Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Skill Prices over the Business Cycle∗
Gonzalo Castex and Evgenia Dechter†
January 2013
Abstract
We examine cyclical behavior of aggregate skill prices and find that the price of school-
ing is weakly procyclical and price of experience is countercyclical. Distinguishing between
technology and neutral productivity shocks, shows that both generate the countercycli-
cal pattern of experience premium whereas procyclicality of schooling premium is driven
by technological innovations. Changing demographics and compositional effects do not
explain the cyclical patterns. To explain these findings, we extend the capital-skill com-
plementarity framework to incorporate education, experience and vintage capital. Em-
ploying predictions of human capital and match quality theories, the model generates
cyclical patterns of skill prices observed in the data.
JEL Classification: E24, J24, J31
Keywords: Returns to education, Returns to experience, Skill prices, Wage structure
1 Introduction
A considerable amount of research is dedicated to study the cyclicality of real wages. More
recent studies, following Stockman (1983) and Bils (1985), show that real wages are pro-
cyclical. The cyclicality of wage differentials has received less attention. On the other hand,∗We would like to thank Mark Bils, Richard Blundell, Denise Doiron, Jesus Fernandez-Villaverde, Michael
Keane, Bob Miller, James Morley, Valentyn Pancheko, Sergio Rebelo, Robert Shimer, Guillaume Vanden-broucke, as well as seminar participants at UNSW, University of Santiago, Central Bank of Chile for commentsand suggestions. We also thank Cristian Munoz for excellent research assistance.†Gonzalo Castex, Central Bank of Chile. Email: [email protected]. Evgenia Dechter, School of Eco-
nomics, University of New South Wales. Email: [email protected].
1
increases in education and experience premia in the 1970s, 1980s and 1990s were extensively
analyzed in the context of technological change and shifts in demand and supply of skills.
Numerous studies document these developments, analyze their causes and implications. Katz
and Autor (1999) provide an extensive review of this wage structure literature. In the con-
text of business cycles, Reder (1955) was the first to formally examine the cyclicality of wage
differentials; his findings suggest that the aggregate skill premium was countercyclical in the
1930s and 1940s. More recently, Keane and Prasad (1993), Young (2003), Lindquist (2004)
and Castro and Coen-Pirani (2008) show that the aggregate skill premium is acyclical or
weakly procyclical, primarily focusing on the education premium.
This study evaluates how schooling and experience premia co-move with the business
cycle. To analyze the cyclicality, we build on the wage structure literature and control for
potential channels that could shift the supply and demand for skills. For example, Katz
and Revenga (1989), Katz and Murphy (1992) and Murphy and Welch (1992) explain the
increases in returns to skills between the 1970s and early 1990s, focusing on skill-biased tech-
nological change and the associated shifts in demand for skilled labor, globalization pressures
and shrinking relative demand for less skilled, declines in unionization and in real mini-
mum wage that primarily affected lower skilled workers. In the real business cycle context,
Krusell, Ohanian, Rios-Rull and Violante (2000) (hereafter KORV), Lindquist (2004), Cas-
tro and Coen-Pirani (2008), in the capital-skill complementarity framework, also identify the
technological change as the main driving force behind the increase in schooling premium and
show that an increase in the relative supply of schooling has a negative effect.1
Less attention is addressed to the post-1990th developments in skill prices. Between the
late 1990s and 2000s there was a slowdown in the increase of schooling premium whereas the
experience premium has declined.2 Jeong, Kim and Manovskii (2012) examine the develop-
ments in wage structure for the entire 1970s - 2000s period. They argue that changes in age
composition of the labor force can explain the rising experience differentials in the 1970s and
1Greenwood, Hercowitz and Krusell (2000) suggest that technological change is the source of about 30%of output fluctuations, thus the co-movement of skill prices with the technological change captures some ofthe prices’cyclicality.
2Figure 2 summarizes these empirical findings.
2
1980s as well as the decline that began in the second half of the 1990s. Some studies also
argue that there is a relationship between the age distribution and the magnitude of cyclical
output volatility; see, for example, Jaimovich and Siu (2009) and Lugauer (2012). We control
for the changing age distribution and for other changing demographics in our analysis.
The estimations are performed in two steps. First, using data from the 1962 - 2010 March
Current Population Surveys (CPS), we estimate the prices of schooling and work experience,
using the Mincer (1974) earnings equation for every year in the sample. Next, we examine the
correlations between the estimates of skill premia and business cycle measures: real output per
capita and unemployment rate. At this stage we also test whether the technological change
and compositional changes in the distributions of observed and unobserved characteristics
can explain the developments in skill prices.3
Simple correlations show that the price of schooling is weakly procyclical and the price
of experience is countercyclical. Further analysis distinguishes between investment-specific
and neutral (or other) productivity shocks. In these specifications, the former is measured by
technological change and the latter by output or unemployment fluctuations. We find that
experience premium is negatively correlated with both types of shocks whereas schooling
premium is positively correlated with technological change and negatively with the neutral
shocks. Thus, both types of shocks generate the countercyclical pattern of experience pre-
mium whereas the weak procyclicality of schooling premium is driven by technological inno-
vations. Changing demographics, although important, do not explain the cyclical patterns of
skill prices. Average education and experience of the work force are negatively correlated with
the education and experience premia, respectively. Decreasing unionization rate is negatively
associated with education and experience premia.
Studies that analyze the cyclicality of wages emphasize the importance of changing work-
ers’quality distribution over the business cycle. Stockman (1983), Bils (1985) and others show
that the distribution of unobservable workers’characteristics changes over the business cycle,
and argue that these compositional changes can generate a countercyclical trend in wages.
3To construct a measure of technological change we follow Cummins and Violante (2002).
3
Keane and Prasad (1993) argue that it is also important to control for individual fixed ef-
fects when evaluating the cyclicality of returns to skills. We find that changes in unobserved
characteristics of the workforce do not explain the cyclical patterns of skill premia.
The weak procyclicality of schooling premium is in line with findings in previous studies.
Keane and Prasad (1993), Young (2003) and Castro and Coen-Pirani (2008) estimate that
the aggregate schooling premium is acyclical. Ziliak, Wilson and Stone (1999) find it to be
weakly procyclical. Lindquist (2004) argues that the weak procyclicality of the schooling
premium is in line with predictions of capital-skill complementarity theory. Lindquist also
shows that investment-specific technological improvements increase the schooling premium
whereas neutral productivity shocks decrease the schooling premium.
The existing literature does not address the cyclicality of experience premium, and to
the best of our knowledge, there is no single theoretical framework that reconciles the weakly
procyclical education premium and the countercyclical experience premium. Countercyclical
experience premium emerges in frameworks that incorporate job match quality or on-the-job
human capital accumulation. Literature on match quality predicts that older workers are
more likely to benefit from a better quality job match and, therefore, hold better contracts
when unemployment is relatively high. Human capital theory predicts that more experienced
workers are more likely to have a higher human capital stock, which might be costly to acquire;
therefore, they are less likely to lose their jobs in a recession. Both theories predict that firms
are more likely to retain workers who have more job experience during recessions and that
firms will offer contracts with smoother wage profiles to more experienced workers.
To explain the empirical findings, we propose an extension to the capital-skill comple-
mentarity model, as in KORV, to incorporate education and experience in a vintage capital
framework. We employ predictions of human capital and match quality theories to make
assumptions about complementarity between experience of labor and vintage of capital. Hu-
man capital and match quality theories predict that more experienced workers have more
stable wage profiles and are less susceptible to job transitions, in our framework this implies
a higher complementarity of vintage capital with experienced labor than with inexperienced
4
labor. KORV estimates that schooling is complementary with capital equipment and tech-
nological innovations. We show that given these complementarity assumptions, our model
generates cyclical patterns of skill prices observed in the data.
The remainder of this paper is organized as follows. Section 2 describes and summarizes
the data. Section 3 discusses empirical methods and reports results. Here we also examine
the magnitudes of potential composition biases and discuss the practical relevance of our
findings. Section 4 outlines the theoretical framework to explain the cyclical variation of
schooling and experience premia. Section 5 concludes the paper.
2 Data
To examine the developments in wage structure, we use data from the March Consumer
Population Surveys (CPS) for 1962 through 2010. The raw sample contains approximately
5.5 million observations, 50,000 to 160,000 observations per year. The subsample used here
includes white males between 18 to 62 years old, who work full time, not self-employed, not
in the armed forces and not enrolled in school. The sample is also restricted to individuals
who were employed for at least 50 weeks in the calendar year prior to the March CPS year.
This restriction is imposed due to data collection methods. Prior to 1975 several variables are
reported in intervals. Weeks worked last year are reported in 6 categories, (the intervals are 1-
13, 14-26, 27-39, 40-47, 48-49 and 50+), to minimize potential measurement errors we choose
individuals with 50 to 52 weeks worked. Our subsample contains 1.1 million observations,
9,000 to 30,000 observations per year, as reported in column (2) of Table 8. Any further
selection criteria is explicitly noted.
Main variables of interest are real hourly wage, education and work experience. March
CPS data contain a wide range of information on labor market outcomes. Some questions
in March CPS refer to usual last year activity and some to labor market activity during the
previous week. Weeks worked and income from wages and salaries refer to last year. There
are two variables that report hours worked, usual hours worked per week refer to last year
but available only for 1976 - 2010, and hours worked last week which are available for the
5
entire period.4 We construct a measure of hours worked per week last year for the entire
sample period based on how they project on hours worked last week, for the available years.
Hourly wage rates are calculated using the last year annual income, divided by 52 and by
the projected hours worked per week. Because the income, weeks and hours data refer to
last year, the actual period covered is 1961 through 2009. Wage rates are expressed in 1999
prices, deflated using the Consumer Price Index.
To construct the schooling variable, we use individual information on highest grade
completed and transform it into years of schooling. Schooling information is not available
in 1962; therefore, this year is excluded from the analysis. The March CPS does not allow
constructing actual work experience, therefore, we apply the standard practice to calculate
the potential work experience: age minus years of education minus 6.5
To measure the business cycle, we use real GDP per capita and the unemployment rate
for 25 to 54-year-old individuals.6 To measure demographic transitions, we employ the CPS
data to calculate the mean experience and the mean education levels of white males between
18 to 62 years old in the labor force.
To obtain a measure of technological change, we follow methodology that was proposed
in Cummins and Violante (2002). The speed of technical change for each capital good in
the equipment and software category (E&S) can be measured as the difference between the
growth rate of constant-quality nondurable consumption and the growth rate of the good’s
quality-adjusted price. Price indexes are from the National Income and Product Accounts
(NIPA).7
4An intervaled measure of usual hours worked per week last year is also available. We do not use thisvariable to minimize potential biases associated with measurement errors.
5Using the potential work experience instead of actual work experience is also the main reason to restrictour sample to males only. There were significant changes in fertility and female labor force participationduring the last five decades.
6GDP data was downloaded from The Conference Board, Total Economy Database,http://www.conference-board.org/data/economydatabase/#Real_GDP. Unemployment data was obtainedfrom the Bureau of Labor Statistics (BLS) database, http://www.bls.gov/cps/data.htm.
7We retrieve data from Tables 1.1.4 and 5.3.4. of the NIPA series to obtain price indexes for nondurableconsumption and equipment and structures (E&S), respectively. For further discussion on construction ofindexes see Cummins and Violante (2002).
6
2.1 Summary Statistics
There were many important changes in the labor market over the last 50 years. We summarize
the developments in detrended real GDP per capita, unemployment rate, potential experience,
education, unionization rate and technological change. These series are depicted in Figure 1,
Table 1 reports correlations between the key variables. Unemployment rates vary significantly
during the 1961– 2009 period and strongly negatively correlated with the detrended GDP.
Mean experience data show the importance of the baby boom cohorts’entry into the labor
market. The average level of schooling is increasing throughout the 1961 - 2009 period and
unionization rate is falling. Figure 1 also reports the three aggregate technological change
measures, the indexes show a substantial decline in technical growth in the 2000s.
3 Econometric Analysis
To examine how labor market rewards productive attributes we use Mincer (1974) wage
equation. The terms "price of experience" and "price of education" refer to the effects of
experience and education on log wage rate. We perform the analysis in two steps. First, for
each year in the sample, using OLS or quantile regression estimation routines, we estimate the
prices of education and experience, controlling for various personal characteristics. Second,
we evaluate the determinants and cyclical variability of skill prices.
For first stage estimations, consider a year t cross-sectional regression of log real wage
rates of individual i on labor market experience, education and other personal characteristics,
logwit = β1tExperienceit + β2tExperience2it + β3tSchoolingit +Xitγt + εt, (1)
where Xit includes a constant, marital status and metro status, εt summarizes the mea-
surement error in the data. Some specifications include spousal schooling level and spousal
ranking in the schooling distribution to proxy for unobserved skills, employing the positive
assortative matching theory, (see for example Lam, 1988). We estimate this equation for
1961 - 2009 using the OLS. Alternatively, to control for compositional changes, we estimate
7
(1) for the 75th percentile using the quantile regresion technique, assuming that those at
higher quantiles are less affected by compositional changes associated with the business cy-
cle, (see for example Lindquist, 2004). Prices of education and experience are defined as
follows, PSCHOOLt = β̂3t and PEXPt = β̂1t + 2β̂2tExperiencet, where Experiencet is the
mean experience level of the workers in the sample in year t.8
The second stage estimations analyze the cyclicality of skill prices,
PEXPt = η1Yt + η2Zt + υt, (2)
PSCHOOLt = µ1Yt + µ2Zt + ωt, (3)
where Yt is a business cycle measure in year t, Zt is a set of control variables and υt, ωt are
uncorrelated measurement errors. For the second stage estimations, throughout the paper,
we report Newey-West robust standard errors with two lags to adjust for serial correlation in
residuals.
Most estimations of (2) and (3) include the technological change index and therefore dis-
tinguish between investment-specific and neutral (or other) productivity shocks. For example,
Greenwood, Hercowitz and Krusell (2000) argue that technological change is the source of
about 30% of output fluctuations. Thus, when controling for the technological change, out-
put or unemployment reflect productivity changes not associated with investment-specific
technology innovations. Further discussion on differential effects of the two types of shocks
is provided in Section 4.
Skill prices, PSCHOOLt and PEXPt , are depicted in Figure 2. We also report series of
prices obtained using the actual number of weeks and hours worked (instead of interval and
projected values, respectively), these estimates are available for the 1975 - 2009 period. The
price of experience is increasing in the 1960s through 1980s but it declines in later years. The
price of schooling is increasing throughout the entire period, but the growth rate is changing
over time and it is higher in the 1980s - 1990s period. Cross-correlations between skill prices,
8The estimates of price of experience for the 75th percentile is obtained using a relevant measure ofExperiencet.
8
business cycle measures and key aggregate indicators are reported in Table 1. Correlations
between the price of schooling and business cycle measures are positive but not statistically
significant. The average price of experience is positively correlated with unemployment and
negatively with detrended output. The countercyclical relationships with lagged business
cycle measures are stronger. Lindquist (2004) makes a similar observation when estimating
the cyclical variability of education premium, arguing that this outcome is consistent with
the capital-skill complementarity theory. In estimations of equations (2) and (3) we use
lagged business cycle measures and report outcomes obtained using contemporaneous values
for comparison.9
Patterns summarized in Figure 2 and Table 1 do not control for important demographic
and economic developments that occurred over the 1961 - 2009 period. Several important
trends are documented in Figure 1: technological progress, increase in average schooling,
decline in unionization, and U-shaped average experience.10 One approach to control for
some of these developments is to construct detrended series of skill prices, depicted in Fig-
ure 3. The correlations between unemployment and skill prices in Figure 3 are 0.60 for
experience and -0.48 for education. Detrended series of wage rates obtained using standard
measures of weeks and hours worked are available for 1975 - 2009 period and show similar
patterns. Results in Figures 2 and 3 suggest that experience premium is countercyclical and
education premium is procyclical. In the regression analysis, alternatively to using the time
trend, we estimate equations (2) and (3) controlling for changing demographics and techno-
logical change, (distinguishing between shocks driven by technological innovations and other
productivity shocks).
Tables 2 and 3 report the results. Columns (1) and (5) report results controlling for the
time trend only. In these specifications, the price of experience is countercylical, 1% increase
in unemployment rate is associated with a 0.16% increase in the experience premium.11 The
9Tables A1 and A2 report estimation results using contemporaneous business cycle measures. Similarcyclical patterns for skill prices emerge.
10See for example Autor, Katz, and Krueger (1998), Bound and Johnson (1992), Mincer (1991), amongmany others, for further discussion on how these changes could affect the skill prices.
11Average price of experience is 1.3%, thus a 1% increase in unemployment rate is associated with 13%increase in experience premium.
9
schooling premium is weakly procyclical. The remaining columns in Tables 2 and 3 report
results when adding control variables.12
The price of experience is countercyclical in all specifications but the magnitude of rela-
tionship is lower when adding more controls. Similarly to Jeong, Kim and Manovskii (2012)
we find a strong negative relationship between the price of experience and supply of experi-
ence, however the changing supply is not suffi cient to explain the countercyclical pattern.13
Unionization rate is negatively correlated with the experience premium, consistent with find-
ings in earlier literature. For example, DiNardo, Fortin, and Lemieux (1996), Freeman and
Katz (1996) and Lee (1999), discuss how the decline in unionization has changed the wage
setting norms and affected skill prices. Specifications in columns (2) - (4) and (6) - (8) include
the inverse quality adjusted relative price of equipment to control for technological change.
The results suggest that there is a negative relationship between the price of experience and
both neutral and investment-specific productivity improvements.
Columns (1) and (5) in Table 3 suggest that education premium is procyclical. Other
studies, Ziliak, Wilson and Stone (1999) and Lindquist (2004) find it to be weakly procyclical
while Keane and Prasad (1993), Young (2003) and Castro and Coen-Pirani (2008) estimate it
to be acyclical. Including the technological change measure in columns (2) - (4) and (6) - (8)
allows to distinguish between investment-specific productivity and neutral shocks. Results
show that investment-specific technological improvements increase the schooling premium
whereas neutral productivity improvements decrease the schooling premium, in line with
Lindquist (2004), who shows a similar outcome using the capital-skill complemetarity frame-
work. The relationship between unionization and the price of schooling is negative and in line
with other studies. The average education level has a negative association with education
premium, we interpret this result a negative supply effect, as for example discussed in Katz
12The estimates are robust in specifications that use contemporaneous business cycle measures, reportedin Tables A1 and A2.
13This negative relationship can be partially induced by using the average experience to construct experiencepremium. Table A3 reports estimation results using experience premium for a representative worker with 20years of experience, the average potential experience of the labor force calculated over the entire period. Theresults are quite similar to those reported in Table 2, implying that the outcome is not driven by constructionof experience premium.
10
and Murphy (1992).14
3.1 Composition effects
Using equations (2) and (3) allows to include controls for changing demographics, but does
not control for changing distributions of unobserved characteristics. Stockman (1983), Bils
(1985) and others show that cyclical changes in the work force composition may induce
a countercyclical bias in the aggregate wage. Aggregate measures of real wages tend to
give more weight to low-skill workers during expansions than during recessions because less-
productive workers are more vulnerable to layoffs in recessions than more productive workers.
Keane and Prasad (1993) find a weak countercyclical bias in experience and schooling wage
differentials.
We use three approaches to evaluate the importance of changing unobserved character-
istics of workers. First, we employ the positive assortative matching theory and estimate
equation (1) controlling for spousal schooling and spousal ranking in the schooling distrib-
ution to proxy for unobserved skills (see for example Lam, 1988). The second stage results
are reported in Tables 4 and 5 and are very similar to those in Tables 2 and 3, the price
of experience is strongly countercyclical and the price of schooling is positively correlated
with technological change but negatively with with neutral productivity improvements. Sec-
ond, we measure skill prices at the 75th percentile of the wage distribution, assuming that
employment is less sensitive to business cycle fluctuations for workers at the high end of
wage distribution. Results are reported in Tables 6 and 7, the cyclicality patterns are very
similar to those obtained using the OLS method in the first stage.15 Third, we compare
estimation outcomes obtained for nearby quantiles (45th, 50th and 55th), assuming that if
there was a shift in the unobserved skills distribution, an individual would not move too far
from the original location. We estimate equation (1) for the 45th, 50th and 55th quantiles
14To examine the robustness of estimates, we estimate equation (2) by education level and equation (3)by experience level. Tables A5 and A6 report the results and show that the cyclical patterns hold for eachsubgroup.
15Appendix Table A.4 reports results for the 75th percentile for price of experience constructed usingaverage experience for the entire period instead of annual averages, the cyclicality patterns are very similar.
11
for t = 1961, ..., 2009, then we test whether βq451t = βq501t = β
q551t , β
q452t = β
q502t = β
q552t and
βq453t = βq503t = β
q553t . For most years t-tests fail to reject the null at the 5% significance level.
Table 8 summarizes the results. There are 5 years for which β1t’s are statistically different,
1 year for β2t’s and 9 years for β3t’s. The probability that the coeffi cients are different is not
correlated with the unemployment rate, we conclude that the effects of composition changes
on estimated skill premia are relatively small.
3.2 Discussion
We find that the price of experience is countercyclical and the price of schooling is positively
related with investment-specific technological improvements and negatively with neutral pro-
ductivity shocks. Schooling and experience measure different types of human capital. Classi-
cal human capital theory suggests that wages reflect the full marginal product from general
training but respond partially to specific human capital (Becker, 1962; Oi, 1962). Therefore,
the costs of specific training are shared between workers and firms while employees bear all
the costs of investment in general human capital.16 To incorporate job match quality and
to distinguish between general and specific human capital in the empirical analysis, equation
(1) can be rewritten as follows,
logwit = λ1tSchoolingit + λ2tTijt + λ3tExpit + θi + ηijt + ξit, (4)
where wit is wage of person i at time t. Tijt is tenure on current job j and measures the
stock of specific human capital. The term Expit is the total work experience, that increases
regardless of the firm at which the worker is employed, while tenure (Tijt) only increases for
job stayers. Thus, Schoolingit and Expit may (partially) reflect the stock of general human
capital. There are three error terms, person specific (θi), match or firm specific (ηijt), and
an idiosyncratic term (ξit). The model is written as linear for expositional purposes. The
literature suggests that Tijt likely positively correlated with ηijt, see for example Altonji
and Shakotko (1987) and Topel (1991). Following the classical human capital theory and
16See for example Hashimoto (1981), Acemoglu and Pischke (1999) and Loewenstein and Spletzer (1999).
12
assuming costly on-the-job training, the larger the ηijt the less likely a firm is to dismiss
a worker. Workers with better matches also tend to have higher tenure since they are less
likely to receive a better outside offer. We would also generally expect ηijt and Tijt to be
positively correlated with experience (Expit), because workers who have been longer in the
labor market would have had more chances to get a better match draw and spend longer on
that job. We do not estimate the extended model as in Equation (4) and do not incorporate
specific human capital measures due to data limitations. In our simplified framework, the
estimated price of experience sums up the returns to tenure and match quality.
It is well established in the literature that the likelihood of job separation is nega-
tively correlated with tenure, see for example Parsons (1972), Mincer and Jovanovic (1981),
Abraham and Farber (1987). Table 9 reports the employment-to-employment, employment-
to-unemployment and job-to-job transitions by work experience, higher work experience is
associated with higher job stability. Another stream of literature, following Bils (1985), shows
that wages of newly hired workers are more procyclical than the wages of workers who stay in
their jobs. Hagedorn and Manovskii (2010) analyze the cyclicality of match-quality and find
it to be procyclical. Thus, there is a cyclical relationship between costly specific job training,
separation probability and match quality. Less experienced workers are more likely to lose
jobs when unemployment is high and accept job offers with a lower match quality. When
unemployment is low, these workers are more likely to move to jobs with a higher match
quality. Such mechanism generates a countercyclical price of experience.
Lindquist (2004) shows that the procyclicality of schooling premium is in line with pre-
dictions of capital-skill complementarity theory, in which the demand for skills is procyclical
and leads to a procyclical trend in the price of schooling. In the following section we offer
a stylized model that incorporates predictions of human capital and match quality theories
into the capital-skill complementarity framework. We show that such model can generate a
countercyclical price of experience and (weakly) procyclical price of schooling.
13
4 A Stylized Model
This section proposes a theoretical framework to explain the cyclical variation of schooling
and experience premia. We extend the capital-skill complementarity model as in KORV,
incorporating education and experience in a vintage capital framework. Our assumptions
about complementarity between different types of labor and capital are based on predictions
of human capital and match-quality theories.
The production process requires four types of labor inputs and three types of capital.
The four types of workers are non-educated inexperienced (Ut), non-educated experienced
(UXt), educated inexperienced (Et) and educated experienced (EXt). The capital inputs are
capital structures (kst), new capital equipment (kent), and vintage capital equipment (keot).
The stock of capital evolves in the standard way, incorporating the possibility that some
fraction of new capital equipment becomes vintage capital equipment.17
The production function is Cobb-Douglas in capital structures and a combination of CES
functions of the remaining factors of production:
G(Ωt) = kαst
[βF γU + (1− β) (τF
σE + (1− τ)F σEX)
γσ
] 1−αγ
(5)
where FU ={λ1U
θt + (1− λ1)UXθt
} 1θ ; FE = {λ2kηnet + (1− λ2)E
ηt }
1η and
FEX = {λ3kµoet + (1− λ3)EXµt }
1µ . The parameters α, β, τ , λ1, λ2, λ3 ∈ (0, 1) govern income
shares and θ, γ, σ, η, µ ∈ (−∞, 1) govern elasticities of substitution.
Equation (5) simplifies into KORV production function if there is no difference in degrees
of substitution within each production factor FU , FE and FEX , i.e., if θ = 1, and σ = η = µ.
On the other hand, σ > η and σ > µ imply that the substitution within FE and FEX is
higher than between the components of FE and FEX . For simplicity of the analysis we assume
one type of unskilled workers by setting θ = 1.
Assuming perfectly competitive factor markets, taking derivatives of equation (5) with
respect to labor inputs (U , E and EX), yields the skill prices (WE , WEX and WU ). Experi-
17A detailed description of the model is provided in Appendix A.
14
ence premium(WEXWE
)and education premium
(WEWU
)are defined as follows,
WEXWE
= κ1[λ3( keoEX )
µ+(1−λ3)]
σ−µµ
[λ2( kenE )η+(1−λ2)]
σ−ηη
(EXE
)σ−1(6)
WEWU
= κ2
[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−σσ×[λ2(kneE
)η+ (1− λ2)
]σ−ηη (E
U
)γ−1 (7)where κ1 =
(1−τ)(1−λ3)τ(1−λ2) and κ2 =
(1−β)τ(1−λ2)β .
4.1 Qualitative analysis
To analyze the cyclical properties of skill prices we make assumptions about the elasticity of
substitution parameters. We assume that σ > η and σ > µ, i.e. a higher complementarity
between new equipment and educated labor or between vintage equipment and experienced
labor than between new equipment and experienced labor or vintage equipment and inex-
perienced labor. Human capital and match quality theories predict that more experienced
workers have more stable wage profiles and are less susceptible to job transitions. As dis-
cussed in Section 3.2, this occurs because more experienced workers accumulate job/match
specific human capital which increases productivity, in our framework this implies higher
complementarity between experience and vintage capital, i.e., σ > µ. KORV estimates that
education is complementary with capital equipment, Lindquist (2004) shows that education
premium comoves with investment-specific innovations, i.e., σ > η.
Education and experience premia in (6) and (7), are functions of keoEX ,kneE and
EXE ,
EU .
Taking derivatives of skills premia with respect to capital-skill ratios we obtain:
∂(WEXWE
)∂( keoEX )
> 0 if σ > µ; and∂(WEXWE
)∂( kneE )
< 0 if σ > η (8)
∂(WEWU
)∂( kneE )
> 0 if γ > σ and σ > η; and∂(WEWU
)∂( keoEX )
> 0 if γ > σ (9)
15
Changes in keoEX andkneE drive capital—skill complementarity effects. In (8), if σ > µ and
σ > η, a rise in keoEX or decline inkneE increase the experience premium. In (9), a rise in
kneE
has a positive effect on schooling premium if γ > σ and σ > η. Where γ > σ implies higher
comlementarity between composites of FE and composites of FEX , than between those and
unskilled labor. If γ > σ, we also obtain a positive relationship between keoEX and the price of
schooling. The ratios EXE andEU in equations (8) and (9) drive relative supply effects. A rise
in EXE reduces the experience premium and a rise inEU reduces the education premium.
To analyze the cyclicality of skill premia in the model we examine how keoEX andkneE move
over the business cycle. To measure the new capital equipment, kne, we use investment in
capital equipment and software over two years, to construct old capital equipment, keo, we use
undepreciated stock of capital equipment and software minus investment in two preceeding
years (using 1 or 2 preceeding years yield similar cyclical patterns). To construct EX we
use total hours worked by individuals with more than 15 years of education and with more
than 30 years of work experience, to construct E we use total hours worked by those with
more than 15 years of education and less than 10 years of experience.18 The correlation
between the unemployment rate and keoEX is 0.3278, between the unemployment rate andkenEX
is -0.3736. Correlation between the total capital stock (without substracting two preceeding
years of investment) is 0.2116.19
We conclude that keoEX is countercyclical whereaskneE is procyclical. Thus, controlling for
relative supply effects, equation (8) predicts a countercyclical experience premium. There is
an ambiguity about the cyclicality of education premium in equation (9) since both capital-
skill ratios have a positive effect. If the effect of kneE is stronger than the effect ofkeoEX , we
obtain a procyclical education premium.
Model predictions are consistent with empirical results reported in Section 3. We find
that experience premium is negatively correlated with neutral productivity shock, i.e. with
18Calculations are performed using data for 1975-2010. Data on capital equipment and software stockand investment are from Bureau of Economic Analysis, Table 4.1 and Table 2.7, respectively. Data on hoursworked is obtained from the CPS sample described in Section 2. Unemployment rate is for 25 to 54-year-oldindividuals, obtained from the Bureau of Labor Statistics (BLS) database, http://www.bls.gov/cps/data.htm.
19We find that keo and ken have similar cycliclal patterns as keoEX andkenEX, respectively.
16
output, or positively with unemployment rate, and negatively with technological change.
Education premium, in specifications that control for technological change, is negatively
correlated with neutral shock but positively with investment-specific technology innovations.
In the model, technological innovations have a direct effect on investment,(kenEX
), whereas
investment has a negative effect on experience premium and positive effect on education
premium. Effects of neutral, or other shocks, are summarized by the relationship between
the skill prices and vintage capital equipment, keoEX . This relationship is positive for both skill
prices, whereas keoEX is negatively correlated with output, (or positively with unemployment).
Relative supply effects in empirical estimations are given by average experience and average
education of the workforce and behave in line with the model predictions. Thus, our stylized
model can explain the cyclical patterns of skill prices observed in the data.
5 Conclusion
Using data from the 1962 - 2010 March Current Population Surveys, we analyze the cyclical
variability of skill prices. We control for a set of channels that may affect the prices directly
or that may shift the relative demand and supply of skills. These channels are extensively
studied in the wage structure literature. The dominant explanations focus on skill-biased
technological change and associated shifts in demand for skilled labor, decline in unioniza-
tion and changing age distribution. In the real business cycles context, in the capital-skill
complementarity framework, the technological change and relative supply of schooling are
also identified as main driving forces behind the developments in schooling premium.
Our results show that the price of schooling is weakly procyclical and the price of expe-
rience is countercyclical. In specifications that distinguish between investment-specific and
neutral (or other) shocks, we find that experience premium is negatively associated with both
types of shocks whereas schooling premium is positively correlated with technological change
(that measures the investment-specific shock) and negatively with the neutral shocks. Thus,
both types of shocks generate the countercyclical pattern of experience premium whereas the
weak procyclicality of schooling premium is driven by technological innovations. Changing
17
demographics and cyclical changes in the workforce composition, although important, do not
explain the cyclical patterns of skill prices.
To our knowledge, there is no single theoretical framework that reconciles the weakly
procyclical education premium and countercyclical experience premium. In the capital-skill
complementarity theory, the demand for skills is procyclical and yields procyclical returns to
schooling but cannot explain the countercyclical experience premium. The countercyclical
experience premium emerges in frameworks that incorporate job match quality or on-the-job
human capital accumulation. The former predicts that older workers are more likely to have
a higher match quality and therefore hold better contracts when unemployment is relatively
high. The latter predicts that more experienced workers are more likely to have higher human
capital, which might be costly to acquire, and, therefore, less likely to lose jobs in recessions.
Both theories predict that firms are more likely to retain workers with higher job experience
and offer them contracts with smoother wage profiles.
To interpret the empirical findings, we incorporate the key outcomes of human capi-
tal and match quality theories into the capital-skill compementarity framework developed
in KORV. Previous studies that use the KORV framework focus on the role of capital-skill
complementarity in the increase and cyclicality of the schooling premium, (see for example
Lindquist, 2004). Our simple model builds on the KORV framework and explains develop-
ments in both schooling and experiece premia. We incorporate three types of labor: un-
skilled, skilled inexperienced and skilled experienced, and three types of capital: structures,
new equipment and vintage equipement. The complementarity between skill and capital
varies with work experience and vinatage of capital, we show that if more experienced work-
ers have higher comlementarity with vintage capital, as human capital and match quality
theories predict, the model generates a procyclical schooling premium and a countercyclical
experience premium.
18
References
Abraham, Katharine G. and Henry S. Farber. 1987. "Job Duration, Seniority, and Earnings,"
American Economic Review , Vol. 77, No. 3, pp. 278-297.
Acemoglu, Daron and Jorn-Steffen Pischke. 1999. "The Structure of Wages and Investment
in General Training," Journal of Political Economy, University of Chicago Press, vol. 107(3),
pp. 539-572.
Altonji, Joseph G. and Robert A. Shakotko. 1987 "Do Wages Rise with Job Seniority?,"
Review of Economic Studies, Vol. 54, No. 3, pp. 437-459.
Autor, David H., Lawrence F. Katz, and Alan B. Krueger. 1998. “Computing Inequality:
Have Computers Changed the Labor Market?,”Quarterly Journal of Economics, 113, pp.
1169-1213.
Becker, Gary S.1962. "Investment in Human Capital: A Theoretical Analysis," Journal of
Political Economy, Vol. 70, No. 5, Part 2: Investment in Human Beings, pp. 9-49.
Bils, Mark J. 1985. “Real Wages over the Business Cycle: Evidence from Panel Data,”Journal
of Political Economy, 93, pp. 666—89.
Bound, John and Johnson, George. 1992. "Changes in the Structure of Wages in the 1980’s:
An Evaluation of Alternative Explanations," American Economic Review, vol. 82 (3), pp.
371-92.
Castro, Rui and Daniele Coen-Pirani. 2008. "Why Have Aggregate Skilled Hours Become
So Cyclical Since the Mid-1980s?", International Economic Review, Volume 49, Issue 1, pp.
135—185.
Cummins, Jason G. and Giovanni L. Violante. 2002. "Investment-specific technical change in
the United States (1947—2000): measurement and macroeconomic consequences," Review of
Economic Dynamics, 5 (2), pp. 243—284.
19
DiNardo, John, Nicole M. Fortin and Thomas Lemieux. 1996. "Labor Market Institutions
and the Distribution of Wages, 1973-1992: A Semi-Parametric Approach," Econometrica, 64
(5), pp. 1001-1044.
Freeman, Richard and Lawrence F. Katz. 1996. “Introduction and Summary”, in Freeman,
R.B. and Katz, L.F. (eds) Differences and Changes in Wage Structure, Chicago: The Uni-
versity of Chicago Press, pp. 1-22.
Greenwood, Jeremy, Zvi Hercowitz and Per Krusell. 2000. "The role of investment-specific
technological change in the business cycle". European Economic Review, 44, pp. 91—115.
Hagedorn, Marcus and Iourii Manovskii. 2012. “Job Selection and Wages over the Business
Cycle,”Working Paper.
Hashimoto, Masanori. 1981. "Firm-Specific Human Capital as a Shared Investment," Amer-
ican Economic Review, Vol. 71, No. 3, pp. 475-482.
Jaimovich, Nir and Henry E. Siu. 2009. “The Young, the Old, and the Restless: Demographics
and Business Cycle Volatility,”American Economic Review, 99 (3), 804—826.
Jeong, Hyeok, Yong Kim, and Iourii Manovskii. 2012. “The Price of Experience.”Working
Paper.
Katz, Lawrence F. and David H. Autor. 1999. “Inequality in the Labor Market.” In Orley
Ashenfelter and David Card, editors, Handbook of Labor Economics, Amsterdam and New
York: North Holland, 1998.
Katz, Lawrence F. and Kevin M. Murphy. 1992. “Changes in Relative Wages, 1963-1987:
Supply and Demand Factors,”Quarterly Journal of Economics, 107(1), pp. 35-78.
Katz, Lawrence F. and Ana Revenga. 1989. "Changes in the Structure of Wages: The United
States vs. Japan," Journal of the Japanese and International Economies, 3, pp.522-523.
Keane, Michael P. and Eswar S. Prasad. 1993.“The Relation Between Skill Levels and the
Cyclical Variability of Employment, Hours and Wages," IMF Working paper, 93/44.
20
Krusell, Per, Lee E. Ohanian, José-Víctor Ríos-Rull and Giovanni L. Violante. 2000. "Capital—
Skill Complementarity and Inequality: A Macroeconomic Analysis," Econometrica, 68,
pp.1029—1053.
Lam, David. 1988. “Marriage Markets and Assortative Mating with Household Public Goods:
Theoretical Results and Empirical Implications,”Journal of Human Resources, 23:4, pp. 462—
487.
Lee, David, S. 1999. “Wage Inequality in the U.S. in the 1980s: Rising Dispersion or Falling
Minimum Wage?,”Quarterly Journal of Economics, 114, pp. 977-1023.
Lindquist, Matthew. 2004. “Capital-Skill Complementarity and Inequality over the Business
Cycle,”Review of Economic Dynamics, 7, pp. 519—540.
Loewenstein, Mark A. and James R. Spletzer. 1999. "General and Specific Training: Evidence
and Implications", The Journal of Human Resources Vol. 34, No. 4, pp. 710-733.
Lugauer, Steven. 2012. "Estimating the Effect of the Age Distribution on Cyclical Output
Volatility across the United States," Review of Economics and Statistics, 94 (4), pp. 896—902.
Mincer, Jacob. 1974. "Schooling, experience and earnings", Columbia University Press, New
York.
Mincer, Jacob. 1991. "Human capital, technology, and the wage structure: what do time
series show?," NBER Working Paper, 3581.
Mincer, Jacob and Jovanovic, Boyan. 1981. "Labor Mobility and Wages, in Studies in Labor
Markets," edited by S. Rosen. New York: NBER, pp. 21-63.
Murphy, Kevin M. and Finis Welch. 1992. "The Structure of Wages," Quarterly Journal of
Economics, 107(1), pp. 285-326.
Oi, Walter Y. 1962. "Labor as a Quasi-Fixed Factor," Journal of Political Economy, vol. 70,
pp. 538-555.
21
Parsons, Donald O. 1972. "Specific Human Capital: An Application to Quit Rates and Layoff
Rates," Journal of Political Economy, Vol. 80, No. 6, pp. 1120-1143 .
Reder, Melvin W. 1955. "The Theory of Occupational Wage Differentials," American Eco-
nomic Review, Vol. 45, pp. 833-52.
Shimer, Robert. 2012. "Reassessing the Ins and Outs of Unemployment", Review of Economic
Dynamics, 15 (2), pp. 127-148.
Stockman, Alan C. 1983. "Aggregation Bias and the Cyclical Behavior of Real Wages."
Conference Paper, Cambridge, Mass.: NBER.
Topel, Robert. 1991. "Specific Capital, Mobility, and Wages: Wages Rise with Job Seniority,"
Journal of Political Economy, Vol. 99, No. 1, pp. 145-176.
Young, Eric R. (2003), "The wage premium: a puzzle," Working Paper, Florida State Uni-
versity.
Ziliak James P., Beth A. Wilson and Joe A. Stone. 1999. "Spatial dynamics and heterogeneity
in the cyclicality of real wages," Review of Economics and Statistics 81 (2), pp. 227—236.
22
23
45
67
unem
ploy
men
t
-2-1
01
2
1960 1970 1980 1990 2000 2010
detrended GDP unemployment
0.0
2.0
4.0
6.0
8.1
0.2
.4.6
d(PC
inde
x)
1960 1970 1980 1990 2000 2010
PC Violante-Cummins E&S
1111
.512
12.5
1313
.5m
ean
scho
olin
g
1960 1970 1980 1990 2000 2010
1819
2021
22m
ean
expe
rienc
e
1960 1970 1980 1990 2000 2010
.1.1
5.2
.25
.3co
llege
%
1960 1970 1980 1990 2000 2010
.1.1
5.2
.25
unio
n %
1960 1970 1980 1990 2000 2010
Summary Statistics
.008
.01
.012
.014
.016
.018
23
45
67
unem
ploy
men
t
1960 1970 1980 1990 2000 2010
unempl (t-1) price of exp (actual hours)price of exp (predicted hours)
Price of experience
.06
.07
.08
.09
.1.1
1
23
45
67
unem
ploy
men
t
1960 1970 1980 1990 2000 2010
unempl (t-1) price of school (actual hours)price of school (predicted hours)
Price of schooling
Edication and experience premia
23
-.00
50
.005
23
45
67
unem
ploy
men
t
1960 1970 1980 1990 2000 2010
unempl (t-1) price of exp (actual hours)price of exp (predicted hours)
Detrended price of experience
-.01
5-.
01-.
005
0.0
05.0
1
23
45
67
unem
ploy
men
t
1960 1970 1980 1990 2000 2010
unempl (t-1) price of school (actual hours)price of school (predicted hours)
Detrended price of schooling
Detrended education and experience premia
24
price of exp.
price of educ.
detrend. output
detrend. output (t-1)
unempl. rate
unempl. rate (t-1)
average exp.
average educ.
tech. change
(1) (2) (3) (4) (5) (6) (7) (8) (9)detrend. output -0.5797 0.1454
(0.0000) (0.3239)
detrend. output (t-1) -0.7696 0.1282 0.8048(0.0000) (0.3853) (0.0000)
unemployment 0.5072 0.0908 -0.8121 -0.5827(0.0002) (0.5393) (0.0000) (0.0000)
unemployment (t-1) 0.7300 0.0649 -0.6320 -0.8024 0.7464(0.0000) (0.6614) (0.0000) (0.0000) (0.0000)
average experience -0.9507 -0.3506 0.5051 0.6328 -0.4727 -0.6101(0.0000) (0.0146) (0.0003) (0.0000) (0.0007) (0.0000)
average education 0.5520 0.8641 -0.1693 -0.1643 0.3643 0.3330 -0.7028(0.0000) (0.0000) (0.2499) (0.2645) (0.0109) (0.0208) (0.0000)
tech. change 0.1467 0.9703 0.2116 0.2267 0.0896 0.0284 -0.3071 0.8723(0.3197) (0.0000) (0.1488) (0.1212) (0.5447) (0.8480) (0.0337) (0.0000)
union rate -0.4506 -0.9419 0.0227 0.0437 -0.2685 -0.2598 0.5834 -0.9619 -0.9354(0.0013) (0.0000) (0.8781) (0.7679) (0.0650) (0.0745) (0.0000) (0.0000) (0.0000)
Table 1: Correlations between skill prices, business cycle measures and aggregate indicators
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2tAvg. exp), the parameters are estimated using 1962-2010 (excl. 1963) CPS, N=48. Output is the real GDP per capita and unemployment is the unemployment rate for 25 to 54 years old individuals. Average experience and education refer to employed averages. Technological change is constructed using Cummins and Violante (2002) methodology for equipment and software (E&S) category. Standard errors in parentheses.
25
(1) (2) (3) (4) (5) (6) (7) (8)
output -0.0033 -0.0023 -0.0014 -0.0015 (0.0003) (0.0004) (0.0002) (0.0003)
unemployment 0.0016 0.0010 0.0005 0.0006 (0.0002) (0.0002) (0.0001) (0.0002)
tech. change -0.0025 -0.0003 -0.0027 -0.0035 0.0000 -0.0034 (0.0008) (0.0004) (0.0006) (0.0006) (0.0005) (0.0006)
avg. experience -0.0026 -0.0026 (0.0002) (0.0002)
avg. education -0.0053 -0.0028 (0.0013) (0.0011)
unionization rate -0.0820 -0.0958 (0.0270) (0.0299)
time trend 0.0015 0.0014 0.0009 0.0008 0.0000 0.0005 0.0001 0.0002 (0.0001) (0.0001) (0.0002) (0.0002) (0.0000) (0.0001) (0.0001) (0.0002)
cons -2.9023 -2.6920 -1.5699 -1.4918 -0.0829 -0.9861 -0.0829 -0.2851 (0.2518) (0.2192) (0.3416) (0.3548) (0.0712) (0.1571) (0.2037) (0.3025)
(1) (2) (3) (4) (5) (6) (7) (8)output 0.0015 -0.0035 -0.0046 -0.0018
(0.0014) (0.0013) (0.0019) (0.0014) unemployment -0.0013 0.0004 0.0002 -0.0006
(0.0006) (0.0006) (0.0007) (0.0007) tech. change 0.0122 0.0094 0.0117 0.0091 0.0106 0.0093
(0.0028) (0.0021) (0.0024) (0.0025) (0.0023) (0.0021) avg. experience -0.0024 -0.0027
(0.0015) (0.0016) avg. education -0.0198 -0.0083
(0.0093) (0.0096) unionization rate -0.1740 -0.2402
(0.0820) (0.0755) time trend 0.0006 0.0011 0.0029 -0.0001 0.0013 0.0001 0.0002 -0.0008
(0.0006) (0.0004) (0.0012) (0.0007) (0.0001) (0.0003) (0.0007) (0.0004) cons -1.2003 -2.2395 -5.5006 0.3053 -2.4554 -0.1041 -0.2100 1.6543
(1.2311) (0.8695) (2.3534) (1.3969) (0.1673) (0.6229) (1.2359) (0.8045)
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2tAvg. exp), the parameters are estimated using 1962-2010 (excl. 1963) CPS, N=48. Output is the real GDP per capita and unemployment is the unemployment rate for 25 to 54 years old individuals. Average experience and education refer to employed averages. Technological change is constructed using Cummins and Violante (2002) methodology for equipment and software (E&S). Newey-West standard errors (lag 2) in parentheses.
Table 2: Price of experience over the business cycle, OLS
Table 3: Price of schooling over the business cycle, OLS
26
(1) (2) (3) (4) (5) (6) (7) (8)
output -0.0030 -0.0023 -0.0013 -0.0016 (0.0003) (0.0004) (0.0002) (0.0002)
unemployment 0.0014 0.0008 0.0004 0.0004 (0.0002) (0.0002) (0.0001) (0.0002)
tech. change -0.0018 0.0005 -0.0020 -0.0030 0.0009 -0.0029 (0.0007) (0.0003) (0.0006) (0.0006) (0.0006) (0.0006)
avg. experience -0.0023 -0.0023 (0.0002) (0.0002)
avg. education -0.0034 -0.0008 (0.0012) (0.0011)
unionization rate -0.0740 -0.0950 (0.0236) (0.0299)
time trend 0.0013 0.0013 0.0006 0.0007 0.0000 0.0004 -0.0001 0.0001 (0.0001) (0.0001) (0.0002) (0.0002) (0.0000) (0.0001) (0.0001) (0.0002)
cons -2.6201 -2.4689 -1.1335 -1.3860 -0.0521 -0.8311 0.3155 -0.1357 (0.2202) (0.2162) (0.3226) (0.3071) (0.0715) (0.1494) (0.2131) (0.3135)
(1) (2) (3) (4) (5) (6) (7) (8)output 0.0021 -0.0024 -0.0020 -0.0010
(0.0013) (0.0012) (0.0017) (0.0011) unemployment -0.0014 0.0002 -0.0001 -0.0005
(0.0005) (0.0005) (0.0006) (0.0005) tech. change 0.0110 0.0119 0.0106 0.0088 0.0124 0.0090
(0.0021) (0.0024) (0.0018) (0.0018) (0.0021) (0.0017) avg. experience -0.0008 -0.0009
(0.0016) (0.0016) avg. education -0.0010 0.0045
(0.0098) (0.0090) unionization rate -0.1370 -0.1833
(0.0657) (0.0580) time trend 0.0004 0.0008 0.0006 -0.0002 0.0013 0.0001 -0.0006 -0.0006
(0.0005) (0.0004) (0.0012) (0.0006) (0.0001) (0.0002) (0.0007) (0.0003) cons -0.6635 -1.6007 -1.1109 0.4028 -2.4506 -0.1719 1.2458 1.1698
(1.0834) (0.7716) (2.2502) (1.1304) (0.1527) (0.4727) (1.1687) (0.6955)
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2tAvg. exp), the parameters are estimated using 1962-2010 (excl. 1963) CPS, N=48. Output is the real GDP per capita and unemployment is the unemployment rate for 25 to 54 years old individuals. Average experience and education refer to employed averages. Technological change is constructed using Cummins and Violante (2002) methodology for equipment and software (E&S). Newey-West standard errors (lag 2) in parentheses.
Table 5: Price of schooling over the business cycle, OLS (with spousal schooling controls)
Table 4: Price of experience over the business cycle, OLS (with spousal schooling controls)
27
(1) (2) (3) (4) (5) (6) (7) (8)
output -0.0038 -0.0035 -0.0020 -0.0019 (0.0005) (0.0007) (0.0006) (0.0007)
unemployment 0.0017 0.0011 0.0004 0.0003 (0.0003) (0.0004) (0.0003) (0.0004)
tech. change -0.0008 0.0026 -0.0013 -0.0030 0.0032 -0.0028 (0.0016) (0.0010) (0.0011) (0.0014) (0.0013) (0.0009)
avg. experience -0.0036 -0.0037 (0.0007) (0.0007)
avg. education -0.0059 -0.0015 (0.0036) (0.0042)
unionization rate -0.1586 -0.1947 (0.0378) (0.0411)
time trend 0.0017 0.0017 0.0008 0.0005 0.0001 0.0005 -0.0004 -0.0002 (0.0002) (0.0002) (0.0004) (0.0003) (0.0001) (0.0002) (0.0003) (0.0002)
cons -3.4034 -3.3323 -1.3612 -1.0121 -0.1378 -0.9059 0.8891 0.5196 (0.3893) (0.3850) (0.7068) (0.5907) (0.1004) (0.3544) (0.6035) (0.3351)
(1) (2) (3) (4) (5) (6) (7) (8)output 0.0004 -0.0039 -0.0047 -0.0023
(0.0013) (0.0014) (0.0018) (0.0015) unemployment -0.0008 0.0005 0.0002 -0.0005
(0.0006) (0.0006) (0.0008) (0.0008) tech. change 0.0108 0.0089 0.0103 0.0073 0.0101 0.0075
(0.0028) (0.0027) (0.0024) (0.0025) (0.0029) (0.0020) avg. experience -0.0016 -0.0019
(0.0021) (0.0021) avg. education -0.0132 -0.0015
(0.0116) (0.0122) unionization rate -0.1609 -0.2360
(0.0985) (0.0867) time trend 0.0011 0.0015 0.0027 0.0004 0.0013 0.0003 0.0000 -0.0005
(0.0006) (0.0004) (0.0012) (0.0008) (0.0001) (0.0003) (0.0009) (0.0004) cons -2.1007 -3.0150 -5.1658 -0.6607 -2.4768 -0.5899 0.1776 1.1373
(1.1460) (0.8601) (2.2618) (1.5280) (0.1536) (0.6179) (1.5749) (0.8404)
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2tAvg. exp), the parameters are estimated using 1962-2010 (excl. 1963) CPS, N=48. Output is the real GDP per capita and unemployment is the unemployment rate for 25 to 54 years old individuals. Average experience and education refer to employed averages. Technological change is constructed using Cummins and Violante (2002) methodology for equipment and software (E&S). Newey-West standard errors (lag 2) in parentheses.
Table 6: Price of experience over the business cycle, 75th quantile regression
Table 7: Price of schooling over the business cycle, 75th quantile regression
28
q=45% q=50% q=55% q=45% q=50% q=55%1961 8921 0.0595 0.0602 0.0614 0.0168 0.0177 0.01821963 9923 0.0580 0.0578 0.0592 0.0171 0.0178 0.01761964 9925 0.0603 0.0604 0.0611 0.0188 0.0194 0.01951965 20963 0.0650 0.0651 0.0644 0.0191 0.0190 0.01951966 13593 0.0616 0.0611 0.0627 0.0180 0.0184 0.01861967 21034 0.0653 0.0656 0.0666 0.0163 0.0164 0.01721968 21344 0.0619 0.0623 0.0623 0.0165 0.0165 0.01671969 20290 0.0634 0.0639 0.0651 0.0162 0.0166 0.01711970 19617 0.0626 0.0631 0.0645 0.0171 0.0175 0.01801971 18912 0.0641 0.0649 0.0654 0.0202 0.0205 0.02071972 19145 0.0612 0.0627 0.0632 0.0218 0.0215 0.02151973 19083 0.0621 0.0632 0.0625 0.0232 0.0233 0.02381974 18255 0.0599 0.0596 0.0606 0.0221 0.0225 0.02331975 18120 0.0658 0.0665 0.0677 0.0245 0.0243 0.02441976 21717 0.0648 0.0659 0.0659 0.0267 0.0268 0.02671977 21640 0.0666 0.0661 0.0656 0.0270 0.0275 0.02681978 21936 0.0669 0.0666 0.0673 0.0266 0.0268 0.02671979 25908 0.0636 0.0637 0.0629 0.0258 0.0262 0.02601980 25472 0.0673 0.0675 0.0676 0.0280 0.0283 0.02831981 22474 0.0672 0.0677 0.0692 0.0271 0.0276 0.02751982 21498 0.0751 0.0755 0.0751 0.0290 0.0286 0.02811983 21683 0.0794 0.0797 0.0790 0.0287 0.0282 0.02831984 22912 0.0796 0.0810 0.0814 0.0304 0.0301 0.03001985 22702 0.0856 0.0854 0.0852 0.0278 0.0285 0.02891986 22498 0.0871 0.0867 0.0872 0.0296 0.0294 0.02971987 23027 0.0847 0.0851 0.0844 0.0276 0.0283 0.02841988 21827 0.0864 0.0867 0.0881 0.0276 0.0279 0.02831989 24129 0.0916 0.0919 0.0915 0.0261 0.0260 0.02671990 23506 0.0908 0.0896 0.0911 0.0257 0.0263 0.02701991 22552 0.0942 0.0940 0.0946 0.0255 0.0260 0.02671992 22136 0.0980 0.0978 0.0976 0.0264 0.0262 0.02671993 21576 0.1010 0.1013 0.1006 0.0274 0.0279 0.02791994 21822 0.0996 0.0984 0.0990 0.0275 0.0280 0.02801995 20059 0.0990 0.0980 0.0988 0.0261 0.0275 0.02781996 20451 0.1030 0.1032 0.1033 0.0254 0.0254 0.02651997 20378 0.1004 0.1017 0.1024 0.0238 0.0241 0.02431998 21198 0.1028 0.1036 0.1035 0.0228 0.0232 0.02371999 21471 0.1046 0.1050 0.1046 0.0236 0.0238 0.02392000 33587 0.1004 0.1010 0.1014 0.0228 0.0227 0.02282001 32810 0.1031 0.1042 0.1051 0.0217 0.0223 0.02262002 31906 0.1045 0.1051 0.1060 0.0226 0.0233 0.02362003 31118 0.1042 0.1052 0.1063 0.0237 0.0241 0.02412004 30677 0.1042 0.1047 0.1059 0.0228 0.0233 0.02352005 31026 0.1043 0.1053 0.1062 0.0244 0.0245 0.02452006 30910 0.1047 0.1063 0.1081 0.0247 0.0251 0.02562007 30352 0.1040 0.1054 0.1063 0.0229 0.0227 0.02342008 29138 0.1053 0.1065 0.1079 0.0211 0.0215 0.02182009 27251 0.1085 0.1093 0.1109 0.0206 0.0206 0.0204
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2tAvg. exp), the parameters are estimated using 1962-2009 (excl. 1963) CPS. Controls include a constant, marital status and metro status. All reported coefficients are significant at the 5% level.
Table 8: Skill prices, 1961 - 2009, by quantile
Year Obs Price of schooling Price of experience
29
(1) (2) (3) (4) (5) (6) (7) (8)
output -0.0024 -0.0006 -0.0008 -0.0002 (0.0005) (0.0007) (0.0003) (0.0004)
unemployment 0.0010 0.0002 0.0003 0.0000 (0.0004) (0.0004) (0.0001) (0.0002)
tech. change -0.0043 0.0004 -0.0039 -0.0047 0.0002 -0.0041 (0.0012) (0.0006) (0.0007) (0.0009) (0.0007) (0.0005)
avg. experience -0.0031 -0.0029 (0.0003) (0.0003)
avg. education -0.0048 -0.0034 (0.0022) (0.0018)
unionization rate -0.1165 -0.1180 (0.0299) (0.0290)
time trend 0.0011 0.0009 0.0005 0.0002 0.0000 0.0007 0.0001 0.0002 (0.0003) (0.0002) (0.0003) (0.0002) (0.0000) (0.0001) (0.0002) (0.0001)
cons -2.1181 -1.7482 -0.8046 -0.4410 -0.0848 -1.3186 -0.0606 -0.3162 (0.4963) (0.4284) (0.5035) (0.4705) (0.0840) (0.2166) (0.3068) (0.2677)
(1) (2) (3) (4) (5) (6) (7) (8)output 0.0024 -0.0020 -0.0034 -0.0012
(0.0012) (0.0016) (0.0016) (0.0012) unemployment -0.0016 -0.0003 -0.0001 -0.0006
(0.0006) (0.0005) (0.0005) (0.0006) tech. change 0.0108 0.0120 0.0114 0.0079 0.0105 0.0092
(0.0029) (0.0026) (0.0024) (0.0022) (0.0023) (0.0019) avg. experience -0.0043 -0.0026
(0.0015) (0.0016) avg. education -0.0210 -0.0074
(0.0092) (0.0103) unionization rate -0.2056 -0.2267
(0.0737) (0.0729) time trend 0.0002 0.0007 0.0021 -0.0004 0.0013 0.0002 0.0002 -0.0007
(0.0006) (0.0005) (0.0009) (0.0005) (0.0001) (0.0003) (0.0007) (0.0004) cons -0.4066 -1.3447 -3.6976 0.9636 -2.4837 -0.4127 -0.1526 1.5122
(1.0930) (0.9984) (1.6979) (1.0789) (0.1617) (0.5548) (1.2179) (0.7957)
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2tAvg. exp), the parameters are estimated using 1962-2010 (excl. 1963) CPS, N=48. Output is the real GDP per capita and unemployment is the unemployment rate for 25 to 54 years old individuals. Average experience and education refer to employed averages. Technological change is constructed using Cummins and Violante (2002) methodology for equipment and software (E&S). Newey-West standard errors (lag 2) in parentheses.
Table A1: Price of experience over the business cycle, OLS, contomporeneous business cycle measures
Table A2: Price of schooling over the business cycle, OLS, contomporeneous business cycle measures
30
(1) (2) (3) (4) (5) (6) (7) (8)
output -0.0020 -0.0016 -0.0014 -0.0011 (0.0002) (0.0002) (0.0002) (0.0002)
unemployment 0.0010 0.0007 0.0005 0.0005 (0.0001) (0.0001) (0.0001) (0.0001)
tech. change -0.0011 -0.0005 -0.0012 -0.0017 -0.0001 -0.0017 (0.0004) (0.0004) (0.0003) (0.0004) (0.0005) (0.0004)
avg. experience -0.0016 -0.0015 (0.0002) (0.0002)
avg. education -0.0054 -0.0029 (0.0013) (0.0011)
unionization rate -0.0515 -0.0601 (0.0151) (0.0167)
time trend 0.0009 0.0009 0.0009 0.0005 0.0000 0.0002 0.0001 0.0000 (0.0001) (0.0001) (0.0002) (0.0001) (0.0000) (0.0001) (0.0001) (0.0001)
cons -1.7725 -1.6819 -1.6517 -0.9288 -0.0185 -0.4656 -0.1709 -0.0258 (0.1635) (0.1498) (0.3488) (0.2108) (0.0391) (0.0995) (0.2096) (0.1650)
(1) (2) (3) (4) (5) (6) (7) (8)output -0.0023 -0.0019 -0.0015 -0.0012
(0.0002) (0.0003) (0.0003) (0.0002) unemployment 0.0011 0.0007 0.0005 0.0004
(0.0001) (0.0002) (0.0001) (0.0001) tech. change -0.0010 -0.0002 -0.0012 -0.0019 0.0002 -0.0018
(0.0007) (0.0005) (0.0004) (0.0006) (0.0007) (0.0004) avg. experience -0.0017 -0.0017
(0.0002) (0.0003) avg. education -0.0052 -0.0023
(0.0014) (0.0016) unionization rate -0.0668 -0.0805
(0.0170) (0.0182) time trend 0.0010 0.0010 0.0009 0.0005 0.0000 0.0003 0.0000 0.0000
(0.0001) (0.0001) (0.0002) (0.0001) (0.0000) (0.0001) (0.0002) (0.0001) cons -1.9666 -1.8816 -1.6897 -0.9046 -0.0154 -0.5094 -0.0125 0.0797
(0.1651) (0.1503) (0.3911) (0.2488) (0.0441) (0.1484) (0.2769) (0.1606)
Table A3: Price of experience at 20 years of experience over the business cycle, OLS
Note: Price of schooling refers to the parameter β3t in Equation (1), price of experience referes to (β1t+2β2t*20), the parameters are estimated using 1962-2010 (excl. 1963) CPS, N=48. Output is the real GDP per capita and unemployment is the unemployment rate for 25 to 54 years old individuals. Average experience and education refer to employed averages. Technological change is constructed using Cummins and Violante (2002) methodology for equipment and software (E&S). Newey-West standard errors (lag 2) in parentheses.
Table A4: Price of experience at 20 years of experience over the business cycle, 75th quantile regression
31
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)output -0.0025 -0.0014 -0.0013 -0.0031 -0.0017 -0.0014 -0.0040 -0.0009 -0.0013 -0.0035 -0.0006 -0.0014
(0.0003) (0.0004) (0.0004) (0.0003) (0.0004) (0.0005) (0.0005) (0.0006) (0.0004) (0.0006) (0.0007) (0.0006)
tech. change 0.0011 -0.0010 0.0008 -0.0016 -0.0027 -0.0053 -0.0033 -0.0053(0.0009) (0.0007) (0.0009) (0.0008) (0.0009) (0.0010) (0.0012) (0.0011)
avg. experience -0.0025 -0.0030 -0.0023 -0.0013(0.0004) (0.0004) (0.0006) (0.0007)
avg. education -0.0057 -0.0076 -0.0029 -0.0001(0.0023) (0.0023) (0.0036) (0.0039)
unionization rate -0.0932 -0.1243 -0.0640 0.0027(0.0272) (0.0303) (0.0325) (0.0375)
time trend 0.0012 0.0008 0.0005 0.0014 0.0009 0.0004 0.0017 0.0008 0.0010 0.0014 0.0005 0.0012(0.0001) (0.0003) (0.0002) (0.0001) (0.0003) (0.0002) (0.0002) (0.0004) (0.0002) (0.0003) (0.0004) (0.0003)
cons -2.4512 -1.5451 -1.0253 -2.6915 -1.7042 -0.7709 -3.3023 -1.3979 -1.9339 -2.7015 -0.9176 -2.2933(0.2746) (0.6047) (0.3705) (0.2837) (0.5555) (0.4576) (0.4451) (0.7084) (0.4391) (0.5637) (0.7651) (0.6307)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)output 0.0015 -0.0070 -0.0020 0.0032 -0.0056 -0.0020 0.0024 -0.0031 -0.0006 0.0010 -0.0021 -0.0014
(0.0023) (0.0035) (0.0024) (0.0019) (0.0028) (0.0020) (0.0009) (0.0015) (0.0010) (0.0009) (0.0010) (0.0014)
tech. change 0.0150 0.0169 0.0105 0.0157 0.0041 0.0087 0.0077 0.0075(0.0037) (0.0038) (0.0033) (0.0030) (0.0021) (0.0018) (0.0022) (0.0023)
avg. experience -0.0045 0.0009 -0.0009 -0.0030(0.0031) (0.0021) (0.0011) (0.0014)
avg. education -0.0302 -0.0132 -0.0186 -0.0138(0.0173) (0.0125) (0.0073) (0.0094)
unionization rate -0.3903 -0.1345 -0.0543 -0.0684(0.1549) (0.1169) (0.0628) (0.0631)
time trend 0.0007 0.0038 -0.0014 0.0001 0.0032 -0.0002 0.0002 0.0030 0.0002 0.0004 0.0014 0.0003(0.0010) (0.0022) (0.0012) (0.0009) (0.0018) (0.0011) (0.0004) (0.0011) (0.0005) (0.0004) (0.0009) (0.0006)
cons -1.2440 -6.9375 2.9294 -0.1016 -6.2118 0.4965 -0.3833 -5.6236 -0.3408 -0.7954 -2.4183 -0.4519(2.0569) (4.2112) (2.4567) (1.6907) (3.3890) (2.2207) (0.8794) (2.0743) (0.9607) (0.8417) (1.6147) (1.2073)
Table A5: Price of experience over the business cycle, OLS, by experience level
0
A Model description, wage premia and busines cycle
Consider an economy with mass one of skilled and unskilled infinitely lived agents, where skills
can be either education or experience. A non-educated worker can be either experienced or
non-experienced, similarly, an educated worker can be either experienced or non-experienced.
Let U , UX, E and EX denote the agent-types measures respectively. Agents are born at
time zero and endowed with the skill type and one unit of time they allocate between leisure
and labor.
There are three types of capital: capital structures, kst; vintage capital equipment, keot;
and new capital equipment, kent. The stock of capital evolves in the standard way, but
incorporating the possibility that some fraction of new capital equipment becomes vintage
capital equipment. Firms make investment decisions on new capital equipment and capital
structures each period. The laws of motion for capital are as follows,
Capital structures:
kst+1 = (1− δs)kst + Ist. (10)
Capital equipment:
kent+1 = ν(1− δen)kent + Iet,
keot+1 = (1− δeo)keot + (1− ν)(1− δen)kent,(11)
We allow for different depreciation rates. Note that a fraction (1 − ν) of new capital
equipment becomes vintage capital equipment each period.
The production process requires four types of labor inputs and three types of capital:
non-educated workers with no experience (Ut), non-educated workers with experience (UXt),
educated and non-experienced workers (Et), and educated and experienced workers (EXt),
capital structures (kst), new capital equipment (kent), and vintage capital equipment (keot).
The capital skill complementarity model incorporates two sectors with two types of final
goods, consumption (ct) and capital structures (Ist), and capital equipment (Iet). The output
33
in each sector is given by:
ct + Ist = AtG(kst, keot, kent, Ut, UXt, Et, EXt) (12)
Iet = AtqtG(kst, keot, kent, Ut, UXt, Et, EXt) (13)
Consumption is denoted by ct, investment in new structures given by Ist, and investment
in the equipment sector by Iet. Inputs of the same factor in each sector are different, we avoid
using superscript indicators to easy notation.
Aggregate technology is neutral to both sectors, At; and there is a specific technology
factor for the capital equipment sector, denoted by qt. For notation simplicity, we define the
set of production factors as Ωt ≡ {kst, keot, kent, Ut, UXt, Et, EXt}.
The production function G is assumed to be homogeneous of degree one and similar in
both sectors. Assuming perfect competition, aggregate output of the economy is given by:
Yt = ct + Ist +Ietqt
= AtG(kst, keot, kent, Ut, UXt, Et, EXt) = AtG(Ωt)
The production function is Cobb-Douglas in capital structures and a combination of CES
functions of the remaining factor of production:
G(Ωt) = kαst
[βF γU + (1− β) (τF
σE + (1− τ)F σEX)
γσ
] 1−αγ
where the factor FU is defined as composites of non-educated labor with different experience
levels (FU ={λ1U
θt + (1− λ1)UXθt
} 1θ ); FE is a composite of production factors for new
capital equipment with educated workers (FE = {λ2kηnet + (1− λ2)Eηt }
1η ); and the term FEX
is a composite of production factors for vintage capital equipment with experienced workers
( FEX = {λ3kµoet + (1− λ3)EXµt }
1µ ).
The parameters α, β, τ , λ1, λ2, λ3 ∈ (0, 1) govern income share and θ, γ, σ, η, µ ∈
(−∞, 1) govern the elasticity of substitution. For simplicity we assume θ = 1.
We derive skill premia and obtain wage ratios as described in equations (6) and (7).
34
A.1 Cyclicality Analysis
Following the procedure described in section 4.1, we analyze how skill premia comove with
capital-skill labor ratio. Equations (14) and (15) shows the cyclicality of experience premium.
∂(WEXWE
)∂( keoEX )
= κ1(σ − µ)λ3[λ3( keoEX )
µ+(1−λ3)]
σ−2µµ
[λ2( kenE )η+(1−λ2)]
σ−ηη
(EXE
)σ−1 ( keoEX
)µ−1(14)
∂(WEXWE
)∂( kneE )
= −κ1(σ − η)λ2[λ3( keoEX )
µ+(1−λ3)]
σ−µµ
[λ2( kenE )η+(1−λ2)]
ση
(EXE
)σ−1 (knrE
)η−1(15)
The cyclicality pattern of educational premium is described as follows:
∂(WEWU
)∂( kneE )
= κ2(γ − σ)τλ2(λ2(kneE
)η+ (1− λ2)
) 2(σ−η)η (kne
E
)η−1 (EU
)γ−1×[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−2σσ+κ2(σ − η)λ2
[λ2(kneE
)η+ (1− λ2)
]σ−2ηη (E
U
)γ−1 (kneE
)η−1×[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−σσ
(16)
∂(WEWU
)∂( keoEX )
= κ2(γ − σ)(1− τ)λ3[λ2(kneE
)η+ (1− λ2)
]σ−ηη (E
U
)γ−1 (EXE
)σ ( keoEX
)µ−1×[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−2σσ×(λ3(keoEX
)µ+ (1− λ3)
)σ−µµ
(17)
Interpretations of the cyclical patterns of experience and educational premia are provided
in Section 4.1
A.2 Full Model: Differentiating between U and UX
In this section we make an extension of our model incorporating non-educated workers with
different experience levels. Results in terms of skill premia and cyclicality do not change.
35
A.2.1 Wages and Skill Premia
Considering the expended model, output is defined as:
G(Ωt) = Atkαst
[βF γU + (1− β) (τF
σE + (1− τ)F σEX)
γσ
] 1−αγ
Where
FU ={λ1U
θt + (1− λ1)UXθt
} 1θ
FE = {λ2kηnet + (1− λ2)Eηt }
1η
FEX = {λ3kµoet + (1− λ3)EXµt }
1µ
We derive skill premia for experience and education: first experience premium for edu-
cated and uneducated workers, then education premium for unexperienced and experienced
workers.
WEXWE
= κ̃1[λ3( keoEX )
µ+(1−λ3)]
σ−µµ
[λ2( kenE )η+(1−λ2)]
σ−ηη
(EXE
)σ−1(18)
WUXWU
= κ̃2(UXU
)θ−1(19)
WEWU
= κ̃3
[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−σσ× [λ2(
kneE )
η+(1−λ2)]
σ−ηη[
λ1( UUX )θ+(1−λ1)
] γ−θθ
(Eγ−1
Uθ−1UXγ−θ
) (20)
WEXWUX
= κ̃4
[τ(λ2(kneE
)η+ (1− λ2)
)ση ( E
EX
)σ+ (1− τ)
(λ3(keoEX
)µ+ (1− λ3)
)σµ
] γ−σσ
× [λ3(keoEX )
µ+(1−λ3)]
σ−µµ[
λ1( UUX )θ+(1−λ1)
] γ−θθ
(EXUX
)γ−1(21)
Where κ̃i is a positive constant.
36
A.2.2 Cyclicality analysis of the full model
We analyze how wage premium, previously derived, fluctuates on a business cycle frequency.
∂(WEXWE
)∂( keoEX )
= κ̃1(σ − µ)λ3[λ3( keoEX )
µ+(1−λ3)]
σ−2µµ
[λ2( kenE )η+(1−λ2)]
σ−ηη
(EXE
)σ−1 ( keoEX
)µ−1(22)
Since the vintage-capital to experienced-labor ratio is coutercyclical, if σ > µ we obtain
a countercyclical return to experience for educated workers. This is in line with a higher
complementarity between vintage capital equipment and educated-experienced workers than
with non-experienced ones.
The experience premium for educated workers when there is a fluctuation in the new-
capital to unexperienced-labor ratio, is as follow:
∂(WEXWE
)∂( kneE )
= κ̃1(η − σ)λ2[λ3( keoEX )
µ+(1−λ3)]
σ−µµ
[λ2( kenE )η+(1−λ2)]
ση
(EXE
)σ−1 (kenE
)η−1 (23)Since the new capital to unexperience labor is procyclical, if σ > η we obtain a counter-
cyclical return to experience. This result is in line with our complementarity assumptions.
Now we analyze the educational premium for non-experienced workers, when there is a
change in the capital-labor ratio as a consequence of economic fluctuations. We consider first
fluctuations on the ratio of new capital equipment to educated and unexperienced workers.
∂(WEWU
)∂( kenE )
= κ̃3τλ2(γ − σ)[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−2σσ× [λ2(
kneE )
η+(1−λ2)]
2(σ−ηη )[λ1( UUX )
θ+(1−λ1)
] γ−θθ
Eγ−1
Uθ−1UXγ−θ
(kneE
)η−1+κ̃3 (σ − η)
[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−σσ×λ2
[λ2( kneE )η+(1−λ2)]
σ−2ηη[
λ1( UUX )θ+(1−λ1)
] γ−θθ
Eγ−1
Uθ−1UXγ−θ
(kneE
)η−1(24)
As(kneE
)is countercyclical, the education premium for unexperienced workers is pro-
37
cyclical if σ > γ and σ > η, which confirms our complementarity assumptions between new
capital equipment and unexperienced and educated workers.
When analyzing the ratio of vintage capital equipment to experienced and educated
workers, the educational premium for unexperienced workers is as follow:
∂(WEWU
)∂( keoEX )
= κ̃3(γ − σ)[τ(λ2(kneE
)η+ (1− λ2)
)ση
+ (1− τ)(λ3(keoEX
)µ+ (1− λ3)
)σµ (EX
E
)σ] γ−2σσ×(1− τ)λ3
[λ2( kneE )η+(1−λ2)]
σ−ηη[
λ1( UUX )θ+(1−λ1)
] γ−θθ
(λ3(keoEX
)µ+ (1− λ3)
)σµ Eγ−1
Uθ−1UXγ−θ
(keoEX
)µ−1(25)
If σ > γ we obtain procyclical educational premium fon unexperienced workers.
Now we show the cyclical pattern of educational premium for experienced workers,(WEXUX
), analyzing first the ratio of new capital equipment to unexperienced and educated
workers.
∂(WEXWUX
)∂( kneE )
= κ̃4(γ − σ)λ2τ[τ(λ2(kneE
)η+ (1− λ2)
)ση ( E
EX
)σ+ (1− τ)
(λ3(keoEX
)µ+ (1− λ3)
)σµ
] γ−2σσ
× [λ3(keoEX )
µ+(1−λ3)]
σ−µµ[
λ1( UUX )θ+(1−λ1)
] γ−θθ
(EXUX
)γ−1 (λ2(kneE
)η+ (1− λ2)
)σ−ηη ( E
EX
)σ (kneE
)η−1(26)
To obtain the procyclical pattern in education premium for experienced workers we need
σ > γ. The wage ratio also fluctuates with changes in the ratio of vintage capital equipment
38
to educated-experienced workers:
∂(WEXWUX
)∂( keoEX )
= κ̃4(γ − σ)[τ(λ2(kneE
)η+ (1− λ2)
)ση ( E
EX
)σ+ (1− τ)
(λ3(keoEX
)µ+ (1− λ3)
)σµ
] γ−2σσ
×λ2(1− τ)λ3[λ3( keoEX )
µ+(1−λ3)]
2(σ−µµ )[λ1( UUX )
θ+(1−λ1)
] γ−θθ
(EXUX
)γ−1 ( keoEX
)µ−1+κ̃4(σ − µ)
[τ(λ2(kneE
)η+ (1− λ2)
)ση ( E
EX
)σ+ (1− τ)
(λ3(keoEX
)µ+ (1− λ3)
)σµ
] γ−σσ
×λ3[λ3( keoEX )
µ+(1−λ3)]
σ−2µµ[
λ1( UUX )θ+(1−λ1)
] γ−θθ
(keoEX
)µ−1 (EXUX
)γ−1(27)
To obtain the procyclicality of the wage premium, we need σ > γ and σ < µ.
We have shown that if the complementarity between new capital equipment and non-
experienced educated workers is higher than complementarity with any other type of labor
force; and the complementarity between new capital equipment with educated workers with
no experience is higher than the complementarity when considering any other type of skills,
the model proposed generates the procyclical return to education and the countercyclical
return to experience we observe in the data.
39
new_tables_structure.pdftable1tables2-7Table8app.tables1-4tables_educ_cell_tables.pdfbyeduc