First Prize Erich

ing from detail, it is instructive to consider what would happen in a perfectly flexible economy, if schooling actually displayed stagnant productivity while all other sectors faced a constant positive rate of productivity growth. The outcome would be what Baumol (1967) has called the cost disease of services. With a functioning labour market, the wages of workers in the stagnantproductivity sector would have to increase in line with those of workers in the other sectors where labour productivity increases. Given an efficient allocation of resources, the price per unit of output of the stagnant-productivity sector would rise with the average rate of productivity growth of the other dynamic sectors. Applying this insight to the case of schooling suggests that a constant-quality unit of schooling output should become more costly over time. In addition, one could even predict the efficient s i z e of the expected cost increase of schooling if the average rate of productivity growth of the dynamic sectors of the economy were known. If average economy-wide productivity grew by 2 percent per year and efficiency conditions prevailed, then the GDP-deflated price of schooling (and of all other stagnant-productivity sectors) should grow by 2 p e rcent as well (Gundlach et al., 2001). That is, public expenditures per student should rise by 2 p e rcent in an economy which grows by 2 percent if schooling productivity had remained constant. H o w e v e r, if public expenditures per student grew by more than average productivity growth, then either schooling productivity should have declined or the quality of schooling output in the form of student perf o rmance should have increased. We begin with the latter possibility. The main problem with an empirical analysis is exactly the difficulty of correctly measuring potential changes in the quality of schooling output. In principle, changes in the quality of schooling output can be measured by the results of standardized student achievement tests at various points in time. However, consistent time-series data on the cognitive achievement of students are available only for the United States (2). These tests suggest that there has been no substantial change in the average performance of US students between 1970 and 1994 (Hanushek, 1997). 2) The National Assessment of Educational Progress (NAEP) began to monitor the performance of students aged 9, 13 and 17 years, in mathematics and science in the early 1970s. The NAEP used the same assessment content and the same administration procedures over time, so the reported average test scores of US students are comparable intertemporally. Schooling is labour intensive and the methods used may not have changed much over the past quarter century.