WikiSummary, the Social Science Summary Database

Ferree: The social origins of electoral volatility in Africa

From WikiSummary, the Free Social Science Summary Database


Ferree. 2005. The social origins of electoral volatility in Africa. unpublished.

In Brief

VOLATILITY (Y) is the number of seats that parties gain and lose each election. When volatility is too high, politics gets messy (short time horizons, thus bargaining breakdowns and coalition failures); when volatility is too low, incumbents get away with too much. But we know little about volatility, except that it is more common in younger democracies. But WHY DOES VOLATILITY VARY even across young democracies?

There are two likely answers: Institutional arrangements and social cleavages. We know a decent amount about institutions (electoral rules, presidentialism, federalism) but very little about the effects of social cleavages (except for Lipset and Rokkan 1967 and a few other European anlayses).

This paper focuses on SOCIAL CLEAVAGES (X), particularly ethnic cleavages (in Africa), as an explanation.



  • When there is no ethnic group that can form a winning coalition on its own, there will be high volatility.
  • When there is a single ethnic group that can form a winning coalition on its own, there will be low volatility.
  • But when there are multiple possible WCs, there will be high volatility.
  • There can be multiple possible WCs because ethnic groups are nested (see Figures 2 and 3). In Benin, for example, 70% of the population are southerners (that's one possible WC) and 55.5% of the population are Fon (that's a second possible WC). (All Fon are Southerners.) Because there are two possible WCs, there will be volatility: Sometimes a Southern alliance might win, other times the Fon will defect.


Assumes that there are, in fact, incentives for aggregation (e.g. electoral incentives).


  • Measures average volatility for each country (cross-sectionally)
  • Counts the number of possible WC's in each country (number of (nested) ethnic groups larger than 50%).
  • Several controls (pg 20-21).
  • Good results (Table 3), though I'm curious why she doesn't include both dummies in the same model (i.e. a dummy for no WCs, a dummy for single WC, multiple WCs as the baseline).


  • She does not adequately defend focusing solely on ethnic cleavages. After all, what about class and religion? These might be nested with ethnicity (e.g. some groups are Christian, others Muslim), but they might not. Presumably that would matter. And Horowitz (or was it Gurr?) noted that class often overlaps with ethnicity. This is merely an empirical critique, though, not a theoretical one.
  • With the count of possible WC's, I presume that she uses 50% as a strict cutoff. This is minor, but perhaps she could use some weighted meausre (along the lines of ENPP) rather than raw count. After all a group with 45% or is pretty darn close, which can matter (depending on the institutions).