Higher Education and Economic Growth

Higher Education and Economic Growth

Higher Education and Economic Growth

Edited by William E. Becker Professor of Economics Indiana University and Darrell R. Lewis Professor of Education Policy University of Minnesota

Springer Science+Business Media, LLC

Library of Congress Cataloging-in-Publication Data Higher education and economic growth/edited by William E. Becker and Darrell R. Lewis. p. cm. Includes bibliographical references and index. ISBN 978-90-481-5792-1 ISBN 978-94-015-8167-7 (eBook) DOI 10.1007/978-94-015-8167-7

1. Economic development - Effect of education ono 2. Higher education-Economic aspects. 1. Becker, William E. II. Lewis, Darrell R. HD75.7.H54 1992 92 6291 338.9-dc20 CIP Copyright © 1993 by Springer Science+Business Media New York Originally published by Kluwer Academic Publishers in 1993 Softcover reprinl of Ihe hardcover 1si edilion 1993 AII rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any torm or by any means, mechanical, photo-copying, record ing, or otherwise, without the prior writlen permission of the publisher, Springer Science+Business Media, LLC.

Printed an acid-free paper.

Contents

Contributors

vii

About the Editors

viii

Preface

ix

Preview of Higher Education and Economic Growth William E. Becker and Darrell R. Lewis

2

Higher Education and Economic Growth Stephen A. Hoenack

21

3

Higher Education, Economic Growth, and Earnings John Pencavel

51

4

Is Public Education Productive? David Alan Aschauer

87

5

The Contribution of Higher Education to R&D and Productivity Growth Walter W. McMahon

105

6 The Effects of Higher Education on Unemployment Rates Wayne J. Howe

129

v

Vl

CONTENTS

7 Quality of Higher Education and Economic Growth in the United States Lewis C. Solmon and Cheryl L. Fagnano

145

8 Higher Education, Business Creation, and Economic Growth in the American States Bryan D. Jones and Arnold Vedlitz

163

Contributors

David Alan Aschauer, Elmer W. Campbell Professor of Economics, Bates College William E. Becker, Professor of Economics, Indiana University Cheryl L. Fagnano, Vice President, The Milken Institute for Job & Capital Formation Stephen A. Hoenack, Professor and Director, Management Information Division, Hubert H. Humphrey Institute of Public Affairs, University of Minnesota Wayne J. Howe, Statistician Clinical Audit Consultant, Aetna Life & Casualty Insurance Company Bryan D. Jones, Professor, Department of Political Science, Texas A&M University Darrell R. Lewis, Professor of Educational Policy, and Professor of Education, University of Minnesota Walter W. McMahon, Professor of Economics, and Professor of Education, University of Illinois John Pencavel, Professor of Economics, Stanford University Lewis C. Solmon, President, The Milken Institute for Job & Capital Formation Arnold Vedlitz, Professor of Political Science, Texas A&M University vii

About the Editors

William E. Becker is a professor of economics at Indiana University. He is editor of the Journal of Economic Education and serves on the editorial board of the Economics of Education Review. His research appears in the American Economic Review, American Journal of Agricultural Economics, Econometric Theory, Journal of Finance, Journal of Human Resources, Journal of Risk and Insurance, Monthly Labor Review, Review of Economics and Statistics, and other journals. He is also the co-author of Business and Economics Statistics, and co-editor of The Economics of American Higher Education, Academic Rewards in Higher Education, and Econometric Modeling in Economic Education Research. Darrell R. Lewis is professor of educational policy and higher education at the University of Minnesota. His research interests have been in the economics of education with a current focus on issues relating to educational efficiency and postschool outcomes. He is the author or co-author of numerous articles, monographs, and books. He has co-edited The Economics of American Higher Education, Academic Rewards in Higher Education, and Faculty Vitality and Institutional Productivity, and coauthored And on the Seventh Day: Faculty Consulting and Other Supplemental Income Activities, and Assessing Outcomes, Costs and Benefits of Special Education Programs.

viii

Preface

After decades of effortless growth and prosperity, America's postsecondary institutions of education have come under increasing financial stress and waning public support. In part, this stress reflects a slowdown in the real rate of national economic growth and the loss of federal and state revenues for education generally. It also reflects a trend of state legislatures simply giving higher education an ever lower ranking on the list of funding priorities. Postsecondary educational institutions in the United States will continue to face increasing financial stress and waning public support as critics question the contribution of higher education to economic growth, which historically has been a major rationale for funding. Unless the trends in education financing can be changed, higher education can be expected to stagnate. What, if anything, can be done? As a starting point, advocates of higher education need to more fully recognize the important ways in which higher education influences technological change and also is influenced by that change. As demonstrated by the chapters in this book, higher education is not a neutral or passive player in economic growth. This volume addresses topics related to the role of postsecondary education in national economic development within the United States. Attention is given to the research, teaching, and service missions of higher education in stimulating economic growth. The social rate of return, aggregate production function, and other aggregate measures of the contribution of education to economic growth are considered in detail. Little attention, however, is given to the importance of colleges and universities in the enhancement of individual students; this topic is covered in our related book The Economics of American Higher Education (Kluwer Publishing, 1992). IX

X

PREFACE

The chapters in this volume summarize the research literature and synthesize what economists and other social scientists have learned about the contribution of higher education to economic growth within the American society. Many of the chapters were originally commissioned by the editors for a seminar on these topics at the University of Minnesota during the spring of 1989. The seminar was financed by grants from the Hubert H. Humphrey Institute and the College of Education at the University of Minnesota. To the deans and faculty of both units, as well as David Berg and Stephen Hoenack (Management Planning and Information Services), we are indebted for institutional support and collegial interaction. Additional support for manuscript preparation was provided by the Joint Council on Economic Education, for which we are most grateful. In Higher Education and Economic Growth we have assembled a group of recognized scholars from economics and the other social sciences who have individually made significant contributions to the literature in higher education. The authors draw heavily from their own research in providing current evidence while focusing on their assigned chapter topics. To the nine contributing authors we are thankful. Without their work this book would not have been possible. Suzanne Becker provided manuscript editing as part of her duties as assistant editor of the Journal of Economic Education. As always, her patience, eye for detail, and skill with the English language are acknowledged and appreciated. Together with the reference librarians at Indiana University, whose services are gratefully acknowledged, Sue did all the final reference work. Manuscript preparation was completed in the Department of Economics at Indiana University through the capable word porcessing skills of Elaine Yarde, to whom we are indebted. William E. Becker Darrell R. Lewis

1

PREVIEW OF HIGHER EDUCATION AND ECONOMIC GROWTH William E. Becker and Darrell R. Lewis

The role of education in the economic health of the nation and the relationship between education and economic growth are increasingly the focus of public debate. In the 1980s, the educational reform literature began warning that our nation was "at risk" because of perceived shortcomings in our educational system. Even the cover of Business Week (September 19, 1988) joined the chorus by stating that "the nation's ability to compete is threatened by inadequate investment in our most important resource: people." Although the early advocates of educational reform aimed their criticism and remedies at primary and secondary education, more recently, the role of higher education has been drawn into the debate. This volume is intended to contribute to our understanding of the relationships between higher education and the economic growth and wellbeing of the nation. What is the role of postsecondary education in the economic development of the United States? What other relationships exist between economic growth and the research, service, and teaching missions of colleges and universities? In short, how and in what ways does higher education in the United States contribute to our nation's economic growth? A significant amount of evidence in the economics literature indicates that an investment in higher education is beneficial both to the individual and to society in general. Our earlier edited book, The Economics of American Higher Education (1992), dealt primarily with the effects of higher education on the gains enjoyed by the individuals. Here, we focus 1

2

HIGHER EDUCATION AND ECONOMIC GROWTH

on the gains to the nation as a whole. What follows is an overview of each of the chapters in this volume and a summary from the literature on the relationships between higher education and national growth. Recognition of Education's Contribution to National Income

The notion that an investment in education and human capital promotes economic growth can be traced to Adam Smith (1776) in his famous study on what constitutes the "wealth of nations," although in his earlier and less well-known work Smith questioned the merits of public education. 1 Not until after World War II, however, did studies attempt to directly link investment in education with incremental gains in national income. In the 1960s, the seminal work of Becker (1960, 1964), Schultz (1961), and Denison (1962) cast light on how, and to what extent, education contributes to the enhanced productivity of the labor force and, in turn, to growth in national income. These important findings have led to hundreds of subsequent studies on the economic value of all forms of education and training, including higher education. For a more detailed discussion of the many issues related to the economics of education, the reader should consult the excellent textbook by Cohn and Geske (1990) and the edited volume by Psacharopoulos ( 1987). Here we will concern ourselves almost exclusively with the role of postsecondary education in economic growth. The contribution of extended education (after high school) to economic growth is presumed to occur through a number of distinct yet interacting functions. First, it is believed that higher education contributes to economic growth through the production of knowledge and that this largely takes place within the major universities through faculty members' and their advanced students' research and creative activities. Second, it is generally acknowledged that colleges and universities contribute to national growth through the diffusion of knowledge, which may result from the external service activities of their faculty, staff, and students. Finally, it is universally accepted that postsecondary institutions contribute to the transmission of knowledge through extensive and varied teaching activities. Economists have focused their attention on this latter set of activities as measured by enrollments, man-years of postsecondary education completed, number of graduates, graduation rates, expenditures, and changes in student earnings. Curiously, smug scholars tend to emphasize only the creation of knowledge as their contribution to the nation's well-being, while eschewing the importance of their teaching in the advancement of economic growth.

PREVIEW OF HIGHER EDUCATION

3

Scholars typically give little thought to the importance of the service function of their college or university. On the other side of the equation, however, a significant portion of postsecondary demand can be attributed to the "consumption" interests of the student and the surrounding community in which the institution is located. That is, the personal satisfaction that the college or university "customers" derive from learning, entertainment, culture, and group activities associated with college life may not be trivial. There is little debate, however, that the primary reason why individuals undertake extended education and training in postsecondary instruction is an investment in self improvement. 2 Economists typically emphasize the screening and human capital value of instructional activities and their ability to increase the future earnings and productivity of the existing student body. Education, by its nature as an investment, has a delayed effect on the growth in national income. Attempts to measure the contribution of education to national growth have tried to capture this investment effect either by a comparison of the ex post financial rates of return associated with alternative levels of education, or the ex post change in productivity measures assigned to education within an aggregate production function, or, more recently, the rate of productivity growth convergence among countries with similar educational characteristics.

Rates of Return for Higher Education

The determination of how much a society should spend on roads, law enforcement, or education would be easy if the benefits that society can expect in return for an investment in these infrastructure components could be accurately calculated. Although numerous economists have attempted to determine the social rate of return to educational expenditures, estimates have been viewed with skepticism by both policy makers and the authors of the studies. There is no question that education benefits the individual student, in terms of both immediate satisfaction and future options and earning possibilities. Education also benefits society above and beyond the private benefits enjoyed by the individual receiving the education: education enables the recipients to become more informed voters, more adaptable to change, more sociable, more equal, more cultured, etc. These added benefits provide the justification for government subsidies to education. At least conceptually, the value of the added benefits plus the value of the private benefits define the social return to education. But there is no general agreement on what are the added benefits of education and how

4


to measure them. Nonetheless, economists have tried with some success, especially in the area of evaluating the effect of education on productivity that accrues to society in addition to the return enjoyed by the individual recipient of the education. When the private return to education is expressed as a percentage of the expenses incurred by the individual (and his or her parents), the resulting percentage is called the private rate of return. Similarly, when the social return to education is expressed as a percentage of the expenses incurred by all of society, it is called a social rate of return. Estimates of the rates of return to various education levels across both developed and developing countries have been summarized and reported by Psacharopoulos (1989) and Jain (1991) among others. The results, with respect to higher education, show a fairly consistent pattern across time, countries, and stage of development. The private and social returns to higher education are generally lower than those to other levels of education, but still above many other rates of return obtainable on alternative forms of physical capital investment. The private returns to education are generally higher than social returns at all levels of education. This disparity is especially pronounced at the postsecondary level within poor countries where the subsidies provided directly to students (for example, grants and loans) and indirectly through government aid to the institutions (tuition subsidies and other legislative appropriations) are easy to measure, but added benefits are few or difficult to measure. During the late 1970s and early 1980s, for example, the social rate of return to higher education averaged about 9 percent in advanced countries, 13 percent in Asian countries, and as high as 16 percent in Latin America. At the same time, the private rates of return to higher education were about 12 percent in advanced countries, 18 percent in Asian countries, and as high as 32 percent in Africa (Psacharopoulos 1987, p. 586). The observation that in developing countries the social and private rates of return to higher education have historically been below both the social and private rates of return for precollege education has led to the belief that there is overinvestment in higher education because of oversubscription and oversubsidization. Within advanced and developed countries, the observation of Freeman (1976, 1980) and others that returns to higher education were falling also led to the popular idea that there had been an "overinvestment" in higher education. If private returns to higher education are artificially high, brought about by the overly large subsidies to higher education, then there will be excessive private demand. There also will be increased political pressure


5

for even greater subsidies and an ever increasing impetus to expand higher education in the public sector well beyond the point where it has a net social benefit or any real impact on growth. 3 According to the overinvestment hypothesis, to the extent that subsidized private demand for higher education takes resources away from other investments with higher rates of return, higher education actually may be detracting from the nation's growth. Especially in the case of developing countries, investment in higher education is alleged to slow general economic growth as more productive investments elsewhere are foregone. To further argue against expanding higher education, the observations by Freeman (1976, 1980) and others in the 1970s that both the social and private rates of return to higher education were falling were used as evidence that higher education needed to be curtailed. Even if social and private returns were brought in line, the fact that both were falling was used as a signal that private and societal funds were better invested where returns were rising. In the 1980s, however, higher education continued to expand in developed countries, such as the United States, as well as in developing countries, such as those in Asia. Did this expansion of higher education detract from or enhance economic growth? Are social rates of return for higher education above or below other alternative forms of investment, and to what extent is this comparison relevant? Does investment in higher education detract from other forms of investment? To what extent is the provision of high skill and technical training, which typically yields a low social rate of return because of the high costs of instruction, necessary for economic growth? Essentially, five countervailing responses have been put forward in the consideration of these questions and against the argument of overinvestment in higher education. First, recent estimates of current and projected returns for higher education in the United States are not low. Evidence by Berger (1988), Murphy and Welch (1989) and others indicates that the returns to higher education reversed their downward trends and began to rise rapidly in the late 1970s and early 1980s. Becker (1992) provides data on incomes received by those with college and high school educations. His calculations (as reprinted here only for males in table 1-1) suggest that returns to higher education may have hit a relative maximum in 1986, although the ratio of incomes of college to high-school educated is still in excess of its previous cyclical high reached in 1970. 4 Second, others (Cohn and Geske 1990; Denison 1962; and Miller 1967) have argued that since expenditures on education (by both governments and individuals) are made largely at the expense of consumption and not savings, such expenditures do not necessarily detract from other physical

6


Table 1-1. Median Income of Twenty-five-to Thirty-four-Year-old Males with College Education or four Years of High School, with Income in Years 1963 to 1990

Years

College (4+ yrs.)

High School (4yrs.)

Ratios

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

6947 7397 7474 8373 8762 9264 10228 10661 10908 11751 12349 12637 13232 13965 14802 15783 17345 18773 20589 21149 21988 23687 26174 27141 28181 28288 29663 29568

5612 5933 6151 6600 6882 7402 8008 8217 8556 9316 10153 10701 10767 11416 12104 13129 14280 15181 15393 15298 15789 17030 16981 17551 18366 19650 20167 20051

1.24 1.25 1.22 1.27 1.27 1.25 1.28 1.30 1.27 1.26 1.22 1.18 1.23 1.22 1.22 1.20 1.21 1.24 1.34 1.38 1.39 1.39 1.54 1.55 1.53 1.44 1.47 1.47

Source: See William E. Becker, Why Go To College? The Value of an Investment in Higher Education, in W. Becker and D. Lewis (Eds.), The Economics of American Higher Education, Kluwer Academic Publishers, Chapter 4, 1992, for data obtained from the U.S. Bureau of the Census. Current Population Reports, Series P-60, Money Income of Households, Families, and Persons in the United States, various issues.

capital formation and growth. As a consequence, additional expenditures on education can make a net contribution to growth even if the social rate of return might be lower for education than for physical capital.


7

Third, still others point out that the efficient system for college loans in the United States has facilitated market efficiencies that narrow the differences between alternative rates of return, thereby justifying investment in human capital. Recent work with field-specific rates of return (Berger 1988, 1992) indicates that enrollment patterns in specific fields of study within higher education reflect their changing market rates of return and that students are sensitive to these changing rates (Catsiapis 1987). Looking at aggregate rates of return overlooks these more meaningful microeffects. Many argue that where systems of public higher education are well developed, and financial markets permit borrowing against future earnings, rates of return on education will, in the long run, tend to equal the rates of return on other productive assets. But whether or not ex post observations on rates that cannot be measured precisely will ever show equality in a rapidly changing environment is open to debate. Fourth, because educational expenses are measured in monetary terms, they are sensitive to the economic conditions of the country in which the education takes place. Educational costs tend to rise quickly with the wealth of the country. They also tend to rise more rapidly than the general rate of inflation within the country. Because the delivery of education is human capital intensive, physical capital technological developments occurring elsewhere in the economy are not easily implemented in education. Technological advances drive down relative prices in the sectors where these advances can be implemented. Within these innovating sectors, wages can rise even though product prices and the cost per unit of output are falling. Because the cost of labor tends to move uniformly with average productivity across the economy, the cost per unit of output and the relative price of the output will rise in the sectors that cannot implement the new technologies. Thus, the apparent social rate of return to education may fall as its price (tuition and fees) rise even though the driving cost increases had little to do with the quality of education or its actual contribution to the overall growth process. Baumol, Blackman and Wolff (1989, chapter 9) refer to this phenomenon as the "cost disease" of education. Finally, as pointed out by Baumol et al., the cost disease phenomenon is insensitive to the contribution that education is making to the innovations taking place in the other sectors. Ironically, the education industry, and higher education in particular, may be the least able to implement the technologies that it is helping to create and disseminate. The research and development aspect of higher education, as well as the more universal instructional and service functions, contain factors that are not amenable to technological improvements. Yet, as the United States and other

8


advanced countries become more and more reliant on technological change and high-tech related activities, education has provided the means and is typically required for rapid implementation across other sectors of the economy. At a minimum, high schools and all of higher education provide the complementary skill and technical training required in the use of modern technology. Thus, the contribution of higher education in promoting growth is even stronger if the complementarities between higher education and other forms of investment are taken into account. Studies by Griliches (1969), Fallon and Layard (1975), and most recently Baumol, Blackman and Wolff (1989) have shown that primary level education is not enough to foster and maintain economic growth. Their work suggests that increased investment in new plants and equipment for the production of new products contributes to national growth only to the extent that human capital exists to identify and facilitate implementation of this capital. Unfortunately, the description of the actual manner in which education contributes to human capital formation, to make it complementary with physical capital, is more stylistic than scientific fact. Nonetheless, given that much of the recent technological change and new capital formation in the United States has come about in high-tech and related industries, it is not unreasonable to assert that higher education has played a major part in facilitating the implementation of this new capital within the U.S. economy. Although it is possible to have too much investment in certain forms of education, especially within higher education, as it is also possible to have too much investment in certain types of plant and equipment- this does not necessarily imply that all of higher education is overextended within the United States. In the development of higher education policy, especially national policy targeted on economic growth, care must be taken to examine the effects of differing types and forms of extended education and training. Growth Accounting for Higher Education

Growth accounting is based on the concept of an aggregate production function that links output to various forms of inputs. Patterned after the early work by Kendrick (1961) and Denison (1962, 1967), most studies of the contribution of education to aggregate economic growth start with a production function in which the total economy's output of goods and services, 0 (measured by gross domestic product), is produced on a fixed amount of land by the inputs of labor, L (measured by total person hours

9


worked), and physical capital, K (measured by inventories of machines and equipment). In functional notation,

0

= f(L,K)

Angus Maddison (1987, p. 653) states that "the oldest index of productivity growth is that for labor productivity, TI," which he defined as the difference between the compound rate of increase in output, 0, and the rate of increase in labor, L, TI(L)

=

o-

L,

where the dots over the variables indicate percentage changes. Similarly, productivity growth associated with increases in capital is defined by n(K)

=

o- L.

These measures of productivity do not show the same patterns over time. Indices of labor productivity tend to be positive throughout the business cycle, while indices of capital productivity tend to move with the business cycle, being negative in recessions. Allowing for these changes in both labor and capital productivity gives rise to the total factor productivity approach

TI(L,K)

=

o - cxL -

(1-cx)k,

where ex is the fraction of national income paid to labor, which under the assumptions of perfect competition and constant returns to scale, is constrained to equal one minus the factor share going to capital. The early work of Douglas (1948), for example, placed ex at 0.75. Because the total factor approach accounted for less than half of the U.S. growth, researchers introduced the "augmented" and "supplementary" factor approaches to take account of improvements in educational attainment and other factors such as changes in the age-sex composition of labor and changes in economic structures. Again, in the notation of Maddison, TI(L*, K*, S)

=

0-

cxL*- (1-cx)k*-

S,

where S stands for factors other than augmented labor, L *, and capital, K*, that are believed to influence growth. Dots over the variables again indicate percentage changes and the asterisks indicate multiple component terms. Maddison identifies nine structural changes that are reflected in S, while augmented capital K* includes adjustments for new technologies and depreciation. In the case of labor, L * stands for two components of labor: the number of workers times their average hours of work and the number of years of schooling completed by the labor force.

10


As an example of how growth accounting is used, consider the work of Smith and Welch (1986) that shows a one year increase in schooling to be associated with 6.5 percent increase in income. From this value King and Smith (1988, pp. 35-36) argue that 21 percent of the 1940-1980 growth in economy-wide U.S. income growth (of 2.4087 percent) resulted from an increase in average schooling levels (of 3.09 years for whites) during this period. Thus, they conclude that only 1.9029 percent of the income growth during this period can be attributed to factors other than education. Pencavel (1991), in his work on the role of education in economic growth, demonstrates how the contribution of the different levels of schooling (primary, secondary, and higher education) can be derived from aggregate measures of the schooling and productivity. Working with: Maddison's indices of average years of schooling of the populations in the United States, United Kingdom, Japan, France, and the Netherlands; Maddison's measure of the contribution to output growth of schooling in those six countries; and Maddison's estimate of ex equal to 0.70, Pencavel calculated higher education's contribution to output growth for three time periods between 1913 and 1984. His estimates, as provided in his chapter in this volume, indicate that although the contribution of overall schooling rose from 14.78 percent to 23.41 percent, the growth in higher education and its relative contribution to economic growth in the United States was much larger. Pencavel estimates that the contribution of higher education to economic growth has increased dramatically in this century, rising from only 1.29 percent in the 1913 to 1950 period to 14.61 percent in the 1973 to 1984 period. Although the technique of growth accounting provides insight into the contribution of education, the technique is not without its critics. Plant and Welch (1984), for example, argue that growth accounting methods may not be appropriate for assessing the contribution of intermediate inputs in production. Education is embodied in the primary input of labor (and physical capital); it uses resources that enhance the labor even though this resource use is not directly observable in the goods and services produced. Thus, an adjustment for alternative uses to which these resources could have been put should be made in accounting for the determinants of growth. Also lacking is a positive adjustment for the contribution that education makes through other intermediate inputs such as health and longevity (Hicks 1980), and the diffusion of ideas (Ruttan 1984)- none of which is reflected in aggregate output measures. Stephen Hoenack also calls into question the macrogrowth accounting method in his chapter (2) in this book that deals with "Higher Education and Economic Growth." He recommends approaches that are based on


11

individual and firm specific microdata. He feels that more prom1smg research avenues for linking higher education with growth can be found in an examination of those who actually create technological change, especially through the roles of scientifically trained personnel in the supply of innovations and in facilitating adjustments to the disequilibria created by innovations. Hoenack also argues that the scientific and technical personnel of universities could significantly increase their contributions to growth, if given appropriate incentives. Contributions of Knowledge Production and Dissemination

Although the growth accounting methods of Kendrick (1961), Denison (1967) and those that followed, including the many rates of return studies, shed light on the relationship between human capital and economic growth, the question persists in political debates as to whether, and how, the infrastructure of higher education actually contributes to the economy. One of the problems in assessing the contribution of higher education to economic growth is that rates of return and growth accounting methods of measurement both tend to ignore the important contributions that the infrastructure of higher education makes to economic growth beyond the directly enhanced value of human capital through education and training. Further confounding our understanding of the contribution of higher education to growth is the reality that higher education both influences and is influenced by economic development. The input of education in production is measured typically by the number of people who complete a given number of years of schooling or college. Unfortunately, this measure ignores the quality of such education and it does not take account of the many other contributions and services performed by higher education. In addition to embodying current knowledge in students, some institutions of higher education create new knowledge. Jaffe (1989), for example, reports on the relationship between research and development activities in universities and those in private industries. A new component or product that might come out of a university lab (such as the development of fluoride in Crest toothpaste at Indiana University, or the invention of Gatorade at the University of Florida) or a new forecasting method or software routine that might be part of the work of an academic economist (such as Bill Greene's academic work with the LIMDEP and ET programs at New York University leading to the founding of Econometric Software, Inc.) creates jobs,

12


increases wealth, and clearly contributes to national economic growth. The direct contribution of such products is easy to see. Although many advances in efficiency appear to come from sources outside the educational establishment (as in the creation of Apple Computer), most high-tech research and development (R&D) is typically associated in one way or another with higher education. In the case of the drug industry, for example, it was the creation of the National Center for Supercomputer Applications at the University of Illinois that enabled more than forty scientists from Eli Lilly & Company to obtain supercomputer training that just a few years earlier was not available because of lack of computer power and human know-how (as reported in The Wall Street Journal, August 14, 1990). In robotics, it was the joint pioneering work at Bell Laboratories and leading U.S. research universities, including the Massachusetts Institute of Technology and the University of California at Berkeley, that gave rise to micromachines that are the size of a human hair (as discussed in Business Week, August 27, 1990). Higher education also adds directly to allocation efficiency (Schultz 1975). Capital-skill complementarity is well known in the economics literature (Griliches 1969). Recent work presented by Bartel and Lichtenberg (1988) and the earlier literature reviewed by Jamison and Lau (1982) suggest that more educated individuals are better able to absorb new information, take initiative, and adopt new technologies. 5 Thus, the effects of expenditure on higher education can be expected to be reflected in more than just the numbers of students admitted, enrolled, or graduating and the private returns they may receive. Regardless of the source, as the economy grows and incomes rise, the demand for enrollment places and other educational services can be expected to increase in differing areas. In addition, as new methods and technology are accepted by industry (as with the computer and information revolution of the past ten years), students quickly get the message that related educational skills are desirable; schools respond by offering that which students demand. These signals are transmitted in a market economy by earnings and income differentials attached to specific jobs and skill bundles. These income premiums are typically associated with individuals with more years of schooling and those in the fastest growing industries where the new methods and technology are being implemented. The question of whether expenditures on public education contribute more to the nation's productivity and growth than other public expenditures is examined in this book by David Aschauer in his chapter, "Is Public Education Productive?" (see chapter 4). His work illustrates the tradition


13

of using macromodels to address this type of question. He does this through the development of an aggregate production function model for estimating the correlation between public spending on education and national productivity levels for a large sample of market -oriented economies over the period from 1960 to 1985. The results are consistent with the notion that variations in per capita output can be largely explained by variations in physical capital and educational capital stocks relative to output. His results indicate further that the implicit share of educational capital in output is in the range of 15 percent to 20 percent, which is consistent with the results from growth accounting discussed earlier. In addition to its effect on the infrastructure of the economy, the contribution of higher education to basic and applied research is reflected in patents, journal articles, and new products. But, there is no precise way to relate these measures to the actual advances in productivity or reductions in costs. In "The Contribution of Higher Education to R&D and Productivity Growth" Walter McMahon (see chapter 5) addresses these measurement issues by defining an overall conceptual framework estimating the contributions, not only of the human capital formed through investment in higher education, but also of investment in academic research to labor productivity growth. Empirical estimates are presented from his model for the United States and ten other OECD countries that suggest substantial contributions. The work of Bartel and Lichtenberg (1987) and Dickens and Katz (1987) suggests that the wages of the highly educated are higher in industries that engage in research and development activities and show rapid technological advancements. The demand for college educated labor appears to move with economic growth and technological advances. Although Pencavel (1991) states that the manner in which highly educated labor relates to advances in physical capital (machines and equipment) is questionable, Hamermesh (1986) believes that the evidence solidly supports the idea that skilled labor and physical capital are complements, while unskilled labor and physical capital are substitutes in production. That is, an increase in the demand for physical capital should be associated with an increase in the demand for college educated workers relative to those with less education. Wayne Howe reports in his chapter on "The Effects of Higher Education on Unemployment Rates," (see chapter 6) that education has become an increasingly important criterion for job market success over the past two decades. Howe concentrates on the relationship between the changing demographic composition and education level of the labor force and the

14


impact of these factors on the structural rise in unemployment in the United States. The primary focus is on how the labor market adjusted to the increased supply of relatively young, inexperienced, and well-educated workers. It is shown that the labor market's response to the rising education level of an expanding labor force was a relative weakening position of employment for high school graduates, compared with those from higher education. 6 The importance of knowledge production and dissemination through education is further suggested by Baumol, Blackman and Wolff (1989) in their study of the rise and fall of nations based on productivity. They found that countries that exploit the technological advances of other countries form a "convergence club" in which those who lag in productivity at any given time tend to catch the leaders as time passes. 7 Countries with like educational levels were shown to converge among themselves, in terms of real gross domestic product, though not catching up with countries whose educational levels were higher (pp. 205-206). This convergence effect was argued to be most pronounced when secondary education was viewed as the explanatory variable. Their finding has implications contrary to that emanating from the rate of return estimates, where the primary levels had the highest social rates of return. Apparently, an emphasis on primary schooling is not sufficient if a nation wants to enhance its use of modern technology in advancing its economic growth. 8 Many of the issues and the work of other economists working on education's contribution to aggregate growth convergence and technology transfer are addressed by Pencavel in this volume (see chapter 3). As state governments across the United States have become increasingly focused on economic development strategies, one of the questions often raised is what effect state expenditures on higher education might have on business and job creation. Bryan Jones and Arnold Vedlitz directly address this question in their chapter dealing with "Higher Education, Business Creation, and Economic Growth in the American States" (see chapter 8). Presenting data from states in a cross-sectional disequilibrium adjustment model for the period 1979-1984, they report that higher education policies (that is, relative to expenditures) in the states do, indeed, influence the net creation of new business, although they acknowledge the ever-present problem of establishing one way causality. They also present evidence indicating that business creation is a critical stimulant in the economic growth process, affecting job growth while job growth fails to stimulate business growth. They conclude that state spending on higher education, not the level of state spending nor the level of federal research dollars flowing to states, may be the most important factor affecting state economic


15

development. The work of Jones and Vedlitz also raises the question of whether states that focus resources on their flagship universities fare better than those employing more of a scattershot approach to higher education. The Effect of Higher Education Quality on Growth

As is evident in the topics addressed in this volume, much of the economics research in higher education to date has emphasized the quantity of human capital embodied in the population and its relationship to productivity and economic growth. As a consequence of this perceived relationship, and our continuing social concerns for equity, much of public policy during the past forty years has been aimed at facilitating expanded access for ever-increasing members of society. The often overlooked issue in these public policy discussions has been the tradeoff between quantity and quality. Is it appropriate, for example, to expand access to advanced levels of education regardless of the quality of the experiences provided? Although this issue has received some attention in the plans of developing countries (Psacharopoulos and Woodhall 1986), surprisingly little concern has been given to the qualitative aspect of higher education in the United States, where it is estimated that higher education participation rates are almost twice those in most other industrialized countries (World Bank 1982). In the case of education, generally, Hanushek (1986, 1989) argues that expenditures on things such as lower class size, teachers' salaries, and other measures of input quality have little effect on output, when measured by test scores. Card and Krueger (1990, 1991), however, show that when the output measure is postschool earnings, expenditures on quality matter greatly. The influence of input expenditures on the relative earnings of blacks versus whites is reviewed by Donohue and Heckman (1991). Lewis Solmon and Cheryl Fagnano address the issue of the "Quality of Higher Education and Economic Growth in the United States" in their chapter (7) for this volume. After examining the evidence on several of the issues relating to the measurement of outcomes and quality in higher education, Solmon and Fagnano conclude that the quality of higher education and economic growth are indeed highly correlated. Their work supports the notion that the distinctive quality of colleges and universities plays a significant role in our nation's economic growth. The chapters presented in this volume suggest important ways through which higher education interacts with other sectors of our economy in the promotion of national economic growth. They also point to import-

16


ant directions in which public policy can be employed to maximize the social returns that accrue to a nation that has a well developed higher education sector.

Notes 1.

In 1759 Adam Smith published The Theory of Moral Sentiments, in which the wrote:

The education of boys at distant great schools, of young men at distant colleges, of young ladies in distant nunneries and boarding-schools, seems in the higher ranks of life to have hurt most essentially the domestic morals, and consequently the domestic happiness, both of France and England ... Surely no acquirement which can possibly be derived from what is called a public education can make any sort of compensation for what is almost certainly and necessarily lost by it (pp. 363-364). 2. Astin, Green, Korn, and Schalit (1985, p. 47) report that 72 percent of the students in a national survey of college freshman agreed that "The chief benefit of a college education is that it increases one's earning power;" 83 percent listed "get a better job" as a very important reason for attending college. Similarly, Willis and Rosen (1979) found the decision to go to college is highly sensitive to monetary considerations, with a ten percent increase in starting salaries inducing almost a twenty percent increase in college enrollments. 3. Pressure to expand higher education that has little social benefit will be extremely pronounced if a collegedegree is used as a screening device for the ranking and identification of job applicants. It will also be great in selected areas of study where the private returns are great but the social returns are low because of the high cost of training. 4. In assessing the human capital enhancing effect of education, labor economists emphasize the ratio of earnings of those with a college versus a high-school education. Income figures include earnings as well as other forms of revenue. To the extent that a college education also affects one's ability to manage the sources of all investments, the college/high school ratio of income might be viewed as a better measure of the overall benefit of college. Of course, if those who pursue college educations have greater endowments initially, then the college/high-school ratio of income will be a biased measure of the overall benefit. 5. An interesting question seldom asked is whether a society can overinvest in education in the sense that the resulting technology transfer is too fast for the realization of expected profits at the time of the investment. That is, if an innovation can be copied by a highly educated pirate before the inventor realizes his or her profit then has society overinvested in education? 6. Simple college-high-school earnings differentials and rates of return to schooling that have no adjustment for labor force participation, are biased. Educational attainment increases the likelihood of being employed at any given time and it also increases the likelihood of working longer hours. For more evidence of this relationship see Mincer (1991). 7. Baumol (1991) reviews his work, Barro's (1991) work and that of others on the implications of different measures of convergence. Although countries that are part of the convergence club will have growth rates that tend to move toward the same long-term level, Baumol identifies many "ancillary variables" that may influence growth. Baumol attaches little importance to headline grabbing events such as "brain drains," "technology thefts,"


17

"job exporting," and the like. Articles such as "Buying the American Mind: Japan's Quest for U.S. Ideas in Science, Economic Policy and the Schools," Chronicle of Higher Education (December 11, 1991)-which reports on a study claiming that universities in the United States arc subsidizing economic progress in Japan through their research that is partially supported by Japanese interests-typically miss the shared growth that is enjoyed by all members of the convergence club. 8. Baumol, et al. 1989, cite W. Arthur Lewis for the observation that unless a country hopes to become involved in extensive research and development and innovation, only a relatively small group of persons with a university education is needed to permit the import and adoption of more sophisticated production techniques. It is the group of trained and skilled technicians that is absolutely required in large numbers, and for them secondary education is usually indispensable (p. 202). Nevertheless these highly skilled workers typically have some postsecondary training that may not be captured in the Baumol, ct al. empirical work. In his chapter, Pencavel argues that the results of Baumol, et al., actually indicate that an increase in enrollment rates in postsecondary education is associated with higher economic growth than the same increase in secondary school enrollment rates.

References Astin, A.W., Green, K.C., Kom, W.S., and Schalit, M. (1985). The American freshman: National norms for falll985. Los Angeles: Higher Education Research Institute, University of California at Los Angeles. Barro, R.J. (1991). Economic growth in a cross section of countries. Quarterly Journal of Economics, 106(2), May, 407-443. Bartel, A.P., and Lichtenberg, F.R. (1987). The comparative advantage of educated workers in implementing new technology. Review of Economics and Statistics, 69(1), 1-11. Bartel, A.P., and Lichtenberg, F.R. (1988). Technical change, learning, and wages. Working Paper No. 2732. Washington, D.C.: National Bureau of Economic Research. Baumol, W.J. (1991). Multivariate growth processes: Contagion as possible source of convergence. Mimeograph dated 10-7-91. Baumol, W.J., Blackman, S.A.B., and Wolff, E.N. (1989). Productivity and American leadership: The long view. Cambridge: The MIT Press. Becker, G.S. (1960). Underinvestment in college education. American Economic Review, 50, May, 345-354. Becker, G.S. (1964). Human capital: A theoretical and empirical analysis with special reference to education. New York: Columbia University Press. Becker, W.E. (1992). Why go to college? The value of an investment in higher education. In W.E. Becker and D.L. Lewis {Eds.) The economics of American higher education. Norwell: Kluwer Academic Publishing {Chapter 4). Becker, W.E., and Lewis, D.L. {Eds.) {1992). The economics of American higher education. Norwell: Kluwer Academic Publishing.

18


Berger, M.C. (1988). Predicted future earnings and choice of college major. Industrial and Labor Relations Review, 41, 418-429. Berger, M.C. (1992). Private returns to specific college majors. In W.E. Becker and D.L. Lewis (Eds.) The economics of American higher education. Norwell: Kluwer Academic Publishing (Chapter 6). Card, D. and Krueger, A. (1990). Does school quality matter? Returns to education and the characteristics of public schools in the United States. Working Paper No. 3358. National Bureau of Economic Research. Card, D. and Krueger, A. (1991). School quality and black/white relative earnings: A direct assessment. Working Paper No. 3713. National Bureau of Economic Research. Catsiapis, G. (1987). A model of educational investment decisions. Review of Economics and Statistics, 20, 33-41. Cohn, E., and Geske, T.G. (1990). The economics of education. New York: Pergamon Press. Denison, E.F. (1962). The sources of economic growth in the U.S. and the alternatives before us. Supplementary Paper No. 13. New York: Committee for Economic Development. Denison, E.F. (1967). Why growth rates differ: Postwar experiences in nine Western countries. Washington, D.C.: The Brookings Institution. Dickens, W.T., and Katz, L.F. (1987). Interindustry wage differences and industry characteristics. In K. Lang & J. Leonard (Eds.), Unemployment and the structure of labor markets. New York: Basil Blackwell. Donohue, J.J. and Heckman, J. (1991). Continuous versus episodic change: The impact of civil rights policy on the economics status of blacks. Journal of Economic Literature, 29(4), December: 1603-1643. Douglas, P.H. (1948). Are there laws of production? American Economic Review, 38(1), 1-41. Fallon, R.P., and Layard, R. (1975). Capital-skill complementarity, income distribution and growth accounting. Journal of Political Economy, 83, April, 279-301. Freeman, R. (1976). The overeducated American. New York: Academic Press. Freeman, R. (1980). The facts about the declining economic value of college. Journal of Human Resources, 14, 289-318. Griliches, Z. (1969). Capital-skill complementarity. Review of Economics and Statistics, 51, November, 465-468. Hamermesh, D.S. (1986). The demand for labor in the long run. In 0. Aschenfelter and R. Layard (Eds.), Handbook for labor economics (Volume 1). New York: North Holland Press. Hanushek, E.A. (1986). The economics of schooling. Journal of Economic Literature, 24, September, 1141-1177. Hanushek, E.A. (1989). The impact of differential expenditures on school performance. Educational Researcher, 18, May, 45-51, 62. Hicks, N.L. (1980). Is there a trade-off between growth and basic needs? Finance and Development, 17(2), 17-20.


19

Jaffe, A.B. (1989). The effects of academic research. American Economic Review, 79(5), 957-970. Jain, B. (1991). Returns to education: Further analysis of cross country data. Economics of Education Review, 10(3), 253-258. Jamison, D.T., and Lau, L.J. (1982). Farmer education and farm efficiency. Washington, D.C.: World Bank. Kendrick, J. W. (1961 ). Productivity trends in the United States. Princeton: Princeton University Press. King, E.M., and Smith, J.P. (1988). Computing economic loss in cases of wrongful death. (R-3549-ICJ). Santa Monica: The Rand Corporation. Maddison, A. (1987). Growth and slowdown in advanced capitalist economies: Techniques of quantitative assessment. Journal of Economic Literature, 25(2), 649-698. Mincer, J. (1991). Education and unemployment. Working Paper No. 3838, (September). National Bureau of Economic Research. Miller, W.L. (1967). Education as a source of economic growth. Journal of Economic Issues, 1, December, 280-296. Murphy, K., and Welch, F. (1989). Wage premiums for college graduates: Recent growth and possible explanations. Educational Researcher, 18(4), 17-26; and reprinted with modification as: Wages of College Graduates. In W. E. Becker and D.L. Lewis (Eds.) The economics of American higher education. Norwell: Kluwer Academic Publishing, 1992, (Chapter 5). Pencavel, J. (1991). Higher education, productivity, and earnings: A review. Journal of Economic Education, 22(4), 331-359. Plant, M. and Welch, F. ( 1984). Measuring the impact of education on productivity. In E. Dean (Ed.), Education and economic productivity. Cambridge: Ballinger Publishing Company (pp. 163-193). Psacharopoulos, G. {1987). Returns of education: A further international update and implications. Journal of Human Resources, 20(4), 581-604. Psacharopoulos, G. (1989). Time trends of the returns to education: Crossnational evidence. Economics of Education Review, 8(3), 225-231. Psacharopoulos, G., and Woodhall, M. (1986). Education for development: An analysis of investment choices. New York: Oxford University Press. Ruttan, V.W. (1984). Social science knowledge and institutional change. American Journal of Agricultural Economics, 66, December, 549-559. Schultz, T.W. {1961). Education and economic growth. In N.B. Henry (Ed.), Social forces influencing American education. Chicago: University of Chicago Press. Schultz, T.W. (1975). The value of the ability to deal with disequilibria. Journal of Economic Literature, 13, 827-846. Smith, A. (1759) (1976). The theory of moral sentiments. Reprint. Indianapolis: Liberty Classics. Smith, A. (1776) (1976). An inquiry into the nature and causes of the wealth of nations. Reprint. Oxford: Clarendon Press. Smith, J.P. and Welch, F.R. (1986). Closing the gap: Forty years of economic progress for Blacks. (R-3330-DOL). Santa Monica: The Rand Corporation.

20


Willis, R.J. and Rosen, S. (1979). Education and self-selection. Journal of Political Economy. (Supplement) 87(5), Part 2, 65-97. World Bank (1982). World development report, 1982. Washington, D.C.: World Bank.

2

HIGHER EDUCATION AND ECONOMIC GROWTH Stephen A. Hoenack

Although economists have long been interested in relationships between education and economic growth, there is still very little empirical evidence about the specific parts of the education system or particular educational policies that contribute cost-effective enhancements to growth. Instead of drawing microlevel inferences that could help guide policymaking, economists have confined their efforts to providing aggregative estimates of the effects of education on growth. These attempts have consisted of assigning to education a residual portion of economic growth that is not accounted for by increases in quantities of labor or nonhuman capital. The microeconomic forces that underlie economic progress determine the most relevant policy connections between education and growth. This chapter begins with a discussion of these microeconomic forces because existing studies of education's contributions to growth make few explicit assumptions about them. These forces are then taken as the context within which studies providing aggregative estimates of the effects of education on growth are interpreted. Subsequently, the chapter turns to research questions regarding higher education's specific contributions to economic progress. Finally, it explores ways in which these contributions can be efficiently increased by institutional changes and new incentive arrangements. Emphasized in the discussion of these questions are the potential contributions of scientific and technical academic personnel and their students to innovative activities within firms. 21

22


The Microeconomic Forces Underlying Economic Growth and Possible Roles of Higher Education

Over the last two centuries, innovations leading to new products and new methods of production and distribution have been more responsible than any other phenomenon for the immense rises in living standards in the developed countries. Increased living standards can be driven primarily by other forces, such as the opening up of a frontier or market, or rebuilding after a war over relatively short periods, during which there are only modest innovations. However, over the long run, innovation is the mainstay of rising living standards. Innovations usually create large and widely distributed externalities; although the innovator's gains are temporary, the resulting longer term social benefits of successful innovations are diffused or bettered by competitors. Consequently, investments augmenting an economy's stock of capital without enhancing innovation do not necessarily contribute as much to sustained rises in living standards as do investments introducing or supporting innovation, even if they achieve the same expected rate of return to investors. Many, although far from all, innovations are based in varying degrees on scientific and technological advances. The activities within firms that support such innovations in turn can depend on higher education's contribution of scientific and technological knowledge and the skills needed to apply it. Innovation also encompasses important economic advances that are, technically, relatively simple. For example, straightforward concepts of marketing and cost-control in retail distribution during the past twenty-five years have revolutionized segments of consumer markets. The higher education system can help augment society's returns from these innovations if more educated persons are more likely to imitate or improve upon them. In addition to creating economic opportunities, innovations can cause the obsolescence of skills, equipment, and capital facilities, a process Schum peter (1950) called "creative destruction." Schultz (1975) hypothesized that education reduces the costs of adjustment to the disequilibria created by innovations. Regardless of higher education's direct contributions to growth, the presence of educated manpower could facilitate adjustments to the disequilibria resulting from innovations by speeding resource owners' responses to the new opportunities. Not much attention has been given to this hypothesis, and little thought has been given to the particular forms of education and types of skills that most facilitate adjustments to disequilibria.


23

In relatively static industries, employees' ability to adapt to change may be relatively unimportant. However, in industries characterized by frequent introduction of new products or frequent adoption of new production techniques, it may be crucial to employers that employees be able to support efficient adaptations to change. Problem-solving skills and an understanding of how one's work relates to that of others may enhance an employee's capability to adjust to changes. For a firm to survive, it may be necessary that employees on their own initiative take actions that support the firm's innovations or its adjustments to innovations introduced by others. This can require employees to decide when to acquire new skills or technical knowledge as required for this purpose, whether or not on the job. When labor is less mobile than capital, firms in innovative industries may base their plant location decisions, in part, on the availability of disciplined workers possessing problem-solving and coordination capabilities and initiative. To the extent that specific types of education can raise the local supply of such workers, the availability of workers possessing these qualities can form a significant part of countries' and regions' economic development policies. The importance of the presence of particular skills and work habits in attracting nonhuman capital could be tested using a sample in which firms' plant location decisions are related to aptitude and achievement tests of available workers and other variables controlling for costs and productivity in alternative locations. The test results would be related to previous educational investments. Elsewhere (Hoenack 1989), I have suggested that there will be increasing pressures on governments to participate in the expanding geographic competition for mobile resources. During the last few decades, there has been a large increase in the number of countries and of areas within them possessing the infrastructure necessary to support modern economic activity. There has also been a rise in many economies' shares-of-valueadded, contributed by economic activities that are not tied to particular geographic attributes such as harbors, rivers, and proximity to minerals. These factors have greatly raised the number of sites capable of supporting a variety of manufacturing, distribution, and business service functions. Citizens possessing location specific resources will expect their local governments to assist in this competition by developing policies that attract mobile resources, including the cost-effective provision of public services that are complementary to growing private market activities. To the extent that the most desired categories of workers attach importance to their own and their childrens' educational opportunities, localities will

24


employ education as a policy tool for attracting and retaining them and the types of businesses that employ them. Rates of Return to Education, Unexplained Residuals, and Economic Growth

The classic studies of Schultz (1961) and Denison (1967) attempted to establish a connection between education and economic growth by interpreting the "residual" calculated from estimated aggregative production functions for the United States. The residual is the amount of growth in national income that is not explained by increments of physical capital and numerical labor inputs in these production functions. Denison divided the labor force into educational categories and used changes in this distribution across the categories to explain education's role in growth. He accomplished this task by attributing the incremental wages received by workers to their respective educational categories. Schultz analyzed levels of total expenditures on education as stocks of educational investments. The role of education in growth was then determined via estimated rates of return on these stocks. The aggregative methodologies employed in these studies make it difficult to use their results to inform policymaking. For example, the results cannot be used to guide choices about alternative educational options such as expanding basic literacy versus increasing the supply of scientific and technological personnel trained by universities. In spite of this limitation, some methodological discussion of these studies is a useful way of drawing attention to the microeconomic relationships between education and growth. Consider the structure of wages that accompanies the major educational categories of the labor force. Such a breakdown of wages according to education levels underlies the Denison study, while in an analysis using Schultz's methodology one would take the extra step of using the wage data for educational categories to calculate rates of return to education. Using this wage structure, let us examine the possibility of a specific reallocation via the education system of a certain number of workers from a lower educational category to one representing more education. Would there be an increase in the economy's rate of growth? Is this desirable? Can available estimated rates of return be used to answer the desirability question? Workers in categories representing larger amounts of education typically receive (at the same experience levels) higher incomes than those in


25

categories requiring lower educational attainment. As a result, the economy's per capita income could rise as a result of the reallocation even if the risk -adjusted rate of return 1 at the margin in the augmented educational category is smaller than the cost of capital. If, for example, the costs of increasing the number of workers in the higher educational category were financed out of present consumption, as with a tax that accomplishes this effect, there would be growth (as defined by a rise in future per capita income). Hence, educational investments that raise per capita incomes are not necessarily efficient. Available estimates of rates of return cannot shed much light on the question of growth and efficiency for two reasons. First, they usually pertain to average, not marginal, returns and, presumably, considerably exceed marginal returns. Further education will not serve both efficiency and growth when all opportunities for educational investments having marginal risk-adjusted social returns in excess of the cost of capital are exhausted. It is interesting to note that estimated private average rates of return to higher education typically do not exceed, and are not infrequently less than, the range of returns reflecting the cost of capital. (See, for example, table 3-7, pp. 56-57 in Psacharopoulos and Woodhall 1985.) The second, and more important limitation of available estimates of returns to education, is that they ignore spillovers. Spillovers, including the effects of higher education on the economic returns to innovation, are the place to look for possible evidence regarding higher education's contributions to efficient growth at the margin. Spillovers can maintain the marginal economic value of educated personnel as the number of such personnel increases; without innovation and the accompanying shifts in economic relationships, efficient investment opportunities for workers with university educations would be quickly exhausted. However, in some industries, the same categories of educated workers may be more prone to the creation of growth enhancing externalities than in other industries. Perhaps higher education, on average, produces more such externalities than lower levels of education or vocational training. There is, of course, the counter example of specific occupations requiring higher education that primarily serve to qualify individuals for work in governmental agencies and parastatals in many countries; these may contribute fewer growth enhancing spillovers than less education intensive occupations that are concentrated in highly innovative industries. There may not be a close relationship between the educational levels of occupational categories and the rate of economic growth unless care is taken to control for the industry in which the occupation is concentrated and the degree of innovative behavior occurring within it.

26


Models that interpret economic growth as being caused by augmented stocks of capital assets, including human capital and the privately captured returns on them, usually leave out the roles of shifts of economic relationships and the externalities that accompany them. Generally, only a small portion of the social returns to these externalities are captured in the earnings of those creating them, including those of the university graduates who contribute to innovations. Indeed, as noted already, growth can make capital obsolete through the process characterized by Joseph Schumpeter (1950) as "creative destruction." Considering growth as an innovative process also brings to attention the fact that the income data used in aggregative production function studies can fail to capture much of the economic growth that is attributable to improvements in product quality and in the quality of inputs. 2 Another conceptual problem arises with the use of the structure of different educational or occupational categories to make inferences about the causes of economic growth. The private and social demands for education are a function of incomes, which are increased by growth, independently of education's returns. It is necessary to control for the effects of growth on incomes in occupations, and the consequent changes in the demand for education, in order to infer the desired reciprocal effects of education on aggregative income and its changes over time. 3 The time-lags over which education affects income suggest one methodology for inferring these effects apart from the impacts of income on educational attainment. To the extent that current income influences future educational attainment, while the impacts of the latter variable are on future income, equations in which income is related to lagged education could provide unique inferences about the effects of education on income. The models employing lags to infer causal directions developed by Granger (1969) and which are applied, for example, by Sims (1974) in the analysis of causality between an economy's money supply and its aggregate income level, may thus apply to education. In this type of analysis, three modeling issues should be considered. First, education is correlated with other variables that can also have lagged effects on growth. Funding for diverse governmentally supplied services often runs in tandem so that expenditures on items such as health care, transportation, and education are correlated over time. When such correlations exist, those publicly provided services other than education, that are complementary with growth enhancing private market activities, should be included in models in order to avoid overestimating the effect of education. The second modeling issue pertains to the composition of educational


27

attainment in the context of a particular country's developmental situation. Ideally, models would include variables controlling for the relative demand for and supply of different types of educated manpower, for the occupations and in the industries where growth enhancing externalities are most important. It may also be the case over time or across different areas taken in a cross section sample, that educational systems vary in their contributions to workers' acquisitions of problem solving and coordination capabilities and their willingness to acquire new skills as needed; the economic values of these outcomes to employers may vary as well. The third modeling issue relates to the outcomes that would have resulted from the alternative expenditures displaced by educational spending; these are affected by public sector priorities. The net growth impacts of increased spending on education depend in part on the effects on economic progress that would have occurred if spending had been directed in alternative ways, including toward other publicly provided services. These should be explicitly modeled. Research Topics for Making Useful Policy Inferences About Relationships Between Education and Economic Growth

Policy relevant inferences about relationships between higher education and economic growth require a focus on the economic effects of specific university instructional and research activities. This section addresses research questions concerning the impacts of these activities on firms' capacities to adapt successfully to an innovative competitive environment. The last section addresses organizational questions concerning the efficient utilization of academic personnel in the scientific and technological work supporting firm related innovative activities. The impacts of a university's activities and outputs on firms' innovative functions can be considered in categories according to the contributions made by faculty and students and the influence of universities and their research functions on firm location decisions. The Contributions of University Personnel

The impacts of academic personnel and of students on innovative activities in firms can be analyzed in case studies of particular firms or industries, or on the basis of samples of a large number of cases designed to permit

28


broader inferences. Under the latter methodology, in order for the costs of the research to be kept to reasonable levels, the study design should not require analysis of each case as intensive as that which often characterizes individual case studies. Ideally, academic activities would not only be linked to firms' innovative activities, but the valuations of their contributions would be measured. However, it would be difficult to attach values to the indirect contributions of academic functions via their direct effects on the productivity of R&D and other employees in firms. Initial studies should focus the task of identifying what the connections are between specific types of academic scientific and technological functions that most tend to contribute to firms' innovative activities. It will be especially important to determine which particular industries or types of processes or products are most frequently affected by academic research. In many cases, institutional or incentive-related problems may be discovered that, if ameliorated, would raise the magnitude and value of these linkages. By studying a sizeable sample of innovations that are broadly representative of particular sectors of the economy, researchers could identify the types of knowledge on which these innovations are based. As a key element of this study, citations in scientific reports and publications of personnel in the firms producing innovations would be analyzed to identify the effect of previous university based scholarship on this work and its disciplinary sources. These citations and citations within the cited studies (including those prepared within universities), could be analyzed to assess the interrelated roles of different types of knowledge underyling the sampled innovations. Preliminary investigation suggests that publication with citations by R&D personnel in firms is frequent. In some industries, patents provide citations, and it is not uncommon for scientific personnel in firms to publish research, perhaps because the rapidity of obsolescence makes information no longer useful to competitors within publication lags. An example of a relevant analysis using citations is provided in Lieberman (1978). Identification of the types of academically trained personnel who contribute to firms' innovative activities can be accomplished in "forward" and "reverse" tracer studies. The former studies start with the point of graduation and could identify the career patterns of scientists and engineers to determine how many end up supporting innovative functions in firms and in what ways. Instead of focusing on recent graduates and waiting to follow them up, cohorts of previous graduates could be identified and found, making it possible to obtain follow-up information without delay. Reverse tracer studies start with cohorts of workers in particular cat-


29

egories of jobs, and study the paths by which different employees reach this point. Such studies would permit inferences about the diversity of backgrounds of the personnel engaged in innovative activities. In these studies, hypotheses could be tested regarding the cost-effectiveness of alternative means such as formal versus on-the-job training to supply highly productive but usually costly skills, and when promising possibilities are not presently available, university scientific personnel could be queried about the potential for providing these skills in specially developed courses or in the context of research projects. Perhaps it would turn out that some skills and experiences could be most efficiently supplied by university personnel directly involved in innovation-supporting research activities and engaged in supervising students specializing in this type of work. For categories of workers other than scientific and technical personnel, the backgrounds of workers in diverse industries could be investigated via reverse tracer studies to infer the types of learning (including basic reading and mathematics skills) and work experiences possessed by different workers. By testing workers or obtaining data on their productivities or career progressions, it would be possible to test hypotheses about how different learning and work experiences contribute to the development of problem solving and coordination skills, which, as suggested earlier, may be important for firms' successful adjustments to the shifts in market demand and supply relationships that often accompany growth. It would be valuable to follow individuals forward and backward in time who have worked in industries undergoing major transitions and to develop measures of personal success and contributions to their firms. Key issues include the impacts on successful adaptation and contributions of basic knowledge in the sciences and in other fields versus more specific technical skills demanded by employers. The Influence of Universities on Firm Location Decisions

The capacity to attract the most productive mobile resources, including managerial and technical personnel and high-value-added service and production facilities, is becoming increasingly important to the strength of the economies of most regions and metropolitan areas throughout the world. It has already been noted that a locality's educational offerings can be a significant element of this capacity. Additionally, local universities may achieve significant economies for firms' R&D enterprises and provide economies for smaller research based firms. It is possible, although difficult, to test these hypotheses. If one could

30


not deal successfully with the modeling problems mentioned below to estimate the impact of the presence of universities on firms' location decisions for R&D facilities, one could observe the sharing of personnel and facilities among the two types of organizations and make a rough estimate of the potential gains. Two modeling issues are important for inferring the degree to which universities serve as an instrument for attracting mobile resources to their localities. The first modeling issue results from the intercorrelation in the availability of different public services across estimation samples, mentioned above in connection with models that use time lags to identify effects of education on income. In this case, multiple publicly provided services influence firms' location choices. It is necessary to specify models to include the complete set of these services in order to obtain unbiased estimates of particular services. For example, higher education institutions are more frequently in areas where support services available to businesses are of higher quality. Therefore, if the quality is not controlled for separately, its influence on firms' choices can be included in the estimated impact of higher education, tending to overstate it. The second modeling issue is identification. In order to make specific inferences about the location choices of owners of mobile resources, it is necessary to specify the reciprocal relationships in which regional economic progress affects the local supplies of public services that influence these choices. For example, states' public funding for higher education services is a function of their own economic progress, and this relationship must be specified exactly in order to determine whether unique inferences can be made about how the availability of these services affects the local supplies of mobile resources. 4 An interesting empirical analysis that illustrates these issues is provided in Jaffe (1989). In a pooled time series cross section sample, this study relates the number of patents awarded by state in each time period to the amount of R&D performed by industry and the quantity of university research. Separate equations are then used to explain the levels of these two independent variables. The author's overidentifying restrictions in the equation used to explain university research include variables that control for the presence of universities and federally funded research centers; and in the equation explaining industry R&D, the overidentifying variables include the values of population and of manufacturing value added. The author did not provide empirical tests of the correctness of these overidentifying restrictions. The justification of each of these equation's exclusions from the other equations can be debated on a priori grounds, and further study of linkages between universities and firms' R&D activities may lead to the specification of variables having stronger justifications.


31

Underlying research questions concerning the relationship between higher education and economic growth include those pertaining to the demand, production, and costs of innovation; the incentives that influence research personnel; and the types of scientific knowledge that affect innovation and the sequences in which it does so. To analyze these questions, two basic issues must be addressed. First, there is the need to model the effects of economic incentives versus scientific developments on innovation. As emphasized by Schmookler (1966), specific advances in particular industries are closely related to the presence of a problem, and, presumably, to the incentives that flow from unsolved problems. Other authors (a good summary is provided in Freeman 1982, pp. 109-112) emphasize the impacts that science has on innovation. The second analytical issue is the endogeneity of the scientific and technological activities that underlie innovation. One promising idea that has the potential of assisting with both modeling problems is isolating those subsets of all scientific applications that relatively frequently contribute to particular industries' innovations. These scientific applications have been referred to as technologi;:li paradigms by Dosi (1988) and technology guideposts by Sahal (1985). To date, such concepts remain ambiguous, and there is little empirical work that specifically measures the particular scientific applications that most often underlie innovations in individual industries. Additionally, very little is known about the degree to which economic incentives can or do influence directions of scientific activity. The fundamental questions are: 1. 2. 3.

What are the available options for researchers' choices among alternative scientific activities, including those which appear to have no direct relationship to economic activity? Is it possible to estimate the relative likelihood of economic payoffs from pursuing these alternatives? What types of incentives now influence the choices among alternative research directions, and can these be improved?

The labels "basic" and "applied" research prejudge the issue of whether scientific activities can be influenced with incentives, and it may be preferable to analyze this question empirically. The use of data on citations in research reports describing the technologies underlying innovations and in various types of technological and scientific publications has promise for addressing these questions usefully, and, generally, for supporting analysis of the interrelationships between scientific and technological activities and innovative activities in firms.

32


Citation data have the advantage over case studies of being applicable to a large variety of industries and scientific and technological developments, and they can permit inferences about specific connections between innovative and scientific or technological activities. Given a problem faced by an individual firm or group of firms, such as a production or transportation bottleneck or a major breakthrough by foreign competitors, different types of organizations may have varying degrees of comparative advantage in producing the components of research or technology underlying appropriate innovative responses. For example, proximity to the product development or marketing stages of production may create informational advantages within the firm that make it more efficient for them to provide specific components of their own research for innovations. In contrast, proximity to more broadly applicable research, the opportunity to share costly equipment and facilities with large sponsored research projects, and the availability of joint efforts of faculty and of student apprenticeship research support may give universities comparative advantages in other areas. For example, a small number of academic personnel might take on the specialty of dealing with certain research areas for one or more firms and encourage some of their students to work in these areas. These individuals would interact with other academic personnel on specific problems when their specialized expertise and interests were most useful. We now turn to the institutional and incentive problems that must be overcome in order to achieve more efficient contributions of academic personnel to firms' innovative activities. Institutional Arrangements, Incentives, and the Efficient Organization of Innovation Supporting Research Activities

Much attention has been given in recent years to the development of university-industry relationships, including industry-funded research projects at universities, research parks located near universities, the participation of faculty members in start up companies, and forums in which staff in firms and universities discuss with each other problems and research of possible mutual interest. A number of such arrangements are described in Powers et al. (1988); two essays in that book by Betz (pp. 281- 327) specifically describe the experiences of the National Science Foundation in supporting and encouraging the establishment of university based research centers in its Industry/University Cooperative Research Program.


33

However, in spite of these widely publicized activities, universities play a relatively small direct role in industrial research, as documented in table 2-1. Faculty consulting and the development of spinoff companies are among the nonfunded activities not represented in these figures. Consulting often has only limited benefits to universities and frequently does not create the joint instructional and research outcomes brought about by many other faculty activities. While they can be very beneficial to the economy, spinoff companies frequently do not provide ongoing research opportunities and can even draw faculty members into management tasks that compete with their research programs. Facilities such as research parks can effectively reduce costs because of proximity to academic personnel and facilities. But beyond these cost savings, they do not necessarily affect the content of faculty research activities. In many industries, academic personnel could make a large contribution to firms' innovative activities. It has been argued (Kash 1989) that major governmentally funded research advances in the defense, health, and agricultural sectors have contributed the most important innovations in our society. The scientific activities occurring within universities that have contributed to these advances frequently go beyond the basic research that is often viewed as the primary domain of science in these institutions. Academic personnel have often provided major applied contributions, although it is perhaps significant that this has been more typically the case when innovation is measured in terms of specific performance criteria as opposed to commercial success and intrafirm efficiency. The question considered here is, why is there not more industrysponsored research within universities that is beneficial to both types of organizations? Considering many universities' depth of scientific capability and their research activities and educational functions (including preparing students interested in scientific and engineering careers in industry) that

Table 2-1. Universities

Percentages of Industry Research* Expenditures Occurring in

Basic Research Applied Research

1980

1987

10.2

13.1

1.2

1.5

* Excludes development expenditures Source: National Science Foundation (1980, pp. 14-15 and 1988, pp. 10-11).

34


are potentially complementary with firm-related research functions, it would be surprising if universities did not have sizeable comparative advantages in supplying many important research outcomes for innovative firms that are not now supplied. The present discussion analyzes some of the organizational and incentive barriers to university research in support of innovation. The organization and content of research presently performed in universities are a function of a system of incentives where academic priorities within the institution and the peer review system for awarding research grants play key roles. Even if there are parties outside of the institution willing to pay for research within it, there will not be an efficient supply response unless these demands translate into incentives for individual faculty to perform this research. The present discussion suggests that when firms have demands for research services from particular academic specialties, it can be the case that such incentives will be lacking even when universities would have a comparative advantage in responding, and such research would be highly beneficial to the institution and its students. The most important reason for this lack of incentives is the difficulty in many cases of specifying in contracts the most important research services that academic personnel could perform for firms. A notable part of many countries' higher education systems has succeeded in obtaining financing and providing incentives for academic personnel to make major direct research contributions to innovative activities. Because much of university-based agricultural research can be valuable to large groups of farmers and agribusinesses, who have acted as coordinated political supporters of research and extension funding at the state and federal levels, there have been significant collective demands for agricultural research in these institutions. The perception that consumers directly benefit from the lower prices that result from agricultural innovation has minimized opposition to such funding. Because large numbers of agricultural producers face common problems, there has not been a problem of idiosyncrasy of particular producers' problems and correspondingly high costs of contracting with these entities for the supply of solutions. Agricultural problems systematically have been amenable to scientific solutions over long periods of time, and ancilliary educational activities have proven productive by transferring knowledge about these solutions to producers. Consequently, long-term, predictable financial support for universitybased agricultural research has been grounded in the solution of common problems faced by producers. It has been possible to establish intrainstitutional incentives tied to success in solving these problems and transferring information about solutions to them. These incentives have


35

permitted the long-term viability of agricultural units within universities and career opportunities for the academic personnel working within them. Firms in much of the rest of the economy usually do not act nearly as successfully as coordinated interest groups seeking university research to support their innovative activities. Due to the diversity of firms' R&D activities, commercial and industrial research in universities that could be directly beneficial enough to be worth lobbying for, would confer specific benefits for individual firms or small groups of them. One or a handful of entities is a less effective pressure group than thousands of well organized farmers and agribusinesses. Individual firms may not contract such research directly from universities either, even when groups of personnel from these organizations would have substantial comparative advantages in carrying out this research. The following paragraphs outline the key demand, supply, and related contracting and institutional issues affecting the degree to which firms employ outside research specialists, specifically university scientific and technical personnel, to support their innovative functions. It is useful to consider separately, ( 1) factors within firms that affect their demand for research services by external suppliers; (2) the supply behavior of academic personnel; and (3) institutional arrangements that can resolve the incentive problems implied by the demand and supply issues. Entrepreneurship in the Firm and the Firm's Potential Demands for Outside Research Activities. Detailed contracts between firms and research sup-

pliers can be cost effective when difficult-to-predict events affecting the firm do not regularly affect what turn out to be the valued research outcomes. 5 In fact, the ultimate productivity of one research direction taken by a firm often depends on researchers' adjustments to the interim outcomes of other research dealing with the same firm-related problems, as well as on evolving outcomes in the product development area and on altered priorities due to changes in the firm's external markets. The most important institutional factor inhibiting firms from directly contracting university personnel for sizeable portions of their research needs is the inordinate amount of staff time that would be required to anticipate, and then specify in detail, the particular research activities under all contingencies that are expected to be most supportive of the firm's potential. The employees directly engaged in ongoing productive activities, as well as long-term staff doing research closely related to these activities, are usually the most knowledgeable about the needs for improvement and about the particular research activities that could have the highest payoffs for a particular firm as events unfold, given its products

36


and technologies. It is most economical for these personnel to work on an ad hoc team basis with other specialists, including scientists and engineers for whom contracts are not needed, because they face incentives as employees of the same firm. The productivity to a firm of scientific and technical staff often depends also on their knowledge about the firm's often idiosyncratic methods of production and organization. Such information is costly to obtain and depends on the degree to which these personnel face incentives to pursue the firm's long-run interest. R&D employees expecting to derive rents from long-term employment with their firms have a stronger incentive to pursue the firm's long-term interests than completely independent suppliers, especially in an innovative competitive environment where the firm's survival (and, therefore, employees' rents from long-term employment) depends on its innovative edge. 6 This incentive can serve to diminish contracting costs by encouraging employees to take the initiative in anticipating the firm's research needs to support innovation. The contracting costs of utilizing independent suppliers of research help explain why many corporations have large internal R&D operations. Universities may, nonetheless, contribute to innovative activity, often importantly, via (1) the use by researchers within firms of published university research and the employment of academic personnel as consultants to transmit or apply such knowledge; and (2) universities' roles as suppliers of trained personnel, even when the economic value of this role is diminished by the absence of institutional arrangements giving students opportunities to carry out projects relevant to research work in firms as part of their training. It is possible to more closely examine the difficulties in providing appropriate incentives for academic personnel to contribute to firm's innovative activities by considering the nature of entrepreneurship within the firm. Decentralized employees often have information cost advantages, vis-a-vis higher level managers, about the firm's benefits and costs of making alternative choices about its product lines and the markets in which to do business. Employees at lower levels of the firm having responsibility for products are often more aware of market developments and are in a better position to interpret them. This informational advantage also applies to decentralized employees' awareness of how and when to make product improvements, whether to tailor them more closely to customer preferences, and whether and how to respond to or stay ahead of competition. The firm's success in an innovative competitive environment can depend on the degree to which employees face incentives to employ their specialized


37

expertise in the firm's interest. Many of the key decisions affecting the firm's product lines, product development, methods of production, and marketing activities are made at decentralized levels, and other decisions made by top managers are based on lower level employees' recommendations which would be prohibitively costly to second guess. While top managers can question the choices made at lower levels and reward employees and groups according to the track record of success of their decisions and recommendations, they often are not in a position to be sufficiently informed to formulate these decisions or to know what they would decide themselves if they were as informed as the employees making them. As a result, entrepreneurship at decentralized levels is key to the firm's success. In this context of limited discretionary behavior based on informational asymmetries favoring lower level employees, the productivity of personnel performing useful research for the firm may be enhanced by their involvement in or close awareness of entrepreneurial activities in the firm that are relevant to their research. The contribution to firm productivity by research personnel can depend on their being included in making choices about the appropriate directions of research activity. It is well known that technically promising advances can be commercial failures. Choices that lead to successful products and production methods often require a combination of technical knowledge, awareness of the firm's strengths, and information about the marketplace. The personnel with specialized expertise in each of these areas cannot therefore operate in isolation; they must to some extent coordinate their actions with those specialized in other areas, and research personnel often must formulate goals for their activities deriving from commercial as well as technical and scientific concerns. It is important for research personnel to have incentives to take into account in their choices, goals related to the efficiency of the production processes, and the market success of the products impacted by their research activities. Incentives for research personnel to make contributions that are not economical to specify in contracts are typically based on evaluations of their contributions over lengthy periods. One can observe in many cases the direct contributions of research personnel, but also available for evaluation are the opinions about these personnel of other personnel engaged in joint efforts with researchers. Cooperation in informal work relationships can be elicited via incentives based on this observation and experience, including each employee's anticipation of need for reciprocal cooperation on projects he or she will initiate in the future. Additionally, in a Schumpeterian competitive environment, persons

38


deriving long-term rents from the firm can be motivated to help their parts of the firm adapt successfully. In effect, an employee expecting to derive long-term rents from his firm has a property interest in its success. When new initiatives introduced by competitors threaten the firm's survival, employees can feel an interest in contributing more than is required by the formally established incentives in order to protect their long-term interests in the firm. The incentive to free ride on the efforts of others is diminished by the factors noted above related to the observability of employees' joint contributions with others and the existence of reciprocal work relationships in which employees having the capacity to contribute to joint projects expect in the future to initiate projects requiring the cooperation of those presently requesting the same. The incentive problems of eliciting useful contributions to a firm's innovative activity from university (or any other external) personnel are essentially the same. When the attributes of their research most important to the firm cannot be contractually specified, they will be undersupplied in the absence of rents received from the firm. The Supply of University Research to Firms' Innovative Activities.

The supply of any kind of research, including that supporting firms' innovative activities, is a function of the institutional arrangements and incentives facing scientific and technical personnel. The strongest incentives now acting on universities' research activities in most fields are: (1) to use internal discretionary resources, including those deriving from revenues from instruction exceeding the costs of instruction, to carry out research functions that raise the professional stature of individual faculty, departments, and the institution as a whole; and (2) to respond to the well funded demands of the federal agencies that sponsor research. It is difficult to create for academic personnel having the most potential interest and capability to work on the firm's innovation related research problems, incentives that are competitive with the long-term incentives these personnel have to work on federally sponsored research and selfgenerated research projects. The problem is effectively summarized in a study for the National Science Board by Peters and Fusfeld (1982, p. 124): First, the university scientist does not see the overall problem facing industry. The industrial research manager must decompose the broad objective or problem into its scientific components, so that the component matches the research interest of the academic researcher. Thus, the university researcher may miss the opportunity to contribute to a broader issue than that within his immediate specialty. The whole may in fact be greater than the sum of its parts, and the university researcher may not be exposed to this broader picture.


39

Second, from a highly pragmatic viewpoint, industry can assign value more easily to a mission or objective than to a research component. Increased funding might be available more readily if the university system could approach industry on this basis. This would not have to take the form of shifting from basic to applied research. It would, however, change the emphasis from the independent scientist doing undirected basic research to greater emphasis on scientists cooperating in what we could term directed basic research.

In order for university personnel interested in being directly active in research supporting firms' innovations to an increased demand for research activity by private firms, it would ordinarily be necessary to devote less effort to other types of research. Over time, work on firm-related problems may require foregoing the development of alternative research programs and the reputational effects of working on them. However, there would be new avenues for building scientific reputations working on problems supporting firms' innovative activities. Presently, there is a large and growing literature that describes scientific and technological work underlying firms' innovative activities, if not the actual connections between this work and specific innovations. One reason is that firms often have to allow scientists and engineers to publish in order to attract and retain them. Another reason is that market developments can be so rapid that scientific information potentially useful to competitors quickly loses its value within publication time delays that already exist. In some industries where scientific information does not lose its value so rapidly, patent protection make publication opportunities available. In many cases, it is very difficult for competitors to draw connections between the information provided in scientific publications and its usefulness for specific innovations. Present university practices already can be flexible with respect to publications. In Peters and Fusfeld (1982, pp. 38- 39) it is pointed out that "Most universities will allow a company sponsoring research some time to review manuscripts resulting from the sponsored research for comment to ensure that they do not contain company proprietary information." And although the authors state that "Academic scientists conducting frontier research are very sensitive to this issue, and are inclined to only allow a company to review the prepublication for one week to one month," they also state that "Most university scientists are generally willing to delay a publication so that it can be reviewed for patent possibilities or, in some cases, for the time it takes to file a patent. Their rationale is that publication of an article often takes place a year or more after it is submitted." The potential responsiveness of university scientific and technical per-

40


sonnel to incentives can be readily noted. These personnel often actively pursue sponsored research funds; the large contract research enterprises within most of this country's major research universities, and individual faculty members' close adherence to the activities that are internally rewarded by their institutions, attest to the influences of incentives on the choices that academic personnel make. In addition, scientific and technical faculty have demonstrated themselves to be responsive to firms' demands for research under consulting arrangements, and faculty members have been willing to invest time in return for equity in potentially profitable start-up companies (Etzkowitz 1983). Judging from the rapid and sizeable response of university research activities that always follow major new Federal research funding initiatives, universities as organization have the capability of responding to external research demands, provided that appropriate incentives for individual academic personnel derive from these demands. Academic personnel considering making a long-term commitment to noncontractual research supporting a firm's innovative activities would be most concerned with the following two questions: (1) whether the firm had an equally long-term commitment to supporting research; and (2) that they would be rewarded by their universities, whether directly based on evaluation of their success in contributing to firm innovative activities, or due to spillovers of this work on their instruction or other research activities. Ensuring appropriate university incentives for firm-related research contributions may require changes not only in traditional rewards for different types of research, but also in the most common modes of organizing research around disciplinary lines. As noted by Peters and Fusfeld (1982, p. 125), "If industry were to make substantial funds available for research in particular areas, but the universities could not make tenure-track appointments, there would probably be little change in the research capabilities of the universities." There can be strong resistance to such changes within universities by those faculty who would not be in high demand by firms for research support, due to their previous investments in response to the existing organization and incentive structure. Organizational Arrangements to Resolve the Incentive Problems of Demand and Supply Issues. The incentive problems occurring when it is not economical to supply research supporting firms' innovative activities via specific contracts can be resolved by making two organizational changes: the establishment of "semi-independent research organizations," staffed by university personnel and students, which are wholly or mostly owned


41

by firms; and changes in universities' organizational and incentive arrangements for the academic personnel who work part-time in these organizations. Semi-Independent Research Organizations: The firm's commitment to establishing and equipping semi-independent research organizations employing academic personnel gives faculty members a considerable degree of assurance of long term commitment to research activities that would support a firm's innovative potential. These organizations would have a measure of independence both of the university employers of their scientific personnel and of the sponsor firm. They could become sizeable part-time employers of university academic personnel and could also be a major source of support for students and a large supplier of equipment having joint usefulness for instruction and other research than that directly benefiting the firm. The typical ownership of these organizations by a single sponsor firm would help provide the noncontractual incentives discussed above for employees, including their part-time academic personnel, to support the firm's innovative activities. High quality research universities typically have many qualified personnel who would attract competing bids to affiliate with semi-independent organizations. In many cases, these bids would be large enough that host universities could assess substantial overhead fees to support expenditures that provide broad benefits for the university community. The possibility of obtaining additional funding for important activities within the institution, such as basic research and support for desired students, could give universities and their scientific and technical departments significant incentives to reward the personnel who are successful in attracting funding for semi-independent organizations. There are many variants on the possible types of semi-independent research organizations as well as, of course, on the particular disciplinary backgrounds and research experiences possessed by university academic personnel affiliated with them on a part-time basis. Among the attributes of these organizations that could vary are their size, scope of research activity, degree of complementarity of their research functions with research and instructional activities within the university, internal organization, formal relationship with one or more universities, possible sponsorship by more than one firm, and organizational ties to established R&D departments within sponsor firms. Variations on the methods by which semi-independent organizations could be financed could also be explored. However, in one way or another these organizations' incomes should be partly dependent upon

42


the sponsor firm's innovative success. Ideally, such incentive financing would be based on the semi-independent organization's scientific and technical contributions to the sponsor firm's long-term profitability. The key technical problems in establishing this financing connection are the same as those of placing appropriate incentives on the firm's internal R&D operation; the contribution of either unit is often difficult to evaluate or anticipate. A possible additional problem with placing incentives on a semi-independent organization is that if the task of evaluating its performance is delegated entirely to internal R&D managers, care must be taken to ensure that evaluation is not influenced by rivalries between R&D units and the semi-independent organization. One possibility is to base funding of semi-independent organizations on a combination of assessed contributions and the firm's overall profitability. There could be a guaranteed funding base, which would decline over time. Investments in costly information are required for a firm to find the subsets of university science and technological activities having the most valuable potential complementarities with its own innovative activities, and to identify the individual academic personnel having the most potential interest in performing research supporting these activities. However, universities' search costs would typically be higher because of the need to learn about productivity relationships within firms in order to assess how their own scientific and technological activities could affect these relationships. Consequently, it would typically be more economical for firms than for universities to make initial proposals. For their part, universities could provide relatively economical information about research grants, faculty publications, areas of increased internal funding, characteristics of students in scientific programs, etc., and arrange interviews and seminars with interested faculty. Once it is known that a university has established a policy of helping to form and actively cooperate with semi-independent organizations employing its own personnel, firms would have a greater incentive to invest in information about whether it would be more economical to use these staff for some new research or for existing research currently performed by in-house R&D departments. Because of the high costs for administrators to duplicate the information held by individual academic personnel, it is important for universities to give incentives to faculty to play the leading roles in negotiations about the structure, equipping and staffing of semi-independent organizations. As noted in a document of the National Science Board (1982, p. 24), "Because of the critical importance of personal interactions at the working level, programs rarely succeed if conceived at top levels of university administration. Of course, senior administrators have an important role in formulating, shaping, and ratifying underlying cooperative agreements, but the


43

process of cooperation almost always begins through personal contacts at the working level." This document adds that "Starting in this familiar way is no guarantee of success, however, because the key individual must also have management capabilities as well as excellence in science" (p. 24). Not all academic personnel would be interested in affiliating with semiindependent organizations. However, many having interests in and a comparative advantage for supplying research supporting firm-related innovations would. Semi-independent organizations can, in addition to providing long term part-time employment relationships and research support for university personnel, offer continuing assistantship and entry career positions for graduate students in disciplines that contribute to innovation. Faculty members in specialties that potentially can make important contributions to innovations would have opportunities to participate in or supervise projects that broaden the applicability of their work. In many cases, a semi-independent organization would be used as an extension of a firm's in-house central R&D operation to deepen its capabilities in specific areas. Mechlin and Berg (1980, p. 95), for example, point out "the value that a centralized research laboratory has as a communication nexus for technical functions throughout a company. Working with the technical staffs of various divisions, laboratory personnel are in an excellent position to know that work in one area ... may have important applications elsewhere.. . . Such knowledge helps avoid both obvious duplication in research and seat-of-the-pants 'fixes' for complex problems. At the same time, it allows the laboratory to define and initiate research beneficial to several divisions but out of their financial reach individually." These authors go on to state that the best way to coordinate research activity is to "put all research personnel together in the same physical place and let the coordination happen over lunch, in the library, in the hallways. Until these people can rub shoulders daily, until they can work together informally on each other's projects, the degree of communication needed for an effective coordination of effort simply will not exist" (p. 96). However, firms with many highly decentralized research units, including one or more university-related semi-independent organizations, might want to introduce competition among them. This competition could be facilitated by employing specialized personnel who broker staff research contributions and the more ongoing activities of the firm. In either case, semi-independent organizations could be instituted as modest-sized annexes of central R&D operations and then allowed to grow either according to these operations' internal evaluations of the performance of their subunits or to the extent that they can compete for corporate-wide funding. Semi-independent research organizations can play important instruc-

44


tional roles for universities. Most personnel working on technological change and science- and technology-based innovation are, of course, university trained. However, semi-independent organizations could enhance the efficiency of supply of these workers by relating their research experiences as part of their training more specifically to firms' innovative activities. Over time, these organizations would increase the relative supply of science graduates having experience performing research underlying firms' innovative activities versus experience performing research on federally sponsored projects or those supported with internally generated university resources. Organizational and Incentive Changes in Universities: Some of the problems to be considered regarding the relationships of semi-independent organizations to universities can be seen in the following example. The administration of Harvard University gave the following reasons, cited in Weiner (1982, pp. 437-8), for withdrawing a proposal for a genetic engineering firm to be partly owned by the institution and certain faculty members: "(a) academic discussion could be impaired because of commercial competition; (b) that professors and graduate students might shirk academic duties and interests to pursue commercial ones; (c) that the administration's authority to protect its academic interests might diminish; and (d) that Harvard's reputation for academic integrity might be damaged by even the appearance of conflict between its academic and financial interests." Although the suggested semi-independent organizations would be owned outright by the sponsoring corporations (in order to create the desired incentives for their personnel to support the firm's innovative success, discussed above) rather than owned by universities or faculty, it is worthwhile to consider the pertinence of these same issues. The possibility of commercial competition among a university's faculty suggests that when there are proposals for more than one semi-independent organization involving faculty in disciplines that work frequently together, the number of organizations should be restricted or organizations should be selected in a manner that avoids such competition. The issue of shirking academic duties relates most importantly to the degree to which research functions in semi-independent organizations would either themselves be academic activities important to the university or highly complementary to such activities. From the university's point of view, this is the most important theme for the design of semi-independent organizations. To the extent that activities in a proposed semi-independent organization do not meet these criteria, there is little justification for the organization. The problem is to determine whether these criteria are met in practice.


45

As objective as possible use should be made of information about training opportunities for students and assessments of spillovers from research in a semi-independent organization on other research activities. Unfortunately, faculty not interested in the type of research performed in semi-independent organizations or who are in fields for which there is little demand from private industry may not make objective judgments. The university's overall financial benefits from semi-independent organizations, if spent in ways that benefit the broader academic community, may raise general acceptance of their presence and enhance objectivity in the evaluation of specific proposals. As for the last two issues considered by the Harvard administration, it is arguable that even though universities would have no ownership interest in semi-independent organizations, they would be influenced by the sizeable financial benefits they could derive from faculty members' affiliations with these organizations. These gains could make administrators reluctant to take actions that conflicted with the interests of the firms that own the semi-independent organizations. To deal with this possible problem, a university should affiliate with a large enough number of semi-independent organizations throughout its scientific and engineering fields so that the costs of terminating a relationship would not be felt keenly by the institution as a whole. Knowing this, and considering the reciprocal benefits of semi-independent organizations to firms, it is unlikely that a firm would threaten to terminate the funding of a semi-independent organization over a policy conflict. It is worthwhile to give more consideration to the issue of determining what are the scientific and technological activities that are acceptable as part of the academy. The mix of scientific activities that in fact occurs in universities is not just a function of the inherent nature of these activities, but also of the demands for research that have succeeded in creating incentives for academic personnel to respond to these demands. Semiindependent organizations are a means of overcoming the contracting problems that prevent firms' demands for research from creating incentives that effectively compete with the incentives emanating from the research demands of governmental agencies. The task for universities is to select semi-independent organizations providing incentives for research activities comparable or superior in intellectual significance to those now taking place. Universities' resistance to change, even that which would enrich the academic environment, is largely a function of the heavy investments made by their academic personnel in the existing structure of incentives. Yet, the potentially large financial gains to universities from semiindependent organizations could be used to provide facilities, research

46


funding, and support for students that would yield benefits to broader groups of faculty than those working on firms' research projects. Foresighted universities should carefully arrange to place charges on the newly affiliated organizations and plan for expenditures of these charges that are broadly and visibly beneficial within the institution. Apart from sharing the university's fiscal gains from semi-independent organizations, the institution could develop procedures providing limited involvement of nonaffiliated university personnel in the governance of these units. The work of the personnel in these organizations should be regularly evaluated by committees including nonaffiliated university personnel, but these committees should have the resources to bring in university and industry scientists having strong academic credentials who can provide information helpful in the evaluations. The nonaffiliated evaluative personnel from within the university should be representative of a variety of disciplines because the semi-independent organizations would often represent interdisciplinary alignments of activities. The resource costs of separate evaluation need to be explicitly taken into account. In many cases, traditional publication channels and professional organizational arrangements cannot be counted upon to provide adequate evaluation information for all of the activities of personnel contributing research that supports firms' innovative activities. The proprietary nature of some research is a problem; it is not the most serious one. As noted earlier, company R&D personnel frequently do publish, and universities would be in a position to negotiate publication policies favorable to their faculty working in semi-independent organizations. The main problem is possible disagreement between university personnel who are and are not involved with semi-independent organizations over the appropriateness and importance of research performed in these organizations, whether published or not. One option for universities is to establish interdisciplinary departments in areas covered by semi-independent organizations as an institutional change that may help to reduce the costs of evaluation as one of their benefits. Conclusion

Three conclusions follow from this discussion. First, while existing research examining the relationships between higher education and economic growth includes interesting insights, this research does not provide a conceptual or empirical basis for policymaking. More detailed inferences about the impacts of specific educational functions on activities in the economy that


47

are growth enhancing are needed. The design of research projects aimed at making these inferences should be based on explicit assumptions about the nature of the economic growth process. Second, there are many promising avenues for research to infer connections between higher education and economic growth characterized by innovative behavior. Feasible projects include those that improve understanding of the roles of scientifically trained manpower in the supply of innovations and the roles of these personnel in facilitating adjustments to the disequilibria created by innovations. Other suggested projects apply methodologies for using data on citations to trace and characterize relationships between university-based scientific work and firms' science and technology-based innovations. The final conclusion is that appropriate incentives could significantly enhance the contributions of universities' scientific and technical personnel to firms' science- and technology-based innovations. Although industryuniversity cooperative relationships are rapidly being implemented in some leading universities, key incentives for academic researchers to support firms' innovative successes continue to be missing. Such incentives are needed to overcome the infeasibility or high costs faced by firms in contracting with external suppliers for certain types of research. The absence of adequate incentives often forces firms to perform research inhouse, which in many cases could be more efficiently provided by university personnel. The establishment of semi-independent organizations may give scientific and technical personnel in universities improved incentives to supply research for firm-related innovations.

Notes I. Two well-known difficulties with the attribution of differences in incomes to education are (I) selectivity bias; and (2) the fact that inefficient or socially unproductive activities can contribute to the incomes of university graduates. These two biases tend to offset each other. The latter bias is especially important in societies where above market wages are paid by governmental agencies or parastatals and positions are rationed on the basis of educational credentials. 2. The technology of quality measurement is in its infancy in national income accounting, and national income accounts are slow even to take into account rises in the productivity of key inputs such as computers used throughout the economy. 3. An example of a study that attempts to deal with this simultaneity problem is Wheeler (1980). 4. The geographically based variables that commonly affect the location decisions of different types of public services are also needed in a model that explains research outcomes and the presence of research supplying organizations. Recent research on initial activities

48


and economies of scale (Arthur 1990) has the potential to provide useful insights into these common locational variables. 5. The problems and costs of specifying detailed contracts and enforcing them when there are such contigencics, and the institutional alternatives in which suppliers receive appropriable rents, are explained in Klein et al. (1978). 6. The tendency for rent-receiving employees to make choices in the firm's interest, when the firm's survival or success is threatened in a dynamic competitive environment, is discussed in Hoenack (1983, 1989). The arguments presented in the following paragraphs arc based on these discussions.

References Arthur, W.B. (1990). Positive feedbacks in the economy. Scientific American, 262, February, 92-99. Bowman, M.J. (1964). Schultz, Denison and the contribution of "Eds" to Economic growth. Journal of Political Economy, 72, 450-464. Boyle, K.A. (1986). Technology transfer between universities and the U.K. offshore industry. IEEE Transactions on Engineering Management, EM-33 February, 33-42. Denison, E.F. and Poullier, J. (1967). Why growth rates differ: Postwar experiences in nine western countries. Washington, D.C.: The Brookings Institution. Dosi, G. (1988). Sources, procedures, and microeconomic effects of innovation. Journal of Economic Literature, 26, September, 1120-1171. Etzkowitz, H. (1983). Entrepreneurial scientists and entrepreneurial universities in American academic science. Minerva, 21, 198-233. Freeman, C. (1982). The economics of industrial innovation. Cambridge: MIT Press. Glazer, S. (1986). Businesses take root in university parks. High Technology, 6, January, 40-47. Granger, C.W.J. (1969). Investigating causal relations by econometric models and cross-sectional methods. Econometrica, 37, July, 424-438. Hicks, N.L. (1987). Education and economic growth. In G. Psacharopoulos (Ed.), Economics of education: Research and studies (pp. 101-107). Oxford: Pergamon Press. Hoenack, S.A. (1983). Economic behavior within organizations. Cambridge and New York: Cambridge University Press. Hoenack, S.A. (1989). Group behavior and economic growth. Social Science Quarterly, 70, 744-758. Jaffe, A.B. (1989). Real effects of academic research. American Economic Review, 79, December, 957-970. Kash, D.E. (1989). Perpetual innovation: The new world of competition. New York: Basic Books, Inc. Klein, B., Crawford, R.G., and Alchian, A.A. (1978). Vertical integration,


49

appropriable rents, and the competitive contracting process. Journal of Law and Economics, 21, 297~326. Lieberman, M.B. (1978). A literature citation study of science-technology coupling in electronics. Proceedings of the IEEE, 66, January, 5~13. Logan, L.B., and Stampen, J.O. (1985). Smoke stack meets ivory tower: Collaborations with local industry. Educational Record, 66, 26~ 29. Louis, K.S., Blumenthal, D., Gluck, M., and Stoto, M.A. (1989). Entrepreneurs in academe: Exploration of behaviors among life scientists. Administrative Science Quarterly, 34, March, 110~ 131. Mechlin, G.F., and Berg, D. (1980). Evaluating research: ROI is not enough. Harvard Business Review, 58, September/October, 93~99. National Science Board. (1982). University-industry research relationships: Myths, realities and potentials. Fourteenth Annual Report of the National Science Board. Washington. D.C.: National Science Foundation. National Science Foundation. (1980). National patterns of science and technology resources: 1980. (NSF 80~308). Washington, D.C.: author. National Science Foundation. (1988). National patterns of science and technology resources: 1987. (NSF 88~305). Washington, D.C.: author. Peters, L., and Fusfeld, H. (1982). University-industry research relationships: Selected studies. Fourteenth Annual Report of the National Science Board. Washington, D.C.: National Science Foundation. Powers, D.R., Powers, M.F., Betz, F., and Aslanian, C. B. (1988). Higher education in partnership with industry. San Francisco: Jossey-Bass Publishers. Psacharopoulos, G. (1985). The contribution of education to economic growth: International comparisons. In J.W. Kendrick (Ed.), International comparisons of productivity and causes of the slowdown. Cambridge: American Enterprise Institute/Ballinger Publishing Company. Psacharopoulos, G., and Woodhall, M. (1985). Education for development. New York: Oxford University Press. Sahal, D. (1985). Technological guide-posts and innovation avenues. Research Policy, 14, April, 61 ~82. Schmookler, J. (1966). Invention and economic growth. Cambridge: Harvard University Press. Schultz, T.W. (1961). Education and economic growth. In N.B. Henry (Ed.), Social forces influencing American education. Chicago: University of Chicago Press. Schultz, T.W. (1975). The value of the ability to deal with disequilibria. Journal of Economic Literature, 13, September, 827~846. Schumpeter, J.A. (1950). Capitalism, socialism and democracy (3rd. ed.). New York: Harper and Row. Sims, C.A. (1972). Money, income and causality. American Economic Review, 62, September, 540~552. Weiner, C. (1982). Relations of science, government, and industry: The case of recombinant DNA. In Policy outlook: Science, technology and the issues of the

50


eighties: A report from the American Association for the Advancement of Science. Volume II, Sources. Washington, D.C.: National Science Foundation. Wheeler, D. (1980). Human resource development and economic growth in developing countries: A simultaneous model. World Bank Staff Working Paper, No. 407. Washington, D.C.: The World Bank.

3

HIGHER EDUCATION, ECONOMIC GROWTH, AND EARNINGS John Pencavel

This chapter provides a selective, nontechnical, review of research on the contribution of education, and especially higher education, to economic growth and efficient resource allocation. Three major questions are posed. First, what is known about the role of education in economic growth? This question has been addressed in different ways. Some economists have applied growth accounting techniques to macroeconomic data on changes in schooling completion levels and changes in aggregate output. These procedures are outlined. Though there is a very large literature in this vein, the methods are arbitrary and the results are fragile. Recently, economists have started to examine more systematically the association across countries between economic growth rates and school enrollment rates. This may well prove to be a fruitful line of research. At present, it is too early to report with any confidence on its conclusions on the impact of education on growth. This research is sketched. Another approach makes use of observations at the level of individual industries to measure the association between the educational attainments of workers and productivity growth. This research confirms the notion that education contributes positively to economic growth, but it has not been successful in quantifying that contribution at all precisely. The reasons for this are discussed. The second major question concerns the earnings premia attached to college education. Such earnings premia permit inferences about the productivity of additional schooling, insofar as earnings differentials among 51

52


individuals reflect productivity differences. In most cases, the productivity of workers is not directly observed and so the correspondence between schooling and productivity cannot be computed, but there are powerful reasons for expecting such an association in markets dominated by competitive, profit-seeking, firms. Virtually all economic models of wages and schooling attribute at least a large part of the earnings premia associated with additional schooling to the greater productivity of more highly educated workers. Where these models differ is over the issue of whether additional schooling causes an increase in labor market productivity. 1 The earnings premia attached to a college education have displayed quite marked movements. The causes of these movements are not well understood. These changes imply that the returns to investments in higher education both for individuals and for society at large may be difficult to forecast accurately: when the college earnings premium is relatively high, the personal and social incentives to attend college and to expand public support for higher education will seem greater; however, by the time the skills have been acquired and additional school resources provided, the returns may have fallen considerably. This presumes, of course, that the schooling choices of individuals respond to market rewards and this constitutes our third question: to what extent is the supply of welleducated labor responsive to earnings? This is discussed below. Whereas the following section concerns the effect of additional schooling on earnings, the discussion in the final section turns this around and inquires into the effect of earnings and other variables on college attendance and on the acquisition of particular professional skills. The general goal of this chapter is to provide an accessible and up-todate statement of our knowledge of the relationships among education, economic growth, and earnings with special attention to higher education and the labor markets of college educated workers in the United States. In this area, there exist certain empirical regularities. The appropriate interpretation of these regularities is less clear, however. Macroeconomic Growth Accounting

Methodology

In the macroeconomic growth accounting literature, the primary issue motivating research has been to quantify each input's, including education's independent contribution to economic growth. An illustration of what is deemed a satisfactory answer is the following: "the contribution of edu-

HIGHER EDUCATION, ECONOMIC GROWTH, AND EARNINGS

53

cation to output growth [in the United States from the Second World War to the mid-1970s] has always been positive and has accounted for about 15 percent to 25 percent of growth in national income per person employed. " 2 (p. 24) The manner in which such estimates are derived may be outlined as follows. An economy is likened to an individual production unit, a "firm," that transforms its purchased inputs into an output, X. Suppose the purchased inputs consist of hired workers, L, and machines, K, in which case the transformation of inputs into output may be described by a production function: X= X( L,K). Provided the same form of the production function applies over time, the proportionate rate of growth of output over time, 11.CnX1, is 11.CnXr

=

(Exdr 11.CnL1 + (ExK)r 11.CnKr

where 11.CnL1 and 11.CnK1 are proportionate growth rates in labor and capital and (Exdr and (ExK)r are output elasticities of labor and of capital at time t. 3 If this firm faces fixed input and output prices and if it maximizes profits, 4 each output elasticity equals the ratio of each factor's total payments to total revenues: (Exdr = (wL!pX)r = su and (ExK)r = (rK!pX)r = sKr· This allows the growth rate to be expressed in terms that are usually observable: 11.CnXr

=

s Lt 11.CnLr + s Kt 11.CnK1

The right-hand side of this equation sometimes goes by the name of the Divisia input index. In practice, the average values of factor shares over a period of time sL and sK, are used in place of su and sKr respectively: 5 (1)

Labor's contribution to economic growth is defined as sL11.CnLr, and capital's contribution to economic growth is defined as sKI1.CnKr. Upon computation, the right-hand side of equation (1) often falls short of the left-hand side. This difference is dubbed the rate of technical change, that is, the rate at which the production function is moving upward through time, permitting more output to be extracted from given inputs. The definition of h 11.CnLr as labor's contribution to economic growth and sK11.CnKr as capital's contribution to economic growth calls for comment. Is this a useful and meaningful definition? It is, if the rate of growth of output may be decomposed exactly into two distinct parts, one that is contributed by labor and independent of capital and the other that is contributed by capital and independent of labor. But is it likely that

54


labor's contribution to the growth in output would be independent of the quantity and type of capital with which labor works? Such independence seems doubtful in view of the experience of certain developing countries where the expansion of the educational system has resulted in disproportionately higher unemployment rates for well-educated workers or in highly educated workers occupying jobs unsuited to their training. Analogously, will the contribution of capital to economic growth not depend upon whether skilled or unskilled workers operate the capital? In fact, whenever incentives are not skewed, particular machines are designed for the attributes of particular workers, while workers are trained to match the requirements of particular machines. In other words, the idea that labor's contribution to economic growth may be assessed independently of the quantity and type of capital seems inconsistent with some basic economic practices. More formally, the notion that the growth in output may be decomposed into two independent parts requires the proportional effect of labor on output, £xL, which is measured by h, to be independent of the quantity of capital available. Also it requires the elasticity of output with respect to capital, £xK, which is measured by sK, to be unaffected by the amount of labor employed. The only production function for which this is true is one in which the inputs are strongly separable (additive) when inputs and outputs are expressed in logarithmic units, that is, the CobbDouglas function: (2)

where A is an unknown constant. Popular production functions such as the constant elasticity of substitution (CES) or the translog do not embody the necessary independence assumption and, therefore, do not permit a clear decomposition between labor's and capital's contribution to economic growth. In short, the search for any input's (including education's) independent contribution to economic growth will be futile if, as seems likely, its contribution is not independent of other factors of production. In the growth accounting literature, the inputs of labor and capital are recognized as having both quantitative and qualitative dimensions. In the case of labor, its quantitative components include hours worked per worker and the number of workers while the educational attainments are included in the quality dimension. Denote by f':...€nLjr the j-th component of the proportionate change in labor input and define f':...€nLr as the sum of the various n-quantitative and qualitative components: n

f':...€nLr =

2:

i=l

f':...€nL;r· Then the growth accounting literature defines the


55

contribution of component j to the growth of output as su1.fnLjr· This is yet again a further restriction on the original production function: not only must labor and capital be additive in a logarithmic production function, but the components that constitute the labor branch must be logarithmically additive with unitary coefficients. In other words, the production function, equation (2), takes the following form: (3)

where n components of labor and m components of capital are identified. Equation (3) implies that the proportional effect on output of a doubling of hours per worker is the same as a doubling of the number of workers and as a doubling of the impact of education on output.

Maddison's Results

The exposition above has been in the context of a single production unit, a "firm." The macroeconomic growth ~ccounting literature treats the aggregate economy as if it mirrored this individual firm. To illustrate, consider Angus Maddison's (1987) recent analysis of six countries' economic growth between 1913 and 1984. With respect to his treatment of education, he categorized the population by years of schooling as in table 3-1 and then aggregated these schooling data, applying a weight of 1.4 to secondary education and 2 to higher education. He suggests these weights roughly correspond to relative wages between each category and primary education, the tacit assumption being that the skills embodied in one year of secondary education and in one year of higher education are 1.4 and 2 times as productive as those in one year of primary education. The average annual compound growth rates of his schooling indices are given in line one of table 3-2. 6 Higher education's contribution to the growth in this educational index is given in line two of table 3-2; for the United States, for instance, the fraction of the growth in the educational index accounted for by higher education was almost 9 percent during 1913 to 1950, 24 percent during 1950 to 1973, and 62 percent during 1973 to 1984. 7 Clearly, with the extension of primary schooling, it has been the growth of higher education in these countries that has contributed most in recent decades to the growth in the schooling indices in line one of table 3-2. In accordance with the reasoning outlined above, the contribution of education to the growth of output is defined as h !!.fnLer where l!.fnLer is the growth in the educational index as given in line one of table 3-2. 8 Line four of table 3-2 expresses the growth in the educational dimension

56


Table 3-1. Sixty-Four

U.S.A

U.K.

Japan

Germany

France

Netherlands

Average Years of Formal Schooling of Population Aged Fifteen to

1913 1950 1973 1984 1913 1950 1973 1984 1913 1950 1973 1984 1913 1950 1973 1984 1913 1950 1973 1984 1913 1950 1973 1984

Primary

Secondary

Higher

Total

4.90 5.61 5.80 5.80 5.30 6.00 6.00 6.00 4.50 5.88 6.00 6.00 3.50 4.00 4.00 4.00 4.31 4.96 5.00 5.00 5.30 6.00 6.00 6.00

1.83 3.40 4.62 5.10 1.90 3.27 3.99 4.50 0.56 2.08 3.79 4.56 3.35 4.37 5.11 5.17 1.77 3.04 4.11 4.89 0.64 1.17 2.49 3.34

0.20 0.45 0.89 1.62 0.08 0.13 0.25 0.42 0.04 0.16 0.39 0.59 0.09 0.14 0.20 0.31 0.10 0.18 0.46 0.90 0.11 0.24 0.39 0.58

6.93 9.46 11.31 12.52 7.28 9.40 10.24 10.92 5.10 8.12 10.18 11.15 6.94 8.51 9.31 9.48 6.18 8.18 9.58 10.79 6.05 7.41 8.88 9.92

Notes: This table is reproduced from table A-12 of Maddison (1987).

of labor input (that is, sL 1lCnLe1) as a percentage of the growth in gross domestic product, given in line three of table 3-2. Correspondingly, the percentage of the growth of output contributed by higher education is given in line five of table 3-2. 9 Simply averaged over these six countries, the contribution of schooling to economic growth was 22 percent in both the 1913 to 1950 and the 1973 to 1984 periods, and it was 7 percent in the golden age of growth from 1950 to 1973. Higher education's contribution to economic growth in these six countries rose from an average of 0.9 percent in 1913 to 1950, to 1.1 percent in 1950 to 1973, and 7.7 percent in 1973 to 1984. As shown in table 3-2, higher education's role in economic growth has always been relatively greater in the United States where it has embraced a larger fraction of people in any age cohort.

-....1

U1

0.63 21.16 3.78 11.71 2.48

0.14 71.19 1.68 5.90 4.20

0.86 38.21 2.18 27.53 10.52

0.78 17.05 1.58 34.61 5.90

two Notes: The entries in lines one and three are annual average compound growth rates. They are given in table 19 of Maddison (1987). Line = 0. 7, expresses the growth in higher education as a percentage of the growth in the educational index. Line four is formed by multiplying line one by h is then dividing by line three, and finally multiplying by 100. Line five is the product of line two and line four divided by 100. Each entry in this table rounded to the second decimal place.

0.46 28.57 1.06 30.23 8.64

0.62 7.70 4.70 9.17 0.71

0.78 62.41 2.32 23.41 14.61

0.52 22.20 5.13 7.08 1.57

1973 to 1984 1. Percent growth in educational index 2. Percent higher education's contribution to (1) 3. Percent growth in output (GDP) 4. Schooling's contribution to output growth (%) 5. Higher education's contribution to output growth (%)

0.28 7.41 5.92 3.26 0.24

0.58 23.88 3.72 10.87 2.60

1950 to 1973 1. Percent growth in educational index 2. Percent higher education's contribution to (1) 3. Percent growth in output (GDP) 4. Schooling's contribution to output growth (%) 5. Higher education's contribution to output growth (%) 0.74 11.08 9.37 5.54 0.61

0.38 11.13 2.43 11.03 1.23 0.51 2.92 1.06 33.88 0.99

0.35 2.55 1.30 19.00 0.48

0.87 3.88 2.24 27.04 1.05

0.47 1.53 1.29 25.49 0.39

0.59 8.69 2.78 14.78 1.29

Percent growth in educational index Percent higher education's contribution to (1) Percent growth in output (GDP) Schooling's contribution to output growth (%) Higher education's contribution to output growth (%)

1. 2. 3. 4. 5. 0.29 16.42 3.02 6.61 1.08

Netherlands

France

Germany

Japan

1913 to 1950

UK

Growth in Schooling and Its Contribution to the Growth in Output, 1913 to 1984 USA

Table 3-2.

58


These estimates are representative of a large literature suggesting that education's contribution to economic growth has been variable across countries and over time, but measured in this way it is estimated to have made a substantial contribution to growth. For example, in Denison's (1979) well-known research, education accounts for between 15 percent and 25 percent of growth in U.S. national income per employee. Its contribution to economic growth has tended to increase over time. Though these growth accounting procedures are the standard macroeconomic methods for quantifying schooling's role in economic growth, in my view these procedures have some serious shortcomings and do not provide accurate estimates of the contribution of schooling to economic growth. Shortcomings

First, if educated labor contributes to the production of goods that are incompletely measured in national income accounts, then the contribution of education to growth will be mismeasured by the methods described above. For instance, conventional methods of output growth are hard pressed to incorporate improvements in the nation's health care. There is a clear positive association between indicators of good health (including longevity) and schooling. This association arises in part through the indirect effect of schooling on earnings that permits the purchase of more investments in health. However, schooling and health are correlated, even holding earnings constant. This probably reflects the ability of more schooled individuals to acquire and process information about healthrelated matters. Schooling is said to produce a more responsible and informed citizenry and to have implications for the diffusion of technology, fertility, and the use of leisure time, outcomes that are incompletely measured in output statistics. Second, identifying schooling inputs with the number of people reaching a particular schooling level is a crude way of quantifying the educational system's role in the embodiment of knowledge in people. The activities of educational institutions are not confined to the dissemination of existing knowledge, but extend to the generation of new knowledge. This is especially the case of universities, many of which in fact, if not in word, place greater emphasis on their research than their teaching activities. These research contributions are not recognized in procedures that do little more than count the schooling completion levels of the population or labor force. Moreover, counting the years of schooling of a nation's work force


59

ignores, as an educational index, movements over time and across countries in scholastic performance and the quality of schooling. The relevance of schooling quality for productivity is suggested from microeconomic data where individuals experiencing better schooling subsequently command higher earnings. 10 This may be because better schools are more effective in reinforcing attitudes (such as perseverance, self-confidence, and punctuality) that are valued in market work because there is not a strong and consistent positive correlation between schooling quality and student test scores. This weak correlation reflects the general difficulty in estimating meaningful educational production functions, that is, stable relationships relating educational outcomes to various schooling inputs. Such relationships are notoriously hard to detect. This is partly because observations on certain critical variables (such as family background) are lacking and because those variables that are measured are often measuring the relevant inputs and outputs with error. That educational outcomes are related to inputs may be viewed as a tautology, but this does not mean we are sufficiently well informed of the precise form of the relationship to make confident statements about the effectiveness of changing different educational inputs. 11 Third, the problems of measuring schooling's contribution to the growth in productivity are compounded by the fact that the educational system responds to (as well as influences) the rate and nature of economic growth. The output of the educational system is not imposed exogenously on the economy, and this implies that the question of higher education's contribution to economic growth is not well defined: in the United States where the educational system interacts closely with the market economy, it is no more meaningful to use the observed movements in output and schooling completion levels to measure the contribution of education to economic growth than to organize the same data to compute the effect of rising wealth on educational attainments. Finally, the growth accounting framework is flawed by not recognizing the implications of schooling being an intermediate, not a primary, input. Educated labor is itself produced through the application of labor and capital, inputs that could have been directed to the production of other things. In effect, in apportioning the determinants of growth, the growth accounting procedures count not only the growth in numbers of educated labor but also the growth in the inputs used to produce educated labor. To understand this point, return to the metaphor of the single production unit, the "firm," and suppose it operated an apprenticeship program to train its new hires. Unskilled workers, L u, are instructed by teachers, L r, and educational capital, KT, and in three months emerge as trained or

60


skilled workers, L 5 . This production of skilled workers is described by the following training function: (4)

These skilled workers then operate other capital, K 5 , to produce useful output: (5)

Now the procedures in the macroeconomic growth accounting literature for quantifying education's contribution to output growth would use equation (5) to define the contribution of skilled labor to this firm's growth in output as gxL !'ienL 5 where gxL = (og/oL 5 ) (L 5/X). However, this skilled labor is generated only by employing labor, Lr, and capital, Kr, that in principle could be converted for use in producing output X. This "lost" output used to train labor is not recognized by defining skilled labor's contribution to economic growth as gxL !'ienL 5. According to equation (4), the proportionate growth in skilled labor may be related to the proportionate growth in the inputs used in the apprenticeship program: !'ienL s

=

fsu !'ienU + [c;r !'ienL r + [c;K fienKr

where fsu = (of/oU)(U/L 5 ), fsr = (of/oLr) (C'!L 5 ), and [c;K = (of/ aKr) (K 1 ! L 5 ). So the purported contribution of skilled labor to economic growth, gxL !'ienL 5, is nothing other than gxL times the weighted sum of the proportionate changes in three other inputs, unskilled labor ( L u), teachers ( L r), and instructional capital ( K~'). This model of a firm's apprenticeship program is formally the same as the nation's educational system in that, in educating labor, the schooling system absorbs resources that could otherwise be allocated to producing current output. Some recognition needs to be paid, therefore, to the opportunity cost of producing intermediate inputs such as educated labor. In fact, in assessing education's role in economic growth, the macroeconomic growth accounting literature usually recognizes that double counting would be involved if changes in L u were included along with changes in L 5 in decomposing the growth in output among its elements. However, it is not recognized that double counting is also involved when both changes in L 5 and changes in the other inputs, L T and K 1 , are included in computing changes in employment and changes in capital, respectively. Growth accounting should be more defensible if education's contribution were calculated by subtracting changes in L r and Kr from changes in v~·. A better procedure than this would involve a comparison between the change in output that has been achieved from the educational system and

61


the change in output that would have emerged if the educational system had not absorbed the resources LT and Kr. That is, using equation (5), the actual output is X= g(Ls,Ks). Now suppose that, contrary to fact, unskilled (or uneducated) labor, L u, instructional labor, L T, and instructional capital, Kr, were applied to producing output according to equation (5): X*= g(Lu,Lr,Ks,Kr)

(6)

denote the proportionate change in output by !}.fnX1 fnX1 - €nX 1_ 1 and the proportionate change in output that would have emerged from equation (6) by /}.fnXt = fnXt - fnX1_ 1 • The contribution of education to economic growth may be defined as !}.fnX1 - !}.fnXt. Of course, this measure of education's contribution to economic growth requires the assessment of a counterfactual statement, equation (6). Nevertheless, this should not be difficult to undertake and is a much more appealing construct. 12 Economic Growth and Schooling Enrollment Across Countries

A different procedure for assessing education's contribution to economic growth is to examine the growth records of a cross-section of countries and determine the association between these growth rates and the educational attainment of the populations or labor forces. This has been undertaken recently both by Barro (1991) and by Baumol, Blackman, and Wolff (1989). As will become evident, the procedure here is quite different from that in the macroeconomic growth accounting literature: in the latter, the observed change in a country's output is decomposed into changes in the factors, in that country supposedly generating the growth in output, including the change in educational attainments; in the work of Barro and of Baumol, Blackman, and Wolff, changes in output across countries are related to levels of schooling across these countries. To illustrate with Barro's work, the following regression equation is representative: !}.fnX; = ~o

+

~~X;

+

~2Ef

+

~3Er

+

~4zi

+

Ej

where !}.fnX; stands for country i's annual rate of growth of per capita real gross domestic product (GDP) from 1960 to 1985. X; is country i's per capita real GDP in 1960, E[ and Er indicate country i's schooling enrollments in 1960 at the primary and secondary level respectively, and Z; represents several control variables. 13 Omitted variables are included

62


in the residual Ei. For the purpose of assessing the impact of educated labor on economic growth, it would have been preferable to use indicators of educational attainments. However, data on the educational attainments of the labor force are not available for all ninety-eight countries in the sample, so EP and Es were used instead: E~' is the number of children enrolled in primary school as a fraction of all children of primary school age, and Es is the fraction of secondary school age children enrolled in secondary school. Barro's least-squares estimate of ~ 1 is -0.0075 with an estimated standard error of 0.0012: Xi is measured in thousands of 1980 US dollars, so ~ 1 = -0.0075 implies that a country with a $1,000 higher per capita real GOP in 1960 experienced a 0.75 percentage points per year lower growth rate. This is evidence in favor of the so-called convergence hypothesis according to which "follower" economies will grow faster than those in the "lead" and productivity levels tend to converge. The estimated coefficients on the schooling enrollment variables are positive and highly significant according to conventional criteria: the coefficient attached to primary schooling, ~ 2 , is estimated to be 0.0250 with a standard error of 0.0056, and the coefficient attached to secondary schooling, ~ 3 , is estimated to be 0.0305 with a standard error of 0.0079. In other words, consider the implications of a 10 percent higher school enrollment rate. If that higher enrollment had been at the primary school level, subsequent economic growth would have been 0.25 percent higher. If that higher enrollment had been at the secondary school level, subsequent economic growth would have been 0.31 percent higher. Given that the average growth rate of per capital real GOP in this sample was 2.2 percent, these implied impacts on economic growth of higher school enrollment rates are of real consequence. Baumol, Blackman, and Wolff (1989) also examined the partial association between economic growth and enrollment in higher education. They report a positive association that is almost always significantly greater than zero by conventional criteria. They find that, holding constant initial per capita real GOP and the annual rate of population growth, a 10 percent higher enrollment rate in higher education in 1965 is associated with a 0.22 percent higher annual growth rate of real GOP between 1960 and 1981. Differences among countries in their enrollment rates in higher education appear to have larger impacts on economic growth than differences in their primary schooling or secondary schooling enrollment rates. 14 This research is more persuasive evidence of the impact of education and of higher education on economic growth than that contained in the growth accounting literature. However, much more needs to be done in


63

this line of work before confident inferences can be drawn from these partial correlations. Many other factors are conjectured to affect economic growth and it is yet to be seen whether the apparent importance of education survives once account is taken of these other factors. Barro's fitted equations remove about 56 percent of the variation in economic growth rates across countries so clearly that there is room for other explanatory variables. Indeed, he recognizes his equations are particularly unsatisfactory in describing the growth performance of subSaharan Africa and Latin America. There is also the question of the causes of the cross-country variation in school enrollment rates. Countries with higher growth rates are likely to allocate some of their greater wealth to the provision of more schooling for their population. If countries with relatively higher economic growth rates in the 1960 to 1985 period also experienced relatively higher growth rates from 1945 to 1960- in other words, if what calls for explanation is the pattern of economic growth rates over a longer period than the 25 years measured by Barro 's dependent variable- then the cross-country variation in school enrollment rates iri 1960, Er and Ef, are in part a consequence (and not simply a cause) of the variation in growth rates. So, at this stage of inquiry, confident inferences about the macroeconomic effects on growth of education and of higher education are not warranted. Microeconomics Research on Education and Technology Change

In addition to the macroeconomic literature reviewed above, the contribution of higher education to the process of economic growth has been examined at the microeconomic level. This literature identifies at least three ways in which schooling serves as a productive input in work performance. First, it may simply enhance the effective labor input from any given hours of work. In this case, a more schooled labor force is equivalent to a larger labor force, enabling more output to be produced from its resources. This enhanced labor input may well increase the productivity of other inputs too. Second, schooling may contribute to allocative efficiency Y A more educated decision maker is more likely to be better informed about the contributions to output of each input and about and output prices. With more schooling, individuals are better able to absorb and follow instructions, to take initiatives and deal with unforeseen contingencies, and to envisage and exploit new possibilities. Indeed, a good deal of

64


evidence supports the notion that more schooled individuals more readily adopt new technologies. 16 For instance, studying the choice of inputs by Iowan farmers, Wozniak (1984) finds that more educated farmers are more likely to adopt new technological innovations although, once adopted, the extent of use of the new input is uncorrelated with farmers' education levels. Third, much of a nation's research and development (R&D) activities, and especially its basic research, takes place in universities and an expansion of the higher educational system is usually accompanied by a complementary increase in expenditures on R&D. Indeed, in graduate education, research and instruction are often joint products and personnel and capital equipment serve both purposes. In many cases, there is a close association between R&D activities in universities and those in research institutes and in private industry and it is reflected in the fact that firms with advanced technologies often locate close to major universities (Jaffe 1989). R&D expenditures are typically linked to technological progress which in turn is identified as a key component of economic growth. In each of these ways, greater schooling may increase an economy's productive potential- by augmenting existing factor inputs, by making more effective use of existing inputs, and by contributing to the development and application of new technologies. In particular, the growing body of microeconomics research on the productivity of schooling indicates that rapid technological change is facilitated when workers have completed more schooling and that the benefits of R&D activities are enhanced when they are complemented by a well-educated labor force. For instance, Bartel and Lichtenberg (1987) report that, in U.S. manufacturing industries in 1960, 1970, and 1980, relatively more well-educated workers were employed in those industries where the (physical) capital equipment was new and that this association was especially strong in industries where R&D expenditures represented a larger fraction of total sales. 17 They argued that highly educated workers are better at adapting to and assimilating new technologies so that the relative demand for well-educated labor is greater in industries with high rates of technological change. This greater demand is consistent with the fact that the relative wages of welleducated workers are greater in industries exhibiting rapid technological change including those undertaking larger R&D expenditures. 18 Though there is a widespread belief that the particular contribution of schooling is to facilitate technological change, less is known about the particular skills acquired at school that are most valuable in work performance. There is some suggestion that it is the mathematical skills of workers that are associated most with workers' productivity (Bishop 1988), perhaps because such skills are most relevant to the problem-solving


65

situations in some work environments. However, at present, much less seems to be known about what types of skills developed at school contribute most to labor productivity on-the-job. Schooling and Earnings

The College Earnings Premium

If schooling contributes to workers' productivity, as argued above, then in an economy where there is room for prices to respond to the demand for productive inputs, at least some of the returns to schooling should be expected to be enjoyed by those acquiring an education. Some of these returns will be monetary and, hence, an extensive literature now exists examining the association between schooling and income and, in particular, the gain in earnings from attending college. The association between schooling and earnings may be decomposed into two parts: first, once they have completed their formal education, more schooled individuals have a higher propensity of being in the labor force at any moment and, among those in the labor force, of working longer hours; 19 and second, holding hours of market work constant, those with more schooling are paid higher wages. The particular form of this earnings-schooling association for U.S. men and women in 1989 is described in table 3-3. The last column shows that at all ages college graduates (those with sixteen years of schooling) earned 62 percent more (if male) or 64 percent more (if female) than high school graduates (those with twelve years of schooling). The magnitude of this premium for attending college varied with age ranging from a low of 46 percent for women aged fifty-five to sixty-four years to 83 percent for women aged twenty-five to thirty-four years. Clearly, earnings differences at any age understate the lifetime earnings differences among schooling groups. The association between age and earnings for any schooling level arises because individuals augment their knowledge acquired from school by accumulating skills on-the-job so that the age-earnings association is better understood as an association between earnings and work experience. There is, indeed, an independent role for age (independent, that is, of experience) because as an individual ages so his physiological and psychological characteristics alter and these may exert an effect on earnings. However, in most surveys, information on both an individual's experience and his age are not collected so that their separate effects cannot be distinguished. Cross-section surveys of individuals consistently reveal age-earnings

66


Table 3-3. Mean Earnings (in Thousands of Dollars) of Women and Men in 1989 by Years of School and by Age

25-34

Years of Schooling 12 12.16 16 22.25 17 or more 24.40 Relative Earnings Y16/Y12 1.83 Y17/Y12 2.01

25-34

Years of Schooling 12 21.33 16 32.47 17 or more 35.48 Relative Earnings Ylf,/ Y12 1.52 Yn/Y12 1.66

35-44

Women Years of Age 45-54 55-64

65+

All Ages

14.30 22.55 29.93

15.21 22.57 30.43

13.55 19.84 25.86

8.42 13.94 13.90

13.44 22.02 27.47

1.58 2.09

1.48 2.00

1.46 1.91

1.66 1.65

1.64 2.04

35-44

Men Years of Age 45-54 55-64

65+

All Ages

26.36 42.08 55.25

30.59 52.95 60.28

26.59 47.68 54.43

16.89 24.78 37.02

24.92 40.39 50.66

1.60 2.10

1.73 1.97

1.79 2.05

1.47 2.19

1.62 2.03

Notes: Y16 / Y12 denotes the earnings of college graduates (those with sixteen years of schooling) as a fraction of the earnings of high school graduates (those with twelve years of schooling). Y17 /Y12 denotes the earnings of those with five or more years of college as a fraction of the earnings of high school graduates. The data describe men and women twentyfive years and over as of March 1990 with positive earnings in 1989. From table 29 of Current Population Reports, Money Income of Households, Families, and Persons in the United States: 1988 and 1989, Consumer Income, Series P-60, No. 172, Department of Commerce, Bureau of the Census.

profiles by levels of schooling with three broad features. First, earnings are concave with respect to age, that is, for each schooling level, earnings rise with age at a decreasing rate up to a maximum and then flatten out or decline. Second, the greater the schooling, the steeper the rate of increase in earnings with age. Third, the peak in earnings tends to be reached later in age for those with more schooling. These are robust empirical regularities found in very many cross-section surveys of individuals. Of course, the age-earnings profile described in any calendar year is


67

conceptually different from the path that any cohort's earnings will follow as that cohort ages, Indeed, it is not difficult to rationalize the general features of the age-earnings profiles reported in table 3-3 by combining assumptions about the change over time in the earnings of different cohorts with linear (not inverted U-shaped) age-earnings profiles. 20 The female earnings-schooling profiles are not the same as those for men. At any schooling level, the increase in earnings with age is smaller for women than for men: the ratio of peak earnings to earnings at age twenty-five to thirty-four years for men is 1.43 for high school graduates and 1. 70 for those with seventeen or more years of schooling; for women, the corresponding ratios are both 1.25. Women in any given age group tend to reveal a greater variation in labor market histories than men so that some of these women at work in 1989 will not have been at work for pay in each year since the completion of their schooling. On average, they have accumulated less work-related experience 21 so that, if the critical variable affecting earnings is experience, then we should not expect the earnings-age profiles for women to be as strongly inclined as those for men. For men, the college earnings premium tends to form an inverted U shape, being highest for those fifty-five to sixty-four years of age. For women, it is U-shaped being highest for young women. In other words, recent cohorts of female college graduates have been earning a larger earnings premium than earlier cohorts. However, the earnings premium for women with graduate training (given in the row Y17 /Y12 in table 3-3) tends to display little variation between the ages of twenty-five to sixtyfour years. Those who believe wage differentials change sluggishly, if at all, will be surprised by the variation in the college earnings premium between 1940 and 1988 as shown in table 3-4 for white men and women. The estimates are drawn from the Decennial Censuses of 1940, 1950, 1960, 1970, and 1980 and then supplemented with the 1988 Current Population Survey. The entries in table 3-4 report the proportionate difference in weekly earnings between college graduates and high-school graduates arranged by years of estimated labor market experience. The college earnings premium ranged from 30 percent for men in 1950 to 81 percent for women in 1940 with the 1970s being largely a period of decline in this premium and the 1980s a period of growth. The premium tends to be most volatile for recent entrants to the labor market (especially those with one to five years of experience) and least volatile for those in mid-career. 22 The natural explanation for the relatively greater volatility of the

68


Table 3-4. Proportionate Weekly Earnings Differential between College and High School Graduates: 1940 to 1988. years of experience

1940

1950

Women 1960 1970

1980

1988

1-5 6-10 11-15 16-20 21-25 26-30 31-35 36 & more All

0.98 0.68 0.69 0.64 0.54 0.98 1.04 1.14 0.81

0.36 0.31 0.38 0.31 0.56 0.40 0.42 0.23 0.42

0.59 0.47 0.33 0.68 0.90 0.81 0.77 0.74 0.66

0.74 0.68 0.72 0.73 0.81 0.84 0.80 0.82 0.75

0.58 0.58 0.56 0.56 0.69 0.59 0.65 0.61 0.62

0.88 0.72 0.75 0.65 0.72 0.40 0.40 0.54 0.75

years of experience

1940

1950

1960

1970

1980

1988

1-5 6-10 11-15 16-20 21-25 26-30 31-35 36 & more All

0.77 0.56 0.60 0.55 0.54 0.33 0.36 0.18 0.56

0.40 0.30 0.38 0.44 0.19 0.10 0.27 0.39 0.30

0.49 0.42 0.44 0.44 0.47 0.45 0.49 0.30 0.45

0.64 0.48 0.51 0.53 0.53 0.53 0.41 0.47 0.52

0.38 0.46 0.51 0.57 0.49 0.58 0.43 0.33 0.49

0.80 0.67 0.57 0.61 0.54 0.48 0.42 0.39 0.64

Men

Note: These estimates are taken from Lydon (1990). Someone with sixteen or more years of schooling is classified as a college graduate and an individual with exactly twelve years of schooling is classified as a high school graduate. Separate least-squares regression equations were fitted to individuals within each of the calendar year-experience cells in this table. The dependent variable is the logarithm of weekly wage and salary income. The independent variables are three schooling dummy variables (one for zero to eleven years of schooling, another for thirteen to fifteen years of schooling, and one for sixteen or more years of schooling) and dummy variables for various years of experience. The figures in the table are the antilogarithms of the estimated coefficients on the dummy variable denoting college graduates minus unity. Actual years of market work experience are not known and are estimated by subtracting estimated age upon leaving school (defined as years of schooling plus six) from current age.

college earnings premium of recent labor market entrants has to do with the type of skills they possess compared with those individuals in midcareer. The skills of recent entrants to the labor market are predominantly those acquired at school and such skills are usually valuable in many


69

different activities and jobs. In other words, the human capital produced by schooling is general and is transferable across firms and jobs. Hence, with many firms competing for such skills and many individuals embodying these skills, a competitive market arises with wages fluctuating according to movements in the demand for and supply of general human capital. By contrast, individuals who have worked for a firm or in a particular job within a firm for a number of years are likely to have accumulated specific human capital, skills specific to that firm or to that job within the firm. These skills are of value when applied in that firm or in that job only: the firm cannot replace the specifically skilled employee with an equally productive new hire while the employee's value in another firm (and, therefore, the wages he can command elsewhere) is lower than that in his present firm. With specific skills, the employee and the employer are each monopolists, one the sole seller of an asset and the other the sole buyer, and the bilateral monopoly setting implies that (implicitly or explicitly) the wage is determined by bargaining between the two parties. The wage cannot fall below the value of the worker's general human capital or else the worker will be inclined to move to another firm that will pay him the value of his general skills; and the wage cannot be so high that the firm would rather hire and train a new worker. But, between these lower and upper limits, the wage will be set by bargaining between two monopolists. Workers typically embody a mix of general and specific human capital, but that mix changes with labor market experience. Recent entrants to the labor market have had little opportunity to acquire specific skills so most of their human capital is general and, therefore, subject to the influences of competitive demand and supply. Individuals with more work experience tend to have accumulated more specific human capital, especially if they have worked for a single firm for a long time. Their wages tend to be the product of bargaining that is sometimes filtered through institutional and administrative rules. Prices and wages set in monopolistic markets tend to show more inertia than prices and wages set in competitive markets. The greater volatility of relative wages among recent labor market entrants compared with more experienced workers reflects the mix of general and specific human capital embodied in these workers and the different types of labor markets in which they operate. The data in table 3-4 indicate a decline in the earnings premium from attending college in the 1970s especially for new labor market entrants and a sharp rise in the 1980s. The reasons for this are not evident. At one time, a popular explanation was that the 1970s witnessed an increase in the number of new labor market entrants, a consequence of the babyboom some twenty years earlier. This increase in the supply of inexperi-

70


enced workers occurred at all schooling levels, but it was more pronounced among college graduates than high-school graduates. Moreover, it was argued (Welch 1979) that the relative value of experience is greater in the market for college graduates than for high-school graduates so that the increase in inexperienced workers caused by the baby-boom cohorts entering the labor market disproportionately depressed the earnings of college graduates. Though the largest of the baby-boom cohorts passed into the older adult age groups in the 1980s, the behavior of the earnings differentials of younger workers in more recent years seems out of proportion to declines in the size of their cohorts so other factors are almost certainly at work. For example, O'Neill and Sepielli (1985) suggest that distinct shifts in the demand for occupations requiring college training have occurred to raise the return to higher education while Murphy and Welch (1989) hypothesize that the loss of manufacturing jobs within the United States has lowered the relative wages of high school graduates. Perhaps the most noteworthy feature of the time-series movements in the college earnings premium in table 3-4 is the absence of any downward trend. It is true that the premium tended to fall in the 1970s, but even in 1980 it exceeded its value in 1950 and since 1980 it has risen considerably. The absence of a trend deserves note because the past fifty years has seen a tremendous growth in the relative number of college educated people: in 1940, of those over twenty-five years of age, only 5.5 percent of men and 3.8 percent of women had completed four years of college or more; by 1987, the numbers were 23.6 percent and 16.5 percent respectively (see table 3-5). Why has the college earnings premium failed to fall in the face of such a substantial increase in the relative numbers of college educated labor? Several explanations have been proposed. One maintains that when new technologies are introduced, physical capital tends to replace the work of unskilled labor; machines are less substitutable for well-educated labor. Hence reductions in the prices of physical capital induce larger declines in the demand for unskilled labor than for well-educated labor. Indeed, the demand for well-educated labor may increase together with the demand for physical capital if the price of capital should fall. As the economy has accumulated more physical capital over time, this argument proceeds, the demand for less-educated labor has fallen relative to the demand for college-educated labor. In other words, the relative demand for college-educated labor has increased in tandem with the increased relative supply of such labor thereby forestalling a reduction in the college earnings premium. In fact, the evidence for this view of the way unskilled labor, skilled labor, and physical capital combine in production is mixed. 23 Other explanations have been put forward such as a relative growth in the

--.)

-

Percent of men completing 4 years of high school or more Percent of women completing 4 years of high school or more Percent of men completing 4 years of college or more Percent of women completing 4 years of college or more Percent of male college graduates to male high school graduates Ratio of female college graduates to female high school graduates Median school years completed, men Median school years completed, women 8.7 9.6

9.0

0.22

0.23 8.6

0.40

5.2

7.3

36.0

32.6

1950

0.45

3.8

5.5

26.3

22.7

1940

10.9

10.3

0.21

0.45

5.8

9.6

42.5

39.4

1960

12.1

12.2

2.22

0.47

8.2

14.1

55.4

55.0

1970

12.4

12.6

0.34

0.64

13.5

20.8

68.1

69.1

1980

Schooling Completed by the Population Aged Twenty-five Years and Over, 1940 to 1987

12.6

12.7

0.39

0.67

16.0

23.1

73.5

74.4

1985

12.6

12.7

0.40

0.67

16.5

23.6

75.3

76.0

1987

Notes: These data are taken from Tables 11 and 12 of Bureau of the Census, Educational Attainment in the United States: March 1987 and 1986, Current Population Reports, Population Characteristics, Series P-20, No. 428. In lines five and six, "college graduates" means those whose highest educational attainment is four years or more of college while "high school graduates" means those whose highest educational attainment is four years of high school.

8.

7.

'6.

5.

4.

3.

2.

1.

Table 3-5.

72


demand for goods and services produced by college-educated labor, which induces a greater demand for such workers, but in general, it has proved difficult to discriminate among alternative hypotheses. Other Aspects of Higher Education and Earnings

To this point, all the research described relates to the association between earnings and years of schooling completed and, in particular, to the earnings premium associated with completion of a college education. In addition to this, there is evidence of an association between earnings and the type of college education obtained. 24 There are two types of variables that have been examined: one set of variables relates to the institutional characteristics of the college or university, such as the college's instructional expenditures per student and whether it is a public or private school; a second set of variables describes the particular features of each individual's education, such as the identify of one's major or the extent of one's mathematical training. Among the institutional characteristics of the college, it seems as if those individuals earn more who graduate from colleges whose entering freshmen score higher on college admission tests, but many other institutional features (such as spending or whether the college is primarily a research institution) are weakly correlated (if at all) with labor market earnings. By comparison, particular aspects of students' college education appear to account for a larger fraction of the variation in earnings among individuals. For instance, other things equal, majors in natural science and engineering (and sometimes undergraduate business majors) earn more than humanities students and those with an education major. Individuals with higher undergraduate grades also earn more. 25 Usually, these relationships are measured holding constant a number of other variables associated with earnings including socioeconomic background variables such as parental income, race, religion, and father's occupation. Nevertheless, it is hazardous to move from these associations and draw inferences regarding cause and effect. Individuals choose the college they attend and they choose the type of education they receive. The association between the earnings of these individuals on the one hand and the characteristics of their colleges and of their education on the other hand may reflect, not the effect of these variables on earnings, but the effect of other (perhaps attitudinal) variables that affect both the educational experience chosen by students and their subsequent earnings. In general, unscrambling causal relations from the accumulation of empirical associations is a difficult and somewhat neglected task.


73

The Supply of College-educated Labor College Enrollments

The previous section has documented a clear earnings premium to a college education. However, this premium has not been constant over time. The movements in the college premium imply corresponding movements in the incentive to acquire a college education and should induce variation in college enrollments. Indeed, table 3-6 presents evidence that college enrollments have fluctuated in response to movements in the college premium. The data on college enrollments in the late 1960s and first part of the 1970s are affected by the incentives to attend college as a means of avoiding the military draft for the Vietnam War. But these effects are largely over by the mid-1970s at which time enrollment rates in higher education fell and then rose in line with the growth in the college earnings premium in the 1980s. They are especially high in 1988. Enrollments in graduate education tend to lag undergraduate enrollments by a few years. Even if students were guided solely by monetary gains and losses, the college earnings premium is only one component of those net returns. Factors such as tuition and scholarships should also affect enrollment and existing research indicates that they do. Thus McPherson (1978) surveys ten studies of the effect of tuition on enrollments and "every single one finds a significant negative relationship between the net price faced by students and their probability of attending college" (p. 180). He estimates that, on average, a 10 percent increase in tuition reduces enrollments by three percent. This effect relates to a change in tuition for all colleges simultaneously; a much larger increase would be expected if tuition were raised at one college with tuition at all other colleges unchanged. McPherson conjectures that it is students from low income families who are most price (tuition) sensitive. In a thorough analysis of almost 23,000 seniors from the National Longitudinal Study (NLS) of the High School Class of 1972, Manski and Wise (1983) found students to be very responsive to tuition, scholarship, and alternative employment opportunities26 in deciding whether to go to college. They reported "the students who did attend college were the most likely to benefit from college education by obtaining a degree" (p. 159). The same comparative advantage finding emerged from Willis and Rosen's (1979) study of 3,611 male World War II Veterans. They concluded, "If we examine a subpopulation of persons with given measured abilities ... the empirical results on selectivity imply that those persons

-...J .+:;.

25.5 27.3 25.7 25.5 24.6 26.7 25.3 25.6 26.6 27.1 27.8 30.3

As a Percent of 18-24 year olds 33.7 35.0 32.7 31.9 30.5 33.1 31.4 31.6 33.0 33.2 33.7 37.3

As a Percent of High School Graduates

Enrollments in All Institutions of 18-24 Year Olds

Enrollment Rates in Institutions of Higher Education, 1967 to 1988

0.57 0.57 0.63 0.67 0.64 0.65 0.68 0.66 0.65 0.64 0.65

Enrollments in First-Professional Degree Programs as a Percent of 20-29 Year Olds

3.31 3.42 3.23 3.38 3.51 3.34 3.27 3.12 3.14 3.22 3.58

Enrollments in Graduate Degree Programs as a Percent of 20-29 Year Olds

Notes: The data for 1988 arc preliminary. These data arc taken from Tables 14, 170, 172, and 173 of National Center for Education Statistics, Digest of Education Statistics 1990, U.S. Department of Education, Office of Educational Research and Improvement, NCES91-660.

1967 1969 1970 1972 1974 1976 1978 1980 1982 1984 1985 1988

Table 3-6.


75

who stopped schooling after high school had better prospects as high school graduates than the average member of that subpopulation and that those who continued on to college also had better prospects there than the average member of the subpopulation" (p. 28)_27 Monetary factors are not the only ones determining college enrollments; for instance, scholastic achievement at high school and parental schooling levels also affect the decision to go to college. 28 However, the sensitivity of this decision to monetary considerations is marked: Willis and Rosen estimated that a 10 percent increase in starting salaries induced almost a 20 percent increase in college enrollments. 29 In general, college enrollments respond to the pecuniary net returns from investing in higher education. Professional Workers

The previous paragraphs have presented some evidence in support of the principle that, other things equal, movements in the monetary costs of and returns from attending college induce movements in college enrollments. This same idea lies behind research examining the flow of college students into different fields of study. The monetary costs and benefits represent the most visible part of these net returns so it is natural for researchers to look for associations between these monetary factors and the relative popularity of particular fields. Usually research has sought to explain the time-series movements in the relative number of graduates such as those in table 3-7 which presents the percent of all master's, first-professional, and doctor's degrees conferred in the major fields of study between academic years 1949 and 1985. The data in this table show that, while the popularity of some fields over time follows strong trends (upward trends in the case of Business and of Computer and Information Sciences and downward trends in the case of Other Social Sciences), in other fields cycles are evident: the relative number of degrees conferred in Education rose up to the mid-1950s, fell to the mid-1960s, rose again to the mid-1970s, and fell again in the mid1980s; the relative number of degrees in Law and the relative number of M.D. degrees fell from the mid-1950s to the early 1970s and have grown since then; Engineering degrees increased to the mid-1960s, then fell to a low point a decade later, and have been rising again recently. Essentially there are two ways to account for these fluctuations in degrees awarded and, in particular, for the suggestion that, in some cases, a cyclical pattern obtains. The first explanation looks for fluctuations in the independent variables determining the relative costs and benefits of

-.J 0\

2.8

1.3

3.1

5.0 3.2 0.9 2.7

1.3 1.9 4.9 3.2 0.9

1.3

23.6 5.5

6.6 8.7 33.0 4.6 7.5

7.9 9.6 37.0 4.0 6.2

6.3

3.7

2.9 4.2 3.3 1.0

4.0 7.0 28.2 7.1 8.4

1965-66

3.0 5.0 31.7 9.1 6.7 0.6 2.1 3.6 3.1 0.9 5.8 2.0 3.5

1970-71

3.3 7.9 33.3 10.6 4.7 0.7 1.2 2.2 2.4 0.7 4.2 3.1 2.1

1975-76

1.5

0.8 2.1 2.4 0.7 3.1 4.2

1.1

3.9 9.1 26.7 14.7 4.8

1980-81

4.0 9.0 21.1 17.2 6.3 2.1 1.0 2.4 2.1 0.7 2.1 5.0 1.5

1985-86

Notes: Health Professions includes hospital and health care administration, nursing, dental specialties, occupational therapy. pharmacy, dental hygiene, public health, veterinary medicine specialties, speech pathology, and medical laboratory technologies. Life Sciences includes anatomy, biochemistry, bacteriology, biology, botany, entomology, physiology, and zoology. Mathematics includes statistics. Physical sciences includes astronomy, chemistry, geology, metallurgy, meteorology, and physics. Other Social Sciences includes anthropology, archeology, history, geography, political science and government, sociology, demography, urban studies, and criminology. The fields listed in the table above represented three-quarters of all master's, first-professional, and doctor's degrees awarded in 1985-86. Fields not listed above include agriculture and natural resources, foreign languages, philosophy and theology, psychology, public affairs, and the visual and performing arts. Before 1960-61, first-professional degrees are not distinguished in the data from bachelor's degrees. For the years 1949-50, 1955-56, and 1959-60, therefore, I estimated the total number of first-professional degrees by assuming they represented the same fraction of the total (bachelor's and first-professional degrees) as in 1960-61 (which was 0.0569). In general, in the earlier years, the information on first-professional degrees is less accurate than that on master's and doctor's degrees. The data arc taken from issues of Digest of Education Statistics, National Center for Education Statistics, U.S. Department of Education and especially from the 1988 issue (CS88-600).

Medicine (M.D.) Law (LL.B or J.D) Education Business & Management Engineering Computer & Information Sciences Mathematics Physical Sciences Life Sciences Economics Other Social Sciences Health Professions English & Literature

1959-60

1955-56

1949-50

Table 3-7. Master's, First-Professional, and Doctor's Degrees Awarded by Field as a Percent of All Such Degrees, between 194950 and 1985-86.


77

spc,cializing in different fields. For instance, we would expect the incentive to acquire an Education degree as a step toward becoming a primary or secondary school teacher to be related to the projected demand for teachers as derived from the number of school aged children. These projections can be estimated from current fertility rates and demographic profiles. Swings in fertility rates may generate marked movements in the supply of teachers. 30 A second class of explanations for the movements in relative numbers of degrees conferred focuses on the internal dynamics of the market structure itself, noting the lag between the date individuals enroll in a college program and the date they graduate and enter the labor market. This lag implies that the costs of and benefits from specializing in different fields are uncertain so that individuals entering a college program must estimate the labor market conditions facing them in the future. When individuals estimate future salaries on the basis of earnings at the time of their enrollment in college, oscillations in earnings and in graduates may result along the lines of a classic cobweb model. 31 Current graduates do not constitute the only source of new supply to a profession. In fact, in a number of cases, a substantial fraction of new entrants consists of individuals returning to the profession after an interruption in their career or entering the profession for the first time after having obtained their qualifications some years previously. This feature of professional labor markets has probably become increasingly important over time as women have made more determined efforts to reenter the labor market after a period of withdrawal from the labor force while assuming primary responsibility for the care of their young children. Thus, 84 percent of new elementary and public school teachers in 1986 consisted of individuals who in the previous year were not attending college nor teaching. In 1966, this percentage was 33?2 In the case of teachers, it appears as if an increase in salaries postpones the date teachers drop out (perhaps temporarily) of teaching. In particular, a $1,000 increase in annual salaries (in 1987 dollars) extends the first spell of teaching by an average of over a year. Those teachers who score higher on the National Teachers Examination (NTE), a general and professional knowledge test, have shorter first spells as teachers reflecting perhaps that such teachers had more attractive alternative opportunities than teachers scoring lower on the NTE. In general, movements in the supply of professional manpower operate not only through the production of new graduates, but also through changes in the rate of attrition of trained personnel and in the rate of reentry into the profession, factors that are rarely accounted for in conventional projections of manpower requirements.

78


Conclusions At least two types of evidence supports the proposition that rising schooling completion levels and movements in the quality of schooling have contributed to the growth in U.S. labor productivity. First, there exists a correlation across countries between economic growth rates and schoolenrollment rates including enrollment in higher education. Second, those industries in the United States whose technical progress has been most rapid are those industries making most use of well-educated labor. This is compatible with the notion that workers with more schooling are more likely to exploit and adjust to new technologies. Well-educated labor is steered to those firms and industries exhibiting a rapid pace of technological change by means of earnings premia, that is, they receive higher wages than they could earn elsewhere. These earnings premia reflect, at least in part, their differentially higher productivity. The fact that these earnings premia have grown sharply in the 1980s suggests the past decade has seen a rise in the productivity of well-educated workers relative to the productivity of less educated workers. These changes in the relative productivity of college educated labor are probably derived in part from developments in product markets (such as the shift away from the output of goods producing industries in the United States) and in part from changes in the type of complementary capital employed (such as the spread of computerized technology). However, the mechanisms through which changes in product markets and in types of capital have affected the college earnings premium are not at well understood. There is good reason to expect these earnings premia will call forth additional supplies of college-educated labor. Existing evidence suggests the supply of new graduates, the rate of attrition from a profession, and the rate of reentry into a profession all respond to earnings differentials and to other monetary incentives. The magnitude of these responses remains highly conjectural, which is why forecasts of the supply of collegeeducated labor are shrouded in so much uncertainty. And here we come full circle: if the supply of well-educated labor rises at a rapid rate, the pace of economic growth is facilitated. In this way, the demand for and supply of college-educated workers, their pay, and the rate of economic growth are variables that are closely interwined. Unravelling these connections is an important task for future research.


79

Acknowledgments

This chapter is a thorough revision of an article published in the Journal of Economic Education. (Pencavel1991). The previous version benefitted from the comments of Moses Abramovitz, Richard Anderson, Kevin Coleman, Hirschel Kasper, Mary Lydon, William Massy, Carol Rapaport, and two anonymous referees. Notes 1. For instance, screening models maintain that schooling completion levels represent indicators of innately more productive individuals who remain at school longer to acquire signals rather than to acquire skills. 2. The statement quoted is from Haveman and Wolfe (1984) summarizing Denison's work. 3. In other words (Exd 1 = [(aXIaL) (LIX)], and (EXK), = [(aX/aK) (K!X)lt. 4. Cost minimization is not sufficient. Under cost minimization, ExL equals wL!mX where m is the marginal cost of output. 5. In some formulations, su and sK, are approximated by a moving average of this year's and last year's factor shares while !'J.€nL 1 and !'J.€nK, are measured by annual rates of change of labor and capital. This is the Tornqvist approximation to the Divisia input index. 6. So there is no confusion, take the United States data as an example. Maddison's educational index in 1913 is formed (using the data in the first line of table 3-1) by computing 4.90 + (1.4) (1.83) + (2.0) (0.20) which equals 7.86. The corresponding value for the United States in 1950 is 11.27. Hence, the annual average compound growth rate is (11.27 /7 .86)< 1137)-1 which equals 0.978 percent. Maddison arques that only some portion of relative wage differentials reflect the relative productivity of educational groups, the other portion being attributed to factors such as natural ability and family connections so he deflates the growth rate in his educational index by 0.6. This yields for the United States for 1913 to 1950 a rate of 0.59 percent ( = 0.6 x 0.978). This is entered into line one of table 3-2. 7. To illustrate how this was computed, again take the United States in 1913 and 1950. In 1913, the value of the educational index excluding higher education is 4. 90 + (1.4) ( 1.83) = 7.46. In 1950, it is 10.37. Hence, the annual average compound growth rate of this index that excludes higher education is (10.37/7.46)(1 137 ) -1 = 0.893 percent. This contrasts with the growth rate of the index that includes higher education of 0.978 percent (as computed in the previous endnote). The percentage contribution of higher education to the growth in the educational index in the United States from 1913 to 1950 is ((0.978-0.893)/(0.978)) 100 = 8.69, which is entered in line two of table 3-2. 8. Maddison set sL equal to 0.7 which broadly corresponds to the share of labor income in total national income in these countries over this period. 9. This is computed by multiplying schooling's contribution to economic growth in line four of table 3-2 by the proportional contribution of higher education to the growth in the educational index, as given by one hundredth of line two of table 3-2. Thus, for the United States from 1913 to 1950, the percentage of the growth of output contributed by higher

80


education is 14.78 x 0.087 = 1.29 percent. This definition of higher education's contribution to economic growth is mine; Maddison did not venture to isolate higher education. 10. The most recent evidence is provided by Card and Krueger (1990). They use the 1980 U.S. Census of Population on white and black men to relate the variations in the slope of the earnings-years of schooling relationship across states and across three cohorts to indicators of schooling quality in these states at the time these people were at school. The schooling quality indicators arc the pupil-teacher ratio, the length of school terms, and the level of teacher salaries relative to average wages in the state. For the most part, these indicators of school quality are associated with earnings in the expected manner and in some specifications the responsiveness of earnings to these quality indicators is sizeable. 11. See, for instance, the studies and discussion in Akin and Garfinkel (1977), Behrman and Birdsall (1983), Bowles (1970), Hanushek (1986), and Solmon (1985). 12. For a full statement of the implications for growth accounting of education being an intermediate input and for a producer surplus measure of education's net productivity, sec Plant and Welch (1984). 13. In the results reported below, Zi consists of the following variables: the average ratio of government consumption to real GDP; the number of revolutions and coups per year; the number per million population of political assassinations per year; and a rough measure of price distortions within a country. Each of these three variables is negatively correlated with !'J.fnXi. 14. Baumol, Blackman, and Wolff appear to place more emphasis on the fact that the t-statistics on the coefficients attached to higher education's enrollment rates arc lower than those on the coefficients associated with secondary education's enrollment rates. This induces them to conclude "that education does matter a good deal for a nation's economic growth, and that what matters is the share of the population with secondary education" (p. 206). In fact, an examination of their results indicates that a given increase in enrollment rates at higher education is associated with a higher economic growth than the same increase in secondary school enrollment rates. Their statements about statistical significance are being confused with statements about the magnitude of the estimated regression coefficients. 15. Note that greater allocative efficiency is not synonymous with increased productivity. Define productivity as output per unit of input(s) and, for illustration, consider a simple case of one output and one input. Suppose the price of this input is relatively high. Then allocativc efficiency will call for using that input sparingly although (if the production function displays increasing returns to scale) output per unit of input (productivity) may well be higher when more of that input is used. 16. The importance of schooling in achieving allocative efficiency has been stressed by Jamison and Lau (1982), Nelson and Phelps (1966), Welch (1970), and Schultz (1975). 17. These relationships held only for younger workers (aged not more than forty-five years) suggesting, perhaps, that the role of education in fostering technological change depreciates as an individual ages. 18. Sec Bartel and Lichtenberg (1988), Dickens and Katz (1987), and Mincer (1989). 19. For instance, in the 1960 Census of Population, holding constant-race, marital status, age, and nonwage income-college educated men had almost two percent higher probability of being in the labor force than high school graduates. For single women, this differential was two and one-half percent, and for married women with spouse present, it was nine percent (where account was also taken of the presence of children and the employment status of the husband). These estimates are from Bowen and Finegan (1969). In the 1980 Census, college educated men worked about eighty-six more hours per year than high school graduates after controlling for a very large number of personal characteristics of these men. See Pencavcl ( 1986). For eVIdence that schooling and unemployment arc negatively


81

correlated, see Ashenfelter and Ham (1979), Nickell (1979), and Mincer (1991). 20. See, for example, Shorrocks (1975). His particular argument relates to the agewealth profile, but it applies equally to the age-earnings relationship. 21. Not only has the average woman spent less of her post-school time at market work, there are discontinuities in her employment history. For evidence that such patterns of work history and labor force attachment are associated with earnings, see Blau and Ferber (1986), Corcoran (1978), Mincer and Ofck (1982), O'Neill (1985), and Salvo and McNeil (1984). 22. The college earnings premia in table 3-4 arc sample point estimates, yet the associated standard errors have not been presented. This means that the reader is advised to look for common variations in these premia and not to attach importance to small or apparently idiosyncratic variations. There seems more volatility across calendar time and across groups with different labor market experience in the college earnings premium for women. At least part of this is probably due to systematic problems in the procedure for controlling for work experience and the true variation is less than this. The estimated standard errors of the female college earnings premia are always greater than those of the male. 23. Hamermesh (1986) describes as "fairly solid" the evidence in favor of the notion that skilled labor and physical capital tend to be complementary with one another (not substitutable) in production. However one of the most careful studies in this literature, that of Weiss (1977), finds little evidence to support it. In a study of a large Canadian telecommunications firm, Denny and Fuss (1983) find that technical change between 1952 and 1972 was capital-using and labor-saving with the least skilled occupations experiencing drastic reductions in demand. 24. Sec, for instance, Daymont and Andrisani (1984), James, Alsalam, Conaty, and To (1989), Ransom (1991), Solmon (1981), Weisbrod and Karpoff (1968), Wachtel (1975), and Wise (1975). 25. Evidence is presented in James, et al. (1989), Solmon (1981), Weisbrod and Karpoff (1968), and Wise (1975). 26. "The primary alternative to postsecondary education for most youth is full-time participation in the labor force. If wage rates are high and unemployment rates are low, expected earnings of people in the labor force are higher, and this will tend to increase the advantages of working versus going to school, other things being equal. Our estimates are consistent with this reasoning: the greater the expected foregone earnings while in school, the Jess likely a person will be to attend school and the more likely to get a full-time job" (Manski and Wise 1983, pp. 19-20). 27. Willis and Rosen also express the point as follows: "plumbers [high-school graduates) may have limited potential as highly schooled lawyers, but by the same token lawyers may have much lower potential as plumbers than those who actually end up choosing that kind of work" (p. 11). 28. One study demonstrating this is Marc (1980). 29. This estimate is calculated at the observed proportion going to college in this sample. By way of comparison, an increase in father's schooling of 1.59 years (this being the difference in average years of father's schooling between the group who attended college and the group who did not) would have increased enrollments by 3.4 percent. 30. A model with these features has been estimated with mixed success by Zarkin (1985) to describe the market for public school teachers. Assuming individuals form unbiased (or "rational") expectations of future earnings, he finds the number of newly certified secondary school teachers in year t to be significantly associated with the number of children in school in years from 5 to t+5. This result did not obtain for elementary school teachers, however. 31. This is exactly how Freeman (1971) proceeds in analyzing occupational choice.

82


Survey evidence of students' expectations about earnings is contained in McMahon and Wagner (1981) who suggest average expectations are quite close to average earnings realizations. 32. This information on teachers is drawn from Murnane and Olsen (1989a, 1989b).

References Akin, J.S., and Garfinkel, I. (1977). School expenditures and the economic returns to schooling. Journal of Human Resources, 12(4), Fall, 460-481. Ashenfelter, 0., and Ham, J. (1979). Education, unemployment, and earnings. Journal of Political Economy, Supplement, 87(5), Part 2, October, 99-116. Barro, R.J. (1991). Economic growth in a cross-section of countries. Quarterly Journal of Economics, 106(2), May, 407-443. Bartel, A.P., and Lichtenberg, F. R. (1987). The comparative advantage of educated workers in implementing new technology. Review of Economics and Statistics, 69(1), February, 1-11. Bartel, A.P., and Lichtenberg, F.R. (1988). Technical change, learning, and wages. Working Paper No. 2732, October. National Bureau of Economic Research. Baumol, W.J., Blackman, S.A.B., and Wolff, E.N. (1989). Productivity and American leadership: The long view. Cambridge: MIT Press. Behrman, J.R., and Birdsall, N. (1983). The quality of schooling: Quantity alone is misleading. American Economic Review, 73(5), December, 928-946. Bishop, J. (1988). Why high school students learn so little and what can be done about it. Working Paper 88-01. Center for Advanced Human Resource Studies, New York State School of Industrial and Labor Relations, Cornell University. Blau, F.D., and Ferber, M.A. (1986). The economics of women, men, and work. Englewood Cliffs. Prentice-Hall. Bowen, W.G., and Finegan, T.A. (1969). The economics of labor force participation. Princeton University Press. Bowles, S. (1970). Towards an educational production function. In W. Lee Hansen, (Ed.), Education, income, and human capital, National Bureau of Economic Research, Columbia University Press, pp. 11-61. Card, D., and Krueger, A. (1990). Does school quality matter? Returns to education and the characteristics of public schools in the United States. Working Paper, No. 3358, May. National Bureau of Economic Research. Corcoran, M. (1978). The structure of female wages. American Economic Review, Papers and Proceedings, 68(2), May, 165-170. Daymont, T.N., and Andrisani, P.J. (1984). Job preferences, college major, and the gender gap in earnings. Journal of Human Resources, 19(3), Summer, 408-428. Denison, E. (1979). Accounting for slower economic growth: The United States in the 1970s. Washington, D.C.: Brookings Institution. Denny, M., and Fuss, M. (1983). The effect of factor prices and technological change on the occupational demand for labor: Evidence from Canadian tele-


83

communications. Journal of Human Resources, 18(2), Spring, 161-176. Dickens, W.T., and Katz, L.F. (1987). Interindustry wage differences and industry characteristics. In K. Lang and J.S. Leonard (Eds.), Unemployment and the structure of labor markets, New York: Basil Blackwell, pp. 48-89. Freeman, R.B. (1971). The market for college-trained manpower. Cambridge: Harvard University Press. Hamermesh, D.S. (1986). The demand for labor in the long run. In 0. Ashenfelter and R. Layard (Eds.), Handbook of labor economics Volume 1, New York: North-Holland, pp. 429-471. Hanushek, E.A. (1986). The economics of schooling: Production and efficiency in the public schools. Journal of Economic Literature, 24(3), September, 1141-1177. Haveman, R.H. and Wolfe, B.L. (1984). Education, productivity, and wellbeing: On defining and measuring the economic characteristics of schooling. In E. Dean (Ed.), Education and economic productivity, Cambridge: Ballinger Publishing Co., pp. 19-55. Jaffe, A.B. (1989). Real effects of academic research. American Economic Review, 79(5), December, 957-970. James, E., Alsalam, N., Conaty, J.C., and To, D-L. (1989). College quality and future earnings: Where should you send your child to college? American Economic Review (Papers and Proceedings), 79(2), May, 247-252. Jamison, D.T., and Lau, L.J. (1982). Farmer education and farm efficiency. The World Bank, Baltimore: The Johns Hopkins University Press. Lydon, M. (1990). Movements in the earnings/schooling relationship, 1940-88. Ph.D. dissertation, Department of Economics, Stanford University, August. Maddison, A. (1987). Growth and slowdown in advanced capitalist economies: Techniques of quantitative assessment. Journal of Economic Literature, 25(2), June, 649-698. Manski, C.F., and Wise, D.A. (1983). College choice in America. Cambridge: Harvard University Press. Mare, R.D. (1980). Social background and school continuation decisions. Journal of the American Statistical Association, 75(370), June, 295-305. McMahon, W.W., and Wagner, A.P. (1981). Expected returns to investment in higher education. Journal of Human Resources, 16(2), Spring, 274-285. McPherson, M.S. (1978). The demand for higher education. In D.W. Breneman and C.E. Finn, Jr. (Eds.), Public policy and private higher education. Washington, D.C.: Brookings Institution, pp. 143-196. Mincer, J. (1989). Human capital responses to technological change in the labor market. Working Paper No. 3207, December. National Bureau of Economic Research. Mincer, J. (1991). Education and unemployment. Working Paper No. 3838, September. National Bureau of Economic Research. Mincer, J., and Ofek, H. (1982). Interrupted work careers: Depreciation and restoration of human capital. Journal of Human Resources, 17(1), Winter, 3-24.

84


Murnane, R.J., and Olsen, R.J. (1989a). Effects of salaries and opportunity costs on duration in teaching: Evidence from Michigan. Review of Economics and Statistics, 71(2), May, 347-352. Murnane, R.J., and Olsen, R.J. (1989b). Will there be enough teachers? American Economic Review, Papers and Proceedings, 79(2), May, 247-252. Murphy, K., and Welch, F. (1989). Wage premiums for college graduates: Recent growth and possible explanations. Educational Researcher, 18(4), 17-26; and reprinted with modification as: Wages of college graduates, In W. E. Becker and D.L. Lewis (Eds.), The economics of American higher education. Norwell: Kluwer Academic Publishing, 1992, Chapter 5. Nelson, R.R., and Phelps, E.S. (1966). Investment in humans, technological diffusion, and economic growth. American Economic Review, Papers and Proceedings, 56(2), May, 69-75. Nickell, S. (1979). Education and lifetime patterns of unemployment. Journal of Political Economy, Supplement, 87(5), Part 2, October, 117-132. O'Neill, D.M., and Sepielli, P. (1985). Education in the United States 1940-1983, Special Demographic Analysis, CDS-85-1, U.S. Department of Commerce, Bureau of the Census, July. O'Neill, J. (1985). The trend in the male-female wage gap in the United States. Journal of Labor Economics, 3(1), January Supplement, 91-116. Pencavel, J. (1986). Labor supply of men: A survey. In 0. Ashenfelter and R. Layard (Eds.), Handbook of labor economics Volume 1. New York: NorthHolland, pp. 3-102. Pencavel, J. (1991). Higher education, productivity, and earnings: A review. Journal of Economic Education, 22(4), Fall, 331-359. Plant, M., and Welch, F. (1984). Measuring the impact of education on productivity. In E. Dean (Ed.), Education and economic productivity. Cambridge: Ballinger Publishing Company, pp. 163-193. Ransom, M.R. (1991). Seniority and monopsony in the academic labor market. Provo, Utah: Brigham Young University, May. Salvo, J.J., and McNeil, J.M. (1984). Lifetime work experience and its effect on earnings. Current Population Reports, Series P-23, No. 136, U.S. Department of Commerce, Bureau of the Census, Washington, D.C., June. Schultz, T.W. (1975), The value of the ability to deal with disequilibria. Journal of Economic Literature, 13(3), September, 827-846. Shorrocks, A.F. ( 1975). The age-wealth relationship: A cross-section and cohort analysis. Review of Economics and Statistics, 57(2), May, 155-163. Solmon, L.C. (1981). New findings on the links between college education and work. Higher Education, 10(6), November, 615-648. Solmon, L.C. (1985). Quality of education and economic growth. Economics of Education Review, 4(4), 273-290. Wachtel, P. (1975). The returns to investment in higher education: Another view. In F.T. Juster (Ed.), Education, income, and human behavior. The Carnegie Commission on Higher Education, Mc-Graw Hill, pp. I51-170. Weisbrod, B.A., and Karpoff, P. (1968). Monetary returns to college education,


85

student ability, and college quality. Review of Economics and Statistics, 50(4), November, 491-497. Weiss, R.D. (1977). Elasticities of substitution among capital and occupations in U.S. manufacturing. Journal of the American Statistical Association, 72(360), December, 764-771. Welch, F. (1970). Education in production. Journal of Political Economy, 78(1), January, 35-59. Welch, F. (1979). Effects of cohort size on earnings: The baby boom babies' financial bust. Journal of Political Economy, Supplement, 87(5), Part 2, October, 65-97. Willis, R.J., and Rosen, S. (1979). Education and self-selection. Journal of Political Economy, Supplement, 87(5), Part 2, October, 7-36. Wise, D.A. (1975). Academic achievement and job performance. American Economic Review, 65(3), June, 350-366. Wozniak, G.D. (1984). The adoption of interrelated innovations: A human capital approach. Review of Economics and Statistics, 66(1), February, 70-79. Zarkin, G.A. (1985). Occupational choice: An application to the market for public school teachers. Quarterly Journal of Economics, 100(2), May, 409-446.

4

IS PUBLIC EDUCATION PRODUCTIVE? David Alan Aschauer

It is commonplace to hear and to read the assertion that productivity and economic growth are adversely affected by the overall level of government expenditure in the economy. Strict conservative economists and policymakers have typically pointed to a large government as a drag on economic performance. Norman B. Ture, a former Treasury official in the Reagan Administration and now the head of the Institute for Research on the Economics of Taxation, has echoed this theme by rhetorically asking "What did Ronald Reagan leave us as a legacy? It was the conviction that we will do better as a society if we constrain the role of government" (National Journal 1989, p. 2144). Such arguments, however, usually do not properly distinguish between public "consumption" and public "investment" in tying government spending to the performance of the private sector. Just as private investment in plant and equipment is important for the process of economic growth and productivity enhancement, so too is public investment in an economic infrastructure. In a series of papers dealing primarily with the United States, Aschauer (1987, 1989a, 1989b, 1989c) has shown that a "core" economic infrastructure- streets and highways, mass transit, airports, sewers and water distribution systems- is directly linked to private sector productivity, profitability, and investment behavior. In brief summary, public infrastructure capital acts as a complementary input to private tangible capital in private production; consequently, a rise in public nonmilitary investment raises the rate of return to private

87

88


capital, thereby boosting productivity, and spurs private investment and economic growth. Public consumption, on the other hand, does seem to be negatively related to productivity growth and private investment. But the notion that public investment has differential effects from public consumption on the private economy can, and should, be extended from investment in tangible capital such as roads, bridges and sewers to investment in intangible capital accomplished through public educational spending. Just as a road may increase the rate of return to constructing a factory, public schooling may augment the returns to private sector work activity. A cursory reading of the results in Aschauer (1989a), however, would lead to the conclusion that educational capital expenditures are not linked to private sector productivity. Table 4-1 shows that for the United States time series, the educational capital stock bears no statistical association with the productivity of the private capital stock. Thus, Alan Blinder was prompted to state that "as a professor, I regret that it isn't educational buildings" that is the most important type of public investment in terms of its influence on private sector productivity (Blinder 1988). Of course, this lack of a time series correlation between the stock of public sector educational buildings and private sector productivity would be expected even if the educational process were truly an important ingredient in the recipe of economic growth and productivity enhancement. The lags between the construction of schools, the actual education of the Table 4-1. Public Capital by Type and Productivity in the United States, 1949 to 1985

Type Core infrastructure (highways, mass transit, airports, electrical and gas facilities, water, sewers) Other buildings Hospitals Conservation and development Educational buildings Source: Aschauer (1989a) Note: Standard errors in parentheses.

Output elasticity .24 (.05) .04 (.03)

.06 (.04) .02 (.02) -.01 (.05)

IS PUBLIC EDUCATION PRODUCTIVE?

89

workforce, and the ultimate impact on productivity are long and varied enough to sever any contemporaneous link between public expenditures on schools and private sector productivity. In order to ascertain the importance of education to the level of national productivity, it is necessary to shift to a long-run analysis where temporal lags will be of little consequence. To this end, this chapter contains an investigation of the linkage between public spending on education and national productivity levels by using a cross-section of countries over the period from 1960 to 1985. The next section discusses the conceptual framework of the study, while the third section contains empirical results. A final section offers conclusions about the role of public spending on education in the determination of productivity and economic development.

Conceptual Framework

The usual neoclassical production technology is expanded to take account of the public provision of educational services to laborers. The production function is assumed (1)

where Y = total output of goods and services, A = an index of the state of economic development, L = "effective" or educated labor and K = physical capital equipment and structures. Effective labor, in turn, is produced by the application of educational "capital," H, to raw labor, N, according to (2)

Substituting equation (2) into (1) yields the following production function in the three variables of raw labor, educational capital, and physical capital (3) Under the maintained assumption that each of equations (1) and (2) is linearly homogeneous, equation (3) will also be linearly homogeneous, which allows the production function to be written in the ratio form: (4) where c = (1 - a) · (1 - b). The last formulation is convenient as it allows us to exploit the steady-state assumption of constancy of the ratio of tangible and intangible capital stocks to output.

90


The straightforward interpretation of the parameters a and 1 - a are as elasticities of output for effective labor and physical capital, respectively. However, under the standard operating benchmark of competition in goods and factor markets, output elasticities are equated to factor shares of total output; accordingly, SK SL

=

a

=1-

a

where si = share of total output paid to factor i, i = K,L. In addition, estimates can be made of the implicit shares of raw labor and educational capital in the total share of output earned by effective labor, sL. Indeed, sN=(1-b)sL bsL.

SH =

Thus, the parameter b is a measure of the proportion of the total return to effective labor which should be attributed to the educational process, and the residual, 1 - b, the proportion to be properly attributed to basic labor.

Empirical Analysis

This framework is used to determine the extent to which variations in physical and educational capital stocks, relative to output, are capable of explaining variations in either labor productivity or per capita output. Several empirical strategies present themselves. Conceivably, time-series data for a single country could be used to estimate a production technology with educational capital as a factor of production. As noted in the introductory section of this chapter, Aschauer (1988; 1989a) found no statistical association between educational structures and productivity, prompting Blinder (1988) to conjecture that there is apparently no connection between education and productivity. While Blinder's conclusion seems unwarranted, it does serve to point out that the lags between educational spending and a productivity impact are likely to render the time-series approach relatively useless. Alternatively, one might hope to accomplish such an investigation by looking across a cross-section of states or localities. Garcia-Mila and McGuire (1987), using state-level data for the United States, obtained results consistent with the view that educational spending is an important input to private output. Still, upon reflection, a difficulty with this and related work becomes apparent; a high degree of labor mobility would


91

attenuate the linkage between the educational process and market returns to effective labor in any given locale which, in turn, would affect the empirical estimates to an unknown extent. However, this latter objection is less likely to pertain to an empirical study across countries as international labor mobility would be of much less quantitative importance than that between cities or subnational governmental jurisdictions. This approach utilizes data on output, investment, and population for a cross-section of 121 market-oriented countries provided by Summers and Heston (1988). A distinct advantage of the Summers-Heston international data set is that an attempt has been made by the authors to convert variables expressed in national currencies to a format allowing meaningful cross-country comparisons by correcting for deviations from purchasing power parity. The basic Summers-Heston data set has been augmented by public educational spending variables from World Military and Social Expenditures. These educational spending data are said to "represent current and capital expenditures by governments for public education and subsidized private education for preschool through university levels." In collecting the data, effort has been taken to include expenditures at the local and intermediate as well as the central levels of government, although the coverage is apt to be more complete at the level of the central government. Unsubsidized private educational expenditures are not included. Summers and Heston provide a subjective ranking of countries in terms of quality of data, running from A (highest data quality, as in the United States) to D (lowest data quality, as in Zambia). Following Romer (1989), the empirical estimates presented below are obtained by a prior weighting of observations according to ratios of the mean squared residuals of preliminary regressions of growth rates of per capita output on a constant, consumption and investment; as the mean squared errors for countries of data quality B, C, and D are roughly the same and twice as large as that of countries of quality ranking A, observations from the last countries are weighted twice as heavily as those from the lower data quality countries. The high-income oil-exporting countries of Kuwait and Saudi Arabia have been excluded from the sample; as argued in Barro (1988), these countries have extremely high per capita incomes (for instance, in 1960, Kuwait has a per capita income of $50,000 (1980 dollars), more than seven times that of the United States) yet low endowments of physical and human capital. Finally, the lack of data on educational expenditures for several countries forced a reduction in the number of countries included in the data set to 107.

92


The basic regression equation to be estimated is (4), after transformation by natural logarithms so as to be linear in its variables:

yin

= Zo

+z

1 •

(h/y) + z

2 ·

(kly) +

e

(5)

logarithm where yin = natural logarithm of output per capita, h/y of the ratio of educational capital to output, k I y = logarithm of the ratio of physical capital to output, and e captures omitted influences on the level of per capita income. Here z 1 = (1 - a) · blc and z 2 = ale, with c = (1 - 1) · (1 - b). Unfortunately, data on capital stocks are not available for most countries in the Summers and Heston sample, and given the short time series for investment flows- from 1960 to 1985- it is not reasonable to generate capital stocks by traditional "perpetual inventory" or similar techniques. Instead, what is known about the behavior of the ratio of tangible capital to output over long periods of time is extrapolated. As early as 1961, Kaldor noted that one of the stylized facts of economic growth was the long run stability in the ratio of capital to output. Data from Maddison (1982) shows that over the period 1950 to 1979 the ratio of the growth rate of labor productivity to the growth rate of capital per hour is very nearly unity for a sample of seven countries (for which there are direct capital stock estimates), implying a close to constant ratio of physical capital to output. Assuming the capital-output ratio to be constant in the steady state allows us to estimate a country's capital-output ratio in the following fashion. The percentage rate of change of the physical capital stock is given by K=IIK-d

where and d

K= =

percentage change in physical capital, I = gross investment physical depreciation rate; equivalently, we have

K

=

(1/Y). (Y/K)- d.

Y = g, the assumption of a constant steady-state ratio of capital to output implies

If the percentage growth rate of output is given as

KIY = (1/Y)/(g

+ d)

(6)

as the steady-state ratio of capital to output. Data is available for the share of gross investment in output, IIY, as well as for the growth rate of output. To obtain estimates of capital-output ratios, therefore, only an estimate of the physical rate of depreciation to apply to the capital stock is needed. In a recent article, Maddison (1987) reports an average ratio of capital consumption to output of between 0.11 and 0.12 for a broad range of countries over a similar time period. This, in turn, translates into ,a

93


depreciation rate equal to 0.03 or 0.04 after use is made of the condition d =(DIY)· (YIK)

= g · (DIY)I[(IIY)-

(DIY)].

For example, the United States, with g = 0.03 and (1/Y) = 0.21 over the period 1960 to 1985, has an implied depreciation rate of0.033 (DIY= 0.11) or 0.04 (DIY = 0.12). As it turns out, the empirical estimates are not particularly sensitive to the choice of depreciation rates within a reasonable range. The implicit maintained assumption of this study, then, is that the variation in steady-state capital-output ratios across countries is reflective, not so much of differences in physical depreciation rates, but rather of differences in technological parameters, conditions of taxation, and other fundamental factors. For example, a country with a low level of capital output is likely to be a country with a technology favoring labor in the production process (low output elasticity of capital, a), while a country with a high capital-output ratio tends to be endowed with a high output elasticity of capital. Thus, the appropriate way to interpret the coefficients, such as a estimated below, is as an average- across countries- elasticity of output with respect to tangible capital. Figure 4-1 depicts estimates of physical capital to output ratios for market-oriented economies obtained by using equation (6) and assuming a physical depreciation rate of 3 percent per annum. The capital-output 7 .-------------------------------------------------~

s r5 4

3

2 :

..

... ... ·. ··. ··.

·.

.·

O L-----~------~------L-----~-------L------L-----~

0

20

40

60

80

100

Heston-Summers Classification KY03

Figure 4- 1. Physical Capital/Output Ratios

120

140

94


ratio averages 2.47 across countries and varies between a high of 6.04 to a low of 0.68; roughly two-thirds of all capital-output ratios fall between 1.35 and 3.59. Since the revolutionary work of Schultz (1961), Becker (1964) and others, economists have come to treat education and training as human capital investments, raising the productive quality of raw labor or tangible labor input. Yet, estimates of intangible human capital stocks exist for only a very select group of countries; see, for example, Kendrick (1976) for such an estimate for the post-World War II United States. Conceptually, however, we may use expenditures on public education to derive an estimate of the stock of public educational capital, relative to output, in a fashion analogous to the procedure undertaken above for tangible capital. Specifically, we define the steady-state ratio of public educational capital to output as HIY = (EDIY)I(g+ded)

(7)

where ED = flow expenditures on public education at all levels of government and ded = depreciation rate of educational capital. As before, assuming a depreciation rate for educational capital allows an estimate of the ratio of educational capital to output. Kendrick (1976) applies a depreciation rate of 3 percent per year to investments in education, which corresponds closely to the same depreciation rate for physical capital discussed above; arguably, however, a lower depreciation rate should be applied to human investment than to physical investment. As it happens, the empirical results appear not to be too dependent on the choice of a depreciation rate between 1 percent and 5 percent per annum. Figure 4-2 shows estimated ratios of educational capital to gross domestic product under the assumption of a 3 percent depreciation rate. The average ratio across countries equals 0.59 and varies between 0.19 and 1.36; approximately two-thirds of the estimates lie between 0.33 and 0.85. Figure 4-3 plots the above estimates of public educational capital against those of physical capital for the same set of market economies. As is clearly evident from the scatter plot, countries that tend to invest more in education- and, therefore, would tend to have a higher ratio of educational capital to output- also tend to invest more in equipment and structures and so would tend to achieve a higher level of physical capital per unit of output. This is consistent with the notion that both types of capital augment raw labor in the production process and that, at least to a degree, the public sector invests in human capital development to complement investment by the private sector in hard capital goods such as equipment and structures.

95


1.4 1.2

..

1 0.8

..

.·

0.6

..

0 .4

.. .

....

0.2 0

0

20

......

.·

40

60

80

100

Heston-Summers Classification

120

140

HY03

Figure 4-2. Education Capital/Output Ratios

KY03 7 ~-----------------------------------------------.

6 5

·.

4 I

3 2

. ... ...

... :

.. . ·..

o L-----J_----~------~-----L------L-----~----~

0

0.2

0.4

0 .6

Figure 4-3. Capital/Output Ratios

HY03

0 .8

1.2

1.4

96


Table 4-2.

Dependent Variable: yin

Depreciation Rate

.02

constant

hly k/y Latin Africa

OECD Rz SER* NOB**

7.50 (.24) .34 (.13) .48 (.16) -.16 (.17) -1.36 (.16) .61 (.15) .980 .503 107

.03

7.43 (.22) .33 (.13) .48 (.16)

-1.29 (.14) .68 (.12) .980 .503 107

7.61 (.22) .39 (.13) .52 (.15) -.15 (.17) -1.34 ( .15) .58 (.14) .981 .489 107

.04

7.55 (.21) .39 (.13) .52 ( .15)

-1.27 (.13) .65 (.12) .981 .489 107

7.83 (.23) .46 (.13) .53 ( .15) -.13 (.17) -1.31 ( .15) .57 ( .14) .982 .479 107

7.77 (.22) .46 (.13) .53 (.15)

-1.26 (.13) .63 (.12) .982 .478 107

Note: Standard errors in parentheses.

* SER: Standard error ** NOB: Number of observations

Table 4-2 shows estimates of equation (5) using estimates of tangible and educational capital stock ratios obtained from equations (6) and (7) over a range of possible depreciation rates of between 2 percent and 4 percent per year. In order to account for differences in the level of economic development, dummy variables for geographical regions (Africa and Latin America, including Mexico) and for the advanced economies composing the Organization for Economic Cooperation and Development (OECD) were added to the estimating equation. In general, positive relationships between both of the estimated capital ratios and the level of per capita output are found to exist; for example, for a depreciation rate of 3 percent, we see that a one percentage point increase in the level of physical capital to output is associated with an increase in per capita output of 0.52 of 1 percent, while an equal percentage point increase in the stock of public educational capital to output is related to a smaller rise in per capita output of 0.39 of 1 percent. Furthermore, the statistical association between capital stock ratios and per capita GOP is robust to reasonable alterations in depreciation rates. Over and above the influence of the stocks of equipment and structures and of education, however, per capita output is low in Africa (relative to the level in Asia with the exception of Japan) and is relatively high in the OECD; the average level

97


Table 4-3. constant

hly k!y pop Latin Africa

OECD R2 SER NOB

Dependent Variable: yin 7.22 .37 .57 .03 -.11 -1.27 .59 .981 .491 107

(.40) (.12) (.15) (.03) (.17) (.16) (.14)

7.13 .37 .58 .04

(.37) (.12) (.15) (.03)

-1.22 (.13) .64 (.12) .981 .490 107

of per capita output in the Latin American countries is statistically indistinguishable from that in the non-OECD Asian countries. Indeed, dropping the dummy variable for the Latin American countries does not discernably erode the fit of the estimated equation. A test for the appropriateness of the restriction that the production function is linearly homogeneous is easily accomplished by adding the (log) level of population to the regression equation; if there is any evidence of increasing (decreasing) returns to scale to be found in the data, population should enter the regression with a positive (negative) and significant coefficient. Table 4-3 suggests that there is no statistical justification for rejecting a working hypothesis of constant returns to scale; although the coefficient estimate on population is positive, it is quantitatively small- a 1 percent increase in the average level of population would be associated with only a 0.03 percent (in an equation with a dummy for Latin America) or 0.04 percent increase (without Latin American dummy) in per capita income- and insignificantly different from zero at standard levels. Thus, from the perspective of economic development, a country cannot expect to raise its level of per capita output unless it supplies workers with equiproportionate increases in capital stocks. It is worth noting that excluding India from the sample does not influence this result in any meaningful manner. Table 4-4 gives direct estimates of the parameters a and b. For the United States, the share of tangible capital in total output is estimated by various researchers to lie between 25 percent and 40 percent (depending upon the particular definitions of capital and of income); consequently, an estimate of the coefficient a in this range would seem reasonable. The coefficient b measures the implicit share of educational capital in labor's share of income; it certainly should lie substantially below 1-a. As shown in the table, the estimate of the share of physical capital in total income

98


Table 4-4. yin = z 0 + [b/(1 - b)] · (h/y) z 3 · Africa + z 4 • OECD + e

7.55 .27 .28 -1.27 .65

Zo

a b Z3 Z4

+ {a/[(1

- a) (1 - b)]} · (k/y)

(.21) (.07) (.07) (.13) (.12)

+

R 2 = .981 SER = .489 NOB= 107

equals 0.27, with a 95 percent confidence interval of (0.13, 0.41). A similar estimate of 0.28 is obtained for the implicit share of educational capital in labor's income share, with a 95 percent confidence interval of (0.14, 0.42). Thus, we can state with some degree of confidence that a large portion of the total returns to labor can be attributed to the skills and knowledge associated with the educational process. Another way of stating these results, of course, is in terms of the shares of total income that may be attributed to each of the three fundamental factors of production: physical capital, educational capital, and raw labor. Capital's share of 0.27 leaves a share of 0.73 for effective labor, which can in tum be apportioned to educational capital and to raw labor. Educational capital's implicit share equals [(0.28) x (0.73)] = 0.20 while that of raw labor equals [(0.72) x (0.73)] = 0.53. The estimates of the average elasticities of output with respect to raw labor, educational capital and physical capital obtained above may be used to calculate an average rate of return to physical and educational capital stocks from the identities r = a · (YI K) = (1 - a) · (1 - b) · (YI ED)

rH

where r = average (across countries) marginal product of physical capital and rH = average marginal product qf educational capital. For the case of 3 percent depreciation of tangible and intangible capital, we obtain estimates of the returns to tangible and intangible capital equal to r = 0.10 and rH = 0.33. The estimate of the rate of return to tangible capital appears reasonable; for the United States, estimates of the gross rate of return to capital are typically in the range of 0.09 to 0.13, and estimates of the net (of depreciation) rate of return in the range of 0.07 to 0.11. On the other hand, the estimate of the return to the educational process of 0.33 is clearly higher than estimates typically obtained in the literature. For example, Freeman (1977) estimated that in the United States the rate


99

of return to education fell from around 0.13 in the 1950s to 0.075 in the mid-1970s. While other studies have found higher rates of return to education in less developed countries, such returns are in the range of 0.16 (see Psacharopoulos 1973) and thus fall short of the 0.33 calculated here. It must be remembered, however, that the flow of expenditure on public education is an incomplete measure of the total resources being devoted to the educational process in any nation, so that a mechanical application of the ratio of output to public educational spending, to convert the estimated output elasticity into an implicit rate of return, will result in upward bias in the estimate of rH . Specifically, in addition to public educational spending, the resources devoted to the educational process would include private spending on education as well as the foregone earnings of individuals while in school. Figure 4-4 shows that in the United States such an expanded notion of educational investment would place it in the range of 11 percent or 12 percent of gross domestic product, substantially higher than the public component of educational spending of some 6 percent of output. We may obtain a direct estimate of the average rate of return to tangible capital, r, and an estimate of the differential or wedge between the rate of return to tangible capital and educational capital, w = rH I r through nonlinear least squares estimation of the equation Rat io to GOP

0 . 14 ,------------------------------------------------,

0 . 13

0 .12

0 . 11

0.1

year

Figure 4- 4. Investment in Education

100


yin= zo + A(H/Y,K/Y;r,w) · (h!y) + B · (HIY,KIY;r,w) · (k/y) + e A(AIY,KIY,rH,r) = r · (K/Y)/{1 - r · [(KIY) + W · (H/Y)]} B(HIY,K/Y,rH,r) = w · r · (H/Y)/{1- r · [(KIY + w · (H/Y)]}. Table 4-5 shows the results of this estimation over the same sample of 107 countries. From equation (1), the estimated rate of return to physical capital equals 5.8 percent and has a 95 percent confidence interval of (2.4, 9.2). The estimated wedge between the rate of return to educational capital and physical capital is 62.4 percent, implying an estimate of the average return to educational capital of some 9.4 percent. However, given the size of the associated standard errors, the hypothesis of equal rates of return to physical and educational capital ( w = 0) cannot be rejected at standard confidence levels. Equation (2) indicates that these results are not affected in any significant way by the exclusion of a separate dummy variable for the Latin American countries. Equation (3) of table 4-5 aggregates physical capital and educational capital by constraining the rates of return to both types of capital to be equal. The overall rate of return to total capital- physical plus educational capital stocks- is then estimated at 6.1 percent, somewhat higher than for physical capital separately, and has a much tighter confidence interval at the 95 percent level, given by (4.1, 8.1). It would be interesting to have separate estimates of the output elasticities of, and rates of return to, both basic (primary and secondary school) and higher education. Unfortunately, for the bulk of the countries investigated here, data on educational spending by level are not readily

Table 4-5. OECD + e

yin A B

Zo

r

w Z3 Z4

Zs

Rz SER

NOB

=

z 0 + A · (hly) + 8 · (kly) + z 3 Latin + z 4

·

Africa

+ z5

((K/Y)/{1 - r · ((K!Y) + w · (HIY)]} w · r · (H/Y)/{1 - r · [(K/Y) + w · (H!Y)]}

= r · =

7.56 .058 1.62 -.14 -1.27 .79 .976 .550 107

(.15) (.017) (3.14) (.19) (.16) (.15)

7.50 .058 1.58

(.13) (.017) (3.17)

7.48 .061 1.00

-1.21 .86 .976 .549 107

(.14) (.13)

-1.21 .85 .977 .546 107

(.10) (.010)

(.14) (.13)

·

101


available, so it is not possible to pursue the approach taken above to accomplish this goal. However, data on school enrollment at the basic and higher educational levels are available from the World Bank publication World Development Report. Thus, we employ the following alteration of the previous strategy. Redefine effective labor as L

=

N · PSu · C

(8)

where, as before, N = raw labor (population) but PS = percentage of primary and secondary school-aged children enrolled in school, and C = percentage of individuals of college age enrolled in higher education. This functional relationship has the desirable homogeneity property that a given percentage point increase in raw labor (and each of its component age groups- school-aged and otherwise) and in enrollment yields the same percentage increase in effective labor. It also allows a potential for differential effects of basic and higher education on effective labor; for instance, a 1 percentage point increase in enrollment in basic education (given the population and its age structure) would produce a u percent increase in effective labor while a one percentage point increase in enrollment in higher education would induce a v percent increase in effective labor. Finally, a given percentage point increase in the population (holding fixed its age structure) will result in a 1 - ( u + v) percentage point increase in effective labor; if 1 > u + v, there are diminishing returns to raw labor in producing effective labor arising as a result of the need for additional education. Combining equations (1) and (8) allows the equation yin

=

zo +

Z1 ·

(kly)

+

Z2 ·

psc

+

Z3 •

ps

+e

where psc = the sum of the natural logarithms of the percentages of primary/secondary and college-aged individuals enrolled in school and ps = the natural logarithm of primary and secondary school-aged children in school. Here, z1 = a/(lla ), where, as before, a= elasticity of output with respect to physical capital, but where z2 = elasticity of effective labor with respect to primary and secondary school-education, and z3 = differential between the elasticity of effective labor with respect to basic versus higher education. Thus, if z3 is insignificantly different from zero, there is no evidence in the data of a differential effect of basic education, relative to higher education, on effective labor or, as a consequence, on per capita output. Table 4-6 provides estimates of the above equation. In equation (1), the estimated value for Zt. 0.46, yields an implicit estimate of capital's share in income of 0.31 and an associated estimate of labor's share of 0.69; accordingly, the elasticity of output with respect to primary and

102


Table 4-6.

Dependent Variable: yin

constant kly psc ps pop Latin Africa OECD R2 SER NOB

5.29 .46 .24 .05

(.55) ( .15) (.07) (.10)

-.06 (.16) -.49 (.20) ( .14) .43 .985 .449 107

5.24 .46 .23 .06 -.47 .46 .985 .446 107

(.53) (.14) (.07) (.10)

5.48 .47 .26 0.0 -.47 -.46 .985 .445 107

(.18) (.12)

(.30) (.14) (.05)

(.18) (.12)

5.39 .48 .26 0.0 .01

(.43) (.15) (.05)

-.47 .46 .985 .447 107

(.18) (.12)

(.03)

secondary education equals (0.69)·z 2 = 0.16. While the estimated value of z3 is positive, quantitatively indicating a somewhat higher degree of importance of basic education relative to higher education, statistically it is not different from zero at conventional significance levels. Equation (2) drops the Latin American dummy without any substantive effect on the overall results. Equation (3) imposes the restriction of equal elasticities for lower and higher education; the relatively tight estimate of z2 (0.26) implies that u+v = 2·z 2 is far below unity, as required for a diminishing return to raw labor in the process of producing effective labor. Equation (4) shows an insignificant separate effect for population on both quantitative and statistical grounds. Finally, Table 4-7 allows a direct estimate of the share of output that can be attributed to physical capital; the value of 0.32 is obtained with a small associated standard error and is of a reasonable order of magnitude.

Table 4-7.

+e

yin= z 0 + [a/(1 -a)]· (k/y) + z 2

5.48 .32 .26 -.47 .46

(.30) (.07) (.05) (.18) (.12)

·

psc + z 4 ·Africa+ z 5 • OECD

R 2 = .985 SER = .445

NOB= 107


103

Conclusion

This chapter has investigated the importance of public educational expenditures to the determination of levels of output per person across a large sample of market-oriented economies. The results are consistent with the notion that variations in per capita output can be largely explained by variations in physical capital and educational capital stocks relative to output. Countries that tend to invest a good deal in both tangible capital, such as machines and plant, as well as in intangible capital such as knowledge- through the educational process- are also countries that attain high levels of per capita output and, by extrapolation, of labor productivity. The empirical results indicate that the share of tangible capital in total output is in the range of 25 percent to 30 percentconsistent with previous estimates for a variety of individual countrieswhile the implicit share of educational capital in output is in the range of 15 percent to 20 percent. The latter estimate may be compared to calculations by growth accountants such as Edward Denison, who has estimated that as much as 42 percent of the growth in labor productivity over the period 1929 to 1956 in the United States may be attributed to (public plus private) education (Denison 1964). Finally, the average rate of return to public education was calculated to equal 9.4 percent. This estimate, while quantitatively higher than the 5.8 percent return to physical capital, statistically is insignificantly different from the latter at conventional levels. Thus, there is no evidence in this data sample to support the argument that a chronic underinvestment in public education exists across countries. While lesser-developed countries may invest a small proportion of their total resources in education and physical capital because they are poor, the results of this chapter suggest that- at least on average- the correlation between capital stocks and per capita output levels is reflective of, and consistent with, a productive relationship between public educational spending and per capita output. In order to raise levels of per capita output, poor countries must seek out ways to expand both tangible and intangible capital which, by raising the productivity of raw labor, will have the desired effect of raising overall standards of living.

References Aschauer, D.A. (1987). Government spending and the "falling rate of profit." Economic Perspectives, 12, May/June, 11-17.

104


Aschauer, D.A. (1989a). Is public expenditure productive? Journal of Monetary Economics, 23, March, 177-200. Initial presentation for NBER, July 8, 1988. Aschauer, D.A. (1989b). Does public capital crowd out private capital? Journal of Monetary Economics, 24, September, 171-188. Aschauer, D.A. (1989c). Public investment and productivity growth in the group of seven. Economic Perspectives, 13, September/October, 17-25. Barro, R.J. (1988). A cross-country study of growth, saving, and government. Unpublished manuscript. Becker, G.S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. New York: National Bureau of Economic Research, distributed by Columbia University Press. Blinder, A. (1988). Are crumbling highways giving productivity a flat? Business Week, August 29, p. 16. Denison, E.F. (1964). Measuring the contribution of education (and the residual) to "economic growth." The residual factor and economic growth. Paris: Organization for Economic Cooperation and Development. Freeman, R.B. (1977). The decline in the economic rewards to college education. Review of Economics and Statistics, 59, February, 18-21. Garcia-Mila, T., and McGuire, T.J. (1987). The contribution of publicly provided inputs to states' economies. Unpublished manuscript. Kaldor, N. (1961). Capital accumulation and economic growth. In F.A. Lutz & D.C. Hague (Eds.), The theory of capital. New York: St. Martin's Press. Kendrick, E. (1976). The formation and stocks of total capital. New York: National Bureau of Economic Research, distributed by Columbia University Press. Maddison, A. (1982). Phases of economic development. Oxford: Oxford University Press. Maddison, A. (1987). Growth and slowdown in advanced capitalistic economies. Journal of Economic Literature, 25, June, 649-698. National Journal, (1989). September 2, pp. 2142-2145. Psacharopoulos, G. (1973). Returns to education. San Francisco: Jossey-Bass. Romer, P.M. (1989). Human capital and growth: Theory and evidence. Unpublished manuscript. Schultz, T.W. (1961). Investment in human capital. American Economic Review, 51, January, 1-17. Summers, R., and Heston, A. (1988). A new set of international comparisons of real product and price levels. Review of Income and Wealth, 34, March, 1-26. World Development Report. (Various issues). Washington, D.C.: World Bank. World Military and Social Expenditures. (Various issues). Washington, D.C.: World Priorities.

5

THE CONTRIBUTION OF HIGHER EDUCATION TO R&D AND PRODUCTIVITY GROWTH Walter W. McMahon

The contribution of higher education to basic and applied research, both at the universities and after university graduates enter industry- as well as the contribution to the outcomes including patenting and technical change to productivity growth- is well recognized and not disputed by economists. The problems come in measuring the value of these contributions of R&D to productivity growth in any precise way, in relating them to the costs, and in isolating that portion attributable to basic research conducted at higher educational institutions. It is illuminating, however, to consider the nature of these measurement problems and then to go on to attempt to define or specify the theoretical framework for what it is that needs to be measured. There is a great deal of descriptive material about the academic research establishment and some of its impacts (some of which is merely anecdotal) but it, nevertheless, helps to define the framework within which academic research is conducted and the nature of some of the outcomes. A notable recent study on the "real effects of academic research" by Adam Jaffe (1989) does undertake a systematic empirical analysis of the effects of geographical proximity. The study finds significant effects of academic research on corporate patents in drugs, medical technology, electronics, optics, and nuclear technology, but it does not undertake to analyze the further steps or to measure overall impacts on productivity growth. Other investigations are also underway on the relation of investment in research to patenting, and on the relation of technical change to productivity. (See, for example, 105

106


Griliches and Berndt (1989), as well as the recent issues of the National Science Foundation's Science Indicators.) After considering briefly some of the problems in measuring the marginal productivity of the academic resources invested in basic research, and the conceptual framework, this chapter will turn to several situations in which efforts have been made to isolate and measure these impacts. Although each situation sheds some light on the problem from different perspectives, there still are many things that are not known and effects that are hard to trace. So, a high degree of precision in the measurement of impacts on productivity growth should not be expected. Basic Problems

There are five basic problems involved in tracing and measuring the contribution of higher education to R&D and hence to productivity growth. The first problem is that there are very long lags before basic research and development conducted by universities and colleges affects economic productivity, producing lags of perhaps seven to forty years or longer. The results of basic research conducted by universities must first be translated into the more highly applied development of products and processes, which is a much larger enterprise conducted largely by private industry. Also the research workers trained in universities must graduate and carry their skills into industry or government. But even after the development of workable products or processes, the dissemination of these to a broader market takes time before there is any economic impact. The second problem is that from the point of view of the econometrician, these long lags and the inherent nature of productivity change result in low frequency effects. Time series data over short time periods, such as from 1960 onwards when relatively good R&D data first became available for most of the OECD nations, are too short to adequately test for these low frequency effects. Using nonlinear estimation with the long lags that are necessary results in wide error margins. It is, therefore, usually necessary to use international comparisons for the period since 1960, for although there are problems of data comparability and other problems, these comparisons do reveal low frequency effects. A third problem is that new technology created by R&D is relatively meaningless in terms of its effect on productivity unless it is embodied. This means it must either be embodied in machines as the new machines (or replacement machines) are produced, or embodied in persons through education. Higher education, in fact, is the major institution embodying this new technology, not only in the research workers trained as new

THE CONTRIBUTION OF HIGHER EDUCATION

107

Ph.D.s, but also in undergraduates who will become the support staff in industry for the new technologies. This creates "vintage human capital" and "cohort effects," to be discussed at some length below. The trouble is that it is difficult to sort out the direct contribution of the education of new workers (human capital production) from the indirect contribution made by the R&D as the new technology is embodied in those new workers. The latter contributes an additional increment to their productivity and gives them an earnings advantage in the labor market. (See the most recent research results on this point by Bartel and Lichtenberg 1985, 1988). A fourth problem is that higher educational institutions conduct only a very small percentage of the R&D. They do only 8.9 percent of the total R&D, according to the National Science Board of the NSF (1987, p. 13). However, the research universities do over 50 percent of the basic research, whereas industry accounts for 80 percent of all development work. If R&D does effect productivity growth, to sort out what portion of this affect is attributable to the basic research mainly done in universities versus the development mostly done in industry is difficult. Furthermore, universities train virtually all of the research scientists who do the research and development work in industry and government. But then if these research workers are paid something approximating their marginal products, that contribution to productivity is included as part of higher education's direct contribution to growth through human capital formation. The fifth problem is that there are major externalities related especially to basic research that even those who question the existence of external benefit spillovers from education do not deny. These externalities mean that the earnings of scientists engaged in basic research do not include rewards for many widely disseminated and long delayed effects from their discoveries on the economy, some of which may not occur for generations. It is not only famous artists who often die poor; it is also famous mathematicians. In light of these externalities, it is not possible to use the earnings of scientific research workers, especially in academia, as a measure of their contribution to national income growth and hence to outputrelated productivity. Production functions, therefore, are more appropriate in addressing this problem than are earnings functions. There are various other problems that arise, but most are less unique to R&D. There is the "simultaneity problem," for example. Investment in R&D contributes to growth, but a feedback effect occurs in that growth of income, and output in a county in turn contributes to larger investment expenditures on both higher education and on R&D. (These effects are developed for higher education and for R&D in McMahon (1974, Chs. 4 and 6).) There is a recursive nature of this relationship in time series data for a single county, given the long lags that are involved,

108


but the simultaneity involved in longer-period inter-county data will be considered further below. A Brief Description of the Patterns

Most of the basic research in the United States is done by higher education in the major research universities (as shown in figure 5-l ). Most other research is done by government, with very little by industry; however, applied research and development is overwhelmingly done by industry. About 15 percent of the applied R&D is done by government, although, if the R&D tax credits given by government are taken into account, the portion supported either directly by government, or through these tax subsidies, is much larger. Since development expenditures loom so large, academic research focusing on basic research (the small 8.9 percent of the total) can be divided among fields as shown in figure 5-2. The largest amount , by far, is supported in the life sciences for research on various diseases and health problems. Research in physics, chemistry, and the computer sciences, currently at $2 billion a year, is less than half that in the life sciences, but considerably more than the amount spent on R&D in engineering and in the social sciences.

(Billions)

80r------------------- ----------------, 70

60

50

Industry Government and Other Universities & colleges

40

30

20 10

0

Basic research

Applied Research

Development

Source: National Science Board (1987, p. 13).

Figure 5-1. Types of Research Performed by Universities, Government, and Industry

109

THE CONTRIBUTION OF HIGHER EDUCATION (Billions) 1 0 r-----------------------------------~

9 8 7

-

Social and olher sciences and psychology 4

Engineering

\ _,..

Physical , environmental . malhematical and computer sciences

3

2

_ _,.. _,..

Life sciences

1........ . ········l

·-·-·-·-·-----·-·

Q L---L-~--~--~--~--~--~--~--~~

1976

78

80

82

84

86

Source: National Science Board (1987, p. 14). Figure 5- 2. University Research by Field

The United States has fallen far behind Japan and Germany since 1971 in the percentage of GNP spent on nondefense R&D, as can be seen in figure 5-3. If defense is included, the United States spends more. Moreover, although there is some spin-off of defense research into civilian applications, there is also a drain of talented scientists away from focusing their energies on civilian production problems in order to produce weapons. On balance, it is doubtful that defense research contributes as much to productivity growth per dollar spent as does R&D that is focused more explicitly on nondefense product_ and process innovation, or on other sources of productivity growth such as education, health, and the design of effective social programs that bear more directly on improvement in living standards. Data relevant to this issue will be presented later in table 5-3. But, in any event , there is continuing concern about the slow productivity growth in the United States and the United Kingdom since 1973 in relation to the higher growth in Japan, Germany, and other fast growing countries such as South Korea, Singapore, Hong Kong, and Taiwan. All have invested larger percentages of their GNP in education and in nondefense R&D since 1960 than has either the United States or the United Kingdom.

110

HIGHER EDUCATION AND ECONOMIC GROWTH (Percent)

3.0.------ -- - -- - - - - -- - - - - ,

-.,Japan

2.5 2.0

•• •

~..,. .. ...,.. ..

West Germany

~"' •• ••

.. '·1'1"!······ • ........ --·~··· ...

······

.. .... .,.

...

United States

I

1.5

United Kingdom France

Source: National Science Board (1987, p. 3).

Figure 5-3. Non-defense R&D expenditures as a percentage of GNP, by country

The Conceptual Framework

In light of these conceptual problems and this factual background, a production function , within which contributions of higher education and R&D to productivity and output can be measured directly , is more appropriate than a focus on proxies for their contributions to productivity, such as the earnings of graduates and the earnings of research scientists. The latter do not accommodate the long lags or the externalities from R&D. The Production Function

The production function that seems most appropriate expresses potential output, Y, as a function of both the quantity and the quality of each input. It allows for increases in the quantity of both capital goods, K , and raw unimproved labor, N , but also for improvement in the quality of labor through investment in human capital, Hand HE, at both basic and higher education levels. It also allows for investment in R&D , le ading to


111

both disembodied technical progre~s and embodiment in both human capital (HE) and physical capital ( K) as investment in the education of each new generation and in each new vintage of capital goods occurs. That is: Y = Y(N,H,HE,K,A,fl),

(1)

where: Y = real output (for example, real Gross Domestic Product, GDP); N = employment, the quantity of raw unimproved labor; H = human capital, formed through basic education, which raises the productivity of the labor force, measured by accumulated depreciated past real investment in primary and secondary education; HE = human capital created by higher education. The new technology created by R&D embodied in each succeeding vintage (or cohort) is embodied as gross new investment occurs, as indicated by the overbar. K = physical capital, with the new technology embodied in each new vintage; A = knowledge-capital created endogenously by investment in R&D, that is, A = A _ 1 + I A - 6 AA _ 1 , where 6 A is the rate of obsolescence of knowledge, and fl = disturbance to productivity growth due to wars, oil price shocks, weather, or other external shocks. All inputs and potential output are treated as at full capacity, with no underutilization. The demand-side (not specified here) jointly determines inflation rates and output, including underutilization. The most crucial point has to do with the embodiment of the new technology by means of new investment in human capital as each new generation of college and university students learn the latest techniques. Thus, the R&D done at universities augments their earnings power as they enter the labor force (Bartel and Lichtenberg 1985, 1988) and makes it difficult to separate the contribution of university-based R&D to productivity growth from the contribution of higher education alone, without the benefit of the embodiment of the new discoveries. Similarly, new investment in physical capital incorporates the more recent advances made as the result of R&D. But in this case, not all of the R&D is university-based, and the portion that is may have been basic research performed many years back. Both of these stocks of human capital formed through investment in higher education and, therefore of physical capital, must be measured in

112


"efficiency units," as shown in equations (2) and (3) below (with an overbar). They embody the new technology by means of gross investment (that is, both net new and replacement investment) at rates aHE and aK, respectively:

+ e0 m 1JHE- ()HEHE_, K = K_, + eaKtJK - ()KK_,,

HE= HE_,

(2) (3)

These exponential rates of embodiment, aHE and aK, where e is the base of Naperian logarithms, are endogenous in the sense that they depend upon the rate of investment in R&D, that is, IAHE- f:JAHE-1 AHE-1

(4)

But they are also endogenous in the sense that the rate at which this disembodied technical progress, a, gets embodied in either human capital or physical capital, or both, aH and aK, depends upon the rates of investment in human and physical capital, respectively. If there is no investment in higher education, there is no embodiment of the technology created by the newest basic research. The Contribution of Higher Education to R&D and Productivity Growth

The contribution of higher education to R&D and to productivity growth can now be defined more precisely with the help of this conceptual framework and measured in terms of the partial and cross partial derivatives of the production function shown below. Defining productivity growth as labor productivity growth (rather than total factor productivity), one can move this out of the production function by first taking the total differential with respect to time, then dividing through by Y (to get percentage rate of growth of output on the left), and then subtracting the percent rate of growth of employment from both sides (to get productivity growth on the left). Assuming that the variables have each been transformed in this way, the elements can be analyzed as follows:

Direct Contributions of Higher Education: 1. From teaching, including the effects on productivity growth from the embodiment of technology: 3Y/3HE 2. From research done at universities: aY/3A (3A/3AHE)

113


Indirect Contribution to Productivity Growth via Embodiment of R&D Done at Universities:

3. 4.

Of new technology in physical capital: oY!oK (oK!oA) (oA!oAHE) Training researchers, no externalities: Zero (included in 1)

Total Contribution of Higher Education and of R&D Conducted by Higher Education to Productivity Growth: 5.

oY!oHE

+

oY!oA (oA!oAHE)

+

oY!oK (oK!oA) (oA!oAHE)

If the production function given by equation (1) could be estimated

accurately, then with one simplifying assumption, ( (oA/oAHE)

=

A;{t.} it

would be a relatively straightforward matter to calculate these partial derivatives, and to compute the social rates of return to both the education and the academic R&D components that they imply.

Empirical Estimates

For the 11 OECD nations for which total investment by government, universities, and industry is available for the 1960 to 1980 period, a production function of the type described above was estimated. The dependent variable is labor productivity growth, measured as the five year percentage rate of change over time within each country in real GOP per person employed. All explanatory variables are also for comparable five year time spans. This intercountry dimension, and the five year time spans within each country, are both quite advantageous in that they allow for changes in productivity as the result of the effects from the supply side of rates of investment in higher education and in R&D as is necessary when testing for low frequency phenomena. The Functional Form

Assuming a Cobb-Douglas form as a first approximation, it is also possible to start with equation (1) above, take the logs (in contrast to the procedure described above), differentiate with respect to time, and convert to perworker terms by subtracting the rate of growth of employment of total labor

114


from both sides. (For the mathematics of this derivation, see McMahon, 1984, Appendix A). The result is a production function that explains productivity growth per person employed in terms of physical capital deepening, human capital deepening, the rate of investment in research and development, plus some variables necessary to control for demandside influences and for other disturbances. Specifically (and corresponding to the empirical results to be shown in table 5-1), the production function estimated for the 11 OECD nations, and then for the five largest nations for which data on investment in R&D is available, is as follows (the lower case letters represent proportional rates of change over time in the corresponding upper case variables shown in equation (1)): y - n where: y - n

=

=

( Y I N) 0 =

h - n

=

he - n

=

k - n =

a

=

e

=

u

=