However, it is not as simple and straight as it sounds. High voter turnout could also be a reflection of a strong pro-incumbency sentiment; when voters feel strongly in favour of the incumbent government, they like to come out in large numbers and express their support.
In a multi-party, coalition-oriented, post-poll & pre-poll alliances, representative politics, West-minister-style democracy such as India, the definition of anti-incumbency is definitely unclear.
For example, if the member of Parliament (MP) of a constituency sat in the opposition benches, and people prefer voting him/ her, is it being pro- or anti-incumbency? How do we define our position with respect to incumbency if a party MP changes sides in the national alliance?
Even simpler, what if our MP decides to cross the floor and seek re-election on the ticket of the opposing party?
As you can see, it is all uncertain and grey. As for the question of the local MP crossing the floor, we will measure incumbency at the party level, and assume that a vote for the erstwhile ruling party is a pro-incumbency vote,
notwithstanding the fact that it is a vote against the incumbent MP. In other words, we are assuming that in a parliamentary election, people vote for the party and not for the candidate (we will resolve that question in a subsequent edition of Election Metrics).
Finally, in case a party splits, the breakaway faction is assumed to be a new party and any votes for the breakaway faction will be assumed to be an anti-incumbency vote.
During 1999 and 2004 we have seen how the alliances shifted sides and one is inconclusive till date whether it was mere mathematics of anti-incumbency!
In the 1999 general elections, about 53% of the constituencies returned the incumbent party. In 2004, about 48% of the constituencies returned the incumbent party.
The overall nationwide turnout in the 1998 elections was about 61%, which reduced to about 59% for the 1999 elections and 58% in 2004, based on the observations. Taking together, the two elections under consideration (1999 and 2004),
there are a total of 559 constituencies where the incumbent MP at the time of elections belonged to the ruling party (this number may not be particularly accurate since we ignore independent MPs.
The difference in voter turnouts between the 1999 and 2004 elections is independent of whether the incumbent was returned to power. Based on this, it is hard to conclude whether turnout actually has an impact on incumbency.
While it is hard to establish a relationship between turnout and anti-incumbency at the national level, could a relationship exist in particular states (with differing effects cancelling out at the national level)?
First, the difference in turnouts as a function of pro- or anti-incumbency vary significantly across states. While West Bengal sees a turnout of 7.5 percentage points higher when the incumbent is being voted out,
Punjab sees a turnout of 5 percentage points higher when the incumbent is voted back!
Secondly, the statistical significance column is full of big numbers—the two above mentioned states and Delhi (where the turnout for an anti-incumbency vote is 8 percentage points higher than for a pro-incumbency vote)
are the only ones where there is a statistically significant relationship between turnout and incumbency.
To summarize, can we conclude that higher turnouts are associated with anti-incumbency? Based on the evidence above, the answer is a resounding NO. While there is some data that supports the conventional wisdom that higher turnouts are linked to anti-incumbency,
the data we have seen is not sufficient to conclusively point evidence in that direction.
There is scope for further work on this though. We have only looked at two pairs of national level elections to test the hypothesis. The story may be different if one were to consider state elections also.
