are mostly related to spending on consumption goods, and the vast majority of policy makers and academicians believe that remittances do not influence investment patterns. They also believe that remittances do not have a positive impact on economic growth. These pessimistic approach, on how remittances are spent, is general (Adams, 2005). Several studies have helped to prove that remittances help households move out of poverty (Adams, 2003), lower mortality rates (Kanaiaupuni, 998) and increase educational and housing spending (Adams, 2005). Furthermore, empirical studies show that remittances can stimulate economic activity and motivate entrepreneurial communities (Durand et al, 996 and Woodruff and Zenteno, 2007). According to Buch et al (2002), remittances can influence economic growth directly or indirectly. However, the degree of the latter channel strongly depends on supporting governmental policies and a supporting economic environment for investment activities. There are only four recent empirical studies that have analyzed the relation between workers remittances and growth. Chami et al (2005) uses panel data to study the moral hazard effect framework on remittance s motivation and their effect on economic activity. They find a negative effect of remittances on economic growth. Glytsos (2005) analyzes the effect of 683
remittances on investment, consumption, imports and output. The author uses a sample of five countries and estimates short and long run multipliers of remittances. He finds that the effect of reducing remittances would be greater than the effect of raising them. Giuliano and Ruiz-Arranz (2005) find a positive effect of remittances on growth, specifically for countries with lower financial development. Finally, Ziesemer (2007) proposes a savings channel that relates remittances with growth. He finds that remittances have a positive impact on growth, due to the ability to increase saving rates in countries with a per capita income of less than US $200. In this paper, path analysis is used, given that the main purpose is to analyze the direct and indirect effects of remittances on economic growth. By using Path Analysis, we add to the current literature, the first academic paper that I know of- on this topic. Other studies with similar hypotheses have used other kind of methodologies. Giuliano and Ruiz-Arranz (2006) and Ziesemer (2006) use Generalized Method of Moments (GMM) using pooled data, while Glytsos (2005) and Chami et al. obtain their results using two stage least squares using instrumental variables. In this paper, as previously stated, I also analyze the relation between workers remittances and economic growth, but I am mainly interested in analyzing the investment channel of remittances. I accomplish this by studying the structural relationship among three variables, remittances, investment and growth. My hypothesis is that there is a causal relation between these variables. I find a positive and significant relationship between remittances and economic growth. In addition, I find a positive and significant relationship between investment and growth, as expected. I also find evidence that the effect of remittances on economic growth is indirect. Investment has an important role as a mediator between remittances and economic growth. The next section presents the description of the data, the following describes the methodology, the section after that present the results and I finalize the paper with a discussion. DATA This paper uses data from Giuliano and Ruiz-Arranz (2005). The sample considers 73 countries over the years 975-2002. The authors make significant changes from data used in other studies. They consider a country-specific measure of remittances by following the definitions of each country in the Balance of Payments Statistics Yearbook and also contacted IMF desk economists and country authorities to increase the number of countries considered for the sample. All remittances (remit) are divided by the country s gross domestic product (GDP). Given that the purpose of this study is to analyze the causal relation between remittances, investment and growth, these and other variables are included to find evidence that supports this hypothesis. Growth of real per capita GDP is used as a measure for economic growth (growth) and the gross fixed capital formation divided by GDP as a measure of investment (inv). The following variables are included given that they are particularly related to a country s economic growth: inflation (inf), as a measure of a country s instability, the degree of openness (open) measured by the sum of exports and imports, and the fiscal balance of the government (fiscal), which is particularly related to economic growth and investment, and the amount of claims of the private sector as a proxy for financial development (Giuliano and Ruiz-Arranz). All the previous variables are measured as a share of GDP, except for inflation, which is the annual percentage change of the Consumer Price Index for each country. Finally, population growth (pop) and years of education (school) measured as the average years of secondary schooling in total population, are also included. All these variables, except for workers remittances, are taken from the World Development Indicators of the World Bank. 684
loans e remit inv inf open e2 growth pop school fiscal Figure. Path diagram from step METHODOLOGY Structure Equation Modeling (SEM) is conducted, to probe the structural relationship among the variables in our model. I employ the term Path Analysis instead of SEM (Hair et al, 2005) from now on given that I am using bivariate correlations among variables. Path analysis and SEM offer several advantages over multiple linear regression. Path analysis allows simultaneous analysis of multiple causal relationships among variables, both direct and indirect. Along with the previous advantages, the estimated path coefficients (arrows) can be used for prediction. Path analysis combines econometric analysis of prediction and statistical tests of a priori hypotheses. Previous studies, such as Durand et al (996) and Massey and Parrado (994), find that remittances are linked to economic growth. In this paper, I propose a hypothesized relation between remittances and economic growth through an investment channel. The relation between remittances and investment is expected to be positive, as well as the relation between investment and growth. The complete recursive model, which includes the direct and indirect effects of remittances on growth, is displayed in Figure. RESULTS The bivariate correlation between all variables is presented in Table. On one hand, remittances are positively related to population growth, openness, financial development, and school. On the other hand, remit has a negative correlation with the inflation and the fiscal balance variables. All these values are consistent with previous research. 685
loans inf Growth inv pop open school fiscal remit loans.00 - - - - - - - - inf -0.0.00 - - - - - - - growth 0.8-0.26.00 - - - - - - Inv 0.38 -.040 0.43.00 - - - - - pop -0.23-0.9 0.0-0..00 - - - - open 0.33-0.28 0.5 0.39-0.9.00 - - - school 0.4 0.2 0.06 0.34-0.48 0.32.00 - - fiscal 0.06-0.00 0.2 0.09-0.06 0. 0.4.00 - remit 0.07 -.60 0.4 0.9 0.4 0.27 0.04-0.9.00 Source: Giuliano and Ruiz-Arranz (2006) Table. Bivariate Correlation matrix Parameter estimate Standardized S.E. remit inv 0.09** 0.052 loans inv 0.232*** 0.056 inf inv 0.052 0.053 open inv 0.249*** 0.058 school inv 0.55*** 0.057 pop growth 0.077 0.058 school growth -0.02 0.060 fiscal growth 0.06** 0.05 remit growth 0.052 0.054 inf growth -0.246*** 0.053 open growth -0.03* 0.060 inv growth 0.453*** 0.055 ***, **,* indicate significance levels at %, 5% and 0% respectively. Table 2. Standardized path coefficients including all variables (model ) Path analysis is performed with the complete correlation matrix using AMOS 7.0. Table 2, shows the results for the standardized path coefficients. The χ 2 test, with three degrees of freedom, is not significant (χ 2 = 3.4, p = 0.334) which implies a good fit. Three additional goodness of fit indices indicate that the interpretation of the estimated paths is adequate. The Root Mean Square Error (RMSEA) = 0.02, the Goodness of Fit Index (GFI) = 0.998 and the Adjusted Fit Index (AGFI) = 0.963. However, four path coefficients are not significant even at a 0% level; one of these paths is the hypothesized direct effect of remittances on growth, which even though has a positive value, it is not significant. In the second step, to improve the results, the insignificant paths are dropped. We do so to improve the model fit (see Figure 2). The differences between the two models are the paths from school to growth and from inflation to investment. Another important change is that we do not include all the variables; we left out the population variable. The fit to the observed data of the parsimonious model is definitely better than the previous one (χ 2 = 3.02, df = 4, p= 0.554, 686
RMSEA=.000, GFI= 0.998 and AGFI= 0.978). The significance of the path coefficients remained the same. From this model, the only path that it is still not significant is the direct effect of remittances on growth, which implies that the effect of remittances is definitely through an indirect channel. Nevertheless, the total effect of remittances on growth is the sum of the direct and indirect effects. This means that the direct effect is 0.067, and the indirect effect is 0.04 x 0.447 = 0.046, giving us a total effect of 0.3. loans e remit inv inf open e2 growth school fiscal Figure 2. Path diagram from step 2 In order to confirm this assumption, the final model is estimated and presented in Figure 3. The fit from this model is similar to the previous one (model 2). This time, all the paths are significant (see Table 3); however, I am only considering that remittances have an indirect effect on growth. The total effect is.048 (0.04 x 0.455), which means that, with an increase of one unit of remittances, the economic growth will increase 4.8 percent. The χ 2 test is not significant (χ 2 =4.63, df = 5, p= 0.462), and the other goodness of fit indices (RMSEA=.000, GFI= 0.996 and AGFI= 0.973) indicate that, overall, we still have a good fit. After establishing that both models have a good fit, we need to decide which one is more adequate to fit the data. In addition to compare both models, we can use other goodness of fit tests, the Akaike s information criterion (AIC) and the Schwarz s Bayesian Information criterion (BIC). From model 2, the AIC is 67.02 versus 66.63 from the third model. The former includes the direct and indirect paths, while the latter only includes the indirect path. Based on these measures, the model seems to improve by deleting the direct path from remittances to economic growth. Finally, to assess the validity of the models, the chi-square difference test is performed. This test can be employed given that model 3 is nested in model 2. The results from this test are presented in Table 4. According to the results from the Chi square difference test, which is not statistically significant (Δχ 2 =.63), the model with one additional path does not have a better 687
fit (Hair, 2005). The nested model has the best fit, considering the chi-square difference test and the AIC comparison. According to Sun (2005), if several indicators give consistent results, a conclusion can be drawn with confidence. loans e remit inv inf open e2 growth school fiscal Figure 3. Path diagram from step 3 Table 3. path Parameter estimate Standardized S.E. remit inv 0.04** 0.05 loans inv 0.227*** 0.056 open inv 0.233*** 0.056 school inv 0.68*** 0.056 fiscal growth 0.09* 0.050 inf growth -0.274*** 0.05 open growth -0.4** 0.056 inv growth 0.455*** 0.053 ***, **,* indicate significance levels at %, 5% and 0% respectively. including only an indirect path (model 3) Standardized coefficients χ 2 df p Model 2 (M2) 3.02 4 0.554 Model 3 (M3) 4.643 5 0.462 Difference (M3- M2)*.63 *χ 2 = 3.84 df = Table 4. Chi-square difference test 688
DISCUSSION The topic of the effect of remittances on economic growth is still open for debate. There have been several studies that have identified a positive relation. However, this is the first paper that analyzes a causal relationship between remittances and economic growth. In this paper, two different channels, for remittances to influence economic growth, are proposed. A direct channel and an investment channel. This indirect channel has been the focus of different studies. Several researchers have come across with this hypothesis but none of them have been able to prove it. Given the characteristics of path analysis and the size of the sample, supporting evidence is obtain of a mediating effect of investment between workers remittances and growth. These results support the hypothesis that remittances only influence growth indirectly through investment. This path, even though it is of low magnitude, it is statistically significant. These results are consistent with other studies that have found a positive relation between remittances and investment. In Taylor (999), where the relation between the role of remittances and the new economic of labor migration (NELM) is analyzed, it is accurately stated that the indirect effects of migration strongly depend on the absence of capital constrains on production. After considering these unique results, countries around the world should consider the increasing importance of this non official source of investment, and make structural changes that could improve the investment environment of this remittance receiving countries. Countries receiving high levels of remittances should be aware that, to have a positive impact on their country s GDP, it is necessary to invest this significant monetary flow. REFERENCES Adams, R. H. and J. Page (2003). International Migration, Remittances and Poverty in Developing Coutries. World Bank Policy Research Working Paper 379, December 2003. Adams, R. H. (2005). Remittances, Household Expenditure and Investment in Guatemala. World Bank Policy Research Working Paper 3532, March 2005. Buch, C. M., A. Kuckulenz and M. Le Manchec (2002). Worker Remittances and Capital Flows. Kiel Working Paper No. 30. Chami, R., C. Fullenkamp and S. Jahjah (2003). Are Immigrant Remittances Flows a Source of Capital for Development? IMF Working Paper, September 2003. Durand, J., E.A. Parrado and D.S. Massey (996). Migradollars and Development: A Reconsideration of the Mexican Case. International Migration Review 30 (2), 423-444. Giuliano, P. and M. Ruiz-Arranz (2006). Remittances, Financial Development and Growth. IZA Discussion Paper No. 260. Glytsos, N.P. (2005). The Contribution of Remittances to Growth. Journal of Economic Studies 32 (6), 468-396. 689
Hair, J. F., W. C. Black, B.J. Babin, R.E. Anderson and R.L. Tatham (2005). Multivariate Data Analysis. Pearson Prentice Hall, 6th edition. Kanaiaupuni, S. M. and K. M. Donato (998). Migradollars and Mortallity: The Effects of Migration on Infant Survival in Mexico. CDE Working Paper No. 98-0. Massey, D. and E. Parrado (994). Migradollars: The remittances and savings of Mexican migrants to the USA. Population Research and Policy Review 3, 3-30. Shipley, B., (2000). Cause and Correlation in Biology: A User s Guide to Path Analysis, Structural Equations, and Causal Inference. Cambridge University Press, Cambridge, UK. Sun, J., (2005). Assessing Goodness of Fit in Confirmatory Factor Analysis. Measurement and Evaluation in Counseling and Development, 37, 240-256. Taylor, J.E. (999). The New Economics of Labour Migration and the Role of Remittances in the Migration Process. International Migration 37 (), 63 88. Ziesemer, T. (2007). Worker Remittances and Growth: The Physical and Human Capital Channels. UNU- MERIT Working Paper. Woodruff, C. and R. Zenteno (2007). Migration Networks and Microenterprises in Mexico. Journal of Development Economics 82, 509-528. 690