Henry Peng Thesis
Transcript of Henry Peng Thesis
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 1/20
Do Labor Market Conditions Affect College Graduation Rates?
United States as a Case
Henry Peng1 [email protected]
Senior Honors Thesis
Department of EconomicsWashington University in St. Louis
February 25, 2011
Abstract: This paper investigates the impact of labor market conditions on college graduation rates.Using panel data of 436 four-year public colleges across the fifty states spanning five years, I examine
the impact of both national and state unemployment on the difference between a school’s four year andsix year graduation rate by cohorts. I use different models to compare the impact of both the national
unemployment rate and the state unemployment rate on the graduation rates. A robust modelcontrolling for both time and school fixed effects result in a positive relationship between both the
national and state unemployment rate by bachelor’s degree and higher and the difference between a
school’s four year and six year graduation rate. Credible assessment of the impact of labor marketconditions on college graduation rates can provide important insight into what factors college studentsconsider when graduating.
I would like to thank my thesis advisor, Professor Juan Pantano, for his valuable time, patience, and guidance. I would
also like to thank Professor Sebastian Galiani, Professor Bruce Petersen, and Professor Dorothy Petersen for their support
and assistance throughout the honors thesis process. I thank Andrew Mary from the Department of Education and Peter
Horner from the Bureau of Labor Statistics with their assistance with the data collection.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 2/20
2
I. Introduction
During times of high unemployment many people worry about their jobs, either holding on to their
current job or finding a new job. For the 1.5 million job seeking college students that graduate this year
into a job market with national unemployment hovering around double-digits, their employment
situation seems bleak. The high unemployment rate is not the only factor that is plaguing college
graduates since many have little or no job experience and are competing with experienced workers.
Research has shown during poor labor market conditions the job matching process deteriorates because
of the difficulty of finding a job that fits. (Oreopoulos, von Wachter, and Heisz 2006) This results in
individuals accepting jobs less preferred and lower salaries, only to leave when labor market
conditions improve (Kahn 2006). What is even more critical is that recent research shows that a
college graduate who enters the labor market during a bad economy negatively impacts that
individual’s long-run wages (Kahn 2006).
On the other hand, poor labor market conditions mean that the opportunity cost of attending school
decreases which is why from the fall of 2007 to the fall of 2008 there was a 6% surge in the freshmen
enrollment rate in post-secondary schools across the U.S, making it the largest increase since 1968. In
this sense, college can be seen as a “safe haven” for incoming students. However, for the college
students whom are about to graduate and do not see graduate school as a viable option, can college still
be a “safe haven”? Given the studies that show long-term consequences in entering the labor market
during a bad economy, one wonders if college students have thought about waiting to enter the labor
market until the economy picks up. This would mean that they would forgo one to two years of earning
but that might be the smarter decision in the long-run.
I am specifically interested in this question and examine the effect of labor market conditions on
college graduations rates. I look at the national and state unemployment rate as a proxy for the labor
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 3/20
3
market conditions since a higher unemployment rate usually means unfavorable job conditions for
students about to graduate.
I follow students from five cohorts, starting with the incoming class of 1999 and the next four
cohorts stopping at the incoming class of 2003. I look at the labor market conditions during the
cohort’s fourth year which is when many students have the option of graduating. For example, I look at
1999’s cohort during 2003 which is the cohort’s four -year graduate time and also 2005 which is the
cohort’s six-year graduation time. The logic is if college students notice the high unemployment during
their fourth year, they may be eligible to graduate but decide to postpone and stay in college longer in
hopes the labor market improves.
I use panel data to control for unobserved school level effects that are likely to be correlated with
the main variable of interest. For example, colleges that have always been sensitive to the state
unemployment rate have a larger difference between the four year and six year graduation rates within
the given year. Since a cross-sectional specification may produce biased estimates for the impact of the
unemployment rate on the difference between the graduation rates, I use a panel data model with
school and period fixed effects to avoid this problem.
I examine 436 U.S public four year bachelor’s degree granting institutions from 2003 to 2007,
collecting data from the Bureau of Labor Statistics and the Department of Education.
II. Prior Research
There has been an increasing amount of evidence that shows the impact of labor market conditions
at graduation affecting long-term wages and subsequent employment opportunities (Ohtake and Inoki,
1997 for Japan; Oreopoulos, von Wachter and Heisz, 2006 for Canada; Brunner and Kuhn 2009 for
Austria; Kahn 2006 for the United States). The standard neo-classical model theorizes that labor
markets function in perfectly completive spot markets however other career models based on frictions
in the labor market and human capital accumulation suggests that shocks in the labor market may have
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 4/20
4
long-run effects on wages (Oreopoulos, von Wachter, and Heisz 2006). Using regional unemployment
rates at the time of graduation Orepoulos, von Wachter and Heisz find that young graduates entering
the labor market during a recession suffer large initial earnings losses which slowly faded away only
after 8 to 10 years. Murphy and Welch (1990) estimate that around 80% of all wage increases accrue
within the first ten years of an individual being in the labor market. Other studies have found that
individuals who face poor local labor market conditions at the time of graduation are at higher risk of
unemployment not only upon graduation but subsequent years also (Raaum and Roed 2006).
Kahn’s work on the impact of labor market conditions on long-run wages show that there is
evidence of negative and persistent effects on wages in the United States. Kahn’s study shows that
using national unemployment as a proxy for labor market conditions result in a negative and
statistically significant relationship with wages both in the first year after graduation and in the long-
run. Interestingly, when the state unemployment rate is used as the proxy variable for labor market
conditions the results are weak and statistically insignificant in most cases. Kahn believes that state
rates only provide support for the national wage results. The case can be made that state
unemployment rates might not result a large effect because highly educated workers may be less
sensitive to local labor market conditions since they are more likely to migrate (Wozniak 2006).
These negative findings on labor market conditions impact on short-term and long-term may be
explained by variations in human capital accumulation at the firm level (Jovanovic 1979a) or in
general, (Neal 1999) developed between graduates during good and bad economies. Other studies
show that workers who enter firms during a recession on average have shorter employment within a
firm which might be indicative of the poor match quality (Bowlus 1995). These conclusions line-up
with studies that show early training/human capital development has positive effects on future wages
(Gardecki and Neumark 1998).
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 5/20
5
While all of the studies mentioned above have made large contributions to understanding the
impact of labor market conditions on wages, the dependent variable they used as the proxy for labor
market conditions for graduating students may not be the best specification. Using the national
unemployment rate may not be the best indictors for the labor market conditions college students face
since a study by the Department of Education shows that 69 percent of college graduates stay within
the state they earned their degree and 87 percent are working one year after graduation (Department of
Education 2000). Although my investigation does not look into the impact of labor market conditions
on wages, I believe the state unemployment rate and specifically the state unemployment rate for
individuals with a bachelor’s degree and higher would be a better proxy for the labor market conditions
college graduates face. That is why I will investigate the impact of both the national and state
unemployment rates (with and without educational attainment) on college graduation rates.
III. Data and Model
The data was obtained from the Bureau of Labor Statistics (BLS) and the Department of
Education. The Current Population Survey (CPS) conducted by the BLS provided the overall state
unemployment rates and the state unemployment rates for individuals with a bachelor’s degree and
higher from 2003 to 2007. The BLS also provided the national unemployment rates and the national
unemployment rates by bachelor’s degree and higher from 2003-2007. I collected the four-year and
six-year college graduation rates and tuition data from the Department of Education from 2003 to
2007. I chose only to look at public four-year colleges granting Bachelor’s degree or higher due to the
higher tendency of those students staying within the state they obtained their bachelor’s degree post-
graduation (Department of Education 2000). I further narrowed down my universe of school by
eliminating the schools that had gaps in their data. Further, military schools were not included due to
the unlikelihood of those graduates entering the labor market.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 6/20
6
Dependent Variable
Difference in Graduation Rate
The difference between the six-year and the four-year graduation rate at a college for a
particular cohort (same incoming class) is the key dependent variable. Both the four-year and six-year
graduation rates are by cohort so 2002’s incoming class would be measured at 2006 to obtain that
cohort’s four -year graduation rate and 2008 to obtain that cohort’s six-year graduation rate. By
subtracting the four-year graduation rate from the six-year graduation rate we get the percentage of
students that graduated not within four years but within five or six years. Both graduation rates account
for transfer students, college dropouts, and joint degree programs.
= –
Four-Year Graduation Rate
I also use the four-year graduation rate as another dependent variable. Looking at the four-year
graduation rate by itself allows us to extend our examination to not only students who postpone their
graduation but look at other possibilities such as the inability of parents to pay for college which might
also be influencing the four-year graduation rate.
Independent Variables
My explanatory variables can be divided into two categories: the national unemployment rate
and the state unemployment rate. Within both of these categories I further look at the overall
unemployment rate and the unemployment rate by bachelor’s degree and higher.
National Unemployment Rate
The national unemployment rate is very general since it includes all fifth states and the District
of Columbia and all individuals currently out of work but actively seeking work out of the total
See Table I in Appendix for Summary Statistics
See Table I in Appendix for Summary Statistics
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 7/20
7
number of employable individuals. Kahn uses the national unemployment rate as a proxy for labor
market conditions and found this measure to have a negative impact on wages and to be statistically
significant.
National Unemployment Rate by Educational Attainment
I believe the national unemployment rate by bachelor’s degree and higher would be a better
proxy compared to the overall national unemployment rate since college graduates are most likely
looking for high-skilled work. Therefore the overall unemployment rate might encompass too wide of
a net since the labor market conditions for high skilled vs. unskilled jobs might vary.
State Unemployment Rate
I look at the overall state unemployment rate since prior studies have also used this as the proxy
for labor market conditions. The rationale in using the state unemployment rate is that many college
graduates tend to stay within the state they earned their bachelor’s degree post-graduation. Logically, it
is easier for firms to recruit at colleges close by. Also since all of the colleges used in our sample are
four-year public institutions a majority of the student population live within the state.
Although previous literature has indicated that highly educated workers may be less sensitive to
local labor markets conditions because of the option to migrate, a study conducted by the Department
of Education of Bachelor’s Degree recipients show that 69 percent of graduates and 74 graduates from
a public college reside in the same state they received their bachelors from one year later (Department
of Education 2000).
State Unemployment Rate by Educational Attainment
The state unemployment rate by educational attainment is for individuals with bachelor’s
degree and higher helps us narrow down the unemployment rate even further. I decided to use this
more narrow definition as a proxy for the labor market conditions because logically college students
look for high-skilled jobs to match their skill level. This is my main variable of interest because I
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 8/20
8
believe it is the best proxy for the labor market conditions college graduates face immediately upon
graduation. This measure is defined by the amount of employable residents in a given state with a
bachelor’s degree or higher currently out of work but actively seeking work out of the total number of
employable individuals in the state with a bachelor’s degree or higher.
Tuition
I look at tuition as a control variable because tuition cost for colleges can vary within and
between states. As the price to college increases the opportunity cost of attending college becomes
higher. I use in-state tuition because most at most state schools the majority of the students are in-state
students. Theoretically tuition should have a negative relationship with the difference between the
four-year and six-year graduation rate since the longer an individual stays in school the more money
that individual will have to pay.
Model
The following equations summarize the two models I use to estimate the impact of labor
market conditions on college graduation rates.
(1)
(2)
Substitutes for : , ,
where is the overall national unemployment rate and is the national unemployment rate by
bachelor’s degree and higher, that corresponds to college i in year t . is the overall state
unemployment rate and is the state unemployment rate by bachelor’s degree and higher both which
also corresponds to college i in year t . is the log of tuition for college i in year t . is the
difference between the four-year and six-year graduation rate and is the four-year graduation rate.
and are school-level and year-level fixed effects, respectively. The key variable of interest is in
all four models.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 9/20
9
IV. Results
I look at two sets of regressions, one for each model. I start with the simplest model and build
in the appropriate controls. All of the regressions control for tuition and use clustered robust errors.
I start by looking at the impact of the national unemployment rate on college graduation.
Model 1 Labor Market Proxy Variable Coefficient
1 Random Effect National Unemployment Rate () .3912113***
2 Random Effect w/ Year
Fixed EffectNational Unemployment Rate () .3328469**
3 College & Year Fixed
EffectNational Unemployment Rate () .4043632***
The national unemployment rate is significant in all three regressions. Using the random effects
panel regression to explain the difference between the four and six year graduation rates shows a
highly significant positive coefficient for national unemployment (0.3912113, p<0.01). However, this
coefficient is flawed in that it is not accounting for upward time trends that might positively bias the
coefficients.
In regression two I add in year fixed effects which slightly lowers the national unemployment
coefficient (0.3328. p<0.05).Yet this model is also flawed in that each school may have inherent
differences that are specific to all and only that particular school which would cause different
differences between the four and six year graduation rate.
For my final regression using this model I add in school fixed effects to the regression. Using
both school and time controls allows my model to look at the data more precisely since it looks at 436
schools over a five year period. This model indicates that with all else held constant with a one
percentage point increase in national unemployment rate we see a ..4 percentage point increase in the
difference between the four year and six year graduation rate (p=0.009). Interestingly, these results
complement prior studies that show the impact of the national unemployment rate on wages.
See Table 2 for regression results.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 10/20
10
Model 1 Labor Market Proxy Variable Coefficient
6 School & Year Fixed
Effect
National Unemployment Rate:
Bachelor’s Degree & Higher ()
.5146441***
7 School & Year Fixed
EffectState Unemployment Rate () .0889683
8 School & Year Fixed
Effect
State Unemployment Rate: Bachelor’s
Degree & Higher ()
.3122881*
For my second regression, I use the national unemployment rate by bachelor’s degree and higher ()
and get similar results. Controlling for school and year effects the impact on the difference between the
four year and six year graduation rate increase to .5146. Using the overall state unemployment rate
() as my main independent variable, controlling for the year and school effects, the coefficient is
insignificant.
Finally, I use the state unemployment rate for individuals with a bachelor’s degree and higher
() as my main dependent variable and notice stay significant even when school and year effects
are controlled for. When controlling for school and year effects my model indicates that with all else
held constant with a one percentage point increase in state unemployment for individuals with a
bachelor’s degree and higher we see a .312 percentage point increase in the difference between the four
year and six year graduation rate (p=0.68).
Model 2 Labor Market Proxy Variable Coefficient
1 School & Year Fixed Effect State Unemployment Rate () -.3345804*
2 School & Year Fixed Effect State Unemployment Rate:
Bachelor’s Degree & Higher ()
-.6339153***
Looking at just the four-year graduation rate we find similar results. However, due to
collinearity, both national rates had to be dropped in my results. However, interestingly both state rates
are significant and for state unemployment by bachelor’s degree and higher our coefficient doubled.
See Table 3 for regression results.
See Table 4 for regression results.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 11/20
11
When controlling for school and year effects my model indicates that with all else held constant with a
one percentage point increase in state unemployment for individuals with a bachelor’s degree and
higher we see a .633 percentage point decrease in the four-year graduation rate. This may indicate that
some students are not only prolonging their stay in college but other students and their parents may not
be able to afford college and may have to leave school.
These results indicate that college students seem highly sensitive to the national labor market
conditions and local labor market conditions for individuals with bachelor’s degrees and higher. Both
the national and state unemployment by bachelor’s degree are significant which logically makes sense
in that college students care the most about the labor market conditions in the industries and type of
jobs they plan to work in.
I use the Hausman test on model 1 regression 4 to verify that using the fixed effect model is the
preferred model. We reject the null hypothesis that there is no systematic difference between the two
models and therefore determine that it is better to use a fixed effects model for our purposes
(p=0.0113).
In all of four models with school and year fixed effects, all else held constant log of tuitions has a
negative coefficient as predicted but is insignificant. Therefore the impact of tuition on the difference
between the four year and six year is not useful however it is still an important control variable.
V. Conclusion
The economic conditions college students face upon graduation can be very different and research
shows that students who plan to enter the labor market during a bad economy may face long-term
consequences, especially negative impacts on their wages. The lucky may find themselves in the job
they desire with relatively high wages and the not so lucky may be forced to a less promising job and
lower earnings potential. So does luck matter? The short answer is yes but the results of this paper
support the hypothesis that college students may be prolonging their stay in college in hopes that labor
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 12/20
12
market conditions improve. Since the opportunity cost of attending school decreases if potential long-
run earnings for working decreases, forgoing a year or two of earnings might not be a bad idea if it
means a smaller net loss to long-run wages.
The results indicate that a sizable number of college students have not let the business cycle control
their luck and the labor market conditions they graduate into. Both the overall national unemployment
rate and the national unemployment rate by bachelor’s degree and higher appear to influence the
difference between a cohort’s four year and six year graduation rate. The state unemployment rate by
bachelor’s degree and higher is significant but does not appear to be as strong of an indicator compared
with the national rate. These results complement Kahn’s findings that the national unemployment rate
has a strong negative relationship with both short and long-run wages. College students may be aware
of the negative relationship between the labor market conditions at graduation and wages and trying to
avoid exiting school during poor labor market conditions. The results in this study are limited since the
data does not cover a recession period and the time horizon is relatively short. Once the data from the
great recession becomes available it will add further clarification to this investigation.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 13/20
13
VI. References
Bowlus, Audra J. (1995): “Matching Workers and Jobs: Cyclical Fluctuations in Match Quality.” Journal of Labor Economics, Vol. 13, No. 2. pp 335-350.
Bureau of Labor Statistics (2011):” Geographic Profile of Employment and Unemployment (2003-
2007)” Department of Labor.
Gardecki, Rosella and David Neumark (1998), “Order from Chaos? The Effects of Youth Labor Market Experiences on Adult Labor Market Outcomes,” Industrial and Labor Relations Review,
pp. 299-322.
Jovanovic, Boyan (1979a). "Job Matching and the Theory of Turnover." Journal of
Political Economy 87 (October): 972-90.
Kahn, Lisa B. (2008): "Job Durations, Match Quality and the Business Cycle: What
We Can Learn from Firm Fixed Effects." print, Harvard University
Kahn, Lisa B. (2009): “The Long-Term Labor Market Consequences of Graduating from College
in a Bad Economy.” print, Yale School of Management.
Murphy, K. & Welch, F., 1990. "Wage Differentials In The 1980s: The Role Of The InternationalTrade," Papers 23, California Los Angeles - Applied Econometrics.
National Center for Education Statistics, (2003): “A Descriptive Summary of 1999– 2000Bachelor’s Degree Recipients 1 Year Later.” print, Department of Education.
National Center for Education Statistics, (2011): ““Four Year and Six Year College Graduation
Rates”.” print, Department of Education.
Oreopoulos, Phil, von Wachter, Till and Andrew Heisz (2006). "The Short- and LongTerm CareerE§ects of Graduating in a Recession: Hysteresis and Heterogeneity in the
Market for College Graduates." print, Columbia University.
Ohtake, Fumio, and Takenori Inoki. 1997. “Cohort effects in the labor market.” Asako, N. Yoshino
and S. Fukuda, 297 – 320. Tokyo: University of Tokyo Press.
Raaum, Oddbjørn and Knut Røed, 2006: Do Business Cycle Conditions at the Time of Labour
Market Entry Affect Future Employment Prospects?, The Review of Economics and Statistics ,
88(2) May, 193-210
Wozniak, Abigail (2006), "Why are College Graduates More Responsive to Distant
Labor Opportunities?," print, Notre Dame.
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 14/20
14
VII. Appendix
Table I: Variables, Definitions, Summary Statistics
Name Definition Mean /
(Standard Deviation)
Min/Max
Difference The difference between a school’s four
year and six year graduation rate.
23.81927
(6.822364)
3/48
ur-Year Graduation Rate The four-year graduation rate by cohort 27.05497
(14.91399)
2/85
onal Unemployment Rate The overall national unemployment
rate for each year.
5.16
(.5390114)
4.6/6
onal Unemployment Rate
y Bachelor’s Degree and
Higher
The national unemployment rate by
educational level: bachelor’s degree
and higher.
2.42
(.4262433)
2/3.1
ate Unemployment Rate The overall state unemployment ratefor each year.
5.06578(1.002893)
2.5/8.1
e Unemployment Rate by
helor’s Degree and Higher
The state unemployment rate by
educational level: bachelor’s degreeand higher.
2.282294
(.6763278)
.9/4.8
Tuition The Log of tuition cost or the change in
the tuition cost from year to year.
9.301638
(.3557994)
1,714/28,94
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 15/20
15
Table II: Regressions using equation 1 (Overall National Unemployment Rate) and clustered-robust
errors
(1) (2) (3)
DEPENDENT VARIABLES Difference (D) Difference (D) Difference (D)
National Unmployment Rate (N) 0.391*** 0.333** 0.404***(0.124) (0.146) (0.155)
LogP (P) 0.256 0.309 -0.568
(0.452) (0.453) (0.521)
t1 0
(0)
t2 -0.176 -0.0737
(0.155) (0.156)
t3 -0.0119 0.161
(0.155) (0.152)
t4 0.245
(0.182)t5 -0.274 0
(0.176) (0)
Constant 19.42*** 19.32*** 26.96***
(4.548) (4.559) (5.227)
Year Fixed Effects No Yes Yes
School Fixed Effects No No Yes
Observations 2,178 2,178 2,178
Number of id 436 436 436
R-squared 0.009Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 16/20
16
Table III: Regressions using equation 1 (National Unemployment Rate by Bachelor’s Degree andHigher) and clustered-robust errors
(1) (2) (3)
DEPENDENT VARIABLES Difference (D) Difference (D) Difference (D)
National Unemloyment Rate Bachelor'sDegree and Higher (H)
0.483*** 0.424** 0.515***
(0.158) (0.186) (0.197)
LogP 0.241 0.309 -0.568
(0.454) (0.453) (0.521)
t1 0
(0)
t2 -0.173 -0.0700
(0.155) (0.155)
t3 0.0274 0.208
(0.156) (0.155)
t4 0.245
(0.182)
t5 -0.274 0
(0.176) (0)
Constant 20.41*** 20.01*** 27.80***
(4.443) (4.422) (5.063)
Year Fixed Effects No Yes Yes
School Fixed Effects No No Yes
Observations 2,178 2,178 2,178
R-squared 0.009
Number of id 436 436 436
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 17/20
17
Table IV: Regressions using equation 1 (Overall State Unemployment Rate) and clustered-robust
errors
(1) (2) (3)
DEPENDENT VARIABLE Difference (D) Difference (D) Difference (D)
State Unemployment (U) 0.346*** 0.216 0.0890(0.0991) (0.147) (0.154)
LogP (P) 0.0918 0.290 -0.575
(0.434) (0.451) (0.520)
t1 0
(0)
t2 -0.124 -0.119
(0.100) (0.0995)
t3 -0.0509 -0.0460
(0.0775) (0.0771)
t4 -0.0505 -0.0532
(0.0691) (0.0718)t5 -0.0928 -0.0905
(0.0566) (0.0589)
Constant 21.22*** 20.24*** 28.93***
(4.283) (4.305) (4.867)
Year Fixed Effects No Yes Yes
School Fixed Effects No No Yes
Observations 2,178 2,178 2,178
R-squared 0.010
Number of id 436 436 436Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 18/20
18
Table V: Regressions using equation 2 (State Unemployment Rate by Bachelor’s Degree and Higher)and clustered-robust errors
(1) (2) (3)
DEPENDENT VARIABLES Difference Difference Difference
State Unemployment Rate Bachelor's Degree and Higher (E) 0.490*** 0.423*** 0.312*
(0.122) (0.164) (0.171)LogP (P) 0.123 0.273 -0.592
(0.427) (0.448) (0.517)
t1 0
(0)
t2 -0.107 -0.0907
(0.0969) (0.0949)
t3 -0.00893 0.00215
(0.0762) (0.0754)
t4 -0.0200 -0.00922
(0.0634) (0.0641)
t5 -0.0664 -0.0531(0.0517) (0.0528)
Constant 21.56*** 20.44*** 28.72***
(4.117) (4.192) (4.799)
Year Fixed Effects No Yes Yes
School Fixed Effects No No Yes
Observations 2,178 2,178 2,178
R-squared 0.011
Number of id 436 436 436
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 19/20
19
Table VI: Regressions using Four-Year Graduation as Dependent Variable
(1) (2)
VARIABLES Four-Year Four-Year
State Unemployment (U) -0.335*
(0.196)State Unemployment Rate Bachelor's Degree and Higher (E) -0.634***
(0.201)
t1 0 0
(0) (0)
t2 0.418*** 0.397***
(0.119) (0.117)
t3 0.416*** 0.359***
(0.0922) (0.0909)
t4 0.481*** 0.439***
(0.0824) (0.0734)t5 0.584*** 0.548***
(0.0699) (0.0638)
Constant 27.37*** 27.24***
(1.163) (0.621)
Year Fixed Effects Yes Yes
School Fixed Effects Yes Yes
Observations 1,910 1,910
R-squared 0.162 0.166Number of id 382 382
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
7/31/2019 Henry Peng Thesis
http://slidepdf.com/reader/full/henry-peng-thesis 20/20
20
Table VII: Hausman test for Model 4
Table VIII: Comparison of the Different Unemployment Rates for a Random State
0
1
2
3
4
5
6
7
2003 2004 2005 2006 2007
U n e m p l o y e n t R a t e
Labor Market Conditions (Alabama)
State Unemployment Rate
National Unemployment Rate
National Unemloyment Rate
Bachelor's and Higher
State Unemployment Bachelor'sDegree and Higher