Non-technical summary

Since doi moi in 1986, the Vietnamese economy has been growing rapidly, with impressive poverty reduction. Yet there is evidence of rising spatial inequality, especially comparing lowland and urban areas with more remote and mountainous areas. There is also a steadily increasing divergence of per capita comes across the 61 provinces. There are two plausible explanations for these spatial disparities. First, the marked differences across provinces in terms of geographical characteristics, human capital, and physical and natural resource endowments might have led to distinctive growth paths. Second, growth and globalization have dramatically altered economic structure in Vietnam, thereby altering factor demands and prices; however, growth “shocks” have been very specific in terms both of sectors and regions. A large share of capital investment has concentrated in a few urban areas, while natural resources suitable for export crops, notably coffee, are concentrated in a few provinces, such as the Central Highlands.

Despite this inherently uneven growth, theory suggests that factor mobility should bring about income convergence. Given the immobility of capital and natural resources, theory suggests that labor markets and migration should become vital channels for transmitting gains of growth and globalization from the booming regions to the rest of the economy. We examine this assertion. We first investigate the patterns and determinants of inter-provincial gross migration flows for two periods, 1984-1989 and 1994-1999, using Census data. We then examine the role of migration as an influence on the speed of per capita GDP convergence across provinces.

Descriptive and historical analyses show that much migration in the 1980s was planned by the government and so not always based on economic incentives. It was mostly rural-rural migration from the North (especially the Red River Delta and North Central Coast) to the New Economic Zones (NEZs) in the Central Highland and the South. Migration in the 1990s, on the other hand, was more spontaneous and responded to income differentials. Rural-rural migration in the 1990s remained important, and its provincial composition was highly correlated with earlier flows. At the same time, rural-urban migration emerged as another dominant flow in the 1990s.

We test econometrically the extent to which these flows can be explained by distance, past migration flows (which proxy for the effect of migration networks), and characteristics of sending and receiving provinces such as per capita income or per capita land. We use ordinary least squares to examine the determinants of inter-provincial gross migration flows for both periods, with 1984-1989 flow being an explanatory variable for the 1994-1999 regression. The results largely confirm theoretical predictions as well as descriptive analysis: people move from low-income to high-income provinces, and from land-scarce to land-abundant provinces. Longer distance is associated with smaller migration flows, although the effect of distance has diminished over time. Past migration is one of the most important determinants of current migration. In addition, we find evidence of a liquidity constraint effect and/or spatial poverty traps; the per capita income of a sending province is found to be a significant determinant of migration flows only when migration cost is high (i.e., when distance is large or when past migration flow is small). In short, econometric analysis robustly confirms economic motives for migration but also suggests the existence of poverty-related labor immobility at the provincial level.

Regarding the link between migration and convergence, we first run a provincial-level growth regression. The dependent variable is average per capita GDP growth 1995-2005, and the independent variables include per capita GDP in 1995 and other initial province characteristics. The result robustly shows that once we control for province characteristics, a convergence rate of about 2% is found. We then add 1990s migration data to the regression to examine its impact on convergence. If migration increases the speed of convergence, then the estimated convergence rate is expected to become smaller when migration is controlled for. To our surprise, we are unable to reject the null hypothesis that migration has an effect on the speed of convergence. Indeed, if anything migration appears to increase somewhat the rate of provincial income divergence.

Theoretically, several models in the literature assert a negative impact of migration on convergence. The ‘new economic geography’ theories consider the possibility of increasing returns to scale and agglomeration effects: if there are economies of scale in production and the wage in the richer province is an increasing function of its population, then labor migration only facilitates income traps in poorer provinces. Even within the framework of a neo-classical growth model, it has been argued that migration, while raising returns to labor, reduces the returns to capital in the sending economy and hence discourages capital accumulation, which is bad for growth. Finally, other theories attribute the divergence result to a kind of “brain drain” from poorer provinces, as outmigrants are typically better-educated, more skilled and perhaps more entrepreneurial than those who stay at home. Whatever the cause, if this result is robust, then it remains to be seen whether migrants’ remittances are sufficient to compensate for the losses of the economies from which they depart.

The current version of the complete paper is available at www.aae.wisc.edu/coxhead/papers