1An extensive literature in comparative politics in the West has looked at the relationship between intra-party centralization within a political party and the impact it has on the overall party system in the country (Schakel 2013). India, with its large-sized national parties, offers a rich case study to investigate how changes within a single political party shape the overall party system. The comparative politics scholarship also acknowledges that a very centralized political federation can have very decentralized political parties (Ware 1996). The Congress Party under Nehru epitomizes the above political context, where a relatively centralized federation had a genuinely decentralized Congress Party (Kothari 1964; Weiner 1967). The decentralized organization of the Congress allowed the party to manage internal conflict and control regional political defections. However, this decentralized organization largely excluded lower castes and other politically marginalized groups from political power (Wilkinson 2000:700). Echoing Wilkinson (2000), India remained a “ranked” society where political power was concentrated mainly in the upper caste elites, both at the national and regional levels. We see the domination of upper caste elites gradually decline only with the growth of new ethnic parties. This process coincided with growing intra-party centralization within the Congress Party under Indira Gandhi, who took power away from regionally dominant political elites.
2Many studies have drawn a parallel between the Bharatiya Janata Party (BJP) under Prime Minister Modi and the Congress Party under Indira Gandhi (Aiyar 2019; Maiorano and Sen 2021; Nakamizo 2020). We have also witnessed a trend towards greater political and economic centralization in India since Prime Minister Modi entered office in 2014 (Aiyar and Tillin 2020; Sharma and Swenden 2022). The growing political and economic centralization that has ensued has resulted in the undermining of democratic institutions through repression, erosion, and penetration (Manor 2021). It was done by keeping the formal democratic institutions intact but decisively tilting the playing field in favor of the incumbent (Maiorano and Sen 2021). Second, the prime minister was credited with welfarism rather than the government or party (Sircar 2020). Third, “chief ministers-including and especially BJP ones-who built their reputation on the delivery of public goods and welfare, would have to find alternative sources of political legitimacy if over-centralization of social policy implementation continue[d]” (Maiorano and Sen 2021:5). The lack of an avenue for chief ministers to rely on the claim for welfarism might push them to resort to ethnic identity as a way to create their independent base of support (Maiorano and Sen 2021); Yogi Adityanath and Hemant Vishwa Sharma, the chief ministers of Uttar Pradesh and Assam, have adopted such a path.
3The above arguments highlight how political centralization within a party negatively impacts the overall party system, from undermining democratic institutions to promoting ethnic outbidding as a form of political competition. While the negative consequence of political centralization is obvious to those of us who study India, there exists a vein of scholarship within the West that argues that centralization within a political party can help promote descriptive representation of historically marginalized groups (Hazan and Rahat 2010; Matland and Studlar 2004; Sobolewska 2013). In the wake of the abrogation of Article 370, even some political leaders like Amit Shah underlined the importance of political centralization in improving representation for India’s minorities. Therefore, even when the claims sound strange to political scientists working on India, it deserves an empirical test.
4In this paper, we test this alternative hypothesis that centralization within a political party can help promote descriptive representation of historically marginalized groups. For example, some recent work in the British context highlights how a centralized party increased the representation of such groups by sidelining the domination of the local selection committee, which is often made up of locally dominant groups. By parachuting historically marginalized candidates in predominantly white safe seats, a centralized Conservative Party was indeed able to select and nominate a large number of ethnic minority and women candidates.
5In the Indian context, we know that the BJP mobilized a diverse set of castes by co-opting local candidates of different caste groups on a constituency-by-constituency basis (Singh and Goel 2019). In this article, we test whether the centralization of the BJP helped the party increase the representation of marginalized social groups. This could be plausible as centralization within the party would have reduced the influence of local office bearers in the candidate selection, whose informal networks could be considered a barrier to the representation of lower castes in the party (Singh 2023). Has centralization actually helped the party overcome such barriers?
6In order to answer this question, we first provide empirical evidence to show that the BJP has become more centralized in the post-2014 era. Second, we build on the comparative politics literature about the implications of party centralization to test whether the BJP’s centralization in India has increased the representation of marginalized groups in politics in the post-2014 era. Here, we provide evidence to show that centralization within the party has not led to improvements in the nomination of women and historically marginalized groups in the Indian national parliament. With this evidence in mind, we revisit the comparative politics literature on the effects of party centralization on the representation of marginalized groups. With the case of the BJP, we highlight how centralization within a party does not always translate to significantly greater representation of historically marginalized groups.
7However, even though, in 2019 compared to 2014, centralization within the party did not contribute to overall improvement in terms of representation of marginalized groups, there were some individual instances where the party did replace male candidates with women candidates and/or upper-caste candidates with lower-caste candidates. In the latter part of the paper, we look at the consequences of these individual changes within the party on the local party system. We rely on aggregate measures of party system change to test if the representation of more women and historically marginalized groups (albeit few) in one major party has had any impact on turnout, party competition, and electoral volatility at the constituency level.
8This paper proceeds as follows: in the first part, we discuss how we measure party centralization in the Indian case and provide some evidence to show increasing centralization within the BJP in the post-2014 period. In the second part, we discuss the implications of party centralization on the representation of marginalized groups in the BJP. And finally, in the last part, we provide some suggestive evidence on the implications of party centralization on the local party system.
9Measuring centralization within political parties in India is a difficult task. This is because political parties in India are relatively weakly institutionalized, with an opaque and non-transparent working culture. The candidate selection procedure is hierarchically driven by patronage networks. The involvement of a host of intermediary networks in the candidate selection process blurs the boundaries between formal party channels and informal networks of party functioning, making it methodologically difficult to measure and codify (Singh 2023).
10While most political parties in India have internal party constitutions, a mere reading of the constitutions will not be enough to determine the internal over-time variation in the level of party centralization within political parties in India. Farooqui and Sridharan (2014) argue that most political parties in India are relatively centralized compared to their counterparts in Europe and the United States. Relying on interview data, they classify the Congress Party, the BJP, and the Communist Party of India (Marxist)—CPI(M), as highly centralized. This is because for all these parties, the subnational party organs merely propose names of potential candidates, but the national central leadership makes the final decision.
11In this paper, we build on Farooqui and Sridharan’s (2014) work to measure party centralization by looking at candidate selection. As underlined by Lundell (2004) and Gallagher and Marsh (1988), candidate selection remains the most important role of a political party. Here, we specifically focus on the renomination of candidates, and in particular, incumbents. It is much more difficult to deny renomination to a sitting Member of Parliament (MP) as opposed to a candidate who lost an election. Hence, we use this as a proxy to measure party centralization. The lower the renomination rate of incumbent MPs, the more centralized the party is. Our assumption is that the incumbent MP would have developed their patronage network in their constituency and outside the constituency amongst voters of their caste and network, which adds to the difficulty in denying nomination to them. The incumbent MPs usually hold good positions in the party organization or have a good network of supporters in the party organization, which makes it difficult for parties to deny nomination to them (Jensenius 2016). They also bring in cash for party organization. We therefore believe that denied renomination is a conservative proxy to measure party centralization within parties in India.
12While the Indian political parties are relatively centralized compared to their western counterparts, some scholars have noted increasing party centralization within the BJP over time, especially since 2014 (Manor 2021). Do we observe this in the BJP’s renomination patterns of sitting MPs?
- 1 We used the stringdist package in R developed by Van der Loo (2014) to name-match candidates in ord (...)
13We use data from 1984 up until 2019 to show the proportion of MPs renominated by the BJP.1
Figure 1
Proportion of MPs being renominated by the BJP over time
- 2 We do not have data on the proportion of BJP MPs being renominated in 2009 due to the delimitation (...)
- 3 The proportion of MPs renominated by the BJP in 2024 also sits at 57 % (Rohmetra 2024).
14In Figure 1, the x-axis indicates the election year,2 whereas the y-axis (left-hand) indicates the total number of BJP MPs in a given election year and the total number of MPs renominated by the BJP from that election year. The orange and red bars thus reflect the total number of BJP MPs in that election year and the absolute number of BJP MPs who are renominated for the following election year, respectively. The y-axis (right-hand) indicates the proportion of MPs being renominated, which is also listed as a proportion for every single election year. In 1989, the BJP renominated both its incumbent MPs from 1984, Chandupatla Janga Reddy from the Hanamkonda seat in Andhra Pradesh and A.K. Patel from Mehsana. In Figure 1, we can also see that the proportion of MPs being renominated by the party has significantly declined since the 1990s, correlating with the party’s strong performance on the national stage, the only exceptions being 1998 and 1999. General elections in 1998 and 1999 were held in close proximity to the previous general elections and presumably reflect higher rerunning rates amongst most candidates across all political parties (Lee 2020). What is notable is the drop in the number of MPs being renominated by the party in the 2019 general elections compared to the 1990s. The party renominated only close to 59 % of its MPs from 2014 for the 2019 general election, which is the lowest proportion of MPs the party has ever renominated.3 The lower proportion of MPs renominated by the party in the 2019 general elections reflects the growing centralization within the party and reconfirms what many political commentators in India have been observing about the BJP in recent years. Of course, the increasing centralization within the BJP should not be seen as the sole cause of the declining renomination of its sitting incumbents. Broader trends—such as the gradual shift in the social composition of India’s political class, with upper-caste MPs increasingly displaced by candidates from Other Backward Classes (OBCs) and lower castes—are also at play (Jaffrelot and Kumar 2009). The BJP is very much a part of this wider transformation of the Indian party system. Nonetheless, the party’s declining renomination of incumbents still reflects the effects of growing intra-party centralization, even if they are shaped by broader structural and electoral dynamics.
15The BJP, which used to take pride in its collective leadership and its consultative process in decision-making, is following a path of centralization of power around Modi (Ashutosh 20234): “The decisions are rumored to be made in the PMO and the party nods in agreement. Young leaders who are unlikely to question this top-down approach are promoted and senior leaders within the party are sidelined.” Modi has overshadowed strong state leaders within the BJP such as Shivraj Singh Chouhan, Vasundhara Raje, Raman Singh, and B. S. Yedurappa. The cabinet ministers have also become figureheads due to the increasing role of the PMO in major decision-making processes (Imtiaz 2025). The form of political centralization observed in Modi’s BJP is the antithesis of a model in which citizens directly hold their representatives accountable for the quality of their economic performance (Sircar 2022). The party has built a direct connection between Modi and the voters, which has minimized the impact of individual candidates on the party’s appeal and electoral outcomes.
16In the next section, we further explore the implications of increasing party centralization within the BJP in the post-2014 era.
17In the comparative scholarship on party organization, there is a growing consensus that party centralization can be used to positively influence the diversity of representation at the local level (Hazan and Rahat 2010; Matland and Studlar 2004). Existing scholarship on party organization highlights how British political parties are relatively decentralized when it comes to candidate selection (Krouwel 1999:11). Sobolewska (2013) shows how centralization of the candidate selection process helped ensure wider representation of ethnic minority candidates in the 2010 general election in the UK. In particular, she shows how national political parties, especially the Conservative Party, were able to select ethnic minority candidates in “safe” predominantly white seats. This was only possible because of centralizing the candidate selection process within the party.
18Centralization increases the influence of national leaders in candidate selection, as opposed to the weight of the regional and/or local leaders (Hazan and Rahat 2010). A centralized candidate selection method allows national leaders of the party to respond to the growing pressure for increasing diversity of representation (Matland and Studlar 1996). The national leaders do that by introducing internal quotas in the list of party candidates for specific social groups and/or reserving constituencies for social groups (Hazan and Rahat 2010). They also scale down the local selectors’ influence (Sobolewska 2013), whose informal network prevents the nomination of women and marginalized social groups (Cheng and Tavits 2011; Kenny 2013; Singh 2023).
19Several scholars have observed a similar process within the BJP in the post-2014 period, that is, the party has become overcentralized in its functioning (Ashutosh 2023; Manor 2016; Sircar 2022). The process of centralization has increased the control of national leaders over candidate selection. It has happened in three ways. Firstly, Modi, who runs a presidential-style election campaign, has brought in private agencies to mobilize votes (Jha 2017; Kumar 2024). These agencies run campaigns against the opposition parties/candidates alongside the BJP’s party organization. They underline how an ideal incumbent should behave and invariably raise the expectations of voters about what a local incumbent can deliver. This puts immense pressure on the incumbent and may decrease their probability of rerunning and winning the election (Banerjee et al. 2013). This should then create an opportunity for new candidates to contest elections. Several political commentators have argued that the party prefers new candidates from traditionally marginalized castes because the pre-election campaign not only generates awareness amongst them but also the central party organization is happy with them as they can control and dictate to them easily, as opposed to candidates from locally dominant castes (Jha 2017).
20Thirdly, the party has increased the appointment of multiple in-charges and deputy in-charges at the state and sub-regional level, who remained involved in the candidate selection process of the party at various levels (Kumar 2024). Their involvement at the local and intermediate levels in the candidate selection process has the potential to remove the obstacle of local leadership of the party.
21Despite the ongoing political commentary on how the centralization within the party pushed for more candidates from the historically marginalized groups, there is very little empirical work to confirm it.
22To empirically test the above hypothesis, we begin by looking at what sort of candidates the party replaced its incumbent MPs with. The party did not renominate 110 of its sitting MPs from the 2014 general elections but recontested in the 2019 general elections in those seats. Did they replace the sitting MPs by nominating more marginalized communities? We know that women have been under-represented in the Indian parliament proportional to their share of the population (Jensenius 2016; Spary 2020). Did the BJP replace more men MPs with more women candidates?
Table 1
|
Year
|
Gender
|
Total
|
|
Women
|
Men
|
|
|
2014
|
14 (12.7 %)
|
96 (87.3 %)
|
110 (100 %)
|
|
2019
|
19 (17.3 %)
|
91 (82.7 %)
|
110 (100 %)
|
Gender and party nomination in 2014 and 2019 general elections
23Out of the 110 MPs who were denied renomination, there were 14 women MPs and 96 male MPs. The party replaced the incumbents with a greater proportion of women candidates. It replaced 5 sitting male MPs with women candidates. While we cannot statistically say that the party was more likely to replace male MPs with women candidates, due to the generally lower level of women MPs and candidates, we do see that the party made an active effort to nominate more women candidates in the 2019 general elections.
- 5 We will make the data publicly available for colleagues to check and cross verify it with their kno (...)
24Next, we look at the candidates’ caste. We begin by coding the jati of all the 110 sitting BJP MPs who were replaced in the 2019 general elections. We have coded the jati of all new candidates that were nominated instead of the sitting MPs. We did this by conducting telephonic interviews of regional experts who are primarily researchers based at local universities and/or anti-caste activists. We have not relied on the information provided by any single informant but cross-checked it with a wide array of researchers and activists to eliminate the possibility of error. Our information is based on a kind of elite interview, which is well accepted in the field of study of the caste of candidates (Jaffrelot and Verniers 2012; Kumar 2024).5
25We classify the jatis into three distinct caste categories—upper caste, backward caste, and Scheduled Caste/Tribe. There was no candidate from the Christian and Muslim communities, hence we did not face any problems incorporating the caste of religious minorities in India. Importantly, while classifying the castes, we have broadly used the official classification of castes used at the state level. If a caste X is a backward caste in a state, but an upper caste in a different state, we have classified it as a backward caste in the former and an upper caste in the latter.
26In Table 2 below, we show the caste make-up of the incumbent MPs that were denied renomination in 2019 and the caste make-up of the candidates that were given nominations instead of the incumbents in 2019.
Table 2
|
Upper caste
|
Backward caste
|
Scheduled Caste/Tribe
|
Total
|
|
2014
|
47
|
27
|
36
|
110
|
|
2019
|
47
|
26
|
37
|
110
|
Caste make-up of the replaced BJP MPs in 2014 and new candidates in 2019
27In Table 2 above, we can see that the majority of the BJP’s candidates came from the upper caste communities, followed by the Scheduled Caste/Tribe communities, and finally from the backward caste communities. We can also see that the party tried to stabilize the caste equation by replacing incumbent MPs with new candidates of the same caste group. In 2019, the number of backward caste candidates dropped by one and Scheduled Caste/Tribe candidates increased by one.
28While Table 1 and Table 2 highlight the overall patterns of candidate replacement within the BJP in terms of gender and caste, in Table 3, we look at the combination of gender and caste when it comes to individual candidate replacements.
Table 3
|
Lower caste to upper caste
|
No change
|
Upper caste to lower caste
|
|
Female to male
|
0
|
12
|
0
|
|
No change
|
5
|
68
|
8
|
|
Male to female
|
2
|
14
|
1
|
Gender and caste makeup of the replaced BJP candidate from 2014 to 2019
- 6 In line with the contemporary consensus on the normative and pragmatic advantages of diversity, we (...)
29In Table 3 above, the cell with the most positive6 change would be the one where a male MP changed to a female candidate and an upper-caste MP to a lower-caste candidate. This cell contains only a single observation regarding Kodarma constituency in Jharkhand, where the BJP replaced its upper caste (Bhumihar) male MP Ravindra Ray with a Yadav female candidate, Annapurna Devi. In most other cases, a positive change in one direction came with no change or a negative change in the other direction. For example, 8 upper caste male MPs were replaced with 8 lower caste male candidates. On the other hand, 16 Male MPs were replaced with a similar caste female candidate and in two cases, the male lower caste MP were replaced with 2 female upper caste candidates.
- 7 The BJP mobilized Jats in Haryana through the induction of powerful Jat leaders such as Chaudhary B (...)
- 8 The BJP gained power in Maharashtra following tactics similar to those used in Haryana, where it sl (...)
30Overall, whether considered independently or together, changes in caste and gender representation do not suggest a meaningful shift away from dominant social groups—particularly men and upper castes—in the BJP’s candidate selection between 2014 and 2019. One possible explanation for this continuity is the BJP’s concern with alienating its core support base. In constituencies where the party has taken even modest steps to enhance the descriptive representation of historically marginalized groups, it has often done so at the potential cost of its traditional upper-caste support. Indeed, early signs of backlash from dominant caste groups, such as Jats7 in Haryana and Marathas8 in Maharashtra, indicate the political risks of significantly altering the party’s social composition. This may help explain the BJP’s reluctance to adopt a more transformative approach in candidate selection, particularly with regard to increasing the representation of lower-caste candidates (Mitra 2024).
31In this section, we test the impact of the BJP’s growing centralization on the local party system at the constituency level. Following Jensenius and Suryanarayan (2022) we define a party system as an enduring pattern of electoral competition between parties/candidates for public office. Party systems are formed based on candidates’ and voters’ incentives to coordinate around common party labels. A local constituency-level party system will be the nature of electoral competition at the constituency level. We look at different characteristics of a party system at the constituency level from turnout change, electoral volatility and party-system fragmentation change. While a party system is formed of all the parties and candidates’ competing in the system, changes within a major party can affect and shape the whole party system (Ware 2007).
32First, in the constituencies where the BJP renominated its incumbent MPs we should see less change in voter turnout, lower electoral volatility, and reduced party system fragmentation, compared to constituencies where the party replaced its sitting MPs. This expectation stems from the idea that replacing incumbents may trigger defections, the entry of new parties, or independent bids, thereby increasing political competition and mobilizing new voters. Second, we examine whether changes in the caste or gender of BJP candidates—particularly when such changes improve descriptive representation (e.g., replacing upper-caste or male candidates with lower-caste or female candidates)—lead to measurable shifts in turnout, electoral volatility, and party system fragmentation. This line of inquiry tests whether symbolic descriptive inclusion of historically marginalized groups can mobilize marginalized voters and reshape local party competition. These hypotheses are important to test because they speak to broader questions about how party organizational decisions and representational strategies influence democratic participation and party system stability in competitive electoral contexts. Understanding these dynamics helps assess whether centralized party control fosters or hinders inclusiveness and stability in large, diverse democracies like India.
33We measure turnout change at the constituency level by looking at the difference in turnout percentages across the two general elections. It is:
34We measure electoral volatility at the constituency level by calculating the Pedersen index. It is
35V is the electoral volatility, pi,t is the proportion of votes that party i received in election t, and pi,t−1 is the percentage of votes that party i received in the previous election t−1. n is the total number of political parties. The summation adds up the absolute changes in the vote shares for each party, and the result is divided by 2 to avoid double-counting the shifts. It looks at the net change in voter support for political parties between 2019 and 2014. We exclude the independents.
36Finally, we also look at the change in the effective number of parties, which measures the number of major parties in a party system (Laakso and Taagepera 1979). The formula for ENOP is
37In the formula, pi is the proportion of votes that party i received in the seat, and n is the total number of parties. ENOP essentially looks at the fragmentation of the party system at the constituency level. We calculate the difference in the ENOP in 2019 from the ENOP in 2014.
38First, we explore the overall impact of the BJP party centralization on the local party system, i.e. at the constituency level. Are there any differences in the level of turnout change, electoral volatility, and the effective number of parties in constituencies where the party reran its MPs and in constituencies where the party did not rerun its incumbent MPs? We expect constituencies in which a major party like the BJP renominated its MPs to have less turnout change, less party system fragmentation, and less electoral volatility in comparison to constituencies in which the party did not renominate its sitting MP. When a major party denies nomination to its incumbent MPs, we could see political defections, where some candidates could go on to form smaller new parties or run as independents. A new party entry, then, could increase turnout (Heath and Ziegfeld 2018) and be reflected in a higher effective number of parties and greater electoral volatility as the defected MP would take some voters with them. In other words, the higher turnout, effective number of parties, and electoral volatility could reflect the personal vote moving with a potentially upset MP who is denied renomination.
Table 4
|
|
Dependent variable:
|
|
|
|
Turnout_change
|
ENOP_change
|
Pedersen Index
|
|
(1)
|
(2)
|
(3)
|
|
|
Renominated (ref: replaced)
|
-0.642
|
-0.084
|
0.933
|
|
(0.616)
|
(0.072)
|
(0.787)
|
|
|
|
|
|
Constant
|
2.828***
|
-0.507***
|
8.731***
|
|
(0.475)
|
(0.056)
|
(0.607)
|
|
|
|
|
|
|
Observations
|
271
|
271
|
271
|
|
R2
|
0.004
|
0.005
|
0.005
|
|
|
Note:
|
*p**p***p<0.01
|
Party system change across seats with rerunning and replaced BJP MPs
39In Table 4, we can see that there is no difference in turnout change, ENOP change, or electoral volatility (measured using the Pedersen Index) in constituencies where the party reran its MPs in comparison to those where it did not. Of course, the measures of party system change we are looking at are aggregate measures where internal party organizational changes in one party should not be expected to make changes at the systemic level. However, when a major party denies nomination to its incumbent MPs, we could still be possible to see political defections of notable candidates, where some candidates could go on to form smaller new parties or run as independents. A new party entry, then, could increase turnout (Heath and Ziegfeld 2018) and be reflected in the positive changes in the effective number of parties or electoral volatility. Nonetheless, we do not see any differences in the aggregate measures of party system change between constituencies where the party rerun its MPs and where it dropped them.
- 9 Once again, we call “positive/progressive” any change that increases the presence of women, Backwar (...)
- 10 Since the independent variables (e.g., caste change, gender change) are binary and unequally distri (...)
40Next, we explore the potential consequences of the key comparative hypothesis we tested in the preceding section: does party organizational centralization lead to increased representation of historically under-represented groups? The aggregate evidence offers limited support for this claim, with no substantial shift toward greater inclusion of such groups. However, there were specific instances, though relatively few, where the party replaced a male candidate with a female one, or an upper-caste candidate with a candidate from a lower caste. While these changes were not widespread, they nonetheless provide an opportunity to examine whether such shifts had any measurable impact on the local party system. In particular, we can compare constituencies that experienced a change toward greater representation of marginalized groups with those that saw either no change or a regressive shift to assess whether progressive candidate selection had any local electoral consequences. Because of the small changes we are looking at, we created a binary coding to increase our statistical power. We code 1 if there is a positive change with respect to caste or gender, and 0 if there is no change or a regressive shift.9 For instance, if the BJP replaces its sitting upper caste MP with a lower caste candidate, we code this as 1, and if the BJP replaces its sitting upper caste MP with another upper caste MP, we code this as 0. We also assign a 0 if the BJP replaces its lower caste MP with an upper caste candidate, i.e., the regressive shift. We do the same for the gender change variable, where 1 = male to female and 0 = no change or female to male. We do this because we are essentially interested in comparing local-level party system changes in constituencies that show a positive change in comparison to constituencies that did not see a positive change. Of course, in an ideal world, we would like to treat this as a nominal variable, but due to the limitations of small n, we settle on this binary approach. We also use robust standard errors to alleviate concerns of heteroskedasticity10 due to the skewed distribution of our main independent variables of caste and gender change. We expect turnout to be greater in constituencies that saw a positive representational change, i.e., upper caste MP to a lower caste candidate or male MP to a female candidate, in comparison to constituencies that did not see a change or saw a regressive shift. Increasing descriptive representation can have the positive effect of mobilizing more voters from the historically marginalized groups to turn up to vote.
41When it comes to electoral volatility, we expect a positive representational change to increase electoral volatility as we expect marginalized voters to move towards the BJP from other smaller caste-based parties or independent candidates that represent the specific caste or gender in comparison to constituencies that saw no change or a regressive shift.
42Finally, when it comes to party system fragmentation measured through the effective number of parties, we expect a positive representational change to decrease party system fragmentation as marginalized voters now consolidate behind the BJP over other regional parties or independent candidates in comparison to no change or a regressive shift in representation.
Table 5
|
|
Dependent variable:
|
|
|
|
Turnout change
|
ENOP change
|
Electoral volatility
|
Turnout change
|
ENOP change
|
Electoral volatility
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
|
|
Caste change
|
-0.181
|
-0.485**
|
7.708***
|
|
|
|
|
(1.970)
|
(0.198)
|
(2.289)
|
|
|
|
|
|
|
|
|
|
|
|
Gender change
|
|
|
|
0.161
|
-0.026
|
1.600
|
|
|
|
|
(1.494)
|
(0.154)
|
(1.818)
|
|
|
|
|
|
|
|
|
Constant
|
2.843***
|
-0.467***
|
8.100***
|
2.804***
|
-0.503***
|
8.483***
|
|
(0.564)
|
(0.057)
|
(0.655)
|
(0.587)
|
(0.061)
|
(0.715)
|
|
|
|
|
|
|
|
|
|
Observations
|
110
|
110
|
110
|
110
|
110
|
110
|
|
R2
|
0.0001
|
0.053
|
0.095
|
0.0001
|
0.0003
|
0.007
|
|
|
Note:
|
*p**p***p<0.01
|
Impact of BJP Candidate Caste and Gender Changes on Turnout, Party System Fragmentation, and Electoral Volatility (2014–2019)
43The regression results in Table 7 provide mixed support for the hypothesized relationships between candidate demographic changes and electoral outcomes. Consistent with expectations, constituencies where the BJP changed the caste background of its candidate between 2014 and 2019 experienced a significant increase in electoral volatility (Model 3), with volatility rising by approximately 7.7 points on average (p<0.01). There is also a statistically significant negative association between caste change and ENOP (Model 2), suggesting that caste-based candidate substitutions may have narrowed the effective number of parties in some constituencies, potentially by consolidating the vote. However, the effect of caste change on turnout (Model 1) is negative but statistically indistinguishable from zero, implying limited evidence that caste-based candidate switching directly affected voter participation.
44In contrast, gender change,—the substitution of a male candidate with a female one or vice versa—does not show a statistically significant relationship with any of the three outcomes (Models 4–6). While the coefficient signs are consistent with theoretical expectations for turnout, party fragmentation and electoral volatility, the estimates are imprecise, likely due to the small number of such changes in the sample and limited variation in the independent variables of caste and gender change. As a result, these null findings should be interpreted cautiously, as the lack of significance may reflect insufficient statistical power rather than the absence of an underlying effect.
45To further investigate the joint effects of caste and gender change on the local constituency-level party system, we include an interaction term between caste change and gender change. Changing both the caste and gender of a candidate could disrupt existing voter alignments more sharply than altering either attribute alone or conversely, it might generate clearer signals to specific voter groups, consolidating support and increasing turnout, reducing party system fragmentation or electoral volatility. Including the interaction term allows us to test whether these combined changes have multiplicative effects on key electoral outcomes.
46Below, we only report plots from the models that have statistically significant effects.
Figure 2
Effect of gender and caste change of the BJP candidate on party system fragmentation
47In Figure 2 above, we can see that there is a statistically significant negative effect on changing the caste and gender of the replaced candidate on party system fragmentation measured through the effective number of parties. When the BJP replaces its incumbent upper caste male MP with a lower caste female candidate, we see the constituency-level party system fragmentation significantly reduce. Of course, the large confidence interval does not help us to substantively interpret the effect size. Nevertheless, despite the small n, we can see that this effect is statistically significant. When the party replaces its upper caste male MP with a lower caste female candidate, we might be seeing specific lower caste and female voters abandoning other smaller parties and independents to consolidate behind the main party. Once again, without individual level data, we cannot be sure of this transfer of votes, but the aggregate data clearly shows that the constituency-level party system is less fragmented in places where the BJP makes these progressive changes than when it does not.
48In this paper, we have provided some tentative empirical evidence to show that the BJP did centralize its organization in the post-2014 period. We show this by using lower rates of renomination of incumbent MPs as a conservative proxy for measuring intra-party centralization. The BJP renominated the lowest proportion of incumbent MPs in the 2019 general election in comparison to all the previous general elections. We have also shown who the party selected instead of the incumbent MPs it did not rerun in the 2019 general elections.
49While most studies underline the negative electoral consequences of intra-party centralization (Shrimankar 2020; Ziegfeld 2016), we have explored an alternative implication of party centralization: does party centralization improve descriptive representation? We know that intra-party centralization improves descriptive representation in the British and Western European context. In this paper we show that intra-party centralization in the BJP did not result in significant improvements in the Indian context, especially in the case of the BJP in 2019. In the 2024 general elections, the party renominated an even smaller proportion of its incumbent MPs. It remains to be seen if the party did replace the incumbent MPs with a greater share of candidates from historically under-represented sections of the population.
50We also know that intra-party centralization contributes to party organizational decay. We have seen this with the Congress Party in the 1970s under Indira Gandhi and are also observing some of its impact on the BJP in the post-2014 era. Centralizing decision-making power leads to an exodus of key party functionaries, and in many cases, senior incumbent MPs who are denied candidature. The party is also not able to account for views that are emanating from party activists and grassroots workers. This often ends up hurting the party electorally. Consequently, the party system sees the rise of regional and smaller parties that are made up of disgruntled party members.
51In the final part of the paper, we also explore the consequences of changes in one party organization on the local party system. We evaluate this using aggregate-level data on turnout, effective number of parties, and electoral volatility. Once again, we must be careful in interpreting the data due to the generally small n changes we are dealing with, and the type of data (aggregate, as opposed to individual-level) we are dealing with. Nonetheless, the aggregate-level changes in the party system in constituencies where the BJP replaced a higher-caste MP with a lower-caste candidate point towards lower party system fragmentation and higher electoral volatility. It could be that in these constituencies; lower caste voters are moving towards the BJP for the 2019 general elections. Once again, we cannot confirm this without individual survey data, but the aggregate data does point in that direction.
52More broadly speaking, to strengthen the theorization of intra-party centralization and its consequences on the party system, it is essential to incorporate more case studies from the Global South. The dynamics of candidate selection, representational trade-offs, and party system responses may operate differently in settings marked by social heterogeneity, weaker party institutionalization, or different forms of party-voter linkages. To this end, India offers a rich but underutilized ground for exploring how centralized party strategies intersect with social cleavages, electoral volatility, and party system fragmentation. In this paper, we have only scratched the surface by only looking at incumbents within a single party for a single election year. Future research should look at all candidates for all the major parties for all the election years both at the national and state level to see how candidate renomination and replacement intersects with descriptive representation and party system change.
53More generally, expanding empirical work beyond established Western cases will allow for more generalizable insights and help refine theoretical frameworks that currently lean too heavily on European or American experiences.