IT has been a great temptation to tell Congress leaders and workers who the Congress voter is. You might say they know it all. Possibly a Congress worker and local leader have a fair sense of who their supporters are in their locality. But it is unlikely if the same can be said about the state level and national level leadership. If they do, their actions often fail to betray this understanding.

While we live in a post-Congress polity, the Congress leaders continue to behave as if the grand old era of the Congress system continues uninterrupted, as if the last 10 years never happened. Some of them still live the day-dreams of a single party majority. They plan and strategise as if they are still a catch-all party representing all sections of society. Hence my desire to write a Rough Guide to a Post-Congress Polity.

I will try to turn this invitation from Seminar into an occasion to look at who the Congress voter is and why this piece of information may be of some relevance. I will draw upon a wide range of survey evidence gathered by the Centre for the Study of Developing Societies (CSDS) over the last four decades, for survey data constitutes the most reliable piece of evidence to discuss voting behaviour.

 

I

Let us begin with a simple overview. Table 1 presents the data for Congress voting by community in the four general elections that took place in the 1990s. This period is often described as the post-Congress polity, for Congress ceased to be the central pole around which political competition was structured in the country. A look at the Congress vote in the 1991 election shows a fairly even spread. The party secured 37% votes in the country.

Its votes among the various social groups classified here is fairly close to this average. The vote share dips by 5 percentage points among the Hindu OBCs, a weakness that goes back to the 1960s when the party began to lose the political support of the middle peasant castes, especially in north India. The only high point in this column is that the adivasi vote is 8 points higher than the Congress average. All other groups, including upper caste Hindus, dalits and the Muslims are within a two percentage point band around the average of 37%.

This may not be a textbook illustration of a ‘catch-all formation’ that draws the same level of support from all sections of society. But the shadows of a catch-all party are clearly visible. In the heydays of the Congress, Prannoy Roy and David Butler had proposed a ‘10% band’. The proposition was that Congress vote among different sections of society falls within a range of 10 percentage points – from 5 points plus to 5 points minus – of its average vote share. The rule could be applied to the picture we see for the 1991 election, with adivasis placed just outside the band.

From this starting point, let us look at the changes depicted in the same table. By the time we come to the 1996 election, the first election without the Nehru-Gandhi dynasty, the Congress votes drop by nearly 8 percentage points. Notably, the drop was not even across different social groups. The loss among the adivasis and Muslims (even in this first election after Babri Masjid demolition) was only 4 points each, while the party lost nearly 10 points each among the upper castes and the OBCs. Adivasis and the OBCs had by then moved outside the 10% band.

 

 

The next two elections came in quick succession. The overall vote of the Congress dipped further by 3 points in 1998 only to spring back to the same level in 1999. But the social composition of the Congress voter underwent a change. Its votes among the Hindu upper caste declined with each election. On the other hand, the Muslim anger subsided and compulsions of political choice steadily brought them back to the Congress. There was a little recovery among the OBCs as well. There were no further losses for the Congress among dalits and adivasis during these two elections.

During these 10 years the Congress has come a long way from the catch-all formation it once was. While it lost about 8 points in overall terms, its loss among the upper caste Hindus was as high as 20 percentage points. The Congress is the weakest not among the OBCs, but among the upper caste, socially the most privileged group. At the turn of the century the Congress no longer enjoyed the confidence of the social group it had worked with throughout the 20th century to bring itself to power. It began to look like a ‘cleavage based party’, a party that draws most of its votes from some social groups and not from others. Its vote share among the adivasis, Muslims and upper castes is now way beyond a 10% or even 20% band.

 

 

The picture of voting by different economic groups in Table 2 tells a similar story. In 1991, the vote for the Congress was evenly spread, even more so than in Table 1. Between 1991 and 1996, the Congress votes underwent steep erosion among the well-to-do or the ‘high’ class in this categorization. This high class is the same as the so-called ‘middle’ class in popular imagination (not to be confused with the middle group in our classification; a member of this group falls truly in the centre of the income and asset based hierarchy of Indian population and dreams of owning a two-wheeler).

The Congress votes in this group went down by 16 percentage points between the two elections. Rarely does such a sharp erosion take place in a short period and that too for a group that is not tied together by primordial bonds. This incidentally was the period when the Congress government had given a U-turn to the country’s economic policy. The ‘high’ class constitute precisely those who benefited from the New Economic Policy. The beneficiaries of the NEP had deserted their political benefactors even before the first generation of economic ‘reforms’ were over.

 

 

Ironically it is precisely the victims of these reforms at the bottom of the class hierarchy who stayed more loyal to the Congress than other groups. The exact pattern varied from election to election. The 1998 election appeared to have reduced class differences in voting. But by the end of the decade, we see a clear pattern emerging. The vote for the Congress has a clear, almost textbook like, slope: the lower a person is placed in the class hierarchy, the higher his or her chances of voting for the Congress. The slope is not steep, but the relationship and its direction is unmistakable. Caste and its association with voting has invited so much popular and professional attention in the last 10 years that we are in a danger of forgetting how clear the line of class division is when it comes to voting.

Let us introduce one more variable at this stage. It is widely believed, and there is some evidence to support this view, that in its years of glory the Congress got substantially more votes among women than among men. Right from the ’60s, we know that a disproportionately large number of supporters of ‘Opposition’ parties were men. This trend continued through the ’70s and ’80s. Table 3 present the trends for the 1990s. By the time we come to the beginning of the ’90s, this gap is not visible. Although the evidence is somewhat deficient on this point, it seems that some political changes in the late ’80s and early ’90s led to an erosion of the Congress support among women.

In 1991 the Congress secured 4 percentage points more among men than women, very unusual for the Congress. Its opponents did better among women. By 1996, the deficit had disappeared, yet the Congress had no gender advantage: it got the same proportion of votes among men and women. But it was already doing a little better in this respect than the BJP and allies. The 1998 and the 1999 elections saw the Congress recover something of its traditional advantage among women.

 

 

By 1999, the Congress vote share was 3 points higher among women as compared to men. One may be tempted to link it with the arrival of Sonia Gandhi to the political centre-stage. But this subject remains to be explored adequately. All we know is that by the end of 1990s, a substantial gender division in voting had opened up. If viewed in terms of the difference between the Congress’ advantage and the NDA’s disadvantage, the gap was around 9 percentage points.

If we review the national picture in terms of all these three divisions, then the Congress turned from a catch-all party to something of a cleavage based party in the 1990s. It was no longer the party of a rainbow coalition, even granting that its rainbow was always thick at the ends and there was a gender dimension to its support. In the 1990s the slices of its rainbow were beginning to come apart.

The Congress was never a party of the OBCs, and it did not become so in the 1990s. But the cumulative result of a process of uneven erosion was that the Congress was now the party whose voters were more likely to be dalit, adivasi, Muslim and poor than ever before and as much likely to be women as in the past. Still, for someone used to looking at cross-country data on voting by social cleavages, the national picture does not appear to be that of sharp social divisions.

 

II

The picture that we have analysed thus far suffers from several limitations. First and foremost, it is presented at an artificial level of aggregation. We know that Indian politics in the 1990s was no longer played at the national level; states had become the effective unit of political choice. The national figures hide the real story at the state level. We need to know if the story at the state level follows the same contours. For the same reason the national level data does not offer us good clues about the why and how of sectional mobilization, for the real action lies at the state level. Finally, stopping the story in 1999 does not tell us much about some of the most pressing questions staring everyone in the face in the run up to the next Lok Sabha elections. What has been the trend since the last Lok Sabha elections? What are the lessons for the Congress in the next Lok Sabha elections?

It is by now an established wisdom that the Congress’ profile at the state level is different from that of its national level. Working on the data from the 1967 Lok Sabha election, Chhibber and Petrocick¹ had argued that the Congress was never a ‘catch-all’ formation at the state level and that its support was more unevenly distributed at that level. Anthony Heath and I had extended that line of reasoning and modified it by looking at the profile of the Congress voter across different categories of states in the 1996 and the 1998 Lok Sabha elections.² The main conclusions of that research are worth recalling here:

‘Congress support changes from one type of contest to another. While the all-India figures show relatively little variation in support from one community group to another, we find much bigger variation once we disaggregate. Moreover, the pattern varies from one type of contest to another, but the differences tend to cancel out at the all-India level. Thus in competition with the BJP including its regional allies, the Congress comes out as a party of the socially and economically marginalized. The same party is supported by the socially and economically privileged when it competes with the left. While it retains a catch-all character in some states where it faces all-regional parties, it declines into a catch-none formation when it is pushed from more than one direction by various cleavage based parties’ (Heath and Yadav 1990: 2526).

We have already noticed one change since this snapshot was taken: unlike then, the all-India picture of the 1999 Lok Sabha election already shows clear cleavage lines. We need to see if this is reflected in the state level competition since then. Besides, both the studies mentioned above have drawn upon the Lok Sabha election data. Let us turn to information at the state level in the last assembly election for all the major states. This information has become available for the first time through a series of surveys based on probability sample undertaken by the CSDS for all the major assembly elections that have taken place since 1996.

Table 4 presents the data on voting by community for 18 states. It uses the same five categories of social communities as used in Table 1 for the Lok Sabha data so as to facilitate a comparison among different states and between the all-India and the state level picture. In order to facilitate quick reading of the data and to neutralize the effect of the ups and downs of the party in different elections, the data is reported here as deviations from the mean. In other words, instead of reporting the vote share of the Congress among different groups (as done in Table 1 and 2), the table simply reports how much the Congress was ‘up’ or ‘down’ in a social group compared to its average vote in that state in that particular election.

 

 

A first look at the table confirms the thesis that variations at the state level are substantially higher than those seen in national level aggregates. There are a large number of double digit variations, indicating the presence of sharper social cleavages at this level. A first glance also reveals some clear overall patterns. The column for ‘upper castes’ shows negative signs in all but four cases; similarly the column for dalits and Muslims reports mostly positive and large entries. The evidence for OBCs and adivasis is more mixed, as there are both kinds of entries.

At a general plane then one can conclude that the Congress is as weak among the upper caste and OBCs and much stronger among the dalits and Muslims than indicated by the all- India data in Table 1. The evidence for adivasis does not affirm the conclusion that follows from the all-India table. Although there are significant state-wide variations here, it seems the Congress cannot take the adivasis for granted in the assembly elections analysed here. We do not know if this is due to a change since 1999 or because of the different level of analysis.

A close look at the table shows three groups of states in terms of their patterns of vote for Congress. The first group of states includes most of the states with a direct Congress-BJP contest or that involving almost a direct contest between Congress and the NDA. But this group also includes Andhra Pradesh and excludes Madhya Pradesh. The social profile of the Congress in these states is characterized by weaker than average support among the upper caste Hindus and Hindu OBCs in that order. Almost without exception the Congress does substantially better among dalits and Muslims than its average support level. All states except Gujarat with significant adivasi populations show that the Congress is weaker among them. There are variations. The distrust of the Congress among the upper castes is not high in states like Haryana and Andhra Pradesh where political divisions among non-dvijas are more salient.

 

 

The Congress performs relatively better among the OBCs in a state like Gujarat with a dynamic OBC leader and in Haryana where the dominant peasant community is not an OBC. The level of support among the dalits and Muslims is consistently and substantially higher than the average in all these states. Uttaranchal and Maharashtra register a lower level for Muslims for there is an alternative pro-Muslim formation in the form of SP and NCP respectively in these two states. Maharashtra, Rajasthan and Andhra Pradesh, three states with substantial adivasi population, register a lower than average support for the Congress among the adivasis, something that must cause anxiety to the Congress leadership.

The second group has states that appear to have a different profile, but once we put the data in the local context we can see that this profile is simply an extension of the basic principle underlying the first group. Madhya Pradesh falls in the second group, for unlike any other state in the first group the dalit vote for the Congress is significantly lower than average. This is due to the rise of the BSP in the northern belt of the state that borders Uttar Pradesh.

Since this phenomenon is confined to a region within the state and the BSP has shown signs of decline after an initial breakthrough, this deviation may not alter the social profile of the Congress voter on an enduring basis. In Punjab, the Congress does better among the upper caste and OBC Hindus, but this has little to do with their caste and more to do with their status as a religious minority. Keeping with the pattern reported for group one, the Congress does very poorly (13 points below the state average) among the real dominant community of the state, the Jat Sikhs.

In Kerala too, the Congress led UDF is the preferred political choice of religious minorities. While the coalition does poorly among OBCs, dalits and adivasis, it makes up with a heavy support from Muslims and Christians (18 percentage points above the state average). In Assam, the Congress makes up for its weakness among all Hindu communities and adivasis by doing very well among Muslims and a large number of ‘others’, presumably outsiders who do not fit into local categories. While deviating from the standard picture, the Congress retains its character as a party of the socially marginalized in all these states.

Finally we have the third category in which the Congress is not a party of those from the social margins. It represents the privileged groups or lacks a clear social profile. In Bihar and Uttar Pradesh, whatever is left of the Congress vote is now more concentrated among the upper castes than among the OBCs or dalits. In Bihar, the Congress did well among the adivasis, but with the formation of Jharkhand that vote will no longer be available to it. Squeezed from both ends by rival political formations, Congress is a catch-none formation in these two states now.

The splits and the marginalisation of the Congress in Tamil Nadu and West Bengal has left it as a party without a clear social profile. It does worse among both the upper castes and dalits and gathers much of its tiny support base from some Muslim pockets. No wonder these distortions in social profile of the Congress have been accompanied by the complete political marginalisation of the party.

 

 

When we turn to class divisions at the state level, we notice a similar though not identical pattern. A first glance at Table 5 shows that the class slope at the state level is steeper than what the all-India table suggested. The pattern is more uniform here. In all but three states, the Congress vote goes up as one goes down the class hierarchy. The class division in voting is sharper in some of the more prosperous states like Delhi, Karnataka, Gujarat and Punjab, though the Congress split in Maharashtra has pushed it outside this category.

It is not an accident that the Congress rules in the top five states in this table characterized by the sharpest class cleavages. Poorer states like Orissa, Madhya Pradesh and Rajasthan have a less sharp division, but the direction of the slope is identical. Some states show a more mixed profile, either because the Congress is stronger among the better-off sections in rural areas as in Maharashtra, Andhra and Himachal or because the rump Congress has no clear profile at all, as in Bihar and West Bengal.

In three out of the 18 states analysed here, we find a reverse slope: Congress does better among the well-off and worse among the poor. In Uttar Pradesh and Tamil Nadu it is accompanied by electoral disasters for the party. Kerala is a genuine exception, like in so many other aspects. The aggressive mobilization of the poor by the Left Front has turned the Congress led coalition into a political formation of the better-off without leading to the electoral demise of the Congress. With the sole exception of Kerala then, it appears that the fortune of the Congress is linked to its capacity to mobilize the poor and the very poor.

The pattern of gender cleavages is straightforward, though the reasons underlying it are far from clear. Table 6 shows that Congress enjoys ‘gender advantage’ (that is, its vote share is higher among women than among men) in all but four states. This advantage exceeds the national average in states like Karnataka, Maharashtra, Delhi, Kerala, Madhya Pradesh and Rajasthan. The Congress won in all these states in the elections that the data is drawn from.

Then there are the other states where the Congress enjoys a more moderate advantage. Out of seven of these states, the Congress won only three. The last category is of states where the Congress gets a lower proportion of votes among women than it does among men. The reasons appear to be very different: from TDP’s and CPM’s successful political mobilization of women to the patternlessness of the Congress in Tamil Nadu and Uttar Pradesh. But the consequence is the same: Congress lost in all these states. This simple index of gender advantage turns out to be the best predictor of the political fortune of the Congress in different states.

 

III

The profile of the new Congress voter can now be easily summarized. The new Congress voter in the post-Congress polity is more likely than ever before to be a socially and economically marginalized citizen: some-one at the receiving end of the caste, communal, class and gender hierarchy. The new Congress voters are not new in the sense of coming from social groups that have never voted for the Congress. All these groups are known to be traditional Congress supporters. What is new about the profile of the Congress voter is the salience and the visibility that these groups have now come to command. An examination of the patterns of the Congress voter at the state level also alerts us to the strong relationship between the social profile of the Congress and its political fate. It does well in states where it is the party of the socially marginalized groups; wherever it has lost its grip on the marginal groups, it has been pushed to political marginalization.

This simple reading of these three variables separately raises some further questions. Is there a relationship between the first two variables? Is the apparent class slope nothing but a reflection of the caste hierarchy? What exactly is the cause-effect relationship? Does a decline of the Congress lead to distortions in its social profile or is the changing social profile the reason why the Congress is going down? How does one interpret the concentration of the Congress vote among the socially marginalized: as a cause for worry or a sign of hope? What exactly are the implications of the evidence presented above for the Congress strategy? Should it focus on regaining what it has lost or on retaining what it has today? Can it gain additional votes from these sections, or has it reached a saturation point?

 

 

Allow me to postpone some of the deeper and complex but academic questions to another occasion and say something about the last set of pressing political questions that relate to the strategy that the Congress must adopt in the short run leading to the next general elections. Theoretically speaking, the Congress has two options. The first option is to try to recreate the rainbow coalition of the days of the Congress system. This would involve working to reduce its deficit areas: upper castes, the well-off and men.

The second option is to focus on retaining and building on what it has got now. This would involve a focus on retaining and strengthening its base in its existing social constituencies: the OBCs, dalits, adivasis, Muslims, poor and women.

The option is only theoretical, for no party can develop its social profile in isolation from what others are doing. During the decade that we have examined, the BJP has developed a ‘social bloc’ of upper caste and upper class that serves as a natural base for its electoral success. The Congress can in the long run work to erode this bloc, but the option of smashing it in the next one year simply does not exist. The Congress has no option except to create a counter-bloc.

 

 

In a party system where the biggest political coalition pursues a cleavage based mobilization, its opponents have no option in the short run except to use the same cleavage to turn the tables. In other words, if the Congress wishes to survive in national politics, it must learn to effectively mobilize the marginalized and the deprived. The Congress may have reached something of a saturation point among the Muslims in a few states, but for all other groups and most of the states it is far from that. It can and must mobilize these groups more than it does today.

One last point. Mobilising the poor and the marginalized is not just a matter of declaring a grand intent. It has clear implications for the party’s political agenda, its leadership choices and its coalition strategies. Such a move is completely at odds with the more than enthusiastic support that the Congress has extended to the second generation of economic ‘reforms’ and disinvestment policies and similar policies being pursued by its own state governments. It does not go well with the old and tired upper caste leadership that still dominates the Congress organization in many parts of the country. And it demands a clear thinking about how to develop enduring coalitions with the political formations that already have mobilized these sections. A recognition of the new Congress voter requires nothing short of the reconstruction of a new Congress.

 

TABLE 1

Congress Vote by Community

Lok Sabha Elections 1991-99

Source for Table 1, 2 and 3: National Election Study (NES) 1996, NES 1998, NES 1999, all conducted by the CSDS. The figures reported here are from the post-poll surveys with a sample size of 9614 (NES 1996), 8166 (NES 1998) and 9418 (NES 1999).

Note: Figures for 1991 are based on respondents’ recall in NES 1996 and are therefore subject to greater than usual errors. The reported vote has been weighted by actual vote share to eliminate survey errors. But this does not make any difference to the pattern of sectional voting as reported here.

 

TABLE 2

Congress Vote by Class

Lok Sabha Elections 1991-99

Note: Class has been defined here with reference to occupation, house type and assets of the respondents. See appendix to Heath and Yadav (1999) for details.

 

TABLE 3

Congress Vote by Gender

Lok Sabha Elections 1991-99

Note: ‘Gender advantage’ here means the percentage point lead for the Congress among women as compared to men in terms of its reported vote share. A negative sign indicates that its vote share was higher among men. The recall of women respondents for 1991 voting was comparatively poor and may have affected the data about their voting pattern.

 

TABLE 4

Congress Vote by Community: State Assembly Elections, 1998-2003

Source for Tables 4-6: State election surveys undertaken at the time of state assembly elections by the CSDS. For Maharashtra, Karnataka and Andhra Pradesh, the data is from the question on the assembly elections in the NES 1999. The data for assembly elections held in 1998 is from exit polls with sample around 5000. All the remaining surveys were post-poll surveys conducted after polling but before counting of votes. The sample size varied between 1,000 and 2,000. The methodological and sampling details for each of the survey are available from CSDS. The reported vote has been weighted by actual vote share to eliminate survey errors. But this does not make any difference to the pattern of sectional voting as reported here.

Note for Table 4 and 5: Table entries are for percentage point deviation in each category from the mean vote share for the Congress in that state in the relevant election. For example, Congress’s overall vote share in Delhi assembly election was 48 per cent. But the survey indicates that it secured only 38 per cent votes among the upper caste. This has been reported in the table as ‘– 10’. Among dalit voters, 63 per cent reported voting for the Congress; it has been depicted here as ‘+ 15’. The same principle is followed in all the table entries.

 

TABLE 5

Congress Vote by Class: State Assembly Elections, 1998-2003

Note: Class has been defined here following the same principle as in Table 2. But the exact definition may vary from state to state depending on the availability of information and on the distribution of respondents.

 

TABLE 6

Gender Advantage for Congress in State Assembly Elections, 1998-2003

Note: As in Table 3, ‘Gender advantage’ here means the percentage point lead for the Congress among women as compared to men in terms of its reported vote share. A negative sign indicates that its vote share was higher among men.

 

 

Footnotes:

1. Pradeep K. Chhibber and John R. Petrocick, ‘Social Cleavages, Elections and the Indian Party System’ in Richard Sisson and R. Roy (eds.) Diversity and Dominance in Indian Politics, vol 1, Sage, New Delhi, 1990.

2. Anthony Heath and Yogendra Yadav, ‘The United Colours of Congress: Social Profile of Congress Voters, 1996 and 1998’, Economic and Political Weekly, 21-28 August 1999.