Confirming Spatial Differences in Andhra Pradesh Villages

A prior examination of 33 villages in Andhra Pradesh helped provide us with data that we utilized to confirm results related to four dimensions of governance: Voice and Accountability, Absence of Violence, Service Delivery, and Absence of Corruption. Following the same process that was used in the previous section, an additive index is constructed for each dimension of governance using the relevant survey questions, and these indices serve as our dependent variables for this analysis. Once again, when necessary the scores are rescaled so that higher scores indicate better experiences with governance.

Table 4 presents the summary data demonstrating the negative relationship between governance scores and distance to town (Dist PHC). Apart from the score for Absence of Violence – which once again proved hard to explain – the other three scores are higher, on average, in villages located closer to towns, while villages further away have significantly lower scores. Remoteness, once again, has a cost in terms of governance outcomes.
-- Table 4 about here --
We utilised varying-intercept multilevel models to identify the determinants of governance. The models are classified as varying-intercept or random-intercept multilevel models, because the group intercepts are allowed to vary and are estimated using group-level variables (Gelman and Hill 2007; Rabe-Hesketh and Skrondal 2008).

Two levels of independent variables are included in the models: individual-level predictors (level 1) and village-level predictors (level 2). It has been argued that multilevel models will perform as well as single-level regression models (Gelman and Hill, 2007:275), and they have the advantage of allowing for within-group correlation, thus reducing a concern related to biased standard errors (Rabe-Hesketh and Skrondal, 2008:129-130). Before presenting multilevel regression results, it is helpful to describe briefly how the independent variables were constructed in this case.

Relative wealth was measured using two separate variables. House Type was coded as 1 for brick homes and zero for mud constructions (Mean = 0.17; SD=0.38). Separately, Land pc measures the total agricultural land possessed by the household divided by the number of household members (Mean = 0.41; SD=0.95). Since these variables are only moderately correlated with each other, we employed both of them in the analysis that follows.

The nature of the dependent variables guides the decision of whether to use linear multilevel models or generalised linear multilevel models for each particular dimension of governance. Our dependent variables respectively have the following number of response categories: Voice and Accountability (10 response categories), Absence of Violence (3 response categories), Service Delivery (3 response categories), and Absence of Corruption (3 response categories). Following DeMaris (1992: 77-78), a linear multilevel model is utilised to examine Voice and Accountability, and generalised linear multilevel models are utilised for examining Absence of Violence, Service Delivery, and Absence of Corruption. In each of these cases, tests of collinearity did not indicate any reason for concern.¹⁵

Among other village-level variables, the percentage of SC residents of a village is negatively associated with Voice and Accountability, Service Delivery, and Absence of Corruption, showing how – while individual SCs may or may not experience poorer governance – villages that have higher proportions of SCs tend to experience poorer governance outcomes on average.¹⁷ Similarly, a greater proportion of STs shows up in significantly lower scores on one dimension: Absence of Corruption. Measures of economic development in a village have mixed effects on governance scores: the existence of a bus facility is negatively associated with Absence of Corruption, a surprising result, but the existence of paved (pakka) approach roads is positively associated with both Absence of Violence and Absence of Corruption.

After controlling for these village-level influences, few individual-level variables gain significance in the analysis. Relative wealth, as measured by house type, is significantly and positively associated only with Absence of Violence, while education is negatively associated with Service Delivery, showing perhaps how more educated people, expecting higher service levels, are more likely to complain about existing modes of service delivery. More frequent interaction with the panchayats and political parties is positively associated with Service Delivery scores – but not with any other governance score.

Overall, this evidence from Andhra Pradesh confirms the view that governance experiences vary across individuals and villages, with distance to nearest town structuring these experiences in significant ways. Rural people who live closer to town enjoy better governance, and those who live further away experience (and come to expect) progressively inferior governance.

An emergent confluence of state and market power has tended to intensify spatial differences within countries, resulting in both lower growth and poorer governance in more remote rural areas. With the advent of globalization, cities have acquired greater importance, and governments, national and state, have followed where markets have led, preferentially locating public services and infrastructure within and close to urban centers.¹⁸

The supply of government functions and functionaries is lower and more erratic in more distant villages. Demand for better governance is also lower: people in more remote locations are less confident that their voices will influence government decisions, less involved in diverse political activities, make contact less often with elected officials, and rely more often, compared to villagers located closer to towns, upon unelected middlemen and vote-broker patrons. We inquired from respondents about their experiences with six types of transactions (with the police, the taluka, banks, hospitals, extension agents, and employment-generating public works), asking how they would initiate contact with the officials concerned. In each case, a greater proportion of closer-to-town villagers responded that they would either contact such officials directly or seek help from political parties or local governments. In the more distant villages, the majority preferred to seek the help of (unelected) village leaders and political middlemen, many among whom are involved in clientelistic and possibly corrupt exchanges (Corbridge, et al. 2005).

Structural factors are partly to account for these outcomes. Processes of state formation commenced (or intensified) in India during the period of colonial rule – and largely persisted with and consolidated in the post-independence period – were of a top-down nature, radiating outward from national and provincial capitals but stopping well short of individual habitations, especially rural ones (Chatterjee 1997), leading one observer to refer to the Indian system as a “thin state” (Kaviraj 1997). Offices and agencies that are important for citizens’ everyday interactions with government – such as police stations and courts, hospitals and primary health centers, talukas and tahsils, high schools and colleges, electricity sub-stations and water supply offices – remain concentrated in towns, often at great distances from rural habitations, where more than two-thirds of all Indians lived and still live.

Political parties, which could have helped fill the resulting vacuum of upward representation, have also been organised from the top-down, being thin on the ground in most parts of rural India (Kohli 1992; Krishna 2010). The virtual absence of government structures linking the vast rural periphery has reinforced the tendency toward centralization of power, accompanied by widespread powerlessness in remoter areas (Kohli 2000).

Democratic decentralization should have served to reduce these effects of distance, but as noted above and as Manor (2010: 77) has stated, “Most state governments in India have been reluctant to empower and fund [panchayats]… So despite a few splendid exceptions…Indian states’ approaches to democratic decentralization may be chiefly remembered as sad examples of what might have been.” ¹⁹ Or as Bardhan and Mookherjee (2006: 21) state, “the extent of genuine functional or fiscal devolution is low on average." Decentralization as implemented so far has not, in any event, sufficed to control the widespread absenteeism of government officials, particularly those – such as schoolteachers, health workers, extension agents, and others tasked with various economic development programs – who, being positioned at the grassroots, can more easily evade monitoring and enforcement from above (Chaudhuri, et al. 2006). Because the agents of a top-down government are themselves appointed and supervised from the top-down, villagers and local representatives have little control over the government staffs deployed to serve these villages. Beyond complaining to higher officials, most often located in towns, villagers can do little or nothing to check absenteeism, enforce accountability, or insist upon norms of professional behavior. Greater distance to towns raises the costs and reduces the frequency of complaints.

Health and education outcomes are both worse in more distant villages compared to others located closer to towns, to a considerable extent a consequence of poorer governance. The risk of disease is significantly higher in villages that lie more than five kilometers from towns: the odds of having malaria are 31 percent greater; those of suffering cholera are 31 percent greater; and the odds of suffering from a catch-all category of “other communicable diseases” are 31 percent greater within further-away villages.²⁰ Learning outcomes at public schools are also significantly poorer in villages located further from towns.

Poorer governance in further villages, including worse service delivery and lower infrastructure provision, results in producing a vicious cycle. Being more often sick and less often educated further reduces remote villagers’ abilities to enforce good governance.

Different experiences and expectations have helped produce diverse levels of political efficacy and dissimilar forms of political behavior, both of which are configured by distance. As a result, nearly 40 percent of the entire Indian population lives at the wrong end of a gradient of governance, residing in villages located at greater distances from towns. To them, “the state can and often does appear distant… a sovereign entity set apart from society” (Fuller and Harriss 2000: 23). Improving governance in India requires attending to such spatial inequalities.

It should be useful to investigate whether the gradient of governance is also significant within other populous developing countries where large parts of the population live in rural areas, often at considerable distances from the nearest towns and government centers. A story of top-down and incomplete state formation has also been told for many of these countries,²¹ and governance outcomes in these countries might well parallel those we found in India, being highly variable and spatially arrayed.

Further research will help uncover the extent of spatial disparities within these and other domains. It will also help uncover contextually-relevant remedies. Whether deepening decentralised local government is the right way forward, or whether empowering civil society organisations, or building political parties – or some combination thereof – will provide a better solution, is itself a weighty question, worth a great deal of empirical research.

In the end, universalism and impartiality, resulting in low variance, is as much an attribute of good governance as achieving a high national average score (Rothstein 2011). Citizens everywhere, even those who do not live in or near big cities, must share equitably in the benefits of good governance.