final project paper

28
Salome Giorgadze Lillian Liang EC393 – A May 12, 2014 The Financial Crisis and Its Impact on Institutional Aid I. Introduction It is no surprise to anyone in America that the cost of attending a private college is on the rise. According to Professor Ronald Ehrenberg of Cornell University this is not a recent phenomenon, “for at least a century, tuition at selective private colleges and universities has risen annually by 2-3% more than the rate of inflation” (2002). However it was not until the 1980’s that college tuition growth finally exceeded the growth of median family income in America (Ehrenberg, 2002). As college becomes more difficult to pay for, more undergraduate students rely on financial aid to pay for tuition. Consequently, about 75% of full-time undergraduate students in the U.S. receive some form of financial aid from colleges – i.e. institutional grants, scholarships, work-study, loans (College Board). The countercyclical nature of the demand for student financial aid poses a related question of whether colleges change their 1 | Page

Upload: salome-giorgadze

Post on 16-Apr-2017

60 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Final Project Paper

Salome Giorgadze

Lillian Liang

EC393 – A

May 12, 2014

The Financial Crisis and Its Impact on Institutional Aid

I.    Introduction

It is no surprise to anyone in America that the cost of attending a private college is on the

rise. According to Professor Ronald Ehrenberg of Cornell University this is not a recent

phenomenon, “for at least a century, tuition at selective private colleges and universities has risen

annually by 2-3% more than the rate of inflation” (2002). However it was not until the 1980’s

that college tuition growth finally exceeded the growth of median family income in America

(Ehrenberg, 2002). As college becomes more difficult to pay for, more undergraduate students

rely on financial aid to pay for tuition. Consequently, about 75% of full-time undergraduate

students in the U.S. receive some form of financial aid from colleges – i.e. institutional grants,

scholarships, work-study, loans (College Board). The countercyclical nature of the demand for

student financial aid poses a related question of whether colleges change their institutional aid

giving during times of financial crisis. The aim of our project is to discover how the Great

Recession of 2008-2009 impacted the amount of institutional aid given to students. We

hypothesized that economic recessions have a negative shock on the amount of institutional aid

colleges provide to their students.

In the subsequent pages we analyze data gathered on forty private liberal arts colleges in

the United States, nine NESCAC colleges and thirty one randomly selected colleges. We look at

real GDP, state unemployment rate, institutional endowment assets from years 2004 to 2011 as

economic variables that impact institutional aid. Additionally we look at institutional expenses

1 | P a g e

Page 2: Final Project Paper

and institutional enrollment from years 2004 to 2011 and having a need-blind admissions policy

as internal factors that might also influence institutional aid giving.

The pooled data and fixed effects models that we built do not provide ample evidence in

favor of our hypothesis. They provide contrary evidence that being in recession and other

economic conditions are unrelated to the changes in institutional aid for colleges in our sample.

We discovered that enrollment and college expenses (less institutional aid expenditures) are

strong predictor variables for our sample, remaining statistically significant across various

regression models. Some curious findings from our pooled regression included that being a

NESCAC college has a negative correlation with predicted institutional aid and that being a

need-blind school has a positive correlation with predicted institutional aid. In the following

pages we provide a detailed analysis of the models, which offer compelling results that might

support the argument that internal factors are more important than external factors for colleges

when determining the amount of institutional aid to be distributed.

II.        Brief Literature Review

Much of the literature that exists in academic and professional journals primarily focuses

on the effect of financial aid packages on student enrollment and retention at universities and

colleges.  A relevant piece of literature to our project is “How University Endowments Respond

to Financial Market Shocks: Evidence and Implications” by Jeffrey Brown, Stephen Dimmock,

Jun-Koo Kang, and Scott Weisbenner.  One way in which an institution can increase its

endowment is by investing in securities.  Thus changes in endowment are sensitive to economic

shocks, which is why we use endowment as an economic indicator.  The paper by Brown et al.

(2014) explores the impact of financial markets on the changes in the endowment payout.  The

2 | P a g e

Page 3: Final Project Paper

researchers establish an “endowment hoarding” hypothesis, stating that sometimes universities

prioritize the maintenance of the size of the endowment fund over the functions of the

endowment, which is to contribute financial support to the operations and programs of the

institution. Brown et al. find evidence that the endowment payout rates grow very slowly

following positive shocks to the endowment, while the payout rates decline significantly after

negative shocks to the endowment (followed by adverse market performance). A compelling

finding is that in order to decrease endowment payouts following a negative shock, universities

cut back on the number of tenure-system faculty, whose maintenance is the most important

component of university expenses.  This asymmetric response to the negative endowment shocks

is exhibited exclusively by the universities where their current endowment is close its value at

the start of the university president’s term, which supports the endowment hoarding hypothesis.

A separate paper by Bradley and Kofoed (2013), titled “The Effect of Business Cycle on

Freshman Financial Aid,” provides empirical evidence that shows that institutional aid, alongside

with state and parental aid, follows the movement of the business cycle. However, the authors

mention the uncertainty about the changes in institutional aid during the business cycle

movements due to the endowment sizes that they initially encountered. This is a very important

point since colleges with large endowments have other means to compensate and to contribute to

the institutional financial aid during economic declines, while colleges with smaller endowments

are more reliant on their endowments to supply the institutional aid. This ambiguity is

demonstrated in one of their logit models, which shows that when the unemployment increases,

the probability of receiving institutional aid increases. The authors interpret this result as

institutions trying to fill in the gap in the aid for students when parental and state aid decreases.

Similar to these results, Long (2013) notes an increased pressure on the institutions as they

3 | P a g e

Page 4: Final Project Paper

received more applications and requests to receive more aid directly from the institution during

the crisis. In her analysis she finds that the percentage of institutional grant recipients and the

mean amount of institutional aid did not change in crisis years.

Our research differs from the one done by Brown et al. in that we focus on the impact of

the financial market shocks on different part of the university expenses - institutional grant

expenses. Furthermore, while Brown et al. are preoccupied with an important question of the

relationship between endowment payouts and endowment returns (meaning changes to the

endowment due to portfolio performance), we examine the impact of recent recession on the

institutional financial aid, one of the target activities of the endowment payouts.

Long’s research poses a research question partially similar to ours, exploring the influence of the

Great Recession on college enrollment and costs to the student families. Although the author

considers financial aid as a mixture of finances from different sources, she looks at the

institutional aid separately as well. In comparison to the three papers mentioned above, our

research poses a narrower question and investigates the impact of the Great Recession only on

the institutional aid since we focus on the decisions made solely by the colleges. Our research

allows for identifying relationships between specific variables (external and internal factors) and

the changes in the amount of institutional aid in a broad time frame from 2004 to 2011, which

helps to track the relationships throughout the time in crisis and non-crisis years.

III.       Data

For this research we gathered data on forty different not-for-profit private liberal arts

colleges in the United States that offer up to a bachelor’s degree; thirty-one colleges were

randomly selected and nine NESCAC colleges were deliberately included. We collected

4 | P a g e

Page 5: Final Project Paper

institutional data, such as expenses and institutional grants and scholarships, from the

institutions’ Form 990s from years 2004 through 2011.  We collected enrollment data and

endowment assets from The Integrated Postsecondary Education Data System.  Ideally we would

like to collect data on what portion of endowment is taken out every year to pay for institutional

grants and scholarships and the investment returns on endowment since it is directly related to

economic conditions.  Unfortunately, the Form 990s only provided such data for years 2008 to

2011.  However, with whatever data we were able to collect on endowment payout toward

institutional aid, investment returns on endowment, and contributions to endowment, we

analyzed it in a sub-data set.  GDP data and state unemployment rates were collected from the

World Bank and Iowa Community Indicators Program database respectively.  In order to account

for inflation over the course of 2004 to 2011 we converted nominal dollars into real dollars

measured in base year of 2004.  Any dollar amounts we discuss in our models are real dollars.

Primary Data Set Summary Statistics

Sub-Data Set Summary Statistics

5 | P a g e

Page 6: Final Project Paper

IV.       Empirical Work

Primary Model

A. OLS Pooled Regression:

Log(Institutional Aid)i = β0 + β1Log(GDP)t + β2Unemploymentit + β3Enrollmentit +

β4Log(Expenses)it + β5Log(Endowment Assets)it + β6I(Recession) + β7RecessionXLog(Endow)i +

β8I(Need-Blind)i + μ

In the initial pooled regression we wanted to assess whether any of the independent

variables were significant in a random effects scenario. Results for this regression can be found

on page 15, column (1). In this model the recession variable was statistically insignificant,

indicating that being in the recession had no predicted impact on the amount of institutional aid

distributed. The same was true for our interaction variable between recession and log of

endowment assets, which is affected by economic conditions, indicating that there was no

difference between the impact of change in log of endowment assets on institutional aid in

recession versus non-recession years.  Furthermore, our state unemployment variable was also

statistically insignificant. Although our recession and unemployment variables were statistically

insignificant, other economic variables, such as log of GDP and log of endowment assets, turned

out statistically significant. We found that a one percent increase in GDP relates to a 1.72 percent

increase in predicted institutional aid, holding all other variables in the model constant, whereas

a one percent increase in endowment assets relates to a 0.18 percent decrease in predicted

institutional aid, holding all other variables in the model constant.  

We found the results for log of endowment to be particularly counterintuitive since one

would expect that as endowment assets of a college increase, the institutional aid it distributes

should increase as well. However, we realize that endowment assets may not be an accurate

variable for assessing institutional aid, since it also includes land and building capital. Ideally we

6 | P a g e

Page 7: Final Project Paper

want to use the portion of endowment that is allocated toward funding institutional aid as one of

our independent variables, in addition to investment returns and contributions to endowment.

Since such data only exists for years 2008-2011, we created a sub-data set for separate analysis.

The results of that analysis can be seen on page 16 and 17 and will be discussed in depth later on.

Other variables that we found to have a statistically significant impact on log of

institutional aid were enrollment, log of expenses, and the college having a need-blind

admissions policy.  Unsurprisingly, colleges with a need-blind admissions policy are predicted to

distribute 20.85 percent more institutional aid than colleges without a need-blind admissions

policy, holding all other variables in the model constant.  We would expect colleges with a need-

blind admissions policy to be able to admit more students with financial needs and therefore to

pay out more in institutional aid than colleges who limit the number of admitted students who

have financial needs.

From this regression alone we were not able to determine whether the Great Recession

had an impact on institutional aid due to the conflicting findings - i.e. the recession and state

unemployment were insignificant, and log of GDP and log of endowment assets were significant

but their coefficients had opposite signs. Thus we produced a second model that contained

college fixed effects and time trends to see if the significance of any of our independent variables

would change.  

B. OLS Pooled Regression and Being a NESCAC College:

Log(Institutional Aid)i = β0 + β1Log(GDP)t + β2Unemploymentit + β3Enrollmentit +

β4Log(Expenses)it + β5Log(Endowment Assets)it + β6I(Recession) + β7RecessionXLog(Endow)i +

β8I(Need-Blind)i + β9I(NESCAC)i + μ

Before we moved on to our fixed effects and time trend model, we were interested in

seeing whether being a NESCAC college had an effect on the amount of institutional aid

7 | P a g e

Page 8: Final Project Paper

distributed. The results of the pooled regression with the NESCAC indicator variable can be

found on page 15 column (2).  The results show that being a NESCAC college is related to a

29.35 percent decrease in predicted institutional aid compared to a non-NESCAC college,

holding all other variables in the model constant.  We found this to be quite an alarming statistic,

and it left us wondering whether having a need-blind admissions policy factored in any way.

After the inclusion of an interaction variable between NESCAC and Need-Blind (page 15

column (3)), being a NESCAC college as opposed to a non-NESCAC college related to a 9.8

percent decrease in predicted institutional aid distributed, holding all else constant in the model.

Looking at the interaction variable, we found that being a NESCAC college with a need-blind

admissions policy is related to a 28.64 percent larger decrease in the predicted amount of

institutional grants compared to NESCAC schools without a need-blind admissions policy,

holding all other variables in the model constant.  We are unsure of why that is the case, since

the results for this seem quite counterintuitive. We suspect that it may have to do with the size of

endowment at each school and with how each school allocates its funds toward institutional aid.

C. OLS Fixed Effects and Time Trends Regression:

Log(Institutional Aid)i = β0 + β1Log(GDP)t + β2Unemploymentt + β3Enrollmentt +

β4Log(Expenses)t + β5Log(Endowment Assets)t + β6I(Recession) + β7RecessionXLog(Endow) +

β8I(Need-Blind) + β9Time + β10Time2 + μ

In this regression time trends were incorporated to take away any effects that time might

have on other independent variables in the model and to assess whether a time trend existed for

institutional aid.  Further we incorporated college fixed effects to control for conditions, such as

being a NESCAC, having a need-blind policy, or attitudes toward institutional aid, etc., that may

vary across different colleges and may have an impact on institutional aid. As shown in column

(6), there exists a quadratic time trend for institutional aid - an additional year relates to an

8 | P a g e

Page 9: Final Project Paper

increasingly larger predicted percent increase in institutional aid, holding all other variables

constants in the model - i.e. institutional aid is not increasing at a constant rate over time.  

After incorporating college fixed effects and time trends, our economic variables that

were once significant in the pooled regression, log GDP and log of endowment assets, became

statistically insignificant. Ultimately none of the variables that are related to economic conditions

are statistically significant, and thus from the evidence in this model we have to reject the

hypothesis that the Great Recession had a negative impact on the amount of institutional aid

distributed.  

Interestingly, enrollment and log of expenses continued to be statistically significant

variables in this new regression.  After running a college and time fixed effects regression that

continues to be the case. According to the final regression with college and time fixed effects, a

one person increase in enrollment is related to a 0.03 percent increase in predicted institutional

aid, holding all other control variables in the model constant. The direction of this association is

to be expected since as enrollment increases, the number of students who may need financial aid

might increase, and thus colleges might have to distribute more institutional aid. Regarding

expenses the results showed that a one percent increase in expenses relates to a 0.45 percent

increase in predicted institutional aid, holding all other variables in the model constant.

Sub-data Set:

In the analysis of the sub-data set, we developed two models: one with institutional aid as

the dependent variable and the other with grants paid out from endowment as the dependent

variable.  Results for the regressions can be found on page 16 and page 17 respectively.

Model 1 with Log of Institutional Aid as Dependent Variable:

9 | P a g e

Page 10: Final Project Paper

A. OLS Regression With College Fixed Effects and Time Trends

Log(Institutional Aid)i = β0 + β1Unemploymentt + β2Log(GDP)t + β3Enrollmentt +

β4Log(Investment Returns)t + β5Log(Contributions)t + β6Log(Expenses)t + β7I(Recession) +

β8RecessionXLog(Investment Returns) + β9RecessionXLog(Contributions) + β10I(Need-Blind) +

β11Time + β12Time2 + μ

The results from this regression can be found on page 16, column (6).  In this regression

it appears that there is a quadratic trend between time and institutional aid.  As time goes on, the

predicted percentage of institutional aid distributed increases as well.  Having incorporated time

trends, we found that none of the variables that are influenced by the economy had an impact on

the predicted amount of aid distributed with the minor exception of the interaction term between

recession and log of investment returns.  The results for the coefficient indicated that a 1%

decrease in investment returns during the recession years is related to a 0.016 percent larger

decrease in the predicted amount of institutional aid distributed as compared to non-recession

years, holding all other variables in the model constant.  This finding provides support for our

hypothesis that the Great Recession had a negative impact on institutional aid.  It is unclear why

log of investment returns itself is not a significant variable in the regression.

B. OLS Regression with College and Time Fixed Effects

Log(Institutional Aid)i = β0 + β1Unemployment + β2Log(GDP) + β3Enrollment +

β4Log(Investment Returns) + β5Log(Contributions) + β6Log(Expenses) + β7I(Recession) +

β8RecessionXLog(Investment Returns) + β9RecessionXLog(Contributions) + β10I(Need-Blind)+μ

Following the assessment of whether institutional aid followed a time trend, we ran a

regression with college and time fixed effects.  Similar to the previous regression, the interaction

term between recession and log of investment returns was statistically significant with a

coefficient of 0.0016.  After controlling for college and time effects, the recession variable was

10 | P a g e

Page 11: Final Project Paper

statistically significant, indicating that being in the recession years is related to a 16.18 percent

decrease in the predicted amount of institutional aid, holding all other variables in the model

constant.  The results from this regression provide more evidence to support our hypothesis.

It is important to make a disclaimer that the accuracy of the models and regressions made

with the sub-data is questionable, even though the p-values for the F-statistics are all less

than .05.  We suspect that this might be the case since the observations are limited to four years,

two years during the recession and two years after the recession.  This dilemma is further

compounded by that fact that there are omitted values in the data set.  It is important to keep this

point in mind in subsequent analyses of models made with the sub-data.

Model 2 with Log of Grants Paid Out from Endowment as Dependent Variable:

In this model we were interested to see whether the Great Recession had an effect on the

amount of grants paid out from the endowment.  The hypothesis is that the Great Recession had a

negative impact on the amount of grants paid out from the endowment.

A. OLS Regression with College Fixed Effects & Time Trends / College & Time

Fixed Effects

Log(Grants)i = β0 + β1Unemployment + β2Log(GDP) + β3Enrollment + β4Log(Investment

Returns) + β5Log(Contributions) + β6Log(Expenses) + β7I(Recession) +

β8RecessionXLog(Investment Returns) + β9RecessionXLog(Contributions) + β10I(Need-Blind)+μ

The results from this regression can be found on page 17, column (6).  In this regression

we denote that there exists a quadratic trend between time and institutional aid.  As time

advances, the predicted percentage of grants distributed from endowment increases as well.

Having incorporated time trends, we found that none of the variables that are influenced by the

economy had an impact on the predicted amount of grants distributed from the endowment,

11 | P a g e

Page 12: Final Project Paper

providing no support for our hypothesis.  Though the interaction variable between recession and

log of investment returns is statistically insignificant, log of investment returns on its own is

statistically significant at the 5% level, meaning that a one percent increase in investment returns

on the endowment is related to 0.15 percent predicted increase in grants distributed from

endowment, holding all other variables in the model constant.  Results further showed that

enrollment is statistically significant with a one person increase in enrollment relating to a 0.16

percent increase in grants distributed from the endowment, holding all other variables in the

model constant.  Similar findings were found in the college and time fixed effects regression.

Results for that can be seen in column (7).  

V. Conclusion

It is a common knowledge that the recent financial crisis put more pressure on all parties

involved in financing the cost of education– families, government, and educational institutions.

In our paper we focused primarily on the decisions made by institutions regarding the allocation

of the institutional financial aid to students during years of economic crisis and non-crisis.  We

investigated the hypothesis that the recent recession had a negative impact on the amount of the

institutional aid distributed by the colleges in our sample. The results from our primary model

tell a different story than our hypothesis. In all of the regression models we used for the main

data set, the variable recession and the interaction term between recession and log of endowment

were consistently statistically insignificant, having no effect on predicted institutional aid.

Moreover, after controlling for school and time effects, economic variables such as GDP and

state unemployment rate became statistically insignificant in predicting institutional aid. These

12 | P a g e

Page 13: Final Project Paper

findings suggest that the nation’s economic well-being is irrelevant to the amount of institutional

aid allocated by colleges in our sample.

Importantly, the log of endowment assets was also a statistically insignificant variable in

our primary model. Taking into account the limitations of the variable itself (endowment assets

encompass all of the land, capital, and monetary property), we investigated an additional subset

of data for years 2008 - 2011. In this additional analysis we used the portion of endowment that

is allocated toward funding institutional aid, investment returns on endowment, and contributions

to endowment as independent variables in place of endowment assets. We considered two

different models within this regression – one using log of institutional aid and the other using the

log of grants paid out from endowment as dependent variables. The model with log of

institutional aid provided some evidence that recession was impactful in predicting institutional

aid. However, we question the truthfulness of the results of these models due to the narrow time

frame of the observations and some missing values. Hence we focused our attention on the main

data set.

Though we were unable to find sufficient evidence to support our hypothesis, we did

come across interesting findings. The enrollment and log of expenses variables turned out to be

statistically significant variables in predicting institutional aid. The fixed effects model predicts

that a one person increase in enrollment is related to a .03% increase in institutional aid, holding

all other variables in the model constant. As Long (2013) notes this outcome is to be expected: as

more students enroll in a college, one can expect that the probability that those students will need

some sort of institutional aid will increase.  Hence, colleges distribute more institutional aid as

enrollment increases. Moreover, the fixed effects model predicts that a 1% increase in

institutional expenses is related to a .45% increase in predicted institutional aid, holding all other

13 | P a g e

Page 14: Final Project Paper

variables in the model constant. This is a positive sign for students of the colleges in our sample

since an increase in college expenses does not relate to a decline in the amount of institutional

aid and actually is associated with its increase.

Taken as a whole, our research has produced an interesting model that could help support

the argument that recession itself is not a significant factor in predicting institutional financial aid

unlike other factors, such as college expenses and enrollment.

References[1] Bradley, Elizabeth and Michael S. Kofoed. 2013. “ The Effect of the Business Cycle

on Freshman Financial Aid.”<http://dx.doi.org/10.2139/ssrn.2353847>.[2] Brown, Jeffrey R., Stephen G. Dimmock, Jun-Koo Kang, and Scott J. Weisbenner.

2014. “How University Endowments Respond to Financial Market Shocks: Evidence and Implications.” American Economic Review 104(3): 931-962.

[3] Ehrenburg, Ronald. 2002. “Tuition Rising: Why College Costs So Much.” <http://net.educause.edu/ir/library/pdf/ffp0005s.pdf>.

[4] “Financial Aid: FAQs.” The College Board. <https://bigfuture.collegeboard.org/pay-for-college/financial-aid-101/financial-aid-faqs>.

[5] Long, Bridget T. 2013. “The Financial Crisis and College Enrollment: How Have Students and Their Families Responded?” Unpublished.

14 | P a g e

Page 15: Final Project Paper

Dependent Variable: Log(Institutional Aid) (1) (2) (3) (4) (5) (6) (7)

VARIABLES PooledPooled+NES

CACPooled+NESCAC

*NB College FE

College FE + Linear

Time TrendCollege FE + Time Trends

College and Time FE

log(GDP) 1.7209** 1.7016** 1.6995** 1.9168*** -0.7874 1.7379(0.7547) (0.7234) (0.7169) (0.3330) (0.8680) (1.6852)

unemployment 0.0043 -0.0000 0.0005 0.0392*** 0.0067 0.0187 0.0166(0.0100) (0.0099) (0.0099) (0.0047) (0.0116) (0.0158) (0.0165)

enrollment -0.0002*** -0.0002*** -0.0002*** 0.0003*** 0.0003** 0.0003** 0.0003**(0.0000) (0.0000) (0.0000) (0.0001) (0.0001) (0.0001) (0.0001)

log(Expenses) 1.3242*** 1.4054*** 1.3989*** 0.3999*** 0.4086*** 0.4222*** 0.4498***(0.1005) (0.0996) (0.0991) (0.1358) (0.1417) (0.1423) (0.1491)

log(Endowment Assets) -0.1846*** -0.1863*** -0.1717*** 0.0162 0.0916 0.0920 0.0486(0.0374) (0.0356) (0.0358) (0.0656) (0.0590) (0.0593) (0.0862)

recession -0.2207 -0.2240 -0.2245 -0.0752 -0.0437 0.0063(0.5768) (0.5215) (0.5162) (0.1431) (0.1512) (0.1515)

recessionXlog(endowment assets) 0.0131 0.0134 0.0135 0.0062 0.0037 0.0050 0.0048

(0.0306) (0.0276) (0.0273) (0.0076) (0.0079) (0.0077) (0.0076)NeedBlind 0.2085*** 0.1136*** 0.2701***

(0.0365) (0.0340) (0.0524)NESCAC -0.2935*** -0.0980**

(0.0480) (0.0488)NESCACXNeedBlind -0.2864***

(0.0700)t 0.0520*** -0.0432

(0.0167) (0.0527)t2 0.0086**

(0.0036)Constant -55.3676** -56.0239** -56.2658*** -49.6099*** 30.5188 -45.8749 6.8420**

(22.7534) (21.7909) (21.5868) (8.7544) (26.3633) (51.9319) (2.8042)

Observations 320 320 320 320 320 320 320R-squared 0.8070 0.8217 0.8251 0.7465 0.7694 0.7777 0.7800Number of id 40 40 40 40

Robust standard errors in parentheses

15 | P a g e

Page 16: Final Project Paper

*** p<0.01, ** p<0.05, * p<0.1

Dependent Variable: Log(Institutional Aid) (1) (2) (3) (4) (5) (6) (7)

VARIABLES PooledPooled + NESCAC

Pooled + NESCAC*Need

Blind College FE

College FE + linear time

trend

College FE + quadratic time

trendCollege + time

FE

unemployment -0.0845*** -0.1051*** -0.1051*** -0.0104 -0.0104 -0.0104 -0.0104(0.0198) (0.0196) (0.0196) (0.0138) (0.0138) (0.0138) (0.0138)

log(GDP) -11.6666 -7.1991 -7.1991 5.7667*(7.3457) (6.3885) (6.3885) (3.1788)

enrollment -0.0001* -0.0001** -0.0001** 0.0002 0.0002 0.0002 0.0002(0.0000) (0.0000) (0.0000) (0.0002) (0.0002) (0.0002) (0.0002)

log(Investment Returns) -0.0883** -0.0580 -0.0580 -0.0122 -0.0122 -0.0122 -0.0122

(0.0391) (0.0357) (0.0357) (0.0176) (0.0176) (0.0176) (0.0176)log(Contributions) 0.1254*** 0.1101*** 0.1101*** 0.0044 0.0044 0.0044 0.0044

(0.0407) (0.0390) (0.0390) (0.0082) (0.0082) (0.0082) (0.0082)log(Expenses) 0.7184*** 0.7855*** 0.7855*** 0.0795 0.0795 0.0795 0.0795

(0.1226) (0.1121) (0.1121) (0.0492) (0.0492) (0.0492) (0.0492)recession -0.7862 -0.5660 -0.5660 0.0606 0.0122 -0.0226 -0.1618**

(0.6813) (0.5645) (0.5645) (0.0855) (0.0659) (0.0556) (0.0738)recessionXlog(Investment Returns) 0.1001 0.0603 0.0603 0.0161** 0.0161** 0.0161** 0.0161**

(0.0803) (0.0728) (0.0728) (0.0077) (0.0077) (0.0077) (0.0077)recessionXlog(Contributions) -0.0789 -0.0415 -0.0415 -0.0123 -0.0123 -0.0123 -0.0123

(0.0749) (0.0700) (0.0700) (0.0081) (0.0081) (0.0081) (0.0081)NeedBlind 0.1326** -0.1288* -0.1288*

(0.0563) (0.0687) (0.0687)NESCAC -0.3914*** -0.3914***

(0.0858) (0.0858)t 0.0870*

(0.0479)t2 0.0174*

(0.0096)Constant 356.6903 220.7535 220.7535 -159.2849 14.7552*** 14.8596*** 15.0162***

(221.3861) (192.5282) (192.5282) (96.4776) (0.9590) (0.9280) (0.8865)

16 | P a g e

Page 17: Final Project Paper

Observations 95 95 95 95 95 95 95R-squared 0.8590 0.8822 0.8822 0.7546 0.7546 0.7546 0.7546Number of id 38 38 38 38

Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

Dependent Variable: Log(Grants) (1) (2) (3) (4) (5) (6) (7)

VARIABLES PooledPooled + NESCAC

Pooled + NESCAC*NeedBl

ind College FE

College FE + linear time

trend

College FE + quadratic time

trend College + time FE

unemployment 0.0741** 0.0648* 0.0648* 0.1161 0.1161 0.1161 0.1161(0.0353) (0.0373) (0.0373) (0.0897) (0.0897) (0.0897) (0.0897)

Log(GDP)69.9463**

* 71.6470*** 71.6470*** 29.4313**(19.0182) (18.9030) (18.9030) (14.4025)

Enrollment -0.0002** -0.0002** -0.0002** 0.0016* 0.0016* 0.0016* 0.0016*(0.0001) (0.0001) (0.0001) (0.0009) (0.0009) (0.0009) (0.0009)

log(Investment Returns) 0.3799*** 0.3927*** 0.3927*** 0.1517** 0.1517** 0.1517** 0.1517**(0.1072) (0.1076) (0.1076) (0.0716) (0.0716) (0.0716) (0.0716)

Log(Contributions) 0.3100*** 0.3002*** 0.3002*** -0.0026 -0.0026 -0.0026 -0.0026(0.1000) (0.0992) (0.0992) (0.0647) (0.0647) (0.0647) (0.0647)

Log(Expenses) 0.3246 0.3611 0.3611 2.3346 2.3346 2.3346 2.3346(0.2586) (0.2607) (0.2607) (1.6813) (1.6813) (1.6813) (1.6813)

recession 1.1296 1.2159 1.2159 0.2153 -0.0318 -0.2094 -0.9197(1.2954) (1.3019) (1.3019) (0.8638) (0.8270) (0.8108) (0.8379)

recessionXlog(Investment Returns) 0.1364 0.1209 0.1209 0.0049 0.0049 0.0049 0.0049

(0.1400) (0.1410) (0.1410) (0.0722) (0.0722) (0.0722) (0.0722)recessionXlog(Contributions) -0.1041 -0.0897 -0.0897 0.0382 0.0382 0.0382 0.0382

(0.1547) (0.1545) (0.1545) (0.0901) (0.0901) (0.0901) (0.0901)NeedBlind 0.0227 -0.0668 -0.0668

(0.1611) (0.1860) (0.1860)NESCAC -0.1423 -0.1423

(0.1541) (0.1541)t 0.4439**

17 | P a g e

Page 18: Final Project Paper

(0.2172)t2 0.0888**

(0.0434)

Constant

-2,115.566

2***

-2,167.4698

*** -2,167.4698*** -922.6715** -34.4240 -33.8913 -33.0922(573.0553

) (569.6529) (569.6529) (435.0604) (30.3470) (30.3539) (30.3685)

Observations 89 89 89 89 89 89 89R-squared 0.9046 0.9052 0.9052 0.2529 0.2529 0.2529 0.2529Number of id 35 35 35 35

Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

18 | P a g e