1 How do Labor Market Conditions Affect the Demand for Law School? January 2004 Jeffrey Greenbaum Massachusetts Institute of Technology Department of Economics 50 Memorial Drive Cambridge, MA
2 Abstract I argue that while most people pursue a graduate degree in order to advance in the labor market, poor labor market conditions can suddenly influence a greater interest in law school. In a tighter job market with fewer career opportunities, adults may evade the labor market by turning toward law school. Such an investment can hopefully provide law school graduates with better salaries and greater job stability. I examine the demand for law school by analyzing both the number of applications that each ABA-accredited law school received and its enrollment from 1987 through Although I find that the average law school receives 167 more applications as its state s job opportunities worsen by 1 percent, I do not find sufficient evidence that labor market conditions affect law school enrollment. An analysis of law school acceptance rates supports the hypothesis that enrollment does not noticeably change because schools successfully anticipate the percentage of students it admits and the state of the economy does not justify changes in enrollment.
3 Introduction Advancement in the labor market often necessitates the pursuit of higher education. As a result of the large opportunity costs associated with higher education, however, individuals are more likely to prefer earning an income than attending graduate school unless they believe that furthering their education can later help them obtain a specific job or salary that they want. Thus, an increase in applications for a given graduate school program implies either an increase in the demand for the careers for which the program prepares its students or that the entry into a career requires additional preparation than in the past. In this paper, I examine whether poor labor market conditions can also motivate an interest in applying to and attending law school. Unlike most other post-baccalaureate programs, the law school admissions process does not require any specific background, such as particular undergraduate majors, college courses, recommendation types, or work experiences. In fact, the Law School Admissions Test (LSAT), unlike any other United States college entrance exam, does not require any prior knowledge since it does not test math, science, vocabulary, or knowledge of any particular academic discipline. As a result of its general application process, a baccalaureate recipient can leave the labor market at any point in his or her adulthood by taking the LSAT, applying to law school, and potentially gaining admissions. In addition to the ease of applying to law school, a high school degree is no longer enough to remain competitive in the economy nor is a bachelor s degree in most cases. Consequently, more people may be applying to law school over time because of the increased popularity of attending some form of graduate school. Furthermore, adults who
4 may have been otherwise attracted to the lucrative pay offered by a career in technology or business may now be discouraged by the slump that many dot-com companies and the United States stock market have experienced. Law, however, often offers a more secure career path, whether one pursues a job in the private sector or in the government. Trends in law school admissions have garnered much attention because of the sudden increase in the number of law school applicants. Specifically, the number of applicants has increased by approximately 30 percent from Autumn 2001 through Autumn 2003, whereas annually, from 1995 to 2000, it remained relatively constant or decreased slightly (Figure 1). This sudden increase in the number of law school applicants coincides with the downturn in the economy and the noticeable increases in the national unemployment rate (Figure 2). Because most other graduate school programs have also seen increases in their applicant pool since Autumn 2001, poor labor market conditions may have been responsible for causing adults to consider applying to graduate schools, such as law school. As a result of this hypothesis, I examine whether this cyclical behavior in the amount of law school applicants has been occurring through multiple phases of the United States business cycle. The existence of a significant and consistent relationship between the demand for higher education and labor market conditions can help policy makers determine how to best allocate funds that support graduate students. In order to examine the extent to which labor market conditions affect the demand for law school, I analyze both whether more people apply to and enroll in law school, as well as how law schools respond to changes in labor market conditions. After analyzing the related literature and the various models that I examine, I discuss how I obtain my
5 data and its limitations. I subsequently detail the methods that I use to analyze the impact of labor market conditions on law school admissions and present my empirical results. In this paper, I find a strongly negative and statistically significant relationship between growth in the employment rate and demand for law school. I find this evidence when I examine the number of applications the average law school receives but not for its enrollment. I likewise find strong evidence that admissions rates vary directly with changes in the employment rate. Finally, I conclude the paper by presenting ideas for further research. Background Several studies have explored whether labor market conditions affect the decision to pursue or discontinue one s education but have generally limited their focus to secondary education. Barbara Duncan  examines teenagers during the first half of the twentieth century and finds that they make the decision to drop out of school based on the job opportunities available to youth and that the dropout rate varies inversely with the unemployment rate. Thus, during an economic boom, teenagers substitute education for the opportunity to leave school early in order to earn an income. Duncan also finds that like the business cycle that models the upward economic growth over time, the fluctuations in the high school dropout rate have been moving along a long-term downward trend. After Duncan s initial study of whether labor market conditions cause teenagers to drop out of school, the results from several analogous cross-sectional studies have been mixed. For example, Rumberger  confirms Duncan s findings that better labor market conditions cause students to leave school early in order to earn money.
6 Rumberger specifically examines the probability that a Black or Hispanic male would drop out of school based on the unemployment rate and also finds a negative relationship. Ehrenberg and Brewer , on the other hand, find evidence that better labor market conditions cause White students from low income families to continue staying in school. Although these results contradict the findings of Rumberger and Duncan for Whites of low income families, they do not find any evidence of this causal relationship in either direction for other White students or for Black or Hispanic students. In contrast to all three studies, Zwerling and Thomason  do not find any evidence that labor market conditions affect the decision of any high school student to drop out of school early. As a result of these mixed findings, Rees and Mocan  use panel data estimation methods to control for the omitted variables bias that could have been causing these different studies to uncover contradictory findings. By making use of data from the academic year through for New York schools, Rees and Mocan find a statistically significant and negative relationship between the unemployment rate and the proportion of a district s high school students who drop out in a given year. However, when they do not control for district fixed effects, the estimated coefficient of the unemployment variable becomes positive. Rees and Mocan thus demonstrate that these prior studies suffered from omitted variable bias by not using a panel data set and controlling for unobservable district effects. I likewise employ panel data estimation methods on the statewide level rather than the national one to examine law schools. Unlike most of the other studies that focused on the elementary or secondary school level, Freeman  examines how changes in the labor market affect the demand for higher education in the United States. Although the economic opportunities
7 in many fields that required a college education, such as academia and teaching, worsened during the 1970s, some business-oriented fields, such as management and accounting, either remained stable or actually improved. In response to the changing labor market, Freeman finds evidence that fields whose labor market worsened experienced decreased graduate school enrollments, whereas professional programs benefited during this time period. Although I examine labor market conditions and not labor markets, the stability of a legal career may continue to attract those who fear a more risky sector, such as technology, during economic downturns. Thus, a combination of all of these reasons may account for sudden changes in the popularity of a graduate school program. Empirical Framework In order to analyze the demand for law school, I examine three phases of the admissions process on the school level. The number of people who enroll in law school each year is also an individual level measurement that captures the demand for law school since an individual who wants to attend law school seeks to enroll in exactly one. However, it does not completely measure the demand for law school since not everyone who wants to enroll in law school receives admissions, and schools can attempt to adjust the size of their incoming class through the acceptance rate. Although more people may want to attend law school during economic downturns, law schools may not have the resources or the interest in accommodating more applicants, and so law schools may strive to not allow labor market conditions to affect their enrollments. Correspondingly, I also examine the number of applications that each law school receives to estimate the demand for law school. This measurement captures changes in
8 the number of law school applicants and in the number applications that individuals are submitting, which are two potential responses to changing labor market conditions that can occur with different magnitudes. I likewise examine the law school s response to changes in the size of its applicant pool by analyzing how law school acceptance rates vary with changes in labor market conditions. Law schools can attempt to control the size of their incoming class by predicting the yield, which is the percentage of admitted students that will matriculate, and they may believe that labor market conditions can affect whether admitted students will be more likely to matriculate or to pursue other opportunities. The independent variable in each model is the annual growth of the employment rate in the state in which the law school is located. The ambitious nature of law school applicants make them more likely to be experiencing slight economic hardship than unemployment when they are applying to law school. Consequently, I use the employment rate in my analysis rather than the unemployment rate since it better reflects the law school applicant pool. Furthermore, I make use of the growth in the employment rate because it isolates the labor market s initial level, which is based on unobservable characteristics. Because applicants apply during the fall and winter of an academic year, x t x t+1, I look at how the growth in the employment rate from year x t-1 to x t affects whether people will submit applications toward the end of year, x t. As a result of the omitted variables bias that Rees and Mocan uncover when they do not control for district fixed effects, I control for law school fixed effects in each model. Likewise, the addition of a time trend to the model captures whether fluctuations in the dependent variable are occurring along a long-term trend, as Duncan finds.
9 Finally, I add year fixed effects to determine whether these fluctuations in the law school admissions process are a result of year specific events. These year specific effects can include policies, such as the creation of on-line applications. Data I obtain data about law schools from the Peterson's Graduate Programs in Business, Education, Health, Information Studies, Law, and Social Work, and the National Science Foundation s CASPAR database to determine whether and how labor market conditions affect the average law school s admissions process. Specifically, the Peterson s guides provide information on the accreditation of a law school, its location, the number of applications that it receives each year, and its acceptance rate. Law school enrollment data and the number of bachelor recipients, both aggregated on the state level, come from CASPAR. Data from the Peterson s guides are available from the academic year through , excluding and , and CASPAR provides data from the academic year through , except for the year, I obtain data on employment and the labor force for the years by state from the United States Department of Labor s Bureau of Labor Statistics. I examine only law schools that the American Bar Association has accredited, which is a status that steadily more schools have been obtaining over time. Most recently, there are 188 such law schools in the United States, including the District of Columbia and Puerto Rico. In addition, the Peterson s guides do not report complete information for each ABA law school in each volume, and these omissions do not systematic. These two factors consequently cause the number of law schools included in each year of the data set to vary considerably and randomly. Furthermore, while labor
10 market data from is available from 1970 to present for the United States on the state level, including D.C., and Puerto Rico, data for some states are unavailable from , thus affecting the states whose law school enrollment I can include in my analysis during that time. In Table 1, I present the annual means and standard deviations of the number of applicants to each law school, law school acceptance rates, and the growth in the statewide employment rate by academic year. The employment data corresponds with the recessions of and as well as the economic boom of the late 1990s. While the average number of applicants rises during the early 1990s and also during the early 2000s, the average acceptance rate of each school decreases during this time. Similarly, I present means and standard deviations across all of the years in the descriptive statistics in Table 2. The law school data is based on what the Peterson s guides report from 1987 through 2002, and the employment level data is based on complete information for the 50 states, DC, and Puerto Rico. The mean of the growth in the employment rate is greater than zero, which indicates that the statewide employment rate generally improved from year to year from 1987 to 2002 throughout the United States. This deviation away from zero can cause the long-term trend to be biased toward fewer people seeking to attend graduate school over time. Methods To examine the role of labor market conditions in the pursuit of a legal education, I apply the methods of Rees and Mocan, who examine how labor market conditions affect the continuation of secondary school. Let e represent the growth in the
11 employment rate from year t 1 to t. I assess the impact of the growth in the employment rate on the number of applications a law school receives by assuming that: ln(applicants ijt ) = β 0 + β 1 e jt + β 2 I i + ε ijt where i is a law school in state j in year t, I j is an indicator for each law school to control for law school fixed effects, and ε ijt is the disturbance term. The logarithm of the number of applicants serves as an approximation for the percent change in the number of applicants from one year to the next. I do not directly calculate this percentage because data is not available for each law school for every year of the sample, which would cause many observations to become lost. I approximate this equation by using an ordinary least squares regression. Furthermore, I add independent variables in separate specifications to account for a time trend, year fixed effects, and lags in the growth in the employment rate. I include two lags of the growth in the employment rate to examine whether the change in the number of applications is deliberate or more instantaneous. Although an applicant may plan well in advance the decision to apply to law school, the ease of the admissions process allows an individual who is experiencing trouble in the labor market to apply within the same year and to easily submit additional last-minute applications. Because each coefficient has a percentage interpretation, an increase in the annual employment rate in a state by 1%, corresponds with an increase in the number of applicants that year that the state s average law school will receive by β 1 %. In addition to the number of applicants, the impact of labor market conditions on law school acceptance rates is also estimated by the equation: acceptance rate ijt = β 0 + β 1 e jt + β 2 I i + ε ijt
12 I include additional specifications in order to study a year trend and year fixed effects. Likewise, an increase in the annual employment rate in a state by 1% causes the average law school in that state to increase its acceptance rate by β 1 %. Finally, because the individual, rather than the law school, is the unit of measurement in law school enrollment, I can control for the change in the number of people eligible to attend law school each year. Accordingly, I include the number of bachelor recipients from the year before to approximate eligible applicants and assume: (enrollment ijt / bachelor recipients j,t-1 ) * 100% = β 0 + β 1 e jt + β 2 I j + ε ij where I j is an indicator for each state. I also include a year trend and year fixed effects separetly in additional specifications. Results We observe in Table 3 that an increase in the statewide employment rate consistently causes law schools to receive fewer applications when controlling for law school fixed effects. Specifically, in the first specification, as the statewide employment rate declines by 1%, the average law school, which receives roughly 1,933 applications per year, will receive an additional 8.66%, or 167 applications. As previously mentioned, the increase in applications by 8.66% captures the additional increase in applicants as well as the additional increase in the number of schools to which each applicant applies. In addition, the inclusion of two lags in the growth in the employment rate shows that while the number of law school applications is strongly and consistently affected by changes in the employment opportunities from the current and past two years, the number of applications is increasingly affected less by each preceding year. Thus, the number of applications each school receives is most strongly a response to the immediate change in
13 the statewide employment rate. This finding indicates that more people are suddenly deciding to submit applications during poorer conditions because of their difficulties with the job market. Another alternative is that applicants who may have planned applying to law school react to the worsened statewide labor market by submitting more applications when they actually apply. They may submit more applications than anticipated since they suddenly perceive a greater sense of competition and fewer alternatives to law school. The addition of a time trend reveals that the number of applications that each law school receives has consistently decreased by roughly 0.5% annually from 1987 to This finding does not necessarily contradict the hypothesis that the number of people applying to law school has generally increased over time. For example, while more people may be applying to law school, rolling admissions policies and the increased popularity of early decision may be causing applicants to submit fewer applications. As a result, the submission of fewer applications per applicant over time may outweigh the increase in the number of applicants. Likewise, the sample is based on a period in which the economy general grew, causing us to expect this coefficient to be biased downward. However, the addition of year fixed effects cause the coefficients and standard errors of the model to noticeably change, which suggests omitted variables bias when these fixed effects are not included. Specifically, the coefficient of the employment variable approaches zero, and there is less statistical evidence that the growth in the employment rate affects the number of applications that law schools receive. These findings support the possibility that specific policies, such as the creation of on-line
14 applications, and other events are significantly determining the number of applications that the average school will receive. In response to improvements in the statewide employment rate, we observe in Table 4 that the average law school consistently increases its acceptance rate. For example, in the first specification, a 1% decrease in the growth of the annual statewide employment rate consistently causes the average law school to decrease its acceptance rate by 2.99%. However, if the average school is also receiving 167 more applicants, then most schools are generally admitting the same number of applicants; in fact, more selective schools with acceptance rates of less than 38% are admitting fewer applicants. These changes in the application rate and the number of accepted students may be a result of a decline in the quality of applicants or an attempt for the school to control the size of its incoming class based on the yield that it predicts. The addition of the time trend reveals that acceptance rates have generally been decreasing each year from 1987 to 2002, signifying that receiving admissions to law school has been more competitive throughout this period. We expect that law school admissions has been more competitive over time because more people are interested in graduate school and are willing to rigorously study for the LSAT as well as seek private counseling for the admissions process. If more people are applying to fewer schools as the applicant data time trend suggests, perhaps as a result of the increased popularity in early decision and rolling admissions, then law schools are responding to these policies by admitting more of their incoming class with people who they anticipate will matriculate.
15 Finally, the addition of year fixed effects decrease the coefficient toward zero, which implies that while labor market conditions may cause law schools to adjust their acceptance rates to an extent, factors specific to each year also cause law schools to significantly change their acceptance rates. For example, an expected applicant yield in one year may cause law schools to have more space or less space the subsequent year. Accordingly, a law school may adjust its acceptance rate to reflect unexpected changes in its previous year s enrollment. Although law schools are receiving more applications as the state s employment opportunities decline, the enrollment specifications in Table 5 indicate that there is not enough evidence to conclude that enrollment has consistently responded to changes in the employment rate. These specifications control for state fixed effects and for changes in the amount of people who are receiving Bachelor s degrees in order to account for the possibility of more people being eligible to enroll in a law school the subsequent year in their state. These findings are not necessarily in contradiction to the hypothesis that more people are interested in attending law school during economic downturns. We know people are demanding a legal education as the economy worsens because law schools are receiving more applications. However, the enrollments are probably not affected because law schools are accurately adjusting the number of people who they accept to reflect the yield that they expect, in light of labor market conditions. Conclusion Adults will choose to invest in their human capital by pursuing an advanced degree when they feel that it will help them advance in the labor market. By examining the demand for a legal degree during the past sixteen years, I find strong evidence that
16 poor labor market conditions provoke an increase in the demand for law school, as approximated by the number of law school applications that the average school receives. Furthermore, these additional applications are most likely a result of immediate labor market conditions rather than that of the past year or two, because either the law school application process does not require much planning or applicants fear rejection during these economic downturns and suddenly submit more applications. However, I cannot determine the extent to which this increase in the number of applications that the average school receives represents an additional amount of applicants or an additional amount of applications that applicants are submitting. Future studies should seek to isolate these two possible responses to determine appropriate policy responses. Likewise, additional studies should incorporate additional potential explanatory variables pertaining to the law school admissions process. For example, the ability to apply on-line may be encouraging people to submit more applications, and so an indicator should be incorporated into the model that would specify whether the year is before or after the creation of on-line applications. In response to changing economic conditions, we also observe that law schools adjust the percentage of students that they accept, and a fewer percentage of students receive admissions at more selective schools during economic downturns. Nevertheless, I do not find enough evidence that changes in the employment rate affect law school enrollment. Thus, law schools may attempt to control their enrollment by anticipating the percentage of students who will enroll in their school through the number of applicants they admit. Successful predications will ensure that they are receiving tuitions from enough students to continue operating but do not have to worry about accommodating
17 students for whom they do not have the resources, such as classrooms, professors, and housing. If a smaller percentage of students are admitted to the more selective schools during economic downturns, then they are anticipating a larger yield. They may justifiably expect a larger yield since applicants have fewer worthwhile alternatives to attending law school, and fierce competition is likely to cause applicants to receive more rejection letters. Although my findings are robust, they may not be applicable to the entire crosssection of law school applicants, such as what Ehrenberg and Brewer find. In addition to various subgroups based on gender and race, it would be interesting to compare applicants who are applying as a college senior with older adults who seek to return to school. Finally, future studies should examine whether these same trends exist in other professional programs as well as other types of graduate programs.
18 Works Cited Duncan, Beverly. Dropouts and the unemployed. Journal of Political Economy 73.2 (1965): Ehrenberg, Ronald G. and Dominic J. Brewer. Do school and teacher characteristics matter? Evidence from High School and Beyond. Economics of Education Review 13 (1994): Freeman, Richard B. Implications of the Changing U.S. Labor Market For Higher Education, NBER Working Paper 697 (1981): Rees, Daniel I. and H. Naci Mocan. Labor Market Conditions and the High School Dropout Rate: Evidence from New York State. Economics of Education Review, 16.2 (1997): Rumberger, Russell W. Dropping out of high school: The influence of race, sex, and family background. American Educational Research Journal, 20.2 (1983): Zwerling, Harris L. and Terry Thomason The Effects of Teacher Unions and Collective Bargaining Laws on Educational Performance. Queen s Papers in Industrial Relations, Queens University, Kingston, Ontario.
19 Figure 1 1 Total Amount of United States Law School Applicants Law School Applicant Autumn Figure 2 2 National Unemployment Rate 6.5 Unemployment Rate Year 1 Source: Law School Admissions Council 2 Source: United States Department of Labor, Bureau of Labor Statistics
20 Table 1: Average Law School Admissions Results and Growth in the Annual Employment Rate from through (standard deviation in parenthesis) Year Applicants Acceptance Rate Growth in Employment Rate (1141) (16.50) (0.5946) (1338) (15.84) (1.1374) (1504) (12.44) (0.8056) (1492) (10.78) (0.7809) (1539) (11.64) (0.7522) (1460) (11.54) (0.6716) (1313) (12.02) (0.4511) (1135) (14.10) (0.4823) (1188) (15.33) (0.4472) (1274) (15.44) (0.4968) (1294) (14.53) (0.5875) (1445) (16.32) (0.4574) (1385) (16.04) (0.5817) (1572) (14.97) (0.5013)
21 Table 2: Descriptive Statistics Variable Observations Mean Standard Deviation Minimum Maximum Applicants Acceptance Rate Growth in the Employment Rate
22 Table 3: Effects of Labor Market Conditions on the Number of Law School Applicants (1) (2) (3) (4) (5) (6) Growth in the -8.66*** *** *** -7.24*** Employment (0.885) (0.898) (1.27) (0.946) (0.959) (1.27) Rate jt Growth in the Employment -5.07*** -5.05*** Rate j,t-1 (1.03) (1.03) (1.18) Growth I n the Employment -4.16*** -4.02*** Rate j,t-2 (1.08) (1.08) (1.15) Year Trend *** *** (0.182) (0.180) Law School Fixed Effects yes yes yes yes yes yes Included Year Fixed no no yes no no yes Effects Included Constant 732*** 1773*** 756*** 912*** 1856*** 931*** (0.787) (362) (3.20) (34.8) (361) (32.2) R N Three asterisks signify statistical significance at the 1% level.
23 Table 4: Effects of Labor Market Conditions on Law School Acceptance Rates (1) (2) (3) Growth in the 2.99*** *** 0.578* Employment (0.269) (0.258) (0.342) Rate jt Year Trend *** (0.1084) Law School Fixed Effects yes yes yes Included Year Fixed no no yes Effects Included Constant 40.00*** -1514*** 29.40*** (0.239) (104) (0.867) R N One asterisk signifies statistical significance at the 10% level.
24 Table 5: Effect of Labor Market Conditions on Law School Enrollment (1) (2) (3) Growth in Employment (4.727) (4.783) (5.384) Rate t Year Trend *** (0.0819) Year Fixed Effects no no yes Included State Fixed yes yes yes Effects Included Constant 175*** (6.03) (1627) (6.85) R N