Appendix 1.
|
Richard Rothstein
|
In 1967, public elementary and secondary schools spent $29.6 billion, or $687 per pupil enrolled in grades K-12. By 1991, spending jumped to $229.4 billion, or $5,566 per enrolled pupil. However, the fact that per pupil spending grew by 710 percent over this quarter century does not tell us the degree to which we have devoted more real resources to education. Much of this increase has been caused by inflation: the prices of most goods and services purchased by schools have gone up each year.For instance, if food prices rise by 5 percent, families must increase their food budgets and expenditures by 5 percent just to maintain their food consumption. Similarly, schools faced with a 10 percent rise in the price of textbooks must increase textbook spending by 10 percent to provide students with the same number of textbooks. To measure historical growth of real per pupil resources requires knowledge of the inflation, or price increases, in goods and services purchased by schools. What we want to understand is the degree to which more real resources are now used by schools and, if so, whether greater resource intensity generates better outcomes.
Examinations of changes in school spending over time must use some measure of inflation to convert 1967 spending to its equivalent in 1991 dollars. We can then speak of "real" (or "inflation-adjusted") as opposed to "nominal" ("unadjusted") school spending growth. Most analysts make this conversion by use of the "consumer price index for all urban consumers" (CPIU), the conventional measure of inflation provided by the BLS.\1\ Using the CPI-U, $687 in 1967 dollars becomes $2,794 in 1991 dollars. In real terms, therefore, per pupil expenditures went from $2,794 to $5,566, or a quarter-century jump of 99 percent. As Benno Schmidt claimed, we "roughly doubled" real school spending.
It is probable, however, that use of the CPI-U for this purpose causes an overstatement of school spending growth. The inflation rate for school purchases is likely to be greater, and will continue to be greater, than the average urban consumer's price inflation that the CPI-U is intended to measure. Table 2 reviews inflation rates for a range of goods and services. These data show that price increases for particular items can be different from price increases for the "average" items included in the market basket of goods and services used to calculate the CPI-U. For instance, inflation in medical care (681 percent) from 1967 to 1991 was much greater than the average for all items, while inflation in commodities like food and manufactured products (344 percent) was less than the average. Because inflation rates vary widely among particular items, it is important to determine carefully the appropriate inflation index to use for converting nominal spending into real changes.
If a family bought the average market basket of goods and services in 1967, and then spent 408 percent more in 1991, it could still buy similar goods and services in 1991 because "all items" inflation was 408 percent. But consider a family that purchased an above average amount of medical care in 1967 and whose total spending also increased by 408 percent by 1991 (i.e., less than the medical inflation of 681 percent). In order to maintain its standard of living in other respects, this family would have been forced to reduce the amount of medical care services (or an equivalent amount of other spending) it purchased by about a third, because medical care prices rose faster than average prices. In contrast, consider a family that purchased an above average amount of commodities in 1967 and whose spending also increased by 408 percent by 1991. This family could improve its living standards, purchasing significantly more commodities (or other items), because commodity inflation (344 percent) was relatively low.
Table 2 also shows that prices for commodities have grown more slowly than prices for all services (344 percent vs. 508 percent). A similar contrast is evident when food and medical care are removed from their respective groups: nonfood commodity (primarily manufactured goods) inflation was 314 percent, roughly two-thirds the 489 percent inflation in "services other than medical care."
Inflation in services exceeds inflation in goods or commodities because productivity (the increase in output per employee hour worked) has grown more slowly in services. Productivity growth in manufacturing, for instance, has allowed industrial firms to reduce their costs (or at least slow the growth in costs) and therefore increase the prices of manufactured products more slowly or not at all. In contrast, many service-sector firms cannot automate their production as manufacturers do; these service firms, for whom it is more difficult to achieve productivity growth, have had to increase prices faster than average. Often cited examples include barbers and orchestras: barbers cannot greatly increase the number of haircuts they perform per hour, and orchestras cannot perform music with fewer musicians each year. These insightsthat disparities in inflation mirror differences in productivity growth, and that industries (i.e., services, barbers, orchestras) in which it is hard to achieve productivity growth will have higher than average inflationare associated with the work of William Baumol (Baumol 1967; Baumol, Blackman, and Wolff 1989). Baumol refers to low productivity sectors as having a "cost disease," and the faster inflation in sectors with relatively slow productivity is generally referred to as the "Baumol effect." Table 3 elaborates how differences in productivity between industries will, in the context of a national labor market, generate differences in inflation rates. Table 3 also illustrates how differences in the price changes (i.e., inflation) of individual industries are driven by differences in productivity growth when all industries increase wages at the same rate, as would be expected in a national labor market, assuming each industry's work force has the same skills and education. Table 3 presents examples of two industries, each of which has 100 workers producing 1,000 units in year one. That is, the examples are constructed so that both industries have the same productivity level of 10 in year one. Because each industry also pays its workers the same (i.e., $20,000), they also have the same price level in the first year of $2,000 per unit.
What happens to the prices of the goods produced in these industries when one industry (Industry A) experiences a 10 percent increase in productivity but the other industry (Industry B) has no productivity growth? We assume that wages increase by 10 percent (reflecting the 5 percent average productivity growth in the economythe average of 10 percent and zero percentand five percent inflation). In Industry A, the productivity growth of 10 percent offsets the 10 percent wage increase so that prices do no increase in year two. Industry B, however, enjoyed no productivity growth but did face 10 percent higher wages, the same as Industry A. The result is that the price of Industry B's goods increased by 10 percent. Thus, an industry that pays comparable wages, for comparable workers, but has low productivity, will experience faster inflation.
Education is subject to the Baumol effect because productivity improvements from cost reductions are difficult to achieve in education. In contrast, manufacturing and telecommunications industries are able to automate work and find efficiencies in use of materials; and thereby reduce the resources needed in production and realize productivity gains. From 1967 to 1991, the private sector achieved productivity growth of 1.1 percent per year, or 30 percent overall. This means that the number of workers necessary to produce an average product fell roughly a third from the beginning to the end of this period. What would a comparable growth in labor productivity look like in schools? Assume that schools use only one resource, teachers, and the pupil-teacher ratio was 20:1 in 1967. Then, if 30 teachers were necessary to educate 600 students in 1967, and if schools could have increased productivity the way the private sector did (by reducing labor inputs and using remaining inputs more efficiently), a 30 percent productivity growth would imply that only 23 teachers were necessary in 1991; in other words, the pupil-teacher ratio would have to rise from 20:1 to 26:1. With only 23 teachers, school cost increases would be in line with the national economy.\2\
While education reform should certainly be on the public agenda, continuous industrial-like realization of cost efficiencies are probably not what the public has in mind. Education costs will rise faster than economy-wide inflation, so real spending per pupil as measured with an average inflation index will rise even though per pupil resources are not growing.\3\ This is illustrated in table 4. This table illustrates how spending per pupil will necessarily rise if there is not any productivity growth or increase in cost efficiencies. For instance, a school with a pupil/teacher ratio of 20:1 that pays teachers $20,000 annually will be spending $1,000 per pupil, assuming, of course, there are no expenses other than teachers. If wages in the economy, and for teachers, grow 10 percent, then spending per pupil will also rise 10 percent, to $1,100. The cost efficiencies necessary to offset higher wages require that the number of pupils per teacher rise to 22.2. Schools are then faced with a continuous rise in number of pupils per teacher or steadily rising spending per pupil, a measure of school costs or inflation, at least when compared to other sectors that can achieve greater cost efficiencies over time.
A related insight of William Baumol is that because productivity improvements are spread unevenly throughout the economy, changes in prices over time will also vary across products. Consumers, therefore, will spend a greater share of incomes to purchase a constant level of products or services in some sectors and a smaller share to purchase a constant level in others. That is, we must increasingly spend a larger share of our incomes on low productivity goods and services that have more rapid price increases (like education) just to maintain the same level of consumption.
It is thus inevitable that inflation in a low productivity industry like education will be higher than inflation in an average industry experiencing average productivity gains. For this reason, use of the average inflation rate for consumer goods and services (the CPI-U) systematically understates the inflation facing school districts. Put another way, a measure of average inflation to deflate school spending trends will systematically mislead by overstating how much "real school spending" has grown. It will give the impression that more of the nominal spending growth represents real new resources provided to school districts for educating students, and that less of the nominal spending growth represents inflation, than was in fact the case. The issue, then, is whether we can select a more appropriate index to use for analysis of school spending.
Despite problems with use of the consumer price index to interpret historical changes in school spending, few researchers have attempted to create an inflation index specifically tailored to education (although the education research community is increasingly sophisticated about regional differences in the cost of living, a conceptually similar issue).\4\ Kent Halstead constructed one index that extends back to 1975 (Halstead 1983 and Research Associates 1993), but no others have attempted to replicate Halstead's work, so its accuracy lacks independent verification. Halstead's index has a theoretical drawback that further militates against its use in the present study.
Halstead constructed his school price index (SPI) by examining price changes for a "market basket" of 42 items typically purchased by elementary and secondary schools in 1975 (Halstead 1983, 138). In 1975, elementary and secondary schools spent 47.68 percent of their budgets on teacher salaries, 3.75 percent on student transportation, 0.7 percent on textbooks, and 1.1 percent on electric power, etc.\5\ By assembling a price series for each of these items, making estimates where necessary, Halstead calculated what it would cost public schools to buy an identical (ignoring most quality improvements) collection of goods and services in each subsequent year. He identified this growth as the school inflation rate, so spending above this rate represented real spending increases.
The Halstead index is not used in this report for two reasons. First, it is not available for the entire 1967 to 1991 period, and second, its treatment of teacher salaries is questionable. Halstead's SPI includes a price series for elementary and secondary teachers based on their actual salary changes. However, what schools pay teachers reflects districts' choices about whether to pay teachers more or less than comparable workers. These choices may be influenced not only by district officials but by legislators and teacher unions as well. When teacher salaries rise relative to salaries of workers with comparable education and experience in other fields, we can presume that schools are upgrading the skill levels of their work force (in other words, providing additional inputs, more "real" resources to students). But if teachers' salaries fall relative to those of similarly educated professionals, then school districts will have a harder time attracting the best qualified teachers, and there will be an erosion in the teacher skill base. Variance from market norms can be considered either an effort to attract a better (or worse) than average quality work force, or the provision of a "rent" (positive or negative) to teachers by either overpaying or underpaying them.
It would have perhaps been more appropriate for Halstead to base his index on all college educated or professional workers, a group "comparable" to teachers. Then, the degree to which schools pay teachers or other school employees more than the market rate would not be obscured by a school price index that ignores the salaries of comparable workers. Conversely, a fall in teacher pay relative to "comparables" would result in a measured decline in real resources provided for students. In the absence of a conceptually correct index, an assessment of real school spending must rely upon some combination of available indices for particular items developed for the CPI-U. One reasonable choice is to use the inflation measure for "services," because schools are a service type industry with "cost disease"/slow productivity characteristics. The actual service index of the CPI-U, however, includes two heavily weighted items that strongly affect the measured inflation rate but that are not relevant to education. Shelter rent (housing) inflation makes up a large part of the service CPI-U and should be excluded. Medical care also has an exceptionally high inflation rate caused by unique characteristics of the health care sector that are not applicable to education. For this reason, the index developed for this reportthe "net services index" (NSI)reflects price increases of services provided to consumers exclusive of shelter and medical care. "Net services" includes items such as entertainment services, personal care services, personal and educational services, public transportation, auto repair, private transportation (other than cars), housekeeping services, and utilities and public services. These tend to be labor-intensive services with low productivity growth (relative to goods or to the average) and therefore are items where increased cost efficiencies are hard to achieve. If schools rely on professional, college educated workers more than do the sectors in "net services" (as is reasonable to believe), then "net services" will still understate school inflation (because wages for educated workers have risen faster than average over the 1967-91 period). Appendix 2 provides technical detail on how the NSI was constructed, nationally, and for each region and local area.\6\
Application of the national net services index to education spending is shown in table 5. These data show that the $687 spent per pupil in 1967 was equivalent to $3,456 in 1991 dollars. Since 1991 per pupil spending averaged $5,566, we conclude that real school spendingreal per pupil resources provided to schoolsincreased by about 61 percent.\7\ Table 5 also shows measured growth in real school spending using the "all items" CPI-U to be 99.2 percentthe much discussed "doubling" of school spending. Selection of the net services index suggests a nearly 40 percent slower growth in school resources than conventional accounts based on the conceptually inaccurate (for this purpose) "all items" CPI-U.
In sum, choice of an inflation measure dramatically affects the portrait of school spending growth. The magnitude of the measurement error from applying the "all items" index cannot be precisely determined because an appropriate school index is not available, but construction of an index from the CPI-U services component, with medical care and housing excluded, seems to be the best alternative. So while it seems certain that conventional estimates have vastly overstated the growth in school resources, the 61 percent growth presented in table 5 is an estimate that, while more accurate than conventional estimates, might still be too high or too low. Development of an improved inflation index for school spending should be a research priority.
References
Barro, S. M. 1994. Cost-of-Education Differentials Across the States. U.S. Department of Education, Office of Educational Research and Improvement. Working Paper No. 94-05.
Baumol, W. June, 1967. "Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis." American Economic Review. 57: 415-426.
Baumol, W., S. A. Blackman, and E. N. Wolff. 1989. Productivity and American Leadership. Cambridge: MIT.
Chambers, J. G. Winter, 1980. "The Development of a Cost of Education Index: Some Empirical Estimates and Policy Issues." Journal of Education Finance. 5.
Chubb, J. E. and E. A. Hanushek. 1990. "Reforming Educational Reform" in Setting National Priorities, edited by H. J. Aaron. Policy for the Nineties. Washington, DC: The Brookings Institution.
Halstead, D. K. 1983. Inflation Measures for Schools and Colleges. Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement, National Institute of Education.
Hanushek, E. A., et al. 1994. Making Schools Work. Washington, DC: The Brookings Institution.
McMahon, W. 1995. "Intrastate Cost Adjustments." Manuscript.
Parrish, T. B., C. S. Matsumoto, and W. J. Fowler, Jr. 1995. Disparities in Public School Spending, 1989-90. U.S. Department of Education, Office of Educational Research and Improvement. NCES 95-300.
Research Associates of Washington. 1993. Inflation Measures for Schools and Colleges, 1993 Update. Washington, DC: Research Associates of Washington.