AGRICULTURAL ECONOMICS Agricultural Economics 34 (2006) 229–242 It is where you are that matters: the spatial determinants of rural poverty in India Richard Palmer-Jones, Kunal Sen ∗ School of Development Studies, University of East Anglia, Norwich, NR4 7TJ, UK Received 24 June 2004; received in revised form 22 November 2004; accepted 12 May 2005 Abstract The spatial patterns of poverty in India are of considerable importance in themselves and for development theory and practice. This article examines the determinants of rural poverty in India using spatial econometric methods. It finds that while agricultural growth is the key determinant of rural poverty declines, there is significant spatial dependence in the growth rates of agricultural output. Irrigation is the primary driver of agricultural growth, and spatial variations in irrigation development seem to be associated with agro-ecological conditions which may be vastly different within Indian states, parts of which may be similar to those prevailing in geographically contiguous states. Poverty reduction strategies need to be designed in the light of spatial factors and using spatial methods. JEL classification: I32, O13, R12 Keywords: Rural poverty; Agricultural growth; India; Spatial econometric methods 1. Introduction The relationship between agricultural growth and rural poverty is one of the most contentious issues in development economics. In the Indian case, there is a vast empirical litera- ture that examines whether agricultural growth trickles down and to what extent (Ahluwalia, 1978; Saith, 1981). Recent stud- ies in this area use cross-sectional or panel data on agricultural growth and rural poverty and take Indian states as units of anal- ysis (Datt and Ravallion, 1998; Besley and Burgess, 2000). The use of Indian states as units of analysis may mask the spatial patterns in the data; as we show in this article that there is con- siderable variation within states in rural poverty, and the level of poverty observed in a region within a state may be different from other regions in the same state but more similar to regions in geographically contiguous states. An important limitation of previous studies is that they do not address the additional complexities in econometric analysis that arise when spatial relationships exist. 1 ∗ Corresponding author. Tel.: 01602 593376; fax: 44 1603 451999. E-mail address: k.sen@uea.ac.uk (K. Sen). 1 A paper that examines the agricultural growth–rural poverty relationship at the sub-state level is Palmer-Jones and Sen (2003), but this paper does not address the econometric problems that arise in the context of spatially dependent data. Spatial relationships between agricultural growth and rural poverty may exist for two reasons. First, as the literature on spa- tial poverty traps in developing countries emphasizes, decisions taken by one agent in a given location may influence decisions by neighbors (Ravallion and Jalan, 1996, Jalan and Ravallion, 2002). In the context of agricultural technologies in developing countries, these spillovers may occur due to learning processes based on information dissemination or to the intrinsic charac- teristics of these technologies that may make these technologies appropriate in regions with a certain type of resource endow- ments as compared to other regions (Brennan, 1989; Traxler and Byerlee, 2001; Conley and Udry, 2002). Second, spatial dependence may also occur in the error structure of the es- timated relationship if, for example, unobserved soil quality in one location is correlated with soil quality in neighboring locations. In this article, we make the argument that rural poverty in India, and its principal proximate correlate, agricultural growth, are spatially correlated, and that underlying agro-ecological and hydro-geological factors may exert an important influence on spatial correlations in rural poverty and agricultural growth. To make this argument, we proceed in three steps. First, we present statistical tests to identify spatial autocorrelation in the key variables of interest—rural poverty and agricultural growth. Second, we use spatial econometric methods to explore the de- terminants of rural poverty and agricultural growth, and show c  2006 International Association of Agricultural Economists