RDP 2002-03: International Financial Liberalisation and Economic Growth 5. Capital Flows and Growth: Some New Results

5.1 Data and Methodology

Our empirical analysis employs annual data for a set of 40 countries, consisting of 20 developed and 20 emerging and developing countries in Asia, Latin America and Africa.^[9] The sample period spans 1976–1995. The choice of countries and the sample period are dictated by data availability. The estimations use a panel regression framework, in which data for each country are averaged over five non-overlapping years. Averaging the data for a number of years helps abstract from short-term business cycle effects and capture the longer-run effects of capital mobility on growth.

The base specification of our model is as follows:

The dependent variable growth is the annual growth rate in real GDP per capita for country i averaged over each 5-year interval t. The first set of explanatory variables includes our state variables – the stock of human capital (proxied by the average years of secondary education in the adult population) and the level of real per capita GDP – both measured at the beginning of each 5-year interval. In the neoclassical framework, the coefficient on the initial per capita GDP captures the rate of convergence (i.e., the rate at which poor countries catch up with rich countries) and is expected to be negative.^[10]

The second set of explanatory variables includes a number of control variables that have been found to be important determinants of growth by previous studies. The coefficient on the openness to trade variable (proxied by ratio of the sum of total exports and imports to GDP) is expected to be positive. Government consumption is expected to have a negative effect on growth. Similarly the black-market exchange rate premium, which we use to proxy financial market distortions, is also expected to negatively affect growth.

The third set of variables includes the different capital flow measures we use. The broadest measure we use is total capital inflows. We also consider the three main components of capital inflows – foreign direct investment, portfolio inflows and bank inflows. The different measures are entered sequentially into the regressions.

An important consideration in these regressions is the possible endogeneity of financial liberalisation and capital flows. As noted by Kraay (1998) there are two main sources of endogeneity. The first is that capital flows themselves may be influenced by economic performance. If a country relaxes controls in ‘good’ times and imposes them in ‘bad’ times, we would find a spuriously large positive effect of liberalisation on growth. Another source of endogeneity is that the extent of capital mobility may be correlated with other fundamental determinants of growth and investment. For example, Grilli and Milesi-Ferretti (1995) observe that countries with small public sectors and relatively independent central banks are less likely to impose capital controls. If having a small public sector and an independent central bank were good for growth, then the benefits of capital account liberalisation would be overstated. In principle, this problem can be addressed by using instrumental variables that are correlated with financial openness, but uncorrelated with the disturbance term. Finding good instruments, however, is difficult.

In selecting the instruments for our estimations we draw on the literature on the determinants of capital flows. Following the work of Calvo et al (1993) a number of studies have sought to explain the movements in capital flows by looking at the relative importance of the external (‘push’) factors and internal (‘pull’) factors. Their findings suggest that US interest rates have played a dominant role in driving capital flows to developing countries. We also use total flows to developing countries to reflect broader supply-side factors. Other instruments include lagged capital flows, lagged GDP growth, and change in the terms of trade.

5.2 Main Results and Discussion

Table 2 presents the main results from our regressions. Regression 2.1 is the base regression without the capital flow variables. The results are consistent with theory and previous empirical findings. The coefficient on initial GDP per capita is negative and statistically significant suggesting strong convergence. Education has a positive effect on growth, but the coefficient is not statistically significant. Openness to foreign trade has a positive and significant effect on growth. The coefficients on black market premium and government spending are both negative and significant.

Regressions 2.2–2.5 augment the base regression with the different measures of capital flows. Total flows have a positive effect on growth, with the coefficient significant at the 10 per cent level. Regressions 2.3–2.5 look at FDI, portfolio, and bank flows individually. Foreign direct investment and portfolio flows have a statistically significant positive effect on growth. Bank flows have a negative but statistically insignificant effect.

Given our focus on the effect of capital flows on developing countries, we next consider the results for the developing countries in our sample (Table 3). The results for the base regression do not differ markedly from those of the full sample. Capital flows, however, are found to have a negative effect on growth, though the coefficient is not statistically significant. As in the full sample case, foreign direct investment and portfolio flows both have a statistically significant positive effect on growth. Bank flows are found to have a statistically significant negative effect on growth. These results are also economically significant. For example, an increase in FDI of 1 percentage point would result in a 0.40 percentage point higher real per capita growth rate per year. A 1 percentage point increase in portfolio flows is associated with a 0.35 percentage point increase, whereas a 1 percentage point increase in bank inflows results in a 0.33 percentage point decline in the real per capita GDP growth rate.

Note that we do not include the investment rate in these regressions, even though investment is an important determinant of economic growth. This has implications for the interpretation of the effect of capital flows on growth. The coefficient on the capital flow variables without investment captures the effect of capital flows on growth through all possible channels, including through investment. The coefficient on capital flow variables with investment on the other hand, captures the effect of capital flows on growth above and beyond its effect on total investment. When investment is included in the regression, the effect of FDI on growth is positive but no longer statistically significant at conventional levels. While the coefficient on portfolio flows becomes marginally smaller, it is statistically significant at the 5 per cent level. This suggests that portfolio flows affect economic growth above and beyond their effect on domestic investment. The coefficient on bank flows remains negative and statistically significant at the 5 per cent level.

In order to check the robustness of these results, we introduce a variety of changes to our specification. These include replacing the black market premium with the measure of the size of the banking sector (bank assets/GDP), adding a measure of institutional strength (proxied by an index of law and contract enforcement), using a currency crisis dummy, and dummies for the 1980s to represent the period of the debt crisis and the ‘lost decade’ for the Latin American countries. Our findings for FDI and portfolio flows remain fairly robust to these changes. While the coefficients on bank flows remain negative, they are not always statistically significant.

Our findings are consistent with the conventional wisdom on the composition of capital flows. Foreign direct investment has historically played a larger role in developing countries than have other forms of capital flows. Though some countries have experienced periods of large bank inflows, they haven't been sustained over time. For the countries in our sample, FDI constituted the largest component of capital flows followed by portfolio flows and bank flows. During 1976–1998 average annual foreign direct investment represented 1.4 per cent of GDP, and portfolio flows and bank flows were approximately 1.1 and 0.5 per cent of GDP. Similarly, simple measures of volatility indicate that FDI was the most stable form of capital flows, while bank flows were the most volatile. For instance, the coefficient of variation of annual FDI, portfolio and banks flows to our sample countries during 1976–1998 was 1.2, 2.8 and 4.8 per cent respectively. Given that bank flows have been small and volatile, it is likely that they have not made a meaningful contribution to investment. Our results also suggest that portfolio flows affect growth above and beyond their effect on investment. While the identification of the exact channels is beyond the scope of this paper, the most likely channel (besides investment) through which foreign investment in the domestic equity and debt markets could contribute to growth is through the development and deepening of these markets.

The hypothesis that the quality of domestic financial and regulatory institutions determines the effect of liberalisation on growth is not firmly supported by the data. Our attempts to test this hypothesis by using alternative measures of institutional strength generally produce results that are either statistically insignificant or contradict the hypothesis. The measures we considered included the ratio of liquid liabilities to GDP, the index of law and contract enforcement, and the index of the quality of countries' accounting and reporting standards. Our guess is that this is a consequence of the incomplete and imprecise nature of these measures, and not because institutions do not play a role in this process.

Footnotes

Details are provided in Appendix A. [9]

This property derives from the assumption of diminishing returns to capital – economies that have less capital per worker (relative to their long-run ratio) tend to have higher rates of return and higher growth rates. [10]