Milanovic (2005, World Bank Economic Review)： 自分のためのノート

備忘録として、読んだ論文の内容。
世銀のWYDという、1988年と1993年と1998年の各国の所得の10分位階級ごとの情報が得られるデータを用いて、
Trade opennessがinequalityに与えた影響を考察している論文。

Milanovic (2005)
 Lundberg and Squire (2003) consider growth and inequality to be determined simultaneously and find that openness, measured by the Sachs-Warner indicator, has either no effect or a mild negative effect on inequality.
 Ravallion (2001) find statistically significant nonlinearity in the relationship between openness and inequality, with openness associated with increased inequality in poor countries.
 Dollar and Kraay (2002) find that openness defined as exports plus imports as a share of GDPP is positively associated with per capita income growth and that this effect is the same for the bottom income quintile as for the mean—trade has no systematic impact on inequality.
 Birdsall and Londono (1997, 1998) report no differences in growth in income between the poorest and other quintiles due to trade variables, although initial distributions of land and education do matter.
 Li et al. (1998) use the ratio of exports to GDP as an explanatory variable for the Gini coefficient and find no statistically significant effect of openness on the Gini coefficient.
 Ravallion (2006) is doubtful about both types of findings because the studies depend on fairly noisy data and work with average only.
 Li et al. (1999), Birdsall and Londono (1998), and Dollar and Kraay (2002) find that openness has no systematic and significant effect on inequality. Lundberg and Squire (1999), Barro (2000), and Ravallion (2001) find that openness has a negative effect on equality in poor countries and that in some of the formulations it has a negative effect on the real income of the poor. There are no results that show openness reduced inequality, that is, no results showing that openness raises the real incomes of the poor proportionately more than the incomes of the rich.
 This article uses World Income Distribution (WYD), which provides information on income levels by deciles or even finer partitions.
 Inequality is assumed to depend on openness measured as (export+import)/GDP, FDI/GDP, M2/GDP and an indicator of democracy.
 Ten-level regressions are estimated with simultaneous decile estimation and GMM estimation.
 Openness has negative impact on mean-normalized incomes of poor and middle-income groups (1st decile – 6th decile) and positive impact on mean-normalized 10th decile income. If regional dummies and the five-year average rate of inflation are added, openness reduces the share of the bottom seven deciles and raises that of the top two.
 The negative effect of openness is smaller in richer countries. Openness would seem to have a particularly negative impact of poor and middle-income groups in poor countries.
 FDI is not statistically significant on mean-normalized incomes of any groups.
 Financial depth (M2/GDP) increases he income share of the poor and middle class and reduce the share of the top decile.
 This study define openness as the trade to GDP ratio, both expressed in nominal dollar terms. Dollar and Kraay (2002) define it as the ratio between trade in 1985 dollars and GDP in 1985 international dollars. Using PPP dollars to express the denominator significantly increase GDP for poor countries and thus reduces the importance of trade in their GDP. When the previous regression is run with the Dollar and Kraay measure of openness, then openness is no longer significant for any decile, nor is the interaction between openness and income, that is, openness does not matter for income distribution.
 The Dollar and Kraay measure substantially underestimate the levels and changes of openness, which makes their measure fail to detect much of a relationship between openness and inequality because the measure is artificially sluggish.