1 Lawyers Lines of Work: Specialization s Role in the Income Determination Process Erin Leahey, University of Arizona Laura A. Hunter, Office of Institutional Research, Evaluation, and Assessment at the University of Massachusetts Medical Center Income inequality has been increasing in the United States, and intraoccupational processes are partly responsible (Kim and Sakamoto 2008; Mouw and Kalleberg 2010). To date, scholars have focused on suboccupational divisions, such as specialty areas, to understand why some members of an occupation earn more than others. In this article we theorize, operationalize, and assess the economic effect of another way in which members of the same profession can be distinguished: by the extent to which they specialize. Using two large secondary datasets on lawyers in the United States, we find that lawyers who specialize earn more. This effect arises partly through two mechanisms individual productivity and firm size and depends upon specialty area prestige: lawyers in low-prestige areas actually benefit more from specializing. It is no surprise to social scientists that income inequality in the United States has been increasing for over 30 years. Since Morris and Western s (1999) call for investigations of this trend, sociologists have begun to analyze the causes of growing wage inequality. 1 Explanations rest largely on the role of occupations. Occupations are deeply institutionalized functional niches (Weeden and Grusky 2005) that largely through social closure processes structure cultural lifestyles and individual identities and help legitimate wage inequality (Grusky 2005). While between-occupation changes account for two thirds of the rise in income inequality between 1983 and 2008 (Mouw and Kalleberg 2010), the remaining third depends on within-occupation characteristics and trends. Indeed, as Weeden (2002) recognizes, occupational-level processes, while important, are not the only ones at work. Explaining income inequality within a single occupation demands a shift in focus from occupations to more fine-grained ways by which labor is divided. Members of the same occupation work in different kinds of organizations and differ in their educational background, skill sets and productivity levels. Previous research demonstrates the importance of these divisions for income attainment within a profession. For example, organizational setting affects compensation in academia, where there is a large divide between research and nonresearch institutions (Tolbert 1986) and private and public institutions (Pfeffer and Langton 1988), and in law, where large law firms pay higher salaries than small firms (Heinz, Nelson and Laumann 2001) and where organizational We thank Joy Inouye for research assistance, and Paul Dimaggio, Robin Stryker and audiences at UCSD and UBC Sociology (especially Mary Blair-Loy and Erin Cech) for helpful comments. Funding for this research was provided by the University of Arizona Rogers Program in Spring The Author Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please Social Forces 90(4) , June 2012 doi: /sf/sos058 Advance Access publication July 10, 2012
2 1102 Social Forces 90(4) hierarchies exploit returns to investments (Garicano and Hubbard 2007). And even though members of the same profession typically have the same credentials, differences in qualitative aspects of education (e.g., type of school attended) and human capital (e.g., experience and productivity) are critical to wage determination (Heinz, Nelson, Sandefur and Laumann 2005; Leahey 2007; Tuckman, Gapinski and Hagemann 1977). These organizational and educational divisions correlate with not only income but also demographic characteristics like gender and race; this has been demonstrated in academe (Bellas 1993), business (Roth 2003) and the legal profession (Kay and Hagan 1995; Noonan, Corcoran and Courant 2005). Consistent with this shift to intraoccupational divisions is a burgeoning literature on the kind of work that members of a profession do, particularly specialty areas. The varying levels of prestige associated with specialty areas make them particularly consequential for careers. Studies of doctors focus on the consequences of specialty choice, for example, whether a doctor chooses pediatrics or internal medicine (Gjerberg 2001; Weeks and Wallace 2002). Studies of academics focus on individuals propensities for particular genres (i.e., book or article publishing) and subfields, and how they shape career trajectories (Clemens, Powell, McIlwaine and Okamoto 1995; Leahey and Rekowsky 2008). Studies of lawyers the profession we study here highlight their practice areas (e.g., tax or patent law), their clientele (e.g., businesses or individuals) and the tasks they engage in (e.g., memo writing, litigation). In the legal profession, working in high-status practice areas (Kay and Hagan 1995), serving business clients rather than individuals (Heinz et al. 2005) and engaging in pure tasks that require not simply diagnosis and treatment but also inference (Abbott 1988) are prestigious kinds of work that benefit practitioners careers and compensation (Dixon and Seron 1995; Gorman 2005, 2006; Hagan 1990; McBrier 2003; Noonan Corcoran and Courant 2005; Sandefur 2001; Seron 1996). To this traditional focus on the content and characteristics of practice areas we add a new focus on the form (i.e., distribution) of practice areas, that is, the extent to which a given lawyer specializes. We expect that lawyers who specialize more (i.e., work in a limited number of practice areas rather than many) will earn more. We use organizational theories about the pressures and concomitant benefits of specializing (Stinchcombe 1965; Zuckerman 1999; Zuckerman et al. 2003) to guide our analysis of individual career outcomes, as moving from an organizational to an individual level of analysis aids theory refinement and elaboration (Vaughan 1992). We also build upon our previous empirical work on another profession academe in which we found that specializing in one or a few subfields (rather than diversifying broadly) is beneficial for a host of career outcomes including earnings (Leahey 2006; Leahey 2007). Our decision to focus on the extent of specialization in practice areas (i.e., domains of abstract legal expertise) rather than the extent of specialization in more concrete realms such as tasks or clientele is theoretically justified in the next sections. Although we focus on within-occupational processes, our site selection acknowledges the continued importance of occupational-level characteristics. Certain types of occupations, such as high-wage occupations, are more closely implicated in rising
3 Lawyers Lines of Work 1103 inequality and thus need to be scrutinized carefully (Mouw and Kalleberg 2010). In this study we focus on one high-wage occupation: lawyers. Lawyers help create new structures of stratification in the larger economy (Hagan 1990:850), especially now that the legal field has grown so large: in 1970, there was one lawyer per 752 persons in the United States (Heinz, Nelson, Laumann and Michelson 1998) and by 2006, there was one lawyer per 145 persons (Bureau of Labor Statistics 2006). The legal field also exemplifies two trends characteristic of most professions: increasing specialization and growing income inequality. Between 1975 and 1995 legal practice areas became more highly specialized and less connected to other practice areas (Heinz et al. 1998:759). Income inequality among lawyers has also increased (Garicano and Hubbard 2009; Heinz et al. 2001:338; Sander and Williams 1989). If these two trends are related, then we expect to see this manifested in individual careers such that lawyers who specialize more also earn more. The Division of Legal Labor Because the kind of work that professionals do is consequential for their careers (and the structure of inequality more generally), research on occupational inequality has long separated workers into functional categories, and research on the legal profession is no exception (Heinz et al. 1998:770). The kinds of work that lawyers do can be classified along four lines: (1. specialty area, (2. client type, (3. professional purity and (4. professional power. Early investigations by Heinz and colleagues (1982) demonstrated that lawyers work in various specialty areas (also called practice areas) that rely on different legal doctrines, and these practice area(s) are highly consequential for their careers. For example, antitrust law has more prestige and accords greater compensation to practitioners than divorce law. Subsequent analyses by Heinz and colleagues suggested an unexpected but strong effect of client type: lawyers who represent businesses are accorded greater prestige and pay than lawyers who represent individuals. Client type bifurcates the field (and correlates with class and ethnic background) so strongly that it effectively creates what Heinz and colleagues call dual hemispheres. Related to but theoretically distinct from practice area and client type is a third way in which legal work is classified: by the professional purity of tasks. Abbott s (1988) professional purity thesis suggests that lawyers who work closely with core, abstract and legal knowledge (i.e., engage in pure legal work, untainted by, say, administrative duties) will receive greater deference and prestige than those whose work is sullied by laypersons and nonlegal matters. Last, some legal environments and tasks imply greater or lesser degrees of professional power, which is significantly related to earnings (Hagan and Kay 1995; Hagan 1990:836). In sum, we know from previous research that lawyers work in different practice areas, serve different kinds of clients and perform tasks that vary in terms of both purity and power, and that all of these kinds of work are related to valued career outcomes like earnings. If we shift our focus from the kinds of work professionals do (i.e., its content) to the form it takes, new distinctions emerge that could illuminate the earnings determination process. The distinction of theoretical interest to us is the extent of specialization.
4 1104 Social Forces 90(4) Some lawyers devote all of their time to a single practice area and thus can be considered specialists. 2 Other lawyers spend a small amount of time in many different practice areas and can be considered generalists. Between these two extremes lies the majority of lawyers who work in a limited set of practice areas. Thus, the degree to which lawyers specialize in their work varies, and we consider this heretofore neglected variation to be theoretically interesting and integral to career outcomes like earnings. This dimension of specialization (extent) is distinct from the dimension more typically studied (area). Indeed, there is no necessary, i.e., imposed, relationship between the two. 3 To assess the extent of specialization, we must consider the number of areas a lawyer practices in, and the amount of time devoted to each, but the substance of the areas is irrelevant. This is demonstrated in Figure 1. Lawyers who share the same four practice areas could be highly diversified (like Lawyer A, who devotes 25% of his time to each area) or highly specialized (like Lawyer B, who devotes 85% of her time to one area, and only 5% of her time to each of the other three). A comparison of lawyers B and C shows that lawyers who have no practice areas in common could easily specialize to the same degree. For this reason, we carefully distinguish both theoretically and analytically between these two dimensions of specialization: areas of specialization (and their respective levels of prestige) and the extent of specialization. This also allows us to consider their interaction, a novel hypothesis we develop in the next section. While previous research has conceptualized and developed ways to measure the extent of specialization (Leahey 2006; Leahey 2007; Heinz et al. 2005), we are the Figure 1. Demonstrating No Necessary Relationship between Areas and Extent of Specialization Extent of specialization GENERALIST Areas of Specialization: SPECIALIST Lawyer A Lawyer B 25% tax law 85% tax law 25% divorce law 5% divorce law 25% immigration law 5% immigration law 25% family law 5% family law Lawyer C 85% patent law 5% anti-trust law 5% securities law 5% business law
5 Lawyers Lines of Work 1105 first to incorporate it into a model of lawyers earnings. Heinz and colleagues (2005:37) use an entropy measure to create a specialization score for each lawyer in their Chicago datafile, but then aggregate these scores to describe differences across time and across practice areas. They do not include the extent of specialization as an explanatory variable in their analyses of the determinants of individuals earnings (see Table 7.2, p.170). In a series of papers, Garicano and Hubbard (2003, 2007) have also explored the extent of specialization among U.S. lawyers, but their focus is on the relationship between individual specialization and characteristics of firms (such as their size, boundaries, and hierarchical structure), for example, how the ratio of partnered lawyers to associates may decrease coordination costs and thus increase efficiencies and returns to skill. The question of whether and how the extent of specialization affects individual earnings remains open. 4 Our theoretical and analytic focus is specialization across practice areas, to the neglect of specialization across other divisions, such as tasks or clientele. We invoke this generalized scope condition for several reasons. First, a focus on specialization across practice areas is consistent with previous research on specialization in the legal field (Heinz et al. 2005; Garicano and Hubbard 2003, 2007). 5 Second, a focus on specialization across practice areas makes our research generalizable to other professions where specialty area is a dominant and consequential form of classification, such as medicine (Weeks and Wallace 2002) and academe (Leahey 2007; Moody 2004). Indeed, a scholar s career, reputation and earnings are defined more by specialty area than by tasks performed (e.g., writing, reading, analysis, mentoring) or by clients served (e.g., colleagues, collaborators, community, students). Third, of all the ways in which legal work can be divided, practice areas are the most visible to outsiders and the most relevant to the decision to hire a particular lawyer or law firm. Outside the legal community, lawyers are known for the kind of law they practice, more so than for the clients they serve or the tasks they perform. Indeed, those outside the legal profession are not privy to the various tasks lawyers regularly perform, like memo writing, negotiation, deposition, and internal administration, especially before legal services and payments are contracted. Specialty area trumps other kinds of divisions like tasks and clientele because it is more visible to, and more consequential for, prospective clients as they make decisions about whom to hire. And as we show, building clients, contracts, and billable hours are critical to economic rewards in the legal profession. 6 Economic Gains From Specializing by Practice Area Although the extent of legal specialization has not been incorporated into an individual level analysis of earnings to date, 7 its effects can be theorized. We incorporate theories of productivity and organizational processes to develop hypotheses. Many of the theories we draw upon are macro-level theories developed to explain institutional behavior, yet we use them to hypothesize about individual careers. We thus agree with Vaughan (1992) that alternating levels of analysis is fruitful for theory elaboration and refinement.
6 1106 Social Forces 90(4) As Adam Smith (1776) noted more than 200 years ago, one mechanism by which specializing may enhance income is via increased productivity. This may be just as true for individual members of a profession as it is for nation states. Focusing in one or a few practice areas leads to efficiency gains (Garicano and Hubbard 2009) and heightened productivity, which (all else equal) should boost earnings. Our previous research on academics (author references removed) found empirical support for this process. The same process may be operating in the legal profession, especially now. In an earlier time, when law firms were small and less likely to be internally organized by practice area, firms rotated new associates across departments to assess their strengths and to allow preferences regarding practice area to develop (Heinz et al. 2001). Increased competition and cost-consciousness among larger firms, however, limited this practice. Firms now realize that most types of lawyers become more productive when they specialize (Heinz et al. 2001:383), likely due to efficiency gains associated with mastering a finite amount of legal doctrine. Thus, firms encourage new recruits to specialize and to hit the ground running once hired to be as productive as possible. Our argument about the gains from specializing in practice area(s) rests largely on the efficiencies gained from mastering a finite amount of abstract legal doctrine; it is unlikely that specializing in more concrete and codified domains such as clients or tasks would boost productivity. Rather, as Gorman and Kmec (2009) find, it is the less codified, less classifiable and more uncertain kinds of work tasks that are rewarded most handsomely. Specialists may also garner higher earnings because they tend to work in large firms. We know that large firms pay higher salaries than small firms because they are more insulated from market fluctuations, and because they emulate the practices of their (predominantly) corporate clients by paying their top lawyers and managers an inordinate share of the firm s income (Dinovitzer Reichman and Sterling 2009; Heinz et al. 2005:165). But why might specialists be more likely to work in large firms? Previous research suggests that the organizational structure of large firms may be more appealing to specialists, and it might also be best able to capitalize on specialists knowledge and expertise. Large firms are best characterized by a corporate model that contains multiple divisions and employs many permanent associates and part-time lawyers (Heinz et al. 2005: 108). And firms with greater hierarchies, that is, associates per lawyer, can better leverage and exploit the depth that characterizes specialists knowledge (Garicano and Hubbard 2007). While we acknowledge that this mechanism is likely relevant to lawyers who specialize in a certain type of client (i.e., lawyers who work primarily with business clients tend to work in large firms), the argument for greater leverage pertains to the abstract legal knowledge specific to practice areas rather than the more concrete and interpersonal knowledge specific to client type. Given these theoretical mechanisms, we hypothesize that specializing by practice area will boost lawyers earnings, even after controlling for other known determinants of earnings. Previous research has identified a number of factors that contribute to earnings in the legal profession and thus are critical control variables. Going to an elite law
7 Lawyers Lines of Work 1107 school and being in the top 25 percent of one s class, having more experience, making partner and working in a prestigious specialty area all promote earnings (Hagan 1990; Heinz and Laumann 1982; Heinz et al. 2005; Noonan Corcoran and Courant 2005). Serving business clients (Heinz and Laumann 1982; Heinz et al. 2005) and networking with notables in the field (Kim and Laumann 2003) are also beneficial. Lawyers who engage in more purely legal work (i.e., inferential work in Abbott s terms) garner more prestige (Sandefur 2001) and presumably the compensation that accompanies it. Lawyers who wield professional power also see financial gains (Hagan, Huxter and Parker 1988; Hagan 1990; Kay and Hagan 1995). Given the recent influx of women and minorities into the bar, we also control for gender and race, with the full understanding that most of our control variables are themselves gendered and racialized, thereby absorbing main effects that would otherwise be associated with gender and race (Ely and Padavic 2007). As noted above, previous research documents the importance of the kind of work professional do, and in the legal field this is typically manifested by lawyers practice area(s) and their associated prestige levels (Heinz et al. 2005; Kay and Hagan 1995; Kim and Laumann 2003). We devote special attention to this variable, distinguishing it from other controls, because it is conceptually and operationally close to our key explanatory variable: the extent of specialization. We expect the effect of specializing to depend on the prestige of one s practice area(s). While it may be intuitive to expect that specializing in a high-prestige area(s) brings added rewards, the benefits of pursuing both a high degree of specialization and working in high-prestige areas may offset each other when pursued jointly. Thus, we hypothesize a negative interaction term: specializing should allow lawyers to compensate for a structurally disadvantaged position, such as working in low-prestige specialty area. For example, personal injury law may not be the most economically rewarding practice area, but lawyers who pursue it almost exclusively might be able to foster efficiencies and become well known for that kind of work, thereby bringing in business and boosting earnings, to a greater extent than lawyers who practice personal injury law on the side, as part of a broader repertoire of legal services. Thus, specializing may be particularly advantageous for those with lower status (Zuckerman et al. 2003). Conversely, diversifying may be most advantageous for lawyers in already advantaged positions, like those working in high-prestige practice areas. In these areas, legal work likely requires a greater amount of task uncertainty and a broader range of skills (Gorman and Kmec 2009) and typically involves working with clients (be they firms or wealthy individuals) with a broad array of legal needs (Fried 1976). Data and Sample We located two secondary data sources perfectly suited to test our hypothesis: the National Survey of Lawyers Career Satisfaction (NSCLS) and the Chicago Lawyers Survey (CLS). In 1989, Ronald Hirsch was commissioned by the American Bar Association (ABA) to conduct the NSLCS. This mail-out survey was fielded to a
8 1108 Social Forces 90(4) probability sample of U.S. lawyers who were members of the ABA; the response rate of 67 percent yielded a sample of 2,189 lawyers. In 1995, John Heinz and colleagues conducted the CLS by mail, randomly sampling lawyers who were in good standing with the Illinois Attorney Registration and Disciplinary Commission and who also had a Chicago address; the response rate of 82 percent yielded a sample of 788 lawyers. Response rates for both surveys were well above average. The topics covered in these two surveys are remarkably similar, including background characteristics, attitudes and, importantly, work histories and experience. While both of these datasets have been used extensively to investigate inequality in the legal profession (Chiu and Leicht 1999; Heinz et al. 1998; Heinz et al. 2005; Kim and Laumann 2003; Sandefur 2001) and both, fortuitously, provide information on allocation of time to different practice areas (which is critical for measuring the extent of specialization), each offers distinct benefits: CLS is more recent and has a direct measure of percent of clients that are business and NSLCS is national in scope. Each datafile contains a rich set of variables; imperfect overlap allows us to control for important variables in at least one, and sometimes both, datafiles. Finding support for our hypothesis using two distinct data sources collected on different samples, at different time points strengthens the external validity of our argument. We impose some restrictions to obtain a sample of lawyers from each datafile. The analysis samples (n = 534 for NSLCS, n = 781 for CLS) include only lawyers who (1. answered all questions about their practice and (2. were working in private practice (i.e., a law firm or as sole-practitioner). We impose this second restriction, which excludes lawyers working for government and for corporations, because the incomedetermination process differs in these practice settings (Dixon and Seron 1995:395) and because billable hours and position in the firm (partner/associate) are important to control for, but are only relevant to private practice. This restriction does not limit our external validity because the majority of lawyers continue to engage in the private practice of law (Curran 1986). In fact, most other studies of the legal field focus on lawyers in private law firms, often further restricting the sample to firms that serve corporations or firms of a certain size (Garicano and Hubbard 2009; Gorman 2006; Gorman and Kmec 2009). Measures Earnings Earnings (i.e., total income derived from legal practice) are measured categorically in both datasets. In the CLS, 1994 earnings were recorded in 20 categories, ranging from less than $10,000 to $500,000 or more. In the NSLCS, 1989 earnings were recorded in eight categories, ranging from less than $15,000 to $200,000 or more. For consistency, we use the minimum value of each income category for both datasets, thereby constructing a more continuous variable. For the few respondents falling in the lowest income category, we assign the value of $1,000, rather than the theoretical minimum
9 Lawyers Lines of Work 1109 of $0. Although our assessment of skewness and kurtosis suggests that a transformation is unnecessary, we follow other users of these datasets (Heinz et al. 2005; Kim and Laumann 2003) and another study of lawyers earnings (Noonan Corcoran and Courant 2005) by taking the natural log of earnings. Our results are robust to alternative measures (e.g., using the midpoint of each income category and assuming a Pareto distribution to obtain the mid-point of the top, open-ended category; Hagan 1990) and to alternative estimation strategies designed for categorical outcome variables (e.g., tobit and ordinal logistic regression; Dixon and Seron 1995). Extent of Specialization We follow Heinz et al. (2005:37) in measuring the extent to which each lawyer s work is specialized. (Recall that Heinz and colleagues constructed a measure of the extent of specialization, but never used it in an individual-level analysis.) In our previous research on the academic profession (Leahey 2006; Leahey 2007; Leahey et al. 2010; Leahey et al. 2008), we measured the extent of specialization in work output; here, we measure the extent of specialization in work input, or the allocation of time. This decision is largely a function of the kind work members of these professions do and how it is quantified in the data available to us. Respondents to both surveys were asked what percent of time they spent working in a fixed set of practice areas in the past 12 months (see Appendix A). From the categorical responses to this question (e.g., 6%-20%, 21%-49%, etc.), we constructed a continuous variable by taking the midpoint of each category. Then, for each individual lawyer, we calculate entropy, a measure of the diversity of one s effort across practice areas: H =ΣP log 1/ P j i i where P i is the proportion of time allocated to practice area i by respondent j. We adjust this measure to make it more interpretable, and also consider modifications. We reverse code the standardized measure of entropy (specialization = 1 Ĥ j / H max ), so that higher values indicate greater specialization. The variable thus ranges from 0 to 1, where a value of 0 indicates that work time is uniformly distributed across practice areas, and a value of 1 indicates complete specialization (devotion of all time to one area). We also constructed two modified versions of the measure: one that adjusted for instances in which total time allocation summed to more than 100 percent 8 and a relative measure that assessed the extent of specialization of each lawyer compared with fellow respondents. Because these various measures are correlated at 0.94 or above, and results do not depend on the measure used, we do not report results based on the modified measures. Our measure of the extent of specialization is, importantly, not confounded with Heinz and colleagues (2005) key variable: percent of clients that are businesses. To ensure this, we consider areas differentiated solely by client base to be the same area. For example, in the CLS, tax, personal and tax, corporate are considered separate categories, given Heinz and colleagues interest in assessing their dual hemisphere
10 1110 Social Forces 90(4) hypothesis. However, for our purposes, we consider both of these areas to be one practice area: tax law. After collapsing categories in this vein (see Appendix A), we computed entropy values based on 36 areas in the CLS and 16 areas in the NSLCS. Differences in practice areas between the two datasets are not problematic, 9 as we standardize the entropy measure by dividing it by its maximum possible value (Ĥ j / H max ), as suggested by Heinz et al. (2005). These differences may in fact be advantageous if our hypothesis is supported regardless of how the substantive structure of the legal field is perceived. Prestige of Specialty Area(s) To capture the prestige of the practice area(s) that each lawyer works in, we begin with the prestige scores for each practice area reported by Sandefur (2001:386-7). The scores range from 0 to 100 and indicate the percent of respondents in the 1995 CLS who reported that the specialty area enjoys above average or outstanding prestige. For example, 84 percent of respondents reported that securities law enjoys above average if not outstanding prestige, and so the prestige score for this practice area is 84. Only 4 percent of respondents rated divorce law similarly, so its prestige score is 4. Because no measure of practice area prestige exists in the NSLCS data, we solicited help from an Urban Lawyers (Heinz et al. 2005) author to map its 16 practice areas to the 36 areas in the CLS data, and then applied the CLS prestige scores (see Appendix B). When a single CLS area matched a NSCLS area (as was the case for the area civil rights ), we used the CLS prestige value for that area. When more than one CLS area matched a NSLCS area (e.g., both general family practice and divorce in CLS match family law in the NSLCS), we took the average of the two values. Finally, to generate a prestige score for each lawyer, we calculated a weighted sum of time spent in each practice area multiplied by the prestige of that practice area. So if a lawyer worked 30 percent of her time in an area with prestige score of 55, and 70 percent of her time in an area with a prestige score of 22, her overall prestige score would be ( ) = Mechanisms We empirically assess whether two of our proposed mechanisms individual productivity and firm size mediate the relationship between specialization and earnings. Recall that larger, more hierarchically organized firms may better capitalize on specialists skills and may also pay more. Individual productivity is measured as billable hours per month and is available in both datasets. Firm size (the number of lawyers in one s establishment) is measured categorically in both datasets; for parsimony s sake, we convert this to a continuous variable by using the minimum value of each category as the value for analysis purposes, and dividing by 10 for ease of interpretation. While measures of other proposed mechanisms (including lawyers rainmaking capacity and their value to employers 10 ) do not exist in either datafile, we control for egocentric
11 Lawyers Lines of Work 1111 network size (the number of notables whom a lawyer knows at the acquaintance level ) in the Heinz (CLS) analyses. 11 Controls A model of income attainment among lawyers would be incomplete without considering the nature of the clientele (Heinz et al. 2005:163). To control for client type in CLS, we use the percentage of each lawyer s time spent with business clients. Although a similar survey item was not asked in the NSLCS, we were able to construct such a measure by mapping practice area data derived from the CLS onto the NSLCS (see Appendix C) and joining this with information on the time that NSLCS respondents devoted to each practice area. For example, if a NSLCS respondent worked in two areas natural resources (72% business) for 20 percent of her time and labor law (44.5% business) for 80 percent of her time her value for percent business clients would be , or 50 percent. Because having fewer clients may increase a lawyer s degree of specialization, we also control for the number of clients in the past year when possible (it is not available in NSLCS). Moreover, we control for professional purity and professional power, both of which have been shown to impact lawyers status and earnings (Abbott 1988; Hagan et al. 1988; Kay and Hagan 1998). We measure professional purity in the CLS using Sandefur s (2001:395) four distinct measures: self reports of days per month in appellate court, frequency of supervision of other lawyers work, frequency of decision making in the organization and the degree to which the work is intellectually challenging. Only the latter two measures are available in the NSLCS. To measure professional power, the available data allow us to use only a few of the indicators that constitute Kay and Hagan s index of professional power, some of which overlap with our measures of professional purity (thus we do not combine measures into an index). For the CLS, frequency of supervision and involvement in decision making may also measure professional power, along with partnership status (binary variables for partner and associate ) and number of employees, which we created by interacting partner status with firm size. For the NSLCS, we use involvement in decision making and number of employees, as well as two additional variables the degree to which a lawyer has control over the selection of cases and percent of time devoted to internal administration to measure professional power. For each variable, higher values indicate greater professional purity or power. We also control for other variables that previous studies have found relevant to lawyers earnings. In both datasets we control for human capital (i.e., type of law school attended (elite or nonelite) and class standing (Hull and Nelson 2000) and work characteristics (aside from productivity and firm size) relevant to earnings, including seniority (years at current firm) 12. We also control for important demographic characteristics like professional age (years since law degree), gender, race, age, and number of children. 13 In the NSLCS, we control for city size because cost-of-living likely affects income.
12 1112 Social Forces 90(4) Analytic Strategy To assess whether the extent of specialization in practice areas affects lawyers salaries, we estimate a series of path-analytic models using the statistical package AMOS. This package can handle missing data by using full information maximum likelihood estimation, which is more ideal than listwise deletion or imputation (Anderson 1957). Path analysis is particularly ideal for specifying the effects of intervening variables (productivity and firm size). We specify three nested models for each sample, the CLS and NSLCS. We begin with a model that includes our key variable the extent of specialization as well as prestige of practice area and all control variables. Our second model assesses the relevance of two mechanisms: productivity (captured by billable hours) and firm size. Our last model includes the hypothesized interaction between specialty area prestige and the extent of specialization. Results A comparison of the national and the Chicago-based samples suggests that the extent of specialization, earnings and the two mechanisms (billable hours and firm size) are associated. In the CLS, average values on all of these key variables are higher than in the NSLCS, suggesting that specializing may indeed be related to earnings (see Table 1). We also assessed whether the area and extent of specialization are empirically related (as we discussed earlier) by running a series of t-tests comparing the mean extent of specialization score between lawyers with and without a particular modal practice area. Using the Bonferroni adjustment for multiple comparisons, we found that two areas in the CLS and four areas in the NSLCS are significantly related to the extent of specialization. 14 Of course, our main interest is less in the practice areas themselves than in their prestige, and how this interacts with the extent of specialization; we investigate this in a multivariate context, below. As hypothesized, the extent of specialization positively affects earnings. We found this to be the case in both samples the CLS (Table 2) and the NSLCS (Table 3): the coefficients are statistically significant and of roughly the same size (standardized coefficients for the direct effect of specialization are and 0.253, respectively). This finding holds in all models, whether or not mechanisms (models 2b and 3b) or the interaction term (models 2c and 3c) are included (indeed, once mechanisms are included, the total effect of the extent of specialization increases). Because we have a logged dependent variable and an unlogged explanatory variable, the exponentiated coefficient gives the percentage increase in the dependent variable for a one-unit shift in the explanatory variable. Among Chicago lawyers (model 2a), a one-unit increase in specialization, that is, shifting from a complete generalist to a complete specialist, increases earnings by over 80 percent (e 0.61 = 1.84). Among lawyers nationwide (model 3a), such a shift doubles expected earnings (e 0.73 = 2.07). These effects are consistent across genders and racial groups. 15 The extent of specialization s effect on earnings is mediated by firm size, and to a lesser extent, individual productivity (see models 2b and 3b). All direct paths that link
13 Lawyers Lines of Work 1113 Table 1: Descriptive Statistics for Chicago Lawyers Survey and National Survey of Lawyers' Career Satisfaction CLS (1995) NSLCS (1990) Mean Median SD Mean Median SD Dependent variable: earnin gs Annual income from legal practice(in 1000s) Key explanatory variables: Specialization Extent of specialization Area(s) ofspecialization: prestige Mechanisms Productivity: billable hours per month Firm size Client type Percent business clients Professional purity and/or power Days in appellate court Frequently supervises other lawyers' n/a n/a n/a work n/a n/a n/a Frequently makes management decisions Challenging jobthat requires high training n/a n/a n/a n/a n/a n/a Number of employees Has control over case selection Percent of time devoted to internal Partner (1 = yes, 0 = no) Associate, Jr. or Sr. (1 = yes, 0 = no) Work characteristics Years at firm Number of clients n/a n/a n/a Continued
14 1114 Social Forces 90(4) Table 1 continued Human capital Top school (1 = yes, 0 = no) Top 25% of law school class (1 = yes, 0 = no) Network endorsement No. high status lawyers acquainted with n/a n/a n/a Demographic characteristics Professional age (survey year is year of Gender (1 = female, 0 = male) Race (1 = White, 0 = other) Age Number of children City size n/a n/a n/a Sample size Note: SD=standard deviation. ean CLS (1995) NSLCS (1990) Median M SD Mean Median SD
15 Lawyers Lines of Work 1115 Table 2: Determinants of Annual Income, Chicago Lawyers Survey (CLS) Model 2a Model 2b Model 2c Controls + Specialization + Mechanisms + Interaction Coefficient SE Coefficient SE Coefficient SE Specialization Extent of specialization a.61**.25.70** ***.72 Area(s) of specialization: prestige a.49**.21.51** ** 1.68 Interaction of area prestige and extent -4.21* 1.87 Mechanisms Extent of specialization productivity Productivity income.004*** ***.000 Extent of specialization firm size 12.45*** ** 3.66 Firm size income.006* *.003 Controls Client Type Percent business clients a.002* Professional purity and/or power Days in appellate court Frequently supervises other lawyers' work.14***.03.13***.03.12***.03 Frequently makes management decisions.09***.03.09***.03.09***.03 Challenging job that requires high training.06* Number of employees.002*** ** **.000 Has control over case selection n/a n/a n/a Percent of time devoted to internal administration n/a n/a n/a Partner (1 = yes, 0 = no).38***.09.35***.09.34***.09 Associate, Jr. or Sr. (1 = yes, 0 = no).27** Continued
16 1116 Social Forces 90(4) Table 2 continued Work characteristics Years at firm Number of clients.001* *.000 Human capital Top school (1 = yes, 0 = no) Top 25% of law school class (1 = yes, 0 = no).17**.06.18**.06.16**.06 Network endorsement Acquainted with high status lawyers.03***.01.03*** ***.01 Demographic characteristics Professional age (survey year - year of degree).02*.01.03**.01.03**.01 Gender (1 = female, 0 = male) Race (1 = White, 0 = other) Age Number of children.05*.02.06**.02.05*.02 City size n/a n/a n/a Intercept 9.18*** *** ***.72 Sample size Model fit R 2 Model 2a Model 2b Model 2c Controls + Specialization + Mechanisms + Interaction Coefficient SE Coefficient SE Coefficient SE AIC BCC Note: SE=standard error; AIC=Akaike Information Criterion; BCC=Browne-Cudeck Criterion. + p <.10, * p <.05, ** p <.01, *** p <.001 aone-tailed test n/a not available in this dataset
17 Lawyers Lines of Work 1117 Table 3: Determinants of Annual Income, National Survey of Lawyers' Career Satisfaction (NSLCS) Model 3a Model 3b Model 3c Controls + Specialization + Mechanisms + Interaction Coefficient SE Coefficient SE Coefficient SE Specialization Extent of specialization a.73* **.63 Area(s) of specialization: prestige a **.74 Interaction of area prestige and extent -4.39** 1.48 Mechanisms Extent of specialization productivity 6.3*** *** 12.3 Productivity income.006*** ***.001 Extent of specialization firm size 13.69*** *** 1.73 Firm size income.028*** ***.005 Controls Client type Percent business clients a.003* * Professional purity and/or power Days in appellate court n/a n/a n/a Frequently supervises other lawyers' work n/a n/a n/a Frequently makes management decisions Challenging job that requires high training Number of employees.003** Has control over case selection Percent of time devoted to internal Partner (1 = yes, 0 = no).70***.09.61***.09.62***.09 Associate, Jr. or Sr. (1 = yes, 0 = no).46***.12.26*.12.28*.12 Continued
18 1118 Social Forces 90(4) Table 3 continued Model 3a Model 3b Model 3c Controls + Specialization + Mechanisms + Interaction Coefficient SE Coefficient SE Coefficient SE Work characteristics Years at firm.06***.01.05***.01.05***.01 Number of clients n/a n/a n/a Human capital Top school (1 = yes, 0 = no) Top 25% of law school class (1 = yes, 0 = * *.09 Network endorsement Acquainted with high status lawyers n/a n/a n/a Demographic characteristics Professional age (survey year is year of.03**.01.05***.01.02*.008 Gender (1 = female, 0 = male) -.17* Race (1 = White, 0 = other) Age -.03** * *.01 Number of children City size.13***.03.09***.02.09***.02 Intercept 2.23*** ***.41.97*.46 Sample size Model fit R A IC B CC Note: SE=standard error; AIC=Akaike Information Criterion; BCC=Browne-Cudeck Criterion. + p <.10, * p <.05, ** p <.01, *** p <.001 aone-tailed test n/a not available in this dataset
19 Lawyers Lines of Work 1119 Figure 2. Interaction Effect in the CLS Predicted Income Figure 3. Interaction Effect in the NSLCS Predicted Income Extent of Specialization Extent of Specialization High prestige Medium prestige Low prestige High prestige Medium prestige Low prestige specializing to earnings via productivity and firm size (in both datasets) are positive, all are statistically significant in the NSLCS and all but one (the path from specializing to productivity) is significant in the CLS. At least in the national sample, specializing boosts productivity, which in turn boosts earnings. In both samples, specialists are more likely to work in large firms, and this also boosts their earnings. Once these two mechanisms are added to the models, the total (i.e., direct plus indirect) standardized effect of the extent of specialization increases from 0.79 to in CLS and from to in NSLCS, further attesting to their role. This suggests a relatively large degree of empirical support for our proposed mechanisms, which help illuminate how specializing boosts earnings.
20 1120 Social Forces 90(4) While we corroborate previous research that finds that prestige of practice area positively affects earnings (Heinz et al. 2005), we also find more nuance to this relationship. As expected, the extent of specialization and the prestige of specialty area interact in both CLS and NSLCS (see models 2c and 3c), and must be considered jointly. While each has a positive effect on earnings, these two factors offset each other somewhat, as suggested by the negative coefficient corresponding to the interaction term. To aid interpretation, we graph these effects (see figures 2 and 3) using only plausible values and plausible combinations of the values (i.e., those that exist in the data). While salaries are higher in the Chicago sample, both figures depict very similar interaction effects. Lawyers working in both medium-prestige (50 th percentile) and low-prestige (10th percentile) practice areas benefit from specializing, but the latter gain the most from specializing (note the steeper slope for this group). Interestingly, specializing actually depresses earnings for lawyers who work in the most prestigious practice areas (90 th percentile); this negative slope is particularly steep in the national (NSLCS) data. Specialists in low-prestige field(s) earn just as much and sometimes more than specialists in high-prestige fields. 16 Exploratory analyses of other possible interaction effects revealed few insights and did not change the results reported here. 17 Most of the controls operate in ways we expected based on previous research. Like Heinz and colleagues recent analysis (2005:170), we find that women s disadvantage disappears once we account for practice setting (firm size). And some (but not all) measures of professional purity and of professional power reach statistical significance, especially for the Chicago sample, corroborating Abbott s and Kay and Hagan s results. The number of clients, which may be related to rainmaking capacity, also positively affects earnings. However, a few effects run counter to previous research. Surprisingly, the percent of clients that are businesses loses its statistical significance in the CLS (but not the NSLCS) after productivity and firm size were introduced (see model 2b), partly because firm size and percent business clients are highly correlated among Chicago lawyers (0.34). We also find, surprisingly, that going to a top law school has no effect on earnings in either sample and being in the top quartile of one s class has a significant and positive effect only in the CLS. Further investigations reveal a significant and positive interaction term between school prestige and class standing (not shown) in both datasets, which suggests that gains from class standing are only manifested in the most prestigious schools. Discussion Why is income inequality growing, even within professions? Intraoccupational differences are partly responsible (Mouw and Kalleberg 2010). Previous work highlights the kind of work that members of an occupation do: their specialty areas, their clientele and the power or purity associated with their tasks. In this article we focused on another way in which lawyers work is differentiated: by the extent to which they specialize by practice area. Is specializing economically beneficial and thus a contributor to wage