Lagos and Benin City are rapidly growing metropolitan areas with urban water markets, where multiple suppliers compete with a state-run water utility. The lack of household access to potable water is a major problem throughout the study area. While the literature on urban services delivery in poor countries has documented the rise of non-state providers and citizen willingness to pay in studies of cost recovery, there is a weak understanding of the variables associated with citizen behavior and consumer choice in urban water markets across different types of households and communities. In this study, three sets of dynamics are identified: (1) How individuals (and communities in the aggregate) respond differently to problems with water supply, (2) the kinds of triggers that produce specific responses, and (3) the characteristics of individuals and households that respond in different ways. In the conceptual model used, access to water shapes attitudes that affect both its consumption and provision. Household behavior is modeled using an expansion of Hirschman's Exit, Voice and Loyalty (EVL) framework to characterize the quality-elasticity of demand. Multinomial logistic regression is used to model the three discrete outcome categories.

Survey results show that exit and loyalty are the most frequently reported response to recent water problems, although exit dominates in Lagos while loyalty is the main response in Benin City. Surveys and interviews show that reported EVL behavior is determined by different access variables, such as usage, availability, affordability, the duration of the problem, and the degree of satisfaction with household water supply in comparison with neighbors. The findings also indicate how neighborhood composition and organization could be related to processes of social and economic deprivation, silencing the preferences of those who are unable to seek alternatives or to speak out. Data are from a household survey (n=784), along with 31 semi-structured interviews, personal observations, and archival materials. Respondents were sampled from 18 neighborhoods in 7 local governments using a multi-level cluster sampling strategy.