604 International Organization TABLE 4.Panel analysis Variables Model 10 Model 11 Model 12 LAGGED FDI 0.364** 0.358** 0.361*** (5.059) (4.952) (5.006) MARKET SIZE -0.554 -0.206 -0.516 (-1.236) (-0.438) (-1.121) DEVELOPMENT LEVEL 0.834* 0.419 0.803* (1.868) (0.886) (1.762) GROWTH 0.024*** 0.024** 0.024*** (2.961) (2.897) 2.867) TRADE 0.006 0.006 0.006 (1.249) (1.402) (1.330) BUDGET DEFICIT -0.023*率 -0.024* -0.024* (-2.187) (-2.272) (-2.261) GOVERNMENT CONSUMPTION -0.039* -0.041** -0.042* (-2.357) -2.444) (-2.508) CAPITAL CONTROLS 0.054** (2.441) FDI INFLOWS CONTROLS 0.002 (0.014) DEMOCRACY 0.021*** 0.021*** 0.019** (2.606) (2.358) (2.224) Time dummies Yes Yes Yes Country dummies Yes Yes Yes Observations 1630 1609 1609 Countries 114 113 113 0.72 0.72 0.72 Note:All regressions are ordinary least squares (OLS)regressions using annual net FDI inflows as a percentage of GDP as the dependent variable. *p<0L,*p<.05,*p<10. els,trade is no longer statistically significant.7 The level of development,when other factors are controlled for,has no statistically significant effects in most models. The picture of economic growth has also changed dramatically from the cross- sectional regressions.Growth is highly significant and positively associated with higher levels of FDI as a percentage of GDP.The difference between this result and the cross-sectional results earlier is most likely attributable to business cycles. When a longer time period is included,it becomes obvious that countries with higher growth rates attract higher levels of FDI. Interestingly,government consumption is highly statistically significant in the fixed-effects models,but only slightly significant in random-effects models.In other words,increases in government consumption have a negative effect on FDI inflows 57.Given the stability of international trade as a percentage of GDP over time,this result is not surprising.Trade is a positive and highly statistically significant determinant of FDI inflows in the random-effects models.els, trade is no longer statistically significant+ 57 The level of development, when other factors are controlled for, has no statistically significant effects in most models+ The picture of economic growth has also changed dramatically from the cross￾sectional regressions+ Growth is highly significant and positively associated with higher levels of FDI as a percentage of GDP+ The difference between this result and the cross-sectional results earlier is most likely attributable to business cycles+ When a longer time period is included, it becomes obvious that countries with higher growth rates attract higher levels of FDI+ Interestingly, government consumption is highly statistically significant in the fixed-effects models, but only slightly significant in random-effects models+ In other words, increases in government consumption have a negative effect on FDI inflows 57+ Given the stability of international trade as a percentage of GDP over time, this result is not surprising+ Trade is a positive and highly statistically significant determinant of FDI inflows in the random-effects models+ TABLE 4. Panel analysis Variables Model 10 Model 11 Model 12 lagged fdi 0+364*** 0+358*** 0+361*** ~5+059! ~4+952! ~5+006! market size 20+554 20+206 20+516 ~21+236! ~20+438! ~21+121! development level 0+834* 0+419 0+803* ~1+868! ~0+886! ~1+762! growth 0+024*** 0+024*** 0+024*** ~2+961! ~2+897! ~2+867! trade 0+006 0+006 0+006 ~1+249! ~1+402! ~1+330! budget deficit 20+023** 20+024** 20+024** ~22+187! ~22+272! ~22+261! government consumption 20+039** 20+041** 20+042** ~22+357! ~22+444! ~22+508! capital controls 0+054** ~2+441! fdi inflows controls 0+002 ~0+014! democracy 0+021*** 0+021*** 0+019** ~2+606! ~2+358! ~2+224! Time dummies Yes Yes Yes Country dummies Yes Yes Yes Observations 1630 1609 1609 Countries 114 113 113 R2 0+72 0+72 0+72 Note: All regressions are ordinary least squares ~OLS! regressions using annual net FDI inflows as a percentage of GDP as the dependent variable+ ***p , +01, **p , +05, *p , +10+ 604 International Organization