Indian Politics & Policy Volume 3, Number 1, Spring 2020 | Page 21

Economic Evaluations and the Incumbent Vote in India’s Parliamentary Elections (2014, 2019)
rural voters are given a value of 1, while
non-rural voters are given a value of 0.
Economic class is captured with four
variables: poor, lower, middle, and rich
(created from an index using household
assets, agricultural land holdings,
and type of housing [pucca, mixed, or
kuccha]).
I pooled the data from the two
election years, which results in an independent
pooled cross-section design
consisting of individuals independently
sampled at different points in time.
There are several advantages to pooling
data. 17 In addition to an increase in
sample size, this empirical strategy allows
me to also examine the effects of
positive and negative evaluations on the
incumbent vote over time.
I present the results of the statistical
estimation as figures. 18 In these figures,
the dots represent the coefficients
or point estimates, and the horizontal
line passing through the dots indicate
95 percent confidence intervals computed
using robust standard errors. A
coefficient is considered statistically
different from zero, i.e. statistically significant,
if its confidence interval does
not intersect the vertical dashed line
(representing 0).
Descriptive statistics for all variables
used in this study is presented in
Table A1. Results of the statistical estimation
are presented in Tables A2-A3.
Statistical Results
In this section, I present the results
of three models. Models 1 and 2
pool the data from both years and
estimate the marginal effects of positive
and negative evaluations of household
economic conditions. In these two
models, in addition to the key variable
of interest—economic evaluations—I
include two other control variables, incumbent
party attachment and a time
dummy variable (Year) that takes a value
of 1 if the election year is 2019 and 0
if 2014. In Model 2, I also include a set
of interaction terms. I do not include
standard socio-demographic controls
in these models. Socioeconomic and
other demographic factors clearly influence
party or candidate choice, but
there is no reason to expect them to
influence the incumbent vote. That is,
there is no reason to expect that Hindu
Dalits systematically choose incumbents
while Christian Adivasis similarly
systematically vote against the incumbent.
Clearly context-specific factors
matter when explaining the outcomes
of specific elections (say, the vote share
received by a particular party or coalition
in a particular election), but are
much less useful for explaining variation
in an incumbency vote, where the
identity of the incumbent party is likely
to change from one election to the
next. Model 3 estimates the same relationship
focusing only on the BJP vote
in 2019. Here I include additional control
variables such as caste-community
identity, economic class, gender, and
rural that are shown to influence the
BJP vote. I interpret these results in the
specific context of the 2019 elections to
the Lok Sabha.
Figure 1A presents the results of
the first model (Model 1) and Figure
1B presents the marginal effects of the
17