The Gradient of Governance: Distance and Disengagement in Indian Villages

Abstract

National governance scores must be seen in light of large within-country variance. Not only being a rural village, but being located at a substantial distance from cities, has great importance for good governance. Analysis of household data from different parts of rural India shows how villages at greater distances to towns tend to have lower scores on multiple governance dimensions. Even after controlling for diverse influences, using both ordinary least square and multilevel regression models, this gradient of governance remains significant, imposing a dual penalty. Already penalised by markets, which have disproportionately rewarded urban and peri-urban areas, residents of villages located further from towns also experience and expect to receive worse treatment from government.

Acknowledgements

The authors would like to thank without implicating Pablo Beramendi, John Booth, Lauren Maclean, Fritz Mayer, Bo Rothstein, Mitchell Seligson, and two anonymous referees for helpful comments on previous drafts of this article. Data collection exercises in Karnataka were partly supported by a grant (number OW2: 205) received from the International Initiative for Impact Evaluation.

Notes

^1. See, for example, Berthélémy and Tichit (2004); Burnside and Dollar (2004); Chong and Calderon (2000); Dollar and Levine (2004); Friedman, Johnson, Kaufman, and Zoido-Lobatón, (2000); Gupta, Davoodi, and Alonso-Terme, (1998); and Mauro (1995).

^2. See http://www.mca.gov/selection/index.php and Radelet (2003).

^3. Studies questioning the necessary (or causal) connection between governance and economic growth include: Andrews (2010); Glaeser, LaPorta, Lopez de Silanes, & Shleifer, (2004); Kurtz and Schrank (2007); Przeworski (2004); Seligson (2002); and Triesman (2007). Works that express doubts about different measures of governance are discussed below.

^4. See Chatterjee (2004, pp. 142–147) for a more complete explication of this argument in the case of India.

^5. See, for instance, Deaton and Dreze (2002); Krishna and Bajpai (2011); and Sen and Himanshu (2004a, 2004b).

^6. We use the terms town, city and urban area interchangeably. In the empirical analysis that follows, we use the definition employed by the Indian government. See http://censusindia.gov.in/2011-prov- results/paper2/data_files/India2/1.%20Data%20Highlight.pdf.

^7. Household surveys (from Gallup, Afrobarometer, Latinobarometer, and LAPOP’s AmericasBarometer, and so forth) provide the only data drawn from consultations with ordinary people. These sources do not cover equally all regions of the world. All of the other data sources used by the World Bank’s team draw upon expert assessments and elite opinions. Heavier weights have been given while constructing indices to data sources that are more highly correlated with the other sources consulted. Elite and expert opinions from different sources tend to be highly correlated with each other. As a result, ordinary people’s opinions carry very little weight.

^8. See Table 3 in Kaufmann, Kraay, & Mastruzzi (2008) for these calculations.

^9. See, for example, Kaufmann, Mehrez, & Gurgur (2002); and Kaufmann, Montorial-Garriga, & Recanatini (2008).

^10. Scheduled Castes are the former untouchables, and Scheduled Tribes are, loosely speaking, India’s aborigines. Other Backward Caste is a more recent administrative listing, and it refers to another group of castes whose members were historically disempowered and oppressed by other groups.

^11. See, for example, Booth and Seligson (2008); Bratton (2008); Lijphart (1997); Rosenstone and Hansen (1993); and Verba, Lehman Schlozman, & Brady, (1995).

^12. Lingayats have long been the most prosperous caste cluster in the two districts examined, dominating society in most villages as a result of their wealth, their control of most of the best land, their educational attainments and numerical strength. They have used their political muscle to get themselves listed officially as ‘Backward’ (since this yields benefits from government), but in reality are anything but ‘backward’. We thank an anonymous reviewer for making this point.

^13. Bivariate correlations among the independent variables are low, with the coefficient nowhere greater than 0.51. Multicollinearity does not appear to be a problem: variance inflation factor (VIF) values for all independent variables were in the range 1.07 to 3.79, with a mean VIF of 1.86. Results from the other regression models are available on request.

^14. The possibility of reverse causation cannot, however, be dismissed: those people, a small number, who feel the most helpless and experience the worst governance outcomes, may more often make contact with political parties and panchayats.

^15. The VIF (variance inflation factor) values for all independent variables were in the range 1.02 to 3.19, with a mean VIF of 1.51, showing low to moderate collinearity.

^16. As in Karnataka, the other dimension of governance, Absence of Violence, remains harder to explain at the village level.

^17. The probability of being SC (or ST) is higher in villages where a greater proportion of the population is SC (or ST), showing how these individual- and village-level variables may be correlated. However, the correlation is not so high as to cause problems for the analysis, as witnessed by the VIF numbers reported above.

^18. An analysis of results from the Indian national census of 2001 shows that while more than two-thirds of all villages located within 5 km of towns have been provided with paved roads, fewer than half of all villages beyond 20 km from towns were similarly endowed through public provision. The corresponding proportions for electric power supply are 85 per cent and 64 per cent.

^19. Karnataka is counted by most observers among these exceptions. More than Andhra Pradesh, it has invested historically in building a stronger structure of panchayats, although these efforts have lagged in recent years. Our results show that even greater efforts will be required in the future to strengthen the roles of gram panchayats, linking them more firmly with the institutional chain of governance flowing from national and state governments through districts and talukas down to the grassroots level.

^20. Author calculations from two rounds of nationally representative data: the District-Level Household and Facility Surveys (DLHS) of 2002–2004 and 2007–08. Similar spatial differences are also apparent for many other types of health and education outcomes. For more information on how to access the survey data, please visit the website for the International Institute for Population Sciences (http://www.rchiips.org/).

^21. See, for instance, Bates (1981); Boone (2003); Herbst (2000); Maclean (2011); and Scott (2010).