








































Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community for help and clear up your study doubts
Discover the best universities in your country according to Docsity users
Free resources
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
Competition for seats in elite U.S. graduate school programs has intensified dramatically over the past 40 years.
Typology: Lecture notes
1 / 48
This page cannot be seen from the preview
Don't miss anything!
Abstract Competition for seats in elite U.S. graduate school programs has intensiÖed dramatically over the past 40 years. In this paper, we study the market for young attorneys to illuminate the role that elite graduate programs play in human capital development. We Önd that attorneys who graduate from law schools ranked in the top ten nationally earn considerably more than those without such a qualiÖcation, even compared to attorneys who graduate from schools ranked 11-
We thank Robert Nelson, Gabriel Plickert, and JeeYoon Park for helping us enhance the AJD dataset, Dan Black and Je§ Smith for the college quality data, David Chambers for analyzing the Michigan Alumni data for us, William Vijverberg for research assistance, Todd Elder for providing programs and technical assistance, and Richard Brooks, Hanming Fang, Dan Ho, Richard Sander, and participants in seminars and conferences for comments. This paper was previously presented under the title "The Returns to Attending a Prestigious Law School." yStanford University Graduate School of Business and NBER. pauloyer@stanford.edu. zDavid Eccles School of Business and Institute for Public and International A§airs, University of Utah and Kellogg School of Management, Northwestern University. scott.schaefer@utah.edu.
1 Introduction
Competition for seats in elite U.S. graduate school programs has intensiÖed dramatically over the past 40 years. For example, despite facing the long odds of an 11% acceptance rate, more than 7,000 people paid $75 each to apply to the Harvard Law School class of 2011. Aspiring lawyers work very hard to get good grades as undergraduates to improve their applications to elite law schools, and a substantial fraction invest in Law School Admission Test (LSAT) preparation classes and materials that typically cost $1,000 or more and require over 50 hours in classroom time alone.^1 Despite this intense competition, there are at least two reasons to question the belief in a causal link between attending an elite law school and attaining career success. First, there are many highly successful lawyers from less prestigious schools. Sullivan & Cromwell and Skadden Arps, for example, rank as the third and fourth most prestigious Örms (according to Vault.com), and both employ many attorneys who graduated from elite law schools such as Harvard, Yale, and Columbia. However, Sullivan & Cromwell also has at least two associates and at least two partners from each of Brooklyn, Catholic, and Ohio State Law Schools. Skadden Arps has ten associates and three partners from Villanova, eight associates and three partners from the University of Connecticut, and eight associates and nine partners from St. Johns.^2 Thus, it is possible to reach the pinnacle of this Öeld without attending an elite law school. Second, any association in the data between attending an elite law school and attaining career success could simply be due to selection e§ects. There is a large literature suggesting that much of the relationship between undergraduate school selectivity and labor market outcomes is due to better schools attracting more talented students. While unraveling the strands of causality here can be a challenge, the causal e§ect of undergraduate school prestige on wages is not generally considered to be large. In this paper, we study the market for young attorneys to illuminate the role that elite graduate programs play in human capital development. We motivate our analysis and provide a preview of our main Öndings in Figure 1. The Ögure shows a non-parametric kernel density estimate of the annual pay earned in 2002 by attorneys who Örst passed the bar exam in 2000, where attorneys are placed into four categories based on their educational backgrounds. First, we categorize attorneys based on the US News ranking of the law school they attended. For this Ögure, we consider only
(^1) A 1989 study of law school applicants (Wightman (1990)) found that about half took an LSAT preparation class. Current LSAT preparation o§erings from Kaplan, an industry leader, include classes that range in price from $1, to $1,500 and involve 51 to 109 classroom hours. The company also o§ers an intensive summer course with 300+ hours in class at a cost of $8,000 and private training packages ranging from $2,300 to $4,500. A Kaplan online self-study class costs $1,150. (^2) This information is derived from lawyer biographies posted on Örmsí web pages during 2008 and 2009, and is based on the data used in Oyer and Schaefer (2012).
Elite Undergraduate Other Undergraduate Top 10 Law School $122,200 $123, Law School Rank 11-20 $113,100 $91,
Table 1: Lawyer Pay By School Prestige. Average Salary in 2002 for lawyers who Örst passed the bar in 2000. Law School rankings based on US News and World Report in 2003. ìElite Undergraduateî schools include 1996 US News Top 25 National Universities and Top 25 Liberal Arts Colleges.
law degrees from elite institutions are close substitutes; the wage impact of attending an elite law school is negligible for graduates of elite undergraduate institutions. To examine this pattern in greater detail, we use a large, representative dataset of lawyers collected by the American Bar Association. We Örst document that attending an elite law school is, on average, associated with a large wage premium and a much higher probability of holding a ìprestigiousî position (which we deÖne as working at a Örm with 100 or more lawyers in one of the top four geographic law markets). Graduates of a top-10 law school earn an average of 25% more than graduates of schools ranked eleven through twenty and over 50% more than graduates of schools ranked 21 through 100. The elite-law-school premium is similar for lawyers two years after they pass the bar exam and seven years after they pass the bar. We then apply various methods to examine the extent to which the elite law school premium is attributable to selection on ability rather than a causal e§ect of law school on earnings. As with much of the literature on undergraduate school quality, we lack an ideal experiment in our data that would allow us to isolate the causal e§ect. We are, however, able to compare the elite law school premium to the wage premium associated with attending a selective undergraduate institution (estimated using data on lawyers and non-lawyers alike from the National Longitudinal Survey of Youth), and examine how both vary as we apply various methods of correcting for selection on ability. Broadly speaking, the elite law school premium is larger and more robust to the inclusion of demographic variables than the selective undergraduate school premium. We Önd, for example, that selection into elite law schools is associated with a variety of de- mographic and background variables that are also likely to independently a§ect a lawyerís career success. Controlling for these factors in a wage regression does not, however, a§ect our estimates of the elite law school premium. Notably, demographic and background characteristics drive selection into selective undergraduate institutions as well, and inclusion of these characteristics in a standard wage regression greatly reduces estimates of the elite undergraduate school wage premium. The one background variable that does impact the elite-law-school premium is attendance at an elite undergraduate institution. There is essentially no premium for attending an elite law school for graduates of elite undergraduate institutions, while the elite law school premium for graduates with
non-elite bachelorís degrees remains large.This suggests that whatever unobservables drive selection into top colleges also drive selection into top law schools. These Öndings are consistent with two possible explanations, which need not be mutually ex- clusive. First, it is possible that there is a direct causal e§ect on lawyersí earnings of an elite baccalaureate or law degree. Such a causal e§ect could operate through a variety of channels; perhaps attending an elite school (law or undergraduate) gives young lawyers access to high-value networks that lead to good career opportunities. It could also be that elite schools are good at teaching some skills that are particularly valuable for attorneys. Second, it is possible that the Öndings are driven by selection on information that is not observable to us; elite schools may sim- ply be good at identifying applicants with unobservable skills that make them successful attorneys. Overall, we believe the evidence is consistent with there being a substantial causal e§ect of attend- ing a top law school ó especially for lawyers that did not attend a top undergraduate institution ó which suggests that it may be wise for prospective attorneys to make substantial investments to improve the odds of being admitted to an elite law school. We present some calculations to this e§ect, but we caution that the Ögures do rely considerably on the assumptions one is willing to make about unobservables. Our paper is related to several other literatures in labor economics, the economics of education, and studies of the legal profession. We discuss prior studies on the e§ects of undergraduate school quality in detail in Section 2. We are aware of just two other papers that relate individualsílabor market outcomes to graduate school quality. In one of these, Arcidiacono, Cooley, and Hussey (2008) study the e§ects of getting an MBA on wages. In speciÖcations similar to ours, they Önd a large premium (20-25%) for going to a Top 25 MBA program relative to other schools but a very small di§erence between Top 10 and Top 11-25 schools. The premiums that they Önd for Top 25 programs are cut roughly in half when they control for individual Öxed e§ects using pre-MBA salary. We cannot use a similar strategy because we do not have pre-law salary data for our sample and because, unlike MBAs, many lawyers have limited or no work experience before law school. Most of our sample went straight from undergraduate school to law school or waited just one year in between. Another paper in this stream is the highly controversial study of a¢ rmative action ìmismatchî in law schools by Sander (2004). In fact, Sander (2004) uses the same dataset that we use and he runs regressions similar to those we run. He also Önds a substantial premium to attending a top law school, controlling for other factors. However, this is not the focus of his analysis and he makes no inquiry into the causal e§ect of law school quality on career success. He further controls for several variables in his analysis that would be inappropriate to include as explanatory variables in our analysis (such as the geographic area where the lawyer works and the
e§ect of college reputation on income. However, a more reasonable model would suggest that
ci = i + 1 xi + 2 zi + i: (2)
That is, the college the person chooses is likely to be a function of her taste for particular types of schools (), the characteristics that a§ect her productivity (x), and other characteristics that are observed by school admission o¢ cers but not by employers (z). The fact that the college choice is determined endogenously would not cause any problems in interpreting wage regressions using the speciÖcation in equation (1) if were independent of c, controlling for the variables in the vector x. This condition seems unlikely to be satisÖed, though. For example, if person i has a positive work ethic, this is likely to a§ect productivity through and make the personís school admission application more attractive through z. In this case, a wage regression that did not have individual Öxed e§ects would attribute some of the e§ects of to c through an upwardly biased . Table 2 summarizes several papers that, in the context of undergraduate institution prestige, have taken di§erent approaches to solving the selection issue. That is, di§erent researchers have chosen di§erent methods to get an unbiased estimate of 2. Behrman, Rosenzweig, and Taubman (1996), who look only at female twins born in Minnesota between 1936 and 1955, use the common background of twins to separate innate ability from the e§ects of schooling. They Önd that, at least for this group, there is a substantial wage premium associated with attending an undergraduate school that grants PhDs, small private colleges, and higher faculty salaries. The magnitude of their estimates is quite large, as they suggest that if a given person receives her undergraduate degree from Wellesley College or the University of Pennsylvania instead of Mankato State University in Minnesota, she can expect approximately a 20% or 36% wage premium, respectively. Brewer, Eide, and Ehrenberg (1999) use a more representative sample and take a more structural approach by specifying a model for selection of college and subsequent earnings. They identify the causal e§ect of college quality on wages by instrumenting for college choice through the costs of the school attended and through the functional form of the school choice and wage equations. They Önd results generally in line with those in Behrman, Rosenzweig, and Taubman (1996). However, the results in Brewer, Eide, and Ehrenberg (1999) are somewhat problematic because, unlike other research in this area and counter to most researchersíintuition, they Önd that selection correction is not important in measuring the e§ect of college quality. Dale and Krueger (2002) and Black and Smith (2004) Önd much smaller e§ects of college reputation on earnings. Dale and Krueger (2002) identify the e§ects of college reputation by comparing earnings of people that were accepted to similar colleges but made di§erent choices about which one to attend. They Önd essentially no e§ect of college prestige on earnings. Black
Paper Comparison Result Behrman, Rosenzweig, and Taubman (1996)
Female twin pairs. Attending Private and PhD- granting universities leads to 10-25% higher earnings. Brewer, Eide, and Ehren- berg (1999)
Model college selection. Attending elite schools increase earnings up to 40% relative to low- ranked public schools. Dale and Krueger (2002) Uses students admitted to same schools but attending di§erent ones.
Little or no e§ect of school SAT scores, but higher tuition leads to higher earnings. Black and Smith (2004) Propensity score matching. Attending a top quartile school in- creases earnings by up to 15% rela- tive to a bottom quartile school. Hoekstra (2009) RDD between state university campuses.
Those attending campus with + SAT points earn 20% more. Table 2: Previous Findings on the Returns to Attending a Selective College
and Smith (2004) use propensity score matching techniques to control for school selection. They Önd that, in most speciÖcations and most subgroups, selection is important. Their estimated causal wage premiums are generally not large, with a maximum of about 15% for a student that attends a top quartile school relative to if she attended a bottom quartile school. Finally, Hoekstra (2009) uses a regression discontinuity approach by comparing students near the margin for getting into the top state university campus in the state. He Önds that getting into this campus, where the average SAT score is 65-90 points higher than the other campuses, leads to a zero to twenty percent increase in earnings at ages 28-33.^3 The variety in the estimated e§ects of college quality suggests that this e§ect can be quite heterogeneous and/or that it is di¢ cult to specify the proper selection correction to separate the selection and value-added e§ects of school quality measures on earnings. But we generally read the results as suggesting that selection is an important component in the correlation between undergraduate school quality and labor market outcomes and think the estimates of the causal e§ect of college quality are generally small on the margins that most students consider. We suspect that few students that attend Wellesley College seriously consider Mankato State University, for
(^3) Studying Colombian students and workers, Saavedra (2008) also uses a regression discontinuity approach. He Önds the highest returns to college quality that we know of, indicating the returns may be higher outside the United States.
or top 25 Liberal Arts Colleges.^4 Panel A of Table 3 provides details on the Wave 1 sample, as well as for those 201 respondents that went to law schools deÖned as being in the Top 10 by US News and World Report in 2003 and the 270 who went to other Top 20 schools.^5 The sample as a whole and each subgroup splits roughly evenly between men and women and averages about thirty years old. Those attending top schools appear to come from somewhat more privileged backgrounds, as their friends and family paid for a higher fraction of their law school expenses and they are more likely to have mothers that continued their education after high school. Those attending more selective schools are, not surprisingly, more likely to have had undergraduate grade point averages above 3.5 and much more likely to have graduated from an elite undergraduate institution. Panel B provides similar information for Wave 2 respondents. They are older at the time of the survey, of course, and they make more money on average. We do not have information on Wave 2 undergraduate GPA and the second wave survey asked about total debt rather than the source of funds for law school, but the two waves look similar and the di§erences across the law school tiers are consistent for the two waves. Our analysis below will focus on two dependent variables. The Örst of these is the log of the personís annual earnings and the second is an indicator variable that takes the value one if the person works at a private law Örm with more than 100 lawyers and in one of the top four legal markets (New York, Washington DC, Chicago, and Los Angeles). The pay di§erences suggest that those going to Top 10 schools earn more than 40% more than the sample as a whole and 25% more than those going to Top 11-20 schools. Figure 2, which displays kernel density estimates of Wave 1 pay di§erences between Top 10, Top 11-20, and Top 21-100 graduates, provides more detail on pay di§erences across law school tiers. The graph shows the well-known bi-modal nature of young lawyer earnings (see discussion of this on www.abovethelaw.com and www.elsblog.org) and large di§erences in what fraction are near the upper mode by law school tier. Figure 3 shows the same graph for Wave 2. Seven years after passing the bar, the distribution of income is no longer bi-modal but the di§erences across law school quality are similar to those for the earlier wave. Panel A of Table 3 also shows that Top 10 graduates are much more likely to work for a large
(^4) We also created an alternative categorization of undergraduate school quality based on average SAT or ACT scores. This led to similar results to those with the US News variable and including both quality measures did not add additional explanatory power. So we use the US News variable throughout the paper. (^5) Because of a tie for number ten, the Top 10 includes the following 11 schools (in order of rank): Yale, Stanford, Harvard, Columbia, NYU, Chicago, Pennsylvania and Michigan (tied), Virginia, and Cornell and Berkeley (tied). ìTop 11-20î throughout the paper includes the following schools (ranked 12-20): Duke and Northwestern (tied), Georgetown, Texas, UCLA, Vanderbilt, USC, and Minnesota and Washington and Lee (tied).
Panel A: AJD Lawyers in 2002 (Wave 1) All Top 10 Law School 11-20 Law School Female 0.4996 0.5174 0. Age 31.183 29.542 29. (3.336) (2.789) (2.960) % from Fam/Friends 0.183 0.270 0. (0.305) (0.365) (0.305) Mother > HS Educ 0.7614 0.8259 0. Live near Mother 0.3579 0.3184 0. Undergrad GPA>3.5 0.5193 0.8101 0. Top Undergrad 0.2400 0.5771 0. Annual Pay $88.4K $122.9K $98.4K (46.2K) (44.0K) (43.6K) Large Firm/Big Mkt 0.1818 0.4577 0. N 1,531 201 270
Panel B: Lawyers in 2007 (Wave 2) All Top 10 Law School 11-20 Law School Female 0.4708 0.4419 0. Age 35.072 34.577 34. (3.268) (2.606) (2.944) Debt at L.S. Graduation $65.2K $76.8K $66.6K (40.5K) (47.2K) (39.9K) Mother > HS Educ 0.7526 0.7778 0. Live near Mother 0.3663 0.3140 0. Top Undergrad 0.2503 0.5543 0. Annual Pay $110.6K $137.8K $116.1K (55.7K) (61.6K) (58.5K) N 1,646 258 296
Table 3: Summary Statistics
(NLSY). We divided the colleges attended into quintiles such that the top and second group are similar proportions of the NLSY sample as the Top 10 and Top 20 groups are of the AJD sample. We also performed analyses using the Baccaleureate and Beyond (B&B) survey. This survey has both advantages and disadvantages relative to the NLSY for our purposes and all our analysis led to similar conclusions to those we draw using the NLSY. More detail on both the NLSY and B&B can be found in the appendix, which is also where we present details of our NLSY analysis. In the main body of the paper, we brieáy discuss how the NLSY results compare to the AJD results, as well as the implications of those comparisons. Appendix Table 1 displays summary information for the NLSY comparable to the AJD summary in Table 3. As with the lawyers, the NLSY respondents from better schools come from families with more education and they are more likely to live somewhere di§erent from where they grew up. Again, those going to better schools both make more after school and show more skill before school (as measured by SAT scores), so it is not entirely clear whether the school quality/wage correlation is due to selection or a causal e§ect of school quality on earnings. Table 4 shows the importance of considering selection issues through analyses where measures of school quality are the dependent variables. Panels A and B display results for Waves 1 and 2, respectively. The Örst two columns of each panel show the results of probits where the dependent variable equals one if the person went to a Top 10 law school. Column 1 uses the whole Top 100 law schools sample while Column 2 is limited to lawyers from Top 20 schools. The results show that selection may be very important. For example, lawyers with at least one parent that graduated from college have a 3 percentage point higher probability of going to a Top 10 school when looking at the whole sample (Column 1). Having an undergraduate GPA above 3.5 also has a highly signiÖcant (statistically and economically) e§ect on whether the person attends a Top 10 law school. Most dramatically, graduating from an elite undergraduate institution is associated with at least a 27 percentage point higher probability of attending a top law school. The third column shows a regression where the dependent variable is 1 if the person went to a US News Top 100 school, 2 if she went to a Top 20 school, and 3 if she went to a Top 10 school. Having a parent that graduated from college is associated with going to a law school that is 0.1 levels higher on this scale and graduating from an elite undergraduate school is associated with more than half a level higher law school. Holding other factors constant, minorities attend higher ranked schools, which could be the result of a¢ rmative action. Reassuringly, given that all the variables in the table do not change once the person goes to law school, the Panel B results for Wave 2 are very similar. The table highlights the potential importance of selection into a top law school and points out the particularly important role of undergraduate institutions. When we run regressions similar to
those in Table 4 without controlling for undergraduate school, the coe¢ cients and signiÖcance on the other background variables increase substantially indicating (not surprisingly) that the same factors a§ect selection into undergraduate and law schools. Appendix table 2 shows similar analyses for undergraduate schools using the NLSY sample. Going to a better undergraduate school is strongly associated with such factors as motherís ed- ucation, high school rank, and SAT scores. This initial look at who attends top undergraduate and law schools indicates that the potential selection problems are similar in these two distinct environments.
We begin by estimating equation (1) for the lawyers in the Wave 1 AJD sample who attended a top 100 school. This includes 1,425 that Örst passed the bar exam in 2000 and that were under 40 years old at the time. The dependent variable is the log of the lawyerís salary in 2002.^6 It is common to use the log of a personís hourly wage as the dependent variable in wage regressions such as these, but about a Öfth of AJD respondents did not provide hours. Our results are similar, but a bit less precise, if we use hourly wages. Regression results are reported in the Örst four columns of Table 5. Column 1 of Panel A reports results with no control variables, so it provides an indication of the average di§erences in lawyer pay across six levels of US News and World Report school rankings. The omitted category in each regression is schools ranked 11-20, so the other ranks are relative to this group. Column 1 makes it clear that there are very substantial di§erences in pay based on where lawyers went to school. Lawyers in schools ranked 11-20 earn approximately 25% less, on average than those in Top 10 schools. Those in schools ranked 21-100 earn another 23% or so less. Lawyers from Top 10 schools average pay of almost $123K, while those from Top 11-20 schools earn about $98K. We know, therefore, that there is a large wage premium associated with going to a higher ranked law school. Column 2 adds controls for gender, marital status, age (indicators for 25-29, 30-34, etc.), and race (indicators for Black, Hispanic, Native American, Asian, and Other). Some of these control variables are important and they add considerable explanatory power to the regression, as measured by the R^2. Women in the sample earn approximately 12% less than men (though this di§erence disappears when we look at hourly pay.) However, adding these controls does not have any e§ect on the relationship between law school rank and pay.
(^6) The question in the AJD survey is ìWhat is your total annual salary (before taxes) including estimated bonus, if applicable, at your current job?î
Panel A Top 10 0.250 0.255 0.241 0.158 0. (0.046) (0.045) (0.046) (0.049) (0.065) Rank 21-100 -0.227 -0.228 -0.228 -0.173 -0. (0.038) (0.038) (0.039) (0.046) (0.059) Female -0.115 -0.118 -0.117 -0. (0.027) (0.027) (0.028) (0.037) Near Mother -0.034 -0.027 0. (0.027) (0.028) (0.037) Undergrad Top 10% 0. (0.042) R-square 0.110 0.132 0.146 0.287 0.
Panel B Top 10 0.314 0.312 0.304 0.234 0. (0.060) (0.061) (0.062) (0.062) (0.0791) Elite Undergrad 0.210 0.211 0.208 0.224 0. (0.036) (0.036) (0.037) (0.038) (0.0490) Top 10 * -0.209 -0.196 -0.196 -0.159 -0. Elite Undergrad (0.078) (0.078) (0.078) (0.078) (0.100) R-square 0.131 0.152 0.165 0.207 0. Controls Demographic no yes yes yes yes Family Background no no yes yes yes School Funding no no yes yes yes Academic History no no no yes yes AJD Sample wave 1 wave 1 wave 1 wave 1 wave 2 N 1,425 1,425 1,425 1,425 1,
Table 5: Lawyer Pay Regressions. OLS ñ Dependent Variable is Log of annual pay. Sample for columns 1-4 is cross-sectional AJD sample in 2002 of lawyers who Örst passed Bar Exam in 2000. ìTop 10îand ìRank 21-100îare based on 2003 US News and World Report rankings. The excluded category is schools ranked between 11 and 20. ìElite Undergradî indicates the lawyer graduated from an undergraduate school ranked as a Top 50 National University or Liberal Arts College using 1996 US News rankings. The speciÖcations in the two panels are identical except for the inclusion of Elite Undergrad and its interaction with Top 10 law school in Panel B and that Panel B column 4 does not control for all the categories of undergraduate college quality (and their interaction with GPA.) ìNear Motherî is an indicator variable for living within 50 miles of respondentís mother. ìUndergrad Top 10%î is a self-reported indicator variable of whether the person was in the top decile of her undergraduate class. Column 5 is similar to column 4, except the sample is the second wave of the AJD.
The speciÖcation in Column 3 adds several controls for family background and the way the lawyer paid for law school, including whether the lawyer lives near her mother, whether her mother was born in the United States, motherís education, fatherís education, whether any of her parents or grandparents are lawyers, and the fraction of law school expenses paid through savings and by parents. These variables add some explanatory power and some are signiÖcant predictors of lawyer income. However, once again the additional controls have no e§ect on the law school prestige relationship with pay. Finally, Column 4 includes our fullest set of controls where we try to capture ability through measures of prior academic success and the cost of law school. Added control variables now include indicator variables for 24 categories of undergraduate school quality, undergraduate GPA (indicators for 3.75-4, 3.5-3.74, etc.), a full set of interactions between these undergraduate quality and GPA variables, an indicator variable for being in the top 10% of her undergraduate class, undergraduate major (indicators for science, business, social science, humanities, and other/missing), whether the person went to a public law school, and an indicator variable for other graduate degrees. The additional control variables make much more of a di§erence now and, speciÖcally, this is driven by the undergraduate quality measures. As the table shows, lawyers in the top decile of their undergraduate class earn 8% more, on average, than other lawyers. Looking at the college quality indicators carefully shows that, with all the Column 4 controls included, lawyers that went to a Top Undergraduate school (as deÖned above) earn 20% more than those that went to other schools. So this regression indicates that Top 10 law school graduates earn about 16% more than Top 11-20 graduates and that lawyers that went to elite undergraduate schools earn an additional 20% regardless of their law school. Thus, the regression in Column 4 leads to two conclusions. First, there is still a substantial premium for going to a top law school, even including all the controls we can. Second, the fact that undergraduate school matters so much means that selection on unobservables is important for lawyers. If all this selection is captured by the undergraduate school categories, then Column 4 captures the causal e§ect of going to a top law school. But that seems unlikely to be an entirely valid assumption and we will use other methods to explore the role of unobservables. The Öfth column performs an analysis similar to the one Column 4, using the Wave 2 sample of lawyers (that is, those who had passed the bar seven years before the time of the survey. Column 5 does not include controls for undergraduate GPA (which we do not have for Wave 2) or undergrad GPA/undergraduate school quality interactions, but is otherwise the same as Column 4. The results are quite similar to those for Wave 1 in terms of the magnitude of the law school quality coe¢ cients. Note that the female wage discount is higher, as many women have pulled back on their hours. Also
Panel A (1) (2) (3) (4) (5) Top 10 0.250 0.251 0.227 0.166 0. (0.048) (0.048) (0.049) (0.058) (0.090) Female -0.182 -0.192 -0.209 -0. (0.049) (0.050) (0.053) (0.076) Near Mother -0.080 -0.078 -0. (0.052) (0.055) (0.082) Undergrad Top 10% 0. (0.074) R-Square 0.055 0.088 0.131 0.370 0.
Panel B Top 10 0.328 0.326 0.314 0.278 0. (0.066) (0.066) (0.068) (0.071) (0.097) Elite Undergrad 0.257 0.270 0.298 0.308 0. (0.067) (0.066) (0.068) (0.070) (0.108) Top 10 * -0.255 -0.255 -0.278 -0.234 -0. Elite Undergrad (0.098) (0.099) (0.100) (0.100) (0.145) R-Square 0.084 0.120 0.168 0.234 0. AJD Sample wave 1 wave 1 wave 1 wave 1 wave 2 N 471 471 471 471 523
Table 6: Top School Lawyer Pay Regressions. Same analysis as Table 4, except limited to graduates of top 20 law schools. Column 2 includes demographic controls. Column 3 adds family background and school funding controls. Column 4 adds academic history controls.
Figure 4: Income by School Quality and Class Rank. Class Rank and income are both self-reported in AJD survey. Wave 1, 2002.
For comparison purposes, Appendix Table 3 shows the results of similar regressions on the NLSY cross-section in 1990. To make the sample comparable to our lawyer sample, we include only people with at least two years of college. The results show that college quality does not appear to be an important determinant of pay in this sample. The average person in the highest tier of colleges earns an average of 6% more than a person in the second tier. But this e§ect is not statistically signiÖcant. The top tier premium grows as college quality drops but the di§erence only becomes statistically and economically signiÖcant when reaching schools with average SAT scores below 840. When all the controls are included, which makes the NLSY speciÖcation comparable to the full set of controls used in Column 4 of Table 5 for lawyers, the school quality/income relationship is very small and the controls have a noticeably larger e§ect on the undergraduate school quality coe¢ cients than they have on the law school quality coe¢ cients. The regressions so far provide at least circumstantial evidence consistent with law school quality having a larger e§ect on lawyer income than undergraduate school quality has on income. Further, they suggest that selection is a larger component of the undergraduate selectivity e§ect than of the law school ranking e§ect. The evidence is consistent with law school quality having a substantial causal e§ect on lawyer income and a bigger e§ect than undergraduate quality has on income. However, we did Önd that undergraduate school quality is related to lawyersíincomes in a way that substantially lowers the implied e§ect of law school quality. Our analysis thus far is limited by the fact that the AJD covers only attorneys with seven years or less of experience and by the lack of information about LSATs.^9 We can partially address these issues by examining data from other sources. We Örst examine whether the relation (whether causal or not) between law school quality and career success continues as lawyers gain experience beyond the level captured by the AJD second wave. One indication that this relationship is long-term is that lawyers from top law schools are highly over-represented in the partnership of top law Örms. The data used in Oyer and Schaefer (2012) include background information for the partners of 285 of the 300 largest law Örms in the U.S. Using this data and data on the number of people that graduated from each U.S. law school, we calculated that, as of the Summer of 2007, 13.4% of graduates of Top 10 law schools between 1970 and 2005 were partners at one of these 285 Örms. 8.9% of graduates of Top 11-20 schools and 3.5% of graduates of other Top 100 schools were partners at these Örms. The 1994-1995 Chicago Lawyers Survey provides another dataset we can use to examine how the school-quality/career-success relation changes as lawyers gain experience. This survey of lawyers
(^9) Note that, while the creators of the AJD have a measure of LSAT success, they have chosen not to make this variable available to insure conÖdentiality and because they do not believe it is reliable for inference in this sample.