ask one question on this: These studies suffer from at least three limitations that may explain the inconclusiveness of their results. First, the inequality data set they use has a very large proportion of missing values. For example, the widely used Gini coefficients data set of Deininger and Squire (1996) - used by Boix (2003) among others-contains only 11 percent of the country-years during the period it covers (Houle 2009). Even the most recent studies typically have around 30 percent missing observations. For example, the democratization models of Freeman and Ouinn (2012) contain a maximum of fifty- four autocracies, and even for these countries many years are missing. Such high levels of missingness are likely to affect results significantly because unavailable observations are not missing at random