Professional Documents
Culture Documents
Women Voters in Indian Democracy: A Silent Revolution: Mudit Kapoor and Shamika Ravi June 27, 2013
Women Voters in Indian Democracy: A Silent Revolution: Mudit Kapoor and Shamika Ravi June 27, 2013
Revolution
∗ †
Mudit Kapoor and Shamika Ravi
Abstract
We discover the trend of steady increase in women voter participation
in Indian elections over time and analyze whether it has any signicant
eect on political outcomes. We compare the turnout of women voters to
men in all state elections from 1962 till 2012. Our analysis reveals a sharp
decline in gender bias in voting over time, across all states, including the
traditionally backward states. This decline is solely driven by increase in
women participation while male turnout remains unchanged over time. We
also nd evidence that women voters are change agents, they vote dierently
from men and aect re-election prospects signicantly.
1 Introduction
The persistence of gender inequality which is embodied in missing women (a
income countries. In this paper, we study the gender bias in political participation
by analyzing women voter turnout in Indian democracy from 1962 till 2012 and
∗
mudit_kapoor@isb.edu
†
shamika_ravi@isb.edu
a steady and a sharp decline in the gender bias in voting over time. In partic-
ular, we nd that the sex ratio of voters (the number of women voters to every
1000 men voters) increased very impressively from 715 in the 1960s to 883 in the
2000s. (2) This phenomenon of declining gender bias in voting is across all the
Pradesh, Rajasthan and Uttar Pradesh. (3) This decline is solely driven by the
dramatic increase in women participation in the elections since the 1990s, while
men participation has remained unchanged. (4) We also nd evidence that women
voters are agents of change - they vote dierently from men and have a remarkable
The fact that more women are voluntarily exercising their constitutional right of
adult surage across all states in India is testimony to the rise of self-empowerment
extraordinary achievement in the world's largest democracy with 717 million voters
tion. Anderson and Ray (2010) provide a decomposition of missing women by age
and cause of death, and their ndings suggest that excess female mortality is a uni-
Duo (2012) provides an excellent overview of the literature by reviewing the re-
of any specic top down policy intervention to raise voter turnout of women, but
This paper also diers from the existing research on women and politics in a
policies that promote women leaders is that they may not be as powerful as the
male leaders. Women leaders might eectively be puppets in the hands of their
husbands or local elites. We do not have similar concerns when we study the role of
women as voters in politics. Women enjoy the right to equality in voting, and a vote
There is evidence to support that women and men have dierent preferences
and Duo (2004) show that the gender of the village council president impacts the
by studying whether there are dierences in voting decisions of men and women.
We nd strong evidence to support that men and women, indeed vote dierently.
we explore whether men and women vote dierently and thereby aect re-election
assembly elections in 2005. When the assembly election was held in Bihar, in
As a result of this, re-elections were held within eight months, in October 2005.
We compare outcomes of all 243 assembly constituencies in Bihar for both the
elections which were held within a short period of time. The results reveal that
In these constituencies, the winning political party of the February election was
not re-elected in October, and a new political party was declared the winner.
formed in Bihar. Our analysis shows that this change in the election outcomes
was fundamentally brought about by the women voters. Our study adds to an
For our analysis, we use constituency level data from the Election Commission
3
of India (ECI). We analyze all state assembly elections held from 1962 till 2012,
for 16 large states. These 16 states together represent more than 93 percent of the
total electors in India. This is a rich data set which has been used by earlier papers
that have explored women's political participation and crime (Iyer et al., 2012),
how voters respond to criminal charges of candidates (Dutta and Gupta, 2012)
2012).
The rest of the paper is organized in the following manner: we describe struc-
ture of Indian democracy in section 2 and the data in section 3. In the data section,
data. The results are analyzed in section 4 where we begin by describing the
consistent trends of reducing gender gap in voter turnout in all states. We then
explore whether men and women vote dierently by studying the 2005 re-elections
in the state of Bihar. We discuss these ndings in section 5 and conclude in section
6.
2 Indian Democracy
India is a federal union of states, it is a Sovereign Socialist Democratic Republic
terms of the Constitution of India that came into force on the 26th of January,
1950. The parliament consists of the President who is the executive head of the
union and the two houses known as the Council of States (Rajya Sabha) and
the House of the People (Lok Sabha). According to the constitution the real
executive power is vested in the Council of Ministers with Prime Minister as its
head to advise the President who shall discharge his functions in accordance to the
advise. The Council of Ministers are collectively responsible to the Lok Sabha.
above the age of eighteen. The President by notication will call upon the par-
liamentary constituencies to elect members of the Lok Sabha and thereafter the
electors of the parliamentary constituency will directly elect the members of the
Lok Sabha on the basis of adult surage. The constitution stipulates that normal
term of the Lok Sabha members is 5 years unless the President dissolves the House
4
of the people earlier. The rst general election in India was held in 1951 - 1952
and so far 17 elections to the House of the People have been held. At present there
India is a federal country with 28 states and 7 union territories (UTs) which
are centrally administered. Every state and the UTs has a legislative assembly and
Governor is the head of the state, however, in some states there is also an upper
house called the State Legislative Council. The executive power of the state is
vested in the Governor who is advised by the Council of Ministers with the Chief
the legislative assembly of the state. The election to the state legislative assembly
is held in the similar manner as the Lok Sabha with state and UTs divided into
single-member assembly constituencies. The normal term for the members of the
state legislative assembly is 5 years. The largest state legislative assembly is for
Uttar Pradesh with 403 members and the smallest is Pondicherry with 30 members.
The Constitution distributes legislative powers between the Parliament and the
State legislatures. For example, all matters relating to defense of India is subject
to the control of the Parliament while public order and police in the state are
3 Data
The Constitution of India has vested the Election Commission of India (ECI)
to oversee, direct and control the entire process of the conduct of free and fair
elections to the Parliament and the Legislative Assembly of the states/UTs. The
The ECI collects and documents election data for each and every parliamentary
and the state assembly constituency. For each constituency it reports data on the
total number of electors and voters which are segregated by gender, the name and
gender of each candidate contesting the election, party aliation of each contestant
and if the candidate is not aliated to any party then the candidate is categorized
as an independent, and the total number votes secured by each candidate in the
election. This data is available for every general election held in the parliamentary
5
and the state assembly constituency from 1951 till 2012.
For our analysis we use data at the constituency level for the state assembly
elections held for 16 large states from 1962 till 2012. These 16 large states rep-
resent more than 93 percent of the total electors in India. Next we describe the
construction of the variables of interest using the data at the constituency level.
PNs !
i=1 f emale votersit
sex ratio of votersst = PNs × 1000, (1)
i=1 male votersit
PNs !
i=1 f emale electorsit
sex ratio of electorsst = PNs × 1000, (2)
i=1 male electorsit
PNs !
i=1 f emale votersit
f emale poll percentst = PNs × 100, (3)
i=1 f emale electorsit
PNs !
i=1 male votersit
male poll percentst = PNs × 100, (4)
i=1 male electorsit
where s is the state, t is the year in which the election is held for the state assembly,
4 Results
The rst set of results are summarized in Table 2a and Table 2b. Before we explain
the results it is important to describe how we construct the average sex ratio of
voters for every state for the elections held in each of the decades from 1960s till
2000s. We illustrate this with the help of an example. In Uttar Pradesh (UP) in
the decade of 1960s 3 state elections were held in 1962, 1967 and 1969 and in each
of these elections the sex ratio of the voters was 538, 675, and 629 respectively. So
we compute the simple mean of these three sex ratios of the voters which is 614
and dene that as the average sex ratio of the voters in the 1960s in UP. For each
6
state we do a similar computation for all the decades. Our results reveal a very
interesting pattern - for every state from 1960s to 2000s there has been a signicant
improvement in the sex ratio of the voters, however, in Haryana and Tamil Nadu
the improvement is marginal compared to other states. Some of the largest gains
happened in the poorest states of India the so called BIMARU states of Bihar,
Madhya Pradesh (MP), Rajasthan and UP. However, we do also observe that there
is a signicant variations in the sex ratio of the voters across the states.
In sharp contrast when we look at the average sex ratio of the electors we do
not observe a similar pattern. As a matter of fact there are some states for which
the sex ratio of electors has marginally worsened from the 1960s till the 2000s.
This suggests that increase in the sex ratio of the voters is not driven by more
women relative to men registering to vote but it is driven by the fact that more
We also plot the sex ratio of voters and electors for every state over the election
years. As expected we observe that the trend in the sex ratio of the voters is
positive for all the states, however, for Haryana and Tamil Nadu (TN) it is more
or less at. In sharp contrast, the trend in the sex ratio of the electors for most
of the states remains more or less at with the exception of Haryana where it has
declined.
Next we do a simple regression analysis to analyze the trends in the sex ratio
of voters and electors. We also do a similar regression analysis for the female and
where yst is the outcome variable of interest: (1) sex ratio of voters, (2) sex ratio of
electors, (3) female poll percent, and (4) male poll percent. The subscript s is the
state, t is the year in which the election is held for the state assembly. post70 is a
dummy variable which equals 1 if the state elections are held on or after the year
7
1970 and before the year 1980, post80 is a dummy variable which equals 1 if the
state elections are held on or after the year 1980 and before the year 1990, post90
is a dummy variable which equals 1 if the state elections are held on or after the
year 1990 and before the year 2000, post00 is a dummy variable which equals 1 if
the state elections are held on or after the year 2000. εst is the error term.
The results of our regression analysis are in Table 3 from column (1) to column
(8). In column (1) and column (2) of the table we report the results for sex ratio
of the voters with and without the state xed eects respectively. From column
(1) we note that in the 1960s the average sex ratio of the voters was 715 and it
increased signicantly by 168 points to 883 in the 2000s. This is the key nding of
the paper. The results are very similar when we control for the state xed eects.
In particular, from column (2) we see that the sex ratio of the voters has increased
very signicantly over time and has improved for every successive decade. When we
compare the 1960s to 1970s we note that the sex ratio of the voters went up by 45,
from 1970s to 1980s it improved by 28, it further went up by 30 from 1980s to 1990s
reect the main point of our paper that gender bias in voting has declined very
signicantly over time. In column (3) and column (4) we report results for the sex
ratio of electors with and without the state xed eects respectively. We nd that
that the sex ratio of the electors has remained more or less the same throughout
the decades. This suggests that more women relative to men are casting their
votes in election. We directly test this by looking at the female and the male voter
turnout. The results for women voter turnout are reported in column (5) and
column (6) with and without the state xed eects respectively. Our results show
that over time the female poll percentage has increased very signicantly and it
has been increasing decade after decade. The rising trend started in the 1980s and
picked up momentum in the 1990s and 2000s. In sharp contrast when we look at
the men voter turnout (results in column (7) and column (8) with and without the
state xed eects respectively) we do not observe a similar pattern. In fact the
men poll percent has remained stable through the decades while increasing only
level of signicance. This result conrms that more women relative to men are
turning out to vote in the elections. This voluntary act of self-empowerment is the
8
key driving force in reducing the gender bias in voting.
by only considering the poorest states of India the so called BIMARU states of
Bihar, MP, Rajasthan and UP. Our results are very similar to the previous analysis.
In particular we nd that in the BIMARU states gender bias has declined very
dramatically over time and this phenomenon is driven by the fact that more women
The data has conclusively shown that over time, the increase in the female
voter turnout is signicant, in all the 16 states. The next line of enquiry then, is
In this section, we ask whether women vote dierently from men. In particular, we
analyze whether re-election prospects are aected dierently by men and women
voters. To address this, we study two closely held elections in the state of Bihar
in 2005. Bihar is one of the largest and the poorest states in India. In terms of
the total number of electors, Bihar is the 4th largest state in India with 55 million
in Bihar. The election was very closely contested but no political party emerged
rule was declared in the state, the state assembly was dissolved and re-elections
were held in all the 243 constituencies, eight months later in October of 2005. The
re-election results revealed a change in the winning party in 87 (36 percent) of the
243 constituencies. This change was sucient for one particular party to emerge
as the winner in the state, and formation of a new government. This setting of a
9
re-election which was held within a short span of time provides us with a natural
experiment, to study the impact of change in women voters vis-a-vis in men voters
October, the female poll percent increased from 42.5 percent to 44.5 percent while
the male poll percent declined from 50 percent to 47 percent (see gures 2 and 3).
Secondly, the mean growth rate of female voters was almost 3 times more in the
constituencies where political parties were not re-elected when compared to those
constituencies where political parties were re-elected. In sharp contrast, the mean
growth rates of male voters across the two types of constituencies were similar (see
Table 5). The two observations put together imply that not only did more women
cast their votes in the re-election, but that on average, they voted for change in
Next we analyze the distribution of the growth rates of female and male voters,
and study if they are dierent in constituencies where outcomes changed in the re-
election. In gures 4a and 4b, we plot the kernel densities of female and male voters
growth rates in the two types of constituencies: (1) the constituencies where the
same party was re-elected and (2) constituencies where political parties were not
re-elected and there was a new winner. We nd that the distribution of the growth
rates of female voters in these two types of constituencies are very dierent. In
sharp contrast, the distribution of male voters in the two groups of constituencies
conclude that while the growth rates of female voters are statistically dierent for
the two types of constituencies, they are statistically similar for the male voters.
This conrms our hypothesis that women voters caused the change in outcomes
in these elections.
We test the re-election prospects at the constituency level more rigorously next.
In particular, we study the eect of the growth rate in female voters and the growth
10
rate in male voters between February and October 2005, on the probability of re-
election at the constituency level using the PROBIT estimation. In our estimation
we also control for district eects (there are 38 districts in the state and on an
average each district has 6 to 7 assembly constituencies). Our results are reported
in Table 6. In column (1) we report the results without district xed eects and
in column (2) we report results with district xed eects. The results clearly show
that the coecient on the growth rate in female voters is negative and signicant,
while the coecient on the growth rate in male voters is positive and signicant.
Our results strongly suggest that growth rate in female voters had a negative eect
on the probability of re-election while the growth rate in the male voters had a
positive eect on the probability of re-election. To ensure that our results are not
rate in female and male voters. The correlation is 0.64 which is less than 0.9 and
to other studies that have analyzed the re-election prospects (Brender and Drazen
(2008)).
The eects of growth rate in female and male voters that we nd are not only
we look at the marginal eects in Table 7, we nd that 1 percentage point increase
in the growth rate of female voters reduces the probability of re-election by 1.4 to 2
percentage points. In contrast, for every 1 percentage point increase in the growth
rate of male voters raises the probability of re-election by 1.3 to 1.5 percentage
point.
rates of female and male voters. In gure 5, we show that the predicted proba-
bility of re-election declines as the growth rate of female voters increases. This
is calculated keeping the growth rate of male voters xed at the mean value. It
is evident from gure 5, that as the growth rate of female voters increases from
-10.2 (10th percentile) towards 13.9 (90th percentile), the predicted probability
of re-election falls steadily from 0.87 to 0.39. In sharp contrast, when we look at
11
the predicted probability of re-election with increases in the growth rate of male
voters, we nd the opposite eect. As gure 6 shows, the predicted probability
of re-election increases steadily from 0.48 to 0.79 when the growth rate of male
Together, these results conrm that men and women voted very dierently in
these elections. They aected the re-election prospects of political parties very
distinctly. More specically, women voters were the agents of change in this con-
text.
5 Discussion
Previous research on female empowerment have analyzed varying outcomes of in-
terest. Some of these include intrahousehold decision making (Ashraf, 2009), lead-
ership in politics (Beaman et al, 2009 and Chattopadhyay et al, 2004), access to
nance (De Mel et al, 2009), and seeking healthcare (Rai and Ravi, 2011). This
There is vast literature in politics which establishes the the necessity of uni-
versal surage for representative democracy. And yet, while advocating universal
surage, James Mill argued that in order to keep expense of representative system
down, women need not have separate voting rights because their interests were
included within those of their husbands and fathers (Alan Ryan, 2012). This was
later criticised by T.B. Macaulay who said, Without taking the trouble to perplex
the issue with a single sophism, he calmly dogmatises away the interests of half
the human race. Our research in this paper establishes that interests of half the
human race (female voters) are not merely signicant but also distinct from the
interest of the other half (male voters) and is reected in their voting behavior.
12
We study the data on Indian elections over the last 50 years and explore the
role of women voters. We nd systematic reduction in the gender bias in vot-
ing, over time and across all states of India. We interpret this as testimony to
self empowerment of women voters. The results of the Bihar re-elections of 2005,
which were held within a short period of 8 months, showed that women voted
dierently from men and signicantly aected the re-election prospects of political
parties. This suggests that men and women have dierent preferences and accord-
ingly, express these in their voting behavior. These ndings strongly complement
Chattopadhyay and Duo (2004) who show that men and women dier in their
So far, we have focused on documenting this silent revolution and exploring the
implications it has for political outcomes. As an extension to this work, we are also
the underlying factors driving this surge in female voter turnouts across all states
6 Conclusion
Existing research has shown persistent gender inequality in outcomes including
for the rst time, we discover a signicant and persistent reduction in gender in-
equality when we analyze voter turnout in all state elections in India, over past 50
this improvement does not seem to be brought about by a deliberate top down
policy action but largely due to voluntary participation of women voters in elec-
election outcomes, we study the Bihar state re-elections of 2005, which were held
within a short span. Our results strongly suggest that an increase in the female
voters turnout negatively eected the probability of re-election for a political party
in a given constituency. And in contrast, the results also show that male voters
13
The two results together show that men and women voted dierently. While
women voted for change, the men voted for status quo. These results highlight
14
References
[1] Anderson, Siwan and Debraj Ray. 2010. Missing Women: Age and
[4] Brender, Adi and Allan Drazen. 2008. How Do Budget Decits and
2220
[8] Dutta, Bhaskar and Poonam Gupta. 2012. How Indian Voters Re-
[9] Gupta, Poonam and Arvind Panagariya. 2012. India: Election Out-
15
[10] Iyer, Lakshmi, Anandi Mani, Prachi Mishra, and Petia Topalova.
[11] Rai, Ashok and Shamika Ravi. 2011. Do Spouses Make Claims? Em-
(6), 913-921
[12] Ryan, Alan. 2012. On Politics: A History of Political Thought from
[13] Sen, Amartya. 1990. More Than 100 Million Women Are Missing.
[14] Sen, Amartya. 1992. Missing Women. British Medical Journal. 304,
587-588
16
Table 1: States and the Election Years
States Election Years
17
Table 2a: Average Sex Ratio of the Voters
18
Figure 1a: Bihar, Madhya Pradesh, Rajasthan, Uttar Pradesh
19
Figure 1b: Southern States
20
Figure 1c: Remaining Large States
21
Table 3: Regression Analysis
(1) (2) (3) (4) (5) (6) (7) (8)
Female Female
Sex Ratio Sex Ratio Sex Ratio Sex Ratio Male Poll Male Poll
Poll Poll
of Voters of Voters of Electors of Electors Percent Percent
Percent Percent
22
post00 168.0*** 156.4*** 25.2 16.0** 12.7*** 12.1*** 2.81 2.45*
(27.2) (16.4) (16.1) (7.43) (2.73) (1.88) (1.77) (1.45)
Constant 715.1*** 846.2*** 903.2*** 996.4*** 52.7*** 60.5*** 65.8*** 71.1***
(23.2) (15.7) (12.4) (5.65) (2.23) (1.88) (1.15) (1.44)
State Fixed
N Y N Y N Y N Y
Effects
Adj-R2 0.218 0.823 0.025 0.850 0.135 0.615 0.028 0.390
Observations 176 176 176 176 176 176 176 176
Note: Standard errors in parentheses, * p<0.10 ** p<0.05 *** p<0.01
post70 is a dummy variable which equals 1 if the state elections are held on or after the year 1970 and before the year 1980 and 0 otherwise,
post80 is a dummy variable which equals 1 if the state elections are held on or after the year 1980 and before the year 1990 and 0 otherwise,
post90 is a dummy variable which equals 1 if the state elections are held on or after the year 1990 and before the year 2000 and 0 otherwise,
post00 is a dummy variable which equals 1 if the state elections are held on or after the year 2000 and 0 otherwise.
Table 4: Regression Analysis for the “BIMARU” States
(1) (2) (3) (4) (5) (6) (7) (8)
Female Female
Sex Ratio Sex Ratio Sex Ratio Sex Ratio Male Poll Male Poll
Poll Poll
of Voters of Voters of Electors of Electors Percent Percent
Percent Percent
23
post00 209.8*** 206.2*** -27.8 -29.9*** 15.0*** 14.7*** -0.59 -0.85
(31.4) (31.5) (21.7) (8.08) (3.41) (3.24) (3.22) (2.98)
Constant 585.0*** 596.2*** 898.6*** 970.1*** 40.0*** 40.3*** 61.0*** 64.6***
(21.0) (30.1) (18.9) (8.03) (1.94) (2.97) (0.93) (1.91)
State Fixed
N Y N Y N Y N Y
Effects
Adj-R2 0.604 0.725 0.097 0.925 0.516 0.610 0.135 0.319
Observations 47 47 47 47 47 47 47 47
Note: Standard errors in parentheses, * p<0.10 ** p<0.05 *** p<0.01
post70 is a dummy variable which equals 1 if the state elections are held on or after the year 1970 and before the year 1980 and 0 otherwise,
post80 is a dummy variable which equals 1 if the state elections are held on or after the year 1980 and before the year 1990 and 0 otherwise,
post90 is a dummy variable which equals 1 if the state elections are held on or after the year 1990 and before the year 2000 and 0 otherwise,
post00 is a dummy variable which equals 1 if the state elections are held on or after the year 2000 and 0 otherwise.
Figure 2: Female and Male Poll Percent in Bihar State Re-election in 2005
52
50
50
48
47
46
44.5
44
42.5
42
40
38
Female Poll Percent Male Poll Percent
Feb-2005 Oct-2005
Figure 3: Kernel density of the Growth Rate of Female and Male Voters between
February 2005 and October 2005
24
Table 5: Comparison of the Mean and Median Growth Rates
Constituencies Constituencies
Same Party Same Party
Not Re-elected Re-elected
Note: The Kolmogorov – Smirnov test rejects the equality of the two distribution functions
25
Figure 4b: Kernel Densities of Male Voters Growth Rate, for Constituencies
where Party was Re-elected and Constituencies where Party was Not Re-
elected
Note: The Kolmogorov – Smirnov test does not reject the equality of the two distribution
functions
26
Table 6: The Effect of Change in Women Voters and Men Voters on the
Probability of Re-election
Dependent Variable:
(1) (2)
RE-ELECTION
Growth Rate in Female Voters -1.4*** -2.0***
(0.004) (0.006)
Growth Rate in Male Voters 1.3**
1.5**
(0.005) (0.007)
District Fixed Effects N
Y
Note: This is the change in the probability of re-election (in percent) for 1
unit increase in the variables.
Standard errors in parentheses. * p<0.10 ** p<0.05 *** p<0.01
27
Figure 5: The Effect of Growth Rate of Female Voters on Predicted Probability of Re-
election
1
Predicted Probability of Re-election
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-10.2 (10th -3.2 (25th 3.8 (50th 8.9 (75th 13.9 (90th
Percentile) Percentile) Percentile) Percentile) Percentile)
Growth Rate
Note: The Growth Rate of the Male Voters is set at the mean value
Figure 6: The Effect of Growth Rate of Male Voters on Predicted Probability of Re-
election
0.9
Predicted Probability of Re-election
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-18.5 (10th -11.6 (25th -7.0 (50th -2.5 (75th 2.0 (90th
Percentile) Percentile) Percentile) Percentile) Percentile)
Growth Rate
Note: The Growth Rate of the Female Voters is set at the Mean Value
28