Enough about Inequality and Unrest Already!

Can we please, PLEASE stop it with the assertions that a country’s income inequality tells us a lot about its propensity for social unrest?

This claim pops up all the time. Exhibit A from an article I read this morning, on China’s official Gini coefficient for 2012:

China’s reality of inequality – and the challenge to narrowing the gap – remains unchanged: the current coefficient of 0.474 poses a high risk for social unrest.

I understand, and am even sympathetic to, the claim that gross disparities in wealth are unjust, particularly in societies where the poorest want for basic needs like food and shelter. I also recognize that organizers of, and participants in, contemporary social unrest often call out economic inequality as one of their chief grievances.

What I’m just not seeing, though, is empirical evidence that countries with higher economic inequality are more susceptible to social unrest.

For starters, there’s the general observation that economic inequality is common and persistent, but large-scale social unrest is uncommon and usually fleeting. What Jim Fearon and David Laitin wrote in 1996 about inter-group tensions and ethnic violence applies just as well here:

Among existing theories of ethnic conflict, accounts focusing on past tensions between groups that are memorialized in narratives of blame and threat tend to dramatic overprediction of violence. Such narratives are almost always present, but large-scale interethnic violence is extremely rare.

The same goes for inequality and popular rebellion. The former is ubiquitous while the latter is scarce, so it’s hard to see how the presence of the one can be said to predict the risk of the other.

Okay, so maybe inequality doesn’t help explain the timing of social unrest, but it does predispose certain societies to erupt when other forces come together. It’s not the spark that starts the fire; it’s the dry tinder that helps the spark catch and spread.

Well, I’m just not seeing this, either.

To look at the association between inequality and unrest, I started by downloading the World Bank’s data on income inequality from Hans Rosling’s Gapminder site. These data summarize occasional national surveys on income or consumption in a Gini coefficient. The higher the Gini coefficient, the more unequal the distribution of incomes in that society. Because the data are only updated occasionally—many countries have just one or two reported values since 1979, the start of the World Bank’s observation period for this measure—I reduced the time series into a single value by taking the maximum (or, in some cases, lone) value for each country. Then I used Erica Chenoweth and Maria Stephan’s data on nonviolent uprisings to identify which countries had seen at least one civil-resistance campaign emerge between 1980 and 2006. Finally, I used the ‘sm‘ package in R to produce kernel density plots that visually compare the distribution of Gini coefficients across those two sets of countries.

The results are shown below. As you can see, there seems to be virtually no difference in the level of income inequality among countries that have and have not produced popular uprisings since 1980. In a bivariate logistic regression model estimated from these same data, the coefficient for the Gini index is <0.01. Not exactly the powerful discriminator we keep hearing about, eh?

inequality_and_uprisings

That chart only looks at nonviolent uprisings, but published research on violent conflict suggests that the association isn’t especially strong there, either. In a 2008 paper (h/t Cyrus Samii), political scientist Gudrun Østby finds only a weak link between income inequality among individuals and the risk of civil-war onset in 36 developing counties. Interestingly, she does find evidence that higher levels of inequality between ethnic groups increase the risk of violent rebellion, suggesting that inter-group comparisons play a role in fomenting conflict. Still, this isn’t the rich vs. poor narrative on which the conventional wisdom about inequality and rebellion depends, and on that score, Østby’s analysis only strengthens my prior belief.

In light of that empirical evidence, it’s hard to put much stock in the oft-heard claim that highly unequal countries are especially prone to social unrest. Given how noisy the data on income inequality are, it seems particularly absurd to treat small fluctuations in a single country’s Gini coefficient as a useful indicator of rising or falling prospects for a popular uprising or civil war. I don’t think this blog post is going to do much damage to the conventional wisdom, but if there are any takers out there, I would be happy to bet against anyone who wants to use Gini coefficients to predict where the next rebellion will occur.

Update: In the Comments, Rex Brynen suggested I also compare the distributions of Gini coefficients in a couple of subsets where inequality would arguably have a stronger effect: poorer countries, and poorer countries with no history of democracy before 1980 (the start of my period of observation). The plots below do that, where “poorer” is defined broadly as countries that weren’t OECD members as of 1980. As you can see, there’s still virtually no separation in the broader non-OECD subset (the plot on the left). When we limit our view to non-OECD countries with no democratic experience before 1980 (the plot on the right), we get a little bit of separation in the expected direction, but the difference is still rather marginal. (In a bivariate logistic regression estimated from this subset, the coefficient is 0.03 with an s.e. of 0.04.)

inequality_and_uprisings_subsets

Leave a comment

25 Comments

  1. Neal Caren and some of his students had a paper at ASA this past year (now looks like its under review at ASR) on the relationship on economic growth and the relationship to protest events. They found a pretty consistent relationship between the two. I’d imagine the World Bank data on change in GDP is a little more stable than Gini coefficients.

    Reply
  2. this guy (http://mortenjerven.com/) at least, thinks the wb data is actually quite terrible. there seem to be many instances where huge 1 year changes in gdp/gdppc are due to changes in national statistical office methodology (esp. wrt measurement of the informal economy). he was on econtalk not long ago (http://www.econtalk.org/archives/2013/01/jerven_on_measu.html)

    Reply
  3. Great post. The bald assertions bother me too!

    Instead of looking at vertical inequalities, people should look at horizontal inequalities, which Frances Stewart has emphasized in her work. I discuss the differences here:

    http://www.fragilestates.org/2012/03/12/horizontal-versus-vertical-social-cohesion-why-the-differences-matter/

    Reply
  4. Beef

     /  February 3, 2013

    You might also find this interesting and relevant. I saw this presented at APSA a few years ago. http://link.springer.com/article/10.1007/s12116-012-9122-7

    Reply
    • Yes, thank you, and that’s a nice hook for me to clarify: I think it’s quite possible that there are contingent effects (like this one), but that’s clearly not the assumption on which these “high-inequality countries are going to explode!” stories are based.

      Reply
  5. Rex Brynen

     /  February 3, 2013

    Not to have you plotting curves all day, Jay, but happens to the curve if you control for 1) democracy, 2) consolidated democracy, 3) absolute income levels, and 4) consolidated democracy + high income? I’m just curious (plus if you plot them I’ll likely pinch the graphics for classroom use!)

    Reply
    • Are you looking for plots of marginal effects or kernel density plots for particular subsets defined by some combination of those elements? It sounds like the latter, but that raises the question of what to do with cases that experience regime changes during the period of observation. Maybe there are some subsets within which you’re especially interested in seeing comparisons—say, only poor countries, or only poor countries with no history of democracy before 1980? If so, let me know, and I’ll see if I can crank out a couple more plots soon.

      Reply
      • Rex Brynen

         /  February 3, 2013

        Similar Kernel density plots, but for those subsets. #1 does generate regime-change-during-observation challenges, and #2 would require some operational definition of “consolidated” (presumably, regular democratic elections for the previous X years).

      • For this exercise, how would you define “poor”? Non-OECD? Bottom two-thirds on income or IM? Bottom half?

      • Rex Brynen

         /  February 3, 2013

        Non-OECD or bottom-two thirds works well (although perhaps OECD membership introduces other implicit criteria.. on the other hand, it would have the advantage of not excluding high-income oil exporters).

  6. Rex Brynen

     /  February 3, 2013

    Thanks for that, Jay–I just inserted those graphs into one of my POLI 227 lectures on the politics of inequality (with appropriate credit, of course!)

    Reply
  7. I believe you would enjoy taking a look at this paper by Carles Boix. Page 207: “The central hypothesis of this essay is that political violence should erupt as income inequality and asset-specificity increase.”

    Reply
    • Thank you. As someone whose primary research interest is comparative democratization, of course I’m familiar Carles Boix’s work on this subject. I decided not to discuss it here, however, because the popular claim I am trying to assess is explicitly un-conditional. Additionally, I’m skeptical that available data on income inequality are sufficiently reliable and precise to test Boix’s argument about the effects of change over time within cases. The theory is more complex than the conventional wisdom, but in my opinion that complexity demands sharper comparisons than available data will allow.

      Reply
      • Thanks for the clarification. I was a bit surprised not to see Boix’s work cited in your post, and understand why now. I also agree that the data are not reliable enough for robust empirical testing (if I remember correctly, Boix uses percentages of small farms as a measure of income inequality — quite an indirect proxy).

  8. Oral Hazard

     /  February 4, 2013

    Last week PBS’s FRONTLINE aired “The Untouchables,” a look at the lack of criminal prosecutions of top Wall Street executives in the aftermath of the financial crash. I’ve suspected for a while that the nonviolent “Occupy” movement response to seeming social injustice and double prosecutorial standard in the US is a function of high per capital GDP and redistribution through social safety net programs.

    I’m wondering whether there is what we might call a biological imperative to revolt against the ruling class only at an appallingly low standard of living — civilization always being a few missed meals away from anarchy type of thing. So that would mean that there may be an objective standard of living above which humans will simply tolerate injustice and adapt because fundamental needs are not perceived as being in severe jeopardy.

    If you continue to move the national GDP (and hence per capital GDP) slider down below the bottom 2/3 non-OECD, do you find a more striking correlation? Or does sample size become too small to be meaningful?

    Reply
  9. Peter

     /  February 4, 2013

    That individual GINI does not have any effect on demonstrations, riots, uprisings or other types of rebellion, is not that surprising, I reckon. As Houle (2009) in World Politics argued, against the theories of Acemoglu and Robinson (2000;2006) and Boix (2003), they don’t consider collective action problems. If analyzing inequality defined as inter-individual inequality (GINI), I see no reason for why they shall overcome their collective action problem. Would it not be more interesting to study the conditional effects of inequality, lets say, when interacted with economic shocks or something else that could facilitate collective action? I reckon inequality in real life actually HAVE an effects – it just needs to be ignited. Thanks for a great blog!

    Reply
  10. Michael Ross

     /  February 4, 2013

    Great post – smart, judicious, clear and straightforward data. I’m sending this to my grad course.

    Reply
  11. Grant

     /  February 4, 2013

    There might come a point at which awful living standards combined with a powerful government actually inhibit rebellion, such as North Korea.

    What about the possibility of the type of regime (especially different types of authoritarian regimes) determines the chance of serious unrest in response to conditions?

    Reply
    • Yes, the empirical evidence I know indicates that regime type is a much better predictor of unrest than inequality. As for how influential and in conjunction with what, that seems to depend in part on what kind of unrest we’re talking about. For nonviolent uprisings, they actually seem to be more likely in countries that aren’t extremely poor but are instead more urbanized with larger and more literate populations, conditional on political institutions and opportunities. Violent rebellion, on the other hand, is more likely in poorer states, other things being equal. And in both cases, the likelihood of rebellion is more likely in countries with national regimes that aren’t solidly democratic or dictatorial. (It’s tempting to read a causal story from that latter fact about relative constraints and opportunities, but there’s so much endogeneity here that I try not to do that.)

      Reply
      • Gyre

         /  February 5, 2013

        I wonder if it’s more education levels than income that determine tactics. The two are obviously related but still distinct.

      • Yes, and this is what makes causal analysis of what Charles Tilly called “contentious collective action” so hard: so much is intertwined, and we don’t see the counterfactuals. Per capita income and education rates co-vary not only with each other but also with urbanization and industrialization, two processes that also shape peoples’ life chances and values and facilitate social mobilization. Meanwhile, long-term global trends in all of these things run parallel to changes in ideas, organizations, and institutions at the national and international levels that also plausibly shape patterns of popular action. Good luck sorting all of that out, right?

  1. Pt I – Understanding San Diego Minimum Wages | Understanding San Diego
  2. Some Thoughts on the Causes of Mass Protest | 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 7,295 other followers

%d bloggers like this: