Anti-government protesters shelter from rain under a makeshift tent made from plastic sheets, on the premises of Parliament House during the Revolution March in Islamabad, Sept. 4, 2014. Pakistan’s protest leaders demanding Prime Minister Nawaz Sharif’s resignation prepared to resume talks with the government on Wednesday, reviving hopes for a negotiated solution to a crisis that has shaken the coup-prone nation. (Akhtar Soomro/Reuters)
Kim Yi Dionne: As part of our continuing collaboration with political science journals, the following is a guest post from political scientist Nam Kyu Kim of the University of Nebraska-Lincoln based on the forthcoming publication of his article, “Revisiting Economic Shocks and Coups,” in the Journal of Conflict Resolution. The article will be available for free (ungated) until Oct. 31.
****
Last week, Naunihal Singh wrote an interesting post following the coup in Lesotho in which he argued that popular opinion has no impact on coups. To provide (indirect) supporting evidence for his argument, he asserted:
For example, it is reasonable to assume that citizens would be happier with a government during periods of high economic growth and unhappier with it during periods of low growth or even decline. However, it turns out that there is no relationship between economic growth rates and the likelihood of a coup.
Is there really no relationship between economic growth and risk of a coup? My research paints a different, and more nuanced, picture.
First of all, existing research is mixed on the effect of economic growth on coups. Some scholars find GDP per capita growth rate can reduce the likelihood of coups, while others including Singh do not (e.g., see the recent two studies: here for ‘no’and here for ‘yes’). Errors in measuring GDP, emphasized by several recent studies (here and here), may be why these studies have different findings.
For instance, I analyze the relationship between annual real GDP per capita growth rates and a coup attempt using the following widely-used data sets on real GDP per capita: Gleditsch, Maddison, the Penn World Table (PWT) 7.0, and the World Development Indicators. For coups, I use Jonathan Powell and Clayton Thine’s dataset. The figures below show how sensitive results are to which measure of GDP per capita is used. I find negative relationships in all models, but the strength and precision of the estimated relationships vary according to measures of GDP per capita. The variation in results are due to weak correlations between these four measures of GDP per capita growth rates, ranging from .49 to .78 – perhaps unsurprising given how hard it is to measure economic growth in poor countries, where most coup attempts have occurred.
The plots show the impact of real GDP per capita growth rates on coup attempts. Higher levels of economic growth seems to lead to a lower likelihood of a coup attempt, but the precision and strength of the estimate of the relationship between the two variables depends on which of four measures of real GDP per capita growth is used.
Second, different economic shocks could have different effects on coups. Earlier research lumped all kinds of economic shocks together. Both permanent shocks and transitory shocks drive fluctuations in aggregate income. Negative permanent shocks to the economy may not only increase the probability of a successful coup but also decrease the prize of the coup. If negative shocks are permanent enough to diminish the expected value of what one can gain from being the new ruler, the probability and the rewards of seizing power become opposing forces. On the other hand, negative transitory shocks do not affect the expected value of what a new ruler can gain from seizing power, and yet negative shocks will increase the probability of a successful coup. The total effect of economic growth on coups then depends on which effect dominates. Though both types of shocks open the window for a coup attempt, potential coup-makers might be dissuaded by permanent economic shocks because they reduce the spoils of office.
Accordingly, in my research, I differentiate between the effects of persistent and transitory income shocks on the likelihood of a coup. I exploit year-to-year fluctuations in rainfall and temperature to isolate transitory shocks to aggregate income (that is, I instrument for GDP growth rates using weather shocks). Research (here and here) finds that low rainfall and high temperature are associated with low income growth. To capture persistent shocks to the economy, on the other hand, I use oil price fluctuations. The response of aggregate income to oil price fluctuations is found to be very persistent. An increase in international oil price raises GDP growth in oil exporting countries. If my conjecture is right, growth in income, driven by rainfall and temperature shocks, is expected to decrease coup risk, while oil price – driven income shocks should have a much smaller impact on coups.
Three main findings stand out. First, GDP per capita growth rates are negatively associated with the likelihood of a coup attempt when using weather shocks to isolate exogenous variation in income, regardless of which GDP measures are used. Second, the magnitude of the relationship is larger here than when I directly look at the relationship between GDP per capita growth and coups. For instance, a one-standard-deviation drop (6 to 8 percentage points) in real GDP per capita growth rate increases the probability of a coup attempt by approximately 9 to 11 percentage points. According to the previous estimates reported in the first figure, contrarily, the same amount of decline in GDP growth increases the likelihood of a coup attempt by merely .08 to .6 percentage points. Last, oil price–driven income shocks have little impact on coups. These results show the importance of distinguishing transitory and persistent shocks as well as that of addressing measurement errors in GDP data.
My research not only poses a challenge to Singh’s argument about the relationship between economic growth and coups but also has implications for his main argument that public opinion has no role in encouraging coup conspirators to act. I find that part of the effect through which economic growth influences coup attempts is channeled through popular protests. The relationship between negative economic growth and protests is also well-established in the literature. In addition, a recent study argues that popular protests provide a public signal, which helps elites contemplating a coup to coordinate, which can help to account for the link from low economic growth rates to coups (actually this emphasis of coordination among elites is consistent with the main argument of Singh’s book). Widespread popular protests can serve as a proxy for public opinion, particularly in autocratic countries. Taken together, there is a reason to believe that military factions are willing to attempt coups when popular opinion, publicly known, is in their favor, even though, as Singh contends, “there is no reason to believe that military factions hesitate to attempt coups when popular opinion is against them.”
****
Nam Kyu Kim is an assistant professor of political science at the University of Nebraska-Lincoln. His research covers the politics of authoritarian regimes, international and internal conflict, human rights, and comparative political institutions.