iii. wage determination and labour market...

Fortin/Lemieux – Econ 561 Lecture 3B

III. Wage Determination and Labour Market Discrimination

7. Investment in Education

1. Motivation and Basic Trends

2. Human Capital Theory a. Theoretical foundations of the Mincer model b. Extensions to the Mincer Model

3. Challenges to the Mincer Wage Equation

4. Challenges to human capital theory as an explanation for education decisions


7.1 Motivation and Basic Trends • Labour supply deals with quantities of labour supplied, but there is also a quality aspect to labour,

which can be enhanced by investment in education o Literacy and numeracy are the first goals of education

• Investment in education have long been seen to be key to economic well-being and growth

o Education policies are important instruments towards economic growth o Basic education is indeed subsidized in most industrialized countries o Moreover, schooling is compulsory until 15-16 years of age in many countries o More recently, policies have also focused on the early childhood learning

• More highly educated workers are paid more than less educated workers

o Why? o Why isn’t everyone getting a Ph.D.?

• The benefits of education are seen not only in term of a more productive workforce, but also in term

a healthier population, a more political engaged one, and a happier one! o There are social as well as private returns to education


• In summary, we can motivate our study of human capital policy with three major worries of our

times: economic growth, inequality and poverty.

• Human capital is not a magic bullet that will solve all social problems, but is a promising avenue for substantial social improvement.

A FEW MOTIVATING FACTS

• There is a lot of empirical evidence that schooling, literacy/numeracy skills, and many other indicators of human capital correlate positively with income and other outcomes

• Positive correlation holds at the individual, group, or country level

• Large increase in educational achievement through most of the 20th century, but much less changes (especially in the United States) in more recent decades, despite increasing returns to education.


AUSAUTBEL

BGR

BRA

CAN

CHE

CHL CZE

DEUDNK

ESP

FINGBR

GRC

HKG

HUN

IRLISL

ISR ITA

JPN

KOR

LIE

LUX

LVA

MEX

NLD

NOR

NZL

POL

PRT

ROM RUS

SWE

THA

USA

AUSAUT BEL

BRA

CAN

CHE

CZE

DEUDNK

ESP

FINGBR

GRCHKG

HUN

IRLISLITA

JPN

KOR

LIELUX

LVA

MAC

MEX

NLDNOR

NZL

POL

PRT

RUS

SVK

SWE

THATUN

TURURY

USA AUSAUTBEL

BGRBRA

CANCHE

CHL

COL

CZE

DEUDNK

ESP

EST

FINGBR

GRCHKG

HRV HUN

IRLISL

ISRITA

JOR

JPN

KOR

LIE

LTU

LUX

LVA

MAC

MEX

NLD

NOR

NZL

POL

PRT

QAT

ROMRUS

SVK

SVN

SWE

THATUN

TURURY

USA AUSAUT BEL

BGRBRA

CAN

CHE

CHL

CHNCOL

CRI

CZE

DEUDNK

ESP

EST

FINGBR

GRC HKG

HRV HUN

IRLISL

ISRITA

JOR

JPN

KOR

LIE

LTU

LUX

LVA

MAC

MEX

NLD

NOR

NZL

PAN

POL

PRT

QAT

ROM RUS

SGP

SVK

SVN

SWE

THA

TTO

TUN

TURURY

USAAUS

AUTBEL

BGR

BRA

CAN

CHE

CHL

COLCRI

CZE

DEUDNK

ESP

EST

FINGBR

GRC

HKG

HRVHUN

IRLISLITA

JOR

JPN

KOR

LTU

LUX

LVA

MAC

MEX

NLD

NOR

NZL

POL

PRT

ROM

RUS

SGP

SVKSVN

SWE

THATUN

TURURY

USA

78

910

1112

log

GP

D p

er c

apita

300 400 500 600Male Mathematics Mean

2000 2003 20062009 2012 Fitted values

GDP and PISA Male Math Scores: All Countries

AUSAUT BEL

BGR

BRA

CAN

CHE

CHL CZE

DEUDNK

ESP

FINGBR

GRC

HKG

HUN

IRLISL

ISR ITA

JPN

KOR

LIE

LUX

LVA

MEX

NLD

NOR

NZL

POL

PRT

ROM RUS

SWE

THA

USA

AUSAUT BEL

BRA

CAN

CHE

CZE

DEUDNK

ESP

FINGBR

GRCHKG

HUN

IRL ISLITA

JPN

KOR

LIELUX

LVA

MAC

MEX

NLDNOR

NZL

POL

PRT

RUS

SVK

SWE

THATUN

TURURY

USA AUSAUT BEL

BGRBRA

CANCHE

CHL

COL

CZE

DEUDNK

ESP

EST

FINGBR

GRCHKG

HRV HUN

IRLISL

ISRITA

JOR

JPN

KOR

LIE

LTU

LUX

LVA

MAC

MEX

NLD

NOR

NZL

POL

PRT

QAT

ROMRUS

SVK

SVN

SWE

THATUN

TURURY

USA AUSAUT BEL

BGRBRA

CAN

CHE

CHL

CHNCOL

CRI

CZE

DEUDNK

ESP

EST

FINGBR

GRC HKG

HRV HUN

IRLISL

ISRITA

JOR

JPN

KOR

LIE

LTU

LUX

LVA

MAC

MEX

NLD

NOR

NZL

PAN

POL

PRT

QAT

ROM RUS

SGP

SVK

SVN

SWE

THA

TTO

TUN

TURURY

USAAUS

AUTBEL

BGR

BRA

CAN

CHE

CHL

COLCRI

CZE

DEUDNK

ESP

EST

FINGBR

GRC

HKG

HRVHUN

IRLISLITA

JOR

JPN

KOR

LTU

LUX

LVA

MAC

MEX

NLD

NOR

NZL

POL

PRT

ROM

RUS

SGP

SVKSVN

SWE

THATUN

TURURY

USA

78

910

1112

log

GP

D p

er c

apita

300 400 500 600Female Mathematics Mean

2000 2003 20062009 2012 Fitted values

GDP and PISA Female Math Scores: All Countries

Source: Algan and Fortin (2015)

10 - ACEMOGLU & ANGRIST

Figure 1 LOG OUTPUT PER WORKER AND YEARS OF SCHOOLING ACROSS COUNTRIES

r FR USA D

ITA FRA

.6#,Wl

1

a) SPA ALPW?LR UKG G1I S

DEN NZE C- SGP HK•E JP 0 VEN SYAiE VEN

A c

OR

URS

P O R

U . .•

r

ALGIR BRA U

CUG

HUN COLM

HU clU U

PAN CZE EGY E~ SRL

- PMN

t.j

PHL -2

P IDN GUY ROM

C)

PA•

IND

ZBW LES CHN SNGIB A ZAM

o MOZRWYOG 0) MAN

)GfGBI M)BRGR jLW

-4

0 2 4 6 8 10 years of schooling

The line shows the fitted OLS relationship. The slope coefficient is 0.29, and the standard error is 0.02.

A simple calculation suggests that for, education to raise income as

steeply as suggested by Figure 1, there must be large human-capital exter- nalities. To see this, note that the private return to schooling, i.e., the increase in individual earnings resulting from an additional year of school-

ing, is about 6-10% (e.g., Card, 1999). If the social return to schooling, i.e., the increase in total earnings resulting from a one-year increase in average schooling, is of roughly the same magnitude, then differences in school-

ing can explain little of the cross-country variation in income. More

specifically, the difference in average schooling between the top and bottom deciles of the world education distribution in 1985 is less than 8 years. With social returns to schooling around 10%, we would expect the

top-decile countries to produce about twice as much per worker as the bottom-decile countries. In fact, the output-per-worker gap is approximately 15. Put differently, a causal interpretation of Figure 1 requires

and Lindahl (1999) for a detailed analysis of the cross-country relationship between education and income.

F

S

Fortin/Lemieux

Source: Fr

– Econ 561

enette, 201

14

Lecturee 3B

! "1!

!"#$%&'?'

@,'58&'A0B'@++0%5$,"5"&-'*,.'/&*%-'01'203+4&5&.'678004",#'

9&10%&'*,.':15&%'20,."5"0,",#'0,';,703&'

'

! ! ! !'

<05&-='$%&!'()*+&!,-.+/0&'!123"2!4&(536+0!D65M&5'!756)!8%&!#9:131;;;!<&-&5(+!=6.,(+!=/5>&4'?!@65-!,-!8%&!A-,8&0!=8(8&'!,-!

#9:;! 65! +(8&5B! N/8.6)&'! (5&! OP65M3Q(+/&! R&'.5,*865'S! 756)! 8%&! N../*(8,6-(+! T&8D65M! FNUTV$G! R(8(@('&B! ! W.%,&>&)&-8?!

,-0&*&-0&-.&?! (-0! 5&+(8,6-'%,*! '.65&'! (5&! 0&',C-&0! 86! )&('/5&?! 6-! (! '&>&-! *6,-8! '.(+&?! 8%&! 0&C5&&! 86! D%,.%! (! D65M&5E'!

6../*(8,6-! *56>,0&'! 6**658/-,8,&'! 86! 7&&+! (! '&-'&! 67! (! (..6)*+,'%)&-8?! 86! 7&&+! ,-0&*&-0&-8?! (-0! 86! '6.,(+,X&! D,8%! 68%&5'!

5&'*&.8,>&+4B!K&765&!.6-0,8,6-,-C!765!,-.6)&?!8%&!6/8.6)&!>(5,(@+&?!,'!5&C5&''&0!6-!(C&?!4&(5?!C&-0&5?!5(.&?!'8(8&!67!@,58%?!(-0!

4&(5!67!@,58%! 7,L&0!&77&.8'?!('!D&++!('! 7(),+4!.6)*6',8,6-!(8!(C&!#I?!)68%&5!(-0! 7(8%&5E'!&0/.(8,6-?!)68%&5E'!D65M,-C!'8(8/'?!

7(),+4E'! 5&+(8,>&! ,-.6)&!(8! (C&!#I?! (-0! 8%&! '.%66+,-C! (88(,-)&-8! .(8&C65,&'! F+&''! 8%(-!%,C%! '.%66+! F;3##G?! %,C%! '.%66+! F#1G?!

'6)&!.6++&C&!@/8!-6!@(.%&+65E'!0&C5&&!F#H3#2G?!(-0!(8!+&('8!(!@(.%&+65E'!0&C5&&!F#IJG?!D,8%!8%6'&!5&*658,-C!%,C%!'.%66+!('!8%&,5!

%,C%&'8! +&>&+! 67! '.%66+,-C! 6),88&0B! ! $%&! .6&77,.,&-8'! (5&! *5&'&-8&0! 5&+(8,>&! 86! 8%&! 6>&5(++! %,C%! '.%66+! C5(0/(8&!)&(-B! ! $%&!

5&'/+8'!(78&5!.6-0,8,6-,-C!765!,-.6)&!,-.+/0&!7,L&0!&77&.8'!765!'&+735&*658&0!,-.6)&!.(8&C65,&'!,-!&(.%!0(8('&8!4&(5>!!

3.5

3.5

3.54

4

44.5

4.5

4.55

5

50-11

0-11

0-1112

12

1213-15

13-15

13-1516+

16+

16+O*NET achievement score

O*NET achievement score

O*NET achievement scoreBefore conditioning on income

Before conditioning on income

Before conditioning on incomeAfter conditioning on income

After conditioning on income

After conditioning on income 4

4

44.2

4.2

4.24.4

4.4

4.44.6

4.6

4.64.8

4.8

4.85

5

50-11

0-11

0-1112

12

1213-15

13-15

13-1516+

16+

16+O*NET independence score

O*NET independence score

O*NET independence scoreBefore conidtioning on income

Before conidtioning on income

Before conidtioning on incomeAfter conditioning on income


After conditioning on income 4.2

4.2

4.24.4

4.4

4.44.6

4.6

4.64.8

4.8

4.85

5

50-11

0-11

0-1112

12

1213-15

13-15

13-1516+

16+

16+O*NET relationships score

O*NET relationships score

O*NET relationships scoreBefore conditioning on income

Before conditioning on income

Before conditioning on incomeAfter conditioning on income



Source: Oreopoulos and Salvanes (2009)

Figure XIII. Educational Attainment Decompositions, Males and Females 1900-1980 Birth Cohorts

0%5%

10%15%20%25%30%35%40%45%50%55%60%65%70%75%80%85%90%

1900

1905 1910

1915

1920

1925

1930

1935

1940

1945

1950

1955

1960

1965

1970

1975

1980

Year of Birth

Per

cent

age

Graduate HSAttend CollegeAttend Given HSGraduate CollegeGraduate Given Attend

Notes: 3-year moving averages based on CPS October, Census, CPS March and NCES data. HS graduates are those who obtained a regular public or private HS diploma (excluding GEDs) from the NCES. "Graduate HS" is the fraction of 8th grade enrollments for a given cohort who report a regular HS diploma. "Attend Given HS" is the fraction of recent HS graduates who report being enrolled the fall of the year following graduation. "Attend College" is college enrollments of recent HS graduates as a fraction of 18 year old cohort size. College graduates are those who report a BA or higher by age 25. "Graduate Given Attend" is those who obtained a four year degree as a fraction of the college enrollment total for that cohort. Two-year degrees are not included. "Graduate College" is the number of college graduates as a fraction of the 18 year old cohort size. Population estimates are from the Census P-20 reports. HS diplomas issued by sex are estimated from CPS October data after 1982.

Source: Heckman and Lafontaine (2007)

Figure XIV. Educational Attainment Decompositions, Males 1900-1980 Birth Cohorts

0%5%

10%15%20%25%30%35%40%45%50%55%60%65%70%75%80%85%90%

1900

1905 1910

1915

1920

1925

1930

1935

1940

1945

1950

1955

1960

1965

1970

1975

1980

Year of Birth

Per

cent

age




Figure XV. Educational Attainment Decompositions, Females 1900-1980 Birth Cohorts

0%5%

10%15%20%25%30%35%40%45%50%55%60%65%70%75%80%85%90%

1900

1905 1910

1915

1920

1925

1930

1935

1940

1945

1950

1955

1960

1965

1970

1975

1980

Year of Birth

Per

cent

age




VOL. 91 NO. 2 YOUTHS AND RISKY BEHAVIOR 99

0.9

1 9 - 0.8

.2 ~ ~ ~ ~ ~ - .0.7

0.6 -S

01.5 X X _ -0.5E

co ~~~~~~~~~~~~0.4~ I1 0.

0.2~ n1.1 - Male/Female College Ratio (left axis) 0.2

---- Male Veteran Rate (right axis) 0.1

0.9 - 0.0 1914 1918 1922 1926 1930 1934 1938 1942 1946 1950 1954 1958

Year of Birth

FIGURE 1. MALE-FEMALE COLLEGE GRADUATION RATE

AND MALE VETERAN RATE, BY BIRTH COHORT

that the relative schooling choices of men and women would follow a smooth inter-cohort trend in the absence of gender-specific factors such as the draft. We also assume that draft avoidance was proportional to the risk of induction faced by a cohort. Under these assumptions, draft-avoidance effects can be measured by regressing the relative education outcomes of men and women in the same cohort on a measure of the risk of induction faced by men in the cohort and an inter-cohort trend function.

As an illustration of the potential insights that can be gleaned from a comparison of male and female relative education outcomes across cohorts, Figure 1 plots the relative graduation rate of white men versus white women for cohorts born between 1914 and 1960, along with the fraction of men in each cohort who served in the military. (The data in this figure are drawn from the 5-percent public use samples of the 1980 and 1990 Censuses). The relative graduation rates show an interesting pattern of deviations between men and women, with "spikes" in the relative graduation rates of men born in the early 1920's and early 1930's, and a smaller but noticeable upward deflection for men born in the late 1940's. All three of these departures were associated with a rise in male veteran rates. The 1920-1925 cohort includes men who had a high rate of service in World War II but were young enough to easily return to school under the GI Bill program (John Bound and Sarah Turner, 1999). The 1930-1935 cohort includes men who were likely to be drafted during the Korean War and were eligible for GI Bill benefits (Marcus Stanley, 1999). Finally,

TABLE 1-ESTIMATED MODELS FOR MALE-FEMALE RELATIVE SCHOOLING

Fraction Fraction Fraction Independent enrolled completed completed variable (A) at ages college some or cohort (B) 20-21 degree college

A. Regression:

Linear trend -2.73 -1.70 -0.89 (X 100) (0.23) (0.08) (0.05)

Induction 1.16 0.46 0.43 risk (0.20) (0.09) (0.05)

R 2 0.96 0.98 0.98

B. Excess Male Enrollment Rate/Graduation Rate Due to Draft Avoidance (Percent):

1941 cohort 2.91 1.00 1.80

1947 cohort 6.47 2.22 4.01

1951 cohort 1.45 0.50 0.90

Notes: Standard errors are in parentheses. The model in column 1 is fit to data for cohorts born during 1939-1959; other models are fit to data for cohorts born during 1935- 1959. Induction risk is the number of inductions during ages 19-22, divided by number of men in the cohort. The risk is 0.082 for the 1941 cohort, 0.178 for the 1947 cohort, and 0.041 for the 1951 cohort. See the text for more details.

the 1944-1950 cohort includes men who were at high risk of service in the Vietnam War and were potentially affected by draft-avoidance behavior (as well as the availability of GI Bill benefits after service). Apart from these three groups, the male-female relative college graduation rate follows a smooth (hump-shaped) inter-cohort trend.

To estimate the effect of draft-avoidance behavior, we assume that draft avoidance was proportional to the risk of induction perceived by men in a cohort. Given our focus on the incentive to stay in college, we use the average number of inductions when the cohort was between 19 and 22 years of age, divided by an estimate of the size of the cohort, as our measure of induction risk. The estimated risk of induction is declining for cohorts born between 1935 and 1942 (from 8 percent to 4 percent), rises quickly to a peak of 11 percent for the 1946 cohort, and then drops steadily to 0 for men born after 1953.

Table 1 A presents a series of regression models which relate the log of the ratio of male to female education outcomes for a cohort to a linear inter-cohort trend and our index of

Source: Card and Lemieux (2001)

Nicole

Text Box

Korean war

Nicole

Callout

Vietnam war

Nicole

Callout

WWII

A. 26-30 Year Old Men

-2.6

-2.1

-1.6

-1.1

-0.6

-0.1

1959 1970 1975 1980 1985 1990 1995

Log

Rel

ativ

e S

uppl

y In

dex

United StatesCanada

United Kingdom

B. 46-50 Year Old Men

-2.6

-2.1

-1.6

-1.1

-0.6

-0.1

1959 1970 1975 1980 1985 1990 1995

Log

Rel

ativ

e S

uppl

y In

dex

United States

United Kingdom

Canada

Figure 3: Age-Group Specific Relative Supplies of College-Educated LaborSource: Card and Lemieux (2001)


7.2. Human Capital Theory • Of the five principal circumstances for which Adam Smith argued there should be compensating wage differentials,

the second one was “ the easiness and cheapness, or the difficulty and expence of learning a business. …

But it is when he compared an educated human to an expensive machine that laid out the basics of human capital theory

“When any expensive machine is erected, the extraordinary work to be performed by it before it is workn out, it must be expected, will replace the capital laid out upon it, with at least the ordinary profit. A man educated at the expence of much labour and time to any of those employments which require extraordinary dexterity and skill, may be compared to one of those expensive machines. ... It must do this too in a reasonable time, regard being had to the very uncertain duration of human life, in the same manner as to the more certain duration of the machine.

The difference between the wages of skilled labour and those of common labour, is funded on this principles.”

• With human capital theory, investments in human resources are considered similarly to other types of investments. Costs are incurred in expectation of future benefits and the investment in the last unit of human capital is made only if the benefits are expected to exceed the costs.


On the supply side: additional schooling entails opportunity costs in the form of foregone earnings plus direct expenses such as tuition. To induce a worker to undertake additional schooling, he must be compensated by sufficiently higher lifetime earnings.

On the demand side: to command higher earnings, more schooled workers must be sufficiently more productive than their less schooled fellow workers. In the perfect competition framework, the wage guides the allocation of workers across firms as to achieve an efficient allocation of resources by equating the wage to the value of the marginal product.

In the long run equilibrium, the relationship between lifetime earnings and the schooling must be such that i) the supply and demand for workers of each schooling level are equated and ii) no worker wishes to alter their schooling level.

• The study of the effects of investment in schooling and on-the-job training on the level, pattern and interpersonal

distribution of life-cycle earnings have been pioneered by Becker (1964, 1975) and Mincer (1958, 1962, 1979).

• This lead to one of the more successful empirical equation called the “Mincer earnings function” εβββ ++++= 2

210)ln( EErSw , where w is earnings (or the hourly wage when available), S is years of schooling, E is years of labor market experience, and ε is an error term.


• Years of schooling are a fairly direct measure of education human capital, while years of labour market experience

is viewed as a proxy for on-the-job training.

• Mincer uses the transformation Experience equals Age minus Schooling minus 6, 6−−= SAE , to capture the interaction between schooling and experience, but assumes multiplicative separability.

• The Mincer equation thus captures the important empirical regularities:

1) increase in earnings with schooling, 2) concavity of log earnings in experience, 3) log earnings experiences profiles are parallel across different education groups (ratio of earnings for persons

with education levels differing by a fixed number of years is roughly constant across schooling levels) no interaction between schooling and experience

4) log earnings-age profiles diverge across different education groups.

• Almost all empirical studies find that schooling has a positive and significant effect on earnings ( 0>r ) that earnings are a concave function of labour market experience ( 01 >β and 02 <β ).

• A simple regression model with a linear schooling term and a low-order polynomial in potential experience explains 20-35% of the variation in observed earnings data, with predictable and precisely-estimated coefficients in almost all applications.

Figure 1

Source: Jacob Mincer, Schooling, Experience, and Earnings

Source: Lemieux (2001)

1104 Journal of Economic Literature, Vol. XXXIX (December 2001)

A. United States B. Sweden 10.5 5.5 -

10 - ~ ~ ~ ~ 05-

,9.5 - 4.5 0

9 0 4 0

8 .5 - _, _ , _ , _ , _ ,_ ,_ ,_ ,_ ,_ , , 3 .5 - , _._ ._ ._ ._ ._ ._ ._ ._ ._ _ 9 10 11 12 13 14 15 16 17 18 19 9 10 11 12 13 14 15 16 17 18 19

Years of Schooling Years of Schooling

C. West Germany D. East Germany 9 8-

8.5 7.5 0

7 - 7 6-l 7.5 - ~~~~~~~~~~~6 .5-

7, , ,, ,6 A _ _ _ _ _ _ _ _ _

9 10 11 12 13 14 15 16 17 18 19 9 10 11 12 13 14 15 16 17 18 19 Years of Schooling Years of Schooling

Figure 1. Unrestricted Schooling-Log Wage Relationship and Mincer Earnings Specification

each year of schooling, controlling for experience and gender, as well as the OLS estimate of the Mincerian return. It is apparent that the semi-log specification provides a good description of the data even in countries with dramatically different economic and educational systems.6

Much research has addressed the question of how to interpret the edu-

cation slope in equation (1). Does it reflect unobserved ability and other characteristics that are correlated with education, or the true reward that the labor market places on education? Is education rewarded because it is a sig- nal of ability (Michael Spence 1973), or because it increases productive capa- bilities (Becker 1964)? Is the social return to education higher or lower than the coefficient on education in the Min- cerian wage equation? Would all individuals reap the same proportionate increase in their earnings from attending school an extra year, or does the return to education vary systematically with individual characteristics? Definitive an- swers to these questions are not available, although the weight of the evidence clearly suggests that education is not merely a proxy for unobserved ability. For example, Griliches (1977)

6 Evaluating micro data for states over time in the United States, Card and Krueger (1992) find that the earnings-schooling relationship is flat until the education level reached by the 2nd per- centile of the education distribution, and then becomes log-linear. There is also some evidence of sheep-skin effects around college and high school completion (e.g., Jin Huem Park 1994). Although statistical tests often reject the log-linear relationship for a large sample, the figures clearly show that the log-linear refationship provides a good ap- proximation to the functional form. It should also Ee noted that Kevin Murphy and Finis Welch (1990) find that a quartic in experience provides a better fit to the data than a quadratic.

Source: Krueger and Lindahl


b. Theoretical foundations of the Mincer model

• The theoretical foundations of Mincer specification can arise from two theoretical framework, i. compensating wage differentials (Mincer, 1958)

ii. accounting identity framework (Mincer, 1974).

• Let )(sw represent the annual earnings of a representative individual with s years of education, and let r be the discount rate, both assumed to be constant over his lifetime, the present value of the stream of earnings (for infinitely lived individuals) is ( ) = 0 + +⋯ ( ) + ( )( ) + ( )( ) + ⋯+ ( )( )

( ) = ( ) [ ( ) + ( ) + ( )( ) + ( )( ) + ⋯+ ( )( ) ]

= ( ) ( )∑ ( ) = ( ) ∙ ( )

• Assuming no direct costs of schooling and comparing an individual with 1 year of schooling, who has to wait that one year before earning a salary, to one with 0 year of schooling, and equalizing present values, we get (1) 1 + ∙ 11 + = (1) = (0) 1 +


(1)(0) = 1 +

• Taking the logarithm on both side, ln (1) − ln( (0)) = ln(1 + ) ≈ for < 0.2 the wage increment for one more year of schooling must approximately be equal to the rate .

• Or in continuous time, let )(sw represent the annual earnings of a representative individual with s years of education,

( )rTrs

T

s

rt

eersw

dtswesV

−−

−

−=

= ∫)(

)()(

where T is the length of working time.

• Letting the compensating differentials take their time to equalize lifetime earnings: ))0(ln())(ln()0()( VsVVsV =⇔=


• The earnings difference between an individual with s years of schooling and 0 years will be given by

( ),)0(ln

)1/()1(ln)0(ln)(ln )(

rSw

eerSwsw sTrrt

+=

−−++= −−−

since the last term goes to zero as T gets large.

• This yields the nice interpretation of r as an internal rate of return to schooling, that is, the discount rate equates the lifetime earnings streams from different educational choices.

• But this framework ignores the direct costs of schooling, assumes that earnings do not vary over the life-cycle,

and that the age of retirement does not depend on years of schooling.

• An alternative framework used by Mincer (1974) builds on an accounting identity initially studied by Becker (1974) and Becker-Chiswick (1966).

• It allows earnings to vary over the life-cycle


• Let tE be potential earnings at time t . Investments in training are expressed as a fraction of potential earnings invested, ttt EkC = , where tk is the fraction invested at time t and let tρ be the return to training investments made at time . Then

( )ttttttt kECEE ρρ +=+=+ 11 .

• Repeated substitution yields ( ) 010 1 EkE jj

tjt ρ+= Π −= .

• Formal schooling is defined as years spent in full-time investments ( 1=tk ). Assume that the rate of return on

formal schooling is constant ( st ρρ = ) and that formal schooling takes place at the beginning of life.

• Also assume that the rate of return to post-school investment, tρ , is constant over time and equals 0ρ . Then, we can write (in logs)

,)1ln()1ln(lnln1

00 ∑−

=

++++=t

sjjst kSEE ρρ

• Which yields the approximate relationship (for small sρ and 0ρ ) using the fact that ρρ ≈+ )1ln( , if 2.0<ρ ,

t


.lnln1

000 ∑−

=

++≈t

sjjt kSEE ρρ

• To establish a relationship between potential earnings and years of labour market experience, Mincer (1974) further assumes a linearly declining rate of post-school investment:

)1(Txk xs −=+ κ ,

where 0≥−= stx is the amount of work experience as of age t and T is the length of working life.

• Under these assumptions, the relationship between potential earnings, schooling and experience is given by:

[ ] .22

lnln 200000 x

Tx

TSEE ssx

κρκρκρρκρ −⎟⎠⎞

⎜⎝⎛ +++−≈+

• Observed earnings equal potential earnings less investment costs, producing the following relationship for observed earnings:

⎟⎠⎞

⎜⎝⎛ −−≈ + T

xExsw sx 1ln),(ln κ

[ ] 200000 22

ln),(ln xT

xTT

sExsw sκρκκρκρρκκρ −⎟

⎠⎞

⎜⎝⎛ ++++−−≈


.2100 xxss ββρα +++≈

• When sE andκ are uncorrelated, the variance of log earnings will be minimized at .0/1 ρ . (See Heckman et al.)

b. Extensions to the basic Mincer model

• In most applications of the Mincer model, it is assumed that the intercept and slope coefficient are the same across persons and do not depend on the schooling level, however allowing κ and sρ to differ across persons, as in Mincer (1974) produces a random coefficient model

[ ],)()()()(),(ln 2221100

2210 xxsxxsxsw iissiisi ββββρρββββρβ −+−+−+−++++=

where the terms in bracket are part to the error term, expressed in deviations from the mean coefficients.

• It turns out that this is a crucial extension to the basic model. Since if individuals were identical, why should they choose different amounts of education?

• However, although heterogeneity explains why we observe different individuals having different schooling, an

additional set of problems is introduced: Schooling is endogeneous and, if this is not allowed for, estimated return may be biased. Returns as well as schooling may vary across individuals.


• This will have important implications (see, Card (1999)) for the econometric estimation of the Mincer equation,

which we will talk about in the estimation section.

Other extensions to the Mincer model:

• Heckman, Lochner and Todd (2003, 2008) examine the importance of relaxing functional form assumptions in estimating internal rates of return to schooling and in accounting for taxes(τ ), tuition ( v ), independence of the length of work-life from schooling ( 1)( =′ sT ), nonlinearity in schooling, and non-separability between schooling and work experience( )()(),( xsuxsw ϕ≠ ):

,)1)(,()()(

0 0

)1()()1(∫ ∫−

−−+−− −−=ssT s

rzsxr dzvedxxswesV ττ τ

• Inferences about trends in rates of return to high school and college obtained from the more general model differ substantially from inferences drawn from estimates based on the basic Mincer earnings regression.

• Heckman et al. also examine the implications of accounting for uncertainty and agent expectation formation. Even when the static framework of Mincer is maintained, accounting for uncertainty substantially affects rate of return estimates.


• They propose a substantial break from Mincer's approach by allowing for the sequential resolution of uncertainty. That is, with each additional year of schooling, information about the value of different schooling choices and opportunities becomes available generating an option value of schooling.

The return in excess of the direct return (the lifetime income received at a given schooling level) is the option value. For example, finishing high school provides access to college, and attending college is a necessary first step for obtaining a college degree.

Each additional year of schooling reveals new information about uncertain college costs or future earnings. Individuals with s years of schooling must decide whether to quit school and receive lifetime earnings of

)(sY , or continue on in school for an additional year and receive an expected lifetime earnings of ))1(( +sVE s .

• When the schooling decision is made at the beginning of life and age-earnings streams across schooling levels are

known and cross only once, then the internal rate of return (IRR) can be compared with the interest rate as a valid rule for making education decisions (Hirschleifer, 1970).

• But the sequential resolution of uncertainty and non-linearity in returns to schooling contribute to sizeable option values. Accounting for option values challenges the validity the internal rate of return.

• They conclude that in the presence of sequential resolution of uncertainty and option values, the internal rate of return-a cornerstone of classical human capital theory-is not a useful guide to policy analysis.


3. Challenges to the Mincer Wage Equation

• The basic Mincer human capital earnings function does not appear to fit the data nearly as well in the 1980s and 1990s as it did in the 1960s and 1970s:

1) Log wages are an increasingly convex function of years of schooling: that is, the log-linearity in schooling no longer seems to hold.

2) The quadratic in experience tends to understate earnings growth at the beginning of the lifecycle and overstate the decline in earnings towards to the end of the lifecyle. Murphy and Welch (1990) find that a quartic fits the data better.

F

S

Fortin/Lemieux

Source: Lemi

– Econ 561

eux (2001)

Lecturee 3B

F

Fortin/Lemieux – Econ 561

Lecturee 3B

Source: Lem

mieux (2001)


3) Experience-wage profiles are no longer parallel for different education groups: that is, the multiplicative separability between experience and schooling no longer holds. For example, the college-high school wage gap is now much larger for less experienced than for more experienced workers.

• The Mincer model is for individuals tracked over school and working ages, and should be estimated by tracking

single cohorts through education and the labour market.

• But it is usually estimated from a cross-section of individuals, which is actually only legitimate if the economy is in a long run steady state.

• Rapid technical change or secular improvement in the quality of education would violate this assumption and

invalidate the use of cross-sectional data for estimation: cross-section and cohort based estimates of returns to education would not be expected to be the same.

• Mincer was well aware that this may be a problem in his early work (Mincer, 1958), and conjectured that the cross-

sectional age-earnings profiles were probably understating life-cycle earnings growth since, in those days, there was substantial secular growth in average earnings.


• Now many researchers believe that the cross-sectional age-earnings profile overstates life-cycle earnings growth. Beaudry and Green (2000) conclude that the slope of the cross-sectional age-earnings profile has increased because new cohorts of men are entering the labour market at increasingly low earnings level in Canada.

• This highlights the well-known problem of distinguishing between age, cohort, and year effects in earnings growth.

• Card and Lemieux (2001) investigate this issue in detail and conclude that the dramatic increase in the college-high

school wage gap for post-1950 cohorts is primarily a consequence of an equally dramatic break in inter-cohort trends in educational achievement.

• Until 1950, each successive cohort of men was more educated than the preceding cohort. This secular trend

abruptly stopped with men born around 1950.

• After 1975, the entry of the post-1950 cohorts were no more educated, and in some case less educated, than previous cohorts.

o The relative supply could no longer offset the rising demand for skilled young workers and, as a result, the college-high school wage gap started expanding dramatically for young workers.

920 P. Beaudry and D. Green

800

750

070 - 650

a}~~~~~~~~~~~~~~~ - - - 197 -n

0

1982 Entry -- 19B6 Entry

350 S * 1990 Entry 990 Cross Section

300 ' ' ' ' . 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54

Age

FIGURE 3b Age-earnings profiles allowing differing slopes by cohort: males, university education

is no indication that recent cohorts, who are starting at lower entry-level wages, should be expected to experience higher wage growth as they age than that experienced by older workers, and thus, there is no basis for predicting that they will attain similar wage levels. Once again, the 1990 cross-sectional profile is substantially steeper than the age-earnings progression actually experienced by any cohort. Given the linear downward trend in cohort earnings experiences, however, there is no suggestion that the cross-sectional profiles are getting steeper with time. This, in fact, is what is found in the cross-sectional literature, which indicates that experience differentials increased substantially in the 1980s and early 1990s for less educated workers but showed at most mild increases for university educated workers.

In columns 3 and 4 in table 2 we explore the robustness of the pattern reported in column 1. Since the estimates on cohort and cohort squared in column 1 are not individually significantly different from zero (although they are jointly significant with a p-value less than 0.00 1), we drop the cohort squared term in column 3 in order to verify the robustness of the observation that cohorts are successively starting at lower wages and are not experiencing higher wage growth. This is, indeed, supported by the results in column 3.

Column 4 of table 2 corresponds to a specification in which the quadratic cohort term is replaced by a full set of cohort-dummies. As in table 1, here we report only two of the fifteen cohort dummies, since that is all that is needed to represent the overall pattern. Both the 1978 and 1992 cohort dummies reported in the table indicate once again the general pattern of deteriorating labour market performance for successive waves of labour market entrants. The point estimates indicate that the 1992 cohort earns approximately 20 percent less than their counterpart did twenty years earlier.


Take-Away on the Mincer Equation • In summary, recent evidence suggests that the basic Mincer human capital earnings model remains a parsimonious

and accurate model in a stable environment where educational achievement grows smoothly across cohorts, as it did in the time period originally studied by Mincer (1974).

• In a less stable environment, however, major shifts in the relative supply of different age-education groups can

induce important changes in the structure of wages that have to be taken into account when estimating a standard Mincer equation.

• Issue to be discussed later in the course: How to interpret the coefficient on schooling? o Ability bias: Upward “bias” o Heterogeneity in effects: ??? o Measurement Error: Downward “bias” o Signalling vs. Human Capital


Challenges to human capital theory as an explanation for education decisions

• One of the clearest implications of human theory is that education decisions should be highly responsive to pecuniary factors such as the rate of return to education, tuition fees, etc.

• However, this is generally not supported by the empirical evidence

• Rates of returns to college/postgraduate educations almost doubled in the United States since about 1980, but enrollments did not increase much, especially for men

• Most studies also suggest that enrollment elasticities with respect to tuition are modest

• Behavioral economics suggest other possible explanations


• For instance, Bettinger, Long, Oreopoulos, and Sanbonmatsu (QJE 2012) “The Role of Application Assistance and Information in College Decisions” use a field experiment (in collaboration with H&R Block) to look at the role of factors beyond the pecuniary ones suggested by HC theory.

• Sample of low income individuals going to H&R Block to get help filling their US tax forms • Some of them were randomized into a treatment where the H&R Block agent helps them fill the

paperwork for federal college loans (free application for federal student aid, or FAFSA) • Another treatment consisted in providing information on how much it would cost to go to college

(tuition) compared to the aid available. • Large effect found: combined treatment (FAFSA plus information) increased probability of

completing two years of college by 8 percentage points (from 28 to 36 percent). But information experiment alone has no impact

TABLE II

MEAN CHARACTERISTICS OF CONTROL GROUP AND DIFFERENCES BETWEEN TREATMENT AND CONTROL MEAN CHARACTERISTICS

Dependent participantsIndependent participants

with no prior college experienceIndependent participants

with prior college experience

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Controlmean

FAFSAtreatmentdifference

Infotreatmentdifference

Controlmean



Controlmean



Female 0.560 0.019 0.015 0.575 �0.003 �0.031 0.638 �0.002 �0.028(0.035) (0.061) (0.011) (0.020) (0.012) (0.023)

White 0.553 0.004 0.097 0.703 0.002 0.001 0.653 �0.014 0.005(0.035) (0.059)* (0.010) (0.018) (0.012) (0.023)

Black 0.379 0.013 �0.079 0.246 0 �0.012 0.285 0.016 0.003(0.035) (0.057) (0.009) (0.017) (0.012) (0.022)

Hispanic 0.023 �0.005 0.002 0.024 �0.001 0 0.023 0 �0.003(0.010) (0.019) (0.003) (0.006) (0.004) (0.007)

Age 17.713 0.029 0.050 25.911 �0.034 �0.191 26.207 0.147 �0.151(0.035) (0.051) (0.067) (0.124) (0.072)** (0.132)

Never in college 0.965 0.015 �0.002 0 0 0 1 0 0(0.012) (0.023)

Married 0.131 �0.002 �0.021 0.128 �0.005 �0.022(0.007) (0.013) (0.008) (0.015)

Single 0.800 0.006 0.019 0.803 0 0.031(0.009) (0.016) (0.010) (0.018)*

AS

SIS

TA

NC

EA

ND

INF

OR

MA

TIO

NIN

CO

LL

EG

ED

EC

ISIO

NS

1221

at The University of British Colombia Library on February 3, 2016 http://qje.oxfordjournals.org/ Downloaded from

Source: Bettinger, Long, Oreopoulos, & Sanbonmatsu (2012)

TABLE II

(CONTINUED)

Dependent participantsIndependent participants

with no prior college experienceIndependent participants

with prior college experience

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Controlmean



Controlmean



Controlmean



Divorced orseparated

0.069 �0.005 0.002 0.069 0.005 �0.009(0.005) (0.010) (0.007) (0.011)

Target degreewould bebachelor’s

0.412 �0.015 �0.025 0.272 0.004 0.028 0.425 �0.009 0.012(0.035) (0.060) (0.010) (0.018) (0.013) (0.024)

Target degreewould beassociate

0.354 �0.018 �0.017 0.488 �0.002 �0.033 0.476 0.009 0.009(0.034) (0.058) (0.011) (0.020) (0.013) (0.024)

Target degreeunsure

0.234 0.033 0.041 0.239 �0.001 0.004 0.099 0 �0.022(0.031) (0.054) (0.009) (0.017) (0.008) (0.013)*

Very interestedin college

0.746 �0.013 0.029 0.510 0.006 0.030 0.655 0.004 0.004(0.031) (0.052) (0.011) (0.020) (0.012) (0.023)

Adjusted grossincome

$23,211 381 �702 $16,404 �226 �587 $17,801 98 �343(816) (1402) (210) (379) (254) (457)

Observations 398 390 80 4117 4389 722 3044 3085 517

Notes. Dependent students are typically under the age of 24 and financially dependent on their parents. Most dependent participants in this sample are high school seniors.Independent participants are over the age of 24 or married, had a child, were a veteran, or were an orphan. ‘‘Prior college experience’’ is defined from surveying participants.

QU

AR

TE

RL

YJ

OU

RN

AL

OF

EC

ON

OM

ICS

1222

at The University of British Colombia Library on February 3, 2016 http://qje.oxfordjournals.org/ Downloaded from


(column [1]), which corresponds to a 40% increase (p-value< .01).The requirement that both parent and dependent sign theFAFSA explains why the filing rate was not even higher amongthe treated. The application first had to be first mailed to thedependent’s household to be signed and then sent to the DOE.It is likely those more interested in college actually followedthrough with the process.

The information-only treatment did not have a substantialeffect on aid application submission. Participants who receivedonly customized information about their likely grant and loaneligibility relative to college costs were no more likely to file a

TABLE III

SUMMARY OF THE RESULTS

Outcome during first year following experiment

(1) (2) (3)

Filed FAFSA(based on

DOE data)

Attendedcollege

(based onNSC and

OBR data)

Attendedcollege and

received PellGrant (basedon DOE data)

Dependent participants (N = 868)Control group mean 0.399 0.342 0.296FAFSA treatment effect 0.157 0.081 0.106

(0.035)*** (0.035)** (0.034)***Info treatment effect �0.012 �0.004 0.004

(0.060) (0.058) (0.056)Independent participants, no prior college (N = 9,228)

Control group mean 0.161 0.095 0.111FAFSA treatment effect 0.267 0.015 0.030

(0.009)*** (0.007)** (0.007)***Info treatment effect �0.019 0.003 �0.016

(0.014) (0.012) (0.012)Independent participants, some prior college (N = 6,646)

Control group mean 0.320 0.263 0.209FAFSA treatment effect 0.195 �0.003 0.017

(0.012)*** (0.011) (0.011)Info treatment effect 0.027 0.013 0.015

(0.023) (0.021) (0.020)

Notes. Treatment effects are mean differences between treatment and control groups (estimated usingordinary least squares). Robust standard errors in parentheses. DOE = Department of Education.NSC = National Student Clearinghouse. OBR = Ohio Board of Regents. Single, double, and triple asterisksindicate statistical significance at the 10%, 5%, and 1% levels, respectively. In the Online Appendix, wepresent the same models with the inclusion of baseline covariates, and the main estimates are robust toincluding these controls.

QUARTERLY JOURNAL OF ECONOMICS1226

at The U

niversity of British C

olombia L

ibrary on February 3, 2016http://qje.oxfordjournals.org/

Dow

nloaded from


iii. wage determination and labour market...

Documents