Transcript Where the Boys Aren’t: Recent Trends in U.S. College

Where the Boys Aren’t:
Recent Trends in U.S. College
Enrollment Patterns
Patricia M. Anderson
Department of Economics
Dartmouth College
And NBER
1
“Just the Facts, Ma’am”
 In 1972, males made up 56 percent
of overall college enrollments
 In 2004, males made up 43 percent
of overall college enrollments
 Similar trends are seen in full-time
enrollment and degree attainment
2
Official Statistics on Fraction of
College Students Who Are Male
0.58
0.56
0.54
Fraction
0.52
0.50
0.48
0.46
0.44
0.42
0.40
0.38
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Full-Time
Source: Digest of Educational Statistics, Table 185
Total
3
The Sample Data
 October supplement to the Current
Population Survey (CPS) collects
information on school enrollment
 Survey covers the civilian, noninstitutionalized population
 I look at those age 17 to 50 who have
not already graduated from college
4
“Just the Facts, Ma’am” Part 2
 While noisier, sample statistics reflect
the same basic trends:
 Fraction male 57 percent in 1972
 Fraction male 44 percent in 2004
 Decline in male enrollments not quite
as sharp when focus only on
“traditional students”
 Fraction male among full-time, 4-yearcollege students drops from 52 percent
to 46 percent over this time period
5
Sample Statistics on Fraction of
College Students Who Are Male
0.6
0.58
0.56
Fraction
0.54
0.52
0.5
0.48
0.46
0.44
0.42
0.4
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
All College Students
College Students Age 18-22
Source: Author Calculations from October CPS for given years
4-Yr College Students Age 18-22
6
Some Descriptive Analysis
 In the early years, the older students are
more likely to be male
 Likely lingering effects of the Vietnam War, as
veterans benefit from the GI Bill
 In the later years, the older students are
more likely to be female
 Likely changing social climate, as previous
investment decision no longer optimal
 Result is a “catch-up” in educational
attainment by earlier cohort females
7
Male/Female Ratio in Enrollment
and Attainment by Birth Cohort
1.6
1.5
Probability Ratio
1.4
1.3
1.2
1.1
1
0.9
0.8
1951
1953
1955
1957
1959
1961
1963
1965
1967
1969
1971
1973
1975
Approximate Birth Year
Enrollment in 4-yr college (age 18-22)
Have Degree by March 98-02
8
Source: Author calculations from October CPS, individual years; March CPS MORG files, pooled 1998-2002
Probability of Enrollment
 Sampled females increasingly more likely to be
enrolled than sampled males
 Note males more likely out of sample due to higher
incarceration and military rates
 Decline in male/female ratio is less steep for younger
individuals
 Recall patterns already seen for earlier cohorts
 Conditional on high school graduation, the probability
of enrollment was almost equal for younger males and
females during the 1990s (later than other groups)
 Note males are more likely to drop out of high school
9
Male/Female Ratio on Probability of
Being Enrolled in College
10
Source: Author calculations from October CPS for given years
A Simple Model of
Human Capital Investment
 Invest in human capital as long as marginal
cost is not greater than marginal benefit
 For annual earnings, Y; annual costs, C;
working life, T; and discount rate, r; attend
if:
Yt  Ct
Yt

0

t
t
t 1 1  r 
t 1 1  r 
T
c
T
nc
11
Implications of the Model –
Less Likely to Invest:
 The higher the costs of schooling
 Decline in male eligibility for the GI Bill
could decrease male enrollments
 Higher psychic costs (males tend to get
worse grades in high school) could
decrease male enrollments
 The higher the discount rate
 If increased future earnings are heavily
discounted, higher current earnings could
decrease male enrollments
12
Implications of the Model –
More Likely to Invest:
 The longer one expects to work
 Social changes that lead to women expecting
longer, less interrupted careers would imply
increased female college enrollments
 The bigger the gap between college and
high school earnings
 If wages for college-educated women have
increased faster than for men, or if wages for
high school-educated men have increased
faster than for women, then would expect
increased female college enrollments
13
Determinants of Enrollment
 Focus on 20-year olds, by cohort
 A cohort is a group of 5 birth years
 1953-1957, 1958-1962, etc.
 For each cohort, a separate linear
probability model is estimated
 In addition to basic demographics,
the explanatory variables are
motivated by the basic human capital
investment model
14
Explanatory Variables
 Demographics
 Marital status, race, veteran status, etc.
 State-level tuition, unemployment rate
 Economic returns
 25th percentile wages for college-educated
workers age 28-32 of your sex, race, state
 75th percentile wages of HS-educated workers
age 23-27 of your sex, race, state
15
Within Cohort Decompositions
 Male-female difference in the probability of
enrollment can be decomposed into 2
parts:
 Unexplained (i.e. the coefficient on male)
 Explained by differences in means:
Pmc  Pfc  ˆ c  ˆc X mc  X fc 
16
Decomposition of Enrollment of 20year-olds born from 1953 to 1957
Difference in Enrollment Rate
Unexplained
(i.e. male dummy)
Explained by
never married
male
0.266
means
female
0.228
male-female
difference
0.038
coefficient
(std error)
coefficient
*difference
-0.019
(0.030)
0.268
(0.012)
0.032
(0.016)
-0.100
(0.039)
-0.014
(0.038)
0.009
(0.019)
-0.145
(0.020)
-0.033
(0.028)
-0.029
(0.018)
0.058
(0.019)
-0.050
(0.046)
0.213
(0.012)
0.188
(0.309)
-0.019
1.000
0.000
1.000
0.804
0.593
0.211
spouse gone
0.013
0.047
-0.034
black
0.115
0.133
-0.017
white
0.868
0.858
0.010
south
0.313
0.337
-0.025
veteran
0.041
0.000
0.041
ln(state tuition)
7.536
7.525
0.010
ln(state UR)
1.852
1.848
0.005
ln(weekly earnings) {25% for college grad}
6.380
5.710
0.670
ln(weekly earnings) {75% for HS grad}
6.739
6.105
0.635
high school graduate
0.821
0.826
-0.004
intercept
1.000
1.000
0.000
Total Explained
Percentage Explained
0.056
-0.001
0.002
0.000
0.000
-0.006
0.000
0.000
0.039
-0.031
-0.001
0.000
0.056
149%
17
Decomposition of Enrollment of 20year-olds born from 1973 to 1977
Difference in Enrollment Rate
Unexplained
(i.e. male dummy)
Explained by
never married
male
0.327
means
female
0.344
male-female
difference
-0.017
coefficient
(std error)
coefficient
*difference
-0.023
(0.023)
0.268
(0.018)
0.055
(0.025)
-0.136
(0.030)
-0.021
(0.029)
-0.040
(0.018)
-0.141
(0.065)
-0.016
(0.035)
-0.112
(0.035)
0.002
(0.031)
-0.035
(0.047)
0.346
(0.017)
0.398
(0.481)
-0.023
1.000
0.000
1.000
0.928
0.819
0.109
spouse gone
0.012
0.030
-0.018
black
0.138
0.156
-0.018
white
0.808
0.795
0.014
south
0.367
0.349
0.018
veteran
0.012
0.003
0.009
ln(state tuition)
8.021
8.024
-0.003
ln(state UR)
1.730
1.728
0.001
ln(weekly earnings) {25% for college grad}
6.314
5.917
0.397
ln(weekly earnings) {75% for HS grad}
6.603
6.185
0.418
high school graduate
0.834
0.859
-0.025
intercept
1.000
1.000
0.000
Total Explained
Percentage Explained
0.029
-0.001
0.002
0.000
-0.001
-0.001
0.000
0.000
0.001
-0.014
-0.009
0.000
0.006
-36%
18
Within Cohort
Decomposition Highlights
 Within cohort, the male dummy is never
significantly different from zero
 Higher marriage rates for females are
important, especially for the early cohorts
 Higher earnings for male high school
graduates are important, especially for the
later cohorts when it is not offset by the
effect of higher earnings for male college
graduates
19
Across Cohort Decompositions
 The change in the male-female difference
in the probability of enrollment can also be
decomposed into the unexplained part and
the part explained by the change in the
differences in means:
ˆct X mc t  X fct   X mc  X fc 
20
Across-Cohort Decomposition
Change in Difference in Enrollment Rate
Explained by
never married
spouse gone
black
white
south
veteran
ln(state tuition)
ln(state UR)
ln(weekly earnings) {25% for college grad}
ln(weekly earnings) {75% for HS grad}
high school graduate
Total Explained
Percent Explained
male-female
difference
in means for
early cohort
0.038
male-female
difference
in means for
late cohort
-0.017
later - early
chort
change in
differences
0.055
coefficents
for
late cohort
coefficient
*change in
differences
0.211
-0.034
-0.017
0.010
-0.025
0.041
0.010
0.005
0.670
0.635
-0.004
0.109
-0.018
-0.018
0.014
0.018
0.009
-0.003
0.001
0.397
0.418
-0.025
0.101
-0.016
0.001
-0.004
-0.043
0.032
0.014
0.003
0.273
0.217
0.021
0.268
0.055
-0.136
-0.021
-0.040
-0.141
-0.016
-0.112
0.002
-0.035
0.346
0.027
-0.001
0.000
0.000
0.002
-0.005
0.000
0.000
0.001
-0.007
0.007
0.023
42%
21
Across-Cohort
Decomposition Highlights
 INCREASES IN RELATIVE FEMALE ENROLLMENT (AS
ACTUALLY OBSERVED) ARE IMPLIED BY
 Narrowing of the male-female marriage gap

Most important contributor to observed change
 Increase of the female advantage in HS graduation
 Contributes slightly (effect ¼ size of marriage gap)
 INCREASES IN RELATIVE MALE ENROLLMENT
(OPPOSITE OF OBSERVED) ARE IMPLIED BY
 Narrowing of male-female HS graduate wage gap
 Exactly offsets HS grad effect
 Decrease in male advantage in veteran status
 Very small effect
22
Summary
 The biggest drop in the male fraction of
college students is likely due to one-time
events  End of the Vietnam War
 End of draft deferments reduces overconsumption of college by males
 Fewer veterans using GI Bill tuition benefits
 Increased opportunities for women
 Enrollment by earlier cohorts at older ages
 Higher age at first marriage
 Lower HS drop-out rates
23