The outcome variable of interest in our analysis is the state-level infant mortality rate (IMR). The IMR is the probability of death faced by infants of age one year or lower, and is measured as the number of deaths of infants under one year old per 1000 live births. The key explanatory variable for our study is state-level public health expenditure (PHE). We measure PHE as the sum of the state-level expenditure on medical & public health, family welfare and nutrition.
In India, public health expenditure is undertaken at all three levels of government: central (federal), state, and local. The central government undertakes expenditures in two forms: direct expenditure, and grants-in aid to state governments. State governments incur expenditures out of the grants-in aid and other resources, for example, tax revenues, available to them. Some state-level expenditure also takes the form of transfers to local government bodies. Local government bodies, in turn, incur expenditures out of these transfers and from other resources that they have. The total of all the expenditures incurred at the three tiers of government provides an estimate of public health expenditure in India. In this paper, we are interested in analyzing the effect of state-level public health expenditure on state-level IMR. Hence, the data for our research project relates only to state-level expenditure on
4In the early 2000s, the states of Jharkhand, Chhattisgarh and Uttarakhand were carved out of
Bihar, Madhya Pradesh and Uttar Pradesh, respectively. For comparability, we keep them as part of the larger states all through.
health care. This includes expenditures incurred by state governments from Central grants-in aid and from other resources available to the state government (for example,
through taxation).5
An important source of variation that might be correlated with both PHE and the IMR is the general quality of policy making and implementation of public welfare schemes at the state level. Previous studies have documented that political compe- tition improves policy making (Besley and Case, 1995; Rodgers and Rodgers, 2000; Besley et al., 2005). Perhaps more pertinent for this paper, many recent studies have found that the degree of political competition has beneficial effects on health status of the population (Besley and Kudamatsu, 2006; Fumagalli et al., 2013). The link in question seems intuitively clear: intense political competition between political parties can lead to an increase in the effectiveness of governance and accountability
5Thus, our data set does not include direct central government expenditure and local-level
expenditure funded by resources other than state-level transfers. According to NHAC (2009),
the local-level expenditure funded by resources other than state-level transfers is a negligible part of total expenditure. While we would have liked to include each state’s portion of total central government expenditure on health, we are unable to do so because of lack of easily available data on the state-wise distribution of central government expenditure on health. According to data compiled by Gupta and Chowdhury (2014), the bulk of health care spending is contributed by states. In 2010–2011, about 64 per cent of total public health care spending came from expenditure by states, about 31 per cent came from expenditure by the central government (direct expenditure and grants-in aid to states), and the rest was accounted for by expenditure of local government bodies. If we count the grants-in aid as part of state-level expenditure, then the contributions of the central government varies between 20 and 30 per cent of total public health expenditure (Choudhury and Nath, 2012). Thus, the figures for state-level public health expenditure used in this paper covers between 70 and 80 per cent of total public expenditure on health care in India. Yet, because our health outcome of interest is the infant mortality rate, which is likely to respond to even low levels of increases in public expenditure, and because, as stated above, under India’s constitution, state governments, rather than the central government, are primarily responsible for health provision, both in terms of health care and public health measures (Gupta and Rani, 2004), our data is expected to cover the part of total health expenditure at the state-level that is most pertinent to our analysis.
of the organs of the state, both of which can lead to an improvement in policy mak- ing. This can not only force state governments to devote more resources to welfare activities (like health care), but also improve the effectiveness of existing delivery mechanisms that have a direct impact on health status for every level of resource allocation. Moreover, higher level of political competition may provide incentives for the political parties to increase public expenditure on education, healthcare, nutrition and so forth, to increase their chances of getting re-elected.
Borrowing from the political science literature, we use an index of the index of effective number of parties in the government as a proxy for the degree of political competition in the state government, a process that is accentuated by the growth of coalition governments over the last two decades in India (Laasko and Taagepera, 1979). For state level assembly election years, the index of the effective number of parties in any state government is computed as
N = Pn1
i=1p 2 i
where N is the effective number of parties in a state government, i = 1, 2, . . . , n
indexes parties in the state government, and pi is the share of party i in the gov-
ernment.6 The value of the index remains unchanged until the next election year,
when a new government is formed and a new configuration of parties emerge as the
6This measure of the effective number of political parties was first proposed by Laasko and
Taagepera (1979). Although many alternatives have been proposed over the years, it is still con- sidered as the standard for comparative political research (Kline, 2009). Chamon et al. (2009, p.4) note that starting with Laasko and Taagepera (1979), the political science literature has used the effective number of parties as ‘the main measure of political competition.’
governing coalition. If a single party forms a government, the value of N is unity, and as the number of parties increase, the value of N increases.
While we have highlighted the possible positive impact of a higher number of effec- tive political parties through more intense political competition, it is worth pointing out that it might also have an opposite effect. If the effective number of political parties is understood as a measure of the ‘hyper-fractionalization’ of political power, then a higher value of the index can lead to higher rent seeking behaviour, and mili- tate against expenditures that are in the long term benefit of the general population. For instance, coalition governments with a higher number of parties in the ruling coalition make the coalition unstable and increase the probability of dissolution of the government before the end of the full five year term. This increases the incentive for each party in the coalition to make expenditure that cater to their narrow support bases, rather than undertake expenditure that would have long term benefits for the population. Thus, depending on the strength of these opposite effects, the effective number of parties might have a positive or negative effect on the IMR.
The additional controls used in the model are the following: per capita real
income, population sex ratio, (adult) female literacy, and urbanization.7