Robustness Checks
First, we estimate equation (1) in logs and consider several variations on the estimation method, OLS, DOLS, and an ECM using the Vintage data. The estimates of the public capital parameter range from 0.37 to 0.66. We also use different sources for the data and find that public capital is productive in every variation, with the elasticity ranging from 0.32 to 0.58 for Modern Data I and from 0.18 to 0.34 for Modern Data II. We disaggregate public capital into equipment and structures and include them separately in the estimated production function. We find the elasticity estimate for public capital structures is 0.24, while equipment does not appear to have an impact on output. We also estimate a VECM with four equations and one cointegrating vector. The VECM estimate of the long-run public capital elasticity is 0.46 and is significant at the 1 percent level. The speed of adjustment parameter for output is significant and negative, indicating the stationarity of the cointegrating relationship. In a block-exogeneitytest, we find evidencethatlagged values of public capital affect the other variables. We also perform Granger’s causality tests and find that lagged values of public capital affect contemporary levels of private output. 6 Public Finance Review XX(X) However, we cannot reject the null hypothesis that lagged values of output do not affect public capital. The primary causality is from public capital to output.
