1 Maloney 1 Does it Pay to Attend an Elite Liberal Arts College? Paul Maloney Abstract: One of the most important decisions in a person s life is what college they will attend. The choice of college can have an effect on future earnings. With the increase in college tuition, it has become even more important to make cost effective decisions. This paper studies how wages are affected by the type of school an individual attends and in particular elite liberal arts colleges. Using data from the National Longitudinal Survey of the Youth 1979 (NLSY79) and the U.S. News and World Report s individual characteristics are linked to characteristics of the college that the individual attended. Individual s wages are regressed on the individual and college characteristics at different points in the individual s career in order to determine the percent change in wages when attending an elite liberal arts college. The findings suggest there is not a statistically significant relationship between wages and attending an elite liberal art college at the beginning of a career but in the middle of an individual s career there is a positive and statistically significant relationship. This research was conducted with restricted access to Bureau of Labor Statistics (BLS) data. The views expressed here do not necessarily reflect the views of the BLS.
2 Maloney 2 I. Introduction One of the most important decisions in a person s life is what college they will attend. The decision will in large part determine the people that they will meet, the passions they will pursue and the career path they will explore. The choice of college can also have an effect on future earnings. With the increase in college tuition, it has become even more important to make cost effective decisions. One type of college that has seen a drastic increase in tuition is liberal arts colleges. This increase leads to the important question; does an elite liberal arts undergraduate education pay off? For the academic year the comprehensive fees of the top liberal arts colleges in the country are over $50,000. Williams College costs $54,560, Amherst College, $53,370, Swarthmore College, $53,250, Middlebury College, $53,420, Wellesley College, $53,250 and the College of the Holy Cross, $52,758 (www.collegeboard.org). With the costs of these colleges rising it is important to determine if the investment pays off in the future. Much of the previous literature has found that there is a lower premium to attending liberal arts colleges in comparison to other types of colleges and universities. This study seeks to improve the literature in two ways. First, it will narrow the focus to elite liberal arts colleges. When looking at the premium to attending liberal arts colleges, the literature groups all liberal arts colleges, regardless of quality into one single category. Second, this study will improve the literature by trying to reduce the omitted variable bias that has resulted by using single proxy measures for the quality of a school. Doing so will demonstrate which college characteristics contribute most to an individual s income. This study will help in determining whether or not the investment of attending an elite liberal arts college is worth it in terms of future earnings. Many top students are often deciding between prestigious liberal arts colleges, and prestigious universities and research institutions.
3 Maloney 3 This study will allow them to see the trade off made by attending one type of school versus the other and shed some light on one of the most difficult decisions that a young adult has to make. The next section presents previous research on this topic. After the relevant literature is discuss, the human capital investment model that this study employs will be presented. This model serves as the framework for analyzing the relationship between school type and wages. The data is then discussed and finally the results of both OLS regression and matching are presented. II. Background Literature As stated above, this topic is something that much of the economic literature has covered, yet the past literature does not use an approach that is entirely adequate in dealing with the nuances of the research question. However, several studies exist that have contributed to this study. The literature uses the human capital investment model, the model that will be utilized in answering the research question. These studies focus on different areas of higher education and different types of colleges and universities serve as the focus. In addition, a study that looks at the flaws in the literature provides motivation for the need for a stronger, improved study. Brewer et al. (1999) explores the impact of attending a certain type of college on future earnings. They looked at six different cohorts of institutions: elite private, elite public, middle private, middle public, bottom private and bottom public. These cohorts are derived from various editions of Barron s Profile of American Colleges. Schools are grouped into the six distinct categories on the basis of entering students class rank, high school grade point average, average SAT scores, and the percentage of applicants admitted. In their first regression they employ the human capital investment model and regress the logarithm of individual earnings on a set of individual characteristics and a set of college characteristics. The evidence suggests that the premium to attend an elite institution increased for the group of individuals that attended
4 Maloney 4 college in the 1980s compared to those who attended in the 1970s (Brewer et al. 1999). This suggests that attending an elite institution increases the amount of money that an individual earns. The human capital investment model applies to this study because education results in an increase in human capital. Similar to how Brewer et al. (1998) and Brewer et. al (1999) look at undergraduate education as an investment in future outcome, this study will look at the investment in undergraduate education where elite liberal arts education is considered and focused upon. As more is invested into education, one s pay, represented by wages, stands to increase. The type of education, and the level of education one receives, will also have an effect on how much can be earned in the future. Brand and Halaby (2005) look at a high school graduation and college cohort across four decades of labor participation and estimate the returns to attending an elite college at distinct points along the career path. They define elite using Barron s Profile of American Colleges. When using regression to determine the effects they found that attending an elite college increases the probability of graduating, attending graduate school, and the salary of a first job (Brand and Halaby 2005). The results imply that there is a premium to attending an elite college at the onset of your career. When looking at the effects on mid-career salaries Brand and Halaby (2005) found by using both regression and matching techniques that there is a positive but statistically weak correlation between elite college attendance and mid-career wages. It appears that over time the effects of attending a certain type of college has little bearing on salary as presumably experience and success within the work force take on a large amount of explanatory power. However, when looking at late-career earnings they found a positive effect to attending an elite college. Their
5 Maloney 5 results are mixed as at different points along the career trajectory there are different effects on wages yielding no conclusive results to the effects of elite college attendance on wages. The findings differ from Brewer et al. (1998) demonstrating the need for more study. Another study that uses the aforementioned human capital accumulation model is one conducted by James Monks (2000). His results shed light on the returns to attending a liberal arts college, the education of focus in this study. Monks (2000) uses the human capital investment model and regresses log of wages on the independent variables of experience, tenure, male, white, Armed Force Qualification Test, public institution, master, doctoral or research institution, specialized institution, non or less competitive, very competitive, highly or most competitive, and net family income. The omitted variables are non-white, female, competitive, private, and liberal arts colleges. Monks makes several conclusions in regards to liberal arts education. He finds that graduates from graduate degree granting research institutions earned approximately 14% more than graduates from liberal arts colleges (Monks 2000). In comparison to specialized institutions, such as undergraduate business schools, he found that graduates from specialized institutions earned approximately 19% more than liberal arts college graduates (Monks 2000). Based on his research it appears that there is a premium for graduating from a research institution relative to graduating from a liberal arts institution. While he does present a compelling argument as to why there is a greater return to attending other types of institutions in comparison to liberal arts institutions, there are some flaws in the paper. The biggest issue is that he groups colleges according to the Carnegie Classification System while not taking into account the fact that colleges in each of the groups are very different. In Monks research a school such as Williams College is considered similar
6 Maloney 6 to that of a school at a much lower tier. The only control he uses is a measure of competitiveness and this single selectivity control does not adequately cover the nuances of what makes a school like William a top tier, elite, liberal arts college. The range in liberal arts colleges from top to bottom is large but in the regression they are all thought of as being equal. While his model does allow the ability to look at liberal arts and competitive simultaneously and determine the effects on wages, a better set of college characteristics is needed so that a greater understanding of the affects of education on wages can be determined. Monks findings are extremely interesting in that they demonstrate that there may in fact be a larger premium to attending other types of universities versus liberal arts colleges. However, it appears that he does not adequately address the fact that the quality of liberal arts colleges covers a wide spectrum. With such differing quality, it seems reasonable to assume that attending an elite liberal arts college would have a far different impact on future earnings than attending an average or below average liberal arts college. In order to better control for different types of schools and different levels of schools, more measures of quality appear to be needed. Black and Smith (2006) look at the idea of including multiple measures of college quality in the human capital investment model used in much of the education literature. Their paper reconsiders the standard education function in a context where multiple measures of college quality are available and provides motivation for the need to enhance the studies that pertain to liberal arts education. The authors look at the current literature surrounding how economists have controlled for college quality in their research. What they realized was that much of the literature uses a single proxy in order to control for varying levels of college quality. One of the most common measures of college quality is the average SAT score of each incoming class. While this is a
7 Maloney 7 measure of college quality, they argue that it is too simplistic. The fact is that schools have multiple dimensions, and the ability to estimate them with one single factor is a very strong assumption. Consider schools such as University of Chicago and MIT. By much of the current literature these schools are considered to be equal as both would fall into the highly competitive, and masters, doctoral and research categories. However, the University of Chicago is known to specialize in liberal arts training where as MIT is known to excel in technical training. Including them in the same category can cause misleading results. The authors claim that using single proxy likely results in underestimates for the effects of college quality on the labor market (Black and Smith 2006). By making use of four additional proxy measures, faculty to student ratio, rejection rate, freshman retention rate, and mean faculty salaries, a GMM estimator suggests a downward bias of around 20 percent relative to using the SAT variable as a single proxy for quality. This shows that the existing literature underestimates the wage effects of college quality. There is a clear value in using multiple proxies in estimating the wage effects of college quality. In one of the more recent papers written about the topic Dale and Krueger (2011) estimate the monetary return to attending a highly selective college. Their findings demonstrate the bias that can result by not using proper controls as was suggested in Black and Smith (2006). What they find is that there is a sizeable return to college selectivity when they controlled for commonly observed variables such as student high school GPA and SAT scores. However, when they include unobserved student ability, by including college quality proxies (in this case average SAT scores of the schools the student applied to), the return fell substantially and was generally indistinguishable from zero. This paper demonstrates how important it is to control for as much as possible so that the results of the study are not skewed.
8 Maloney 8 Black and Smith (2006) demonstrate the need to improve the current literature. Both Brewer et al. (1999) and Monks (2000) will serve as frameworks for my study. I will utilize the model that these papers employ in order to determine the premium of attending an elite liberal arts college. The way in which Brewer et al. (1999) and Brand and Halaby (2005) classify the schools they are studying by elite, middle, and bottom show the value of focusing on elite liberal arts colleges. Approaching my question with the Brewer et. al. and Monks studies as frameworks will allow me to better understand my results and how using multiple proxies for quality, as suggested in Black and Smith (2006) and demonstrated in Dale and Krueger (2011), affects the estimated premium of attending certain types of colleges. III. Method Defining Key Terms A key element to this study will be defining the research question: Does it pay to attend an elite liberal arts college? Elite will be defined as any liberal arts college that appears in the top 50 of the U.S. News & World Report s National Liberal Arts Ranking. Pay will be determined solely in relation to wages. This study in no way claims that success is achieved by a higher level of pay; therefore we limit the study to pay in terms of monetary compensation, and do not look at success. Finally, a Liberal Arts College will be defined using U.S. News & World Report s system of rankings. Human Capital Investment Model The human capital investment model that this study will employ is a model that relates wages to labor market experiences, individual characteristics, and college characteristics. The model can be seen by the equation:
9 Maloney 9 (1) Ln(Wages)= f(labor market experiences, individual characteristics, college characteristics) By investing in education, a form of human capital, the equation shows that this will have an effect on future wages. The functional form, log-linear, is the form accepted by the literature as the way in which this particular model behaves. The above equation can be viewed as the population model: (2) Ln W i = β 0 + β 1 L i + β 2 X i + β 3 C i + ε Where W i is equal to an individual s wage, L i is equal to labor market experiences such as experience and tenure, X i equals individual characteristics such as gender and race, and C i equals college characteristics such as type of college, tuition, total enrollment, acceptance rate, freshman retention rate, six year graduation rate, percent of classes under 20 students and SAT/ACT national percentile. The ε represents the random error resulting from the inability to account for all variables that affect an individual s wage. The variable of interest in this study will be the variable pertaining to elite liberal arts colleges (included in the college characteristics in the above equations). IV. Data The college characteristic data utilized in this study comes from the U.S. News and World Report s College Compass The data set contains information on over 1600 colleges and universities in the United States including metrics on academics, cost and financial aid, selectivity, student satisfaction, and the student body. This data set will be linked to data from the National Longitudinal Survey of the Youth 1979 (NLSY79). The NLSY79 contains responses from 12,686 men and women who were ages when they were first surveyed in They were surveyed annually until 1994, and biannually until The size of this data
10 Maloney 10 set will be crucial in making sure that the results of the study are as accurate as possible. From the NLSY79 the independent variables such as gender and mother s and father s education are made available as well as the dependent variable, wages. Through NLSY Geocode data Federal Interagency Committee on Education (FICE) codes are provided for each individual who attended college. Every college and university in the country has an FICE code which will allow the two data sets to be linked. Due to the fact that much of the previous literature finds that there is a premium to attend an elite university, it is important to compare the metrics of the elite universities and elite liberal arts colleges. It would makes sense that if they are not all that different statistically then there may in fact be a premium to attending an elite liberal arts colleges just as there is for attending an elite university. A sample of the top 50 liberal arts colleges in the U.S. News and World Report s rankings was taken in order to obtain summary statistics of these elite liberal arts colleges. The results can be seen in Table 1. When comparing this sample to a sample of the top 50 national universities (see Table 2) we can see how elite liberal arts colleges measure up to the elite national universities. Since much of the previous literature has used average SAT score as the main control variable for college quality it is interesting to note that the schools in the top 50 national universities had an average SAT score of 1365, only 25 points higher than that of the top 50 liberal arts colleges. The average SAT score for the top 50 national universities ranged from 1195 to 1525 where as it only ranged from 1240 to 1485 for the top 50 liberal arts colleges. While the top national universities had the highest SAT scores, on average there was not much of a difference between the two.
11 Maloney 11 One area in which it appears that liberal arts colleges score better than the national universities is in financial aid. While on average the two samples of schools have an almost identical total cost, $49, for liberal arts versus $50, for national universities, liberal arts colleges, provide on average, more financial aid than national universities. The average financial aid package of the liberal arts college s was $34, compared to just $29, for national universities. With the increasing financial burden that college places on families, a $5,000 difference can be a deciding factor in which school a student ends up attending. In addition to financial aid, the sample of liberal arts colleges also receives greater support from their alumni. On average, approximately 39% of all alumni from the top 50 liberal arts colleges donate to their alma mater, where as only 25% of those having attended a top 50 national universities donate. While the 25% stands to represent a greater monetary value, as national universities have far more students, and therefore far more alumni, it is never the less interesting to note. It indicates that alumni feel a strong connection to their school and value the education and experience that they received. In addition to feeling more of a connection, presumably more are donating because they are doing better off financial. If this were the case then it may appear that there is a premium to attending an elite liberal arts college. One metric of note in which there is a significant discrepancy between national universities and liberal arts colleges is the diversity index. In calculating the diversity index, U.S. News factors in the total proportion of minority students as well as the overall mix of groups. The formula they utilize produces an index ranging from 0 to 1, with a score closer to 1 indicating a more diverse student population both in terms of number of minority students and mix of ethnic groups. The top national universities had a diversity index of.524 and liberal arts
12 Maloney 12 colleges an index of.396. The national universities are doing a much better job of attracting students from a multitude of different backgrounds. To further the comparison between the elite liberal arts colleges and the top national universities the sample of 50 schools in each category was downsized to include only the top 10 in each. The 20 schools that were looked at include the elite of the elite in the county. They include schools such as Williams, Amherst, Swarthmore, and Middlebury on the liberal arts list, and schools such as Harvard, Princeton, Yale, and MIT on the national universities list. The summary stats of these schools are shown in Table 3 and Table 4. In reducing the sample size to the top 10 in each category we see that there is a widening in the gap between the average SAT scores. The top 10 national universities had an average SAT score of 1479, over 60 points higher than the top 10 liberal arts colleges which had an average of This may be attributed to a greater focus on SATs in the admissions process at the top national universities versus liberal arts schools which have smaller applicant pools and thus do not need to focus as heavily on standardized testing when reviewing applications. Aside from the SAT scores, which were more comparable when looking at the top 50 in each category, there are several metrics that speak to college quality that are practically equal. The most interesting statistic as it relates to college quality being the percent of classes with less than 20 students. One can assume that students learn better in smaller classes where they are able to receive more attention from professors and are able to learn from other students through class discussion. When comparing the top 50 in each category we see that on average liberal arts colleges had 68% of their classes with less than 20 students versus only 58% percent for national universities. When looking only at the top 10 schools in each category these numbers become
13 Maloney % and 72.5% respectively. These numbers are basically equal indicating that students are receiving similar classroom experiences at both types of schools. When looking at the top 50 we saw that liberal arts schools did a much better job in providing aid. When looking only at the top 10 we find that both types of schools give very similar aid packages with an average of approximately $38,000 for liberal arts colleges and $39,000 for national universities. It seems as if students have equal opportunities to attend both types of schools when it comes to the financial decision. Something that was rather surprising to find that did not change from looking at the top 50 schools versus the top 10 schools was that a larger percentage of alumni donated to the liberal arts colleges versus the national universities, 50% versus 38%. While both of these numbers are much higher than they were when looking at the top 50, the gap between the two is even more surprising. As indicated by more classes having less than 20 students it would appear that students attending top 10 universities would feel a similar connection to the school as those who attended a liberal arts college with similar class sizes, but the alumni giving rate does not indicate that. It continues to indicate, as it did before, that individuals who attend a liberal arts school have a higher probability that they donate to their alma mater. The summary statistics support the fact that while elite liberal arts colleges and elite national universities are very different, they provide very similar qualities of education. With similar education it would stand that the pay off would be similar in terms of wages. Prior to looking at summary statistics of the individuals being surveyed, it is important to look at the summary statistics of all schools that are in the final data set. Table 5 includes summary statistics of the college characteristics that are used in the OLS regressions. These summary statistics correspond to the colleges that individuals in the NLSY attended. The table
14 Maloney 14 indicates that the mean total undergraduate enrollment is almost 16,750, over 14,000 more than that of the top 50 liberal arts schools shown in Table 1. The cost of these schools is also far less than that of both the top 50 liberal arts and top 50 national schools. This makes sense as the distribution of tuition is not even. The top schools cost significantly more than the average schools. As both top 50 liberal arts and top 50 national appear to be far different than the average institutions of higher education, it would imply that they have different effects on wages than average schools. In addition to looking at the summary statistics in regards to institutional level characteristics, it is important to take a look at the individual level data provided by the NLSY79. Table 6 displays some of the key metrics that will be utilized in this study. Table 6 indicates that there is a relatively even split between males and females in the survey. A very large majority, nearly 80%, of those surveyed lived in either a city or town versus a rural area. When looking at the educational level of the respondents parents, the average highest level of schooling completed for both parents was approximately the 11 th grade. On average, the parents of those surveyed do not have a high school diploma. This is not entirely shocking given the time frame in which their parents would have been attending school (1950s) but it is still interesting to note never the less given its effect on the schooling of the individual. Table 6 displays the summary statistics of wages, reported as annual salary, earned by respondents over the period of time surveyed. As wages is the dependent variable of this study, it is important to take a closer look at these statistics. The wages that are reported in the NLSY have been truncated so that the top 2% of wages are replaced with the group average. This was done in order to give a more accurate portrayal of the wages of the respondents and remove some
15 Maloney 15 of the outliers. As can be expected, the average salary of respondents increased as time went on. In looking at Table 6 we see that the standard deviations are rather large in all years and in both 2002 and 2008, larger than the mean. This indicates an extremely large fluctuation in wages of the respondents. Part of this could be due to people leaving the work force and thus reporting a wage of 0. V. Results I begin by regressing the log of yearly wages against all of the individual and college characteristics, and a dummy variable for top 50 liberal arts and top 50 national. These results are presented for the years 1985, 1990, 1996, 2002, and 2008 in Model 1 of Tables These years correspond to different points in an individual s career and allow us to see how type of college attendance influences wages over time. As individuals surveyed were between the ages of 14 and 22 in 1979, 2008 wages reflect a mid-career salary when individuals are in their mid to late 40s. As evident in all five years, males appear to earn significantly more than females and there appears to be a difference for white versus non-white college graduates at the onset of their career (as indicated in Table 7 and Table 8) and towards the middle of their career (as indicated in Table 11). Model 1 does not result in any statistical significance for the variable of interest, top 50 liberal arts for any of the five years. This may be due to a high level of multicollinearity as all of the college characteristics are used in determining which schools are top 50 liberal arts and top 50 national. In an effort to avoid multicollinearity, Model 2 drops out the dummy variables of top 50 liberal arts and top 50 national in order to look at how the college characteristics relate to wages. Table 7 (1985) shows that the 6 year graduation rate is negatively correlated and statistically significant to wages and all others are not. In the following years the only college characteristics
16 Maloney 16 that have statistical significance are tuition in 1990, 1996, and 2002, total enrollment in 1990, and percent of classes under 20 students in Model 3 attempts to do the reverse of Model 2 where all college quality measures are dropped and the dummy variables for top 50 liberal arts and top 50 national are put back into the regression. Top 50 national is statistically significant in all years except 1985 whereas top 50 liberal arts is statistically significant only in This indicates that an individual receives the benefits from attending a top 50 national school earlier in the career path than an individual who attended a top 50 liberal arts school but in the end they are both positively correlated to wages. Table 11 indicates that the coefficient on top 50 liberal arts is 0.63, almost three times that of the coefficient on top 50 national, Therefore, holding all other independent variables constant, attending an elite liberal arts college versus all others is associated with a 63% increase in yearly wages. This is in contrast to attending a top 50 national university over all others where there is only a 23% increase in yearly wages. This would indicate that not controlling for any measures of college quality results in a higher premium to attending a top liberal arts school over all other schools versus attending a top 50 national institution over all other schools. While this result seems to support the claim that these top schools are similar and it does in fact pay to attend an elite liberal arts college, it is important to use some measures of college quality as to not present a bias result. Models 4-6 attempt to reduce the bias by introducing some of the college quality characteristics as control variables. As indicated by the summary statistics related to top 50 liberal arts and top 50 national, national schools are much larger than liberal arts schools and thus it is important to control for size. Model 4 includes the variable percent of class under 20 students as a means to control for this institutional characteristic. Controlling for size does not change any of the results from
17 Maloney 17 Model 3 as top 50 national is still statistically significant in all years except 1985 and top 50 liberal arts is statistically significant only in The coefficient on top 50 liberal arts in Table 11 is still much greater than that of top 50 national. Model 6 also attempts to control for size by adding total enrollment in addition to the percent of classes under 20 students as control variables. This allows not only the classroom size to be controlled for but also the total student population. Controlling for size in this way causes some shifts in the significance of top 50 national schools as it is now significant at a higher p-value in 1990 (Table 8) and 2008 (Table 11) and there is no longer a significant relationship in The only change in regards to top 50 liberal arts is that the coefficient is now significant in Model 5 uses the college quality measure that much of the previous literature uses as a measure of intelligence, average SAT/ACT scores. In this study SAT/ACT scores are presented as a national percentile where each individual institution s average standardized test scores were converted to their national percentile equivalent. This allows schools that only report ACT scores or only report SAT scores to be compared. Using standardized testing as a proxy for intelligence has rather drastic effects on the coefficient of the top 50 national schools. Controlling for intelligence in this way results in the variable, top 50 national, being insignificant in all years included in this study. In looking specifically at Table 11, the t-stat drops from 2.44 in Model 3 to 0.61 where as liberal arts remains statistically significant. These results are rather eye opening as they support the notion that it is not the fact that an individual went to a top 50 national university that causes them to earn more money, but it may be due to the fact that the individual is a more intelligent person and thus earns a higher wage. The same cannot be said for top 50 liberal arts as even when intelligence is controlled for liberal arts is still statistically significant.
18 Maloney 18 The results of Model 5 demonstrate something that much of the literature attempts to control for, a selection bias. It is reasonable to assume that the top students choose to attend the top schools. The problem is that this study and studies like it are trying to look at how the college type itself affects wages and when top schools attract top students it is likely that much of the positive effects on wages result not from the institution but from the individual. Model 5 indicates that the reason that top 50 national has a positive, statistically significant effect on wages in most other models is a result of the fact that the most intelligent students, who stand to earn the most money regardless of the school they attend, are going to top 50 national schools. In an effort to control for this selection bias this study looks at the average treatment effect on the treated (ATT). ATT controls for selection as people choose the school they attend for different reasons. ATT matches individuals with similar characteristics with the only difference being that one attended a top 50 liberal arts college and the other did not. This allows the effect of which school the individual attended to be determined. All of the individual characteristic variables including whether or not a foreign language was spoken at home, where the individual lived, mother s and father s education, race, gender, and tenure were used in the matching process. With selection on observables taken into account, Table 12 indicates that attending a liberal arts college increases the natural log of wages in 2008 by.775 over if the student did not attend a liberal arts college. This can be interpreted as a 78% increase in yearly wages for an individual who attended an elite liberal arts school. This matching result is similar to that of the OLS results found in Table 11, Model 5 where SAT/ACT national percentile is controlled for. There appears to be a statistically positive relation between attending an elite liberal arts school and wages in the middle stages of an individual s career.
19 Maloney 19 VI. Conclusions This study adds to the literature in that it demonstrates that not only is there a payoff to attending an elite college, but specifically a payoff to attending an elite liberal arts college. While this benefit is not realized at the beginning of an individual s career, as time goes on the effects are realized. This study displays that there is not a statistically significant relationship between wages and attending an elite liberal art college at the beginning of a career but in the middle of an individual s career there is a positive and statistically significant relationship between the two. This is supported by both OLS regression as well as ATT matching which controls for selection bias. This finding is similar to that of Brand and Halaby (2006) where the effects of attending an elite college are found to increase over time. They claim that the education that you receive from an elite school may not be realized until later in a career. This study is limited in that the number of individuals in the NLSY79 that attended an elite liberal arts college is rather small compared to the number that attended other types of schools. This may limit the scope to which these results can be extended. It is clear that a data set that has a more even balance of individuals attending different types of schools would yield results that could be extended to all of higher education. Never the less, the findings of this study demonstrate that more research is needed to determine the effects of attending an elite liberal arts college on future earnings.
20 Maloney 20 References Black, Dan A., and Jeffrey A. Smith. Estimating the Returns to College Quality with Multiple Proxies for Quality. Journal of Labor Economics 24.3 (2006): Brand, Jennie E., and Charles N. Halaby. Regression and Matching Estimates of the Effects of Elite College Attendance on Educational and Career Achievement. Social Science Research 35 (2006): Brewer, Dominic J., Eric R. Eide, and Ronald G. Ehrenberg. Does It Pay to Attend an Elite Private College? Cross-Cohort Evidence on the Effects of College Types on Earnings. The Journal of Human Resources 34.1 (1999): College Board. 23 Mar < Dale, Stacy, and Alan B. Krueger. Estimating the Return to College Selectivity over the Career Using Administrative Earning Data, Princeton University Working Paper No. 563, <http://dataspace.princeton.edu/jspui/bitstream/88435/dsp01gf06g265z/1/563.pdf>, (accessed November 2, 2011). Eide, Eric, Dominic J. Brewer, and Ronald G. Ehrenberg. Does it Pay to Attend an Elite Private College? Evidence on the Effects of Undergraduate College Quality on Graduate School Attendance. Economics of Education Review 17.4 (1998): Monks, James. The Returns to Individual and College Characteristics Evidence from the National Longitudinal Survey of Youth. Economics of Education Review 19.3 (2000): National Longitudinal Survey of the Youth (1979). 6 Dec <https://www.nlsinfo.org/investigator/pages/search.jsp>.
21 Maloney 21 U.S. News College Compass. U.S. News and World Report. 8 Nov <http://premium.usnews.com/best-colleges>.
22 Maloney 22 Table 1: Top 50 Liberal Arts Colleges Summary Statistics Variable Mean Std. Dev. Min Max % Private % Religiously Affiliated % Urban % Suburban % Most Selective Undergraduate Enrollment Total Cost Student-Faculty Ratio (# to 1) Classes <20 Students Classes >=50 Students yr Graduation Rate Average % of need met Average Financial Aid Package Fall 2010 Acceptance Rate Average SAT Average Freshman Retention Rate Average Alumni Giving Rate Degree-Seeking Women Diversity Index
23 Maloney 23 Table 2: Top 50 National Universities Summary Statistics Variable Mean Std. Dev. Max Min % Private % Religiously Affiliated % Urban % Suburban % Most Selective Undergraduate Enrollment Total Cost Student-Faculty Ratio (# to 1) Classes <20 Students Classes >=50 Students yr Graduation Rate Average % of need met Average Financial Aid Package Fall 2010 Acceptance Rate Average SAT Average Freshman Retention Rate Average Alumni Giving Rate Degree-Seeking Women Diversity Index
24 Maloney 24 Table 3: Top 10 Liberal Arts Colleges Summary Statistics Variable Mean Std. Dev. Min Max % Private % Religiously Affiliated % Urban % Suburban % Most Selective Undergraduate Enrollment Total Cost Student-Faculty Ratio (# to 1) Classes <20 Students Classes >=50 Students yr Graduation Rate Average % of need met Average Financial Aid Package Fall 2010 Acceptance Rate Average SAT Average Freshman Retention Rate Average Alumni Giving Rate Degree-Seeking Women Diversity Index
25 Maloney 25 Table 4: Top 10 National Universities Summary Statistics Variable Mean Std. Dev. Min Max % Private % Religiously Affiliated % Urban % Suburban % Most Selective Undergraduate Enrollment Total Cost Student-Faculty Ratio (# to 1) Classes <20 Students Classes >=50 Students yr Graduation Rate Average % of need met Average Financial Aid Package Fall 2010 Acceptance Rate Average SAT Average Freshman Retention Rate Average Alumni Giving Rate Degree-Seeking Women Diversity Index
26 Maloney 26 Table 5: College Characteristic Summary Statistics for those Schools in OLS Regressions Variable Mean Std. Dev. Min Max Undergraduate Enrollment Tuition Classes <20 Students yr Graduation Rate Fall 2010 Acceptance Rate SAT/ACT National Percentile Average Freshman Retention Rate
27 Maloney 27 Table 6: NLSY79 Individual Characteristic Data Summary Statistics (R=12,868) Variable Mean Std. Dev. Min Max % Male % who spoke Foreign Language at Home % who lived Urban Area Highest Grade Level R's Mother Completed Highest Grade Level R's Father Completed Wages Wages Wages Wages Wages
28 Maloney 28 Table 7: Wage Regressions in 1985 on Individual and College Characteristics Model Independent Variables Foreign Language Urban Mother s Education Father s Education Race Male Tenure Tenure squared Top 50 liberal arts Top 50 National Tuition Total enrollment Fall Acceptance Freshmen retention 6 yr. graduation rate % Classes <20 students SAT/ACT Percentile (-0.59) ** (-2.50) (-0.41) (-1.33) ** (4.36) ** (4.17) ** (15.14) ** (-9.97) (-0.05) (0.45) 7.71E-06 (1.59) -1.78E-06 (-0.71) (0.69) (1.21) * (-1.86) (-0.36) (1.28) (-0.56) ** (-2.49) (-0.40) (-1.36) ** (4.35) ** (4.17) (15.16) ** (-9.99) 7.95E-06 (1.66) -1.50E-06 (-0.62) (0.55) (1.16) * (-1.80) (-0.30) (1.29) (0.53) ** (-2.11) (-0.13) (-1.05) ** (5.67) ** (4.42) ** (15.96) ** (-10.28) (0.37) (0.83) (-0.18) ** (-2.31) (-0.38) (-1.02) ** (5.09) ** (4.31) ** (15.37) ** (-9.99) (0.47) (1.01) (-0.41) (0.16) ** (-2.16) (-0.28) (-1.08) ** (5.07) ** (4.65) ** (15.49) ** (-10.00) (0.33) (0.66) (0.43) (0.02) ** (-2.62) (-0.43) (-0.96) ** (5.29) ** (4.16) ** (15.46) ** (-10.05) (0.38) (1.01) -1.00E-06 (-0.48) (-0.33) ** p<0.05 * p<0.10
29 Maloney 29 Table 8: Wage Regressions in 1990 on Individual and College Characteristics Model Independent Variables Foreign Language Urban Mother s Education Father s Education Race Male Tenure Tenure squared Top 50 liberal arts Top 50 National Tuition Total enrollment Fall Acceptance Freshmen retention 6 yr. graduation rate % Classes <20 students SAT/ACT Percentile (1.06) (0.31) (0.74) (0.76) ** (2.70) ** (11.22) ** (12.53) -5.41E-06** (-8.25) (0.04) ** (2.83) 3.65E-06* (1.91) (-1.23) (0.58) (0.46) (-1.28) (-1.33) (1.03) (0.28) (0.70) (0.79) ** (2.74) ** (11.22) ** (12.55) -5.41E-06** (-8.26) ** (2.79) 3.30E-06* (1.79) (-1.03) (0.67) (0.32) (-1.39) (-1.35) ** (2.11) (-0.11) (0.31) ** (2.00) ** (4.10) ** (11.67) ** (13.48) -5.22E-06** (-8.83) (1.05) ** (2.63) * (1.73) (0.21) (0.53) (1.65) ** (3.55) ** (11.69) ** (12.97) -5.07E-07** (-8.53) (1.23) ** (2.80) (-1.62) * (1.86) (0.05) (0.75) (1.39) ** (2.95) ** (11.87) ** (13.57) -5.36E-06** (-8.98) (0.72) (1.56) * (1.74) (1.61) (0.05) (0.71) (1.33) ** (3.48) ** (11.49) ** (12.87) -5.04E-06** (-8.43) (1.33) * (1.89) 3.81E-06** (2.37) (-0.19) ** p<0.05 * p<0.10
30 Maloney 30 Table 9: Wage Regressions in 1996 on Individual and College Characteristics Model Independent Variables Foreign Language Urban Mother s Education Father s Education Race Male Tenure Tenure squared Top 50 liberal arts Top 50 National Tuition Total enrollment Fall Acceptance Freshmen retention 6 yr. graduation rate % Classes <20 students SAT/ACT Percentile ** (2.48) (0.65) ** (2.33) (0.19) (0.69) ** (11.47) ** (7.13) -1.56E-06** (-4.15) (-1.25) (0.09) 8.64E-06* (1.84) 2.19E-06 (0.95) (-0.47) (0.41) (0.30) (-0.63) (0.41) ** (2.48) (-0.34) ** (2.45) (0.13) ** (2.12) ** (13.57) ** (5.28) -7.75E-07 (-3.13) 7.93E-06* (1.71) 2.70E-06 (1.21) (-0.50) (0.32) (0.37) (-0.59) (0.40) ** (3.36) (1.05) ** (2.42) (1.19) ** (2.02) ** (11.99) ** (8.07) -1.66E-06** (-4.82) (-0.18) ** (3.43) ** (3.03) (0.98) ** (2.55) (1.27) (1.41) ** (12.05) ** (7.83) -1.68E-06** (-4.70) (-0.07) ** (3.41) (-0.92) ** (2.77) (0.81) ** (2.42) (0.75) (0.40) ** (11.87) ** (7.69) -1.60E-06** (-4.58) (-0.81) (1.53) ** (3.30) ** (3.20) (0.72) ** (2.58) (0.92) (1.49) ** (11.84) ** (7.77) -1.67E-06** (-4.66) (0.02) ** (2.40) 4.65E-06** (2.37) (0.47) ** p<0.05 * p<0.10
31 Maloney 31 Table 10: Wage Regressions in 2002 on Individual and College Characteristics Model Independent Variables Foreign Language Urban Mother s Education Father s Education Race Male Tenure Tenure squared Top 50 liberal arts Top 50 National Tuition Total enrollment Fall Acceptance Freshmen retention 6 yr. graduation rate % Classes <20 students SAT/ACT Percentile (1.37) (-0.25) ** (2.98) (-1.16) (0.90) ** (13.96) ** (5.47) -8.17E-07** (-3.31) (-0.04) (-1.66) ** (4.03) 2.40E-06 (0.97) (-0.92) (0.57) (0.01) (-1.26) (-0.70) (1.31) (-0.38) ** (2.89) (-1.06) (0.99) ** (13.92) ** (5.50) -8.26E-07** (-3.35) ** (3.92) 1.59E-06 (0.66) (-0.27) (0.72) (-0.28) (-1.48) (-0.78) ** (2.87) (0.08) * (1.83) (0.79) (1.60) ** (13.41) ** (5.71) -8.88E-07** (-3.49) (1.54) ** (1.97) ** (2.22) (-0.14) ** (2.23) (0.35) * (1.82) ** (13.63) ** (5.34) -8.04E-07** (-3.24) (1.67) ** (2.13) (-1.58) ** (2.07) (-0.13) ** (2.32) (-0.38) (0.36) ** (13.30) ** (5.98) -9.39E-07** (-3.73) (0.94) (0.03) ** (3.68) ** (2.48) (-0.34) ** (2.45) (0.13) ** (2.12) ** (13.57) ** (5.28) -7.75E-07** (-3.13) * (1.71) (1.14) 5.19E-06** (2.32) (0.28) ** p<0.05 * p<0.10