[sacw] Anatomy of BJP's Rise to Power (EPW)

[Harsh Kapoor]

September 9, 1999
FYI
(South Asia Citizens Web)
---------------------------------------
Economic and Political Weekly
August 21-28, 1999
Special Articles

Anatomy of BJP's Rise to Power
Social, Regional and Political Expansion in 1990s

Oliver Heath

I Social Base

II Social Expansion

III Regional Expansion

IV Political Expansion

Notes

THE spectacular rise of the BJP is one of the major political stories of
the 1990s. During the last decade the party has undergone a rapid
geographical and political expansion, the like of which has never been
seen before. From its lowly position in 1989, when the BJP was a small
localised party with a political presence restricted to just a few
states in the Hindi heartland, the acquisition of new territory and new
allies has transformed it into being the main political force in India
with a mass national following. The aim of this paper is to explore the
effect that these expansions have had on the social base of the BJP. Has
its appeal widened to attract voters from different social backgrounds?
Or has it merely strengthened its hold on its traditional support base?
Do the allies give the party an entrance in to new states and new
sections of society? These are the questions which need to be answered.

National aggregations provide a useful starting point from which to
carry out our investigation. By looking at the overall picture we can
gain a sense of what, if anything, has changed. However, the BJP's rise
in popularity has not been evenly spread across the country. The
regional expansion of the party can be seen to have taken shape in three
distinct waves. The first wave took place in the 1950s and early 1960s
when the BJP's historical predecessor, the Jan Sangh, first became a
significant political force at state level. These 'primary' states are
the states where the Jan Sangh emerged as a viable opposition party in
the 1950s and 1960s. The second wave includes the 'secondary' states,
where the Jan Sangh, and then after 1980 the BJP, were present before t
he 1989 boom, but were only minor opposition, winning just a few seats
in the 1960s, 1970s and 1980s. The third wave is marked by the 1989
election, when the BJP emerged as a force in many of the 'tertiary'
states. By looking at the social expansion with reference these
different levels of regional expansion, we can attempt to identify the
forces that lie behind the party's emergence as a national party.

I Social Base

It is well known that support for the BJP has tended to be more
concentrated among the upper castes and the wealthy, and it's weakest
among the Muslims and the underprivileged. But how do these different
sources of support compare? Is caste a more significant factor than
class? How much of a role do age and gender play? It is important to
answer these questions so that we have reference point to relate social
expansion to.

Bivariate analyses may to some considerable extent duplicate the
significance of factors. For example, educational level is closely
related to social class, therefore, education and class tables may
simply be redescribing the same phenomena rather than telling us
anything new. What is interesting is to examine the impact of one
variable on the vote for BJP and it's allies (BJP+), controlling for the
other variables. In other words, among people of the same community or
class, does education still make a difference to the way people vote?

To explore this question we carry out a multivariate logistic regression
in which we simultaneously include community, class, education and so on
rather than looking at them separately. We use logistic regression,
which is the appropriate technique when we have a binary dependent
variable.

We include in our model community, class, education, age, gender and
locality. These variables are all treated in exactly the same way as in
cross-tabulations, but all are included simultaneously. Table 1 shows
the parameter estimates. These estimates can be interpreted as fitted
log odds ratios. If we exponentiate them, we obtain the usual odds
ratios which are more easily interpreted. These are the ones shown in
the second of the split columns. These odds ratios show the fitted odds
of supporting the BJP and its allies in the particular group in question
relative to the overall odds. These are what is known as 'deviation
contrasts'. If the fitted odds ratio is greater than 1, this indicates
that the group in question is relatively likely to support BJP+, while
if the odds ratio is less than one, it indicates that that group is
relatively unlikely to support BJP+.1

The magnitude of the parameters tell us by how much the particular
groups differ from the overall electorate (controlling for the other
variables in the model). Thus if the fitted odds ratios are close to 1,
either just above or just below, then we can say that the groups in
question are relatively similar in their support for BJP+. The further
apart they are, the more different the group's voting patterns are.

Table 1 shows the parameter estimates for each of the variables. To some
extent the support base of the BJP resembles, if not the elite, then
definitely the middle and upper classes. That is to say, that although
the most advantaged members of society may not always be the most likely
to vote BJP, the most disadvantaged are at any rate the least likely.
Although it may be tempting to say that the BJP is therefore a party
that represents social and economic privileges, none of the evidence is
really strong enough to support this. On the whole the degrees of
polarisation within the variables are relativly weak, that is with the
noticeable and distint exception of community.

Controlling for all the variables, it is the upper castes that are by
far the most likely to vote BJP. This means that although graduates are
more likely than illiterates, and men more likely than women, and so on,
caste over-rides their influence.

Table 2 gives a summary of how significant each of the variables are in
effecting the vote for BJP+. Chi2 is a measure for the difference
between the expected and the observed results. The expected results
assume a null association. That is, one would expect all groups within
the variable to give the same degree of support to the BJP+. The greater
the difference between the relative levels of support, the greater the
Chi2. Community and class, and to a lesser extent sex, age and locality,
all play a part, although community is by far and away the most
significant. Therefore, if the social base of the BJP has expanded, the
key variable that we need to measure it with reference to is community.

II Social Expansion
By looking at the internal composition, or column percentages, of the
BJP+ we can analyse the distribution of the different community's within
the party. This tells us what percentage of BJP voters are upper caste,
OBC and SC, etc. Secondly, we can look at the degree of support that
each community gives to the BJP. Although all the communities may vote
BJP to some extent, the row percentages and odds ratios tell us which
community offers the greatest and least support. However, the row
percentages only show absolute increases, and not relative increases. To
see whether the communities exhibit any change in their voting patterns,
or are just being swept along with the tide, we need to look at the odds
ratios. It is important to look at the picture from both sides. Inflow
tables tend to highlight the cross-cutting of cleavages, whereas the row
percentages show to what degree people vote along caste lines. Although
similar information, in terms of change, can be gained from both
approaches; the distinction remains important. Thirdly, we can look at
the overall effect of community on voting. Phi and Chi2 both measure its
aggregate significance by quantifying the level of polarisation along
community lines in one election.

As the multivariate analysis led us to suspect, it is the upper castes
that form the most dominant section of BJP+ support. This was still the
case in 1998 as it was in 1991, so in that respect it would seem that
little has changed. However, the degree to which they predominate has
undergone major changes. Table 3 shows that there has been a real and
significant shift in the social composition of the BJP+. The proportion
of BJP+ voters who are upper caste has fallen by 10 per cent since 1991.
This shift away from the heavy reliance of upper castes has been matched
by gradual growth in representation from all of the other communities.
The growth that has been most pronounced is amongst the OBCs, whose
presence has increased by 4 per cent, although there has also been an
increase in the other social groupings of SC (3 per cent), ST (2 per
cent) and Muslim (1 per cent). However, these counter-growths have been
relatively small, and could just be the result of standard sampling
error. We should not therefore read too much into them at this stage.

The row percentages in Table 4 show the degree of support that each
community gave to the Jan Sangh in 1967 and to the BJP+ in 1991, 1996
and 1998. The 1967 data need to be treated with some caution. Not only
is the overall sample size quite small and restricted to only male
respondents, but also the Jan Sangh share of the vote is small.
Therefore, the sample may not have accurately picked up the Jan Sangh
voters. However, broadly speaking we can say that the upper castes gave
far more support to the Jan Sangh than the other community groups did,
which is also true for the BJP 30 years later. In this sense then, there
has been a strong degree of continuity between the respective social
profiles of the two parties.

However, in terms of finer analysis, we are unable to draw any other
robust links between the two parties. Instead, what we can do is to see
what has happened to the BJP since 1991. The row percentages show that
during the 1990s there has been absolute growth in support for the BJP+
across all the communities. However, to see whether this rise in
popularity has been evenly dispersed across the board we need to look at
how the odds ratios have changed.

In the hierarchy of support by community little has changed since 1991.
Relative to the overall electorate, the upper castes still give the
greatest support to the BJP, followed by the OBCs, the scheduled tribes,
the scheduled castes and the Muslims, respectively. However, the odds
ratios show that there has been a relative decrease in the support from
the upper castes. Even though more upper castes voted BJP+ in 1998 than
in 1991, their overall dominance has been reduced by the greater
propensity of the other community's to vote BJP+.

The odds ratios for the upper castes were their highest in 1996.
However, they significantly fell in 1998. Even though the odds ratios
for the upper castes voting BJP+ were higher in 1991 than in 1998, the
overall effect of community was lower. This indicates that the social
cleavages were not as pronounced as they later became. The Chi2 for
community was at its highest in 1996, and then dropped a little in 1998.
However, the disadvantage with using Chi2 as a comparative measure
between samples, rather than a comparative measure of variables within
the same sample as was used in Table 2, is that it is sensitive to the
size of the sample. Thus bigger values are expected when big samples are
used. To some degree then the rise in Chi2 might merely reflect the over
all rise in popularity for the BJP and not indicate a strengthening of
social cleavages as might first be assumed. To verify this we use Phi.
The Phi results support what Chi2 told us.

Therefore social profile needs to be explored from two angles. Firstly,
what is the relationship between the move away from upper castes, and
the rise in the other communities? And secondly, what is the overall
effect of community? Having examined how the overall shape of the BJP
and its allies has changed, we next need to examine how these profiles
compare to those in the three waves of BJP expansion.

III Regional Expansion
Does the social profile of the BJP remain constant from the time of its
establishment? Does the social profile take the same shape in each wave?
How do the different social profiles change over the course of time?

To answer these questions we will test the validity of three hypotheses.
Firstly, that expansion takes the form of a top-down conversion. That
is, it enters through the upper castes, who have traditionally been the
most likely to vote BJP and would therefore be the most receptive to its
arrival, and then draws in support from the other community groups
afterwards. The second hypothesis is that the social profile on entry
reflects the BJP+'s profile at that time. In this respect we would
expect to see similar profiles in each of the regions. Thirdly, in each
phase, or penetration, the party redefines itself and explicitly tries
to appeal to new voters. In this instance there would be markedly
different profiles in each of the regions. Perhaps unsurprisingly, the
states where the BJP has had the longest political presence are the
states where the BJP is still most popular. The secondary states are
also the second most popular. In 1996 the BJP+ vote share increased
noticeably, and the party is now only marginally less successful than it
is in the primary states. In the tertiary states the BJP had an almost
non-existent presence before 1989. However, it is in these states that
its growth has been most dramatic, rising from 10.2 per cent in 1991 to
35.4 per cent in 1998.

Table 6 shows the inflow of where the BJP+'s support comes from. Whereas
Table 5 showed the percentage of people in each region who voted BJP+,
this table shows each region as a percentage of the total number of
people who voted BJP+. That is, in 1998 31.4 per cent of the people who
voted BJP+ came from the primary states. The table illustrates how the
BJP's reliance on the primary states has dwindled over the years.
Although the BJP is most successful in the primary states, in 1998 the
largest number of its votes came from the tertiary states. This makes
the party's expansion into this area all the more significant.

Table 7 shows the column percentages for BJP+ support in each of the
regional groupings. At first glance the table seems to paint a rather
confusing picture. However, by focusing on the changes that have taken
place in each region, we can gain some indication of what each ones
enduring characteristics are, and thus begin to test our hypotheses.

The proportion of upper castes within the BJP fold in the primary states
has not changed significantly over the three elections, with a slight
fall between 1991 and 1998 the values have merely oscillated 1
percentage point either side of the average, and 1996 value, of 41 per
cent. Similarly its proportion within the tertiary states has not
altered much either. That leaves us with the secondary states. It is
within this region that the big changes have occurred. The proportion of
upper castes in 1991 was 54 per cent, marking it as the most upper caste
reliant region of all. However, in 1996 its share fell by 12 percentage
points, which was further consolidated by an additional fall of 2
percentage points between 1996 and 1998. In terms of a move away from
the upper castes it is therefore in the secondary states that the major
changes have taken place.

How do these changes compare with what has been happening with the other
castes? The OBC share has been slightly erratic in the primary states. A
slight rise in its share between 1991 and 1996 was followed by a
slightly bigger drop. The overall impression though is of relative
continuity, in at least in sofar as all the changes have basically
cancelled each other out. In the tertiary states there has been a
noticeable fall in its share. However, in the secondary states its share
has consistently gone up, rising from 34 per cent to 37 per cent to 40
per cent. The scheduled castes and Muslims have remained fairly stable
in the primary states, and made some gains in the secondary and tertiary
states. The scheduled tribes have increased their share in the primary s
tates, and remained fairly stable in the secondary and tertiary states,
although their share did drop a fraction in 1998. However, for the small
groups, such as the scheduled tribes and Muslims, even small changes in
their proportion within BJP can be the result of significant changes in
their voting behaviour. Therefore, it is a good idea to see how their
odds ratios have changed.

The row percentages offer an interesting insight into the voting
behaviour of the Muslims. In 1991 a negligible number of Muslims voted
BJP in the secondary and tertiary states. They are the last group to
join the BJP bandwagon. However, as the BJP established itself in these
states they voted for it in greater numbers. The voting seems to go in
waves. In 1991 they only voted BJP in the primary states, where the
party was already firmly entrenched. In 1996 they voted in all regions,
in greater numbers in the primary states and less so in the tertiary
states. In 1998 they again voted in all regions, but this time just as
much in the tertiary states.

This pattern leads to two possible conclusions. Firstly, that as the
last to enter, they are led by the other communities, thus following a
top-down conversion process. But the fact that their growth was most
evident in the tertiary states indicates that the BJP+ appeals to them
(or vice versa) when it sets itself up anew.

The odds ratios provide us with the firmest basis from which to examine
our hypotheses of social expansion. In order for the top-down process to
give an accurate representation of what has happened, we would expect to
see the highest odds ratios for the upper castes, and the lowest odds
ratios for the Muslims and scheduled castes, in the tertiary states.
This is clearly not the case as both Muslim and scheduled caste support
is weakest in the primary states and strongest in the tertiary states.
And the upper castes odds ratios in the primary states are substantially
higher than their odds ratios in the tertiary states. Thus, due to the
lack of corroborating evidence we must refute the top-down hypothesis.

The second hypothesis, that the party carries its present profile into
the states that it penetrates, is slightly more problematic. It is
somewhat difficult to ascertain whether the BJP takes its profile with
it when it enters new regions for two reasons. Firstly, we do not have
adequate data for the emergence of the party in the secondary states,
and given what we have, we cannot assume that the profile at entry
follows the same course in each of the regions. That is, hypothetically,
even if the early profile of the secondary states reflected the profile
of the primary states of that time, they may both have evolved in
different ways over the course of time. Therefore, even though the
profile of the secondary states does not closely resemble the profile of
the primary states now, or even since 1991, we cannot rule out the
possibility that it once did. However, our data series does manage to
capture the initial expansion into the tertiary states, and its gradual
consolidation, so this gives us a more robust footing from which to
speculate. Given this, a further question arises. That is, which profile
would carry through to the tertiary states, that of the primary states,
or that of the secondary states? The question becomes redundant on
analysis. The profile in the tertiary states bears little resemblance to
either primary or secondary in 1991, 1996 or 1998, so we must refute the
theory. This leaves us with third and most likely alternative. That is,
the party redefines itself and tries to appeal to different sections of
society each time it establishes itself on new territory.

In 1991 and 1996 the tribal vote becomes relatively stronger the further
the region is away from the core. However, this pattern is clearly upset
by the 1998 results, where it is strongest in the primary states and
weakest in the secondary states. The major growth has obviously taken
place in the primary states, but why is this? Perhaps this tells us
something about the power of mobilisation. The tribals' natural impulse
to vote for the BJP more is when the party is an outsider. However, as a
group to be mobilised, the tribals are relatively untouched. Thus the
BJP is able to appeal to them in a way that it could not do to the OBCs,
SCs and Muslims while still maintaining the support of the diehard upper
caste faithful. In this way it follows the classic upper caste --
scheduled caste mobilisation strategies that were employed by Indira
Gandhi in the 1970s. Thus, in light of the changes in the composition of
the BJP+ in the different regions that we noted from Table 7, we must
accept that the party shows a preference to mobilising different
communities in different regions. In this sense then it does redefine
itself.

In terms of the overall effect of community we can see that the Chi2 and
Phi values decrease as the region moves further away from the primary
states. This indicates that caste is less significant in the states the
BJP has most recently infiltrated, and is still most significant in the
old guard states where the BJP first emerged. What does this tell us?
Firstly, that the cleavages that were set up when the Jan Sangh emerged
have by and large remained in tact. Although the party has grown in
these states, it has never managed to rid itself of its initial
cleavages, but then, neither has had to. The tertiary states are still
less caste-based than other states, but whereas the primary states have
only increased marginally since 1991, and the secondary states have even
decreased, there has been a somewhat alarming rise of community
polarisation in the tertiary states.

IV Political Expansion
Allies are an operational indicator of expansion and political
credibility. The biggest electoral gains that the BJP combine made were
in the regions where it had allies. How does the profile change with
this inclusion? Do the allies give the party an entrance in to other
social groups? Or do they appeal to traditional BJP voters? Do they play
a pivotal role in the party's reinvention? Do the allies offer a
significant account for the widening social base of the BJP in the
secondary and tertiary states? In the same state how does support for
the BJP and its allies compare? To examine how respective sources of
support for the BJP and its allies differ we disaggregate the state
groupings in which the BJP has fought alongside allies. In 1991 and 1996
this was only in the secondary states, and in 1998 it was the secondary
and tertiary states.

In the secondary states in 1991 all the communities gave greater support
to the BJP than to its allies. However, in 1996 the balance of power
swung, with the allies claiming roughly equal, and marginally greater
support amongst the OBCs and upper castes, respectively. The scheduled
castes and scheduled tribes still gave greater support to the BJP than
to the allies, and the Muslim vote was divided evenly between the two.
In 1998 however, the allies increased their support relative to the BJP
amongst the scheduled castes as well, leaving only the scheduled tribes,
and the Muslims to a small extent, preferring to vote for the BJP than
its allies.

Table 11 shows the complete picture of support by region for the BJP and
its' allies. The social base of the allies needs to be analysed both
internally, comparing the differences between regions and over time, and
externally, comparing it to the profile of the BJP.

The allies have two types of social base. The first wave of allies,
those in the secondary states, has much more support amongst the upper
castes and OBCs than they do amongst the other communities. Compared to
the tertiary allies, the odds ratios for the SC, ST and Muslims are very
low in the secondary states. However, the profile is somewhat different
in the tertiary states. Although the odds ratios are still highest, by
quite a margin, for the upper castes, there are a number of communities
with odds ratios around, or just below one. The scheduled castes,
scheduled tribes and Muslims, all have odds ratios that, relative to
their other odds ratios in other regions and years, are high. Although
the odds ratios for these communities do not set individual precedents,
they do mark a significant collective increase. This is reflected by the
very low values of Phi and Chi2, of which the Phi value is the lowest
ever, second only to the value for the tertiary states in 1991.

By comparing the social profile of the BJP with that of its allies in
the tertiary states in 1998 we can see how the inclusion of allies
effects the social base of the BJP. In the tertiary states the upper
castes, SC, ST and Muslims, all offer relatively greater support to the
allies than to the BJP. This leaves the BJP somewhat OBC dominant, and
weaker among the other castes. This suggests two things. Firstly, in the
tertiary states the upper castes do not have the same affinity with the
BJP as they do in the primary states, which may be partly due to the
fact that in the tertiary states the BJP is not the preserve of the
upper castes. Thus, the upper castes, the SC, ST and the Muslims all
prefer the allies to the BJP, although probably each for somewhat diffe
rent reasons.

To what extent has the BJP gained from its acquisition of allies? The
social base of the BJP cannot be isolated purely by looking at who votes
for the BJP. Depending upon seat sharing arrangements, voters who are
committed BJP supporters may be forced into voting for one of their
allies. Therefore, an interesting exercise is to see which party
respondents would have voted for if there had been no alliances, in
1998. Table 12 shows the parties that respondents said they would have
voted for if there had been no alliances. In the secondary states the
BJP not only retains more of its own vote, but also claims more of the
allies vote compared to the tertiary states. This indicates that the
leaning of those who voted for the allies is much stronger towards the
BJP in the secondary states. Due to the reduced sample size that table
produces, there are not enough cases to say anything meaningful about
the STs and Muslims. Therefore they have been discounted. However, what
is clear is that a significantly higher percentage of upper castes would
have voted BJP in the secondary states than in the tertiary states. This
gives greater support to the view that the allies in the tertiary states
appeal to the less traditional BJP voter, thus giving the BJP an entry
into social groups that would not normally vote for them.

Conclusion

So how has the BJP expanded? What effect has its regional and political
expansion had on its social base? The regional expansion of the BJP has
been inter-twined with a distinct three tiered growth in its social
appeal. To a large extent the presence of its political allies have
aided this process, and allowed the party to gain a foothold in new
territory. The relationship between the three forms of expansion is
undoubtedly connected. However, as far as causal relationships go the
direction is unclear. One can safely assume though that the BJP has been
an active participant in determining which course its social expansion
takes. And as for those who say that the BJP has little in common with
most of its allies, it would seem that the selection criterion relies
more on the allies social appeal than on their ideological stance. In
this respect the mobilisation strategies that the BJP have employed have
been very carefully orchestrated.

In each successive step that the BJP makes away from its homeland of the
primary states, the groups that have expanded the most also move a step
down the ladder of the party's traditional support base. Thus in the
primary states, which represent the core of the party's stronghold, its
core source of social support, that of the upper castes, has remained in
tact. The only other community that has been significantly mobilised in
this region is the scheduled tribes.

Moving one step away, into the secondary states, it is the OBCs, who
overall are the second most likely group to vote BJP, who have made the
most significant increases in their propensity to vote for the party.
The states in this region have a profile more similar to the primary
states than the tertiary states do. This is illustrated by the social
base of the allies, which is relatively strong amongst the upper castes
and OBCs and relatively weak amongst the other communities. It thus
reinforces the overall mobility drive, with both BJP and allies working
in tandem to appeal to similar sorts of people.

The third tier of scheduled castes and Muslims have emerged most
strongly in the tertiary states. In these states the BJP's reliance on
its allies, both as a vote winner and as an entry point into other
social groupings, is at its strongest. Whereas in the secondary states
the partnerships mainly took the form of BJP led alliances, in the
tertiary states the alliances are generally regionally led, with the BJP
supporting from the outside. Party's such as the AIADMK, TDP and
Trinamool Congress, and to a lesser extent the BJD, carry more weight
locally than the BJP. Thus the BJP prospers by association, and without
them would most probably become marginalised.

It is in the tertiary states that the less traditional BJP voters have
become more enfranchised, similarly it is the presence of the allies
that brings most support from these social groups. With the BJP on its
own gaining little favour with anyone other than the upper castes and
the OBCs, its political future lies in the hands of its' allies. The
biggest gains that the BJP made have been in these states, which now
also constitute its largest source of support. So, as the BJP pushes to
establish itself at the centre once again, the role of the allies in the
these states will become more important than ever.

In a sense then, there are three parties. Admittedly they are not
completely independent ones, but nontheless - they are each distinctive
in their own way. Each one has played a significant role in helping the
BJP become the national party that it is today. It has only been by
delicately redefining itself and its social base that the party has been
able to spread its wings and leave its nesting place of north India.
Although much of the momentum for this change has come from within the
party, the transition has only been finally possible because of the help
it has received from parties outside. This help though, seems to be
somewhat illusory. The allies, especially in the tertiary states, have
not provided a push for the BJP to reach a firm hand-hold, but have
provided the hand-hold itself. If the allies were to let go then in all
liklihood the BJP would have a long way to fall.

Notes
[Many thanks are due to Yogendra Yadav and Anthony Heath, who were both
great sources of help in preparing the outline for this paper, and to
the CSDS Data Unit in general, and Himanshu Bhattacharya in particular,
for assistance with the aggregate data.]

1 The SPSS syntax used to run this logistic regression is:
Logistic regression BJP
/Method = enter comm occup educ age sex locality
| /Contrast (Comm) = Deviation
/Contrast (Educ) = Deviation
/Contrast (Age) = Deviation
/Contrast (Sex) = Deviation
/Contrast (Locality) = Deviation
/Criteria pin (.05) Pout (.10) iterate (20).
Note that as Sex is a binary variable we do not need to write a Contrast
statement for it.

2 For a detailed description of how the community variable was
constructed, see appendix XX of Congress paper by Anthony Heath and
Yogendra Yadav.

3 Similarly, for explanation of the Class variable see appendix XX.

4 Similarly, for a methodological note on the National Election Surveys
see appendix XX.