Improving the Quality Versus Increasing the Quantity of Schooling: Evidence for Rural Pakistan

Interest in estimating the impact of school quality on earnings has increased. But most studies on this topic have important limitations, particularly in studies for developing countries. They tend to ignore behavioral decisions regarding schooling and individual and family background characteristics, use school quality measures aggregated to the regional level, and rely on crude indicators of teacher quality. These limitations may explain why the micro evidence of important school quality effects is scant. Moreover, the question of the relative rates of return, in terms of earnings, to improving school quality versus raising quantity has not been addressed. The data demands for estimating such rates of return are considerable. This paper presents a conceptual framework for undertaking such estimates, uses special data collected for this study, and makes estimates within a framework that controls for important individual and household choices, including the duration of schooling and subsequent participation in the wage labor market. Subject to qualifications because such an ambitious strategy stretches the limits of the even the special data that we collected, the estimates suggest that in rural Pakistan rates of return are much higher for investing in primary school quality or quantity than for investing in middle schools and, at the primary school level, somewhat higher for expanding low-quality schools (predominantly for girls) than for increasing quality in existing schools. More generally, the methods adopted here permit a more satisfactory assessment in developing economies than previously of the rates of return to improving school quality versus increasing quantity and their implications for educational policy. *Behrman, University of Pennsylvania; Ross, Bryn Mawr College; and Sabot, Williams College. This paper builds on results generated by the Human Capital Accumulation in Post Green Revolution Pakistan Project of the International Food Policy Research Institute (IFPRI). We are grateful to the World Bank and USAID for financial support; to Mary Bailey, Meg Ewing, Emily Mellott, and Matthew Tropp and Amy Whritenour for able research assistance; and to participants in the North East Universities Development Consortium Conference at Harvard and in a seminar at Williams College for helpful comments. The views presented here are those of the authors and should not be interpreted as reflecting the views of IFPRI, USAID, or the World Bank. 1 Studies of the impact of school quality on earnings include Akin and Garfinkel (1977), Anderson, Shugart and Tollison (1991), Behrman and Birdsall (1983), Behrman, Birdsall and Kaplan (1996), Behrman, Rosenzweig and Taubman (1996), Betts (1995, 1996a,b), Card and Krueger (1992a,b), Daniere and Meechling (1970), Grogger (1996a,b), Heckman, Layne-Farrar, and Todd (1996), James, Alsalam, Conaty and To (1989), Johnson and Stafford (1973), Jud and Walker (1977), Link and Ratledge (1975a,b), Link, Ratledge, and Lewis (1976, 1980), Margo (1986), Morgan and Sirageldin (1968), Nechyba (1990), Reed and Miller (1970), Ribich and Murphy (1975), Rizzuto and Wachtel (1980), Solmon (1973), Solmon and Wachtel (1975), Wales (1973), Wachtel (1976), Weisbrod and Karpoff (1968), Welch (1966, 1973a,b), and Wolfle (1973). Virtually all of the existing studies of the impact of school quality on earnings treat schooling as predetermined (Behrman, Rosenzweig and Taubman 1996 is an exception in which schooling is posited to respond to individual and family endowments). All of the aggregate studies and many of the micro studies of the impact of school quality on earnings ignore family and individual characteristics (exceptions among the micro studies include Akin and Garfinkel 1977, Altonji 1988, Altonji and Dunn 1996a,b, Behrman, Rosenzweig, Taubman 1996, Link and Ratledge 1975a,b Solmon 1973, Taubman 1975, Wachtel 1976). Betts (1995) is an example of a study that uses school-level characteristics and does not find much impact. Examples of studies of the impact of school quality on earnings that use schooling measures defined for districts or states include Behrman and Birdsall (1983), Behrman, Birdsall and Kaplan (1996), Card and Krueger (1992a,b), Ribich and Murphy (1975), Rizzuto and Wachtel (1980), Wachtel (1976) and Welch (1966, 1973a,b). 1 Estimation of the impact of time spent in school on earnings and other outcomes has been a major area of research. But time spent in school is an input, not an output of schooling. School quality also is likely to influence such outputs as cognitive and other skills and the earnings they yield. There may be important tradeoffs between increasing time spent in school (higher school quantity) and improving school quality. Three separate, but related, literatures address dimensions of the school quality-cognitive achievementearnings nexus and continue to attract much scholarly and policy interest (see, for example, the Symposium on School Quality and Educational Outcomes introduced by Moffitt 1996, and Card and Krueger 1996). First, there is a fairly large literature on the impact of school quality on earnings. However, the studies in this literature usually rely on crude proxies as measures of school quality, ignore human capital endowments and inputs provided in the home, and treat all aspects of schooling, even time in school, as predetermined. The estimated impact of school quantity and quality may be biased if home characteristics and ability affect earnings through choice of time devoted to schooling. Bias also can result if indicators of school quality do not include such direct measures of teachers' quality as their cognitive skills, even though how much a teacher knows is recognized as critical to how much a child learns. Perhaps as a result, many micro studies do not find much of an impact of school quality on earnings, or find negative effects of school attributes about as frequently as they find positive effects. Many studies also suffer from measurement error resulting from aggregation: Despite considerable within-geographical area variation in school inputs, they use the average of school characteristics for school districts or states rather than school characteristics for the schools actually attended by the individuals studied.1