More on Democratic Consolidation and Time

In yesterday’s post, I used survival analysis to look at the relationship between the age of a democratic regime and the risk of democratic breakdown. From those estimates, I concluded that traditional thinking about that relationship was probably wrong. Other things being equal, democracies are actually more likely to fail as time passes, at least up until their late teens or early 20s.

That post drew some incisive comments from Joe Wright, a professor at Pennsylvania State University who has done some terrific work with similar data and methods. One of Joe’s recommendations was to look at the same relationship in subsets of democracies to see if there are any telling variations.

So, let’s do it.

In his comments, Joe suggested subsetting either by the type of democracy (presidential vs. parliamentary) or by the type of authoritarian regime that immediately preceded each episode of democracy. Unfortunately, the data set I’m using doesn’t include information about either of those, so I’m going to take a different tack. Here, I’m going to compare baseline hazards (the term of art for the relationship we’re plotting) across groups defined by: a) whether or not there has been at least one transfer of power between political parties since the “birth” of the democratic regime; and b) the degree of protection for civil liberties (categorized as low, medium, and high).

The figure below juxtaposes plots for groups defined by any alternation in power. In democracies where no alternation in power has yet occurred (the plot on the left), we see essentially the same pattern we saw in the full sample. In democracies where at least one alternation has already occurred (on the right), we see no real association between regime age and risk.

The next figure does the same thing with groups defined by the strength of civil liberties. Here, I split the sample into three groups: low (5-7 on Freedom House’s scale); medium (3-4); and high (1-2). As it happens, there are only two instances of democratic breakdown during my period of observation in the last of those bins–which Freedom House calls “consolidated democracies”–so I’ve set that group aside and just plotted baseline hazards for the low and medium groups. The results are similar. Once again, the pattern in the higher-risk subset–the democracies with the weakest protections for civil liberties–is of increasing risk over time, while the pattern in the lower-risk subset is of no real association. (When comparing these plots to each other and to the preceding ones, note that the scales on the vertical axes sometimes vary.)

I read those results as evidence that traditional thinking about the relationship between the passage of time and democratic consolidation is biased by a selection effect. Yes, the risk of a reversion to authoritarian rule is lower in older democracies than it is in younger ones, but that’s really because the democracies most susceptible to breakdown have already been weeded out. For fragile democracies, the risk of breakdown actually increases over time, unless and until they manage to transform themselves into lower-risk cases by producing an alternation in power or deepening protections for civil liberties. After that, there is essentially no association between the passage of time and the prospects for regime survival. (For a classic illustration of selection bias at work, see here.)

Leave a comment


  1. I’m a bit curious about the dataset – can you provide examples of democracies that collapsed in their fifth or sixth decade?
    And how are “odds of democractic breakdown” calculated?

    • According to this data set, the democracies that were “oldest” at the time of their breakdowns in the past half-century were: Venezuela in 2005 (47 years old), Sri Lanka in 1982 (35), Lebanon in 1976 (34), Uruguay in 1972 (31), Philippines in 1972 (29), and Honduras in 2008 (29). The log odds of democratic breakdown are estimates from a generalized additive model with the form of a logistic regression and smoothing splines for regime duration alongside a few other control variables. This is a form of survival analysis for discrete-time data (i.e. where observations about survival time and status are made at routine intervals).

  2. (oops, my name was left in the Hebrew for some reason. This is the English spelling).

  1. Democratization by Heart Attack? The Peculiar Case of Malawi « Dart-Throwing Chimp
  2. Transitions: Worrying about South Africa | EthioSun
  3. Indonesia’s Elections Offer Some Light in the Recent Gloom | Dart-Throwing Chimp

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

  • Author

  • Follow me on Twitter

  • Follow Dart-Throwing Chimp on
  • Enter your email address to follow this blog and receive notifications of new posts by email.

    Join 13,612 other followers

  • Archives

%d bloggers like this: