Bundling of Basic Public Services and Household Welfare in Developing Countries: An Empirical Exploration for the Case of Peru

Abstract

Using panel data for Peru for the period 1994–2000, we found that increases in household welfare, as measured by changes in consumption, are larger when households receive two or more services jointly than when services are provided separately. Such increases appear to be more than proportional, as F-tests on the coefficients of the corresponding regressors confirm. Thus, we found that bundling of services may help realize welfare effects. This finding is particularly robust in the case of urban areas.

Notes

¹ The Living Standards Measurement Study was designed by the World Bank in 1980 to explore ways of improving the type and quality of household data collected by statistical offices in developing countries and to explore ways to increase the accuracy, timeliness and policy relevance of household survey data collected in developing countries. Specific modules include: consumption, education, health employment, anthropometry, non-labour income, housing, price data, environmental issues, fertility, household income, savings, household enterprises and time use.

² Since these surveys were not planned as panel surveys beforehand, the sizes of samples, while always representative at the national level, were allowed to vary. In fact, a weakness of these data is that when disaggregating between rural and urban areas the econometric power of the data is sometimes reduced significantly, as seen later.

³ The 1986 panel is not representative at the national level, and is only so for Lima, the capital.

⁴ The prevalence of poverty is highest in rural areas where about two-thirds of households are poor and about 50% of the population is considered extremely poor (World Bank, 1993).

⁵ Glewwe and Hall argue that the fixed effects in the error terms disappear, and bias is reduced. They explain that this approach increases efficiency.

⁶ For the sake of economy and since they are very similar to the findings presented in this paper, we do not report these findings but would be more than willing to provide them upon request. We also tested an instrumental variables approach but, unsurprisingly, we were not able to come up with credible instruments. Our thanks to an anonymous referee for comments on this section.

⁷ A second approach was to control not only for same household head, but also for same household size. Another approach was to use a somewhat different definition of our dependent variable. Instead of using the log of the real change in consumption, we used the relative change in consumption. To do this, we ranked the households according to consumption per capita in 1994 and calculated a cumulative distribution function for the households. We repeated the procedure for 1997. Then we calculated the ratio between both years, and used this series as our dependent variable. This approach has been used by Lanjouw (1998). Further sensitivity analysis was performed by using total wealth in 1994 and 1997 instead of consumption. For the most part, our results do not change.

⁸ Combinations of three services were also tested and yielded non-statistically significant coefficients.

⁹ These results are available upon request.

¹⁰ Notice, however, that all our regressions in Table 1 included a dummy for urban area.

¹¹ We also tested a panel from 1994 to 2000, which yields positive and statistically significant coefficients for the case of bundling in two, three and four services in the case of urban areas but yields positive although weakly statistically significant coefficients for rural areas in the case of two-utility bundling and no statistical significance when combining three utilities.

¹² We gratefully acknowledge the comments of an anonymous referee in the development of this section.

¹³ We were not able to apply a household fixed effects approach as we ran of out of degrees of liberty.

¹⁴ For reasons of space we do not report full regressions, although they may be provided upon request.

¹⁵ As before, the marginal impact of adding one public service to the welfare of the household tends to be statistically significant at conventional levels, which is confirmed when F-tests are applied.

¹⁶ Tt is somewhat unclear whether receiving multiple services as a household is equivalent to the government providing a bundle of services. Having multiple services may not reflect differences in the form of government provision but differences in household wealth. This issue is beyond the scope of our research.

Additional information

Notes on contributors

Alberto Chong

Anonymous referees provided useful suggestions on previous versions. We are also grateful to Vajeera Dorabawila, Virgilio Galdo and Pablo Suárez for excellent research assistance. The views and opinions in this paper should not be attributed to the Inter-American Development Bank, the World Bank, or their corresponding executive directors. The standard disclaimer applies.