Do Assassinations Change History?

03/01/2008
Featured in print Digest

A country whose autocrat is assassinated is 13 percentage points more likely to move toward democracy in the following year than a country where the assassination attempt on the autocrat failed.

Do assassinations change history? It's a tangled question, given the many political, military, socioeconomic, and other forces at work. But a new study suggests that political assassinations can change the course of individual nations.

"We find that assassinations of autocrats produce substantial changes in the country's institutions, while assassinations of democrats do not," conclude authors Benjamin Jones and Benjamin Olken in Hit Or Miss? The Effect of Assassinations on Institutions and War (NBER Working Paper No.13102). They also find that the killing of leaders intensifies small-scale conflicts but may hasten the end of large-scale ones. Their analysis demonstrates a violent means through which societies democratize, and more generally illustrates the important role of individual leaders in shaping institutions and conflict. The analysis further suggests the key role that random events - such as whether a bullet hits or misses its target -- can play in shaping events.

One of the startling findings in this paper is how common assassinations are. The authors focus narrowly on nations' leaders - the most powerful individual within a given country. They don't include in their study "coup d'etats," where a group kills the head of state in order to seize power. Also, they only look at "serious" attempts on leaders' lives - incidents where the weapon was actually discharged. (Note: Guns are the most frequently used and effective weapons; bombs are also frequently used but are not very effective, the study finds.) Even with all those restrictions, there have been 298 serious assassination attempts on leaders worldwide since 1875, the authors calculate. Of those, only a fifth - 59 - succeeded.

The number of political assassinations has risen to record levels in recent decades. A national leader has been assassinated in nearly two of every three years since 1950, according to the study. But that's mostly a factor of there being so many more independent countries than a century ago. Individually, leaders are safer than they were a century ago, the authors write. "At the peak in the 1910s, a given leader had a nearly 1 percent chance of being assassinated in a given year; today, the probability is below 0.3 percent."

Given the difficulty of separating cause and effect in history, one innovation of this paper is the development of a method for analyzing whether assassinations actually cause change. To do this, the authors compare successful assassinations with failed assassination attempts. Their key assumption is that, once the weapon is actually engaged (the gun fired, the bomb detonated), whether the attempt succeeds in killing the leader is driven largely by chance. To validate this assumption, the authors show that, once the weapon is discharged, death or survival is largely unrelated to features of the attack (other than weapon) or the situation of the country at the time of the attack. That being the case, they use failed attacks as a "control group" for successful attacks and ask whether national outcomes differ substantially depending on the result of the attack.

Their findings are striking: A country whose autocrat is assassinated is 13 percentage points more likely to move toward democracy in the following year than a country where the assassination attempt on the autocrat failed, the authors calculate. Also, the successful assassination of an autocrat is 19 percentage points more likely than a failed attempt to lead to subsequent leadership changes being made through regular, institutional means. These effects are not just short-term changes. They're still seen a decade or more later.

Assassinations also have an effect on conflict, at least in limited contexts. The researchers find that successful assassinations lead to an intensification of small-scale conflicts and, perhaps, hasten the end of large-scale conflicts. Of course, the authors note, these findings are based on the differences between failed and successful attempts. Thus, it's difficult to tell whether the observed phenomena are caused by successful assassinations, failed assassinations, or both.

To gain some insight into which type of outcome - success or failure - tends to matter, this study includes additional analysis using a method called propensity-score matching. Though less conclusive, that analysis backs up the commonsense notion that successful assassinations have a much bigger impact than failed ones. However, there are some signs that failed attempts trim the chances of a move toward democratization somewhat, perhaps because autocrats can crack down on opposition movements after a failed assassination attempt.

Besides providing evidence for certain theories of democratization, institutional change, and war, this study emphasizes how random events can lead to profound change. "Had Hitler lingered 13 minutes longer in a Munich beer hall in 1939, he would likely have been killed by a waiting bomb," the authors write. "Our tests provide evidence that small elements of randomness -- the path of a bullet, the timing of an explosion, small shifts in a leader's schedule -- can result in substantial changes in national outcomes."

They conclude: "Whether or not assassinations change 'the history of the world' in [British statesman Benjamin] Disraeli's words, they do appear to change the history of individual countries."

-- Laurent Belsie