Appendix A
METHODOLOGY
This study began with three related questions. How does Wall Street’s system of pay for performance allow gender inequality to persist? How do some women manage to become highly successful on Wall Street? And what can we learn from the Wall Street case about gender inequality more generally? At a general level, these are questions about the processes that lead to a well-known outcome: gender inequality in the workplace. At a more specific level, these questions hone in on the effects of a growing trend in advanced capitalism, which is the movement toward flat organizations that pay variable amounts based on performance evaluations.
These research questions called for an exploratory design that was intended to compare otherwise similar men and women and to generate theory. As it turned out, some well-developed social psychological theories provided insights that, coupled with an understanding of the variable pay system, suggested why performance-based pay systems can be bad for women. Removed from the laboratory, these processes interacted with social structures to produce uniquely determined outcomes. But these processes and the importance of the pay system for how they operated on Wall Street could not be uncovered using existing data or exclusively quantitative data. For this reason, I combined qualitative and quantitative approaches using a carefully designed sample.
I chose a “grounded theory” approach as the most appropriate research strategy. (See Glaser and Strauss 1967 or Strauss and Corbin 1998 for a description of this approach.) In-depth interviews with a carefully targeted group of men and women who started careers on Wall Street during the same period facilitated the process of discovery by enabling theoretical concerns to emerge and develop throughout the interview stage of the research. It was through this process of discovery that the performance-based bonus system and the degree of subjectivity in performance criteria emerged as pivotal influences on gender inequality in this setting.
Wall Street is an extreme case rather than a typical or representative one. The extreme nature of this case is useful for three reasons. First, the sharp contradiction between time demands in Wall Street’s culture and the personal needs of workers calls attention to much-debated time-money trade-offs. These trade-offs are underlined in an industry that requires such intense time commitments from workers. Second, the extremely high pay of this setting offers Wall Street workers more choices than most other workers in terms of childcare and domestic arrangements and the outsourcing of personal and household care. As a result, their struggles with these work-life issues accentuate the constraints that these issues pose in contemporary workplaces. Third, and most important, the proportion of annual pay that takes the form of a variable bonus highlights the effects of a performance-based reward system on gender inequality. The importance of performance evaluations in this setting throws subjective and nonmerit influences into sharp relief. All of these phenomena may be found in other organizations in less extreme form, especially as greater numbers of them move toward variable pay and long hours. Understanding how gender inequality emerges among extremely similar men and women in this setting reveals the processes that reinforce inequality in a growing segment of the labor market.
Sampling
The study design required a sample that could compare men and women from a specific cohort of entrants to Wall Street who were similar in their credentials and background characteristics, the prestige of the firms where they started, and the market cycles that affected their careers. As a result, the sampling procedure had to locate a delimited group of financial professionals. Accomplishing this task required finding the population of interest and developing a strategy for sampling within it.
Traditional random sampling methods were unacceptable for obvious reasons. Attempting to draw a random sample within census tracts (the preferred method for securing an unbiased and representative sample) would have yielded too few Wall Street workers even in the New York City area, and any that did appear in the sample would be likely to have started at different times and to differ widely in their work experience and backgrounds. But abandoning the principle of randomness altogether and using a snowball sample, whereby initial respondents provide contact with others in their own networks, was also undesirable and involved unacceptable risks. The potential for self-selection and systematic bias would have compromised the findings. As a result, I rejected this type of sampling method as well.
The sampling strategy that I ultimately accepted was one that used placement reports and alumni information from graduate schools of business. Major Wall Street investment banks hire 75–85 percent of their incoming associates from the top four MBA programs in finance, with some associates coming from an additional two or three schools. Because the firms of interest hire the majority of their incoming associates from the top MBA programs, I drew my sample from the placement reports and alumni directories of five elite graduate programs in finance for the years 1991–93. Most members of these cohorts were in their early to mid-thirties at the time of the interview, a prime time in the life cycle for family formation.
These years were also chosen because these graduates would have passed their first promotion and would have been in the position of vice president or higher at the time of the interview if they had remained in the industry. While there would be advantages to examining professionals who are more advanced in their careers, especially in terms of differences in compensation, locating a cohort that had passed two or more promotional levels would have been difficult because women have higher rates of attrition and are harder to locate, making it even more difficult to obtain a suitable comparison sample of more senior women that did not include only the most successful. This would probably also have negatively affected the response rate and created greater selection bias. Since the early years of an investment banker’s career are the most likely to result in attrition, and reaching the VP level is a substantial accomplishment in itself, this sampling design was optimal for addressing the primary research questions of this study.
In MBA programs, women represented approximately 25–30 percent of all graduates in the 1990s. Employees in major investment banking firms claimed that approximately 15–20 percent of incoming associates in an average year were women. Confirming this, among the names obtained through the five graduate schools of business, women were 19.8 percent of the graduates who entered investment banks upon graduating. This suggests that my sampling strategy obtained a reasonable approximation of the population of interest.
From the cohort of graduates, I compiled a list of those who obtained jobs as associates in the major investment banks to provide the sample population.1 These organizations are primary underwriters of initial public offerings and mergers and acquisitions. They developed as distinct from commercial banks because of the separation of securities from depository banking activities in the Glass-Steagall provisions of the 1933 Banking Act.2
For the purposes of this study, I used Eccles and Crane’s original framework and accounted for mergers by including employees and former employees at Morgan Stanley, Goldman Sachs, Merrill Lynch, Lehman Brothers, Salomon Brothers, Smith Barney, Credit Suisse First Boston, J. P. Morgan (a commercial bank that expanded into investment banking in the 1980s), Bear Stearns, and DLJ. Because it was important to delimit the scope of the study and traditional investment banks maintained a higher level of prestige than commercial banks, I selected only workers who began their careers at the above ten firms. An MBA graduate who started in any of these firms was eligible for selection.
The five graduate schools of business contained 621 graduates for the three years included in the sample. To obtain comparable samples of men and women from the lists of MBA graduates, I separated all eligible respondents’ names by gender to create lists of 498 men and 123 women.3 I then selected samples of men and women using a random numbers table. I attempted to locate and contact as many potential respondents as possible. I sought those who had left the companies where they started investment banking careers using telephone directories, Internet searches, and alumni directories. Women were more difficult to locate because many had married and changed their surnames since business school and many did not have a telephone number listed in their name. In another strategy to locate more potential respondents, at the end of each interview I asked respondents if they knew the whereabouts of their classmates or individuals who were in their incoming class of associates.4 Among those selected for an interview, 29 were impossible to find (17 women and 12 men), and 10 refused to participate (5 women and 5 men). So the response rate is 66 percent and the cooperation rate is 88 percent. This compares favorably with many other studies, especially of professionals.
The final sample contained 44 women and 32 men. The initial target of equal size samples of each gender was abandoned due to the greater interest in women’s experiences in a historically male-dominated field, as well as women’s greater willingness to be interviewed. The sample included those who remained in major investment banks, those who moved into other financial organizations, and those who left finance careers entirely.
This sampling strategy proved to be very effective and provided a number of advantages. First, these men and women were very similar in their background characteristics, education, skills, and qualifications. Because they graduated from elite MBA programs, all of them shared the most important educational credential for a career on Wall Street. These MBA programs also tend to admit students who are similar in their class background, previous educational and work experience, and their performance on the GMAT. This group of individuals entered a small set of organizations with comparable prestige within a single industry during a short period of time; they faced similar market conditions and their initial employers had similar organizational prestige. As a result, this study can assess gender differences among very comparable men and women. Also, I was able to study these individuals without the involvement of the firms. It would have been extremely difficult to gain the cooperation of even one of these highly secretive organizations. Even if they had provided access, a firm-generated sample would likely have contained less variation in areas of financial services, and respondents might have been less candid if they believed that their employer could have access to the data.
The Interview
In 1998 and 1999, I conducted interviews that combined fixed-choice (quantitative) and open-ended (qualitative) questions. Responses to fixed-choice questions were used to construct the measures used in the statistical analyses, while open-ended qualitative responses revealed the processes underlying the quantitative results. Interviews lasted an average of one hour and were partially structured to elicit respondents’ career histories from the time they completed their bachelor’s degrees until the time of the interview, their responses to events, their motivations for making career decisions, and their total compensation for 1997. By collecting quantitative and qualitative data, I was able to analyze both outcomes (the degree of gender inequality in pay) and processes (the mechanisms that produced this inequality) for men and women who were highly similar in their background characteristics, educational and work qualifications, and the organizations and market cycles in which they began their careers.
The interview schedule was developed over a period of months, during which eleven pretests helped to refine and reorder a complete career history questionnaire. (See appendix C.) This structured interview format was necessary to guide respondents’ stories and to facilitate recall. The interview assumed the form of a career history from before the MBA until the time of the interview. The final interview schedule was structured to ensure comparability across respondents while remaining open-ended enough to permit probing and discovery, and to accommodate new insights as they occurred. I made only a few minor changes after the interviewing began, largely to make the career histories easier for respondents to recall and to accommodate emerging insights about the subjectivity or objectivity of performance criteria.
I approached respondents by telephone, usually with an initial voice-mail message indicating the purpose of the call. In some cases, home addresses were available through telephone directories, and respondents received letters in advance indicating the nature of the study. The initial contact described the purpose of the study, confidentiality of responses, and the nature of participation—a onetime interview lasting approximately forty-five minutes. Because the gender component of the research question might be expected to intrigue female respondents, I told women that I was particularly interested in women’s experiences, including how they balance work and family life. For men, I also mentioned an interest in work and family life, but focused more on job changes and career paths. I did not inform men that I was studying gender differences. This small degree of deception was necessary in order to reduce rejections from potential male respondents.
The interviews took place at a time and location of each respondent’s choice. I conducted all interviews myself, and they lasted between forty minutes and one hour and fifteen minutes. Some people allotted me a very short time slot, in which case I had to selectively ask the most important questions and omit some probes. I tape recorded most interviews,5 which were later transcribed verbatim and analyzed. Fixed-choice questions were fully structured, while open-ended questions were semi-structured but also flexible to allow for probes into transition points, motivations for career decisions, and choices concerning work-life balance and other aspects of their experiences.
Because of the time-intensive single-interviewer design, the sample size is smaller than might be desirable for quantitative analysis. But the in-depth interview offers some unique advantages over a large-scale survey with exclusively fixed-choice answer categories. Most important, it captures the processes involved in professionals’ careers as they developed over time rather than providing a snapshot of the outcomes (as in a purely quantitative survey). Admittedly, there may be problems with memory recall, and respondents may subjectively reinterpret their pasts. But the in-depth interview offered insights into the meaning of people’s lives as they experienced them. Responses to open-ended questions revealed how social psychological processes interacted with the compensation system to heighten a pattern of gender inequality. The insights from the open-ended responses complemented the quantitative analyses using the same sample to present an overall picture of young financial professionals, the culture of finance, and gender differences on Wall Street.