The Long Term Evolution of Female Human Capital Audra Bowlus and Chris Robinson University of Western Ontario Presentation at Craig Riddell s Festschrift UBC, September 2016
Introduction and Motivation I Labor market attachment of females has increased dramatically over the last half century (Blundell and Macurdy (1999), section 3, Goldin (2006), Blau and Khan (2013).) Large and very interesting literature trying to explain why - not the topic of this paper Focus of this paper is on an important implication of human capital theory, given this greater labor market attachment: an increased incentive for females to invest in human capital
Introduction and Motivation II One manifestation of increased incentive to invest: convergence in share of successive female birth cohorts that are college graduates to that of males Only a partial indication of increased human capital Standard composition adjustment methods show increase in efficiency units of female human capital through changes in education and experience cell composition Cannot capture within cell increases for females: an implication of human capital theory given increased labor market attachment Consequence: under-estimate of the contribution of female human capital to postwar growth
Introduction and Motivation III Human capital is not directly observed: problem of identification Alternative approaches and data sources: (1) Wage-based approach uses data on changes in wages to infer changes in quantities of human capital - classic Ben-Porath (2) Standard composition adjustment uses data on changes in education and experience (3) Job-skills-based approach uses data on changes in occupations Paper uses and compares estimates from (1) and (3); by definition (2) cannot capture the within cell increases for females the paper seeks to explore
Outline of the Presentation: Wage-based Approach Description of Wage-based approach and identification - (Ben-Porath style with cohort effects) - extension of Bowlus and Robinson (AER 2012) Discussion of three sources of cohort effects: (1) selection on ability in education choice; (2) technological change in human capital production functions; and (3) expected attachment to the labor force Cohort data on education and participation relevant for (1) and (3) and some discussion of (2) Estimates of life-cycle human capital (efficiency units supplied) profiles - (wage-based approach) - and comparison of cohort patterns for males and females Complications due to (time path of) discrimination in comparing males and females
Outline of the Presentation: Job-Skills-based Approach Description of Job-Skills-based Approach Problems of identification of life-cycle and cohort effects with the Job-Skills-based Approach Construction of a single dimension measure to compare with the single dimension measure of the wage-based approach Estimates of life-cycle (single dimension measure) profiles - (job-skills-based approach) - and comparison of cohort patterns for males and females
Outline of the Presentation: Summary and Next Steps Wage-based approach yields: pattern of life-cycle human capital profiles for females that shift up for the cohorts with higher labor market attachment for all education groups contrast with pattern for males where there is much less change in labor market attachment across cohorts Job-skills-based approach yields mostly similar qualitative pattern in the contrast between females and males, though different magnitudes Next steps: (1) refinement of both methods; (2) use of panel data including LISA; (3) estimate of under-estimation of true aggregate labor input of females
Wage-Based Approach: Framework Heterogeneous human capital - type determined by education group - separate prices Each education group has only one type - quantities can be aggregated within type (efficiency units) - similar to canonical model Ben-Porath style optimal life-cycle accumulation problem with a first stage choosing education group (type) (Heckman, Lochner and Taber (1997)); solution follows from comparison of marginal cost of production of human capital and marginal benefit Bowlus and Robinson (2012) framework with cohort effects extended to allow for variation in labor market attachment Identification of human capital quantities (efficiency units) via implementation of a flat-spot method based on the Bowlus and Robinson (2012)
Wage-Based Approach: Three Sources of Cohort Effects Two sources of cohort effects in the original Bowlus and Robinson (2012) framework: (1) heterogeneity in ability and correlation of ability and education level (effects through the production functions and marginal cost) (2) technological change, broadly interpreted, in the production of the different types of human capital (again, effects through the production functions and marginal cost) Third source of cohort effects particularly important for this paper: increased incentive for females to invest in human capital following the large secular increase in their participation rates (effect through marginal benefit)
Wage-Based Approach: Data Patterns Relevant for Cohort Effects Data relevant for cohort effect (3) through increased marginal benefit for cohorts with greater labor market attachment are life-cycle participation patterns by cohort Data relevant for cohort effect (1) through heterogeneity in ability and correlation of ability and education level are completed education levels by cohort More difficult for cohort effect (2) through technological change in human capital production (currently inferred in Bowlus and Robinson (2012) and Agopsowicz, Bowlus and Robinson (2016)) Possibility of some direct data from other sources, e.g. Green and Riddell (2013), Barrett and Riddell (2016)
Female Life-Cycle Participation by Birth Cohort
Life-Cycle Participation by Cohort: Females Summary Participation rates are higher for more educated groups More recent cohorts for all education groups show higher participation levels at all ages Interesting changes in the life-cycle pattern of participation: For early cohorts the peak in participation rates does not happen until relatively late in the life-cycle Most dramatic difference in the shifts in profiles across the cohorts is that participation at the earlier part of the life-cycle increases For all but the dropouts, by the 1958 cohort the rate is at its peak and relatively flat shortly after the end of the formal education period Potential, therefore, for large cohort effects for females in all education groups
Male Life-Cycle Participation by Birth Cohort
Participation Rates by Age: Summary for Males By comparison with females: Males show only quite minor changes across cohorts Tendency for participation to fall slightly compared to the strong increase for females College graduates for all birth cohorts show flat participation rate at high level from their mid to late 20s until their mid-50s and still show participation rates of 80% or more until age 60 Some college males show same pattern but begin a slow decline earlier and start to fall below 80% by their late 50s High school graduates similar to some college except for more variation by cohort and dropouts show most cohort variation with lower participation for the most recent cohorts and generally lower participation at each age
Increasing Educational Attainment for Females
Lower Education Groups: Comparison of Males and Females
Higher Education Groups: Comparison of Males and Females
Increasing Educational Attainment for Females Completed education level shares of four education groups: dropouts, high school graduates, some college and college graduates for successive birth cohorts of females from 1931 to 1967 obtained from using observations on the same age group (point in the life-cycle), 31-35, for each cohort to control for life-cycle effects in reporting The two lowest education levels show declining shares, while the two highest education levels show increasing shares Overall, pattern of completed education levels for females relative to males shows the features expected from the increased incentive for females to invest in human capital implied by their increased labor force participation
Wage-Based Approach: Maintained Assumption for Male-Female Human Capital Comparisons There is no specifically male or female human capital - for example, female and male college graduates have the same type of college graduate human capital Same assumption made in constructing college graduate human capital in standard implementation of the canonical model using composition adjustment - males and females may have different amounts of the college graduate type human capital, but they do not have different types Same assumption in much of discrimination literature
Wage-Based Approach: Possible Distinction between Market and Home Human Capital Goldin (2006) discusses shift in the college majors for women: Whereas in 1970 a standard dissimilarity index for college majors between men and women exceeded 0.5, it fell to about 0.3 in 1985 Women s majors shifted from those that were consumption related to those that were investment related. There may be two types of human capital: one useful largely in market production (Goldin s investment major); and another more useful in home production (Goldin s consumption major) Shift in production at the university level as seen in major choice towards more market oriented human capital would show up as technological improvement in producing (market oriented) human capital for females relative to males This technological improvement would be tied to the pattern of participation increase and major shift Amend assumption to: there is no specifically male or female market oriented human capital
Wage-Based Approach: Identification of Male and Female Human capital The wage-based analysis in this paper uses the price series derived in Bowlus and Robinson (2012) from data on full time and full year males to back out the quantity of human capital profiles by cohort for both males and females In a standard competitive market males and females would face the same price for any given type of human capital - price series derived in Bowlus and Robinson (2012) based on data for males can be used to compute efficiency units of human capital for females Males used because their participation pattern for each cohort and across cohorts, and their (college graduates) education pattern across cohorts are important elements of the implementation of the flat-spot method Patterns for females make similar implementation of the flat-spot method on female data infeasible Complications from discrimination
Wage-Based Approach: Complications from Discrimination I Large literature that studies discrimination against females - can take many forms If it results in a different (lower) price for females for same type of human capital, male based price series would be an over-estimate of the price series females faced resulting in an under-estimate of the female human capital when the female wage is divided by this over-estimated price Using male price series not a problem if level of discrimination constant - changes in cohort pattern of efficiency units for females could still be identified and contrasted with changes in the cohort pattern for males Primary concern is that there may have been a secular decrease in discrimination resulting in secular decline in amount of under-estimation of female human capital, imparting an upward bias on estimated change in female human capital
Wage-Based Approach: Complications from Discrimination II Possible that the main effect of discrimination is not to put a large wedge on the price but rather to prevent females acquiring the levels of human capital acquired by males by reducing access to training and promotion opportunities or other barriers Also possible that a large part of the disappearance of the gender wage gap, at least on labor market entry for female college graduates is due to the shift in college major to produce the investment rather than consumption type human capital referred to by Goldin (2006) and that this shift could be in part a response to reduced barriers from discrimination In this case use of the price series based on males would still permit the estimation of changes in the actual amount of efficiency units supplied by females in the different cohorts even in the presence of declining discrimination
Wage-Based Approach: Data MCPS for the period 1964-2009, same as Bowlus and Robinson (2012) Provides source of annual data for large age range covering wide range of pre- and post-war cohorts Issues of time varying top coding and allocation treated same way as Bowlus and Robinson (2012) Allows for selection of full time and full year
Wage-Based Approach: Estimated Profiles and Cohort Differences The comparisons presented are for estimates based on the evidence for full time and full year males and females They are not designed to measure relative changes for females via changes in education or labor supply per se They are designed to examine whether a composition adjustment approach to measuring female human capital is likely to under count the growth in female human capital following a participation increase That is, we are looking for evidence of relative changes for females compared to males even within age and education group and full-time and full year workers
Wage Based Human Capital Profiles for Female
Wage Based Human Capital Profiles for Males
Wage-Based Approach: Cohort Pattern Differences in Male and Female Profiles The figures show a clear contrast for males and females: Apart from the dropout group, which for much of the period can be expected to have a negative selection effect based on the change in the fraction of the cohort that are dropouts, there is clear evidence of an upward shift in the profiles for females, consistent with the increased participation In contrast, male dropouts, high school graduates, and to a lesser extent the some college group show a decline in the recent cohorts For male college graduates there is an increase for the post 1949 cohorts, interpreted in terms of cohort effects from selection and technological change in the human capital production function in Bowlus and Robinson (2012) For female college graduates, however, there are clear increases across all three cohorts
Wage-Based Approach: Summary Estimates of Cohort Differences Bowlus and Robinson (2012) presented a simple measure of cohort differences in the profiles for male college graduates by regressing the log efficiency units from the previous plots for a particular age range (30-45) on a quadratic in age and cohort dummies The age range restriction means that all ages are represented for all cohorts used We extend this to males and females for all education groups and use the individual level data: the estimates of the cohort dummies are insensitive to the inclusion of the age quadratic the age quadratic is only necessary to pick up the life-cycle shape The cohort dummies are plotted separately for low and high education groups
Wage-based Cohort Patterns: Lower Education Groups
Wage-based Cohort Patterns: Higher Education Groups
Wage-Based Approach: Male-Female Comparison Results Clear evidence from wage-based approach of different cohort pattern for females consistent with increased labor market attachment Females human capital increases generally in education/experience cell, including lower education groups, unlike males Magnitudes are large: the 1961 cohort full time and full year college graduate is supplying on average over 20% more than her counterpart from the 1946 birth cohort Suggests conventional labor input measures result in significant under-estimate of the contribution of females to post-war growth
Job-Skills-based Approach: Description Infers the amount of human capital from occupation data instead of from wage data thus avoiding the problems associated with declining discrimination Growing literature uses measures of job skills to study human capital and wage patterns We assume skills can t be unbundled (Heckman and Scheinkman (1987), Firpo, Fortin and Lemieux (2013) - same skill types are held by all education groups but in different amounts and ratios Same objective function as in Ben-Porath framework - some optimal path for the bundle - but multi-dimensionality of bundle breaks simple link between quantity and wages In order to compare human capital from a multi-dimensional bundle of skills to the single dimension measure from the wage-based approach we scale the bundles along a single dimension
Data for the Job-Skills-based Approach Same as for the wage-based analysis - MCPS - except drop years of the data set before the start of three digit 1970 occupation coding (1970 coding began in 1971) Emphasis on consistency over time in the data source for occupations - four different census occupation coding periods in the data covering 1970, 1980, 1990 and 2000 census coding Strong similarity between 1980 and 1990 codes - more major differences between 1970 and 1980 and between 1990 and 2000 IPUMS project constructed a set of three digit occupation codes, based on census 1990 coding, that aims for consistency across 1970 through 2000 coding For consistency number of codes is reduced Final number of codes is less than the close to 500 in the later period original coding schemes, but still provides for a lot of variation with close to 400 codes
Job-Skills-based Approach: Definition of the Bundles Literature typically takes two steps to assign job skills to workers: (1) job skill ratings are taken from sources such as the DOT or ONET and average skill ratings for a low dimension vector of skills (skill portfolio) for each occupation are constructed (2) skill portfolios are assigned to individuals in the main workers data sets on the basis of the worker s 3-digit occupation code (Poletaev and Robinson (2008), Yamaguchi (2012), Gathman and Schonberg (2010), Bowlus, Mori and Robinson (2016).) Could make a grid of the separate skills from a low dimension skill portfolio obtained from (1) and treat these as the bundles Instead we treat each of the approximately 380 IPUMS consistent codes as a unique bundles of skills and bypass (1) so that scaling the bundles is the same as scaling the IPUMS occupations In practice, discretizing each skill in a three vector of skills and allowing for even a modest number of discrete categories for each skill soon results in a grid where there are as many points as unique occupations
Job-Skills-based Approach: Construction of Single Dimension Measure Analogous to composition adjustment approach to scaling different age/education cells to compute single total for a given type of human capital in the canonical model Standard implementation computes efficiency units for a single skill held by different age and education cells by pricing the cells using the wages of the cell averaged over the entire period Could price bundles to scale them and produce quantities on a single dimension in same way Preferred approach modifies this - does scaling using efficiency units for males from the wage-based approach Argument: male efficiency units are correctly estimated and can therefore provide a direct baseline quantity scaling for the bundles
Job-Skills-based Approach: Identification Wage-based method can capture both life-cycle and cohort effects on quantities of human capital - simply divide the observed wages across the life-cycle or across cohorts by the appropriate price For job-skills-based approach observed data across the life-cycle or across cohorts are occupations If there was a very fine grid of occupations in which all workers had identical skills within occupation we could capture changes in skills across the life-cycle and across cohorts by observing changes in these occupations Moreover, the price wedge between males and females for the same type of human capital could be estimated from wage differences within this fine grid of occupations Unfortunately, in practice the occupation codes and allocation of workers to these codes are very far from this ideal situation
Job-Skills-based Approach: Single Occupation Career Problem Example Suppose individuals start in an occupation corresponding to their initial bundle and then make investments that simply scale up the bundle The ideal fine occupation grid would represent sequences on job ladders such that each higher level bundle of any type corresponded to a different occupation - with enough occupation codes the growth in skills over the life-cycle or cohort changes could be inferred from occupation data Actual occupation coding is such that many careers in which the bundle does evolve through human capital investment are represented by a single occupation code rather than a sequence of codes on a job ladder (many professions such as lawyer, doctor, and professor) neither growth in skills over the life-cycle, nor differences in the human capital across cohorts of lawyers, can be identified from the occupation data
Job-Skills-based Approach: Summary Estimates of Cohort Differences The same regressions are run as for the wage based method with the log of efficiency units as the dependent variable and cohort dummies However, the log of efficiency units is now measured by the single dimension measure from the job-skills-based approach: each individual in the data set is assigned a value of efficiency units based on their 3 digit IPUMS consistent 1990 code The results are again insensitive to the inclusion of an age quadratic The cohort dummies are again plotted separately for low and high education groups
Skill-based Cohort Patterns: Lower Education Groups
Skill-based Cohort Patterns: Higher Education Groups
Job-Skills-based Approach: Male-Female Comparison Results for Low Education Groups Compared to the wage-based results, for the low income groups the male-female difference is still apparent, though there are some significant differences: The earlier cohorts for females again show improvement while for males their human capital declines However the later cohorts decline for both males and females, albeit the rate of decline is slower for females Job-skills-based approach can only capture cohort effects to the extent that cohorts change occupation patterns - cannot capture within occupation changes: magnitudes are, therefore, likely to be different pattern may be affected by differential changes in the amount of occupational change across cohorts (within education) for males and females
Job-Skills-based Approach: Male-Female Comparison Results for High Education Groups For the high education groups the results are qualitatively quite similar to the results from the wage-based approach, though again the magnitudes are different: Male college graduates show a very similar pattern to the wage-based estimates and the females show a similar continuous growth over cohorts Male some college group also show a similar pattern to the wage based estimates and their decline in the pre-war cohorts again contrast with the increase for females Unlike the wage-based estimates, however, the some college group for females shows no growth after the 1949 cohort
Outline of the Presentation: Summary and Next Steps The relative patterns for males and females are similar in both the wage-based and the job-skills-based approach For the higher education groups there is consistent evidence from both approaches of increased human capital in the more recent female cohorts Suggests that there is enough change in occupation patterns, especially for female college graduates, for the job-skills-based approach to capture at least some cohort effects wage-based approach suggests that there is also substantial within occupation increases in the human capital for females, but the potential bias induced by declining discrimination makes it difficult to assess the true magnitude Next steps: (1) refinement of both methods; (2) use of panel data including LISA; (3) estimate of under-estimation of true aggregate labor input of females