Acemoglu et al. (2019) provide evidence that democracy has a significant and robust positive effect on log GDP per capita. They estimate a dynamic panel data model of log GDP per capita for 175 countries over the period 1960-2010 using 4 lags on the dependent variable. They introduce a new dichotomous measure of democracy that consolidates the information from several sources. Their results show that democratizations increase GDP per capita by about \(20 \%\) in the long run. The data set and Stata programs are available on Acemoglu's web site https://economics.mit.edu/faculty/acemoglu/ data.

(a) Replicate the first 4 columns of Table 2 of Acemoglu et al. (2019) which run the FE dynamic panel model with one, two, four, and eight lags of log (GDP per capita) and democracy. Both country and time dummies are included. In particular, verify that their preferred specification of 4 lags, reported in column 3 , yields a positive and significant effect of democracy on log (GDP per capita) with a long-run effect of 21.24 and standard error of (7.22).

(b) Replicate the next 4 columns of Table 2 of Acemoglu et al. (2019) which run the Arellano and Bond (1991) one-step dynamic panel model with one, two, four, and eight lags of log (GDP per capita) and democracy. Time dummies are included. In particular, verify that their preferred specification of 4 lags, reported in column 7, yields a long-run effect of democracy on log (GDP per capita) of 16.45 with a standard error of (8.43).

\[\tilde{\delta}_{W 3 S L S}=\left[\widetilde{Z}^{\prime}\left(\Sigma_u^{-1} \otimes P_{\tilde{X}}\right) \tilde{Z}\right]^{-1}\left[\tilde{Z}^{\prime}\left(\Sigma_u^{-1} \otimes P_{\tilde{X}}\right) \tilde{y}\right]\]

\[\begin{equation*}
\ln C_{i t}=\alpha+\beta_{1} \ln C_{i, t-1}+\beta_{2} \ln P_{i, t}+\beta_{3} \ln Y_{i t}+\beta_{4} \ln P n_{i t}+u_{i t} \tag{8.43}
\end{equation*}\]