cases where the degree of income variance can be gleaned from published research. He finds that in systems where income variance is high, 17 percent exhibit high levels of rule conformance and maintenance; where income variance is moderate, 75 percent exhibit high levels of these performance measures; and where variance is low, 89 percent exhibit high rates of performance. (Tang cautions against inferring too much from these results, because the degree of income variance could not be identified in a significant fraction of the case studies he compiled.) Lam’s (1998) regression analysis shows that income inequality (measured by a zero-one variable indicating either “low/medium” or “high” variance in average annual family income) is significantly and negatively related to water delivery performance. Income inequality is also significantly and negatively related to productivity in the Nepali systems, but it is not significantly related to physical condition of the system. Khwaja (2000) finds a U-shaped relationship between inequality in one form of income—project returns—and project maintenance in his Pakistan study. Starting from a low level of inequality in project returns among beneficiaries, increased inequality tends to reduce project maintenance. Beyond a certain level of project-return inequality, however, maintenance improves as inequality widens.

Wealth inequality. Wealth inequality is likely quite highly correlated with income inequality, and its effects are similar here. Bardhan (2000) and Dayton-Johnson (2000a, 2000b) compute the Gini coefficient based on irrigated land holding for their Indian and Mexican studies. The Gini coefficient is related to performance: The relationship, where it is significant, is negative in the Indian study. Bardhan finds that landholding inequality is significantly and negatively associated with canal maintenance in the Tamil Nadu systems. For Bardhan’s indicator of intravillage conflict over water, he finds evidence of a U-shaped relationship between the Gini coefficient and this indicator of performance.19 That is, at low and high levels of inequality, there is little intravillage conflict, but for inequality in the middle range, conflicts are more likely. Bardhan finds no statistically significant effect of inequality on rule conformance. For the Mexican study, the full effect of landholding inequality on maintenance (accounting for the indirect effect on the choice of rules) is negative, but complicated.20 Khwaja (2000) once again finds a U-shaped relationship between landholding inequality and project maintenance in his Pakistan study: Starting at perfect equality, increasing inequality reduces maintenance, while at high inequality levels, increasing inequality raises maintenance.

Head-enders and tail-enders. Another source of inequality is the asymmetry between those at the head and tail ends of the canal network. As noted, this is probably only imperfectly correlated with inequalities in landholding wealth given that land markets do not function very well. Tang (1992:60-63,73-74) considers the impact on rule conformance and maintenance of the presence of “disadvantaged groups.” In most cases, this refers to tail-enders, although in a few instances it refers to groups against which system rules systematically discriminate. In a



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement