Workplace mentoring in the legal profession.

Lentz, Bernard F.

I. Introduction

The purpose of this article is to direct the attention of economists to the phenomenon of workplace mentoring. There are indications that the incidence of mentor-protege relationships in certain occupations and industries is as high as 75 percent or more. This fact notwithstanding, the economics literature generally (and labor economics literature specifically) is virtually silent with respect to mentoring. By way of introducing the subject matter to the profession, we focus attention on mentoring in the legal profession. We are able to provide lower-bound estimates of the incidence of mentoring among lawyers. We then propose and discuss two explanations of the practice. In brief, mentoring (generally) may serve to: (1) enhance the efficiency of job matching, and (2) reduce employee turnover and enhance worker investment in firm-specific human capital by bonding workers and firms more closely together. At this early stage, we are able to report some evidence in support of function (2).

In a survey of top business executives, Roche [22] found that 63.5 percent of the 1,250 respondents had a mentor, defined as "a person who took a personal interest in your career and who guided or sponsored you." Nearly as many of these executives (61.6 percent) had served in a mentoring capacity themselves. America and Anderson [1] found that 53 percent of 1,050 female corporate officers surveyed reported having had a mentor/protege relationship. The evidence suggests that mentor-protege relationships are a common labor market phenomenon, at least in certain occupational classifications.

Roche found that initial attachment to a mentor generally occurs during the first five years of an individual's career. However, since respondents who reported having such a relationship averaged 2 mentors, he noted considerable attachment to mentors during the proteges' 6th-10th years of career. Most respondents reported that their mentors exerted a substantial influence on their career. Those executives who had a mentor averaged 47.3 years in age and $118,900 in salary, whereas executives who did not have a mentor averaged 49.2 years in age and $114,200 in salary. Finally, Roche noted that 50 percent of executives who had a mentor reported being highly satisfied with their career progress, while only 40 percent of executives who had no mentor reported that level of career satisfaction.

Table I. Mentoring in the Legal Profession - 1984

Category Percent with Mentors

Private Practice 38.36
Corporate Counsel 33.82
Federal Agency 40.54
State/Local Agency 31.11

Solo Practice 13.56
2-3 Lawyers 30.94
4-9 Lawyers 39.52
10-20 Lawyers 55.80
21-30 Lawyers 38.46
31-60 Lawyers 51.25
61-90 Lawyers 57.78
90+ Lawyers 48.25

Associates 57.04
Partners 28.60

Arguably, Roche's survey suffers from selectivity problems. Since the focus of his survey is on the mentor-protege relationship, the response rate may be weighted in favor of individuals who participated in such a relationship. However, even if we assume the extreme case that none of his (2,726) non-respondents ever had a mentor-protege relationship, his numbers imply a minimum rate of mentoring among executives of 20 percent.

Our figures, derived from the American Bar Association's National Survey of Career Satisfaction/Dissatisfaction, conducted in 1984, reveal a high incidence of mentoring in the legal profession. The focus on general career satisfaction/dissatisfaction mitigates the selectivity problem inherent to Roche's survey. One specific question (out of many) posed was: "Do you have a mentor in your place of work who furthers your career and gives you advice?" Note that the question refers to the respondent currently being protege to a mentor, as opposed to a more encompassing question asking whether the respondent ever had such a relationship. The former undoubtedly understates the true incidence of mentoring. Still, the figures are illuminating.

Roughly 35 percent of the lawyers who responded to the survey regarded themselves as proteges in this form of relationship. The supply of mentors in a solo practice is not very great; thus the low incidence. Well over half of all associates were proteges to mentors, as were over one quarter of the partners. These numbers suggest that Roche may not suffer much of a selectivity problem. With 35 percent of lawyers reporting currently being in a mentor-protege relationship, it is not difficult to imagine that 50-60 percent or more of all lawyers have participated in such a relationship at some point in their careers.

The handful of empirical investigations of the impact of having a mentor on subsequent career success find that individuals who have had a mentor experience greater job satisfaction and earn more than unmentored individuals. However, all of the previously-reported findings with respect to returns to being mentored have been based on data collected either through surveys that ask about mentorship specifically (e.g., Roche [22]; Ochberg, Tischler, and Schulberg [18]), or retrospective interviews (Levinson, et al. [17]; Collins and Scott [3]; and Zey [26]). The former suffer from potential selectivity problems, the latter suffer from small samples.

Yet numerous aspects of the mentor-protege relationship remain unexplored. Why, for example, would senior members of an organization ever agree to mentor junior members? Presumably, such mentoring absorbs time and other resources commanded by the mentor which have an opportunity cost. Why do mentored individuals experience greater career success than unmentored individuals? Is there something inherent to the relationship that makes the protege more productive or does mentoring merely identify those individuals who will be more successful anyway? Why would organizations encourage or discourage the practice of mentoring? These questions and others have simply never been addressed before, to our knowledge, in the economics literature.

II. Theories of the Mentoring Function

Job Matching

The M-P relationship is compatible with the theory of job matching (Jovanovic [8; 10]; Barron, Black, and Loewenstein [2]). In a world characterized by costly information about worker quality, firms may be willing to invest resources in identifying high quality employees (for our purposes, workers with high productive capacity). Similarly, would-be employees who have high productive capacity may be willing to invest resources in communicating to firms their high quality status. One means of satisfying both parties' job matching wishes is through the M-P relationship. The employee signals his quality through his willingness to bear the cost of "training" under a mentor; the firm invests the (unrecoverable) time of senior personnel in identifying high quality junior employees. Since matching mistakes are costly to the firm, the profit-maximizing firm will differentially reward efficient matchers. Any incentive to shirk on the part of mentors is mitigated by the reward structure (which we take to include possible adverse career consequences of systematically misjudging the abilities of their proteges).(1) A straightforward implication of the job-matching explanation is that mentored individuals will experience greater earnings and career success than non-mentored individuals. The facts, as reported by Roche for top corporate managers, are consistent with job-matching. The interpretation, however, is quite different. In the management literature mentoring creates more productive workers; in a job matching context more productive workers are identified by the mentoring process.

Mentoring as job matching implies: (1) that individuals with relatively high productive capacity will seek out mentors and vice-versa, (2) that the productivity (and hence earnings) of proteges while they are being mentored will be higher than the productivity of unmentored workers, and (3) that earnings (and other observed measures of career success) of mentored individuals will exceed those of unmentored individuals. The higher earnings and greater career success of proteges noted by Roche is consistent with implication (3). However, the job-matching explanation is inconsistent with our observation that 25 percent of partners in law firms are proteges to mentors. We expect that matching would be a fait accompli by the time an individual is promoted to partner in a law firm. Moreover, if job matching was the sole function of the mentor-protege relationship, we would expect all new employees to be mentored. This does not occur.

Firm-Specific Human Capital

Mentoring also may be an efficient vehicle whereby firms encourage employees to develop firm-specific human capital (SHC). Employers prefer workers to develop SHC, because it bonds them to the firm, which implies less turnover than occurs among workers with general human capital (GHC). However, workers prefer to develop GHC because of its portability across firms. To induce workers to invest in SHC, firms typically pay some portion of the worker's acquisition cost. SHC binds workers to firms because of the VMP loss (reflected in lower wages) implied by switching employers and binds firms to workers because of the costliness of replacing capital. The inverse relationship between workforce turnover and specific human capital acquisition has been documented by Parsons [19], Pencavel [20] and Jovanovic [9].

There is, however, a potentially nontrivial moral hazard problem with respect to the firm compensating the individual for his investment in SHC. If the worker's SHC is not fully transferable to another firm, his opportunity wage is something less than his current VMP. Reneging by the firm, in terms of not paying an individual the full value of his/her VMP or a slower-than-justified promotion schedule, is mitigated by the need to induce a continuous stream of workers to invest in SHC. The firm communicates its commitment to honor its payoff promises by permitting senior personnel to become mentors. The firm absorbs a loss equal to the value of lost productivity (billable hours) of senior personnel during the period of active protegeship. The firm's willingness to incur this expense acts as a Klein-Leffler [11] type of performance bond.

Although discussions of investment in SHC typically focus on possible reneging by firms, there is always a double moral hazard problem. Even if there is no other firm currently in existence willing to bid a positive price for an individual's SHC, that individual always has the option of leaving the firm and starting up a competing enterprise, drawing upon the SHC acquired from his previous employer. This may be doubly damaging in the context of law firms because the employee who quits may simultaneously draw off the portion of his old firm's client base that he serviced.

A General Model

Let [Mu] = the probability individual i, who has no mentor, is retained by the firm in period t. For simplicity, we assume that [Mu] is invariant across t. For an individual who retires in year R, with a discount rate r, the present discounted value of his earnings can be represented as:

[Mu] [integral of] [W.sub.t][e.sup.-rt] dt between the limits of R and 0. (1)

We assume that the marginal cost of acquiring GHC equals zero and that SHC is acquired only by transfer from mentors to proteges.(2) The would-be protege acquires a mentor at some opportunity cost of time that is reflected in his wage rate during the period of protegeship (from t = 0 until t = p, where p is the end of the mentoring period). We assume that proteges earn (1 - [Alpha])[W.sub.t] during the period when the M-P relationship is actively fostered and (1 + k)[W.sub.t] thereafter. In addition, there is a return to SHC in the form of an increased retention probability, ([Delta]). We assume that [Delta] depends upon the length of time an individual receives SHC from a mentor, i.e., [Delta] = f(p), where f[prime] [is greater than] 0 and f[double prime] is unsigned.

The general decision calculus for a new employee deciding whether or not to become protege to a mentor is framed as:

[Mathematical Expression Omitted].

Equation (2) indicates merely that the net present value of being a protege must exceed the individual's opportunity cost. However, the analytic solution to the optimization problem stated in general form in equation (2), given even simple functional forms for the variables, is extremely complex.

The decision calculus from the standpoint of an individual contemplating becoming a mentor is virtually identical to that of the would-be protege. Having a protege may influence the individual's productivity, his probabilities of retention, or both. During the time spent as mentor, (s)he would suffer some loss of earnings, due to loss of productivity, that would have to be more than made up for by increased earnings and/or retention probability in future periods.

Several empirically testable propositions emerge from the human capital development theory of mentoring. First, reported incidence of protegeship should decline with an individual's experience. With increasing experience (s)he is more likely to have already acquired much of the workplace-relevant human capital, thus obviating the need for a mentor. Second, since reduced turnover is what the firm has at stake in encouraging development of SHC, high-turnover individuals (in the statistical sense) are not likely to be approached by mentors. For example, we expect that individuals with a history of changing jobs are less likely to have a mentor than individuals with a more stable employment history. Similarly, since young women are more likely to exit the workforce than young men, we expect the latter to be more likely to report having a mentor than the former. Third, during the period of active protegeship, the protege's (mentor's) wage will be less (greater) than his/her opportunity wage. Fourth, growth in earnings with experience will be greater for employees with mentors than those without mentors. Finally, turnover among mentored individuals should be lower than turnover among nonmentored individuals.

III. Empirical Evidence Regarding Mentoring

Data

The incidence of mentoring among lawyers reported in Table I and the empirical results reported in this section are based on responses to the National Survey of Career Satisfaction/Dissatisfaction (American Bar Association Young Lawyers Division, 1984).(3) A random probability sample of 3,018 lawyers of all ages was drawn from ABA member and nonmember lists coveting 569,706 lawyers. The sample included individuals employed by government agencies and corporations, as well as attorneys engaged in private practice. The survey questions addressed aspects of the respondents' current work environment, job history, educational background and personal characteristics.(4) The total survey response rate was 76.9 percent; there is no evidence of any statistically significant response bias. Hirsch [7, 18] argues that the survey findings are "very close to being representative of the entire legal profession."

There is a fundamental difference between the nature of a partner's ties to a law firm and the nature of an associate's ties to that same firm. Partnership status normally implies (partial) residual claimancy whereas associates are employees with no claims to the firm's profits. Tax minimization incentives may motivate partners to not take out the full value of current-period earnings as current-period wages (see Leibowitz and Tollison [16, 385]), but rather, to defer some fraction of earnings until a later period. This would have the effect of increasing measurement error in a (reported) earnings function. For this mason, in our empirical work we distinguish between respondents who are associate attorneys in private practice and those who are partners in a private practice. We had information on all variables of interest for 1,461 respondents.

Question #16 of the survey asked about the protegeship status of the respondent: "Do you have a mentor in your place of work who furthers, your career and gives you advice?" Note that the respondent is asked whether (s)he currently is a protege as opposed to asking whether (s)he had ever been in such a relationship. This wording influences the strength of the empirical results to be reported, an issue we discuss presently.

Factors That Influenced Whether an Individual Had a Mentor

We began our empirical investigation of the mentor-protege relationship in the legal profession by searching for factors that were associated with protegeship. Our general model of the choice function included the following variables: respondent's educational background (to test for matching), type of organizational setting (since the absence of a profit motive in the public sector and the attenuated profit motive for corporate attorneys, given that in-house legal counsel typically is not structured as a profit center, may mitigate the incentive to engage in efficient matching or human capital development through mentoring), firm size (for supply-side reasons), partnership status (to test for both matching and SHC development), gender and family status (to test for SHC development), experience, including internship experience (to test for both matching and SHC development), city size and minority status.

We emphasize the duality of the choice function; not only does a would-be protege search for an appropriate mentor, the mentor (with the tacit or explicit encouragement of management) chooses the protege. Thus, demand by one party is a necessary but not sufficient condition for a mentor-protege relationship to materialize. A double coincidence of wants determines the incidence and duration of mentor-protege relationships.

The specific model we estimated is identified in Equation (3):

[Protege.sub.i] = [summation of] [[Beta].sub.i][a.sub.i]. (3)

Educational Variables

[a.sub.1] TOPVERY = graduated in the top quartile from a very prestigious school

[a.sub.2] SECONDVERY = graduated in the 2nd quartile from a very prestigious school

[a.sub.3] THIRDVERY = graduated in the 3rd quartile from a very prestigious school

[a.sub.4] FOURVERY = graduated in the 4th quartile from a very prestigious school

[a.sub.5] TOPSOME = graduated in the top quartile from a somewhat prestigious law school

[a.sub.6] SECONDSOME = graduated in the 2nd quartile from a somewhat prestigious law school

[a.sub.7] THIRDSOME = graduated in the 3rd quartile from a somewhat prestigious law school

[a.sub.8] FOURSOME = graduated in the 4th quartile from a somewhat prestigious law school

[a.sub.9] VERYREVIEW = worked on the law review at a very prestigious school

[a.sub.10] SOMEREVIEW = worked on the law review at a somewhat prestigious school (the omitted control category is not very prestigious law school)

[a.sub.11] LSAT = Law School Aptitude Test score

Work Experience

[a.sub.12] EXPERIENCE = number of years since respondent finished law degree

[a.sub.13] EXPERIENCE SQUARED

[a.sub.14] #DIFFJOBS = number of different jobs respondent has had since completing his/her law degree

[a.sub.15] SAMEFIRM = respondent's experience has been with the same firm

[a.sub.16] PRELAW EXPERIENCE = number of years respondent spent working after college but before going on to law school

[a.sub.17] PRIVATE INTERN = respondent interned in a private law firm

[a.sub.18] LEGAL AID = respondent interned in a legal aid clinic

[a.sub.19] GOVT INTERN = respondent interned in a government agency

[a.sub.20] OTHER INTERN = respondent interned in some other setting

Personal Attributes

[a.sub.21] FEMALE = respondent was female

[a.sub.22] MINORITY = respondent was not a white caucasian

[a.sub.23] MARRIED = respondent was married

[a.sub.24] NUMCHILD = number of children respondent indicated as having

[a.sub.25] FEMARR = respondent was a married female

[a.sub.26] FEMCHILD = number of children a female respondent indicated

[a.sub.27] FEMEXP = female x experience

[a.sub.28] FEMEXP2 = female x experience squared

Current Job Setting

[a.sub.29] CORPEXEC = respondent worked in an executive position in a corporation

[a.sub.30] CORPSUPVSR = respondent worked in a supervisory position in a corporation

[a.sub.31] CORPJUNIOR = respondent was a junior attorney in a corporate job

[a.sub.32] CORPOTHER = respondent worked in a non-executive, non-supervisory, non-junior job in a corporation

[a.sub.33] GOVTEXEC = respondent worked in an executive-level position in a government setting

[a.sub.34] GOVTSUPVSR = respondent worked in a supervisory-level position in a government setting

[a.sub.35] GOVTJUNIOR = respondent was a junior attorney in a government job

[a.sub.36] GOVTOTHER = respondent worked in a non-executive, non-supervisory, non-junior government job

[a.sub.37] JUNIORFIRMSIZE2 = respondent was a junior attorney in a firm of 2-3

[a.sub.38] JUNIORFIRMSIZE3 = respondent was a junior attorney in a firm of 4-9

[a.sub.39] JUNIORFIRMSIZE4 = respondent was a junior attorney in a firm of 10-20

[a.sub.40] JUNIORFIRMSIZE5 = respondent was a junior attorney in a firm of 21-30

[a.sub.41] JUNIORFIRMSIZE6 = respondent was a junior attorney in a firm of 31-60

[a.sub.42] JUNIORFIRMSIZE7 = respondent was a junior attorney in a firm of 61-90

[a.sub.43] JUNIORFIRMSIZE8 = respondent was a junior attorney in a firm of 90+

[a.sub.44] PARTNRFIRMSIZE2 = respondent was a partner in a firm of 2-3

[a.sub.45] PARTNRFIRMSIZE3 = respondent was a partner in a firm of 4-9

[a.sub.46] PARTNRFIRMSIZE4 = respondent was a partner in a firm of 10-20

[a.sub.47] PARTNRFIRMSIZE5 = respondent was a partner in a firm of 21-30

[a.sub.48] PARTNRFIRMSIZE6 = respondent was a partner in a firm of 31-60

[a.sub.49] PARTNRFIRMSIZE7 = respondent was a partner in a firm of 61-90

[a.sub.50] PARTNRFIRMSIZE8 = respondent was a partner in a firm of 90+

(solo practitioners are the omitted control category)

[a.sub.51] PETITCITY = respondent lived in a town whose population was 25,000-49,999

[a.sub.52] SMALLCITY = respondent lived in a town whose population was 50,000-249,999

[a.sub.53] MEDIUMCITY = respondent lived in a town whose population was 250,000-1,000,000

[a.sub.54] LARGECITY = respondent lived in a town whose population was over a million

(population less than 25,000 is the omitted control category)

We looked for a number of relationships to hold. For supply-side reasons we expected incidence of protegeship among solo practitioners to be significantly less than the reported incidence among respondents working in any type of multi-attorney setting. In solo practices there is, by definition, no other within-firm attorney to mentor the respondent. In any other multi-lawyer setting (private practice, corporate or government) there are attorneys that one might approach to serve as a mentor. We suspect that transactions cost considerations limit the incidence of mentoring by someone from a different firm.

We expected PROTEGE to be influenced positively, on the demand side, by predictors of productive capacity (such as LSAT scores, whether the individual worked on the law review and his/her academic performance at prestigious law schools) and previous experience (such as relevant internship experience). That is, would-be mentors choose as proteges those individuals who enter the finn with certain credentials from law school that indicated high productive capacity. However, we expected respondents with a history of switching employers to be viewed as unattractive proteges by potential mentors. Thus we expected incidence of protegeship to be related negatively to the respondent's number of previous employers (#DIFFJOBS).

Table II reports logistic regression estimation results for protegeship status, where 1 = protege. Variable means for proteges and nonproteges are detailed in Appendix A.

Our expectation about the (negative) relationship between general experience in the law and incidence of current protegeship was amply borne out by the data. However, the fact that an individual had spent his/her entire career with the same law finn had no significant bearing on [TABULAR DATA OMITTED] the probability of protegeship. This suggests that the important human capital transmitted from mentor to protege consists of general legal skills, as opposed to informational and power structure networks within the firm.

The regression results indicate that the probability of being mentored does not differ significantly between highly credentialed individuals and employees without such credentials. With the exception of serving on the law review at a somewhat prestigious law school (which is negatively associated with having a mentor), none of the educational or internship variables demonstrates a significant impact on the probability of having a mentor. Thus we find no evidence, at this stage, that mentoring serves a matching function.

Jumping ahead to Table III, we note that certain employee characteristics are associated with positive wage differentials, including whether the respondent graduated from a very prestigious law school in the top quartile of the class, whether the respondent completed an internship with a government agency or served on the law review at a somewhat prestigious law school, and LSAT scores. Other employee characteristics are associated with negative wage differentials, including graduating in the fourth quartile of a very prestigious or somewhat prestigious law school class and serving a legal aid internship. The former credentials apparently identify workers who are highly valued by the firm, presumably because their productivity significantly exceeds the productivity of workers without these characteristics. The converse presumably holds for the latter type of credentials.

It is not unreasonable to suggest that, in a matching-based model of mentor/protege relationships, would-be mentors would screen for talented new attorneys on the basis of their performance in law school, with the mentor serving as a more intense judge of good matches. Perhaps would-be mentors do make such offers. However, for the junior attorneys with strong educational credentials the returns from being mentored arguably will be lower than for junior attorneys without these credentials. This assumes, as we show in Table III, that in the large, performance in law school is strongly related to subsequent performance on-the-job. Thus, the best and brightest law school students are high opportunity cost workers; their greater ease of exit from their current employer to another by itself helps guarantee proper treatment by their current employer, in a world without mentors. This suggests that an adverse selection process may be at work, with some fraction of the very best junior employees turning down offers to serve as proteges to senior members of the firm.

By the same token, the less obviously talented junior attorneys probably have the most to gain from being mentored, but potential mentors may be reluctant to service these individuals. We are unable, at this stage, to separate the demand-side effects from the supply-side effects empirically. [TABULAR DATA OMITTED] Our limitations in this respect mitigate our finding empirically a significant relationship between Protege status and these matching variables.

None of the variables that focus on the race, gender or marital status of respondents demonstrate a significant impact on the probability of having a mentor. In separate, but unreported, regressions, we found no evidence of discrimination in Protege status against lawyers over age forty or those who attended law school at night with respect to incidence of mentoring. These results are available upon request.

The results presented in Table II also reveal that, as expected, mentor-protege relationships are much more prevalent among lawyers working in a multi-lawyer setting than among solo practitioners. Moreover, the pattern of coefficient estimates suggests that more mentoring occurs in private practice law rinns than among corporate or government attorneys and that incidence of protegeship is surprisingly high among partners in private practice. This is surprising because the matching and human capital development functions are not likely to be relevant for partners. We investigate this phenomenon more closely in a separate paper (Laband and Lentz [14]).

Earnings

We turn our attention next to the impact of having a mentor on earnings. As reported previously, Roche [22] found that mentored corporate officers earn, on average, more than their unmentored colleagues. Strober [24] reported no differences in earnings of mentored versus unmentored individuals, based on a survey of 251 MBA graduates from Stanford University in 1974. Roche's work suffers, at a minimum, from selectivity problems. Although the women respondents in Strober's sample worked significantly fewer hours than men and reported a higher rate of protegeship, she only examined possible differences in annual earnings. We suspect that this cuts against finding significant differences that might exist with respect to hourly earnings. Neither author controlled in a rigorous fashion for experience and other variables that might influence reported earnings. Although we think that a good theoretical case exists for expecting mentored individuals to earn more than unmentored individuals (with ceteris paribus conditions in effect), no previously-reported empirical evidence that we are aware of supports the theoretical argument rigorously.

In the ABA Survey, earnings in 1983 were reported as a series of (8) category variables, ranging from less than $15,000 to $200,000 or more. We constructed a continuous earnings variable by taking the midpoint of each category for categories 1-7, and assigning $200,000 as annual income for respondents who reported earning at least that much.

Respondents also reported (as a continuous variable) the number of hours worked per month that were "considered by you or your employer to be part of your employment/job regardless whether defined as "billable" or not. We multiplied the reported number of hours worked per month by 12 to obtain hours worked per year. We adjusted this figure by subtracting out vacation hours, calculated by multiplying the reported number of weeks of vacation time actually taken by the number of hours worked per week (pro-rated from hours worked per month). This should be a fairly close estimate of actual hours worked per year.

Simon and Warner [23] provide the intellectual foundation upon which we based our expectations regarding the relationship between protegeship and wages. To the extent that mentors help to match proteges with jobs, through what Simon and Warner term a "referral" process, the per se impact of having a mentor should be to increase earnings of those individuals relative to the earnings of unmentored individuals. If, on the other hand, mentors transmit general and specific human capital to their proteges, some earnings loss for proteges relative to nonproteges is implied. This earnings loss derives from the lost productivity experienced by proteges during the time they spend engaged in the human capital transmission process. These two theoretical constructs are not mutually exclusive. A young attorney might signal his/her suitability for a job by agreeing to be trained by a mentor. Thus an empirical finding of no impact of having a mentor on earnings might reflect the co-mingling of a positive screening effect and a negative opportunity cost impact.

With respect to the impact of experience on earnings our expectations were less ambiguous. If mentoring performs a job-matching function only, we should observe lower returns to experience among proteges as compared to nonproteges (Simon and Warner [23]; Corcoran, Datcher and Duncan [4]). That is, matching implies differential starting wages, not differential returns to experience. However, to the extent the mentor-protege relationship serves as a vehicle for transmitting human capital from the former to the latter, we should observe differential returns to experience, for the protege, during the period of protegeship. To be precise, the necessary condition for observing statistically significant returns to protegeship experience is that proteges receive human capital that otherwise identical nonproteges do not receive and/or proteges receive the same human capital at a faster rate than otherwise identical nonproteges.

We generally expected attorneys working in the private sector to cam more than those working in the public sector. Among the attorneys in private practice, we expected partners to earn more than associates. We also expected measures of the respondent's academic performance while in law school and his/her internship experience to influence earnings, in a positive manner. That is, we expected respondents who graduated in the top quartile of their class to earn more than those who graduated in the 2nd, 3rd and 4th quartiles; those who graduated in the 2nd quartile to earn more than those who graduated in the 3rd and 4th quartiles, and so on. We also expected earnings to be positively related to city size, for cost-of-living reasons.

Table III reports results of least squares regression estimation of the determinants of (log) hourly earnings, defined as annual earnings divided by number of hours worked per year.(5) Earnings were measured per hour worked as an automatic control for work effort.

Proteges per se received lower hourly earnings than unmentored individuals. The coefficient estimate for PROTEGE is shown to be significant at nearly the 5 percent level. However, due to the presence of multicollinearity between PROTEGE and other explanatory variables (principally EXPERIENCE), the variance inflation factor for PROTEGE is 5.482, which means that the standard error calculated for this estimated coefficient is inflated by 2.34 times.(6) This suggests that the true statistical significance of the PROTEGE variable may be considerably greater than the reported significance level. In additional, but unreported, regression analyses, we found no evidence of significantly different annual earnings between mentored and unmentored respondents. Rather, proteges' work effort, in terms of the number of hours worked per year, exceeds that of nonproteges.(7) This finding is consistent with our assumption of an opportunity cost of training during the period of active protegeship.

The regression results reveal the expected positive and diminishing returns to experience. Hourly earnings increase at a little less than 8 percent per year and peak at approximately 31.5 years of experience. On top of these returns, however, proteges exhibit additional positive but diminishing returns to experience with a mentor (protege x experience).(8) The variance inflation factor for PROTEGE x EXPERIENCE is 14.331, which means that the standard error calculated for the coefficient estimate is inflated by 3.79 times; the standard error around the coefficient estimate of the squared term is inflated by 2.81 times. Again, this suggests that the true statistical significance of the PROTEGE x EXPERIENCE variables may be considerably greater than the reported significance level. Since the greater work effort of proteges is controlled for by the PROTEGE variable, and passive acquisition of human capital is controlled for by the EXPERIENCE and EXPERIENCE SQUARED terms, the significant returns to experience with a mentor evidently results from the mentored respondent's enhanced productivity. We can only assume that this reflects general and specific human capital acquired from the mentor. Our findings in this regard clearly support a human capital transfer function for the mentor-protege relationship.

The matching function, by contrast, implies no positive differential earnings growth with experience for mentored individuals as compared to unmentored individuals. Indeed, a matching explanation of mentoring implies lower earnings growth for proteges than for nonproteges. Since the only difference between matched and unmatched workers is the firm's uncertainty regarding the workers' productivity, matched workers, about whom the firm's managers have less uncertainty, receive higher starting salaries. Since there are no productivity differences between the two groups, earnings growth, as a percent of starting pay, should thus increase faster for the unmentored individuals. Thus our results on hourly earnings do not support a matching function for mentoring but do support a training function and/or a bonding function vis-a-vis the firm's commitment to reward productivity.

Other variables in the regression estimation conform to our general expectations. Solo practitioners earned significantly less than every other category of attorney with the exception of junior attorneys in private firms with 2-3 members. The relative sizes of the coefficient estimates suggests that both corporate and private practice attorneys earned more than public sector attorneys in 1983; with compensation for both associates and partners in private practice showing an upward trend as firm size increases. There is a positive relationship between city size and hourly pay.

Respondents who were in the bottom quartile of their law school class at very prestigious and somewhat prestigious law schools earned less than respondents who did not attend those law schools. By the same token, respondents who were in the top quartile of their class at very prestigious law schools earned more than respondents who did not attend very or somewhat prestigious law schools. Earnings were positively related to the respondent's LSAT scores, his/her having been an intern in a private practice law finn or in a public sector agency and working on the law review at a somewhat prestigious law school.

Turnover

We argued previously that the M-P relationship may serve to cement long-term relationships between workers and firms by increasing workers' investment in firm-specific human capital. We used two questions included on the Survey to shed empirical light on this proposition. One question was phrased, "do you plan to change your employment within the next few years?" Those respondents who answered this question in the negative, were then asked, "if you felt you had a reasonable alternative option, would you change your employment?" We collapsed these two questions into one variable which was assigned a value of one if respondents either planned to or were willing to change employment for any reason. Table IV reports logistic regression estimation of this respondent attachment variable. We expected proteges to report greater attachment to their firms, i.e., to observe a negative sign on the PROTEGE coefficient.

The significant, negative sign on PROTEGE in the attachment model indicates that mentored individuals at least say they will be more loyal to their firms than nonmentored individuals. We suspect this is true for complementary reasons. The protege likely develops at least some specific human capital that is nontransferable to other firms. By itself, this is not sufficient to induce loyalty, as the firm may renege on promises to reward the individual for investing in SHC. It is the firm's willingness to absorb some loss of productivity by the mentor during the protegeship period that serves to guarantee its good intentions vis-a-vis proper compensation of the protege, thus making the investment a positive net present value project for the protege. The more significant the general human capital component of the job relative to specific human capital, the greater the need for a mechanism that bonds workers to firms and vice-versa.(9)

Our findings on loyalty to the firm are consistent with Roche [22], who found that unmentored executives reported a willingness to change jobs substantially in excess of the reported willingness [TABULAR DATA OMITTED] to change jobs among mentored executives. Similarly, both Dalton, Thompson and Price [5] and Kram [12] found that mentoring reduces turnover among young professional corporate managers.

IV. Concluding Comments

We have argued that the mentor-protege relationship may be an efficient means for firms to (1) induce junior employees to invest in acquiring general and firm-specific human capital, thus (2) reducing employee turnover (enhancing employee loyalty). Our findings of positive differential returns to experience among mentored individuals and negative per se returns to being mentored provides indirect evidence that proteges are trained by mentors. We also find that mentored individuals are less likely to leave the firm given a viable alternative option (i.e., they are more loyal); this also is consistent with a human capital development function for the mentor-protege relationship.

As we noted previously the protege identification question asked on the survey is framed so as to understate the true incidence of mentoring in the legal profession.(10) We can identify only those individuals who had a mentor when the survey was administered, not those who ever had one. Some fraction of the respondents who reported having no mentor undoubtedly had a mentor at some point in their legal career. This under-reporting of overall mentoring works against our finding significant differences between the two groups of individuals.

Consider the search for possible earning differentials. If, as our data seem to indicate, mentored individuals experience greater earnings than unmentored individuals, the inclusion of unidentified but mentored individuals in the group identified as not currently having a mentor implies that measured earnings of that group exceed the true earnings of unmentored individuals. Any estimated earnings differential between mentored and unmentored individuals will be smaller than the true differential. Our consistent findings with respect to the impact of being mentored on earnings and attachment to the firm are all the more impressive given the imprecision in identifying mentored individuals.

Future research might profitably focus upon the period of active protegeship for clues to the function of the M-P relationship. For example, we would expect that if human capital augmentation is the primary function of the relationship, then faster learners ought to be proteges for less time than slower learners. Similarly, drawing from our previous work on occupational following in the legal profession (Laband and Lentz [13]), we would expect fewer second (or nth) generation lawyers to contract with mentors for human capital augmentation reasons regardless of where they work. In the context of the family law firm, nth generation lawyers would be even less likely to be proteges, since their family ties serve a bonding function equivalent to contracting with a mentor. Similar implications fall out with respect to incidence of mentoring.

Appendix. Variable Definitions and Means (All Attorneys, Proteges,
Nonproteges)

PROTEGE: Respondent indicated (s)he was currently a
 protege (.33, 1, 0)

EXPERIENCE: Number of years since receipt of the JD/LLD
 (9.91, 5.19, 12.27)

FEMALE: 1 = female; 0 = male (.17, .19, .16)

MINORITY: Respondent a member of a minority (Black,
 Hispanic, Oriental) (.02, .02, .03)

HOURLY WAGE: (23.98, 19.54, 26.58)

ANNUAL HOURS: Hours worked per year (2359.98, 2420.02, 2324.36)

MARRIED: Respondent was married (.72, .67, .74)

NUMCHILD: Number of children respondent had (1.32, .89, 1.53)

#DIFFJOBS: Number of different jobs held prior to respondent's
 current job (1.54, 1.14, 1.74)

VERYVIEW: Respondent worked on the Law Review at a very
 prestigious law school (.05, .04, .05)

SOMEVIEW: Respondent worked on the Law Review at a somewhat
 prestigious law school (.10, .11, .10)

TOPVERY: Respondent placed in the first quartile
 academically at a very prestigious law school
 (.10, .09, .11)

SECVERY: Respondent placed in the second quartile
 academically at a very prestigious law school
 (.09, .09, .08)

THRVERY: Respondent placed in the third quartile
 academically at a very prestigious law school
 (.03, .03, .03)

FOURVERY: Respondent placed in the fourth quartile
 academically at a very prestigious law school
 (.01, .01, .01)

TOPSOME: Respondent placed in the first quartile
 academically at a somewhat prestigious law school
 (.20, .22, .19)

SECSOME: Respondent placed in the second quartile
 academically at a somewhat prestigious law school
 (.16, .16, .16)

THRSOME: Respondent placed in the third quartile
 academically at a somewhat prestigious law school
 (.07, .07, .07)

FOURSOME: Respondent placed in the fourth quartile
 academically at a somewhat prestigious law school
 (.02, .03, .02)

LSAT: Respondent's score on Law School Aptitude Test
 (542.55, 592.82, 517.29)

PRIVATEINTERN: Respondent did a legal internship in a private law
 firm (.54, .66, .48)

GOVT. INTERN: Respondent did a legal internship in government
 (.22, .26, .20)

LEGALAID: Respondent did a legal internship in a legal aid
 clinic (.16, .17, .16)

OTHERINTERN: Respondent did a legal internship in some other
 context (.16, .15, .16)

CORPEXEC: Respondent was in an executive position in a
 corporation (.03, .02, .04)

CORPSUPVSR: Respondent was in a supervisory position in a
 corporation (.02, .02, .02)

CORPJR: Respondent was a junior attorney in a corporation
 (.04, .06, .03)

CORPOTHER: Respondent was in a non-executive, non-supervisory,
 non-junior position in a corporation
 (.00, .00, .00)

GOVT.EXEC: Respondent was in an executive position in a
 government agency (.01, .00, .01)

GOVT.SUPVSR: Respondent was in a supervisory position in a
 government agency (.03, .02, .03)

GOVT.JR: Respondent was a junior attorney in a government
 agency (.01, .01, .02)

GOVT.OTHER: Respondent was in a non-executive, non-supervisory,
 non-junior position in a government agency
 (.06, .06, .05)

JRFIRMSIZE2: Respondent worked in firm having 2-3 lawyers
 (.02, .05, .01)

JRFIRMSIZE3: Respondent worked in firm having 4-9 lawyers
 (.08, .13, .03)

JRFIRMSIZE4: Respondent worked in firm having 10-20 lawyers
 (.05, .09, .03)

JRFIRMSIZE5: Respondent worked in firm having 21-30 lawyers
 (.02, .03, .01)

JRFIRMSIZE6: Respondent worked in firm having 31-60 lawyers
 (.03, .06, .02)

JRFIRMSIZE7: Respondent worked in firm having 61-90 lawyers
 (.02, .04, .01)

JRFIRMSIZE8: Respondent worked in firm having 90+ lawyers
 (.05, .08, .03)

PARTFIRMSIZE2: Respondent worked in firm having 2-3 lawyers
 (.09, .06,. 11)

PARTFIRMSIZE3: Respondent worked in firm having 4-9 lawyers
 (.09, .07, .09)

PARTFIRMSIZE4: Respondent worked in firm having 10-20 lawyers
 (.04, .06, .03)

PARTFIRMSIZE5: Respondent worked in firm having 21-30 lawyers
 (.01, .00, .02)

PARTFIRMSIZE6: Respondent worked in firm having 31-60 lawyers
 (.02, .02, .02)

PARTFIRMSIZE7: Respondent worked in firm having 61-90 lawyers
 (.01, .01, .01)

PARTFIRMSIZE8: Respondent worked in firm having 90+ lawyers
 (.03, .03, .03)

PETITCITY: Firm is located in 25,000 [is less than] Town
 [is less than] 49,999 (.08, .06, .09)

SMALLCITY: Firm is located in 50,000 [is less than] Town
 [is less than] 249,999 (.21, .21, .21)

MEDIUMCITY: Firm is located in 250,000 [is less than] Town
 [is less than] 1,000,000 (.24, .28, .22)

LARGECITY: Firm is located in Town [is greater than] 1,000,000
 (.25, .32, .21)

1. Fitt and Newton [6] recognize the potential adverse consequences to the mentor from poorly identifying the capabilities of their proteges: "The mentor accepts a big risk that the protege's performance may fail to meet expectations. In this case, his peers and superiors may question his judgment."

2. These results are consistent with Roche's [22] findings. He reports: "In ranking the characteristics most important for a mentor to have, respondents gave the highest value by far to a mentor's 'willingness to share knowledge and understanding'."

3. The data utilized in this study were made available (in part) by the Interuniversity consortium for Political and Social Research. The survey was conducted by Ronald L. Hirsch. Neither the original collector of the data nor the Consortium hear any responsibility for the analyses or interpretations presented here.

4. The questions pertaining to current work experience included number of hours worked per week, tasks performed, type of law engaged in, compensation, vacation days allowed and taken, size of firm, treatment by superiors and plans to change employment, to name a few. The respondent was not asked to identify his/her current employer or to rank the prestige of the organization.

5. To check the sensitivity of our results to estimation technique, we also estimated the earnings equation using an ordered logit technique, given the categorical nature of the dependent variable. Like Ribitzer and Taylor [21, 35], who used the same data set that we employed, we found the ordinary least squares estimation results and the ordered logit results to be nearly identical. We also estimated selectivity-adjusted earnings functions, only to find that the coefficient estimate of the selectivity variable was consistently insignificant. We do NOT report results of an earnings decomposition because our inability to identify proteges in the post-protege period necessarily implies that returns to experience for each group will be mis-estimated. White's [25] test statistic for homoskedastic error terms is a chi-square equal to 1217.57, with 1226 degrees of freedom; we fail to reject the assumption of homoskedastic errors.

6. In determining the variance inflation factor for explanatory variables, each regressor is regressed against all other regressors, with a unique R-square value calculated for each one. The variance inflation factor for explanatory variable i equals: [Mathematical Expression Omitted]. For further discussion, see Lardaro [15, 445].

7. These results are available upon request.

8. The test statistic for the test that the coefficients on Protege, Protege x Experience and Protege x [Experience.sup.2] exert a jointly significant influence on earnings is a chi-square equal to 21.7962, which is significant at the .01 level.

9. We investigated empirically the relative importance of general versus specific human capital in the legal profession, by estimating the returns to experience (GHC) and tenure with the same firm (SHC) for attorneys in private practice. Estimated returns to experience were sizable and significant; those for tenure were not. These findings suggest that general human capital is of greater importance than specific human capital, at least among this subset of attorneys.

10. Strober's survey question is framed similarly to ours: "In your current job, do you have a mentor (or mentors), that is, a person in a higher position in your company who is particularly helpful and supportive to your career?" Fifty-two percent of the women and 46 percent of the men respondents reported that they had a mentor.

References

1. America, Richard R. and Barbara E. Anderson. Profile of a Woman Officer. New York: Heichrick and Struggles, 1979.

2. Barron, John M., Dan A. Black, and Mark A. Loewenstein, "Job Matching and On-The-Job Training." Journal of Labor Economics, January 1989, 1-19.

3. Collins, Eliza and Patricia Scott, "Everyone Who Makes It Has a Mentor." Harvard Business Review, July/August 1978, 89-101.

4. Corcoran, Mary, Linda Datcher and Greg J. Duncan. "Information and Influence Networks in Labor Markets," in Five Thousand American Families - Patterns and Progress, edited by Greg J. Duncan and James N. Morgan. Ann Arbor, Michigan: Institute for Social Research, 1980.

5. Dalton, Gene W., Paul H. Thompson, and Raymond L. Price, "The Four Stages of Professional Careers - A New Look at Performance by Professionals." Organizational Dynamics, Summer 1977, 19-42.

6. Fitt, Lawton W. and Derek A. Newton, "When the Mentor Is a Man and the Protege a Woman." Harvard Business Review, March/April 1981, 56, 58, 60.

7. Hirsch, Ronald L., "Are You on Target?" Barrister, Winter 1985, 17-29.

8. Jovanovic, Boyan, "Matching, Turnover, and Unemployment." Journal of Political Economy, February 1984, 108-22.

9. -----, "Firm-Specific Capital and Turnover." Journal of Political Economy, December 1979a, 1246-60.

10. -----, "Job Matching and the Theory of Turnover." Journal of Political Economy, October 1979b, 972-90.

11. Klein, Benjamin and Keith B. Leffler, "The Role of Market Forces in Assuring Contractual Performance." Journal of Political Economy, August 1981, 615-41.

12. Kram, Katherine E. Mentoring at Work, Glenview, Ill.: Scott-Foresman, 1985.

13. Laband, David N. and Bernard F. Lentz, "Self-Recruitment in the Legal Profession." Journal of Labor Economics, April 1992, 182-201.

14. ----- and -----. "Implicit Contracts and Performance Bonds: The Mentor-Protege Relationship in Legal Partnerships." Working paper, 1994.

15. Lardaro, Leonard. Applied Econometrics, New York: Harper-Collins, 1993.

16. Leibowitz, Arleen and Robert D. Tollison, "Free Riding, Shirking, and Team Production in Legal Partnerships." Economic Inquiry, July 1980, 380-94.

17. Levinson, Daniel J., Charlotte N. Darrow, Edward B. Klein, Maria H. Levinson, and Braxton McKee. Seasons of a Man's Life, New York: Knopf, 1978.

18. Ochberg, Richard L., Gary L. Tischler, and Herbert C. Schulberg, "Mentoring Relationships in the Careers of Mental Health Administrators." Hospital and Community Psychiatry, September 1986. 939-40.

19. Parsons, Donald O., "Specific Human Capital: An Application to Quit Rates and Layoff Rates," Journal of Political Economy, November/December 1972, 1120-43.

20. Pencavel, John H., "Wages, Specific Training, and Labor Turnover in U.S. Manufacturing Industries." International Economic Review, February 1972, 53-64.

21. Rebitzer, James B. and Lowell J. Taylor, "Incentive Structures and Market Outcomes: The Case of Law Firms." Industrial Relations Research Association Series, January 1993, 32-41.

22. Roche, Gerard R., "Much Ado About Mentors." Harvard Business Review, January/February 1979, 14-28.

23. Simon, Curtis J. and John T. Warner, "Matchmaker, Matchmaker: The Effect of Old Boy Networks on Job Match Quality, Earnings, and Tenure." Journal of Labor Economics, July 1992, 306-30.

24. Strober, Myra H. "The MBA: Same Passport to Success for Women and Men?" in Women in the Workplace, edited by P. A. Wallace. Boston: Auburn House, 1982.

25. White, Halbert, "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity." Econometrica, May 1980, 817-38.

26. Zey, Michael G. The Mentor Connection. Homewood, Ill.: Dow Jones-Irwin, 1984.