In the aftermath of this weekend’s Japanese elections, we’re once again hearing a lot of discussion about sweeping elections or even political earthquakes. When political scientists want to think about changes in election results across successive elections, we turn to measures of electoral volatility, or, more specifically in most cases, the Pedersen Index of Electoral Volatility.
The Pedersen Index is a fairly neat measure that calculates the change in each party’s results from the first election to the second election, and then summarizes this volatility across all of the parties (see here for more details.) Given its (general) ease of calculation and its fairly close approximation of the underlying concept, it is no wonder that the measure has become fairly common in the comparative politics literature.
The measure, however, suffers from a fairly serious flaw, especially when we move to newly competitive party systems. Namely, the measure does nothing to distinguish between volatility caused when voters shift between two existing parties and volatility caused by new party entry and old party exit. Put another way, both the 2009 Japanese elections and the 2001 Polish elections will generate high volatility scores; however, in the Polish case, over 40% of the vote went to essentially new parties. This in turn has important implications for both policy – e.g., foreign investors might have a very different take on making long term investments in countries where votes swing between established parties and where new parties are constantly entering the political scene – and theory, i.e., we might expect different variables to explain swings in votes between established parties as opposed to decisions regarding new party entry.
With this in mind, Eleanor Powell and I have a new paper that develops rules for calculating these two separate types of volatility – which we have creatively labeled Type A and Type B volatility – and then compares the two volatility measures across 80 elections from 21 post-communist countries. (For anyone who is interested, we’ll be presenting the paper at APSA this Friday, September 4th, at 2:00 PM on Panel 11-31.)
Some of the initial findings are quite interesting.
For example, here’s a graph of the total volatility (as calculated by the standard Pedersen Index) by election over time:
We can clearly see a gentle downward slope, leading us to conclude that volatility is falling in the post-communist world, much as we might expect in consolidating party systems. However, here’s a graph of just the Type B Volatility (i.e., the volatility caused by voters switching between parties that contested both elections):
Now our downward slope has disappeared. So in terms of volatility across existing political parties, there does not seem to be much stability developing over time. So where did the decrease in volatility come from in the overall Pedersen Index? Here’s the Type A volatility (i.e., the volatility caused by new party entry and old party exit):
So what is apparently going on is that there is a decline in the volatility caused by new party entry and old party exit. This is also an encouraging sign for the development of stability in new party systems, but it is one we would not have been able to identify using the Pedersen Index alone. For more, see the paper.
And bonus points for anyone who can come up with a better name for the two types of volatility measures…
[Hat tip to Mik Laver, whose comments on another paper during a reading group were the impetus for this research.]
{ 1 comment }
Very interesting stuff. I’ve done some work on volatility in Israel, and would like to contest your choice of handling splits and mergers. Primarily, I’m not clear on the under 5% rule for mergers – why would two parties, one with 6% and one with 4% of the vote who merge and get 10% of the vote, count as 2% type A volatility and 2% type B? It seems to me that we should have 0 volatility in this scenario, regardless of the size of the smaller party.
I also came across similar difficulties with coding Israeli parties, and found that treating mergers as if the two parties were already merged in the first elections to be more sensible.
Also, the Israeli case gives even more confusing situations. For example, in 2003 there was a party, Mafdal, and a Party-bloc (that ran under one ticket) of Ikhud-Leumi/Israel-Beitenu. In 2006, the party bloc split into its two constituent parties, but one of them, the Ikhud Leumi, formed a new bloc with the Mafdal. How do you calculate volatility for this mess? I simply treated all three as a single bloc for the two elections, so any vote transfers within the three were not counted towards volatility. That seems most reasonable when we think of what volatility really means: people shifting political allegiances from one party to another. Your system creates “virtual shifts”, where people who didn’t change their preferred party at all are counted as part of electoral volatility.
Comments on this entry are closed.