INCREASING RETURNS TO EDUCATION IN THE U.S 1967-97 - AN INFORMATION ECONOMY CONTEXT
A DISSERTATION
SUBMITTED TO THE SCHOOL OF EDUCATION AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FUFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Trang 2UMI Number: 3160848
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy submitted Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction
In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted Also, if unauthorized copyright material had to be removed, a note will indicate the deletion ® UMI UMI Microform 3160848 Copyright 2005 by ProQuest Information and Learning Company
All rights reserved This microform edition is protected against unauthorized copying under Title 17, United States Code
ProQuest Information and Learning Company 300 North Zeeb Road
Trang 4I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy
Martin Carnoy (PrincipaYAdvisor)
I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy
toe ¥ — eee eee eee eee oe
Susanna Loeb
Trang 5INCREASING RETURNS TO EDUCATION IN THE U.5 1967-97 - AN INFORMATION ECONOMY CONTEXT
Sunny Xinchun Niu ABSTRACT
This study investigates increasing returns to education in an information economy context It uses an occupation classification scheme based on the dominance of symbolic content (Schement and Lievrouw 1984) to classify occupations into six information occupation categories
tt
(“primary producer,” “secondary producer,” etc) and one non-information occupation category “non-information”) It applies OLS regression models to examine returns to education in these occupations and a cross-time decomposition technique to separate the effects on increasing wage differentials by education of increasing returns to education and a shift from non-information to information occupations It also uses a multinomial logit model to examine individuals’
information occupation choices
Using the Current Population Survey 1968-98 data, the estimates produce four major findings (1) Employment in information occupations is growing, especially in “primary
producer” occupations Employment in “non-information” occupations is declining In "primary producer” occupations, women’s wage earnings caught up with men’s across education, yet women have less of a chance than men to work in these occupations (2) Higher education is not rewarded with higher wages in “non-information” occupations and there is little increase in this relationship over time On the other hand, rewards for higher education are substantial and are increasing in “primary producer” and “secondary producer” occupations (3) In “non-
Trang 6except for Asian-American men, minorities have much lower probability than Anglo-Americans to work in “primary producer” occupations (4) Except for older women, both increasing returns to education in information occupations and increasing employment among the more educated in “primary producer” occupations contribute to increasing wage differentials by education Except for older college graduates, increasing employment among women in “primary producer” occupations accounts for much of narrowing wage differentials by gender
Trang 7ACKNOWLEDGEMENT
Writing a dissertation proves to be far more difficult, challenging, and rewarding than I ever thought it would be It was a journey full of frustration, confusions and struggles, as well as learning and successes; there were moments of depression and helplessness, and of excitement and joy As I look back at the time of finishing my last revision, my heart is filled with
thankfulness, to people around me during the past few years, especially during the past three years, and to the loving God, whom I became to know in person in the past three years
I thank first my dissertation committee members — principle advisor, Professor Martin Carnoy; committee members, Professor Susanna Loeb, Professor Tom MaCurdy, and Professor Myra Strober; and committee chair, Professor Norman Nie
Martin has played such a critical role in my dissertation project; this work would neither get started nor get completed without him Leading me to get interested in the changing
information economy, Martin enthusiastically supported my proposal on further examining increasing returns to education in the context of the changing information economy Walking through with me the analysis results, Martin helped me to see overall pictures and to be able to tell stories Knowing well that most Asian students are weak in expository writing, Martin challenged me to go beyond the point of being a student simply finishing a problem set or conducting a data analysis, but to learn to think as an independent researcher As I had been constantly challenged throughout my dissertation work, I have learned a critical skill as a researcher — to examine an economic phenomenon, look for plausible explanations, and take a stand based on careful analyses Literature review, modeling and data handling, analyses and result interpretation cease to be separate work difficult to be united afterwards but well-
Trang 8Although I learned the concept from research method classes, it only became a personal knowledge and experience after I was constantly challenged to think as an independent researcher to tell a coherent story in my dissertation work Martin also guided my dissertation work in very practical ways My work covers a period of 30 years, conducting analyses for every single year seems to be burdensome and unnecessary; Martin helped me to choose appropriate years as to cover both booming and recession periods, thus the results provide relatively clear picture without diving into too much details Having advised many Ph.D dissertation projects, Martin knows so clearly which work to refer so I can learn how to write my dissertation in the right style Understanding my difficulties in English writing, Martin personally connected me with a native speaker with good writing skills on the subject to help edit my earlier draft And by the very end, he himself spent hours and hours editing my final draft Knowing clearly what is - expected in an oral defense, Martin suggested me to be brief in introduction and focused in
presenting major findings, and be strict to keep within the time limit
Trang 9when I was in the middle of the frustration and confusions Besides my amazement of how quickly he responded to often confusing results with new insights, his enthusiasm over economic research in the education field is contagious and often seized me with the excitement That excitement, along with the challenges he presented, led me to go beyond technical data analysis to the meaning of conducting research — to understand the changes occurred in this society and how these changes affected people’s life in our chosen field
Besides her close work with Martin in advising my writing, lam very grateful to Susanna for her practical guidance in modeling and data analysis With her rigorous training in theories and methods in economics, Susanna helped me to understand the always used regression modeling and its interpretation in such a crystal clear way that I never imagined before She offered detailed guidance on running different models to obtain necessary results, which otherwise would involve complicated computations Susanna also questioned the sophisticated economic modeling I used in my work and pointed out the potential problems, which many other researchers also faced When the problems did arise, she analyzed with me together the specific situation in my research Patient and diligent, young and energetic, with the good spirit of encouraging and protecting students, Susanna served very well on the committee as both a role model and a peer Iam very thankful to have her on my dissertation committee
Trang 10techniques and well experienced in economic research, Tom reminded me that the nature of
economic research is to explain complex economic phenomena, even empirical studies should not make the sophisticated modeling the goal Seemingly the greatest discouragement at the
moment, the suggestion turned out to be one of the most important decisions made in my dissertation work Then upon my earlier draft, Tom again challenged me to ask what
contribution my work made to current knowledge of the subject, which every serious researcher should start the research with Offering concrete examples, I have learned when to draw further conclusions based on results and general economic theories, and when to stop to make my own speculations legitimate conclusions Tom also helped me to understand problematic practices in cross-time decomposition though they are normally done so in similar research Instead of just going through an exercise, I was able to make a real contribution through a more careful analysis
A warm appreciation is to Myra Myra accepted gracefully the invitation to be on my committee and commented carefully on my 200 pages of an earlier draft She reminded me not to confuse but communicate clearly to my readers, and pay special attention to the details of tables and figures Thanks, Myra
Special thanks to Norman! Having worked with Norman on various research projects at Stanford Institute for the Quantitative Study of Society (SIQSS) in the past two years, I felt very comfortable to have Norman as my committee chair Regardless his fanatic schedule, Norman managed to fit into my oral date (Thanks to Emily Borom, Norman’s administrative assistant, for she helped the schedule.) Norman also made a great comment of comparing the new information occupation classification scheme I used with the traditional Bureau of Labor
Trang 11I would like to thank Matthew Crawley, Jeff Marshall, Peter Mollis, and Raymond Shen for their editorial help I have learned from them about the common mistakes I usually make and more appropriate phrases to present the same ideas Now I am able to spot my own mistakes and edit my own writing for the first round!
People in my personal life, my fiancé then husband, my parents, bothers and sisters in Christ, my friends, have loved me and sustained me through this journey, and the long journey of life from the past to the days to come With Hong, my fiancé when this work started and my husband now, I have walked into a kind of married life that I never thought it would belong to me, and I have gained personal understanding why martial status always matters in social studies! I may not work as crazy as before, such as stayed up all night, I work more efficiently and more effectively, and a regular schedule over time leads to a steady progress I still
experience difficulties and frustrations, I return to solve problems more quickly with a renewed spirit Besides the strong emotional support, Hong supported me in various life details He proof-read my drafts for grammar mistakes, he brainstormed with me about the stories the results might tell, he helped insert page numbers to tables and figures, he cooked and sent me dinner boxes when I stayed late in the officer, he went grocery shopping, and he helped set up my oral exam reception With such a love and support, we were able to prepare and celebrate our wedding in the midst of my dissertation work; Hong shared my struggles as well as joys in this learning process What a blessing!
Trang 12In God’s love, brothers and sisters at our campus fellowship and at the church we attend prayed for my oral exam and dissertation completion Particularly, I thank Chris and Esther Hsiung for their parent-like love and prayers, and Uncle Simon and Auntie Betty Yu for their prayers and love I thank Yvonne Chen for her prayers and listening ears, Lillian Huang for her prayers and encouraging words, Haiying Li for her prayers and accompany in various situations, and Marshall Yan for his prayers and valuable assistance in data analysis
Iam greatly indebted to Dr Weifang Min at Beijing University He encouraged and supported me in applying for Ph.D programs at top American research universities, from which this long journey could ever start However crazily busy as the Vice-President of the University, he continues giving his personal attention to his students I am deeply touched and thankful!
I thank my friends in Beijing, Balin and Wujing, Fenghua Guan, Zili and Yingcheng Yang, Rongging Yang, for their love and friendship over many years And I thank my friends here, Kai Jiang and Lin He, Bobai Li, Tatiana Melguizo, Yang Su, Christine Wotipka, Zhi Liu and Lei Zhang, Li-an Zhou, at Stanford for their enduring friendship
Finally, I thank God for He gives me a life through my parents, guides me grow and leads me to personally know Himself through people around me I look back to my life, I know everything happened to me and every one coming into my life did so by His good, pleasing and perfect will In God, every aspect of life —- studying for the Doctorate degree, completing the dissertation work, working with other graduate students at SIQSS, building and nurturing a friendship far and near, learning to love each other in marriage — starts to unite together in God for one goal that I may learn to know God more and grow to be more like Jesus Everything reveals its noble meaning beyond itself, makes struggles worthwhile and rewards lasting I thank God for ” in all things God works for the good of those who love him, .”
Trang 13CONTENTS 03.020 21a IV ACKNOWLEDGEMENITTS -.- HH HH TK TH TH nà Hà HH Tà Ki KH Tà Tà Hi tin nh nàn VỊ 0/900.) —— e ee ee Eee e eS EE SE SESE Eee EEE EEE EEE EEE XIV | 8 ©) ci ok (1) 50 She Rr Ki HT nà Ki Ki HH Tà Hà Hà Hà tà tin nh it XV LIST OF APPENDIX - - HH HH HH TT Ki KT Ki HH nà Ki HH Ki tà tà thi HH XVI [ NNTRODUC THIONN Ác ng TH HH HH TH 03030110 101 1 i8đ§§2:2.9080/28:42412 255 e 6
Three Major Trends in Wage DifferentiaÌs -. .- nhàn 7 Debates on the Causes and the Eourth 'Trend -. ch nh 9 Increasing Returns to Education Across and Within Industries - The Result of Technological Change - SH HH HH HH HH Hà nh nh nh nh nà 15 Computer se as a Measurement and the roblem -.-. c 19 đ0019ỡ016iỡ162150 55:77 1211 ae 22 jie@â)/60213007.90822.0.02./9) 41 24 The Concept of the Information Econormy -. ng ng 24 (0-1800 18: 1 ằẮee 26
The Information Occupation C ]lassification Scheme and Hypothesis 28 IV METHOD AND DATA WL EEE EES EEE EEE DEES EEE EE EE EEE 31 REQTESSIONS 0 cece cece eee eee ee ne een EEE EEE EEE EEE EEE EEE EE EE EEE EEE EE EEE SEES EEE SHEE EEE EEE: 31 A Cross Time DecompOSItIOn - on SH HH nh HH ni nh nh ng 37 A Switching Regression Model with Endogenous Switching and a Multinomial Logit
Trang 14V INCREASING RETURNS TO EDUCATION SH HH ngàn 51
The Changing Wage StruCUT€ - -.- cọ HH nen nh xa 51 Increasing Returns to Education and Wage Disadvantages Facing Educated Women 52 Wage Disadvantages Facing Educated Minorities -. .c cà sen 56 VI THE INFORMATION ECONOMY CONTEXT c:ccccece cece eee e eens nese nese eeenseenenneeseeeeeass 61 The Portrait of the Information Economy and Its Wage Structure 62 Increasing Within-Group Differentials Measured by Information Occupations 71 Higher Rewards to the More Educated and Women in Information Occupations 78 VII INCREASING RETURNS TO EDUCATION IN THE INFORMATION ECONOMY
Ằ® 6) 002912 5 82 Hiph and Increasing Returns to Education In Information Occupations 83 Equally Compensated Minoritles In “Primary Producer” Occupations 92 Increasing Returns to Education Accounting for Increasing Educational Differentials 99 VII SELECTION BIAS IN THE CASE OF WOMEN AND INFORMATION OCCUPATION
CHOICES .- QC Q HH SH HH nh Ki ni nh HH nh Ti ng nà tà nàn ng Tà nà nà ti hinh và 104 Selection Bias in the Case of Womeni - SH HH ng ưe 104 The More Educated Working In Information Occupations - 107 Minorities Are Less Likely to Work in “Primary Producer” Occupations 111 IX SUMMARY AND CONCLUSIONG - ch nghe 113 Summary of Eindings and Eour ConcÏusions -.- «sec LO
Future Research IĐirectiO'nS - - QC Q Q0 Ợ Ợ ng, ng HH HH nh hs 120 APPENDICES - QC Q9 Ợ ng g4 ng ng ng ĐH ng HH n ĐH ĐH ĐH nh nh nh v 123
Trang 15LIST OF TABLES
4.1 Variable DesCrIDtIOT - cm HH ni nh TK nh Hi nà Tà Hà nà Ki Hàn nh nà cà nà ni HEY 33 5.1 Percentage Differences (%) In Weekly Wages, 1967-97 -.- SH nh 53 5.2 Education Coefficients, The Base Model, Men and Women, 1967-97 54 5.3.a Wage Earning Differences Relative to Anglo-Americans by Race, by Education, Men,
6.2 Percentage Differences in Weekly Wages Between “Primary Producer” and “Non-
Information” Occupations (%) by Percentiles, All and by Gender, 1967-97 Ó8 6.3 Percentage Differences in Weekly Wages Between the 90 and the 10' Percentiles (%) by
Education, All and by Gender, 1967-97 -.- - c9 SH HH HH HH KÝ nh nh he 73 6.4.a Information Occupation Coefficients, The Unrestricted Model, Men, 1967-97 75 6.4.b Information Occupation Coefficients, The Unrestricted Model, Women, 1967-97 76 7.1.a Wage Earning Differences Relative to Anglo-Americans by Race in “Non-information” and
“Primary Producer“ Occupations, Men, 1967-97 - c SH nà, 93 7.1.6 Wage Earning Differences Relative to Anglo-Americans by Race in “Non-information” and
“Primary Producer” Occupations, Women, 1967-97 -.- nà 95 7.2 Percentage Differences in Predicted Weekly Wages (%) Between College Graduates and High
School Graduates for Two Age Groups, 1971, 1997 -.- HH nh 100 8.1 Education Coefficients, the Switching Model and the Restricted Model, Women,
1978, 1986, 1997 Q00 200 2 2n ng HH ng HE ĐH HH HE HH HH BH 0 0 00006 0 6 6 6 6 ki 6 4 106 8.2 Race and Education Coefficients, The Multinomial Logit Model, Men and Women,
Trang 16LIST OF FIGURES
Trang 17LIST OF APPENDICES TO CHAPTERS
4.1 Computation for Rewards to Information Occupations and Returns to Education in the Unrestricted Model with Interactions cece ccc cc eee ce ee ee eee ee cence eee ee ee ene ee nese eee eeea eas 124 4.2 Computation for Wage Earning Differences Relative to Anglo-Americans by Race in the Base
ed NUTS Coe C-) nnn 126
4.3 Computation for Wage Earning Differences Relative to Anglo-Americans by Race in the Dnrestricted Plus Model With InteraCtIOnS HH HH HH HH HH Hành 127 6.1 Employment Distributions in Information Occupations (%) by Education, All, 1967-97 128 6.2 Employment Distributions in Information Occupations (%) by Race, All, 1967-97 130 6.3 Employment Distributions in Information Occupations (%) by Weekly Wage Percentiles,
All, 1967-97 e4 4 HT HH TH TH HH HH TH TH 0100180 111110171135 132 6.4 Distributions of “Officials and Managers” and “Professionals” in Information Occupations,
Men and Women, 1967-97 0 ccc c ccc cceccesncccecccceeeeensscececueeseceseeuceceeensucesceseseseueess 134 8.3 Probabilities of Working in Information Occupations with Given Characteristics, Men and
Trang 18CHAPTER I INTRODUCTION
Research on the economic value of education and work experience has a half-century tradition, but economists continue to try to understand why returns to education increase and decrease over time As early as the 1950s, Stigler and Bland (1957) and Schultz (1959) did pioneering research on the relationship between wages and investment in education Becker (1962) expressed returns to education in a general theory of human capital Mincer (1974) further developed the human capital model These researchers consider that education increases
knowledge and skills thereby improving an individual’s productivity More years of work experience means more on-the-job training, which improves productivity as well
The most interesting puzzle for human capital theory over the past 15 years is to explain why wage differentials are increasing Wage differentials by education level increased greatly in the 1980s and 1990s, with a particularly sharp rise in the relative wages of college graduates At the end of the 1970s, college graduates earned 37-38 percent more than high school graduates in hourly wages By 1989, they earned 58 percent more (Murphy and Welch 1992) Wage
differentials by work experience have also expanded substantially (Blackburn, Bloom and
Trang 19(R&D) and high-tech capital (Bartel and Lichtenberg 1987; Mincer 1989; Blackburn, Bloom and Freeman, 1990; Allen, 1996) Also, whereas wage differentials by education are low in goods industries, the differentials are high and increasing in service industries (Bluestone 1990) However, most studies attribute the lion’s share of increasing wage differentials to increasing differentials within industries (Blackburn, Bloom and Freeman 1990; Katz and Murphy 1992) Employees using computers on the job appear to earn 10 to 15 percent more than otherwise similar individuals This computer differential is greater for more educated individuals, and one researcher finds that nearly 40 percent of the increase in returns to education can be attributed to increased computer use (Krueger 1993) Although treating "computer use" as an independent variable has its own problems, the emergence of the concept, along with those of R&D and high- tech capital, reflects what happened during the past 30 years — a profound technological change in the economy characterized by a rapid development and application of information technology and increased computer use (Mincer 1989) Meanwhile, the finding that computer use for nonproductive activities does not enhance earnings suggests that the content of work is important in the new economy
This study investigates increasing returns to education Literature on the subject argues that this is the result of a skill-biased labor demand, driven by technological change, favoring the more educated and women (Mincer 1989, 1991; Bound and Johnson 1992; Berman, Bound and Griliches 1994; Allen 1996) This study asks the following questions: is there a measurement available to capture technological change occurring in the information economy? If there is, how does this new measurement help us to understand increasing returns to education in the
Trang 20dominance of symbolic content (Schement and Lievrouw 1984) to classify occupations into six
Ƒ/ // WM
information occupation categories (“primary producer,” “secondary producer,” “recycler,”
I) /
“maintainer,” “technology producer,” and “technology maintainer”) and one non-information occupation category (“non-information”) It applies OLS regression models to examine returns to education in these different occupation categories and a cross-time decomposition technique to separate the effects on increasing wage differentials by education of increasing returns to education and a shift from non-information to information occupations It also uses a switching regression model to examine selection bias in estimating returns to women’s education, and a multinomial logit model to examine individuals’ information occupation choices
Using the Current Population Survey (CPS) 1968-98 data, the estimates produce four major findings (1) Employment in information occupations is growing, especially in “primary producer” occupations Employment in “non-information” occupations is declining This trend holds for both men and women, and across education, work experience, race groups, and wage percentiles Information occupations have favored the more educated and women In "primary producer” occupations — those occupations associated with the creation of new information — women’s wage earnings caught up with men’s — across education levels — beginning in the early 1980s, yet women still have less of a chance than men to work in “primary producer”
occupations (2) Higher education is not rewarded with higher wages in “non-information” occupations and there is little increase in this relationship over time for both men and women On the other hand, rewards for higher education are substantial and are increasing in “primary
Wel
Trang 21graduate education have been able to approach and even surpass their Anglo-American counterparts in wage earnings In “primary producer” occupations wage differences between minorities and Anglo-Americans are small and no-existent even for lower levels of education Yet except for Asian-American men, minorities have much lower probability than Anglo- Americans to work in “primary producer” occupations (4) Except for older women, both increasing returns to education in information occupations and increasing employment among the more educated in “primary producer” occupations contribute to increasing wage differentials by education Except for older college graduates, increasing employment among women in
“primary producer” occupations accounts for much of narrowing wage differentials by gender In summary, wage earnings in “primary producer” occupations seem to be more merit- based, characterized by large variations across education but not across gender and race; wage earnings in “non-information” occupations seem to be more attribute-based, characterized by large variation across gender and race but not across education
Given that the information occupation classification scheme classifies occupations based on the dominance of symbolic content (Schement and Lievrouw 1984), different information occupations thus likely reflect different skill requirements The above findings lead to a further understanding of increasing within-group differentials and returns to education (1) Re-
classifying occupations based on the dominance of symbolic content seems to be useful in explaining the process by which wage earnings became more unequal within education groups and more equal between men and women If technological change is accurately reflected in this new classification scheme, the results suggest that the dominance of symbolic content is
associated with the changing wage structure (2) The high rewards (after controlling for education levels) associated with working in information occupations are likely to be
Trang 22educated and women in information occupations indicate that they are, in some sense, more skilled (3) It seems that high and increasing returns to education reflect less the inherent productivity associated with more education than high and increasing returns to the skills that the more educated acquire (4) These skills — especially those associated with “primary producer” occupations, of effectively and creatively using knowledge to generate new
information — seem to be equally rewarded across gender and almost equally rewarded across
Trang 23CHAPTER II LITERATURE REVIEW
The increase in wage differentials in the U.S labor market from the 1970s to the 1990s is remarkable While the trends are well-documented, the causes are still controversial The wage differentials by education and experience have increased greatly (Blackburn, Bloom and Freeman 1990; Bound and Johnson 1992; Murphy and Welch 1992; Katz, Loveman and Blanchflower 1995), whereas gender wage differentials have narrowed substantially (O’Neill and Polachek 1993) A more interesting trend, however, is that wage differentials have increased within narrowly- defined education and experience groups for both men and women (Juhn, Murphy and Pierce 1993; Katz, Loveman and Blanchflower 1995) There is a substantial difference in the timing of a rise in inequality across education and experience groups and in the timing of a rise in inequality within education and experience groups The difference suggests that the rising wage
differentials and the rising education and experience premiums are actually distinct economic phenomena Much of the increase in wage differentials is thus interpreted as being due to
Trang 24in different industries and especially within industries It leads to an increasing demand for skilled labor, which favors the more educated and women (Mincer 1991; Allen 1996; Bound and Johnson 1992) Computer use seems to be a very compelling measure of such skill-biased technological change (Mincer 1989; Krueger 1993) However, the finding that using a computer for nonproductive activities does not enhance earnings challenges the validity of incorporating computer use such a measurement, and suggests that the content of work is more important A measurement is needed to better capture technological change, and to lead to a further
understanding of increasing returns to education, thereby shedding light on increasing education differentials, increasing within-group wage differentials, and narrowing gender wage
differentials
2.1 Three Major Trends in Wage Differentials
Identifying major trends is the starting point of the literature on the subject Many have noticed and investigated the phenomenon that wage differentials increased enormously from the 1960s to the 1980s, and continued increasing in the 1990s Scholars have reached a consensus concerning three major trends Education differentials are increasing greatly, so are experience differentials, yet gender wage differentials are narrowing substantially More recent studies identify a fourth trend: wage differentials are increasing within groups with similar education and experience for both men and women!
First, education differentials have increased, with a particularly sharp rise in the relative wages of college graduates Murphy and Welch (1992) examine the structure of wages among white men for the period from 1963 to 1989 based on the March Annual Demographic
Trang 25
Supplements to the Current Population Survey (CPS) They show that hourly wages of college graduates exceeded hourly wages of high school graduates by 40 percent in 1966, 48 percent in 1971, 37-38 percent by the end of the 1970s, and then by 58 percent in 1989 Also using March CPS data, Blackburn, Bloom and Freeman (1990)? find that the gap between the earnings of college graduates and the earnings of high school dropouts fell from a 47 log point difference in 1973 to a 46 log point difference in 1979 and then rose to a 63 log point difference in 1987 Based on the CPS earnings data for 1973-74, 1979, and 1988, Bound and Johnson (1992) confirm that the rise in the average relative wages of more-educated workers is higher than that of less-educated workers Using more recent CPS data, Katz, Loveman and Blanchflower (1995) report that the earning differentials widened between college graduates and high school graduates in the 1980s and the early 1990s However, the pace of increase in education wage differentials was much slower in the period from 1987 to 1991 than in the period from 1979 to 1987
Second, experience differentials expanded substantially from 1963 to 1989 In Murphy and Welch’s study (1992) of young men with between 1 and 10 years of experience, the college premium fell from 48 percent in 1971 to 28 percent in 1979 and then rose rapidly to 69 percent in 1989 Among those with relatively low levels of education, the average wages of more
_ experienced workers increased relative to the wages of less experienced, especially for men from 1979 to 1988 (Bound and Johnson 1992) Katz, Loveman and Blanchflower’s (1995) study
compares the movement in the earnings of experienced workers (those with 26 to 35 years of experience) to that of new entrants in the work force among college and high school graduate men Experience differentials for college graduates expanded between the early 1970s and 1978 and declined after 1978 The gap between wages of experienced workers and new entrants for
Trang 26
less educated men increased sharply from 1979 to 1987 and exhibited a minor decline at the end of the 1980s until 1991
Third, after remaining fairly stable in the 1960s and 1970s, gender differentials narrowed throughout the 1980s O’Neill and Polachek (1993) find that women’ earnings represent
approximately 60 percent of men’ earnings throughout much of the post-World War II period After 1976, however, gender gap on average declined by about 1 percent per year Another study concludes that, relative to the average wage of men, the average wage of women increased by about 8 percent from 1979 to 1988, resulting in a fall in the average wage “disadvantage” for women from 30 percent to 24 percent (Bound and Johnson 1992) Katz, Loveman and
Blanchflower (1995) also find that the 1980s witnessed a substantial narrowing in gender wage differentials
2.2 Debates on the Causes and the Fourth Trend
The identification of these three major trends motivates researchers to search for the causes of the phenomena they document Although researchers more or less agree on the major trends, they debate about the causes of these trends For example, there are several hypothetical explanations for the increasing wage differentials.3 From a supply side point of view, Welch (1979) and Dooley and Gottschalk (1984) claim that the baby boom generation entered the labor market and drove down the college premium in the 1970s Consequently, fewer college graduates or slower growth in college graduates may be responsible for the increase in the college premium in the 1980s Another supply-side argument centers on the declining quality of public education That is, high school graduates are less productive than previous cohorts with similar levels of education (Dooley and Gottschalk 1984) Alternative explanations focus on
Trang 27
changes in wage-setting institutions such as the decline in unions (Freeman 1991), changes in pay norms (Mitchell 1989), and on the erosion of the real value of the minimum wage (Blackburn, Bloom, and Freeman 1990) Still others argue that shifts in product demand was largely associated with large trade deficits in the 1980s, which led to a sharp decline in manufacturing employment and a shift in employment toward sectors that are more education and women intensive (Borjas and Ramey 1995; Katz and Revenga 1989; Murphy and Welch 1991)
Blackburn, Bloom and Freeman (1990) evaluate these hypotheses in their study They look into the shifts in the demand and supply of workers with different levels of education and experience, and then examine the declining importance of two major wage-setting institutions the minimum wage and labor unions They examine how changes in measured demand, supply, and wage-setting institutional variables contributed to the changes in the earning gaps in the 1980s They also focus on the differences in the changes in these variables between the 1970s and the 1980s They find that the contribution of the changes in the allocation of labor across
industries was far from negligible This reallocation of labor affected relative earnings in part because employment for all skill groups shrunk in the industries in which the less skilled were relatively well-paid, but more importantly because the shrinkage in employment in high-wage industries involved a disproportionately large share of less skilled jobs Furthermore, less skilled workers tended to shift disproportionately into industries in which the relative wages of less skilled workers fall over time For the period of 1973-79 and 1979-87, the supply of college graduates increased relative to that of high school dropouts and graduates, although the rate of increase was lower in the 1980s than in the 1970s Time-series regressions‘ show a negative
Trang 28
association between earnings differentials and relative labor supply for all age and education
groups Declines in unionization were highly concentrated among the less educated in the period of 1980-88, and the proportion of high school dropouts and graduates who belonged to unions fell by 13 percent, while the proportion of college graduates fell by only 4-6 points An analysis of union wage premiums reveals large union wage effects for the less educated and insignificant (actually negative) effects for college graduates De-unionization and union wage premiums together substantially widened the earnings gap between those with college education and those without in the 1980s The authors conduct source-of-change analyses indicating that industrial shifts and de-unionization accounted for non-negligible portions of the overall increase in educational wage differentials in the 1980s.° Relative supply movements, which differed sharply in the 1970s and in the 1980s, accounted for a large part of the accelerated pace of change in the wage gaps At the same time, the authors argue that since labor supply and institutional changes seem reasonably well-measured in the study, the 1970s and the 1980s differed
importantly in ways not captured by the analysis Relative demand shifts were thus inferred as being accelerated in the 1980s There is little evidence, though, that the increasing wage
differentials across educational groups were due to the decline in the real value of the minimum wage in the 1980s, to the increase in the pace of technological change, to the changes in the labor supply of white men below age 25 and of women in different educational groups, or to the changes over time in the quality of less-educated workers
Offering an additional account of this increased demand hypothesis, Katz and Murphy (1992) use a simple supply-demand framework to analyze changes in the U.S wage structure
Trang 29
from 1963 to 1987 They show that there is substantial long-run growth in the relative supply of more-educated workers, young workers, and women, and the increase in the average educational attainment of the labor force is striking For the period from 1963 to 1987 as a whole, and most strongly for the 1980s, those workers with the largest increases in relative supplies tended to have the largest increases in relative wages A negative relationship between growth in labor supplies and growth in relative wages only appeared during the period of 1971-79 The authors argue that these findings indicate that demand growth was an important component of the change in labor prices over the period from 1963 to 1987 (particularly during the 1980s) Across the three decades, differences in the rate of the growth of the supply of college graduates as a fraction of the labor force appeared to play an important role in explaining the large differences in the behavior of the relative earnings of college graduates — the college wage premium increased moderately in the 1960s, declined in the 1970s, and expanded dramatically in the 1980s
Adams (1994) assesses the impact of domestic technology, foreign technology and import penetration on the U.S wage structure from the 1970s to the 1990s He finds that domestic technology was associated with rising wage differentials between high school graduates and college graduates, yet foreign technology and trade had the same effect on wages of high school graduates and college graduates His results do not support the opposite effects as suggested by factor propositions that foreign technology and trade led to the increasing education differentials
Trang 30workers with 1-10 years of experience, those in the 10 percentile of wages lost about 7 percent in
real log wages, whereas those in the 90" percentile gained almost 25 percent For workers with 21-30 years of experience, those in the 10* percentile lost about 3 percent, while those in the 90" gained almost 39 percent High school graduates in the 10" percentile lost about 15 percent in real wages, whereas those in the 90" percentile gained about 9 percent The relative wage
changes for college graduates showed a similar increasing disparity, with college graduates in the 10* percentile losing about 5 percent and those in the 90" percentile gaining about 25 percent Katz, Loveman and Blanchflower (1995) confirm the finding, showing that within-group inequality® expanded enormously, with the 90-10" percentile differentials in log weekly wages expanding by 31 percent for men from 1967 to 1989 and by 13 percent for men and 15 percent for women from 1979 to 1989
An important finding detailed in Juhn, Murphy and Pierce (1993) is that, for men, little change occurred in within-group wage inequality over the 11 years of 1959-70, however the period of 1970-88 was characterized by an enormous increase in inequality That is, within-group inequality started to expand in the early 1970s and continued increasing in the 1980s It sharply contrasts with the pattern of education differentials, where the wage gap between college graduates and high school graduates fell in the 1970s, then rose sharply between the late 1970s and the late 1980s then slowed down The authors thus find a substantial difference in the timing of the rise in inequality across observable dimensions of skills (i.e., education and experience) and the rise in inequality within schooling and experience groups While rapid increases in education differentials and returns to experience for less educated groups accounted for more than half of the total wage differentials in the 1980s, the increase in inequality based on “unobserved” skills
° For both studies, within-group differentials are empirically examined by looking at the
Trang 31operated from the late 1960s The authors conclude that the increased wage differentials and the rise in the education premium are actually distinct economic phenomena The earlier increase in within-group wage differentials suggests that the rise in the demand for “unobserved” skills predates the rise in returns in education in the 1980s Katz, Loveman and Blanchflower’s (1995) analysis of college graduates and high school graduates from 1970s to the early 1990s leads to the same conclusion
The very finding that the general rise in within-group inequality and the rise in education and experience premiums are distinct economic phenomena requires a new approach for
understanding the increase in wage differentials Juhn, Murphy and Pierce (1993) suggest an approach that assesses the impact of biased rates of technological progress in different industries and occupations on increasing wage differentials across education, experience, and within groups Actually, earlier studies already pointed in this direction Although Blackburn, Bloom and Freeman (1990) focus on explaining the widening of educational wage differentials, their industry-shift analysis find that certain industries paid low wages to workers in all educational groups.’ When the trend in wages by education group is disaggregated by industry category, it became clear that increasing education differentials were confined mainly to the service sector (Bluestone 1990) Howell and Wolff (1991) find that the high-employment growth segment had substantially higher skill and education requirements than the low-growth segment in both the goods and the services sectors But the low-growth service industries had a higher job skill level than the high-growth goods industries These studies, along with Juhn, Murphy and Pierce (1993), lead the following important points regarding the increasing educational and within-
7 The authors examine the industrial differences in the returns to education by including industry dummies in regressions The hypothesis is that change in earning differentials in the regression that includes industry dummies should be smaller than the change in the regressions without dummies
Trang 32group differentials First, within-group differentials have increased for all education levels Second, however, within-group differentials have not increased uniformly across education levels Third, biased rates of technological progress in different industries and occupations possibly have led to the increasing within-group wage differentials for all education groups, but differently at different education levels Fourth, more educated workers are in higher demand, resulting in increasing returns to education and thus increasing education differentials
2.3 Increasing Returns to Education Across and Within Industries — The Result of Technological
Change
The literature that focuses on the third and fourth points addresses returns to education across industries and within industries using several measurements of technological change Several studies examine how returns to education and experience differ when industries have different levels of research and technology (R&D) investments and other technological inputs Mincer (1989)? explores the effects of different paces of technological change on industry demand for educated and experienced workers as reflected in Panel Study of Income Dynamics (PSID) data during the period of 1968-83 The author finds relative increases both in demand for and in wages of skilled workers in the more progressive sectors where there was a faster pace of technological change.’” The increases held for both younger and older workers, and the rate is maintained over time Allen (1996) uses 1979-89 CPS data to estimate changes in the wage structure across industries An estimation of the conventional wage equation for each of 44
? Both Mincer (1989) and Mincer (1991) use total factor productivity growth indices (PG) for 28 U.S industries calculated by Conrad and Jorgenson Jorgenson’s measures contain detailed adjustments of labor inputs for their “quality” components, such as education, age and gender composition The productivity growth residuals are thus largely purged of human capital components
Trang 33industries finds that there was considerable dispersion in wage equation parameters across industries at any point of time The rate of the returns to education in 1979 was 5.7 percent, with a standard deviation of 1.4 percent across industries, ranging from 3.0 percent in the industry of “eating and drinking places” to 9.4 percent in the industry of “business services.” In 1989, the average rate of the returns to education increased to 8.0 percent with an increased standard deviation of 2.1 percent, ranging from 3.3 percent in “eating and drinking places” to 11.2 percent in “medical services.” There was also a tendency for industries with high returns to education to have greater returns to experience Allen also relates the CPS data to R&D intensity, usage of high-tech capital, recentness of high technology, growth in total factor productivity, and growth of capital-labor ratio He reports that three technology variables (R&D intensity, high-tech capital and growth of capital-labor ratio) accounted for 30 percent of the increase in the wage difference between college and high school graduates.'! The returns to education are larger in industries that were intensive in R&D and high-tech capital.’ The rate of return to education increases the most in the industries where R&D intensity grows most rapidly and high-tech capital is used most intensively.'’ The results confirm the results in earlier studies using CPS data in 1960, 1970 and 1980 (Bartel and Lichtenberg 1987) A very recent study by Bartel and Sicherman (1999) matches a variety of industrial-level measures of technological change to a panel of young
variables: growth of capital-labor ratios and growth of employment The coefficients on the interaction variable (education*PG) are positive in the period of 1970-79 and in the period of 1960-79
11 R&D intensity is measured by the ratio of scientific and engineering employment to total employment Usage of high-tech capital is a measured by the share of high-tech relative to total investment, where high-tech capital consists office, computing, and accounting machinery, and scientific and
engineering instruments The ratio of net to gross capital stock in 1979 and 1989 from Fixed Reproducible Tangible Wealth in the United States is used to measure the recentness of the technology
12 The author regresses the returns to education on R&D intensity, high-tech capital intensity, and other technology indicators
Trang 34workers, observed between 1979 and 1993 (NLSY) The authors find that the wage premium associated with technological change was primarily due to the sorting of more able workers into industries with a faster pace of technological change, and this premium was unrelated to any sorting based on gender or race In addition, the education premium associated with
technological change was the result of a greater demand of innate ability or other unobserved characteristics of more educated workers
It is important to note, however, that returns to education and experience differ by industries is a minor part of the story of why educational and experience differentials and within- group differentials increase greatly and why gender differentials narrows significantly Most studies attribute the lion’s share of increasing wage differentials to increasing differentials within industries Although focusing on reallocation of labor between industries, Blackburn, Bloom and Freeman (1990) show that the widening of the earning gaps for the work force as a whole was associated largely with growing gaps within industries About 70 to 80 percent of the increase in the earnings gap between college graduates and high school dropouts in the period of 1979-87 occurred within industries Davis and Haltiwanger (1991) use the information in the
Longitudinal Research Datafile (LRD)" to investigate changes in the plant wage structures over the past three decades Combining plant level wage observations in the LRD with wage observations on individual workers in the CPS, and estimating the between-plant and within- plant components of overall wage dispersion, they find wage differences even between plants within the same firm In Katz and Murphy (1992)’s demand-supply analysis, between-industry shifts in employment increased the demand for college graduate men by over 30 percent relative to men with 12 or fewer years of education Demand shifts in favor of women were much greater
Trang 35
for high school graduates and those with some college education than for high school dropouts and college graduates However, measured changes in the allocation of labor between industries can only account for a minority of the secular demand shift in favor of college graduates and women with rising relative wages throughout the period They argue that if within-industry relative factor demand is stable so that the changes in wage structure are entirely explained by between-industry shifts in labor demand and relative supply changes, then the shares of industrial employment of groups with rising relative wages should tend to fall inside every industry Since the share of aggregate employment of women and college graduates increased over the period of 1963-87, the stable within-industry demand scenario required a substantial shift in employment into industries that intensively employ women and more-educated workers However, the employment shares of women and college graduates increased in almost every industry in the period of 1963-87
A couple of studies focus on examining and explaining this within-industry wage differentials Berman, Bound and Griliches (1994) use Annual Survey of Manufactures (ASM) data during the 1979-1987 period to investigate the shift in demand away from unskilled to skilled labor in U.S manufacturing over the 1980s They highlight three findings (1) The shift was due mostly to increasing use of skilled workers within the 450 industries in U.S
manufacturing, and not to a reallocation of employment between industries, as would be implied by a shift in product demand due to trade or defense buildup (2) The trade and defense demand
were associated with only small employment reallocation effects (3) Increasing use of non- production workers was strongly correlated with investment in computers and in R&D Relating to the last point, the authors review recent work on examining the extent to which technological change explains increases in the demand for skilled labor They also find that computer and R&D
Trang 36
measures picked up a statistically and quantitatively significant effect, and that together they account for 70 percent of the move away from production labor The results support the notion that biased technological change was an important contributor to the within-industry skill upgrading
Sector shift combined with between-industry analysis therefore agree on one common point with within-industry analysis: technological change leads to an increased skill-biased labor demand, which favors the more educated and women.’* Using data from March CPS
1963-87, Mincer (1991) finds that year-to-year educational wage differentials can be
predominantly explained by R&D expenditures per worker.'” Allen (1996) finds that for the period of 1979-89, R&D had the greatest impact on wage growth among college graduates, and for workers with 20 or more years of experience For both men and women wage growth strongly linked with R&D intensity for college graduates with less than 10 years of experience R&D coefficients are generally larger for women than men Bound and Johnson (1992) speculate that women’s faster wage growth is attributable partially to improvement in the “unobserved” labor quality of women, and partially to technological change relatively favoring women Technological change comprises the principal source of the increase in education differentials and the decrease in gender differentials For older non-college workers, technological change account for an even larger proportion of the increase in their relative wages
2.4 Computer Use As a Measurement and the Problem
16 See Blackburn and Bloom (1988) for an excellent discussion of the impact of technological change on wage differentials Bartel and Lichtenberg (1987) present cost function estimates for 61 manufacturing industries that suggest that skilled labor is a component to new technology
Trang 37Even though they agree that biased technological change is the principal source of increasing wage differentials by education, researchers use different ways to measure
technological change that occurred in many industries at different paces Some examine R&D investment across industries and within industries, and investment in computers within industries Computer use itself has attracted focused attention as a measure of the contribution of technological change to increased wage differentials Mincer (1989) points out that as newer vintages of capital contain new technology, the skill-bias of capital intensity partly reflects the skill-bias of technology This supports the notion that while the computer revolution has been highlighted as a prototypical example of technological change occurring in the past three decades, computer use can be a legitimate measure of such skill-biaséd technological change Krueger (1993) actually investigates the impact of using computer on wages based on data from October 1984 and 1989 CPS His analysis shows that employees using computers at their jobs earned 10 to 15 percent more than otherwise similar individuals The computer differential was greater for more educated workers A further examination of the effect of computer use on the return to education shows that nearly 40 percent of the increase in the returns to education can be attributed to the expansion in computer use However, Krueger (1993) reports, “these results show the most highly rewarded task computers are used for is electronic mail, probably reflecting the fact that high-ranking executives often use e-mail On the other hand, the results indicate a negative premium for individuals who use a computer for playing computer games the coefficient on computer games virtually negates the coefficient of using a computer at all.” The results suggest that using computer for nonproductive activities does not enhance earning
Trang 38occupations A careful examination of other studies also leads to focused attention on
occupations For example, in Berman, Bound and Griliches’ study (1991), production labor and non-production labor distinction closely mirrors the distinction between blue-collar and white- collar occupations Production labor saving technological change was the major reason for the shift in demand from unskilled labor to skilled labor Katz and Murphy (1992) examine the industries and occupations together, and show that, over the 1967-87, industrial employment shifted out of low-tech and “basic” manufacturing into professional and business services, and out of production and service occupations into professional, technical and managerial
occupations '® These shifts favored college graduates and of women and work against less- educated men Allen (1996) finds the returns to education were much higher in the industries that are R&D and high-tech capital intensive, and his measures of R&D intensity is based on employment of scientists and engineers Also, when he drops scientists and engineers from the sample to examine the effects of technological change on the returns to education in nonscientific occupations, he finds that R&D and high-tech capital intensity had an impact on relative earnings across a broad range of occupations Bound and Johnson (1992) argue that technological change relatively favors women, because women tend to work in occupations that on average impose higher intellectual demands than do men Krueger (1993) further points out that the
characteristics of workers and employers in an occupation are likely to change slowly over time, yet some occupations adopted computers extremely quickly in the 1980s The author thus estimates the payoff to computer use with an alternative approach estimating the relationship
Trang 39
between the growth in wages and the growth in computer use at the level of the occupation." The results indicate that computer growth was positively associated with wage growth in an occupation
2.5 Concluding Remarks
The basic arguments of the literature can be summarized as follows The introduction of new technologies such as computerization from the 1960s to the 1990s have changed the relative demand for labor, which favors “more-skilled” (unobserved skills) workers over “less-skilled” workers, more educated workers over less educated workers, and women over men This favoring of the more educated and women have resulted in increasing within-group and
education differentials, and narrowing gender differentials Regarding skill-biased technological change, there are two important points First, it has been reflected in different industries and different occupations R&D intensity measures change across industries quite well, but computer use seems to be a plausible index to classify categories of occupations to reflect the differences in technological change across occupations Second, using computer use as a criterion to classify occupations has a serious flaw The finding that computer use for nonproductive activities does not enhance earnings indicates that the content of work is more important
The next chapter will discuss in more detail the industry and occupation classifications used in the literature that go along with traditional industry concepts, where manufacturing sector is the core sector and physical capitals are the dominant factor Technological change shifts the economy to be information based The dominance of symbolic content thus serves as
Trang 40
the new criterion of classification to better capture and measure technological change, and different occupations across all industries are consequently classified into six information occupation categories and one non-information occupation category
The Methods and Data chapter discusses regression models that incorporate the information occupation variables and interactions between these occupations and education It also discusses a decomposition technique to separate the accounts of increasing returns to education and the shift among information and non-information occupations on increasing education differentials However, estimating rewards to information occupation choices has a potential sample selection problem because the choices are possibly nonrandom, and they may be associated with other unmeasured characteristics contributing to higher earnings This sample selection problem would pose a problem in interpreting regression results In addition to this potential nonrandom selection problem, only wages of working women are observed, which means a truncated wage distribution Estimations based on a sample of working women deviate from those of the women population, which has important implications for understanding women’s work behaviors The Method and Data chapter then discusses a switching regression model with endogenous switching in the case of women, which separates the effect of self- selection and the effect of occupation choices However, the switching model is very restrictive on error term structure, and the parameters are often estimated with a great deal of error A separate multinomial logit model thus is discussed to understand the effect of explanatory variables on choices among information and non-information occupations