A Panel Data Analysis of General Elections Under Multi Party Electoral College

Abstract

We analyze the probability of an incumbents winning in the consecutive election, under the assumption that all individual candidates are equally likely (i.e. random selection) when that they are from the same party. We are trying to analyze the probability of winning by ruling party (i.e. the party which has won the last election from that constituency), irrespective of whether that party forms a coalition during election, formed the last government in center/state or was the part of the last central/state government in case of coalition government. In particular, we model the voters behavior by Binary Choice Model (Logit Model) using panel data of India's general (parliamentary) elections from 1967 through 1999, without taking into account the effect of the central government's ruling, while we are emphasizing on the effect of constituency wise ruling and the results of that particular constituency. We find evidence that incumbents have better log-odds of getting re-elected if they contest on behalf of the national parties, from reserved seats, from constituencies with large valid votes polled with respect to electorate, if their ruling duration is more. We don't find any significant evidence of getting re-elected if the sex of the competing candidates is same or if the number of candidates contesting election from the constituency is higher.

Keywords: electoral system, GEE population-averaged model, panel data and incumbents

JEL Classification: C25, N45