Are MLAs Different than the Candidates They Defeat? Evidence from the Haryana Vidhan Sabha (1991-2009) Adam Ziegfeld University of Chicago 1 October 2011 1 Postdoctoral Fellow, Department of Political Science, University of Chicago. Email: ziegfeld@uchicago.edu. The author gratefully acknowledges Jyoti, Kashmir Dhankhar, Sankar Prasad Duriya, Mohit Kumar, Jyoti Mishra, and Rohit Sharma for their invaluable contributions in the collection of the data described in this chapter. Financial support for the data collection was provided by grants from the British Academy and the John Fell OUP Research Fund at the University of Oxford. 1
Are MLAs different in their socio-demographic characteristics than the candidates whom they defeat? So far, scholarly attention has focused almost exclusively on successful MLA candidates that is, those who ultimately populate the legislature. Losing candidates have received scant attention, despite the close margins that often separate winning and losing candidates. 2 But, losing candidates matter too, for they can potentially say a great deal about the broader stock of political leaders from which legislators are drawn. This chapter examines whether winning candidates systematically differ in their socio-demographic characteristics from runners-up. It does so using data on candidates for the Haryana Vidhan Sabha in the five elections from 1991 through 2009. Comparing winning candidates to runners-up reveals few major differences. Differences evident in one election with respect to a specific socio-demographic characteristic do not, as a rule, persist over time; rather, differences are typically limited to a single election. Broadly speaking, the sociodemographic profile of the Haryana legislature therefore reflects the overall stock of viable candidates running for office. In other words, if runner-up candidates replaced winning candidates as members of the Haryana legislature, this hypothetical legislature would look very similar to chamber s actual socio-demographic profile. Data This chapter draws on a unique dataset collected in the winter of 2010-2011 by the author and a team of India-based research assistants. As part of this data collection effort, the research team gathered information on the caste, religion, place of birth and residence, occupation, political experience, and familial political connections of candidates contesting Haryana state elections between 1991 and 2009. The difference between this and many other prior data collection efforts is the focus on losing candidates. The research team attempted to 2 In Haryana Vidhan Sabha elections from 1967-2009, approximately 27.1% of the 963 constituency-level races were won by a margin of less than 5%. 2
collect data on all candidates winning more than 10% of the vote in their constituency as well as candidates winning less than 10% of the vote but who contested on the label of a major party. Major party label was defined broadly to include not only Congress and the Indian National Lok Dal (INLD), 3 but also the Bharatiya Janata Party (BJP), Bahujan Samaj Party (BSP), Haryana Janhit Congress (Bhajan Lal) (HJCBL), Haryana Vikas Party (HVP), and Janata Dal (JD). Consequently, the dataset includes information not only on winning candidates but also on a large number of losing candidates. Information was gathered through interviews with politicians and political activists in each of Haryana s 21 administrative districts. It total, approximately 165 respondents provided information. Wherever possible, multiple respondents were interviewed about the same candidates to ensure, as much as possible, the accuracy of the data. Obtaining information second-hand was necessary because systematic data are not available in any published sources on non-winning candidates. Press sources could be consulted for some candidates but would not provide information for a large number of individuals. Candidates could not, as a rule, provide information on themselves because of the time and resources that would have been required to contact and meet hundreds of candidates. Additionally, many candidates from prior elections are now deceased. Occasionally, however, information came directly from the candidates themselves. Fortunately, for much of the information collected, respondents tended to agree with one another. Agreement was particularly high on caste, religion, and relatives in politics, especially when family members had held prominent political office. Respondents were also consistent in identifying the same candidate over multiple elections. In election reports from the Election Commission of India, names are reported inconsistently. Not only are 3 Throughout, I use the INLD party name to refer not only to the INLD but also its predecessors, the Samajwadi Janata Party (SJP) in 1991 and the Samata Party (SAP) in 1996. The immediate predecessor to the INLD was the Haryana Lok Dal (Rashtriya) (HLD(R)). The HLD(R) never contested state elections, only the1998 Lok Sabha elections. The HLD(R) subsequently changed its name to the INLD, which was first used in Lok Sabha elections in 1999 and in Vidhan Sabha elections in 2000. 3
transliterations inconsistent, but the same individual s name often varies in the use of last names and in the choice of whether to abbreviate a name with initials or use the full expansion. Inconsistencies across reports make it difficult to know whether a similar name refers to the same person or different people. With the help of respondents, I was able to identify when the same individual contested multiple MLA and MP races, thereby creating a reliable measure of previous experience as an MLA or MP candidate. Respondents were also reasonably consistent in identifying candidates places of birth and residence. Information on occupation and previous experience in local politics is somewhat more prone to inconsistency or lack of information and should therefore be treated with greater caution. This chapter focuses on the 900 candidacies that represent the winners and runners-up in the five most recent state elections: 1991, 1996, 2000, 2005, and 2009. In addition, the chapter refers to the 92 successful third-place candidates, understood here as third-place candidates winning more than 20% of the vote. Table 1 presents the number of winning, runner-up, and successful third-place candidates by party. About two-thirds of the candidates come from Congress and the INLD, Haryana s two largest parties, with the HVP and BJP together accounting for about 17% of the candidates who were winners, runners-up, or successful third-place candidates. The remaining 16% are either independents or from smaller political parties. Caste in Reserved Constituencies Caste is arguably the most important characteristic of an Indian state legislator. Table 2 examines the caste of MLAs and runners up in Haryana s seventeen SC reserved constituencies. 4 Across all five elections in the reserved constituencies, the number of winners from each caste broadly matches the number of runners-up. Chamars account for the 4 Haryana has no ST reserved seats. 4
majority of winners and runners-up, distantly followed by Balmikis. Dhanaks and others comprise the rest. The precise number of winners and runners-up from each caste varies somewhat from year to year. In 1991 and 2009, a slightly larger number of Chamars won than came in second place and the number of Balmikis winning was somewhat lower than the number of Balmikis who were runners-up. But, in the intervening elections the number of Chamar winners and Chamar runners-up was almost identical. Taking the five elections from 1991-2009 together, the caste composition of the winners in reserved seats in the 1990s and 2000s is similar to the caste composition of the runners-up. The number of winners and runners-up was 52 and 45 among Chamars; fifteen and seventeen among Balmikis; seven and four among Dhanaks; and nine and thirteen among other SCs, with missing data for two winners and seven runners-up. Unless the missing data disproportionately represent one of these castes (or caste groups), then Chamars and Dhanaks are somewhat more numerous among winners than among runners-up. But, since the number of missing observations is greater among runners-up than winners (seven and two, respectively), it is dangerous to make too much out of this fact. Furthermore, because the total number of candidates is small, modest differences between the number of winners and runners-up for each caste or caste group could be attributable to only one or two candidates who contested multiple times and consistently won or lost. With those caveats in mind, it is nevertheless worth noting that the group with the biggest advantage in terms of the number of winners relative to runners-up is the numerically largest Scheduled Caste in Haryana as well as the major caste with the highest socioeconomic status. Chamars account for about half of Haryana s SCs and are more educated and more literate than other SC groups in Haryana. If the larger number of Dhanak winners to Dhanak runners-up is, in fact, indicative of a more general pattern (and not an artefact of missing data), then it cannot be explained as a function of numerical size or educational 5
advancement. Compared to Balmikis, Dhanaks are less numerous and have very similar levels of literacy and education (Census of India 2001). Well performing third-place candidates do not substantively change the picture painted by winners and runners-up. Among the twelve third-place candidates winning 20% or more of the vote on whom data are available, seven are Chamars. The remaining five are all from different castes: Balmiki, Dhanak, Khatik, Oad, and Sodha. Given the small numbers involved, these figures do not indicate noteworthy overrepresentation or underrepresentation of any group. Chamars constitute just over half of the successful third-place candidates, just as they represent just over half of the winners and runners-up. To summarize, in SC reserved constituencies the caste-wise breakdown of winners is similar to the caste-wise breakdown of both runners-up and successful third-place candidates. The main exception to this generalization is the somewhat larger number of Chamar winners than runners-up. This modest difference, driven mainly by the 1991 and 2009 elections, may well be the product of the Chamar numerical preponderance and comparatively high socioeconomic standing among SCs. If Chamar candidates are, indeed, somewhat more successful in winning seats than candidates from other castes, then this finding may suggest that caste plays an important part in electoral politics, even in reserved constituencies. Caste is generally thought to be less important in reserved constituencies because reservation severely constrains the range of castes from which parties can draw candidates and forces parties to field candidates who, because of their caste identities, might find it otherwise difficult to get elected. However, the numerical preponderance of Chamars in most constituencies in Haryana may still confer an advantage on Chamar candidates by giving them a large base of co-ethnics on which to build their candidacy. 5 5 Based on district profiles from the 2001 census, the category Chamars, etc. (which includes allied castes) is the largest SC category in all of Haryana s districts. In 2001, Chamars constituted a majority of the SC population in nine districts and more than 40% in seventeen districts. Sirsa and Fategarh were the only districts 6
Caste in General Constituencies In Haryana s 73 general constituencies, the caste breakdown of winners is very similar to the runners-up, particularly when viewed at a very high level of aggregation. Figure 1 presents the caste-wise breakdown of all winners and runners-up in elections from 1991 through 2009 in Haryana s general constituencies. Among winners and runners-up, the castewise percentages are virtually identical. In this figure, all candidates fall into one of five caste categories (in descending order of size): intermediate, upper, OBC B, OBC A, and SC. Intermediate castes are those castes that are neither considered ritually upper caste nor classified by the government as OBC or SC. The lion s share of this group consists of Jats, but also includes Rors and Bishnois. The upper caste category includes the ritually-defined upper castes Brahims, Rajputs, and Banias as well as Punjabis. Although Punjabis are not a caste and, in fact, include multiple castes, they are popularly conceived of as a community on par with caste groups. For example, when asking the caste of a candidate, respondents would repeatedly offer Punjabi as an answer even though, in the same breath, they used jati terms like Jat, Brahmin, or Yadav to describe other candidates. Since the majority of Punjabi candidates belong to the upper castes (Brahmins, Khatris, and Banias), they are included among the upper castes. The similarities between the winners and runners-up in Figure 1 would not change appreciably if Punjabis were treated as a category of their own. In Haryana, OBCs are classified into two groups. The OBC B category consists of only five castes: Gujjars, Lodhs, Meos, Sainis (or Malis) and Yadavs (or Ahirs). With the exception of Lodhs, these castes are numerically large and well represented among Haryana s political class. The OBC A category consists of a very large number of much smaller castes. Finally, where they accounted for less than 40% of the SC population. At the time of the 2001 census, Haryana had only nineteen districts. Today there are twenty-one. 7
the Scheduled Caste category consists of those classified as Scheduled Caste by the state of Haryana. The columns in Figure 2 do not sum to 100% because of missing data. Looking at individual castes and specific years does not produce a radically different picture of the caste breakdown of winners and runners-up than that which is shown in Figure 1; however, doing so yields some interesting observations. First, the figures for Bishnois highlight the need to treat these data with some measure of caution. Across the five elections, the number of Bishnoi winners is six as compared to two runners-up. All eight Bishnoi candidacies are from the same family. Four of them are Bhajan Lal, former chief minister of Haryana, all of which were winning candidacies. Two are Bhajan Lal s nephew Dura Ram, who won one election and was the runner-up in the other. The remaining two candidacies are Bhajan Lal s son Kuldeep Singh Bishnoi (who won) and Bhajan Lal s wife Jasma Devi (who did not). The larger number of Bishnois among winners than among losers is unlikely related to caste and more likely related to Bhajan Lal s status as one of the state s leading politicians. This example underscores the extent to which small differences between the number of winners and losers ought not to be given undue emphasis. Such differences can often result from only one or two individuals whose success might stem from factors entirely unrelated to their socio-demographic characteristics. A second important finding concerns the number of Jat winners and runners-up. Not only is the number of total Jat candidates strikingly similar in the aggregate across all elections, but the number of winners and runners-up is very similar in each individual election. The number of Jat winners and runners-up is identical in 1991 and 2005 and nearly identical in 2009. In 1996 and 2000, there are slight differences, but in opposite directions. In 1996, there were a few more runners-up than winners, but more winners than runners-up in 2000. As a result, the overall number of winners and losers is almost identical in the 8
aggregate: 154 winners and 152 runners-up. Given the relatively large number of candidates, this similarity is noteworthy. Third, for other castes, the number of winners and losers is not as even across years. Among the upper castes, a noticeable feature of the data is its unevenness. Over time, the total number of winners and losers for each of the upper castes is fairly similar. But, more often than not, in any given election the number of winners and losers is fairly different. For instance, the number of winners exceeds the number of runners-up by three or more for Punjabis in 1991, Banias in 1996, and Brahmins in 2005. Meanwhile, the number of runnersup exceeds the number of winners by three or more for Banias and Rajputs in 1991, Brahmins in 2000, Punjabis in 2005, and Punjabis and Rajputs in 2009. Across elections, no caste consistently produces more winners than runners-up or vice versa. There are, however, hints of election-specific patterns among the upper castes. In 1996, all of the four upper caste groups produced more winners than losers. In 2000, all four produced either identical numbers of winners and runners-up (Punjabis, Banias) or more runners-up than winners (Brahmins, Rajputs). The 2009 election was also not particularly good for the upper castes. One possibility is that year-specific patterns reflect the varying performances of political parties. The BJP fields more upper caste candidates than any of the other parties, and in 1996, the BJP had more winners than in any other election, most of which (eight of eleven) were upper caste. By contrast, in 2000, the BJP did well enough that it had a larger than normal number of runners-up, but not well enough that many of its candidates actually won. Among its runners-up, eleven of thirteen were upper caste. To some extent, variation over time may be a product of the BJP s performance. Another possibility with respect to the 2009 election is that delimitation may have had an important impact. The constituencies in use through 2005 were established in the mid- 1970s. By the mid-2000s, these constituencies exhibited considerable malapportionment. 9
Some urban constituencies had very small electorates as did some rural constituencies, while many constituencies located in newly urbanizing areas had very large populations. Not only was 2009 a poor election for upper caste candidates (Punjabis in particular), but it was also a very good election for Yadav and Gujjar candidates. Because small overrepresented urban constituencies were often heavily populated by Punjabis, expanding these urban constituencies to include more outlying semi-urban or rural areas likely had the effect of making several constituencies far more favorable to OBC candidates and less favorable to upper caste candidates. If parties did not adjust their candidate selection appropriately in 2009, then this could explain the sudden jump in the number of OBC B winners (and concomitant drop in runners-up) at the same time as the number of runner-up upper caste candidacies increased and the number of winners decreased. While delimitation may have had a negative impact on upper caste candidates and a positive impact on candidates from the large OBC castes (that is, OBC B castes), these effects were not necessarily felt in the same constituencies. Among constituencies that existed prior to and after delimitation, Narnaul was the only one that had an upper caste MLA in 2005 followed by an OBC B MLA in 2009. Rather, almost all of the seats that were won by upper caste candidates in 2005 but not 2009 were won in 2009 by candidates who were not OBC B s. Similarly, most of the seats that OBC B candidates won in 2009 but not in 2005 were not previously held by upper caste legislators. Instead, the seats were lost to or won from other castes. These other castes were mainly Jats or, because of changes in a constituency s reservation status, Scheduled Castes. This is not to say that delimitation did not have an important impact by creating more constituencies favorable to the numerically large OBC B castes and diminishing the number of constituencies in which upper caste candidates were viable. Delimitation may well have had precisely these effects, but these ideas remain hypotheses whose verification requires further research. 10
A final observation about the relative number of winners and runners-up across castes concerns the poor performance of Saini candidates. Sainis, an OBC B caste, are the only caste for which there is evidence of a decisive disadvantage. For every one Saini MLA there are three Saini runners-up. In every election, the number of runners-up is greater than the number of winners, and the runner-up candidates are distributed across the main parties. Including successful third-place candidates, there were 22 Saini candidacies in the five elections. They were defeated by candidates from a variety of castes: three OBCs, six upper castes, and six Jats. Three individuals account for half of the Saini candidacies: Hari Singh Saini in Hisar (three times a runner-up), Balbir Singh Saini in Pehowa (four times a runnerup), and Bishan Lal Saini in Jagadhri and Radaur (twice a winner, twice a runner-up, and one time a third-place candidate). Excluding these three candidates, there are still twice as many Saini runners-up as winners (as well as a third-place candidate), meaning that the losing pattern is not merely confined to these three candidates. The exact reason for the preponderance of Saini runners-up relative to winners is somewhat unclear, but perhaps due to a combination of the caste s size and location in the caste hierarchy. Sainis are not numerically dominant in any part of the state in the same way that Jats, Yadavs, and Meos constitute large pluralities or majorities in some parts of rural Haryana. At the same time, in certain areas, Sainis are concentrated in towns, which are often the political preserve of the upper castes. One plausible hypothesis is that Sainis are sufficiently numerous and socioeconomically advanced as a community that they are an important force in politics (unlike, for example, many of the smaller OBC castes), but insufficiently dominant in either rural or urban areas to do well in elections. The Representativeness of the Candidate Pool 11
Since the caste-wise breakdown of runners-up roughly matches the caste-wise breakdown of winners (albeit with a few exceptions), an important question is whether the distribution of candidates across castes approximates the population as a whole. The percentage of Scheduled Caste winners and runners-up almost exactly mirrors the SC share in the population. According to the 2001 census, 19.3% of the Haryana population belongs to the Scheduled Castes, while the percentage of winners and runners-up belonging the Scheduled Castes is 19.3% and 19.8%, respectively. Of course, the correspondence between the SC population and the number of SC winners and runners-up is mainly a function of political reservations. As Table 3 highlights, the number of SC winners and runners-up in general constituencies is negligible. 6 Within reserved constituencies, the distribution of castes among the winners and runners-up roughly corresponds to the SC composition of the Haryana population. Based on the 2001 census, Chamars account for about 51% of Haryana s SC population, Balmikis 19%, and Dhanaks, 12%, leaving the remaining 18% divided among very small castes. Among winners and runners-up in reserved constituencies, Chamars are 57%, Balmikis 19%, Dhanaks 6%, other castes 12%. Missing data constitute an additional 5%. Thus, Chamars are somewhat overrepresented in the aggregate, while Dhanaks and smaller castes are underrepresented. The underrepresentation of the smaller castes is, in some ways, unsurprising. Throughout India, viable candidates usually come from numerically large castes. However, in reserved constituencies one might expect to find a different dynamic. One possible expectation would be that candidates from very small castes would appear less threatening to non-sc voters because they lack a large caste base of their own and need to secure the support of large share of non-sc voters to win. Given the stigma that still attaches 6 The five winning or runner-up candidacies in general constituencies represent three individuals (Aman Kumar Nagra in Chachhrauli, Mani Ram in Darba Kalan, and Pawan Kumar in Naraingarh). Four of the candidacies were in constituencies in Ambala District, meaning that in the remaining portions of the state, there was only one SC winner or runner-up in five elections. 12
to SC status and arguments about non-sc voters preferring pliable, non-assertive SC candidates, one might expect that SC candidates from small castes would actually do better than candidates from larger castes because the candidates from the smaller casters are the preferred candidates of the non-sc majority. This appears not to be the case as these data suggest a correlation between caste size and overrepresentation. The largest caste is overrepresented; the second largest caste is almost perfectly represented; and then smaller castes are underrepresented. Assessing the representativeness of candidates in the general constituencies is complicated by the absence of census data on caste. Table 4 presents the caste-wise percentage of winners and runners-up aggregated over the five elections along with four estimates of the caste-wise population of the states. The first estimate comes from Joshi and Rai (2004), who report the caste-wise breakdown of the population of respondents from Haryana in the 2004 Indian National Election Study. The second and third estimates come from recent published works by M.S. Rana (2006) and S.S. Chahar (2004) that discuss the caste demography of Haryana. Whereas the CSDS estimates are generated through a stratified random sample of the population, the sources of the Rana and Chahar estimates are unclear. They likely reflect varying perspectives on the conventional wisdom. The Rana estimates at least have the virtue of summing to 100%. The Chahar estimates do not and include references to caste groups whose size is never enumerated. I therefore treat the CSDS data as the most reliable, followed by the Rana estimates, and then the Chahar estimates. But, since the various data sources included different groups in their estimates, they are all worth including. Finally, the table includes figures from the 1931 census, which was the last Indian census to include data on caste. The percentages reflect the caste population of the districts of Ambala, Gurgaon, Hisar, Karnal, and Rohtak and the princely states of Dujana, Jind, Kalsia, 13
Loharu, Nabha, and Pataudi. These administrative units closely approximate contemporary Haryana. 7 The percentages for each caste in the 1931 census are generally lower than the estimates in the other columns. This is to be expected. According to the 1941 census, the five districts that later came to comprise Haryana had a Muslim population of 27.6%. The number of Muslims in these districts in 1951 was 1,114,813 fewer than in 1941, while the number of displaced persons from Pakistan was 775,253. Since the number of outmigrants exceeded the number of inmigrants, the predominantly Hindu castes who remained in Haryana during Partition should account for a somewhat larger share of the population post-partition than they did pre-partition. With this in mind, the 1931 figures corroborate to a large extent the figures in the other columns. Interestingly, comparing the percentage of winners and runners-up from each caste to the caste estimates reveals only a modest overrepresentation of Jats. Jats comprise just about a third of the winning and runner-up candidates. According to the four estimates in Table 4, the Jat population ranges between 21% and 35%. The CSDS figure is 29.5%, but a large share of the 5.8% Sikh population is also Jat, suggesting a figure somewhat above 30%. News reports typically offer a somewhat smaller range of estimates, with some putting the population between 20-25% (PTI 2009, Ramachandran 2009) and others suggesting that Jats constitute between 25-30% of the population (PTI 2005, Joy and Shankaran 2009, ET Bureau 2010). As with the Rana and Chahar estimates, the origin of these figures is opaque, and it is unclear whether they include Jat Sikhs as well as Hindu Jats. Only the Chahar estimate of 35% of the population would indicate an underrepresentation of Jats among winners and runners-up. All other estimates indicate varying degrees of overrepresentation. However, considering that Haryana s politics is very commonly referred to as being heavily dominated by Jats, the overrepresentation of Jats is, based on most estimates, fairly modest. Taking the 7 Parts of Ambala District and the states of Jind, Kalsia, and Nahba are located in contemporary Punjab. Part of the state of Patiala was located in present day Haryana, but it is excluded from these calculations. 14
CSDS estimates (which are consistent with many estimates in the press), the overrepresentation of Jats is modest or non-existent. The number of Jat winners and runnersup is only 3-4% more than the population would warrant if one does not consider the Jat Sikh population. Of course, in general constituencies Jat winners and runners-up constitute a much larger share of the candidates. Since a number of SC reserved constituencies are located in heavily Jat districts, were in not for SC reservation, the share of Jat winners and runners-up could well reach 40% or more. In this way, SC reservation not only ensures that SCs are represented but also prevents the overrepresentation of other groups. Most of the large OBC B castes also appear modestly overrepresented among the ranks of winners and runners-up. The CSDS data club together Yadavs and Gujjars, and that estimate is fairly close to those offered by Rana and Chahar. Relative to these figures, Yadavs and Gujjars are either somewhat overrepresented or perfectly represented among winners and slightly to somewhat underrepresented among the ranks of the runners-up. Sainis and Meos are also represented reasonably well as compared to their shares of the population. Meos may actually be somewhat overrepresented. The CSDS data do not distinguish between Meos and other Muslims castes. But, back of the envelope calculations suggest that 2.5%-3.5% of the population is Meo. According to the 2001 census, Haryana s population was 5.8% Muslim. Of that, approximately 70% are in Gurgaon and Faridabad districts, which are overwhelming Meo. If Meos are between 60% and 90% of the Muslim population of the districts, then 2.5%-3.5% of the state population ought to be Meo. Based on this estimate, the Meo percentage of winners and runners-up is either nearly representative or a modest overrepresentation. Turning to the upper castes, the degree of overrepresentation is potentially greater, at least relative to the CSDS and Rana estimates. The CSDS data put the upper caste figure at 19.3%, or about 5-6% lower than the share of winners and runners-up. The Rana estimates 15
include a large Other category and does not include figures for Rajputs and Banias, so it is unclear what his full estimate of upper castes is. By contrast, the Chahar estimate which does not include precise figures for Brahmins and Banias would suggest the possible underrepresentation of the upper castes, though it ought to be recalled that the Chahar figures sum to more than 100% and therefore must overestimate the size of certain castes. The 1931 census data also suggest the possibility of upper caste underrepresentation. Based on these data, the upper castes were about 15.3% of the population. In addition, according to the 1951 census, 13.8% of the population consisted of displaced persons, most of whom would today constitute the Punjabi population. Together, the population of Punjabis and other upper castes comes close to 30%. If this is accurate, then the upper castes are slightly underrepresented. Though the precise judgments are difficult to establish, Jats, most of the OBC B castes, and the upper castes appear to be represented in rough proportion to their share of the population or somewhat overrepresented. One unambiguous conclusion is that the OBC A castes are underrepresented. Their share among winners and runners-up is meagre, even though the CSDS data and Rana estimates suggest that they constitute at least 12-13% of the population. The underrepresentation of the OBC A castes is unsurprising. The category includes a large number of castes that are, for the most part, quite small. In any constituency, an individual OBC A caste is unlikely to constitute a large share of the electorate. Consequently, the probability that the major parties will field a candidate from one of the OBC A castes should be low. Finally, turning to successful third-place candidates, the figures for caste in Table 6 indicate a high degree of representativeness, especially among Jats and upper castes. Most of the underrepresentation of the OBC A castes benefits the OBC B castes. It total, considering the historically dominant position of the upper castes in north India and the current characterizations of Haryana s politics as Jat-dominated, the pool of successful 16
candidates looks very much like the population of the state as a whole. Of course, this statement comes with two important caveats. The first is that the OBC A castes are substantially underrepresented, and the second is that the Scheduled Castes are represented in proportion to their population thanks to reservation. If not for reservation, then both the OBC A castes and Scheduled Castes would likely be highly underrepresented. Other Characteristics With regard to characteristics other than caste, winners and runners-up are sometimes quite similar and sometimes not. As Table 5 shows, for several characteristics, no consistent pattern emerges. With religion, for example, the number of winners and runners-up from a religious group often differ but not in a consistent direction. In 1996, 2000, and 2005, there were more Hindu winners than runners-up, but more Hindu runners-up than winners in 1991 and 2009. Similarly for Muslims and Sikhs, the ratio of winners to runners-up varies from election to election, sometimes favoring the winners (Muslims in 1991 and 2009; Sikhs in 1991), sometimes the runners-up (Muslims in 1996 and 2005; Sikhs in 1996 and 2000), and sometimes identical (Muslims in 2000; Sikhs in 2005 and 2009). For occupation, and gender, the comparison of winners and runners-up is also inconsistent across years. The occupation variables in Table 5 are coded in the following way. Professionals are those candidates who are doctors, lawyers, engineers, teachers, professors, or high-level civil servants. Businesspeople are those engaged in business of any form. Many of those coded as professionals and businesspeople are also described by respondents as having some ties to agriculture. Many professionals and businesspeople have agricultural plots inherited from their families but do not derive most of their income from or invest most of their time in agriculture. For this reason, the agriculture variable only codes as an agriculturalist those candidates for whom agriculture is described by respondents as their 17
sole profession. Since many candidates are listed as having multiple occupations, and there is a fair amount of inconsistency across respondents, this set of variables is likely to have the highest degree of measurement error, and the figures should therefore be treated with caution. Broadly speaking, the occupational figures are fairly similar for winners and runners up. Notable exceptions are the low number of professional winners and high number of professional runners-up in 1996 and the large number of businesspeople runners-up in 2009 as compared to winners. Otherwise, occupational differences between winners and runnersup are modest, with no clear patterns across elections. For instance, the number of professionals is higher among winners than runners-up in 1991, 2005, and 2009 but lower in 1996 and 2000. For businesspeople, the number of winners exceeds runners-up in 1991, 1996, and 2000, but is lower in 2005 and 2009. The figures for agriculturalists are similarly inconsistent across elections in terms of the ratio of winners to runners-up. Gender too exhibits inconsistent patterns over time. The number of female winners is greater than the number of female runners-up in 1991 and 2005 but lower in 1996, 2000, and 2009. One noteworthy feature of the data is the larger number of female candidates, both winners and runners-up, in 2005 and 2009. Future elections will determine whether this represents a long-term increase in female candidates. Potential longitudinal trends aside, the data indicate no systematic pattern in whether women tend more frequently to be winners or runners-up. With respect to place of origin, the most obvious feature of the data is the large majority of candidates who are, in one way or another, natives of the areas in which they contest. The rows listed under place of origin in Table 5 represent the number of candidates who are, first, natives of the assembly constituency in which they contest and, second, natives of the district in which they contest. The number of district natives is quite large, and the figures are fairly similar across winners and runners-up. The 2005 election is the only one in 18
which runner-up district natives outnumber winning district natives. The number of constituency native winners is greater than the number of constituency native runners-up in three of the five elections. Perhaps the broader point is that the overwhelming majority of successful candidates are from the districts in which they contest, and a healthy majority of winners and runners-up are also constituency natives. The same is true of successful thirdplace candidates. The characteristics on which differences between winners and runners-up are consistently evident are those related to political experience and relatives in politics. Candidates political experience is coded in two ways. The first is whether the candidate held local office prior to contesting an MLA election. The second is whether the candidate was previously an MLA or MP before the election year in question. To be clear, this variable is not a measure of incumbency, only an indicator of prior legislative experience. Interestingly, these two measures of experience present very different pictures with respect to winners and losers. Overall, a comparatively small share of winning and runner-up candidates has had prior experience holding local office. The numbers appear to increase somewhat over time, which is not surprising since the advent of panchayati raj in the 1990s should have increased the opportunities to hold local elected office. Perhaps even more interestingly, a greater share of runners-up has experience in local elected office than winners. In some ways, this finding is surprising as legislative experience is typically viewed as a positive candidate attribute, and one would expect that winning candidates would be more likely to have such characteristics relative to those candidates who fail to win election. One potential explanation for the relative preponderance of runners-up with local experience is that high quality candidates may entirely bypass local elections and begin their political careers in state-level elections. If so, since high quality candidates should also be more likely to win elections, the number of 19
winning candidates with experience in local elected office should be somewhat lower as compared to runners-up. In contrast to local office, winning candidates are more likely to have previously been an MP or MLA. In 1991 and 2005, the differences are particularly large, and in 2000 and 2009, they are fairly small. The 1996 election represents an important exception as the number of runners-up with prior experience as an MLA or MP is significantly higher than the number of winners. This variation is, for the most part, correlated with the partisan breakdown of the legislature. The 1991 and 2005 elections, when winners were much more likely than runners-up to be former MLAs or MPs, were the elections that produced the largest number of Congress legislators, and 1996 was the election in which the smallest number of Congress legislators won. In general, Congress tends to nominate a large number of candidates with previous experience in either the Haryana Vidhan Sabha or the Lok Sabha. The number of Congress candidates who were former MPs or MLA was 43, 59, 49, 50, and 66 in the 1991, 1996, 2000, 2005, and 2009 elections, respectively. In 1996, despite the large number of former MLAs and MPs nominated by Congress, only 32 Congress candidates managed to either win or come in second in the election. Instead, a particularly large number of BJP and HVP candidates won. The BJP has historically won few MLA seats, and the HVP was a relatively new party in 1996, so these two parties nominated few candidates with prior state- or national-level legislative experience. The advantage that winners with experience as MPs and MLAs seem to enjoy may actually be more a function of which parties win the most seats than a preference among voters for candidates with experience. When a party has been successful in past elections, it has more former MPs and MLAs that it can field in subsequent elections. Finally, the number of winning candidates with relatives that preceded them in politics is consistently higher than the number of runner-up candidates with relatives in 20
politics. As Table 5 shows, the indicator of family in politics is decomposed into three subindicators, measuring whether a candidate had family in politics 1) at the local level or who contested but failed to win state-level office, 2) who won state- or national-level elections, and 3) who constitute one of Haryana s major political dynasties. The dynasties include the families of former chief ministers Rao Birender Singh, Devi Lal, Bansi Lal, Bhajan Lal, and the current chief minister Bhupinder Singh Hooda. The number of winners with family in local politics or who failed in state politics is not consistently greater than the number of runners-up with family of similar experience. But, the number of candidates with relatives in state or national politics is consistently higher among winners than losers. The distinction between winners and runners-up is, surprisingly, less stark among candidates from the major dynasties. Overall, however, winners are more likely to have family in politics than runnersup. Summary The data described in this chapter on the characteristics of winners and runners-up in Haryana state elections show that, on the whole, runner-up candidates look very much like the candidates who actually win. Even among successful third-place candidates (Table 6), the percentage of candidates with a particular characteristic over the five elections is not so different from the percentage of winners or runners-up with that characteristic in any of the individual elections. The caste breakdown of winners and runners-up is very similar. When differences between winners and runners-up emerge in specific elections, they tend not to indicate broader patterns. More often, winners from a caste are more numerous than runners-up in some elections, but in other years the reverse is true. Sainis are the only caste for which runners-up always outnumber winners. Otherwise, replacing winners with runners-up in the 21
Haryana legislature would result in a legislature that is very similar in its caste breakdown to the actual legislature. Interestingly as well, the pool of candidates is broadly representative of Haryana s population. The only major exception is the OBC A castes who are noticeably underrepresented. Aside from caste, on most other characteristics the pool of runner-up candidates looks similar to the winners. And, as with caste, where differences emerge, they are not consistent across time. Two exceptions in this regard are political experience and family in politics. As a group, runners-up more often have experience holding local office than winners, but the reverse is true for state- and national-level office. However, differences between winners and runners-up with respect to state- and national-level office may well be driven by which parties win elections. When previously successful parties win elections, winning candidates are more likely to have earlier been MLAs or MPs. When newly successful parties win elections, their candidates are unlikely to have been MLAs or MPs before. The other difference between runners-up and winners comes in the realm of family in politics; winning candidates more often have family connections in politics than runners-up. Implications What light can these data potentially shed on the electoral process in India and on the importance of candidates characteristics in shaping election outcomes? There are four distinct possibilities about how candidates characteristics can affect election outcomes. First, candidate characteristics might not matter at all. If voters are highly partisan, then they may be indifferent to candidates characteristics. If this is the case, then winners and runnersup should be very similar because neither voters nor parties distinguish candidates on the basis of these characteristics. The number of winners and runners-up with a particular characteristic should simply reflect the broader pool of candidates (which may or may not be 22
representative of the wider electorate). Second, candidate characteristics might matter a great deal to voters. In anticipation of candidate characteristics mattering, parties might all select candidates with similar characteristics so as to maximize their chances of winning and not put themselves at a disadvantage relative to other parties candidates. As a result, the entire pool of candidates should look very similar in their characteristics and differences between winners and runners-up should be minimal. Third, candidate characteristics per se might not matter, but they might be correlated with characteristics that do matter, some of which may be easy to measure, others of which might not. If so, then winning candidates should more often exhibit certain (positive) characteristics as compared to runners-up. Fourth, candidate characteristics per se might matter, in which case candidates with certain characteristics should tend to do better in elections than those without those characteristics. Consequently, winners who, by definition, fare better in elections than runners-up should differ from runners-up in their characteristics. Since possibilities one and two and possibilities three and four are observationally equivalent, the data presented in this chapter cannot distinguish between them; however, the data are capable of indicating which of the two sets of possibilities is more likely. As the chapter makes clear, on occupation, gender, place of origin, and (for the most part) caste, winners and runners-up tend to be very similar. From this, one can infer that either these characteristics do not matter much to voters or that they matter to voters but parties do their best to neutralize the importance of these characteristics by fielding similar candidates. Since there is relatively little evidence to suggest that voters place great weight on a candidate s gender or occupation, the first possibility that characteristics do not matter much might apply. By contrast, caste is widely thought to matter; it is more likely a characteristic for which parties attempt to field similar candidates in an effort to avoid putting themselves at a competitive disadvantage. Finally, given the large number of candidates who are natives of 23
the constituency in which they contest, it seems plausible that there is a strong norm in favor of fielding local candidates on the presumption that local candidates enjoy some advantage over candidates from farther afield. By contrast, political experience and family in politics are characteristics on which candidates differ more systematically. As compared to runners-up, winners tend more often to be former MLAs and MPs and to have relatives in politics. But they are less often former elected officials at the local level. Since experience in local-level government is unlikely to be viewed as an actively negative attribute of a candidate, this characteristic is likely correlated with some other characteristic that matters. As noted above, candidates who begin their careers in local government may be of generally lower quality or have fewer ties to important politicians within their parties, which prevents them from beginning their political careers competing for higher level office. State- and national-level political experience and family in politics may matter per se, as voters might value prior experience. Or, these characteristics may matter because they are indicators of other characteristics, such as proximity to power or the likelihood of access to patronage resources. Whether these characteristics matter per se or because they are strong indicators of something else that matters to voters, one question is why parties do not field candidates accordingly. If parties can make a habit of fielding local candidates and can, at the constituency-level, frequently field candidates belonging to the largest caste so that voters cannot discriminate between candidates on the basis of these two characteristics, why can they not nominate candidates with political experience and family in politics so as to potentially neutralize the impact of these characteristics as well? Notably, what distinguishes these characteristics from the others described in this chapter is that they depend on a party s history. Assuming that a candidate and her family remain within a party (which is not necessarily a straightforward assumption), then that candidate can only have prior experience 24