and performance. Both may vary systematically with the degree of heterogeneity. With respect to institutions, Tang (1994:231) describes two types of “rules-in-use”: boundary rules (“the requirements one must fulfill before appropriating water”) and authority rules (the procedure and basis for withdrawal, including fixed shares or rotating turns). Bardhan (2000) and Dayton-Johnson (2000a) consider cost-sharing rules for mobilizing canal-cleaning and maintenance efforts, as well as water allocation rules that define households’ claims on irrigation resources.
Performance is measured in various ways. An obvious dimension is the degree to which irrigators adhere to the rules established above: rule conformance. Bardhan (2000) measures whether water allocation rules are frequently violated by one group; Tang (1994) codes more generally whether the irrigation rules are followed. Alternatively, one could measure not rule conformance per se, but rather the level of infrastructure maintenance. Bardhan (2000) uses a categorical-variable index of maintenance of distributaries and field channels. Dayton-Johnson (2000b) uses disaggregated variables and estimates statistical models of three dimensions of maintenance: the degree of definition of canal side slopes, state of repair of field intakes, and degree of control of leakage around the canals. Lam (1998) uses the overall physical condition of the system. Another class of performance variables measures the adequacy of water delivery; Lam aggregates information on adequacy of water delivery at various points in the system, equity among users, and reliability of water supply at the tail end. An imperfect indicator of the success of irrigation is crop yields, considered by Dayton-Johnson (1999) and by Lam (1998), who aggregates information on output per hectare, and cropping intensity at the head and tail ends of the system. Lam subjects the three dimensions of his performance measure (condition, water delivery, and productivity) to confirmatory factor analysis, finding, among other things, that the dimensions are not highly correlated. Fujita et al. (2000) propose a four-dimensional concept of irrigation system performance, based on the existence of rules for maintenance, coordination in rice-cropping schedules, practice of water rotation, and organized monitoring of rules. They perform a principal components analysis on the four measures to derive appropriate weights for an index of performance. Finally, Bardhan (2000) also considers the absence of water-related conflicts as a measure of performance.
All of the studies considered here are multivariate analyses, but the list of independent variables differs considerably from study to study. To some extent, then, we are comparing estimated coefficients that are not strictly comparable: This should be borne in mind when considering the results reviewed in the following paragraphs.
Income inequality. What then, are the effects of heterogeneity? Consider first inequality in incomes. Tang (1991) finds that “a low variance of the average annual family income among irrigators tends to be associated with a high degree of rule conformance and good maintenance.” Tang (1992:72-73) identifies 27