# This table reports instrumental variable iv

• 55

This preview shows page 49 - 52 out of 55 pages.

This table reports instrumental variable (IV) regression results for analyses that examine the effect of country-level ETR on country-level growth in real GDP or employment. The sample period is 1995 to 2013. All non-indicator variables are winsorized at the top and bottom one-percentiles. The t -values, reported in parentheses, are based on standard errors clustered at the country level. In the second stage regressions, Aggregate ETR it is instrumented by Log(Tax visit) it and by Other ETR it . Log(Tax visit) it is calculated as natural logarithm of one plus number of visits or required meetings with tax officials in country i and year t . Other ETR it is calculated as the weighted average of aggregate ETR of other countries (other than country i ) in our sample in year t, weighting by the inverse of the logarithm of the distance between the two countries’ capital cities. All other variables are defined in Table 2. The t - values, reported in parentheses, are based on standard errors clustered at the country level. Other variables are defined in Table 2. First Stage Regression Second Stage Regression Dependent variable = Aggregate ETR it Dependent variable = Real GDP growth it+1 Dependent variable = Employment growth it+1 1 2 3 Aggregate ETR it 0.105 (−3.18) 0.034 (−2.36) Statutory tax rate it 0.085 (1.42) 0.154 (3.46) 0.017 (1.03) Control variables Real GDP growth it 0.061 (−0.77) 0.277 (4.92) 0.082 (3.18) Log(Population) it 0.077 (1.45) 0.030 (1.15) 0.003 (0.22) Log(Tax visit) it 0.037 (3.10) Other ETR it 31.702 (−9.67) Intercept Yes Yes Yes Country fixed effects Yes Yes Yes Year fixed effects Yes Yes Yes N 511 511 511 Adj. R 2 0.792 0.665 0.304 Kleibergen-Paap LM statistic 26.106 (p-value = 0.00) Kleibergen-Paap F statistic 55.704 Hansen's J statistic 0.309 (p-value = 0.58) 2.403 (p-value = 0.12) Electronic copy available at:
49 Table 7 Replication of Mendoza, Milesi-Ferretti, and Asea (1997) This table reports replicates results in Table 5 of Mendoza, Milesi-Ferretti, and Asea (1997). The sample period is 1995 to 2013. All non-indicator variables are winsorized at the top and bottom one-percentiles. The t -values, reported in parentheses, are based on standard errors clustered at the country level. Taxcon, Taxlab and Taxcap are computed as described in Mendoza et al. (1997) and are the effective tax rate on consumption, labor income and capital income respectively. All other variables are defined in Table 2. The t -values, reported in parentheses, are based on standard errors clustered at the country level. Dependent variable = GDP per capita growth it Observation level 5-year averages Annual 1 2 3 4 Aggregate ETR it 0.099 (−2.96) 0.074 (−2.45) Taxcon it 0.029 (1.07) 0.030 (1.15) 0.052 (1.17) 0.046 (0.92) Taxlab it 0.031 (−1.41) 0.016 (−0.75) 0.036 (−0.48) 0.052 (−0.67) Taxcap it 0.019 (−0.67) 0.025 (−1.03) 0.055 (0.62) 0.103 (1.41) Log(GDP 1995) it 0.004 (−2.24) 0.001 (−0.67) Intercept Yes Yes Yes Yes Country fixed effects No No Yes Yes Year fixed effects Yes Yes Yes Yes N 96 96 451 451 Adj. R 2 0.205 0.289 0.547 0.563 Electronic copy available at:
50 Table 8 Additional tests of assumptions underlying H1 Panel A reports sub-sample results from OLS estimation models that examine the effect of country-level ETR on country-level growth in real GDP or employment. The model is estimated separately for poor and strong control of corruption score countries. Control of corruption scores data are obtained from the Worldwide Governance Indicators and sample is divided into poor- and strong-control of corruption at the median of corruption index.