Forecasting Democratic Transitions in 2012

2011 was a year of remarkable democratic ferment, as citizens in an unusually large and diverse set of countries took to public spaces to demand more dignity in their lives and more accountability from their governments.

In nearly all cases, the democratization those protesters are demanding remains incomplete. While Occupy participants in the United States rightly decry the occasional act of police brutality against them, the gap yawns widest in countries ruled by authoritarian regimes, which still “occupy” more than two of every five countries, including some of the richest and most populous.

Which of those authoritarian regimes are “ripest” for transitions to democracy in 2012? To help answer that question, I used a statistical technique called Bayesian model averaging to identify and weight a number of risk factors and then applied those weights to the most recent data available. The result is a set of probabilistic forecasts of democratic transition for all countries worldwide currently under authoritarian rule.

For purposes of this forecasting exercise, political regimes are categorized in “either/or” fashion. A regime is considered to be a democracy when it meets all of the four conditions enumerated below. A regime that fails to satisfy any of these conditions is considered to be an autocracy.

1. Elected officials rule. No unelected individuals (say, a king, like Abdullah II of Jordan or Mohammed VI of Morocco) or organization (say, a military junta, like Egypt’s SCAF) determine or direct policy outcomes.

2. Elections are fair and competitive. Elections offer voters a meaningful choice between candidates and are free of widespread fraud and abuse.

3. Politics is inclusive. All adult citizens–male and female, without regard to racial or communal identity–have equal rights to vote and participate in politics.

4. Civil liberties are respected. The government generally recognizes and protects freedoms of speech, association, and assembly.

A transition to democracy occurs when a government chosen by fair, competitive, and inclusive elections takes office (assuming the other conditions enumerated above hold as well). The transition is dated to the installation of the new government, not the elections. This rule avoids treating aborted transitions, such as the one that occurred in Algeria in 1991, as equivalent to the establishment of democracy. Conceptually, the idea is that the authoritarian regime remains in place until a new government is actually installed, and as such, that authoritarian government may veto the transition at any moment until that handover of power.

The chart below plots the estimated likelihood of transition in 2012 for all autocracies worldwide, based on preliminary data from 2011. One thing that’s immediately noticeable about these scores is that they are all pretty low. If you check the scale on the bottom axis, you’ll see that most scores are under 10%, and many are approximately zero. To some extent, that’s an artifact of the rarity of these events. On average, only a few democratic transitions happen worldwide each year, so the easiest way to make a forecast that’s about 95% accurate is simply to say they won’t happen anywhere. The point of an exercise like this one is not to identify precisely which countries will transition when, a task that’s still well beyond the reach of current data and methods (and will probably remain so forever). Instead, it’s better to think of the list as an attempt to identify which of the world’s authoritarian regimes are most likely to experience the few transitions we might expect to see over the course of 2012.

I hope the forecasts stand on their own, but I’ll offer comments on some the results that I found most surprising or intriguing.

* Most surprising to me, Syria ranks among the 10 countries most likely to transition in 2012, while Egypt lands much farther down the list, barely cracking into the top 40. Beyond the nonviolent movements that arose in both countries in 2011, those estimates don’t account for recent events, and any subjective assessment would probably flag Egypt as the more likely case. Nevertheless, I think these estimates do hint at near-term potential for political transformation in Syria while reinforcing the need for caution and concern on the prospects for democratic government in Egypt. (Note: the Egypt forecast assumes that civil liberties improved in 2011 to a 4 on the Freedom House scale. The estimated probability would be slightly higher if that score were a 3, and it would be noticeably lower if that score were a 5 or worse.)

* The forecasts suggest that prospects for a democratic transition in 2012 in Russia improved substantially with the emergence of a nonviolent protest movement after fraudulent legislative elections earlier this month. It ranks 13th on the list, in the same neighborhood as Armenia and (surprising to me) Sudan.

* For China, the analysis confirms the prevailing view that the structural potential for a democratic transition remains low, but it also underscores the point that China’s transition prospects will improve if and when civil liberties expand or the economy suffers a sharp downturn.

* Many of the countries found to be most likely to transition soon are in sub-Saharan Africa: Burkina Faso, UgandaMali, Kenya, and Senegal represent five of the top 10, while MozambiqueMalawi, Madagascar, and Guinea-Bissau can also be found in the top 20. Also notable is Nigeria‘s presence just a few notches further down the list, in the 23rd spot. These forecasts suggest that the good-news story of accelerating economic growth on that continent may coincide with another regional wave of fresh attempts at democracy.

* Where sub-Saharan Africa looks especially promising, Central Asia looks especially bleak. Uzbekistan, Kazakhstan, and Turkmenistan land at the bottom of the list, and Tajikistan perches just a few notches higher. These grim forecasts are driven by those countries’ lack of democratic experience, their exceptionally repressive regimes, and their oil and gas wealth.

For readers who want a peek under the hood, I can tell you that these estimates are generated by an algorithm that accounts for just a few things. (Of course, all of the parenthetical statements about relative risk get the caveat, “Other things being equal.”)

* Whether or not the country has ever had a democratic regime (a transition is more likely if so)

* The age of the current authoritarian regime (the effect depends on prior democracy; risk increases over time for countries without democratic experience, but it’s more or less constant over time for countries with democratic experience)

* The scope of civil liberties the previous year, per Freedom House’s index (a transition is more likely with more liberties, but the association is non-linear)

* The share of the country’s gross national income generated from oil & gas extraction, per the World Bank’s World Development Indicators (a transition is less likely with more oil)

* Whether or not the country had a nonviolent popular movement the previous year, according to Erica Chenoweth‘s NAVCO data set (more likely if so)

* The annual rate of economic growth, according to the IMF’s September 2011 World Economic Outlook (more likely when growth is slower)

That list isn’t short because I lacked ideas about what else might help predict these events. Many other factors were also considered but were found to be poor predictors, while still others were left out of the analysis because there simply wasn’t enough data for enough countries or years (sometimes historical, sometimes current) to use them. Some of the factors I included in the call to BMA but found to be poor predictors of democratic transitions include:

* Per capita income (ditto for infant mortality rates)

* Literacy rates (normalized to the annual global median)

* Percent of the population in urban areas

* Youth bulge

* Trade openness (imports plus exports as a share of GDP)

* Mobile phone subscriptions per capita (normalized to the annual global mean)

* Ethnic or cultural diversity

* Ongoing civil war

I’d love to have readers use the Comments field to offer their own views on prospects for democratic transitions in 2012 and the factors and forces that will drive those events. Meanwhile, I’ll close with a wish: I hope I’m wrong and every one of these countries gets a democratic government very soon.

Leave a comment

17 Comments

  1. Courtney Carter

     /  December 18, 2011

    I’m glad to see an update on your work in this area! I’m in the process of writing my master’s thesis on a similar topic (transitions away from autocracy); it’s been heavily influenced by the PITF’s work, as well as your paper on contentious collective action and Jack Goldstone’s recent writing on sultanistic regimes. In your current research, have you looked into whether results vary based on the type of authoritarian regime (personal/military/single-party/hybrid) that is in place prior to the transition?

    Generally speaking, the factors that you found to be significant seem to be systemic properties of a state that would predispose it toward democracy, whereas several of the non-significant variables (mobile phone subscriptions, youth bulge) may act as catalysts for regime change but would not reflect the underlying causes or eventual outcome of the transition.

    I’m very curious to see how much of a role economic complexity plays in the demise of autocratic regimes. It seems as though it might be a stronger predictor than GDP growth, so I’m eagerly awaiting the release of Hausmann & Hidalgo’s ECI data.

    Reply
    • Thanks for reading and commenting, Courtney. I did not test whether distinctions between types of autocracies help predict transitions because the transitions data I’m using don’t include a measure like that. I know there are a few other data sets that do, but I’m not really satisfied any of them are reliably capturing those distinctions, especially over time within countries. As it happens, I’m hoping to find funding for some work to develop a fuzzy-set solution to that problem, but it hasn’t happened yet. Oh, and ditto on the economic complexity data; that would be really interesting to test here.

      Reply
  2. Peter Dörrie

     /  December 19, 2011

    I am very excited to see Burkina on top of the list (by a huge margin). I will spend my next year there and while I know that these are statistics and not 100% forecasts, I am really looking forward to an exciting (and hopefully peaceful) time there!

    Generally I think you experiment is very interesting and you should definitely revisit it in a year, to see what really happened. The results coincide with my gut feeling (for what it’s worth) that Sub-Saharan Africa is the likeliest place to see substantial political reform over the next years.

    Looking at your data, I would like to know if mobile phone ownership was excluded because there was not enough data, or because it was found to be a lousy predictor. The same goes for literacy rate, youth bulge and population in urban areas, all of which I would have thought would contribute towards a “democratising environment”.

    That income per capita and trade openness didn’t make the list in turn validates my personal guess that not so much “national wealth” matters (as liberals would argue), but distribution of wealth, when it comes to political changes. Speaking of which, why did you not test/include the Gini coefficient or a similar metric into the experiment?

    Reply
    • Thanks, Peter. Yes, it would be great (and proper) to revisit the list in a year to see how the forecasts fared. Hopefully, I’ll have a chance to do that.

      All of the variables I listed at the bottom were ones I included in initial iterations of the BMA exercise that got posterior probabilities of approximately 0. In other words, I tried them and the results suggested they weren’t especially useful predictors. I did not include an inequality measure because even the best cross-national time-series data are full of large, non-random holes and otherwise pretty noisy.

      Good luck with your work in Burkina Faso, and please keep us posted on political developments there when you can.

      Reply
  3. Nick Jorgensen

     /  December 20, 2011

    Interesting project — I’m glad to see someone applying Bayesian modeling to these kinds of forecasting exercises. I’m curious about the inclusion of the Philippines on this list — while I realize that the past decade has been disastrous there from a human rights standpoint (as a geocoded data set of extrajudicial killings I’ve assembled bears out), and the 2004 election was pretty fishy, but the 2010 election was reasonably clean, there are multiple opposition parties both inside and outside the legislature (and yes, I know, Filipino political parties are not especially credible, but they do exist), and there is no official press censorship (that task has been outsourced to private assassins, er, “contractors”).

    Reply
  4. Richard Bridger

     /  January 24, 2012

    Would it be worth including a variable for whether an election is due in 2012, or perhaps more broadly for whether a transition is likely / approaching? It’s obviously not a cause of a democratic transition, but I would imagine that in countries where the fundamentals appear to be aligning, it could be the stimulus for it crossing the threshold. For example, Armenia has parliamentary elections this year and presidential ones next year – so the chances of the conditions being satisfied will probably increase there simply by virtue of the fact the elections are happening. (Plus, though I’m relying on my impression rather than a source for this claim, it does seem that elections tend to act as a stimulus for transitions in countries that are otherwise well suited to them.)

    Reply
    • Yes, it certainly would, for all the reasons you mention, and I wish there were a simple way to do that. Surprisingly, though, there isn’t an off-the-shelf cross-national time-series data set that includes an indicator for election years. The closest and best shot comes from the Institutions and Elections Project (IAEP), but it’s not routinely updated, so it can’t be used for forecasting. The absence of a good measure for such an important variable is a good argument for using a simpler checklist guided by the results of the statistical analysis. I’ve been a skeptic of evidence-based checklists for a long time, but Daniel Kahneman lays out some solid evidence in his new book Thinking, Fast and Slow that they actually work pretty well.

      Reply
  5. Kimberly Curtis

     /  January 29, 2012

    I realize there is probably no way to data plot this, but the possibility for regional domino effect might be an interesting thing to keep in mind as well. For example, Mali and Mauritania are ranked higher by a few spots than Senegal, but if a transition were to happen in one it would likely impact the others due to geographic closeness.

    Reply
    • You know, when I’ve included a measure like that in models in the past, I’ve found that it didn’t improve the model’s predictive power at all. Other studies have shown a significant contagion effect (e.g. Brinks and Coppedge), so I won’t claim my results are definitive, but that’s why you don’t see anything in the model about recent transitions in nearby states.

      Reply
  6. Hmm, I guess your model didn’t work for Libya seeing as it meets your 4 requirements.
    Freedom of speech/assembly
    Free and fair elections
    All elected officials
    And elections were inclusive

    Reply
  1. For Political Forecasting, Deep Knowledge of Specific Cases Is Overrated « Dart-Throwing Chimp
  2. A Step Back for a Closer Look at the Philippines and Development | In Asia
  3. On the what and how of my freelancing plans | peter dörrie
  4. Джей Ульфельдер: почему одни авторитарные режимы выживают, а другие нет | Толкователь
  5. The Libyan Surprise « Dart-Throwing Chimp
  6. New Leaders and Political Change in Authoritarian Regimes (with an Eye on Ethiopia) « Dart-Throwing Chimp

Leave a Comment

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 7,816 other followers

%d bloggers like this: