The World Bank Economic Review, 36(4), 2022, 972–998 https://doi.org10.1093/wber/lhac020 Article Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 The Timing of Elections and Neonatal Mortality: Evidence from India Shampa Bhattacharjee Abstract This paper uncovers evidence of political cycles in developmental outcomes in the Indian context. Comparing children born to the same mother, it shows that children born 0–11 months before scheduled state legislative assembly elections have a significantly lower risk of neonatal mortality. The effect of being born just before elections is higher in politically more competitive regions. The paper provides some evidence of the channels behind this result. The usage of prenatal care increases before elections and mothers of children born before elections are more likely to have antenatal checkups and tetanus injections during pregnancy. Components of antenatal checkups, like the probability of having a blood test or an abdominal examination during pregnancy, also increase before elections. The improvement in child health outcomes before elections seems to be driven by a transfer of resources from non-election to election years rather than an overall improvement in child health outcomes. JEL classification: I12, I15, D72, O17 Keywords: neonatal mortality, child health; political cycles, democracy, elections 1. Introduction Theoretical research on political cycles emphasizes that opportunistic politicians create favorable condi- tions before elections to increase their chance of being elected (Nordhaus 1975; Lindbeck 1976; Rogoff and Sibert 1988; Persson and Tabellini 1990). This can have significant implications in developing coun- tries since the impact of economic conditions on the lives of poor people is likely to be high (Pallage and Robe 2003; Paxson and Schady 2005; Bhalotra 2010). This paper investigates the effect of the timing of scheduled state legislative assembly elections1 on neonatal mortality or deaths in the first month of life. The identification strategy compares neonatal mortality between siblings born 0–11 months before scheduled elections and those born in other years.2 Comparison across siblings controls selection issues that can arise from the fact that women who give Shampa Bhattacharjee is an assistant professor of Economics at Shiv Nadar University, Greater Noida, India. Her email ad- dress is shampa.bhattacharjee@snu.edu.in. The author thanks the editor, the referees, Siwan Anderson, Nicole Fortin, Patrick Francois, Ashok Kotwal, Amartya Lahiri, Marit Rehavi, Arka Roy Chaudhuri, and the participants at the Economics Empir- ical Lunch at the University of British Columbia, CEA conference, Montreal, and Annual Conference on Economic Growth and Development, Indian Statistical Institute, New Delhi for their constructive comments and suggestions. A supplementary online appendix is available with this article at The World Bank Economic Review website. 1 Scheduled elections are held after a gap of five years as per constitutional mandate. 2 The results are robust to using an instrumental variable approach (See the section Instrumental Variable Analysis of the supplementary online appendix). © The Author(s) 2022. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com The World Bank Economic Review 973 birth before elections can differ from those who give birth in off-election years. In addition to mother fixed effects, the estimating equation includes year, month, and order of birth fixed effects and state-specific linear year of birth trends. Using data from 25 Indian states, the paper shows that neonatal mortality is lower by about Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 0.5 percentage points for children born 0–11 months preceding scheduled elections. This is 11 percent of the mean neonatal mortality and implies about 0.15 million fewer annual deaths in the first month of life. The timing of unscheduled elections has no influence on neonatal mortality.3 The paper also shows that the effect of being born just before scheduled elections is more pronounced in districts where the state ruling party or coalition has a narrower absolute margin of victory. The results are robust to the inclusion of district-specific linear time trends and state-year fixed effects.4 The remaining threat to the identification strategy can arise because of unobservable time-varying mother- level characteristics. To see whether these vary before scheduled elections, the paper examines the effect of scheduled elections on fertility, sex composition of births, sex composition of previous births, birth interval, birth order, and mother’s age at birth. The results show that these do not change significantly before scheduled elections. India is a particularly relevant place to analyze the implications of election cycles. It is a developing country with high participation in state assembly elections.5 The elections in different states do not occur simultaneously, which allows me to identify the impact of elections separately from time-invariant effects. The constitution of India requires state elections to be held every five years. However, there have been some unscheduled elections. These are likely to be unanticipated and their timing can be endogenous (Khemani 2004). Thus, the paper focuses on scheduled elections. The paper explores medical care usage as a possible channel behind the effect of political cycles on neonatal mortality and shows that mothers of children born before scheduled elections are likely to have more frequent antenatal checkups and at least one tetanus injection during pregnancy. The paper also tests the effect of scheduled elections on specific components of antenatal checkups and finds that moth- ers of children born before scheduled elections are more likely to have a blood test and an abdominal examination during pregnancy. The results obtained in the paper imply that there is no overall improvement in child health. The paper finds no effects of the scheduled election cycle on state health expenditure, implying that the capacity of the health-care system does not increase before elections. The results are indicative of a transfer of resources from non-election to election years. This study has several contributions. Firstly, to the best of my knowledge, this is the first paper that analyzes the effect of political cycles on individual-level developmental outcomes either in developed or in developing country contexts. Previous research has focused on macroeconomic outcomes and policy in- struments.6 Politicians can take a number of measures before elections to improve their electoral prospects (Alesina, Roubini, and Cohen 1997). This can range from direct manipulation of government policy to outright illegal measures like the distribution of cash in exchange for votes. Thus, the effect of the electoral cycle on particular policy instruments might underestimate its overall effect. The combined effect of these measures might be better reflected in developmental outcomes. Secondly, compared to previous research on political cycles in India, this study uses a significantly larger sample: more than 150,000 children span- ning over 22 years and covering 81 scheduled elections.7 Moreover, since the data have information on 3 The insignificant correlation between the timing of unscheduled elections and the outcome is consistent with the litera- ture on political cycles in India (Vadlamannati 2015, 2009). 4 Since this paper uses monthly data, the variation is at the state-year-month level, allowing the inclusion of state-year fixed effects. 5 The states in my sample have an average voter turnout of about 60 percent. 6 See the section Existing Literature on the Impact of Political Cycles for a review of the literature. 7 One problem with Cole (2009) is that the paper covers a period of only eight years. Sáez and Sinha (2010) and Khemani (2004) use a longer time period, but their unit of observation is a state. 974 Bhattacharjee the birth histories of mothers, the regressions can include mother fixed effects.8 Another advantage of this study is that while the previous studies in India used yearly data, it used monthly data to uncover evidence of political cycles. This allows for the clean identification of the pre-election effects. This is not possible using yearly data since the data corresponding to the electoral year are composed of both pre- Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 and post-election months. This also allows me to divide the year before elections into finer intervals of 0–6 months and 6–12 months. This paper is divided into eight sections. The next section briefly outlines the institutional background. The third section describes the data used in this study. The fourth section outlines the empirical strategy. The fifth section discusses the results. Robustness checks are presented in the sixth section. The seventh section provides evidence for some of the possible mechanisms. The eighth section finally concludes the paper. 2. Institutional Background 2.1. State Assembly Elections in India State assembly elections are held in India to elect the Members of State Legislative Assemblies (MLAs). The party or coalition with a majority of seats in the state assembly elections is invited by the governor to form the state government. State elections are held in assembly constituencies. The number of constituencies can differ across states and is determined by the state population.9 The constitution of India requires that elections for state assemblies be held at five-year intervals.10 However, elections are not always held at five-year intervals. Unscheduled elections before five years are possible when legislators defect from the ruling party, or a coalition government breaks down. Unsched- uled elections can also arise because of political pressure from the central government.11 Article 356 of the constitution gives power to the president of India to dismiss the elected state government on the rec- ommendation of the central government and impose President’s rule in the state. This continues for a few months and is followed by unscheduled elections. Existing literature has documented the partisan use of President’s rule to dismiss state governments that are not aligned with the central government (Dua 1979; Khemani 2004). Khemani (2004) observes that in her sample,12 85 percent of elections, that occurred before the full expiry of the term, occurred in states where the state ruling party was not aligned with the party at the center and about 45 percent of these elections occurred after the imposition of President’s rule.13 In my sample, there are 139 state elections. Of these, 48 elections (34 percent) took place before the completion of the five-year term. When an unscheduled election takes place, the timing of the next scheduled election changes to five years from the unscheduled election. The fact that state elections are not perfectly synchronized is primar- ily due to the presence of unscheduled elections, since the first election in all states was in 1951.14 Khemani (2004) argues that the timing of unscheduled elections is both potentially endogenous and unanticipated 8 The unit of observation is district or state in most of the previous papers, allowing them to include district and state fixed effects only (Khemani 2004; Cole 2009; Sáez and Sinha 2010). 9 There are separate national elections to elect the members of the Lok Sabha, the lower house of the Indian Parliament. 10 Scheduled elections do not always occur in the same calendar month, which is precisely 60 months since the last election. In my sample, about 30 percent of the scheduled elections were not held in the same month as the previous election. 11 A few unscheduled elections, like the 1992 Punjab election, took place after the constitutionally mandated interval of five years. 12 Her sample consists of 16 major states of India over the period 1960–1994. 13 The imposition of Article 356 has significantly reduced after the Supreme Court of India judgment in the case of S. R. Bommai v. Union of India in 1994. The Supreme Court specified the conditions under which President’s rule can be imposed, restricting its scope (Joseph and Reddy 2004). 14 This is true for the states which were in existence in 1951. Some new states were created after 1951. Their year of the first election can also lead to a state-level difference in election cycle. The World Bank Economic Review 975 by the incumbent state government. Since scheduled elections take place after the completion of a full five- year term, the electoral calendar is exogenous and known in advance by all agents. Hence, opportunistic manipulation by politicians before scheduled elections is possible. On the other hand, unscheduled elec- tions are more likely to be sudden and thus allow politicians much less time to manipulate policies. The Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 timing of unscheduled elections can also be endogenous. For example, if unscheduled elections are held in years when the state is undergoing an economic boom, the effect of elections on neonatal mortality will be overestimated. Similarly, if unscheduled elections are held during a downturn, the effect of elections will be underestimated. The constitution of India assigns the powers and functions to the center and states. The central govern- ment is responsible for the items mentioned in the union list (like defense and foreign affairs). Similarly, state governments have exclusive jurisdiction over items mentioned in the state list (like public order, po- lice, and agriculture). The concurrent list includes areas of joint jurisdiction of the center and the states like education. The delivery of public health services is essentially a state responsibility (Berman 1998).15 Health-care workers almost always are state government employees (Singh 2008). Therefore, state governments can take measures to discipline health-care workers. 2.2. Existing Literature on the Impact of Political Cycles There exists a large literature on opportunistic political cycles. Empirical evidence on political cycles in developed countries is mixed. Most of these studies analyze the effects of electoral cycles on macroeco- nomic outcomes and fiscal instruments like taxes and government expenditure (McCallum 1978; Alesina 1988; Klein 1996; Berger and Woitek 1997; Galli and Rossi 2002; Veiga and Veiga 2007; Grier 2008; Katsimi and Sarantides 2012). Apart from these papers, previous studies have also analyzed political cycles in violence (Newman 2013; Aksoy 2014; Daniele and Dipoppa 2017), judicial sentencing (Dyke 2007; Berdejó and Yuchtman 2013; Park 2017), public sector hiring and earning (Borjas 1984; Levitt 1997; Tepe and Vanhuysse 2009; Mechtel and Potrafke 2013; Aaskoven 2016; Cahan 2019), and bank lending (Dinç 2005; Englmaier and Stowasser 2017). Specifically, in the context of health, a few papers have analyzed the effect of political cycles on health expenditure using cross-country data and show that health expenditure increases before elections in OECD countries (Potrafke 2010; Herwartz and Theilen 2014) and European countries (Castro and Martins 2018). Takaku and Bessho (2018) find that the hiring of doctors increases in public hospitals before mayoral elections in Japan. There is a growing literature documenting the presence of political cycles in developing countries. Several papers find evidence of political cycles in fiscal and monetary policies (Magloire 1997; Block 2002; Gonzalez 2002; Akhmedov and Zhuravskaya 2004; Brender and Drazen 2005; Guo 2009; Vergne 2009; Drazen and Eslava 2010; Aidt et al. 2020). Existing studies have also analyzed the effects of election timing on public sector employment (Inoue 2020), natural resource rents (Klomp and de Haan 2016), violence (Harish and Little 2017), corruption (Sidorkin and Vorobyev 2018), and crime (Meloni 2018) in the context of developing countries. In the Indian context, Cole (2009) shows that bank lending follows the electoral cycle, with agri- cultural credit increasing by 5–10 percent points in an election year. Khemani (2004) finds evidence of political cycles both in public service delivery such as road construction and in fiscal instruments such as commodity taxes. Fagernäs and Pelkonen (2018) find that teacher hiring and teacher transfers in- crease after assembly elections. Davies (2021) finds evidence of reduced absenteeism of public school teachers in the year before elections and an increase in teacher absenteeism in the year after elections. 15 The union list includes health-related research and scientific and technical education. The concurrent list includes issues with wider national ramifications, like preventing infectious diseases from spreading across state boundaries (Gupta and Rani 2004). 976 Bhattacharjee Baskaran, Min, and Uppal (2015) and Min and Golden (2014) show that electricity provision increases before elections. Vadlamannati (2015) provides evidence of increased efforts by incumbent governments to control corruption during scheduled election years and Vadlamannati (2009) shows that communal violence increases before scheduled elections in India. No such effects on corruption and violence are seen Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 for unscheduled elections. Sáez and Sinha (2010) analyze the effects of electoral cycles on different types of government expenditure and find that state health expenditure falls in election years. 3. Data The micro-data used in this survey come from the second round of the National Family Health Survey of India (NFHS-2) conducted in 1998–1999 (IIPS 2000). This is a nationally representative survey. The data set contains complete fertility history for ever-married women aged 15–49 in 1998–1999. The data include information on the year, month, and order of birth of all live births and the time and incidence of all child deaths. The survey has information on the state of residence of the household during the time of the survey. The state of birth of the child might be different from the state of the residence of the household in 1998–1999. To avoid this problem, the study excludes women reported as visitors in the survey and considers births only after the woman started living in their current residence (Bhalotra and Clots-Figueras 2014). The results are robust to the inclusion of these births (see the section Robustness to the Sample Restrictions). Since twins have a significantly high mortality rate compared to singletons (Buekens and Wilcox 1993), the study considers only women with singleton births (Anukriti 2018; Bhalotra 2010). The study also excludes women who gave birth before the age of 13 since they are likely to be significantly different from other women (Anukriti 2018). The results are robust to the inclusion of women with twin births and underage births (see the section Robustness to the Sample Restrictions). This data set is used to construct individual child-level indicators of neonatal mortality across time. The neonatal mortality variable is defined as a dummy indicating whether the child died by the age of 1 month. I multiply the neonatal mortality dummy by 100 for ease of representation in tables. The estimation sample contains more than 150,000 children, born to 45,687 mothers over the period 1977– 1999 across 25 Indian states.16 Out of these 45,687 women, 15,908 women gave birth both in the interval corresponding to 0–11 months before scheduled elections and in other years. The study only considers post-1977 data because constituency boundaries remained fixed over this period. In addition, Bhalotra et al. (2014) state that the nature of political parties was very different before 1977. Restricting the data to the post-1977 period also limits recall issues (Bhalotra 2010). The study excludes children born less than a month before the date of the survey since these children have not been completely exposed to the threat of neonatal mortality. The unit of observation in this paper is an individual child. Table 1 shows that the average neonatal mortality over the sample is 47 per 1,000 individuals.17,18 We also see that the proportion of first births as a fraction of all births is smaller than the proportion of second births. This might be due to the fact that I am considering children born after 1977 and thus first-born children for some mothers might not be included in the sample. The sample averages of all individual-level controls used in the neonatal mortality regressions are reported in table 1. The election data come from the official website of the Election Commission of India (ECI 2017). The data include information on the year-month of the election, the identity of the contestant and their party 16 The main analysis includes all states that were in existence in India during the time of the survey. The results are robust to the inclusion of only those states that existed for the entire period 1977–1999 (see the section State Definition of the supplementary online appendix). 17 This can be compared with the global average of 32 per 1,000 in 1990. 18 There is substantial variation in neonatal mortality across states. Kerala has the lowest incidence of neonatal deaths at 18 per 1,000 while Uttar Pradesh has the highest rate at 62 per 1,000. The World Bank Economic Review 977 Table 1. Summary Statistics Variable Mean Std. dev. Observations Neonatal mortality (per 100) 4.652 21.062 169,527 Infant mortality 8.061 27.223 169,527 Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election 0.113 0.317 169,527 Born 12–23 months before scheduled election 0.126 0.332 169,527 Born 1–12 months after scheduled election 0.117 0.321 169,527 Born 13–24 months after scheduled election 0.121 0.326 169,527 Born 0–11 months before unscheduled election 0.112 0.315 169,527 Girl child 0.482 0.5 169,527 Mother’s age at birth 23.174 5.144 169,527 Birth order 1 0.226 0.418 169,527 Birth order 2 0.253 0.435 169,527 Birth order 3 0.196 0.397 169,527 Birth order 4 0.133 0.34 169,527 Birth order 5 0.084 0.277 169,527 Birth order 6 0.051 0.22 169,527 Birth order 7 0.028 0.166 169,527 Birth order 8 0.015 0.121 169,527 Birth order 9 0.008 0.087 169,527 Month of birth 1 0.071 0.257 169,527 Month of birth 2 0.065 0.247 169,527 Month of birth 3 0.08 0.272 169,527 Month of birth 4 0.078 0.268 169,527 Month of birth 5 0.081 0.273 169,527 Month of birth 6 0.087 0.281 169,527 Month of birth 7 0.085 0.279 169,527 Month of birth 8 0.103 0.305 169,527 Month of birth 9 0.09 0.286 169,527 Month of birth 10 0.096 0.294 169,527 Month of birth 11 0.088 0.283 169,527 Month of birth 12 0.075 0.264 169,527 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: The table reports means and standard deviations of variables used in the neonatal mortality regressions. Neonatal mortality is defined as the number of deaths in the first month of life per 100 births. Figure 1. Scheduled Election Cycle Dummies Source: Author’s analysis. Note: The figure shows the construction of the scheduled election dummies. Each point on the line represents a time point. End of months are marked on the line. SE denotes scheduled elections and USE denotes unscheduled elections. 0 indicates 0–11 months before scheduled elections, −1 indicates 12–23 months before scheduled elections, +1 indicates 1–12 months after scheduled elections, +2 indicates 13–24 months after scheduled elections, and −2 indicates more than 23 months before scheduled elections. affiliation. Using this data, I construct dummies for whether a given month-year is 0–11 months before scheduled elections, 12–23 months before scheduled elections, 1–12 months after scheduled elections, and 13–24 months after scheduled elections. Figure 1 illustrates the construction of the scheduled election cycle dummies through an example. The line represents different time points. Each tick denotes the endpoint of a distinct month. We start from the time point 0 which represents the endpoint of month 0. SE denotes a 978 Bhattacharjee scheduled election and USE denotes an unscheduled election. The interval 0 denotes 0–11 months before a scheduled election. The interval −1 indicates 12–23 months before scheduled elections. The category +1 denotes 1–12 months after scheduled elections. The interval +2 indicates 13–24 months after scheduled elections and −2 indicates more than 23 months before scheduled elections. By construction, the biggest interval is −2. Thus, the proportion of children born more than 23 months before scheduled elections is Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 likely to be higher than children born in the other categories. The paper tests whether the effect of the timing of scheduled elections on neonatal mortality is higher in politically more competitive regions. A region is defined to be politically competitive if the state ruling party/coalition has a narrower margin of victory in that region. The NFHS data report the district of residence of the respondent but not the political constituency. A district is composed of a number of con- stituencies, on average nine in my sample. Thus, the margin of victory of the state ruling party/coalition is defined at the district level. The margin of victory of a district is the weighted average of the constituency level absolute margin of victory of the ruling party or ruling coalition in the previous election.19 In con- structing the weighted average, the weights used are the number of electors in the constituency divided by the total electors of the district. Competitive districts are defined as districts with margin of victory less than the 10th percentile.20 The Election Commission of India website provides information on party affiliations of individual contestants in an election. However, the names of the ruling party or parties in the ruling coalition are not available on the election commission website. This information is essential for computing the political competition variable which is used in this analysis. Reports from the Times of India newspaper are used to ascertain the name of the party (or all the parties in the case of coalition governments) that ruled a state.21 Table 1 reports the distribution of births across scheduled election years and other years. If there were no unscheduled elections, it is expected that the proportion of children born in each of the five years of the electoral cycle will be more or less equal, each of them being close to 1/5 or 0.2. The presence of unscheduled elections implies that the proportion of children born more than 23 months before scheduled elections will be higher than children born in the other categories, as suggested in fig. 1. Table 1 shows that the proportion of children born in each reported category is less than 0.2. This implies that the proportion of children born more than 23 months before scheduled elections is more than 0.2, which is higher than those born before in any other years of the electoral cycle. In order to analyze the possible mechanisms behind the fall in neonatal mortality, the paper examines the impact of the timing of scheduled elections on the following outcomes: number of antenatal checkups during pregnancy, whether the mother had a tetanus injection during pregnancy, whether the mother was given iron and folic acid supplements during pregnancy, whether the woman received assistance from doctors or nurses or traditional birth attendants during delivery, and whether the birth took place in a hospital. The paper also finds the effects on the components of antenatal checkups, like the likelihood of having blood and urine tests, abdominal examination, internal examination, ultrasound, amniocentesis, and weight checks during pregnancy. Information on these variables is available only for the last two children born after 1995 in NFHS-2 data. 19 The incumbent’s margin of victory in a constituency is defined as the difference in vote share between the state ruling party/coalition and the maximum vote share among all other candidates. Margin of victory is defined as 1 if the incum- bent ran uncontested and negative of the vote share of the winner if no incumbent candidate ran for the election (Cole 2009). 20 Instead of a continuous measure, the paper uses a dummy variable for political competition since targeting resources across districts will only happen if the margin of victory is sufficiently narrow. The results are robust to other definitions of competitive districts (see tables 9 and S10.4 in the supplementary online appendix). 21 The Times of India is India’s “paper of record” (Wilkinson 2006). The World Bank Economic Review 979 Table 2. Summary Statistics: Channel Regressions Variable Mean Std. dev. Observations Outcome variables Neonatal mortality (per 100) 3.598 18.624 44,969 Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Tetanus during pregnancy 0.753 0.431 31,895 Antenatal visits 2.831 3.236 31,889 Iron or folic acid supplements 0.587 0.492 31,985 Weight checked during pregnancy 0.364 0.481 31,979 Blood test during pregnancy 0.4 0.49 31,979 Urine test during pregnancy 0.375 0.484 31,979 Abdomen examined during pregnancy 0.521 0.5 31,978 Internal examination during pregnancy 0.236 0.424 31,979 Ultrasound during pregnancy 0.122 0.328 31,978 Amniocentesis 0.013 0.115 31,954 Assistance during delivery: doctors/nurses 0.427 0.495 31,988 Assistance during delivery: traditional birth attendant 0.387 0.487 31,988 Hospital delivery 0.342 0.474 31,983 Independent variables Born 0–11 months before scheduled election 0.141 0.348 44,969 Girl child 0.479 0.5 44,969 Scheduled castes 0.182 0.386 44,572 Scheduled tribes 0.16 0.366 44,572 Muslim 0.145 0.353 44,926 Urban 0.245 0.43 44,969 Mother’s age at birth 23.827 5.414 44,969 Single births 0.987 0.114 44,969 Source: The sample includes children born in the years 1995-1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: The table reports means and standard deviations of the variables used in regressions showing the mechanisms. All outcome variables, except Antenatal visit, are dummy variables equal to 1 if the mother had used the particular medical care during pregnancy or delivery. Antenatal visit is the total number of antenatal visits during pregnancy. Membership in the Scheduled Caste and Scheduled Tribe categories is indicated by the dummy variables Scheduled caste and Scheduled tribe. The dummy variable Muslim is equal to 1 if the mother is Muslim. Urban is a dummy variable indicating urban area of residence. Single births (as opposed to multiple births) are indicated by the dummy variable Single birth. Table 2 shows the mean of the outcome variables as well as the individual-level controls used in these regressions. The mean neonatal mortality reported in table 2 is lower than the mean neonatal mortality reported in table 1. This is expected since the sample for these regressions corresponds to the latter part of the original sample used in table 1. Thus the mean neonatal mortality in the sample used for the channels regression is lower since neonatal mortality shows a declining trend over time. 4. Empirical Strategy The basic estimating equation is yimdstn = α + β Estn + φ Ximdstn + μm + τt + θn + πs × t + imdstn , (1) where yimdstn is a dummy variable that indicates whether the child i, born to mother m, in district d of state s in year t and month n died in the first month of life. The variable of interest is Estn , which is a dummy variable equal to 1 if the month of birth n is between 0 and 11 months before the month of a scheduled election. The equation includes mother fixed effects (μm ), month of birth fixed effects (θ n ), and year of birth fixed effects (τ t ). Mother fixed effects take account of the selection issues associated with the type of mothers who give birth before elections. The equation also includes state-specific year of birth trends, π s 980 Bhattacharjee × t. The vector Ximdstn denotes other controls like a dummy for the sex of the child, order of birth fixed effects, and mother’s age at birth. These controls account for the variation in death risk among children born to the same mother. The standard errors are clustered at the state level. The number of states is 25 in the sample. Since the number of clusters is small, the paper reports the p-values obtained by implementing Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 a wild cluster bootstrap, which accounts for the relatively small number of clusters.22 Here, it is important to distinguish between scheduled elections and unscheduled elections. The timing of an unscheduled election is likely to be endogenous since its timing, unlike that of a scheduled election, is not predetermined; rather, it is determined by the political players involved. Unscheduled elections are also likely to be sudden to the incumbent government. So the government has less time to adjust policies. Thus, the paper uses only scheduled elections.23 We expect the coefficient of the election dummy, β in equation (1), to be negative and statistically significant. The paper uses the 0–11 months interval as the relevant window for the scheduled election dummy. The Election Commission of India imposes the Model Code of Conduct just before elections, which makes it difficult for political parties to indulge in unfair practices like bribing voters and using official resources for political gains (Singh 2012). Thus, taking short intervals just before elections might not capture the full extent of pre-election political manipulation.24 The retrospective nature of the NFHS data used to generate the neonatal mortality variable might lead to recall bias. Restricting the data to the post-1977 period controls for this to some extent. More- over, several quality checks are included in NFHS to ensure that the data is of reliable quality (Bhalotra 2010). Another problem with this data is that as we go back in time, we observe a greater proportion of younger mothers (Bhalotra 2010) in the sample, which might lead to an increased proportion of high-risk teenage births. This problem is less severe in this paper since the sample is restricted to women whose first childbirth was above the age of 13. In addition, all regressions control for the mother’s age at birth. The paper also estimates the effect of the entire election cycle on neonatal mortality. The estimating equation is yimdstn = α + β0 Estn + β−1 E−1stn + β+1 E+1stn + β+2 E+2stn + φ Ximdstn + μm + τt + θn + πs × t + imdstn , (2) where Estn is a dummy variable which is equal to 1 if the month of birth n corresponds to 0–11 months before a scheduled election, E−1stn is a dummy variable which is equal to 1 if the month of birth n lies between 12 and 23 months before a scheduled election, E+1stn is a dummy variable equal to 1 if the month of birth n is between 1 and 12 months after a scheduled election, E+2stn is a dummy variable equal to 1 if the month of birth n is between 13 and 24 months after a scheduled election. The omitted category includes the children born two or more years before scheduled elections. 5. Results 5.1. Neonatal Mortality Table 3 shows the effect of the scheduled election cycle on neonatal mortality. Column 1 of table 3 presents the estimates of equation (1). Column 1 shows that neonatal mortality is lower for children born 0–11 months before scheduled elections. Children born 0–11 months before scheduled elections have over 22 The p-values are obtained using the Stata command boottest. This method does not produce standard errors (Roodman et al. 2019). 23 Previously Brender and Drazen (2005) and Katsimi and Sarantides (2012) have considered the effects of only sched- uled elections to account for the endogeneity of the timing of elections. Section Instrumental Variable Analysis of the supplementary online appendix analyzes the effect of elections on neonatal mortality using an instrumental variable approach. 24 Section Role of Political Competition shows that the distribution of resources is more targeted closer to elections. The World Bank Economic Review 981 Table 3. Scheduled Elections and Neonatal Mortality (1) (2) (3) (4) (5) (6) District-specific State-year Baseline linear time trends fixed effects Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.5172** −0.5833** −0.5094** −0.5751** −0.5856** −0.8601* (0.0253) (0.0234) (0.0284) (0.0238) (0.0048) (0.0388) [0.0361] [0.0368] [0.0403] [0.0379] [0.0280] [0.0723] Born 12–23 months before scheduled election −0.0364 −0.0360 −0.3181 (0.8387) (0.8458) (0.5931) [0.8507] [0.8603] [0.6084] Born 1–12 months after scheduled election 0.0354 0.0292 −0.2236 (0.8951) (0.9156) (0.7693) [0.8984] [0.9158] [0.7811] Born 13–24 months after scheduled election −0.3160 −0.3093 −0.5324 (0.2203) (0.2235) (0.3296) [0.2572] [0.2564] [0.4059] Observations 154,334 154,334 154,334 154,334 154,334 154,334 Mean of dependent variable 4.8456 4.8456 4.8456 4.8456 4.8456 4.8456 Number of mothers 45,687 45,687 45,687 45,687 45,687 45,687 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each column represents a separate regression. The dependent variable is neonatal mortality. Columns 1 and 2 report coefficients on dummies for children born before and after scheduled elections. Columns 3 and 4 include district-specific linear time trends in place of state-specific time trends used in columns 1 and 2. Columns 5 and 6 include state-year fixed effects in the regressions. Apart from the reported variables, all regressions include mother fixed effects, dummies for year of birth, month of birth, order of birth, and sex of the child and control for mother’s age at birth. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. 11 percent more survival chance in the first month of life. Column 2 of table 3 shows the estimates of equation (2), where the regression includes dummies for the entire election cycle. Column 2 shows that children born 0–11 months before scheduled elections have significantly lower mortality risk. The coefficients on the dummies indicating other years in the electoral cycle are all statistically insignificant. To benchmark my results with the existing literature, the effect size is comparable to Kudamatsu (2012), who analyzes the effect of democratization in Sub-Saharan Africa on infant mortality in the post–Cold War period. The paper compares the infant mortality rates of children born to the same mother before and after democratization and finds that democratization led to a decline in infant mortality by 12 percent of the sample mean. The baseline specification includes state-specific linear time trends. Columns 3 and 4 show that the results are robust to the inclusion of district-specific time trends. Similarly, columns 5 and 6 show that the results are robust to the inclusion of state-year fixed effects. It is to be noted here that since this paper uses monthly data, the variation in the election dummy comes at the state-month-year level and not at the state-year level. This allows me to include state-year fixed effects, since the election dummy corresponds to 0–11 months before an election and is not specific to a particular year. Figure 2 illustrates the effect of scheduled elections over shorter intervals. The horizontal axis indicates quarters of birth relative to scheduled elections. Each quarter on the horizontal axis denotes children born between x − 1 and x quarters before or after scheduled elections. Quarter 0 on the horizontal axis de- notes children born 0–2 months before scheduled elections. Similarly, quarter −1 denotes children born 3–5 months before scheduled elections and quarter 1 denotes children born 1–3 months after scheduled elections. The vertical axis shows estimates from a specification similar to (1) but includes 16 dummies indicating the quarter of birth of children born two years before and two years after scheduled elections. 982 Bhattacharjee Figure 2. Election and Neonatal Mortality 2 Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 1.5 1 .5 Coefficients 0 −.5 −1 −1.5 −2 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8 Quarters Relative to Scheduled Election Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: The graph reports coefficients and 95 percent confidence intervals of estimates obtained by regressing neonatal mortality on distance in quarters from the month of birth to the month of election. The figure shows that while children born 0–2 months before scheduled elections do not experience any decline in neonatal mortality, children born 3–11 months before scheduled elections experience a signif- icant decrease in neonatal mortality. The insignificant effect for the quarter immediately before elections might be due to the imposition of the Election Commission of India’s Model Code of Conduct just before elections. 5.2. Accounting for Unscheduled Elections One problem with comparing children born before scheduled election years with children not born before scheduled elections is that the comparison group consists of children born before unscheduled elections and those born in off-election years. The paper addresses this issue in two ways. Firstly, equations (1) and (2) are estimated, controlling for the effect of unscheduled elections by in- cluding a dummy variable equal to 1 if child i is born 0–11 months before unscheduled elections. Columns 3 and 4 of table 4 present the results. The results are again similar to the baseline results, presented in columns 1 and 2. Moreover, we see that the coefficient on the unscheduled election dummy is statistically insignificant. The insignificant correlation between the timing of unscheduled elections and the outcome is consistent with the literature on political cycles in India (Vadlamannati 2015, 2009). Khemani (2004) finds that the correlation between unscheduled elections and outcome variables is opposite in sign to the effect of scheduled elections and observes that this is due to the possible endogeneity in the timing of unscheduled elections, combined with its sudden nature. Factors correlated with political volatility can be correlated The World Bank Economic Review 983 Table 4. Accounting for Midterm Elections (1) (2) (3) (4) (5) (6) Including unscheduled Dropping unscheduled Baseline election dummy elections Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.5172** −0.5833** −0.4843** −0.5296* −0.4939* −0.5432* (0.0253) (0.0234) (0.0311) (0.0319) (0.0391) (0.0452) [0.0361] [0.0368] [0.0447] [0.0514] [0.0536] [0.0695] Born 12–23 months before scheduled election −0.0364 0.0101 −0.0640 (0.8387) (0.9531) (0.7427) [0.8507] [0.9587] [0.7565] Born 1–12 months after scheduled election 0.0354 0.0894 0.0651 (0.8951) (0.7455) (0.8438) [0.8984] [0.7594] [0.8581] Born 13–24 months after scheduled election −0.3160 −0.2936 −0.2716 (0.2203) (0.2437) (0.3601) [0.2572] [0.2809] [0.4056] Born 0–11 months before unscheduled election 0.3043 0.3153 (0.3098) (0.2894) [0.3671] [0.3514] Observations 154,334 154,334 154,334 154,334 134,723 134,723 Mean of dependent variable 4.8456 4.8456 4.8456 4.8456 4.8437 4.8437 Number of mothers 45,687 45,687 45,687 45,687 42,317 42,317 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each column represents a separate regression. The dependent variable is neonatal mortality. Columns 1 and 2 are similar to columns 1 and 2 of table 3. Columns 3 and 4 include a dummy for whether a child is born 0–11 months before unscheduled elections in addition to the scheduled election dummies. Columns 5 and 6 also report coefficients on the scheduled elections dummies but the sample excludes children born 0–11 months before unscheduled elections. Apart from the reported variables, all regressions include mother fixed effects, dummies for year of birth, month of birth, order of birth, and sex of the child. The regressions also include state-specific linear time trends and control for mother’s age at birth. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. with the timing of unscheduled elections as well as the outcome variables. This paper finds an insignificant effect of unscheduled elections. However, the sign of the estimate is positive. This might be due to the fact that political volatility is correlated with the worsening of health outcomes. Secondly, equations (1) and (2) are estimated after dropping all children born 0–11 months before unscheduled elections so that the control group consists only of children born in the off-election years. The results are presented in columns 5 and 6 of table 4. It can be seen that the results are robust to the exclusion of children born before unscheduled elections. 5.3. Infant and Post-Neonatal Mortality Infant mortality is mortality during the first 12 months of life and post-neonatal mortality is mortality during months 2–12 of an infant’s life. Table 5 presents the results separately for infant and post-neonatal mortality dummy variables.25 , 26 Similar to neonatal mortality, I multiply these dummy variables by 100 for ease of representation in tables. The first two columns of table 5 show the baseline results for neonatal mortality and are the same as columns 1 and 2 of table 3. Columns 3 and 4 show estimates for infant 25 Infant mortality is defined as a dummy equal to 1 if the child dies during the first 12 months of life. Post-neonatal mortality is a dummy equal to 1 if the child dies by the age of 2–12 months. 26 Section Alternate Mortality Measures of the supplementary online appendix presents the effect of scheduled elections on mortality over shorter intervals. 984 Bhattacharjee Table 5. Neonatal, Infant, and 2–12 Month Mortality (1) (2) (3) (4) (5) (6) Neonatal Infant 2–12 months mortality mortality mortality Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.5172** −0.5833** −0.6653 −0.6885 −0.1957 −0.1837 (0.0253) (0.0234) (0.1061) (0.1171) (0.4376) (0.4729) [0.0361] [0.0368] [0.1441] [0.1723] [0.4981] [0.5355] Born 12–23 months before scheduled election −0.0364 0.0518 0.0254 (0.8387) (0.8589) (0.8878) [0.8507] [0.8767] [0.8918] Born 1–12 months after scheduled election 0.0354 0.1551 0.0396 (0.8951) (0.6141) (0.8534) [0.8984] [0.6397] [0.8624] Born 13–24 months after scheduled election −0.3160 −0.2578 0.0121 (0.2203) (0.5152) (0.9499) [0.2572] [0.5465] [0.9567] Observations 154,334 154,334 154,334 154,334 146,568 146,568 Mean of dependent variable 4.8456 4.8456 8.4112 8.4112 3.7441 3.7441 Number of mothers 45,687 45,687 45,687 45,687 44,796 44,796 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each column represents a separate regression. The dependent variable is neonatal mortality in columns 1 and 2, infant mortality in columns 3 and 4, and 2–12 month mortality in columns 5 and 6. The table reports coefficients on dummies for the scheduled election cycle. Apart from the reported variables, all regressions include mother fixed effects, dummies for year of birth, month of birth, order of birth, and sex of the child. The regressions also include state-specific linear time trends and control for mother’s age at birth. Children who died in the first month of life are excluded from the sample in columns 3 and 4. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. mortality and columns 5 and 6 show the effect on post-neonatal mortality. The results show that the effects are statistically insignificant for infant and post-neonatal mortality. Since I am interested in the effects of being born before elections on child-level mortality outcomes, it makes more sense to focus on neonatal mortality since the post-neonatal period of a child born before an election is likely to have some exposure to the post-election months. It is to be noted that the estimates of post-neonatal mortality can suffer from selection issues since children will enter the post-neonatal period only if they survive the neonatal period. This might be a serious issue given the high rate of neonatal mortality in India. 5.4. Role of Political Competition Existing literature suggests that political competition affects the allocation of public funds and public service delivery (Bardhan and Mookherjee 2010; Gupta and Mukhopadhyay 2016). Politicians may have an incentive to behave more opportunistically when their margin of victory is small. To examine the role of political competition, the paper analyzes whether the intensity of political cycles depends on the margin of victory of the current state ruling party or coalition in the previous election. Targeting of resources across districts is expected to happen if the margin of victory is sufficiently narrow. Thus, for analyzing the effect of political competition, low-margin districts are defined as districts with the district-level weighted margin of victory being less than or equal to the 10th percentile and high- margin districts as districts with the district-level weighted margin of victory being greater than the 10th percentile. The World Bank Economic Review 985 Table 6. Role of Political Competition (1) (2) (3) (4) 0–11 months before 0–5, 6–11 months before baseline with interactions baseline with interactions Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.5172** −1.6129* (0.0253) (0.0712) [0.0361] [0.0917] Born 0–11 months before × high margin dummy (>10th percentile) 1.2858 (0.1542) [0.1739] Born 0–5 months before scheduled election −0.3559 −2.7404*** (0.2288) (0.0091) [0.2538] [0.0064] Born 0–5 months before × high margin dummy (>10th percentile) 2.7526** (0.0176) [0.0257] Born 6–11 months before scheduled election −0.6765** −0.5696 (0.0148) (0.5606) [0.0384] [0.5713] Born 6–11 months before × high margin dummy (>10th percentile) −0.0690 (0.9398) [0.9405] High margin dummy (>10th percentile) −0.4582** −0.4514** (0.1259) (0.1293) [0.0417] [0.0446] Observations 154,334 150,677 154,334 150,677 Mean of dependent variable 4.8456 4.8775 4.8456 4.8775 Number of mothers 45,687 45,047 45,687 45,047 Source: Sample includes children born in the years 1977–1999 in 25 states. Data come from the second round of National Family Health Survey and Election Commission of India website. Note: Each column represents a separate regression. The dependent variable is neonatal mortality. Column 1 is similar to column 1 of table 3. Column 3 reports the coefficients of dummy variables for children born 0–5 months and 6–11 months before scheduled elections separately. Columns 2 and 4 report the coefficients of the election dummies as well as the interactions of the election dummies with the high margin of victory dummy. Apart from the reported variables, all regressions control for mother’s age at birth and include mother fixed effects, dummies for year of birth, month of birth, birth order, and sex of the child, and state-specific linear time trends. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. The following equation tests the role of political competition: Yimdstn = α + β Estn + ψ Estn × Adstn + ρ Adstn + γ Ximdstn + μm + τt + θn + πs × t + imdstn , (3) where Adstn is a dummy variable denoting districts with a high margin of victory in the previous election. In this specification, β denotes the effect of scheduled elections on close-election districts and ψ denotes the additional effect of scheduled elections on districts with a high margin of victory in the previous election. We expect β to be negative and statistically significant and for ψ to be positive and statistically significant. The results of targeting based on political competition are presented in table 6. Column 1 of table 6 is similar to column 1 of table 3 and shows the effect of being born 0–11 months before a scheduled election. Column 2 of table 6 shows the estimates from equation (3). It can be seen that the coefficient 986 Bhattacharjee Table 7. Heterogeneity Tests (1) (2) (3) (4) Panel A Muslim Non-Muslim SC/ST Non-SC/ST Born 0–11 months before scheduled election −0.4576 −0.5479** −0.9450*** −0.4174 Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 (0.3329) (0.0169) (0.0091) (0.2330) [0.2945] [0.0306] [0.0054] [0.2672] Observations 22,389 131,805 50,501 76,542 Mean of dependent variable 3.9374 5.0122 5.1339 4.9788 Number of mothers 5,786 39,861 14,279 23,896 (1) (2) Panel B Female Male Born 0–11 months before scheduled election −0.1869 −1.0534** (0.5975) (0.0181) [0.6089] [0.0414] Observations 58,107 63,558 Mean of dependent variable 4.8944 5.6878 Number of mothers 21,793 25,013 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each cell represents a separate regression. The dependent variable is neonatal mortality. The table reports the coefficient of the scheduled election dummy. Panel A reports the results for Muslims in column 1, non-Muslims in column 2, SC/STs in column 3, and non-SC/STs in column 4. Results for females and males are reported in columns 1 and 2 of panel B. All regressions include mother fixed effects, dummies for year of birth, month of birth, order of birth, and sex of the child. The regressions also include state-specific linear time trends and control for mother’s age at birth. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. of the scheduled election dummy is negative while the coefficient of the interaction between Adstn and the scheduled election dummy is positive, as expected. However, the estimate of the interaction term is statistically insignificant at conventional levels of significance. As the year before an election starts, there may be an overall increase in politicians’ efforts, resulting in increased availability of resources across all constituencies. However, as elections draw near, resources can be specifically targeted to close-election constituencies. This is particularly relevant because of the Election Commission of India’s Model Code of Conduct, imposed just before elections. Since the allocation of resources to all constituencies might be difficult just before elections because of the Model Code of Conduct, the diversion of resources can be targeted to areas of high political competition. Thus politicians might target only constituencies with a smaller margin of victory just before elections. To test whether districts are targeted based on the level of political competition just before elections, I divide children born 0–11 months before scheduled elections into those born 0–5 months before and those born 6–11 months before scheduled elections. Column 3 of table 6 shows the estimates from equation (1), where the 0–5 and 6–11 month dummies are included separately. Column 4 of table 6 includes the inter- actions of Adstn with the 0–5 and 6–11 month dummy variables. The results show that neonatal mortality falls for children born 0–5 months before a scheduled election in close-election districts. 5.5. Heterogeneous Effects Table 7 analyzes heterogeneity in the effects of scheduled elections across different groups. Panel A presents the results for Muslims, non-Muslims, Scheduled Castes (SC) and Scheduled Tribes (ST), and non-SC/ST.27 Across these four groups, neonatal mortality is highest for SC/STs. The table shows that 27 Muslims are the second-largest religious group in India. The constitution of India recognizes some traditionally dis- advantaged castes (Scheduled Castes) and certain culturally distinct tribes (Scheduled Tribes) as requiring additional consideration. The World Bank Economic Review 987 the effect of electoral cycles is higher for SC/STs, which might be because they have a higher margin for improvement. Panel B shows the result separately for male and female children. The results show that the effect is primarily driven by males. Females have lower neonatal mortality rates in my sample, which might reflect Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 their biological advantage (Naeye et al. 1971). Thus the higher effect for males might be the result of the fact that they have higher scope for improvement. However, the result can also reflect gender bias against female children (Murthi, Guio, and Dreze 1995; Gandhi Kingdon 2002; Pande 2003; Jayachandran and Kuziemko 2011; Azam and Kingdon 2013; Bharadwaj and Lakdawala 2013). 6. Robustness Checks 6.1. Mother and Birth-Level Characteristics This section explores the possibility that the results presented in this paper are driven by the sex composi- tion of births, fertility, and pre-existing birth characteristics, such as sex composition of previous children, birth interval, mother’s age at birth, and order of birth. For analyzing the effect of scheduled elections on the sex composition of births, equation (1) is estimated with the dependent variable being a dummy for whether the child born is a female. To check the effect on fertility, NFHS data is used to generate a woman year-month panel such that each woman is observed for each month-year, starting from the beginning of the year she turns 15 and till the month-year of the survey. The dependent variable is a dummy variable equal to 1 if the woman gave birth in a given month-year. I estimate an equation similar to (1) with the birth dummy as the dependent variable. The independent variable is a dummy equal to 1 if the month-year corresponds to 0–11 months prior to scheduled elections. For each child, sex composition of previous births is defined as the proportion of females among the children born to the mother previously. Birth interval is the difference in months between the last birth and the current birth. The paper considers two dummy variables for the order of birth: whether the child born is of birth order greater than 1 and whether the birth order is greater than the mean birth order in the sample. Table 8 shows the estimated results. The estimates are statistically insignificant. 6.2. Robustness to the Definition of Competitive Districts Previously, competitive districts were defined as districts with margin of victory less than the 10th per- centile. This section shows that the results are robust to alternate definitions of competitive districts. Table 9 shows that the effect size is similar if we define close-election districts as those with margin of vic- tory less than the 15th percentile and the 1st percentile. However, the effects become imprecisely estimated if we define the competitive districts as districts with margin of victory less than the 1st percentile.28 The paper also examines the effect at different levels of margin of victory (Callen et al. 2016). For this, the following variables are defined: a dummy indicating margin of victory between the 10th and 50th percentiles, a dummy indicating margin of victory between the 50th and 90th percentiles, and a dummy indicating margin of victory greater than the 90th percentile. Column 4 of table 9 includes the interaction of these dummy variables with the election dummy. We can see that the coefficient of the election dummy is positive and the interaction between the election dummy and the three margin of victory dummies are all positive. Since the interaction terms are almost equal in magnitude to the main effect of the election 28 In the main analysis, the margin of victory is defined based on percentiles; hence, it is a relative measure. Table S10.4 shows that the results are similar if we define close-election districts based on absolute measures. The table shows that the results are similar to the main results if close-election districts are defined as districts with a margin of victory of less than 5 percentage points or 10 percentage points. However, the results become smaller if we take the median as the threshold. 988 Bhattacharjee Table 8. Robustness Check: Mother and Birth-Level Characteristics (1) (2) (3) (4) Female Birth Previous sex Mother’s age Panel A dummy dummy composition at birth Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.0013 −0.0002 0.0029 −0.0000 (0.8623) (0.3760) (0.4159) (0.6691) [0.8727] [0.4048] [0.4825] [0.4557] Observations 154,334 9,414,516 113,377 154,334 Mean of dependent variable 0.4847 0.0164 0.5121 23.2783 Number of mothers 45,687 67,077 35,182 45,687 (1) (2) (3) Birth Birth order Birth order Panel B interval = 1 dummy ≤ mean dummy Born 0–11 months before scheduled election 0.0590 −0.0072 −0.0007 (0.7431) (0.0637) (0.6786) [0.7471] [0.1103] [0.6750] Observations 113,377 154,334 154,334 Mean of dependent variable 30.9066 0.1962 0.6556 Number of mothers 35,182 45,687 45,687 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each cell represents a separate regression. The dependent variables are mentioned at the top of each column. The table reports the coefficients on the scheduled election dummy. Apart from the reported variables, all regressions include mother fixed effects, dummies for year of birth, month of birth, and order of birth. The regressions also include state-specific linear time trends and control for mother’s age at birth. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. dummy, the pre-election decline in neonatal mortality only occurs for children born in close-election districts. Column 5 includes the interaction of the continuous absolute margin of victory variable with the election dummy. The results are small in magnitude and statistically insignificant since we expect that targeting of resources across districts will only happen if the margin of victory is sufficiently narrow. 6.3. Robustness to the Sample Restrictions In the paper, the main specification excludes mothers who had underage births or multiple births. The study also excludes women who report themselves to be visitors and births that took place before the woman migrated to her current place of residence. This section relaxes these conditions. Table 10 presents the results. Columns 1 and 2 include women with underage births and multiple births and the estimates are not very different from the baseline results. Columns 3 and 4 include women who report themselves to be visitors and births that took place before the woman migrated to her current place of residence. The magnitudes of the estimates fall. However, the results are similar in sign and significance to the baseline results. 6.4. Excluding West Bengal We expect the effect of the timing of elections on neonatal mortality to be weak in states with very little political competition. West Bengal is the only state where the ruling party did not change during the sample period. The Left Front came to power in West Bengal in 1977 and remained in power until 2011. Thus, one party ruled for the entire sample period (1977–1999). West Bengal has very little political competition and so we expect political cycles to be significantly muted there. The World Bank Economic Review 989 Table 9. Robustness Check: Role of Political Competition (1) (2) (3) (4) (5) Born 0–5 months before scheduled election −2.6668*** −2.0635** −2.2176 −2.6571*** −0.3376 (0.0084) (0.0036) (0.0528) (0.0085) (0.4445) Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 [0.0051] [0.0242] [0.1502] [0.0049] [0.4593] Born 0–5 months before × high margin dummy (>10th percentile) 2.7551** (0.0154) [0.0216] Born 0–5 months before × high margin dummy (>15th percentile) 2.1636** (0.0106) [0.0389] Born 0–5 months before × high margin dummy (>1st percentile) 2.0352 (0.0696) [0.1716] Born 0–5 months before × medium-low margin dummy 3.3182** (0.0069) [0.0256] Born 0–5 months before × medium-high margin dummy 2.3139** (0.0424) [0.0263] Born 0–5 months before × high margin dummy (>90th percentile) 2.0650 (0.2010) [0.2573] Born 0–5 months before × absolute margin 0.7767 (0.7892) [0.8089] High margin dummy (>10th percentile) −0.4536* (0.1401) [0.0697] High margin dummy (>15th percentile) −0.3867* (0.0049) [0.0813] High margin dummy (>1st percentile) 0.0281 (0.9588) [0.9647] Medium-low margin dummy (10th–50th percentile) −0.4382* (0.1691) [0.0954] Medium-high margin dummy (50th–90th percentile) −0.5550* (0.0936) [0.0588] High margin dummy (>90th percentile) −0.0539 (0.8871) [0.8938] Absolute margin −0.1691 (0.7808) [0.7715] Observations 150,677 150,677 150,677 150,677 150,677 Mean of dependent variable 4.8775 4.8775 4.8775 4.8775 4.8775 Number of mothers 45,047 45,047 45,047 45,047 45,047 Source: Sample includes children born in the years 1977–1999 in 25 states. Data come from the second round of National Family Health Survey and Election Commission of India website. Note: Each column represents a separate regression. The dependent variable is neonatal mortality. The table reports the coefficients of dummy variables for children born 0–5 months before scheduled elections, interactions between the election dummy and dummy variables indicating different levels of margin of victory, and the margin of victory dummies. All regressions include mother’s age at birth, mother fixed effects, dummies for year of birth, month of birth, birth order, and sex of the child, and state-specific linear time trends. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. * p < 0.10, ** p < 0.05, *** p < 0.01. 990 Bhattacharjee Table 10. Robustness Check: Sample Restrictions (1) (2) (3) (4) (5) (6) Including mothers with No migration Excluding multiple and underage births restrictions West Bengal Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.5332** −0.5927* −0.4247* −0.4875* −0.6129** −0.7139** (0.0336) (0.0327) (0.0583) (0.0613) (0.0201) (0.0118) [0.0455] [0.0509] [0.0706] [0.0855] [0.0408] [0.0304] Born 12–23 months before scheduled election −0.0499 −0.1187 −0.1764 (0.7473) (0.5256) (0.2009) [0.7688] [0.5688] [0.2136] Born 1–12 months after scheduled election 0.0518 0.1917 −0.0573 (0.8298) (0.4564) (0.8423) [0.8326] [0.4920] [0.8497] Born 13–24 months after scheduled election −0.2839 −0.3399 −0.3400 (0.2844) (0.1540) (0.2289) [0.3214] [0.1833] [0.2695] Observations 161,417 161,417 196,959 196,959 148,540 148,540 Mean of dependent variable 5.2597 5.2597 5.0064 5.0064 4.9002 4.9002 Number of mothers 47,260 47,260 57,746 57,746 43,846 43,846 Source: The sample includes children born in the years 1977–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each column represents a separate regression. The dependent variable is neonatal mortality. The table reports coefficients on dummies representing the scheduled election cycle. Columns 1 and 2 include mothers who had teenage births or multiple births. Columns 3 and 4 include mothers who were reported as visitors and births before the mother moved into the current area of residence. Columns 5 and 6 exclude the state of West Bengal. Apart from the reported variables, all regressions include mother fixed effects, dummies for year of birth, month of birth, order of birth, and sex of the child. The regressions also include state-specific linear time trends and control for mother’s age at birth. Columns 1 and 2 additionally include two dummy variables indicating underage births and multiple births. Columns 3 and 4 include a dummy indicating whether the mother reported herself to be a visitor or whether the birth occurred before the mother moved into the current area of residence. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. Columns 5 and 6 of table 10 show the results excluding West Bengal. The magnitudes of the estimates increase with the exclusion of the state of West Bengal, as expected. 6.5. Inclusion of Other NFHS Rounds Table 11 includes data from the other NFHS rounds. Column (1) includes data from NFHS-1 (1992– 1993) and NFHS-2. The magnitude and significance of the estimates are similar to the baseline estimates using only NFHS-2 data. Column (3) includes results from NFHS-3 (2005–2006) and NFHS-4 (2015– 2016). The magnitude of the effect falls and the result turns statistically insignificant. While the zero results with the inclusion of the last two rounds are surprising, there is a problem with including the last two NFHS rounds. In my original sample, the state boundaries were consistently defined for the entire period, 1977–1998. Three union territories got state status in 1987. However, this did not change the existing state boundaries. On the other hand, three new states were carved out of the existing states in India in 2000.29 This altered the state boundaries of three major states: Bihar, Uttar Pradesh, and Madhya Pradesh. Thus comparison of siblings born to the same mother can be across two states with different political structures. In order to check whether this accounts for the differential effects, these new states are dropped from NFHS-3 and 4, and the districts of the parent state from which these states were created are dropped from NFHS-1 and 2. The results are presented in columns 2 and 4 of table 11. The 29 Another state of Telangana was carved out of the state of Andhra Pradesh in 2014. In order to account for this, the analysis in this section excludes the years 2014 and 2015 from NFHS-4. The World Bank Economic Review 991 Table 11. Other NFHS Rounds (1) (2) (3) (4) NFHS-1 and 2 NFHS-3 and 4 including excluding including excluding Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 new states new states new states new states Born 0–11 months before scheduled election −0.5421*** −0.4889** 0.0772 0.0648 (0.0009) (0.0036) (0.2638) (0.3059) [0.0080] [0.0135] [0.1934] [0.3316] Observations 285,325 272,797 937,215 830,726 Mean of dependent variable 4.9905 5.0246 3.6466 3.6318 Number of mothers 86,204 82,478 291,666 258,289 Source: Data come from four rounds of the National Family Health Survey (NFHS-1, 2, 3, and 4) and the official website of the Election Commission of India. The sample is mentioned at the top of the columns. Note: Each cell represents a separate regression. The dependent variable is neonatal mortality. The table reports the coefficient of the scheduled election dummy. All regressions include round-mother fixed effects, round-year of birth fixed effects, dummies for the month of birth, order of birth, and sex of the child. The regressions also include state-specific linear time trends and control for mother’s age at birth. Errors are clustered at state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. results from the restricted sample are similar to those from the entire sample, shown in columns 1 and 3. Thus the differential results are not driven by the change in state boundaries. While the exact reason for the reduction in the incidence of political cycles in the later NFHS rounds cannot be ascertained with certainty, there exists literature on the gradual weakening of the effect of political cycles as a democracy matures (Akhmedov and Zhuravskaya 2004; Brender and Drazen 2005). 7. Mechanisms The government can influence neonatal mortality in many ways. Firstly, government expenditure on health can increase before elections. Secondly, the public health system might work more efficiently before elec- tions and so the usage of medical services might go up. Thirdly, if government workfare schemes run more efficiently before elections, employment rates can go up in the pre-election year.30 Fourthly, income can also go up because of the distribution of cash by political parties to buy votes. Banerjee et al. (2011) note a large-scale distribution of cash before elections in Delhi. The political cycle in neonatal mortality is unlikely to be influenced by an increase in state health expenditure. Sáez and Sinha (2010) document a fall in health expenditure in India in election years. Using yearly data on state health expenditure, the paper finds no effects of the scheduled election cycle on health expenditure. The results are shown in table S10.5. One problem with this analysis is that state health expenditure data is yearly. Thus, it is not possible to clearly distinguish between the pre- and the post-election period in the election year. Analysis of the direct effect of elections on health-care access or disposable income is difficult because of data limitations. Instead, this paper examines whether health-care usage goes up before elections. Use of prenatal care is an important determinant of neonatal mortality (Gortmaker 1979). Thus, this paper analyzes the effects of scheduled elections on a large number of inputs during pregnancy. It also tests the effect of being born before elections on the usage of medical care during the time of birth. 30 There have been a number of rural employment guarantee schemes in India during my sample period, like the Ma- harashtra Employment Guarantee Scheme (EGS), the National Rural Employment Programme (NREP), and the Rural Landless Employment Guarantee Programme (RLEGP). The NREP and RLEGP were merged into the Jawahar Rozgar Yojana (JRY) when it was launched in 1989. 992 Bhattacharjee Table 12. Channels (1) (2) (3) (4) (5) Neonatal Antenatal Iron/folic acid Weight Panel A mortality Tetanus visits supplements checked Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Born 0–11 months before scheduled election −0.3831* 0.0261* 0.0736* −0.0007 0.0150 (0.0610) (0.0101) (0.0629) (0.9335) (0.0651) [0.0956] [0.0737] [0.0920] [0.9343] [0.1607] Observations 44,531 31,593 31,588 31,683 31,678 Mean of dependent variable 3.7478 0.7571 2.7624 0.5784 0.3679 (1) (2) (3) (4) (5) Abdominal Internal Panel B Blood test Urine test examination examination Ultrasound Born 0–11 months before scheduled election 0.0082* −0.0044 0.0112* 0.0115 −0.0050 (0.0620) (0.3475) (0.0388) (0.1692) (0.4613) [0.0860] [0.3452] [0.0683] [0.1790] [0.5189] Observations 31,678 31,678 31,677 31,678 31,677 Mean of dependent variable 0.3890 0.3642 0.4956 0.2529 0.1196 (1) (2) (3) (4) Assistance during delivery Traditional Hospital Panel C Amniocentesis Doctors/nurses birth attendants delivery Born 0–11 months before scheduled election 0.0045 −0.0025 0.0124 −0.0027 (0.1328) (0.8328) (0.1491) (0.7139) [0.1343] [0.8411] [0.2089] [0.6978] Observations 31,654 31,686 31,686 31,681 Mean of dependent variable 0.0157 0.4188 0.3932 0.3375 Source: The sample includes children born in the years 1995–1999 in 25 states in India. Data come from the second round of the National Family Health Survey and the official website of the Election Commission of India. Note: Each cell represents a separate regression. The dependent variable is mentioned at the top of each column. In addition to the scheduled election dummy, all regressions include state fixed effects, mother’s age at birth, and dummies for year of birth, multiple births, urban area of residence, membership in Scheduled Caste and Scheduled Tribe categories, and Muslims. Errors are clustered at the state level. The p-values are shown in parentheses. Wild bootstrap cluster p-values are reported in square brackets. Significance stars are based on wild bootstrap cluster p-values. * p < 0.10, ** p < 0.05, *** p < 0.01. The above information on take-up of health-care services by mothers before and during delivery is only available for children born in the period 1995–1999. Because of the reduced sample size, the sample restrictions imposed in the main analysis are relaxed in this section. Thus, I do not exclude mothers with underage births and multiple births and relax the migration restrictions. Since the information on these variables is not available for the majority of children born to a mother, the regressions could not include mother fixed effects. Instead, the regressions include state fixed effects and a set of additional controls. These controls are dummies for membership in Scheduled Castes and Scheduled Tribes, which are the disadvantaged socioeconomic groups, a dummy for Muslims, a dummy for the urban area of residence, a dummy for multiple births, and a dummy for female children.31 State-specific time trends are not included because of the shorter time span of the data. Table 12 shows the estimated results. Column 1 of panel A shows that the main result, which is the neonatal mortality result, is robust to these changes in the original specification. However, the effect becomes marginally significant. Column 2 of panel A shows that mothers of children born 0–11 months before scheduled elections are about 2.6 percentage points (3 percent of its mean) more likely to have at least one tetanus injection during pregnancy. Column 3 of panel A shows that the number of antenatal 31 Controlling the sex of the child is important even for prenatal investments (Bharadwaj and Lakdawala 2013). The World Bank Economic Review 993 visits increases by about 3 percent of its mean for mothers of children born before elections. There is no significant effect on the probability of having iron or folic acid supplements. The paper also tests the effect of the timing of scheduled elections on the components of antenatal checkups and shows that the probability of a blood test during pregnancy increases by 0.8 percentage Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 points (2 percent of its mean) and the probability of having an abdominal check increases by 1 percentage point (2 percent of its mean). Among the other components of an antenatal checkup, the probability of having a weight check increases by 1.5 percentage points (4 percent of mean), having an internal examination increases by 1 percentage point (4.5 percent of mean), and having an amniocentesis increases by 0.45 percentage points (28 percent of its mean). However, the results for the probability of having a weight check, internal examination, and amniocentesis are statistically insignificant. Columns 2 and 3 of panel B show the estimates for the presence of any medical help during delivery. Column 2 shows the effect of being born before elections on the likelihood of the presence of doctors or nurses. Column 3 shows the effect on the likelihood of the presence of traditional birth attendants. The effect is small and statistically insignificant for the presence of doctors/nurses.32 The presence of traditional birth attendants increases by about 1.3 percent points (3 percent of its mean) for children born before elections but the effect is imprecise. The increased presence of traditional birth attendants might be due to the increased income of households before elections. Column 3 shows that the likelihood of hospital delivery does not change prior to elections.33 8. Conclusion The economics and political science literature provide evidence that politicians have strong incentives to improve the economic conditions of voters before scheduled elections. These incentives to provide special favors can be higher in politically more competitive regions. However, there has been very little empirical research on the welfare implications of these political cycles. The results presented in this paper show that the impacts of electoral cycles are not confined to economic policies and macroeconomic outcomes alone. Rather, they can have important effects on individual-level developmental outcomes. This paper shows that neonatal mortality is significantly lower for children born one year before sched- uled elections. The results further provide some evidence that the reduction in neonatal mortality is higher in politically more competitive districts for children born just before elections. My results also show that medical care utilization goes up before elections. These can have significant positive effects on both mother and child health. Health in general, and particularly child health, is an extremely important issue in developing countries. Child deaths occur due to insufficient health-care facilities or inadequate nutrition (Jones et al. 2003). The availability of health-care facilities and nutrition depends on government policies. Thus political variables play an important role in influencing child survival. These variables can be manipulated by politicians to improve the economic conditions of voters before elections to improve their electoral prospects (Nordhaus 1975; Lindbeck 1976; Rogoff and Sibert 1988; Persson and Tabellini 1990). My results show that the maximization of political gains by politicians leads to extremely different results for children born to the same mother. The welfare implications of the electoral cycle depend on whether the improvement in child health outcomes before elections is because of an overall improvement in the health of the children born before elections or a transfer of resources from non-election years to election years. Overall improvement in the health of children can result from an increase in the capacity of the health-care system. The construction of 32 The regressions of the effects of being born before elections on the presence of doctors and nurses separately give small and statistically insignificant results. 33 The effects are small and statistically insignificant if we consider the probability of birth in private and government institutions separately. 994 Bhattacharjee new hospitals is time consuming and is unlikely to be altered in the short run. Given the low government expenditure on health, state capacity to provide health services is limited in India. Sáez and Sinha (2010) and Waknis (2014) document a fall in health expenditure in India in election years. The paper finds no effects of the scheduled election cycle on health expenditure.34 The low government expenditure on Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 health combined with the unchanged state health-care capacity before elections implies that electoral manipulation is likely to occur through the distribution of resources across the different years of the electoral cycle (Berenschot 2010). Diversion of resources from non-election years to election years can occur in several ways. For ex- ample, doctors can take leave strategically in the off-election years to avoid better monitoring during election years. Davies (2021) finds evidence of reduced absenteeism of public school teachers in the year before an election and an increase in teacher absenteeism in the year after an election. Like public school teachers, public health-care workers usually are state government employees in India (Singh 2008). Thus, absenteeism of public health workers can also be lower before elections, under pressure from the state government. Using experimental evidence from the neighboring country, Pakistan, Callen et al. (2016) find that politicians can play an instrumental role in reducing the absence of doctors in hospitals. Thus, political incentives can affect the timing of doctor absenteeism. It is also possible that politicians hoard essential resources like medicines until the year before an election.35 These come at the cost of worsen- ing health outcomes in off-election years. Existing literature also suggests a decline in corruption before elections and an increase in corruption immediately after elections. Vadlamannati (2015) provides ev- idence of increased efforts by incumbent governments to control corruption during scheduled election years. Fagernäs and Pelkonen (2018) note that increased corruption in the post-election year can increase the hiring and transfers of primary school teachers after assembly elections in India. This indicates that elected legislators are more accountable before elections, but this comes at the cost of reduced account- ability in other years, which can affect child health outcomes deferentially across the different years of the electoral cycle. Thus the pre-electoral increase in health outcomes seems to reflect a reallocation from the off-election years to election years rather than an overall improvement in child health. Data Availability Statement The individual-level health data used in the paper comes from the National Family Health Survey (NFHS). It is available free from the Demographic and Health Surveys website but requires their permission for download. I can make available the rest of the data that I use in the paper, the programs which can replicate the regressions, and share instructions on how to access the NFHS data on request. References Aaskoven, L. 2016. “Fiscal Transparency, Elections and Public Employment: Evidence from the OECD.” Economics & Politics 28 (3): 317–41. Aidt, T., Z. Asatryan, L. Badalyan, and F. Heinemann. 2020. “Vote Buying or (Political) Business (Cycles) as Usual?” Review of Economics and Statistics 102 (3): 409–25. Akhmedov, A., and E. Zhuravskaya. 2004. “Opportunistic Political Cycles: Test in a Young Democracy Setting.” Quarterly Journal of Economics 119 (4): 1301–38. Aksoy, D. 2014. “Elections and the Timing of Terrorist Attacks.” Journal of Politics 76 (4): 899–913. Alesina, A. 1988. “Macroeconomics and Politics.” In NBER Macroeconomics Annual 1988, 13–62. Cambridge, MA: MIT Press. 34 The results are shown in table S10.5. 35 There is some evidence of medicine hoarding by politicians during the COVID pandemic (Sharma 2021), which indicates the possibility of such hoarding for political gains in the past. The World Bank Economic Review 995 Alesina, A., N. Roubini, and G. D. Cohen. 1997. Political Cycles and the Macroeconomy. Cambridge, MA: MIT Press. Anukriti, S. 2018. “Financial Incentives and the Fertility-Sex Ratio Trade-Off.” American Economic Journal: Applied Economics 10 (2): 27–57. Azam, M., and G. G. Kingdon. 2013. “Are Girls the Fairer Sex in India? Revisiting Intra-Household Allocation of Education Expenditure.” World Development 42: 143–64. Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Banerjee, A. V., S. Kumar, R. Pande, and F. Su. 2011. “Do Informed Voters Make Better Choices? Ex- perimental Evidence from Urban India.” Working Paper, Harvard University. Available at: http://citeseerx. ist.psu.edu/viewdoc/download?doi=10.1.1.221.1405&rep=rep1&type=pdf. Bardhan, P., and D. Mookherjee. 2010. “Determinants of Redistributive Politics: An Empirical Analysis of Land Re- forms in West Bengal, India.” American Economic Review 100 (4): 1572–600. Baskaran, T., B. Min, and Y. Uppal. 2015. “Election Cycles and Electricity Provision: Evidence from a Quasi- Experiment with Indian Special Elections.” Journal of Public Economics 126: 64–73. Berdejó, C., and N. Yuchtman. 2013. “Crime, Punishment, and Politics: An Analysis of Political Cycles in Criminal Sentencing.” Review of Economics and Statistics 95 (3): 741–56. Berenschot, W. 2010. “Everyday Mediation: The Politics of Public Service Delivery in Gujarat, India.” Development and Change 41 (5): 883–905. Berger, H., and U. Woitek. 1997. “How Opportunistic Are Partisan German Central Bankers: Evidence on the Vaubel Hypothesis.” European Journal of Political Economy 13 (4): 807–21. Berman, P. A. 1998. “Rethinking Health Care Systems: Private Health Care Provision in India.” World Development 26 (8): 1463–79. Bhalotra, S. 2010. “Fatal Fluctuations? Cyclicality in Infant Mortality in India.” Journal of Development Economics 93 (1): 7–19. Bhalotra, S., and I. Clots-Figueras. 2014. “Health and the Political Agency of Women.” American Economic Journal: Economic Policy 6 (2): 164–97. Bhalotra, S., I. Clots-Figueras, G. Cassan, and L. Iyer. 2014. “Religion, Politician Identity and Development Outcomes: Evidence from India.” Journal of Economic Behavior & Organization 104: 4–17. Bharadwaj, P., and L. K. Lakdawala. 2013. “Discrimination Begins in the Womb: Evidence of Sex-Selective Prenatal Investments.” Journal of Human Resources 48 (1): 71–113. Block, S. A. 2002. “Political Business Cycles, Democratization, and Economic Reform: The Case of Africa.” Journal of Development Economics 67 (1): 205–28. Borjas, G. J. 1984. “Electoral Cycles and the Earnings of Federal Bureaucrats.” Economic Inquiry 22 (4): 447–59. Brender, A., and A. Drazen. 2005. “Political Budget Cycles in New versus Established Democracies.” Journal of Mon- etary Economics 52 (7): 1271–95. Buekens, P., and A. Wilcox. 1993. “Why Do Small Twins Have a Lower Mortality Rate than Small Singletons?” American Journal of Obstetrics and Gynecology 168 (3): 937–41. Cahan, D. 2019. “Electoral Cycles in Government Employment: Evidence from US Gubernatorial Elections.” Euro- pean Economic Review 111: 122–38. Callen, M., S. Gulzar, S. A. Hasanain, and M. Y. Khan. 2016. “The Political Economy of Public Sector Absence: Experimental Evidence from Pakistan.” Working Paper, National Bureau of Economic Research. Available at: https://www.nber.org/papers/w22340. Castro, V., and R. Martins. 2018. “Politically Driven Cycles in Fiscal Policy: In Depth Analysis of the Functional Components of Government Expenditures.” European Journal of Political Economy 55: 44–64. Cole, S. 2009. “Fixing Market Failures or Fixing Elections? Agricultural Credit in India.” American Economic Journal: Applied Economics 1 (1): 219–50. Daniele, G., and G. Dipoppa. 2017. “Mafia, Elections and Violence against Politicians.” Journal of Public Economics 154: 10–33. Davies, E. 2021. “Absence: Electoral Cycles and Teacher Absenteeism in India.” Working Paper. Available at: https://www.emmerichdavies.com/files/Davies,%20Emmerich%20-%20Absence.pdf. Dinç, I. S. 2005. “Politicians and Banks: Political Influences on Government-Owned Banks in Emerging Markets.” Journal of Financial Economics 77 (2): 453–79. Drazen, A., and M. Eslava. 2010. “Electoral Manipulation via Voter-Friendly Spending: Theory and Evidence.” Journal of Development Economics 92 (1): 39–52. 996 Bhattacharjee Dua, B. D. 1979. “Presidential Rule in India: A Study in Crisis Politics.” Asian Survey 19 (6): 611–26. Dyke, A. 2007. “Electoral Cycles in the Administration of Criminal Justice.” Public Choice 133 (3–4): 417–37. Election Commission of India (ECI). 2017. “Statistical Report on General Election.” Technical Report. Available at: https://eci.gov.in/statistical-report/statistical-reports/. Englmaier, F., and T. Stowasser. 2017. “Electoral Cycles in Savings Bank Lending.” Journal of the European Economic Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Association 15 (2): 296–354. Fagernäs, S., and P. Pelkonen. 2018. “Teachers, Electoral Cycles, and Learning in India.” Journal of Human Resources 55 (2): 699–732. Galli, E., and S. P. Rossi. 2002. “Political Budget Cycles: The Case of the Western German Länder.” Public Choice 110 (3): 283–303. Gandhi, Kingdon. 2002. “The Gender Gap in Educational Attainment in India: How Much Can Be Explained?” Journal of Development Studies 39 (2): 25–53. Gonzalez, M. D. L. A. 2002. “Do Changes in Democracy Affect the Political Budget Cycle? Evidence from Mexico.” Review of Development Economics 6 (2): 204–24. Gortmaker, S. 1979. “The Effects of Prenatal Care upon the Health of the Newborn.” American Journal of Public Health 69 (7): 653–60. Grier, K. 2008. “US Presidential Elections and Real GDP Growth, 1961-2004.” Public Choice 135 (3–4): 337–52. Guo, G. 2009. “China’s Local Political Budget Cycles.” American Journal of Political Science 53 (3): 621–32. Gupta, B., and A. Mukhopadhyay. 2016. “Local Funds and Political Competition: Evidence from the National Rural Employment Guarantee Scheme in India.” European Journal of Political Economy 41: 14–30. Gupta, M. D., and M. Rani. 2004. India’s Public Health System: How Well Does It Function at the National Level?, Volume 3447. Washington, DC: World Bank Publications. Harish, S., and A. T. Little. 2017. “The Political Violence Cycle.” American Political Science Review 111 (2): 237–55. Herwartz, H., and B. Theilen. 2014. “Health Care and Ideology: A Reconsideration of Political Determinants of Public Healthcare Funding in the OECD.” Health Economics 23 (2): 225–40. Inoue, C. 2020. “Election Cycles and Organizations: How Politics Shapes the Performance of State-Owned Enterprises over Time.” Administrative Science Quarterly 65 (3): 677–709. International Institute for Population Sciences (IIPS). 2000. “National Family Health Survey (NFHS-2), 1998-99.” Technical Report. Available at: http://rchiips.org/NFHS/nfhs2.shtml. Jayachandran, S., and I. Kuziemko. 2011. “Why Do Mothers Breastfeed Girls Less Than Boys? Evidence and Impli- cations for Child Health in India.” Quarterly Journal of Economics 126 (3): 1485–538. Jones, G., R. W. Steketee, R. E. Black, Z. A. Bhutta, and S. S. Morris. 2003. “How Many Child Deaths Can We Prevent This Year?” Lancet 362 (9377): 65–71. Joseph, J. V., and K. J. Reddy. 2004. “Executive Discretion and Article 356 of the Constitution of India: A Comparative Critique.” Electronic Journal of Comparative Law 8 (1). Katsimi, M., and V. Sarantides. 2012. “Do Elections Affect the Composition of Fiscal Policy in Developed, Established Democracies?” Public Choice 151 (1–2): 325–62. Khemani, S. 2004. “Political Cycles in a Developing Economy: Effect of Elections in the Indian States.” Journal of Development Economics 73 (1): 125–54. Klein, M. W. 1996. “Timing Is All: Elections and the Duration of United States Business Cycles.” Journal of Money, Credit and Banking 28 (1): 84–101. Klomp, J., and J. de Haan. 2016. “Election Cycles in Natural Resource Rents: Empirical Evidence.” Journal of Devel- opment Economics 121: 79–93. Kudamatsu, M. 2012. “Has Democratization Reduced Infant Mortality in Sub-Saharan Africa? Evidence from Micro Data.” Journal of the European Economic Association 10 (6): 1294–317. Levitt, S. D. 1997. “Using Electoral Cycles in Police Hiring to Estimate the Effect of Police on Crime.” American Economic Review 87 (3): 270–90. Lindbeck, A. 1976. “Stabilization Policy in Open Economies with Endogenous Politicians.” American Economic Re- view 66 (2): 1–19. Magloire, F. S. 1997. “Political Monetary Cycles and Independence of the Central Bank in a Monetary Union: An Empirical Test for a BEAC Franc Zone Member Country.” Journal of African Economies 6 (1): 112–31. The World Bank Economic Review 997 McCallum, B. T. 1978. “The Political Business Cycle: An Empirical Test.” Southern Economic Journal 44 (3): 504–15. Mechtel, M., and N. Potrafke. 2013. “Electoral Cycles in Active Labor Market Policies.” Public Choice 156 (1–2): 181–94. Meloni, O. 2018. “Is There an Electorally-Motivated Crime Rate Cycle? Evidence from Argentina.” Economics Bul- Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 letin 38 (2): 1102–10. Min, B., and M. Golden. 2014. “Electoral Cycles in Electricity Losses in India.” Energy Policy 65: 619–25. Murthi, M., A.-C. Guio, and J. Dreze. 1995. “Mortality, Fertility, and Gender Bias in India: A District Level Analysis.” Population and Development Review 21 (4): 745–82. Naeye, R. L., L. S. Burt, D. L. Wright, W. A. Blanc, and D. Tatter. 1971. “Neonatal Mortality, the Male Disadvantage.” Pediatrics 48 (6): 902–6. Newman, L. S. 2013. “Do Terrorist Attacks Increase Closer to Elections?” Terrorism and Political Violence 25 (1): 8–28. Nordhaus, W. D. 1975. “The Political Business Cycle.” Review of Economic Studies 42 (2): 169–90. Pallage, S., and M. A. Robe. 2003. “On the Welfare Cost of Economic Fluctuations in Developing Countries.” Inter- national Economic Review 44 (2): 677–98. Pande, R. P. 2003. “Selective Gender Differences in Childhood Nutrition and Immunization in Rural India: The Role of Siblings.” Demography 40 (3): 395–418. Park, K. H. 2017. “The Impact of Judicial Elections in the Sentencing of Black Crime.” Journal of Human Resources 52 (4): 998–1031. Paxson, C., and N. Schady. 2005. “Child Health and Economic Crisis in Peru.” World Bank Economic Review 19 (2): 203–23. Persson, T., and G. E. Tabellini. 1990. Macroeconomic Policy, Credibility and Politics. Fundamentals of Pure and Applied Economics Series 38. London: Harwood Academic Publishers. Potrafke, N. 2010. “The Growth of Public Health Expenditures in OECD Countries: Do Government Ideology and Electoral Motives Matter?” Journal of Health Economics 29 (6): 797–810. Rammohan, A., K. Iqbal, and N. Awofeso. 2013. “Reducing Neonatal Mortality in India: Critical Role of Access to Emergency Obstetric Care.” PloS One 8 (3): e57244. Rogoff, K., and A. Sibert. 1988. “Elections and Macroeconomic Policy Cycles.” Review of Economic Studies 55 (1): 1–16. Roodman, D., M. Ø. Nielsen, J. G. MacKinnon, and M. D. Webb. 2019. “Fast and Wild: Bootstrap Inference in Stata using Boottest.” Stata Journal 19 (1): 4–60. Sáez, L., and A. Sinha. 2010. “Political Cycles, Political Institutions and Public Expenditure in India, 1980–2000.” British Journal of Political Science 40 (1): 91–113. Sharma, N. 2021. “Hope Covid Medicines Not Being Hoarded for Political Gains, Says Delhi HC.” India Today. May 17. Available at: https://www.indiatoday.in, accessed September 21, 2022. Sidorkin, O., and D. Vorobyev. 2018. “Political Cycles and Corruption in Russian Regions.” European Journal of Political Economy 52: 55–74. Singh, N. 2008. “Decentralization and Public Delivery of Health Care Services in India.” Health Affairs 27 (4): 991– 1001. Singh, U. K. 2012. “Between Moral Force and Supplementary Legality: A Model Code of Conduct and the Election Commission of India.” Election Law Journal 11 (2): 149–69. Takaku, R., and S.-I. Bessho. 2018. “Political Cycles in Physician Employment: A Case of Japanese Local Public Hospitals.” Social Science & Medicine 216: 97–106. Tepe, M., and P. Vanhuysse. 2009. “Educational Business Cycles.” Public Choice 139 (1): 61–82. Vadlamannati, K. C. 2009. “Does Timing of Elections Instigate Riots?: A Subnational Study of 16 Indian States, 1958-2004.” Indian Journal of Political Science 70 (2): 529–52. ———. 2015. “Fighting Corruption or Elections? The Politics of Anti-Corruption Policies in India: A Subnational Study.” Journal of Comparative Economics 43 (4): 1035–52. Veiga, L. G., and F. J. Veiga. 2007. “Does Opportunism Pay Off?” Economics Letters 96 (2): 177–82. Vergne, C. 2009. “Democracy, Elections and Allocation of Public Expenditures in Developing Countries.” European Journal of Political Economy 25 (1): 63–77. 998 Bhattacharjee Waknis, P. 2014. “A Political Economy of Sub-National Government Spending in India.” International Journal of Public Policy 10 (1–3): 27–83. Wilkinson, S. I. 2006. Votes and Violence: Electoral Competition and Ethnic Riots in India. New York: Cambridge University Press. World Health Organization (WHO). 2006. “WHO Child Growth Standards: Length/Height-for-Age, Weight-for-Age, Downloaded from https://academic.oup.com/wber/article/36/4/972/6754316 by Sectoral Library Rm MC-C3-220 user on 10 December 2023 Weight-for-Length, Weight-for-Height and Body Mass Index-for-Age: Methods and Development.” Geneva: World Health Organization. Technical report. Available at: https://www.who.int/publications/i/item/924154693X.