If it seems like impeachment has made the news more often of late, that’s true, and not just in the United States. Although impeachments are still rare events, they have occurred more frequently in the past several years than at any other time in the past half-century, and the outcomes of those events seem to be shifting as well.
Koto’s Knowledge Team recently compiled a dataset on all impeachments of national heads of state since 1970. The chart below shows when those events occurred and how they turned out. We sorted the outcomes into three categories.
- Unsuccessful. Impeachment proceedings began and maybe came to a vote, but the head of state survived the process and remained in office (e.g., Bill Clinton in 1998, or Romania’s Traian Basescu in 2007 and again in 2012).
- Successful. The head of state was impeached and removed from office (e.g., in 2016, Brazil’s Dilma Rousseff and South Korea’s Park Geun Hye).
- Resignation. The head of state resigned when impeachment proceedings were threatened or underway (e.g., Zimbabwe’s Robert Mugabe in 2017 and Jacob Zuma of South Africa in 2018).
Impeachments and their outcomes worldwide since 1970
We see a couple of trends in that chart. One is the aforementioned increase in the frequency of these events over time. All but two of the 35 impeachment processes in the last 50 years occurred after the early 1990s, and 15 of those have happened since 2012.
The increased incidence of impeachments coincides with growth in the share of countries worldwide under partially or fully democratic governments, and those two trends are almost certainly linked. The chart below shows the distribution of democracy scores for countries worldwide over the same period, using the Varieties of Democracy Project’s annual polyarchy index as our measure of that concept. Each translucent dot represents one country’s score in a given year, while the solid blue line shows change over time in the global distribution of those dots. The uptick in impeachments coincides with a significant upward shift in the blue line, and a simple statistical model confirms that impeachments are more likely to occur in countries with higher democracy scores.
Global trend in degree of democracy, 1970-2017
The other trend we see in the timeline of impeachment events is an apparent shift in the outcomes of these processes. From the early 1990s until the mid-2000s, slightly more than half of these cases ended with the successful removal of the impeached executive from office. Since 2007, however, the most common outcome has been the resignation of the threatened executive (eight times), followed by his or her survival in office (six times). During that period, chief executives have been removed by impeachment only four times.
This apparent shift could occur by chance, but it may also result from a learning process. After seeing so many successful impeachments in the 1990s and early 2000s, today’s leaders may have adjusted expectations about the odds of their survival in office, and they may be acting accordingly.
We would love to use more advanced statistical modeling or machine learning techniques to test our theories about these trends over time and their causes, but there are too few of these events to get the analytical traction we would need to learn much from that exercise. We do plan to add these events to our Analogy Engine, however, so we and other users of that tool can more confidently compare cases where impeachments happened, and see how those events affected politics and the economy in those countries. And, last but not least, we expect in the not-too-distant-future to have indices that let us track changes over time in the prevalence of news about impeachments and other forms of government and social instability across countries, as an efficient way to summarize and maybe even predict these trends around the world.
- We used a logistic regression model with country-years as the units of observation, the occurrence of an impeachment event as the target, and lagged values of V-Dem’s polyarchy and official corruption indexes as the features. Both higher democracy scores and higher levels of corruption were strongly associated with a higher probability of an impeachment event, although the coefficient for democracy was about 50% larger than the one for corruption. We also tried a logistic regression model with a multiplicative interaction term, and a generalized additive model with smoothing splines for both terms (to allow for nonlinear effects), but neither improved on the goodness of fit of the simpler model.