Institutional quality and the tenure of autocrats

Reviewing Svolik’s list of autocrats, 11 cases were identified in which leaders may not have been autocratic, some because they were democratically elected, and some because their countries were assigned relatively high polity2 scores. The questionable autocrats were Mogae of Botswana, Gustavo of Ecuador, Zedillo of Mexico, Zia of Pakistan, Noriega of Panama, Macchi of Paraguay, Fujimori of Peru, Garcia of El Salvador, Diouf of Senegal, Lee of Taiwan, and Armalino of Uruguay. As a robustness check, all observations for those countries were dropped and the regressions below re-run. One can debate whether they should be included (we favor including them to rely on Svolik’s data without our input), but the results are qualitatively the same whether or not they are included, so the results do not depend on them.

This was also calculated as a percentage change, and the results remain essentially the same regardless of the way EFWrate is calculated.

Data for Oil is from indexmundi.com; for Distance, the latitudes of country capitals are found at http://​lab.​Imnixon.​org/​4th/​worldcapitals.​html, and the calculator used to convert that information into distance from the equator is found at http://​www.​eaae-astronomy.​org/​eratosthenes/​index.​php?​option=​com_​content&​view=​article&​id=​47&​item&​id-68; Died is from Svolik & Akcinaroglu (2007).

This was suggested by a reviewer, to make the coefficients easier to interpret.

Data for PCIrate come from the World Bank Per Capita Income, PPP (Purchasing Power Parity). As with EFWrate, PCIrate was calculated both this way and as a percentage change in PCI, and in specifications not shown in the paper, the results are essentially the same regardless of the way that PCIrate is calculated.

The number of observations varies in the regressions because of data limitations. In regression (1) there are observations for 99 different autocrats, but in regression (2) only 80 of them have observations for all of the independent variables. In regression (3) only 71 autocracies have data for both PCIrate and EFWrate. Regression (4) uses only those countries in regression (2) that have data for PCIrate.

Note that one of the reasons Niskanen (1971) gives for his budget maximization hypothesis is that those in government gain more power with larger budgets, so the idea that power over others generates utility is not completely foreign to public choice analysis.