Introduction
Infant mortality varies nearly as much within India as it does across the entire globe. For instance, in 2011 the infant mortality rate in India’s capital city Delhi was 28 deaths/1000 live births, comparable with developing countries such as Guatemala and Dominican Republic.Footnote 1 Infant mortality more than doubles in the state of Madhya Pradesh, where infant mortality was 59 in 2011, comparable with low-income countries such as Haiti and Liberia.Footnote 2 On the Southern tip of India, Kerala has an infant mortality rate of 12, which is comparable with infant mortality rates in the developed world.Footnote 3 The presence of such extreme variation within a single country demonstrates the importance of understanding the subnational determinants of infant mortality.
Yet, the bulk of the existing political science literature primarily focuses on national-level conditions such as regime type. At the heart of the literature is the idea that democracies outperform non-democracies (Moon and Dixon, Reference Moon and Dixon1985; Przeworski et al., Reference Przeworski, Alvarez, Cheibub and Limongi2000; Zweifel and Navia, Reference Zweifel and Navia2000; Lake and Baum, Reference Laasko and Taagepera2001; Bueno de Mesquita et al., Reference Bueno de Mesquita, Smith, Siverson and Morrow2003; Besley and Kudamatsu, Reference Besley and Kudamatsu2006; Wigley and Akkoyunlu-Wigley, Reference Wigley and Akkoyunlu-Wigley2011; Gerring et al., Reference Gerring, Thacker and Alfaro2012, among others). However, contradictory scholarship questions if there is any relationship between regime type and infant mortality at all (Ross, Reference Ross2006). This paper offers a different perspective that not only sheds light on cross-national and within-regime variations in infant mortality but also accounts for subnational variations within a single democracy – India. I argue that variations in three core democratic attributes, political representation, citizens’ participation, and electoral competition motivate political representatives to reduce infant mortality.
Several prominent scholars such as Dahl (Reference Dahl1971), Verba et al. (Reference Verba, Schlozman and Brady1995), Manin et al. (Reference Manin, Przeworski and Stokes1999), and Lijphart (Reference Lijphart2012), among others, emphasize the significance of these democratic attributes. Manin et al. (Reference Manin, Przeworski and Stokes1999: 4) stress the significance of representation as they note ‘… a central claim of democratic theory is that democracy systematically causes governments to be representative’. Verba et al. (Reference Verba, Schlozman and Brady1995: 1) refer to citizens’ participation as ‘the heart of democracy’. Dahl’s (Reference Dahl1971) polyarchy emphasizes attributes of both representation and participation. Lastly, presence of free and fair elections is yet another important feature of democracy (Dahl, Reference Dahl1971), where electoral competition ensures responsiveness of elected officials. Overall, there is little disagreement among scholars about the significance of these three democratic attributes.
This paper makes two primary contributions to existing research. First, the concept of democracy is disaggregated into three core democratic attributes to more precisely specify the link between democracy and infant mortality. This enables us to understand why democracies perform better than non-democracies. As the three democratic attributes are prevalent to a larger degree among democracies as compared with non-democracies, the paper also speaks to the larger debate in the existing literature about whether democracies perform better than non-democracies. Second, the paper identifies the determinants of infant mortality that vary at both the national and subnational level, thereby shedding new light on the controversy over the role of democracy at the national level as well as providing a subnational theory on the politics of infant mortality within democracies.
The implications of this research stretch beyond the domain of infant mortality to address how democracy affects the plight of the poor in particular and human welfare in general. Infant mortality is particularly prevalent among the poorest sections of society and also reflects basic living conditions of the poor such as poverty, housing quality, resistance to diseases, sanitation, air quality, availability of clean water, accessibility to neonatal, and prenatal health service, among others (Victora et al., Reference Victora, Wagstaff, Schellenberg, Gwatkin, Claeson and Habicht2003) for which cross-national time-series data as well as subnational time-series date are not widely available.
Moreover, infant mortality has been used as an indicator of human well-being by several prominent cross-national studies (Moon and Dixon, Reference Moon and Dixon1985; Przeworski et al., Reference Przeworski, Alvarez, Cheibub and Limongi2000; Lake and Baum, Reference Laasko and Taagepera2001; Gerring et al., Reference Gerring, Thacker and Moreno2005, Reference Gerring, Thacker and Moreno2009, Reference Gerring, Thacker and Alfaro2012; Ross, Reference Ross2006, among others).
I employ a time-series data set from 15 major Indian states between 1980 and 2011 to assess how the core features of democracy affect infant mortality. Democratic attributes of representation, participation, and competition are measured with the effective number of parties (ENP), voter turnout, and the electoral margin of victory, respectively. Overall, the results demonstrate the significance of democratic politics, particularly political representation, in influencing infant mortality.
The Indian case
India presents a particularly useful case to analyze the relationship between democratic attributes and infant mortality for two primary reasons. First, the vast variation in infant mortality among Indian states is suggestive that subnational politics plays an important role in India. Second, a subnational research design provides methodological leverage by holding constant prominent national-level determinants of infant mortality.
Indian states present comparable patterns of variation in infant mortality as observed cross-nationally and among democracies and non-democracies alike. Figure 1 displays the dispersion of infant mortality rates through box-plots from 1980 to 2011 across four groups: global sample of countries, democracies, non-democracies, and Indian states.Footnote 4 The x-axis identifies these four groups and the y-axis shows the variation in infant mortality rates.Footnote 5 What is particularly interesting is that the variation in infant mortality among Indian states is comparable with the variation among all democracies at the national level.Footnote 6 Thus, understanding the subnational variation in infant mortality at the state level is imperative.Footnote 7
The distribution of infant mortality: globally, non-democracies, democracies, and Indian states, 1980–2011.
A subnational study on India also offers an important methodological advantage by providing relatively greater comparability between cases such as common historical, cultural, and socio-economic characteristics, which enhances our ability to make valid inferences (Snyder, Reference Snyder2001; Gerring, Reference Gerring2004). Despite differences among Indian states, a subnational study provides a better control for measurable and immeasurable factors as compared with a cross-national study. In particular, this subnational research design allows one to hold constant two primary cross-national determinants of infant mortality – regime type and political institutions. The literature emphasizing regime type argues that democracies perform better than non-democracies (Przeworski et al., Reference Przeworski, Alvarez, Cheibub and Limongi2000; Lake and Baum, Reference Laasko and Taagepera2001; Bueno de Mesquita et al., Reference Bueno de Mesquita, Smith, Siverson and Morrow2003; Gerring et al., Reference Gerring, Thacker and Alfaro2012, among others). However, India’s democratic legacy cannot account for the large disparities in infant mortality among Indian states. The political institutions literature identifies centripetal (Gerring et al., Reference Gerring, Thacker and Moreno2005) and parliamentary institutions (Gerring et al., Reference Gerring, Thacker and Moreno2009) as two institutional alternatives that produce lower levels of infant mortality than their counterparts.Footnote 8 However, as political institutions at the subnational level in India mirror those at the national level, they cannot explain the subnational variation in infant mortality either.Footnote 9 This study holds democratic regime type and political institutions constant and focuses on alternative factors that vary within democracies.
Democratic attributes and infant mortality
Infant mortality has been a heavily studied subject among social scientists. Mosley and Chen (Reference Mosley and Chen1984) provide a broad theoretical framework to analyze mortality from a social science perspective by identifying three categories of socio-economic determinants: individual-level, household-level, and community-level factors. This paper focuses on community-level determinants by analyzing the effect of three democratic attributes among Indian states. The relationship between democratic attributes and infant mortality in contemporary times is integrally linked to historical differences in socio-cultural factors across regions within India. There is a discernable regional pattern in welfare outcomes where the Southern states seem to perform better than the Northern states.Footnote 10 These regional disparities may reflect distinct socio-cultural factors that have influenced the development of democracy across the country. Dyson and Moore (Reference Dyson and Moore1983) discuss the distinct kinship structures between the Northern and Southern states and its consequences for autonomy of women. In contrast to the Northern states, among Southern states, marriages can be endogamous (spouses can belong to the same kinship group) such that women are able to maintain links with their kins and there are fewer restrictions on inheritance of property rights by women. Serra (Reference Serra1999: 9) argues that these kinship features in the South could arguably account for a ‘more “progressive” and open culture’, which is suggestive of a more democratic society in the South as compared with the North.
Yet, another explanation focuses on differences in the position of the lower castes between the Northern vs. Southern states. The numerical strength of the lower caste and caste-based politics in the South facilitated their mobilization (Church, Reference Church1984; Varshney, Reference Varshney2000).Footnote 11 In contrast, politics among Northern states revolved around the Hindu–Muslim cleavage, where the relationship between the lower and upper caste members was clientelistic with the lower caste dependent on the upper caste members (Rudolph and Rudolph, Reference Rudolph and Rudolph1967).Footnote 12 The difference in status of the lower caste may have influenced the degree to which democratic attributes are prevalent in the two regions.Footnote 13
Several studies focus on state- and district-level variations in welfare outcomes and services within India (Nag, Reference Nag1983, Reference Nag1989; Measham et al., Reference McGuire and Frankel1999; Betancourt and Gleason, Reference Betancourt and Gleason2000; Besley and Burgess, Reference Besley and Burgess2002; Gaudin and Yazbeck, Reference Gaudin and Yazbeck2006; Bose, Reference Bose2007; Bhattacharya, Reference Bhattacharya2009; Pushkar, Reference Pushkar2012) and this study contributes to this vast literature by proposing a democratic theory of infant mortality among Indian states.
Existing development and health-related scholarship sheds light on various policy alternatives that influence health outcomes such as infant mortality. For instance, hospital infrastructure and quality (Aguilera and Marrufo, Reference Aguilera and Marrufo2007), whether or not the mother is educated (Hobcraft et al., Reference Hobcraft, McDonald and Rutstein1984; Song and Burgard, Reference Song and Burgard2011), time between childbirths (Trussell and Pebley, Reference Trussell and Pebley1984), water pollution (Jorgenson, Reference Jorgenson2004), environmental factors and condition of mothers (Folasade, Reference Folasade2000), fertility levels, family health programs and female illiteracy (Macinko et al., Reference Macinko, Guanais, de F. and de Souza2006), and infrastructure (Fay et al., Reference Fay, Leipziger, Wodon and Yepes2005), among others, are some of the proximate determinants of infant mortality. However, an imperative question arises, namely, what motivates elected officials to adopt any of these policies? I focus on three core democratic attributes – namely, political representation, citizens’ participation, and electoral competition – that fundamentally shape the incentives of political representatives to adopt appropriate policies to reduce infant mortality.Footnote 14
Political representation
Parties are the quintessential agents of representation in a democracy where they play the key role of linking citizens with their elected officials (Diamond, Reference Diamond1997). Scholars such as Schattschneider (Reference Schattschneider1942: 3) assert, ‘democracy is unthinkable save in terms of parties’, while Lipset (Reference Lipset1996: 169) refers to parties as ‘the core institution of democratic politics’. Even though most democracies have political parties, the nature of party system varies between democracies.
The significance of party systems in democratic governance has been emphasized by prominent subnational studies on India. For instance, Chhibber and Nooruddin (Reference Chhibber and Nooruddin2004) argue that two-party states are more likely to spend on public goods than multiparty states because political parties in the former have a larger winning coalition to please and private goods become unaffordable for political parties as the size of the winning coalition increases. Proposing a different explanation, Sáez and Sinha (Reference Sáez and Sinha2009) argue that as multiparty states signal uncertainty, political parties spend more in these states to keep their constituents happy as compared with two-party states. Interestingly, these two studies make theoretically plausible yet contradictory predictions about the relationship between party systems and government expenditure. In contrast to the two studies, this paper focuses on a different theoretical perspective, the representativeness of party systems and on a different outcome, infant mortality.
Drawing on institutional models of democracy such as the consensus (Lijphart, Reference Lijphart2012) and centripetal models (Gerring et al., Reference Gerring, Thacker and Moreno2005), I argue that political representatives in multiparty states are more likely to pursue policies that reduce infant mortality because a multiparty system is more inclusive as compared with a two-party system. Multiparty states enable inclusion of diverse ideas and preferences in the legislature by providing different groups an opportunity to collectively influence policy-making and better represent varied issue dimensions within society as compared with two-party states. The motivation for all parties to work together in multiparty systems can be attributed to the accountability mechanism in a democracy that encourage parties to cooperate, compromise, and negotiate with one another and enact comprehensive policies that are beneficial for the society as a whole. Moreover, parties can also hold each other accountable as multiple parties can closely observe actions of others and keep the populace better informed about the performance of their elected representatives. While both inclusion and accountability play an important role, inclusion is the distinguishing feature between a two-party and a multiparty state.Footnote 15
Indeed parties in a two-party system may have an incentive to cater to the median voter and propose policies with a broader appeal that are centrist in nature (Downs, Reference Downs1957). This is likely in a homogenous society but in a diverse society such as India, parties in a two-party system are unlikely to cater to the interests of myriad groups because the median voter may not be representative of all the groups within society. Thus, interests of smaller groups are more likely to be overlooked in a two-party system. It is plausible for parties in a two-party system to reach out to minorities but they are likely to do so only to get the necessary majority support and win the election in a two-party state. Minorities whose support is not required to win office will find it relatively difficult to get their voices heard in the policy-making process. This is less likely to happen in a multiparty system.
One potential criticism that can be made against a multiparty system is that it may give too much influence to smaller parties who could behave in an intransigent manner and make it difficult to pass comprehensive policies. While this is a plausible scenario, incentives to do so will be lower in a society with cross-cutting cleavages such as India where parties often have to appeal to different issue dimensions to expand their support base (Chandra, Reference Chakraborty and Kim2005). The prospect of alienating additional supporters in future elections will encourage parties to work with each other.
As an illustrative example let us consider the state of Tamil Nadu. The Congress Party had a stronghold over the state since independence from 1947 to the 1960s, with a support base that primarily consisted of the landed and the upper castes (Manivannan, Reference Manivannan1992). Gradually, regional parties such as the Dravida Munnetra Kazhagam (DMK) and the All Indian Annadurai DMK emerged to represent the interests of the poor, lower castes, and women and smaller caste-based parties such as the Pattali Makkal Katchi (Toiling People’s Party) party, the Pudhiya Thamizhagam party, the Makkal Tamil Desam Katchi party, among others, emerged to represent different caste-based groups (Wyatt, Reference Wyatt2002). What is noteworthy is that changes in infant mortality correlate with changes in the ENP within the state.Footnote 16 Between 1980 and 1990, ENP was over 2 and infant mortality during this period reduced by 36%.Footnote 17 ENP fell to under 2 between 1991 and 2000 during which time infant mortality reduced by only 10%. From 2001 to 2011, ENP gradually increased to over 2 and during this period the state witnessed the steepest decline in infant mortality, 55%. This was especially significant considering that national infant mortality was reduced by 31% between 2001 and 2011, a lower rate of reduction than that of Tamil Nadu. This can be attributed to the state government’s emphasis on universal coverage rather than targeted schemes (Kumar et al., Reference Kumar, Chen, Choudhury, Ganju, Mahajan, Sinha and Sen2011) and provision of better maternal and neonatal care in the last decade such as increasing the number of institutional deliveries, skilled personnel during deliveries, and providing incentives for trained personnel to work in rural areas, among others (Padmanaban et al., Reference Padmanaban, Raman and Mavalankar2009).
In contrast, Madhya Pradesh exhibits poorer health outcomes with an average infant mortality of 98 from 1980 to 2011. One fundamental difference between the two states is that while lower castes/classes have been able to mobilize and gain representation in Tamil Nadu leading to a multiparty system, upper castes/classes have been dominant in Madhya Pradesh (Harriss, Reference Harriss1999; Varshney, Reference Varshney2000; Wyatt, Reference Wyatt2002). Even though smaller parties such as the Gondwana Ganatantra Party and the Bahujan Samaj Party, among some others, have been trying to garner support in Madhya Pradesh (Ramshankar, Reference Ramshankar2004; Shankar and Sisodia, Reference Shankar and Sisodia2009), the state is still primarily characterized by a two-party system where the two main parties are the Congress and the BJP. Indeed while there has been an increase in the representation of lower castes and classes, this has primarily taken place under the auspices of the two main parties (Gupta, Reference Gupta2005, Reference Gupta2006). The consequence of the relative lack of representation of these groups can be seen in the relative lack of access to health services and prevalence of worse health outcomes especially among these groups in the state (Pallikadavath et al., Reference Pallikadavath, Foss and Stones2004; Nagdeve, Reference Nagdeve2014). This discussion leads to the first hypothesis: an increase in the number of parties reduces infant mortality.
Citizens’ participation
Citizens’ participation has been referred to as ‘the heart of democracy’ (Verba et al., Reference Verba, Schlozman and Brady1995). Indeed, widespread citizens’ participation is essential for a vibrant and flourishing democracy where the masses have the opportunity to influence the governing process. However, democratic societies display tremendous variation in citizens’ participation. Cleary’s (Reference Cleary2007) study provides evidence of a positive relationship between citizens’ participation and responsiveness of elected officials among Mexican municipalities. I emphasize that participative citizens are able to induce accountability from elected officials regardless of the institutional or cultural differences between countries. This study on India provides an opportunity to assess the relationship between citizens’ participation and infant mortality in an Indian context.
Citizens’ participation can take various guises, ranging from voting in elections, referendums, holding peaceful protests, demonstrations, and attending public hearings, among others. I specifically focus on citizens’ electoral participation and argue that high levels of voter turnout reduce infant mortality by signaling the presence of an overall active and vigilant populace. Voting is a frequent and collective activity that can influence broad outcomes within society (Verba et al., Reference Verba, Nie and Kim1978). It provides a mechanism through which political representatives can be held accountable. However, even though majority of the adults can vote in a democracy (Dahl, Reference Dahl1989), not all eligible voters exercise this right.
High voter turnout mirrors a society where citizens take advantage of the electoral mechanism to either reward representatives or penalize political representatives for their poor performance in office (Powell, Reference Powell2000).Footnote 18 More importantly, high turnout is also indicative of a participative citizenry who are active in other aspects of politics (Inkeles, Reference Inkeles1969; Verba et al., Reference Verba, Schlozman and Brady1995) such as participating in protests, demonstrations, among others, to induce better government performance. Kerala provides evidence of a politically active populace, where the people vote in large numbers and participate in various other aspects of politics as well.Footnote 19 The state has the highest levels of unionization in both the formal and informal sectors, which include the upper and lower classes (Heller, Reference Heller2000). Moreover, citizens have been known to act collectively to monitor the functioning of schools and health centers (Franke and Chasin, Reference Franke and Chasin1997), so much so that the absence of a doctor at a primary health center has resulted in demonstrations being held by people at the local government office (Mencher, Reference Mencher1980).Footnote 20 The state is indicative of an active citizenry (Nag, Reference Nag1983, Reference Nag1989), which can be attributed to the social and religious reforms that weakened the caste system and enabled the poorer sections of society to join forces and influence democratic politics (Sen, Reference Sen1992). Emphasizing the role of participation, Sen notes ‘… that government action, in order to be successful, requires popular participation’ (Reference Sen1992: 276). Thus, an active citizenry induces political representatives to meet the needs of their supporters, thereby motivating them to perform better and reduce infant mortality.Footnote 21
Lower voter turnout, on the other hand, indicates the presence of voters who do not use the electoral mechanism, making it easier for poor performing incumbents to remain in office. It is also more generally indicative of a passive populace who are not as participative in other aspects of politics. This may be tantamount to political representatives taking an indifferent attitude toward the needs and preferences of their supporters because the masses do not hold them accountable through political participation, resulting in higher levels of infant mortality.
For instance, the average voter turnout in Uttar Pradesh is ~51% from 1980 to 2011, one of the lowest in the country compared with the national average of 69%. Uttar Pradesh has an average infant mortality of 99, one of the highest in India, while the national average is 62. In contrast, the state of Kerala has a high voter turnout in the country, averaging 74% from 1980 to 2011 with an average infant mortality of 19, one of the lowest in the country. This discussion suggests an increase in voter turnout reduces infant mortality.
Electoral competition
Free and fair elections are the basic prerequisites for a democracy (Dahl, Reference Dahl1971).
Elections provide an opportunity for citizens to choose the candidates who are best suited to satisfy the needs of the masses. While elections are a regular phenomenon in all democracies, here again there is variation in electoral competitiveness.
Existing studies on electoral competitiveness and government performance provide inconclusive evidence. While Cleary’s (Reference Cleary2007) study does not find a significant relationship between competition and performance, subnational studies of the United States do find evidence of a positive link between the two (Griffin, Reference Griffin2006; Konisky and Ueda, Reference Konisky and Ueda2011). These contradictory findings are intriguing as both Mexico and the United States have a SMDP electoral system.Footnote 22 This study on India enables one to revisit the impact of electoral competitiveness in a similar electoral context. I argue that higher levels of electoral competition, as captured by a smaller margin of victory, leads to lower levels of infant mortality as it signals the presence of viable competitors, thereby providing incentives for all candidates to perform well.Footnote 23
Competitive elections provide incentives for the incumbents and the challengers to appeal to the electorate with the promise of better performance. The incumbents are aware of alternative challengers who could potentially displace poor performing officials. Competitiveness prevents incumbents from governing poorly and exercises the necessary pressure on them to be proactive. Challengers, on the other hand, constantly strive to come to power. Here again, given the competitive nature of elections, they need to appeal to the electorate with the promise of better performance than the incumbents.Footnote 24 If elected, political representatives have to satisfy the needs and preferences of multiple segments of population by implementing policies that enhance societal welfare as a whole. Moreover, the presence of cross-cutting cleavages in India makes it easier and imperative for contestants to make broader appeals to mobilize a larger number of supporters by emphasizing different cleavage dimensions (Chandra, Reference Chakraborty and Kim2005). This adds a dimension of fluidity to the support-groups of contestants as citizens can identify with alternative political parties. Competitive elections propel all political representatives to perform well, regardless of the electoral system. The closer or more competitive the election, the easier it becomes for challengers to replace incumbents. Thus, political parties cannot get complacent in a competitive environment.
For instance, extreme electoral competitiveness in the state of Andhra Pradesh in the 2004 elections propelled political parties competing in the state to provide greater welfare benefits (Elliott, Reference Elliott2011). The margin of victory (vote share) between the winner [Indian National Congress Party (INC)] and its primary challenger [the grand alliance led by the Telugu Desam Party (TDP)] in the state election was <1%. Between 2004 and 2011, the reduction in infant mortality in Andhra Pradesh was 27% as compared with a 21% aggregate reduction in infant mortality among all Indian states during the same time period, indicating a steep decline in infant mortality in the state following competitive elections. In the 2009 election campaign, the INC proposed to provide free health-care services for serious illnesses, subsidized housing projects, pension scheme for widows, scholarships for relatively deprived sections of society, among others. The TDP in an attempt to appeal to the electorate, proposed to provide free health-care to a broader range of illnesses than those proposed by the INC and give cash transfers to the poor families as well, among other things. Even though the INC won the 2009 elections, the competitiveness of the election motivated the challengers to appeal to the electorate with the provision of greater welfare goods and services.
Alternatively, in a relatively uncompetitive system, there are fewer incentives for incumbents to be responsive to the needs of the populace. The lack of government initiative to generate development in Bihar demonstrates the consequences of an uncompetitive system. Bihar was under the rule of the Rashtriya Janata Dal (RJD) party from 1990 to 2005.Footnote 25 Elections were relatively uncompetitive with the incumbents winning elections over these 15 years with an ~12.5% margin of victory. There was an 18% reduction in infant mortality during this time period in the state as compared with a 38% reduction in infant mortality among all Indian states. Moreover, there was no discernable improvement in the condition of the incumbents’ supporters, namely the rural poor (Robin, Reference Robin2004) and 15 years of misrule and poor governance adversely influenced the development of the state (Sharma, Reference Sharma1995).Footnote 26 Thus, even though there were periodic elections, they were insufficient to motivate the incumbents to satisfy the needs of the populace.Footnote 27 This discussion suggests a smaller margin of victory reduces infant mortality.
Data and methods
This paper analyzes the impact of three core democratic attributes on infant mortality on a sample of 15 major Indian states from 1980 to 2011.Footnote 28 The primary dependent variable is infant mortality, which is measured by the number of infant deaths (of 1 year or less) per thousand live births. As the variable is skewed, the log of infant mortality is used in all the models. The data for infant mortality are available from the Sample Registration Systems (various years) surveys published yearly by the Office of the Registrar General. The primary independent variables are political representation, citizens’ participation, and electoral competition. Political representation is measured with the effective number of parliamentary parties, which takes into account each party’s seat-share in the ‘Vidhan Sabha’ (lower house of the state legislature).Footnote 29 Citizens’ participation is measured by voter turnout to ‘Vidhan Sabha’ elections, which is calculated as a percentage of registered voters.Footnote 30 Electoral competition is measured with electoral margin of victory, which is the percentage difference in vote share between the largest and the second largest recipients of votes.Footnote 31 The data for the three primary independent variables have been calculated from the Election Commission of India (various years).
Over the 32-year period of analysis, infant mortality has been declining across all the states in India and this trend is evident globally as well, plausibly because of the advancement in science and technology, improvements in health-care services and products, and greater economic development. However, there is a discernable regional pattern within India where the Southern states display lower levels of infant mortality as compared with the Northern states. In the case of the three democratic attributes, while the ENP, voter turnout, and margin of victory do change from election to election, in general, we do observe the Southern states displaying a more democratic society with a higher number of parties and voter turnout as compared with the Northern states. However, not all states neatly fit into the North–South dichotomy, which is why a state-level analysis such as this is especially beneficial.
Figures 2, 3, and 4 present scatter-plots with fitted values between the theoretical variables of interest.Footnote 32 In Figures 2 and 3, the downward sloping fitted lines indicate a negative correlation between ENP, voter turnout, and infant mortality suggesting that multiple parties and higher levels of voter turnout are associated with lower levels of infant mortality. In Figure 4, the upward sloping fitted line indicates a positive correlation between margin of victory and infant mortality suggesting that a larger margin of victory is associated with higher levels of infant mortality, although the correlation is weaker in Figure 3 than in Figures 1 and 2. Overall, the scatter-plots provide some insight into the relationship between the theoretical variables of interest.
Effective number of parties and infant mortality.
Voter turnout and infant mortality.
Margin of victory and infant mortality.
The theory is also put to a more stringent test by holding other variables constant. I draw on the existing cross-national literature to account for alternative explanations of infant mortality. Income per capita is a standard control in most studies (Przeworski et al., Reference Przeworski, Alvarez, Cheibub and Limongi2000; Gerring et al., Reference Gerring, Thacker and Moreno2005, Reference Gerring, Thacker and Moreno2009; Ross, Reference Ross2006; Gerring et al., Reference Gerring, Thacker and Alfaro2012). Income may influence infant mortality where states with higher income have more resources to address the welfare needs of their residents. I measure income with the log of per capita state domestic product at constant price. The data are available from the Reserve Bank of India (various years). Population size is yet another factor that may influence infant mortality where states with larger populations find it difficult to provide health services to all its residents (Zweifel and Navia, Reference Zweifel and Navia2000; Enikolopov and Zhuravskaya, Reference Enikolopov and Zhuravskaya2007). I use the log of population and the data are available from the Census of India (various years), published by the Office of the Registrar General.Footnote 33 State and year dummies have been included in the baseline models to guard against the possibility of state or time-specific effects influencing the relationship between the primary dependent and independent variables.
An ordinary least squares panel-corrected standard error (PCSE) estimation technique with a lagged dependent variable is used to assess the relationship between the theoretical variables of interest. The PCSE technique enables me to address the problem of unit heteroskedasticity (Beck and Katz, Reference Beck and Katz1995). A lagged dependent variable is able to control for alternative factors that may influence infant mortality as well as address serial correlation. As it has been argued that a lagged dependent variable may be atheoretical and could overshadow the effect of substantively significant variables (Achen, Reference Achen2000), robustness tests include findings without a lagged dependent variable as well.
Findings
Table 1 shows the impact of ENP, voter turnout, and electoral margin of victory on infant mortality. Model 1 shows that ENP has a statistically significant and negative effect on infant mortality, suggesting that a state with a higher ENP leads to lower levels of infant mortality. This is consistent with the hypothesized relationship between the two variables. The paper argues that states with more parties present an opportunity for multiple political parties to influence policy-making, thereby encouraging parties to work together toward comprehensive policies. Moreover, cross-cutting cleavages provide an additional incentive for parties in multiparty states to formulate policies with a broader appeal to the masses. Thus, higher ENP captures a more representative society that enables diverse societal interests to be incorporated in policy-making and the findings provide support to this theoretical proposition.
Democratic attributes and infant mortality
L.=one year lag; ENP=effective number of parties; SDP pc=state domestic product per capita.
Standard errors in parentheses.
***P<0.01, **P<0.05, *P<0.1.
Model 2 shows the effect of voter turnout and the finding indicates that it has a statistically significant and negative effect on infant mortality, suggesting that high voter turnout leads to lower levels of infant mortality. This is also consistent with the theoretical perspective proposed in this paper. High levels of voter turnout signal the presence of an overall active and vigilant populace, thereby increasing the accountability of elected officials. Political representatives, in turn, perform well as they are under the constant scrutiny of the public.
Margin of victory is statistically insignificant in model 3. The variable reaches statistical significance in some of the models discussed below, however, its effect is not robust to alternative specifications.Footnote 34 The theoretical argument proposed in this paper is that greater electoral competition motivates incumbents and challengers to perform well, given a higher degree of competitiveness. However, while an uncompetitive environment may make the incumbents complacent, the challengers still desire to win office and thus may strive to influence policy and enhance well-being of the masses. These contradictory theoretical perspectives could possibly account for the lack of significance of the variable. Clearly, the effect of electoral competition needs to be explored further to understand the ways in which this core democratic feature influences welfare outcomes.
Model 4 includes all three primary independent variables in one model. The results remain unchanged for ENP and voter turnout – where both variables retain statistical significance. Margin of victory, once again, ceases to have a significant effect on infant mortality. In terms of substantive significance, increasing the ENP and voter turnout from their minimum to maximum values decreases infant mortality by approximately 7 and 4%, respectively.Footnote 35 Number of parties and voter turnout capture attributes of political representation and citizens’ participation and the findings demonstrate the significance of these two core features in influencing performance of political representatives.
State domestic product per capita and population do not reach statistical levels of significance in any of the models in Table 1. While they are significant in some of the models discussed below, their effect is not robust across alternative model specifications. This suggests that political factors play a particularly important role in influencing public health outcomes. The baseline models are also very stringent with a lagged dependent variable as well as time and state dummies, which may partially account for the reduced influence of control variables. The lagged dependent variable remains significant in all four models, demonstrating that the previous years’ level of infant mortality is an important predictor of infant mortality in the next year.
Robustness tests
Various sensitivity tests were carried out to analyze the robustness of the findings. The baseline models were estimated with a lagged dependent variable that addresses the problem of serial correlation and captures any other factors that could influence infant mortality. As a first robustness test, instead of a lagged dependent variable I estimated the baseline models with AR(1) to address serial correlation and control for additional determinants of infant mortality. Party dummies are included for states where one of the three primary national parties held a plurality of seats in state legislatures, namely the BJP, the Congress, and the leftist parties, namely, Communist Party of India (CPI) or Communist Party of India, Marxist [CPI(M)]. These parties indicate distinct ideological underpinnings where the BJP is considered right of center, the Congress is considered centrist, while CPI/CPI(M) is considered leftist (Chhibber and Nooruddin, Reference Chhibber and Nooruddin2004). A regional party dummy variable is also included for states where regional parties attained a plurality of seats in the state legislature. Data for the party variables were calculated from the Election Commission of India (various years). In order to capture federal assistance to state governments, grants from the center is included in the models as well (Reserve Bank of India, 2010). Female literacy and percentage of urban population are two other factors that could influence infant mortality where higher values of female literacy and a higher percentage of the urban population may be associated with lower levels of infant mortality. The data are available from the Census of India (various years), published by the Office of the Registrar General.
The findings are presented in Table 2.Footnote 36 Among the control variables, higher income and greater grants from the center are associated with lower levels of infant mortality while states where the BJP has held a plurality of seats is associated with higher levels of infant mortality. Among the theoretical variables of interest, ENP and voter turnout remain significant in all the models and while margin of victory is significant in model 3, it ceases to have a significant effect in model 4 where all three primary independent variables are included. This reconfirms the effect of political representation and citizens’ participation.
Democratic attributes and infant mortality (without a lagged dependent variable and including additional control variables)
L.=one year lag; ENP=effective number of parties; BJP=Bhartiya Janata Party.
Standard errors in parentheses.
***P<0.01, **P<0.05, *P<0.1.
The next two tables present findings without time dummies (Table 3) and without state dummies (Table 4). Both ENP and voter turnout remain statistically significant across all the models in Tables 3 and 4. Similar to the baseline findings, margin of victory remains insignificant.
Democratic attributes and infant mortality (without time dummies)
L.=one year lag; ENP=effective number of parties.
Standard errors in parentheses.
***P<0.01, **P<0.05, *P<0.1.
Democratic attributes and infant mortality (without state dummies)
L.=one year lag; ENP=effective number of parties.
Standard errors in parentheses.
***P<0.01, **P<0.05, *P<0.1.
A series of additional robustness tests were carried out and the results are presented in the appendix. Table 5 in supplementary material shows the effect of the three democratic attributes on infant mortality among 28 Indian states.Footnote 37 Extending the analysis to 28 states enables us to see if the baseline results on the 15 major states are applicable to all states in India as well, despite the differences among the states. While ENP remains significant across all the models, voter turnout is significant in model 2 but insignificant in model 4. This may suggest that the effect of voter turnout may not apply to all the states in India. A plausible explanation could be that voting is not very informational for political representatives, as it does not inform officials about the specific needs and preferences of the masses (Verba et al., Reference Verba, Nie and Kim1978). Margin of victory is significant in model 3 but remains insignificant in model 4.
Table 6 in supplementary material presents findings with 2, 3, and 4-year lags. I did not extend the number of lags beyond 4 since elections are generally held every 5 years within states when values of the three independent variables of interest are likely to change. The findings indicate that the ENP is the only variable that remains statistically significant across all models. This is consistent with the baseline and all other robustness tests as well. Thus, the representational benefits of multiple parties are also prevalent over a longer period of time. Even though voter turnout is significant in the majority of models discussed above, it fails to reach statistical levels of significance in the extended lagged models. Voting only takes place during elections and while it is indicative of an active citizenry (Inkeles, Reference Inkeles1969; Verba et al., Reference Verba, Schlozman and Brady1995) that may not always be the case. Perhaps a sustained indication of citizens’ participation on a day-to-day basis such as participation in protests, contacting representatives, attending public meetings, may be better able to capture an active citizenry who can hold representatives accountable over a longer period of time. However, data constraints make it difficult to do so here. Margin of victory fails to reach statistical levels of significance in two of the three models. Contrary to expectations, it is negative and significant in model 1, which reiterates the lack of a definitive conclusion about the variable. Tables 7 and 8 in supplementary material assess the baseline models without state domestic product and population, respectively. Table 9 in supplementary material presents models with a one-step generalized method of moments dynamic estimation technique (Arellano and Bond, 1991). ENP continues to have a statistically significant effect on infant mortality while voter turnout and margin of victory do not reach statistical levels of significance in any of the models. Overall, the number of parties has the strongest effect on infant mortality, further attesting to the crucial influence of political parties in particular and political representation in general, in enhancing societal welfare.
Conclusion, implications, and future research
Understanding how democratic politics affects government performance is a core theme in comparative politics and political science at large (Dahl, Reference Dahl1971; Bueno de Mesquita et al., Reference Bueno de Mesquita, Smith, Siverson and Morrow2003; Lijphart, Reference Lijphart2012) and the spread of democracy around the world makes this research question as substantively important as it is theoretically intriguing. I argue that these incentives to perform well are contingent on the variation in three core democratic attributes – political representation, citizens’ participation, and electoral competition. The empirical analyses demonstrate that political representation plays the most important role in this regard. The paper contributes to the growing research on subnational studies that focus on democratic attributes and welfare outcomes within India (Besley and Burgess, Reference Besley and Burgess2002; Chhibber and Nooruddin, Reference Chhibber and Nooruddin2004; Sáez and Sinha, Reference Sáez and Sinha2009; Pushkar, Reference Pushkar2012).
This paper presents a first-cut analysis in exploring the links between three democratic attributes individually and infant mortality within a democracy, and the results demonstrate the significance of ENP. Analyzing the possibility of an interactive relationship between the three attributes and welfare outcomes could be a promising follow-up project. Alternatively, much like Indian states, democratic countries also vary in essential democratic attributes of representation, participation, and competition and future research can empirically test to see if variations in core democratic attributes can account for disparities in human welfare outcomes between democracies.
Like most studies, this paper has its limitations. One limitation is that while it attempts to explain state-level variations in infant mortality, it is unable to account for variations within states, for instance at the district level. Future research could explore the role of politics at a more disaggregated micro level. Additionally, as this paper primarily focuses on the role of politics within a democracy, it is unable to account for variations within non-democracies. For instance, non-democratic countries such as China and Cuba have made significant strides in improving public health outcomes. Provision of broad-based health services played an important role in reducing infant mortality rates within both countries. Presence of a strong labor movement and other mass organizations, a brief stint with democracy from 1940 to 1952, and consolidation of state power post-1959 in Cuba (Waitzkin, Reference Waitzkin1983; McGuire and Frankel, Reference Measham, Rao, Jamison, Wang and Singh2005) while Mao’s policies of universal welfare coverage, provision of health services at a low-cost, and health awareness campaigns in China (Hesketh and Zhu, Reference Hesketh and Zhu1997; Banister and Hill, Reference Banister and Hill2004) played an important role. Future research can further explore the determinants of infant mortality among non-democracies to get a comprehensive understanding of political dynamics and welfare outcomes across regime types.
The primary implication of this study is that instituting a minimalistic democratic regime that barely permits representation, participation, and competition may be insufficient to motivate political leaders to perform well. Democratic regimes need to be more representative, participative, and competitive in nature to meet the welfare needs of the citizens.Footnote 38 Ensuring that multiple segments of society can voice their concerns during the policy-making process through their elected representatives can enhance political representation. Existing research identifies the nature of the electoral system (Duverger, Reference Duverger1954) or societal cleavages (Eckstein, Reference Eckstein1963; Lipset and Rokkan, Reference Lipset and Rokkan1967, among others) as two determinants of party systems. While societal cleavages may be difficult to manipulate, adoption of a proportional electoral system may influence the number of parties. Additionally, reforming existing rules within proportional electoral systems may also increase the number of parties such as increasing district magnitude or reducing the electoral threshold so as to make it easier for additional parties to gain representation. The onus of ensuring high levels of citizens’ participation eventually lies with the citizens who need to use the opportunity given to them in a democratic society to hold their political representatives accountable through various forms of participation. Prevalence of free and fair elections can increase the likelihood of competitive elections. Additionally, parties that make a broader appeal to the masses can challenge the dominance of an incumbent party, thereby facilitating electoral competitiveness.
Acknowledgments
The author would like to thank the editors of the Journal, the anonymous reviewers, Paul T. Bellinger Jr, Dr Jonathan Krieckhaus, Dr Moises Arce, Dr Leslie Schwindt-Bayer, and Dr Byunghwan Son for their valuable feedback on earlier versions of this article.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1755773916000138
