6. Sensitivity tests
In this section, we conduct several sensitivity tests to make sure that our major findings are robust to alternative specifications. We first check how sensitive our main findings are when we use different weights in calculating the average growth rates. We then test whether nonlinearity of the tenure variable and the potential measurement error in the age65 variable could cause any bias. Finally, we conduct some sensitivity tests to make sure our findings are not mainly a result of the new retirement policy implemented in early 1980s.
6.1. Weights in average performance
Our previous measure of average growth performance implicitly assumes that each year during a leader's tenure has the same weight in the personnel evaluation. In reality, the weight on each year may not be the same. To test how sensitive our major findings are when we use different weights in calculating the average growth rates, we allow the weight to differ across years.20 More specifically, we create a weighted measure of the GDP growth rate over the tenure T, gwT, which is defined as