Sunday, January 15, 2012

When do states cut subsidies?

AP is reporting that the negotiations between Nigerian labor unions and the Jonathan government have broken down and that protests against the government's fuel subsidy cuts are likely to continue.

Independent of any potential economic benefits, it is clear that the cuts are politically costly. And, as the first rule of political economy dictates, politics trumps economics. Indeed, its remarkable that someone like Paul Collier so quickly dismisses the Nigerian protests as "poor people tricked into lobbying for greedy elites" and equates protesters with the Tea Party while briefly skimming over the fact that food and transportation costs for ordinary Nigerians have doubled as a result of the shock. I cannot understand how the reasonable response to seeing one's cost of living skyrocket in a span of a few days as a result of government policy is to just accept it, move on and not protest (but again, I'm not an economist). Yes, the subsidy was distortionary, unsustainable, and a means for fueling corruption, but removing it in one sudden swoop, without meaningful consultation and without a real plan for ameliorating the massive adjustment costs will not eliminate corruption and will only serve to reinforce the sentiment that Nigeria's nascent democracy is not responsive to its people. All of the government's rhetoric about how the subsidy affects the rich more than it does the poor is a small consolation (and blindly ignores diminishing marginal returns to income). 

Given the ongoing political fallout, the real question is why did the Nigerian government choose to enact such a drastic policy if it (likely) knew that the cuts would generate such a heavy public backlash? Colleague Daniel Solomon recently hinted at national-regional elite struggles as one rationale. Indeed, given Nigeria's history of center-periphery tensions and the recent attacks by Boko Haram, the Nigerian government may have thought of subsidy removal as a way to curb the revenue streams of regional elites and strengthen the authority of the center. But even here, the policy could backfire as these same regional elites can take advantage of public anti-government anger to increase their personal legitimacy. This appears to be taking place. Local governors - particularly those in the north - are already negotiating for a more advantageous subsidy regime that would shift more of the burden onto the more urbanized southern states. Moreover, these leaders are also seeking to boost their own public support by appearing responsive to public concerns and trying cushion the blow of higher transportation costs. Its unclear in this case whether the benefits to the center outweigh the public opinion costs (especially given the ongoing strike). 

This provokes another interesting question. What makes a government more or less likely to cut subsidies? The logic above suggests that the level of democracy plays an important role. Leaders that depend on public support for survival will be more heavily deterred from slashing transfer programs than those that only need to satisfy a limited group of elites.

My Google Scholar searches could not find any previous papers that dealt systematically with the domestic politics of subsidy cuts, so I put together a quick test. Using the IMF's Government Finance Statistics dataset (available at the ICPSR) I calculated the year-to-year changes in government spending on subsidies and other transfers for each country. I added a dummy variable that took on the value of 1 when the changes were negative (representing cuts) and 0 when they were positive or nil. I then merged the IMF data (stupid IMF country codes) with the Quality of Government time-series dataset and ran a series of logit regressions with the cuts variable as the response and Polity IV democracy score (higher = more democratic), International Country Risk Guide quality of governance score (higher = less corrupt), and real per-capita GDP as the independent variables. The results are below (coefficients are in terms of the log-odds):
Variable Coefficient Std. Error Z-score p > abs(z)
Democracy -0.1664171 0.0327781 -5.08 0
Quality of Governance -0.9525243 0.7546757 -1.26 0.207
Real GDP per Capita 0.0000372 0.000021 1.77 0.077
Of the three variables, democracy is the only one significant at the 95% level (though RGDP per Capita is significant at 90%). The negative effect is as expected - higher levels of democracy reduce the odds of a subsidy cut in any given year. Replacing democracy with the winning coalition measures in Bueno de Mesquita et. al. (2003) gives similar results, suggesting that the effect may be due to democratic leaders having to appeal to the population at large as opposed to regime insiders in order to survive.

One avenue for further research is to ask is whether corruption has an effect independent of democracy. The results weakly suggest that lower levels of corruption decrease the odds of subsidy cuts but the effect is not significant. However, the two measures are highly correlated. Including quality of governance but not democracy in the regression returns a negative and significant coefficient. The effect also makes sense - more corrupt governments may prefer to funnel funds into private coffers as opposed to spending them on large-scale public transfer payments. In the case of Nigeria, extreme levels of corruption (it's ranked 143rd on Transparency International's Corruption Perceptions Index) make it difficult for the Jonathan government to credibly assure Nigerian citizens that savings from the subsidy will not merely be distributed among elites. High levels of corruption may therefore be a barrier to subsidy removal. Amegashie (2006) suggests that:
If economic agents do not trust their governments and believe their most of their taxes are used for the private gain of politicians, then they might agitate for subsidies  on certain commodities as a way of getting a piece of the national pie. Subsidies, no matter how inefficient, may then be used by the electorate as a way of getting politicians to commit to some form of credible redistribution.  
Although a deterrence effect is not apparent in my tests, it may be that subsidy cuts, ceteris paribus, spark greater public protests in countries with higher levels of corruption. Unfortunately, I don't know any other good time-series measure of domestic unrest other than the Banks dataset (which Georgetown, to my knowledge, does not have), so I was unable to do a quick regression with STATA.

Edit: 6:34 PM 1/15/2012

1 comment:

  1. I think this is a very good and under-researched question and your first-take is cool. Using IMF data may not be ideal, but certainly adequate. You should publish this. Also, this is a really nice blog. Keep up the good work!