What Brings Female Professionals to Entrepreneurship? Exploring the Antecedents of Women’s Professional Entrepreneurship

Abstract

This study explores how female professionals engage in starting their own businesses, known as professional entrepreneurship. In particular, this study specifies what factors foster the likelihood of self-employment of female professionals. Drawing upon the push and pull theories of entrepreneurship, we argue that individual capabilities (as a pull factor) make the self-employment of female professionals less likely, while discrimination experiences (as a push factor) make the self-employment of female professionals more likely. Given such bifurcated effects of these factors, we examine the combinatory effects of individual capabilities and discrimination experiences (which are specified as attribute-based and family-based discrimination experiences) on the rate of self-employment of female professionals. With a sample of 1356 female lawyers in the U.S., we test our hypotheses predicting the rate of self-employment with respect to prior salary and discrimination experiences. Our results reveal that prior salary (a pull factor) motivates female lawyers to stay at the traditional law firms, whereas attribute-based discrimination experiences (a push factor) motivate them to open their own office. Furthermore, we find that such a push effect is pronounced only among the female lawyers with lower salaries. Then, the empirical findings are discussed to elaborate the process of female professionals’ entrepreneurship.

1. Introduction

Literature on entrepreneurship has focused mainly on idea recognition, generation, and implementation [1,2,3], whereas it is still developing the entrepreneurial processes of minorities [4,5,6]. In fact, the entrepreneurial processes of minorities are understood as topics related to self-employment [7,8,9]. Studies on self-employment pay much attention to the question of why people who are disadvantaged choose to be self-employed, and there are two hypothetical arguments that are popularly known as the “pull” and “push” theories [10,11,12]. The former is that attractive business opportunities will pull individuals with higher levels of capability, whereas the latter is that factors such as low levels of education or discrimination, which are hard to overcome, push workers into self-employment [8,12,13,14]. Studies on women and minorities address the fact that such minority populations may be pushed toward self-employment because they are more likely to experience discrimination, such as the glass ceiling, compared to the majority [13,15]. The glass ceiling reflects a barrier that prevents minorities from professionally advancing into higher levels of an organization [16]. Since the glass ceiling effect is closely related to an individual’s career when embedded in traditional organizations, self-employment of female professionals will be influenced by such social barriers [17].

Based on this understanding, we have more opportunities to develop the individual aspects of women’s self-employment. For example, even though individuals are structurally disadvantaged, they might be able to craft their own ways to cope with such circumstances. This brings us to the question of whether and how women who have enough capabilities to cope with disadvantaged social structures are motivated to be self-employed. In particular, since professional women’s career capabilities are legitimated by accreditation processes (e.g., academic degrees or professional certificates), female professionals are influenced by not only the push factors (e.g., discrimination or the glass ceiling) but also the pull factors (e.g., professional capabilities). Given that self-employment is understood as an action taken to avoid the traditional hierarchical organizations [17], if women’s abilities are accredited by society (e.g., professionals), they might be reluctant to leave their organizations and be willing to persist in their positions there [18]. On the other hand, it can be argued that with the accredited abilities, women who start their own businesses as professions are afforded greater personal autonomy [8,9]. Then what significantly motivates female professionals to start their own businesses? Or when and how do female professionals decide to leave the traditional organizations to start their own businesses? Our main goal for this study is to address these questions.

In addressing these questions, we postulate that the process of profession-driven self-employment should be differentiated from self-employment driven by female professionals. Drawn from the theories on entrepreneurship such as pull and push, this study focuses on investigating factors that motivate the self-employment of female professionals. As such, this study provides insights into how female professionals cope with their ambivalence between professional accreditation and social barriers. To explore the impacts of the pull and push factors on women’s choice of self-employment, we consider the population of female lawyers in the U.S. As the degree of Juris Doctor (JD) represents credibility in practicing law, female lawyers reflect women who have professional capabilities as well as the possibilities to encounter social barriers within an organization. With a sample of 1356 female lawyers in the U.S., we examine the effect of a pull factor instantiated by prior salary and push factors instantiated by discrimination on the rate of self-employment.

In this paper, we first review the literature on professional entrepreneurship and on the push and pull factors affecting the self-employment process of female professionals. With these literature reviews on the push and pull factors, we develop our hypotheses in Section 2. In Section 3, we specify how we collected the data and employed a statistical method to examine our hypotheses. In Section 4, we present our findings based on the Cox proportional hazard regression models, followed by robustness checks. In Section 5, we discuss our findings, and then draw conclusions by providing future research suggestions in Section 6.

2. Theory and Hypotheses

2.1. Professional Entrepreneurship

Professional entrepreneurship refers to “self-employment to join professional partnerships and establish a professional practice” [19] (p. 12). Simply put, professional entrepreneurship is the foundation of a business-oriented organization driven by professionals, such as freelancing or self-employment. In particular, professionals choose the form of self-employment as a type of entrepreneurship and have some commonality, such as autonomy and seeking better earnings [8,9]. As such, the professionals encompass “not only the traditional classic professions (physicians, lawyers, and university professors) but also includes other abstract workers and symbolic analysts such as accountants, pharmacists, engineers, and scientists” [20] (p. 431–432).

For a more comprehensive understanding of professional entrepreneurship, we need to determine the nature of professions. Professionals are organized groups that perform their own functions for the society and play a significant role in a society [18]. Greenwood [18] enumerated the attributes of a profession, which are “(1) systematic theory, (2) authority, (3) community sanction, (4) ethical codes, and (5) a culture” (p. 45). Related to this, professional practices are characterized by “(a) the application of theoretical and scientific knowledge to tasks tied to core social knowledge or to tasks tied to core societal values (health, justice, financial status, etc.), (b) considerable autonomy and freedom from oversight, except by peer representatives of the professional occupation, and (c) exclusive or nearly exclusive control over a task domain linked to the application of the knowledge imparted to professionals as part of their training” [20] (p. 431). These characterizations of professions and professional practices imply that people who hold the position of a professional can function with autonomy and exclusivity. Since professionals provide specialized knowledge or skills, their activities tend to be self-disciplined. As such, through professional entrepreneurship, professionals can provide their specialized knowledge or skills to the public by expanding the scope of autonomy.

2.2. Self-Employment of Female Profesionals

2.2.1. Pull Theory

The pull theory postulates that “the existence of attractive, potentially profitable business opportunities will attract (“pull”) alert individuals into entrepreneurial activities” [10] (p. 46). According to the definition of Amit and Muller [14], the “pulled” entrepreneurs are referred to as “individuals who are attracted to start his or her own business due to his or her personal unforced desires.” The personal unforced desires include job achievements [21,22], recognition of business opportunities [23,24], expected earnings [25,26], and the aspiration for job autonomy and independence [11,27]. Rindova et al. [28] stated that the most significant driver of entrepreneurship is the hope to achieve autonomy, or desire for independence. In their study, they contended that “seeking autonomy,” “authoring,” and “making declarations” are the core elements in becoming self-employed.

By nature, professionals have greater autonomy than other employees, and thus accreditation of professional practices defines the range of jobs in which the professionals are engaged [18,20]. Because of these, professional entrepreneurship is motivated by the individual attributes needed to accomplish their professional works (e.g., high level of education and experience) [12]. As evidence, Carr [29] and Robinson and Sexton [30] consistently showed that education and experience lead to women’s self-employment.

However, when these internally-constructed pull factors are embodied in traditional organizations, they may also encourage professionals to remain at the organizations. For example, the individual capabilities (and thus their achievements) in professional organizations such as law firms are usually reflected by salary contracts. Based on what the professionals accomplished during previous time periods, the salaries for the next time periods are determined. As such, if one professional earns a higher salary than the other, this means that the former has more competitive advantages at the organization. If a professional is recognized within the organization, he or she is likely to stay at the organization.

This is more likely to occur for female professionals (e.g., [17]). As Fossen [31] and Leveine and Rubinstein [32] specified, women tend to be more risk-averse, which will drive them to remain in the system of salaried employment rather than turn to self-employment. Accordingly, the choice of self-employment by female professionals with high prior salaries is unlikely to unfold. That is, the more prior salaries signify their competitive advantages within the traditional organizations, the less likely the female professionals will participate in self-employment. Thus, we hypothesize the following:

Hypothesis 1.

If a female professional had a high salary in prior organizations, the rate of self-employment participation will be slower.

2.2.2. Push Theory

The push theory argues that people are pushed into entrepreneurship by “negative situational factors such as dissatisfaction with existing employment, loss of employment, and career setback” [10] (p. 46). In a similar vein, Amit and Muller [14] defined “pushed” entrepreneurs in terms of “an individual’s dissatisfaction with his or her current position, for various reasons unrelated to his or her entrepreneurial characteristics, pushing him or her to start his or her own business.” Push factors are negative motivations that move individuals towards forms of entrepreneurship [33]. These types of entrepreneurs are sometimes called “forced entrepreneurs” [34].

In addition to direct studies of the relationship between job satisfaction and the decision to become an entrepreneur, psychological evidence supporting the push theory includes studies that variously describe entrepreneurs as “misfits,” “rejects from society,” and “displaced individuals” [10] (p. 46). That is, when entrepreneurs perceive their environment as hostile and tumultuous, these negative factors will push them into business activities [10].

There are two factors related to the push theory for female professionals. Overall, the push theory signifies that the dissatisfaction with existing employment, loss of employment, and career setback will make female professionals consider self-employment participation. One of the salient push factors is attribute-based discrimination. For example, female professionals may experience discrimination when they are omitted from critical projects in the organization due to their gender identity. If a female professional experiences discrimination at her workplace because of her identity, it can be a critical impetus for self-employment [8,9]. In this sense, we hypothesize that the attribute-based discrimination, as a push factor, will motivate female professionals to consider self-employment.

On the other hand, the family-related aspects can be sources for discrimination [12,29]. For example, for married professionals the adverse treatment by the organization can occur because the organization perceives that married professionals may not spend as much time on their jobs as single professionals [12]. Childcare is another example of family-related discrimination [35]. Since childcare is costly in terms of time spending, having a child can cause a certain type of discrimination. To reconcile this work–life balance, female professionals will be likely to consider self-employment with the increase in family-related discrimination. Based on the two aspects of discrimination related to the push theory, we have two hypotheses:

Hypothesis 2a.

If a female professional experiences attribute-based discrimination at prior organizations, she will turn to self-employment participation faster.

Hypothesis 2b.

If a female professional experiences family-based discrimination at prior organizations, she will turn to self-employment participation faster.

While many studies treat the theories of pull and push for entrepreneurship distinctively and independently, the factors for pull and push should be simultaneously considered [27]. Dawson and Henley [27] acknowledged that the push and pull factors are convoluted in practice and those boundaries may be blurred. In particular, there might an endogenous process across the pull and push dimensions. That is, the pull factors can influence the push factors and vice versa.

For female professionals, when we consider the pull factor instantiated by prior salary level, and the push factors, i.e., attribute-based and family-based discrimination, we may have a specified mechanism for self-employment. The female professionals with higher salaries may feel less attribute-based discrimination because the higher level of salary means that they are likely to hold critical roles in the organization. Because of this, the female professionals will be more motivated to remain at the traditional organizations to enjoy their competitive advantages over other professionals. As a result, even though female professionals may experience either type of discrimination (i.e., either attribute-based or family-based discrimination) at their workplaces, if they have earned a higher salary level, they will not be motivated toward self-employment, preferring to stay at the traditional organization. From this standpoint, we hypothesize:

Hypothesis 3a.

When a female professional with a higher level of prior salary experiences attribute-based discrimination, she will be less likely to choose self-employment participation.

Hypothesis 3b.

When a female professional with a higher level of prior salary experiences family-based discrimination, she will be less likely to choose self-employment participation.

3. Methods

3.1. Data and Sample Population

To test the hypotheses, we considered the career development process of female lawyers in the U.S. and collected the relevant data from the After the JD project. The After the JD is a longitudinal study of lawyers across 18 legal markets in the U.S., administered and published by the American Bar Foundation. Wave I of After the JD was launched in May 2002 and contains a sample of new lawyers from 18 legal markets—ranging from the four largest markets (New York City, District of Columbia, Chicago, and Los Angeles) to 14 other areas ranging from small metropolitan areas to entire states; Wave II was conducted from 2007–2008 as a follow up of Wave I [36]. The database captures the data of more than 5000 lawyers during their first 10 years after law school. Specifically, the After the JD data have six subcomponents, based on Wave II: (1) current professional employment, (2) professional employment history, (3) current employment and career transition, (4) training, education, and debt, (5) social, political, and community participation, and (6) other background information. Hence, this database was appropriate for our study in testing the timing of self-employment by tracing career movement within the profession and exits from the profession. Further, the database provided each individual lawyer’s employment status, including when he or she started working for their current employer, the organizational type, and their position within the organization as well as demographical attributes.

The sampling procedure of the After the JD study consisted of two stages. First, selection was within metropolitan areas to obtain wide geographic and population distributions of geographic areas. Next, individuals who met the eligibility criteria within the chosen geographical areas were selected [36]. In the first stage, the nation was divided into 18 strata by region and size of the new lawyer population. Within each stratum, one primary sampling unit (PSU)—consisting of either a metropolitan area, a portion of a state outside large metropolitan areas, or an entire state—was selected. The PSUs included all four major markets (Chicago, Los Angeles, New York, and Washington, DC) (i.e., those with more than 2000 new lawyers), five of the nine large markets (i.e., those with between 750 and 2000 lawyers), and nine of the remaining smaller markets. Through the two waves of surveys between 2003 and 2006, individual information and career history were traced and collected from 5353 lawyers in the U.S. After removing cases with missing or inconsistent responses from the pool of female lawyers, we had a total sample of 1356 female lawyers.

3.2. Estimation Model

To examine how female lawyers’ self-employment is motivated, we focused on the historical event where the given female lawyer leaves a traditional organization (i.e., law firm) and becomes a solo practitioner (or self-employed). This study operationally assumed that the emergence of self-employment is commonly followed by a Poisson process with Equation (1):

$${\mathsf{\lambda }}_t=\underset{\Delta t\to 0}{\lim }\frac{P(t<T\langle t+\Delta t|T\ge t)}{\Delta t}.$$

(1)

Based on this, we adopted a Cox proportional hazards model to estimate the self-employment rates of female lawyers. The Cox proportional hazards model is useful to explicitly to test time-dependent events [37]. Also, compared to the parametric hazards models, which explicitly require a specified baseline model, the Cox proportional hazards models implicitly assume a base model. Thus, the latter is more flexible to build estimation models, and also helps prevent making arbitrary assumptions about the baseline model. The estimation model is seen as follows:

$${\lambda }_{it}={\lambda }_{0t}\exp \left(X_{it-1}\tilde{\gamma }+{\beta }_{1}PUL{L}_{it-1}+{\beta }_{2}PUS{H}_{it-1}\right).$$

(2)

Based on this, we adopted a Cox proportional hazards model to estimate the self-employment where X_it−1 is a matrix of control variables in region i at time t−1; PULL_it−1 and PUSH_it−1 denote pull and push factors for female lawyer i at time t−1, respectively; $\tilde{\gamma }$ is a vector of parameters for control variables and β₁ and β₂ indicate the parameters for the predictors; λ₀ corresponds to the baseline hazard function, which is left unspecified. To control the possible reverse causality, as indicated in the equation, we fully utilized the two waves of the After the JD survey. For the left hand side of Equation (2), i.e., the hazard rate of self-employment, we used the second wave of the survey, conducted in 2006. Meanwhile, for the right hand side of the equation, i.e., independent variables and control variables, we used the Wave I survey data, collected in 2003. By separating the data for independent variables from those for the dependent variable, we attempted to avoid the cases where the dependent variable (the state of self-employment) could influence the independent variables (i.e., pull and push factors).

In such a semi-parametric event history model, the estimated hazard ratios are assumed to be proportional over time. To confirm this assumption, the Grambsch and Therneau test was performed and we found that the Schönfeld residuals were not significant for the full model. This indicated that our estimation model held the assumption that the effect parameters multiply the hazards [38]. Figure 1 validates this proportionality.

3.3. Measures

3.3.1. Dependent Variable

As we were seeking to understand the job transition toward self-employment of female lawyers, the dependent variable was the rate of female lawyers’ move to self-employment from traditional law firms. From the After the JD database, we obtained data about the female lawyer’s job transition year and type of organization. Self-employment was operationally defined as the point in time when the given female lawyer engaged as a solo practitioner or started her own legal office. We coded the year in which the female lawyer opened her own office after leaving a law firm or other organization. This indicated whether and when the event of entrepreneurship was reached. Time was calculated from the year the female lawyer earned her JD. The female lawyers who did not have a catalyzing event were right-censored at the end of the observation period (i.e., the year 2006). Over the observation period, 8.6% of the female lawyers became solo practitioners or self-employed. On average, it took the female lawyers 10.56 years to reach self-employment (s.d. = 2.81).

3.3.2. Independent Variables

As a pull factor, we considered the personal earnings that the given female lawyer made in the past, for women with high skill levels can earn more money in a traditional salary-and-wage employment [17]. Likewise, given that lawyers contract their annual salaries based on their achievement or work experience, the level of salary can represent the extent to which the lawyers accomplish their tasks or effectively perform legal practices. In this sense, operationally, the prior salary was defined as the maximum dollar amount of pay for work that the given lawyer earned from prior organizations. To measure the variable, we first collected the all the salary information of each female lawyer and found the maximum level of salary she made in her employment history.

As an independent variable representing the push factor, attribute-based discrimination was operationally defined as the situations that happened in prior workplaces by virtue of the respondents’ attributes. In this study, attribute-based discrimination was calculated in two ways. First, we used a four-item scale assessing whether the given respondent had ever “experienced demeaning comments or other types of harassment,” “missed out on a desirable assignment,” “had a client request someone other than you to handle a matter,” and “experienced one or more other forms of discrimination.” Second, we collected the contingencies where the given female lawyer left her former job because of discrimination. We traced the career history of each female lawyer and coded 1 if the given female lawyer answered that she left her former job due to discrimination. Then, we summed up all the binary codes from the two approaches to measure discrimination to construct a variable of attribute-based discrimination. This variable ranged from 0 to 7.

Another independent variable representing the push factor was family-based discrimination. Family-based discrimination is related to household responsibilities and mostly women tend to take those responsibilities [29,39,40]. In fact, previous studies showed that marriage encourages women rather than men to consider self-employment [12,29,41]. However, as Patrick et al. [12] pointed out, childcare is much more related to the self-employment of women because there is a “higher burden placed on women’s time as a care giver” (p. 369). Given that the pressure for childcare can “push” women toward self-employment [12], the discrimination from organizations caused by the childcare can be another push factor, called family-based discrimination. The family-based discrimination was, in this study, operationally defined as the situations where the given lawyer experienced any adverse consequences at work as a result of having a child. From the database of After the JD, we coded 1 if the given female lawyer experienced unfair treatment or discrimination due to having a child. There were 11 questions related to the family-based discrimination: (1) delay in promotion, (2) loss of seniority, (3) questioning of commitment to work, (4) loss of office space, (5) pressure to work while on parental leave, (6) difficulty in obtaining flexible hours or part time work, (7) loss of clients, (8) unreasonable work load following parental leave, (9) loss of challenging assignments, (10) loss of opportunities to travel, and (11) loss of income. We summed up the 11 binary codes and constructed the variable of family-based discrimination.

3.3.3. Control Variables

In predictions of the rate of female lawyers’ self-employment from traditional law firms, we controlled for (a) race, (b) number of turnovers, (c) age, (d) marital status, (e) family influence, (f) legal practice experience, and (g) employment by law firms. To control for demographic attributes, race, age, and marital status were collected and included in the model. Race and age were used as the respondents answered. For age and number of turnovers, we took the logarithm of these variables, as the distributions of these variables were skewed. The marital status was measured as a binary code indicating whether the given female lawyer was in a couple or not. The survey item asked the lawyers to choose their marital status among: (1) never married or never in a domestic partnership, (2) married first time, (3) remarried after divorce, annulment, or being widowed, (4) domestic partnership, (5) divorced or separated, (6) widowed, and (7) other. Given that family characteristics are crucial for the decision on the self-employment participation of women [29], we made a dummy variable to indicate whether the given female lawyer had a spouse or domestic partner. Accordingly, we coded 1 if the given lawyer answered her marital status as “married first time,” “remarried after divorce, annulment, or being widowed,” or “domestic partnership.”

Family influence, as another family characteristic, was considered as a (potential) antecedent from family members for legal practice. If family members, especially parents or grandparents, were lawyers, they could share their client bases or social networks as well as constructive advice for career development, and this would facilitate the self-employment. We counted the number of family members (i.e., parents and grandparents) who were lawyers. This variable ranged from 0 to 3. Legal practice experience indicated the time duration in which the given lawyer worked as a legal professional. The variable was measured as the time difference between the year of bar admission and the year 2006, the year for the second wave of After the JD. The mean legal practice experience was 7.4 (s.d. 0.58) and its values ranged from 5 to 9.

Gender consciousness referred to “recognition of the power imbalance based on gender that pervade women’s lives” [42] (p. 884). We assumed that if a given female lawyer participated in a gender-specified organization, the lawyer would be conscious of gender identity. Thus, in the context of legal professions, gender consciousness was operationally defined as the extent to which a given female lawyer had participated in activities of gender-based organizations during law school. In the survey of After the JD, female lawyers were asked as to whether or not they participated in any gender-based organization during law school. Based on the answers from the question of After the JD, we coded 1 if the given female lawyer was not a member; 2 if the lawyer was a member; 3 if the lawyer was an active participant/officer in 2003; 4 if the lawyer had been an active participant/officer before 2003. The variable, as an interval scale, ranged from 1 to 4.

Lastly, we considered the probability that the given female lawyer had worked at the traditional law firms. Since the pull and push factors assume that female lawyers build up their abilities or experience discrimination within their organization, affiliation in at least a law firm was essential. To consider the possibility where female lawyers directly participated in self-employment after earning the JD or went through unemployment, first, we created a dichotomous variable indicating whether the given female lawyer previously worked at a law firm. By using a logistic regression model, this dichotomous variable was estimated with respect to race, location, age, the ranking of the law school from which the given female lawyer graduated, and the number of other accredited or certificate degrees (such as Master of Laws (LLM), other Master’s degree, doctoral degree, MBA, and others). The estimated value, which indicated the probability that a given female lawyer had worked at a law firm, was used to control for the cases where female lawyers directly opened their legal offices without any work experiences at law firms.

Table 1. Descriptive statistics and correlations.

N=1356	Mean	SD	Min	Max	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1. Solo practitioners	0.09	0.28	0	1
2. ln Prior Salary	11.28	0.57	3.91	13.53	−0.15 ***
3. Attribute-based Discrimination	0.95	1.24	0	7	0.05 †	−0.03
4. Family-based Discrimination	0.94	1.80	0	11	0.02	0.02	0.20 ***
5. Race (Hispanic)	0.09	0.29	0	1	0.03	0.01	0.02	−0.01
6. Race (Native)	0.01	0.10	0	1	0.07 **	−0.05 †	0.03	0.08 **	−0.03
7. Race (Asian)	0.11	0.31	0	1	−0.06 *	0.10 ***	−0.04	−0.05 †	−0.11 ***	−0.04
8. Race (White)	0.65	0.48	0	1	0.00	−0.05 †	−0.08 **	0.05 *	−0.44 ***	−0.14 ***	−0.48 ***
9. Race (Others)	0.02	0.13	0	1	0.00	0.00	0.02	0.03	−0.04	−0.01	−0.05 †	−0.19 ***
10. ln # Turnovers	−0.71	1.43	−2.30	1.63	0.04	0.14 ***	0.05 †	0.04	0.00	0.05 †	−0.03	0.02	0.04
11. ln Age	3.57	0.13	3.37	4.19	0.14 ***	−0.09 ***	0.03	0.00	−0.05 †	−0.01	−0.09 **	0.06 *	0.00	−0.05 †
12. Marital Status (Married)	0.53	0.50	0	1	0.04	-0.01	0.01	0.27 ***	−0.05 †	0.05 †	−0.05 †	0.15 ***	0.00	−0.01	0.07 *
13. Family Influence	0.15	0.41	0	3	−0.03	0.04	−0.03	0.04	−0.06 *	0.00	−0.08 **	0.16 ***	−0.04	0.02	-0.04	0.02
14. Legal Practice Experience	7.40	0.58	5	9	0.05 †	0.01	0.06 *	−0.03	0.02	−0.01	0.03	−0.10 ***	0.02	−0.03	0.19 ***	−0.04	−0.03
15. Employment by law firms	0.93	0.27	0.00	5.27	0.07 **	−0.15 ***	0.05 †	−0.03	0.16 ***	0.00	−0.12 ***	−0.15 ***	−0.04	−0.08 **	0.29 ***	0.01	0.00	0.03

^†p < 0.1 * p < 0.01, ** p < 0.05, *** p < 0.001; two-tailed tests.

presents the descriptive statistics and correlations for the analyses predicting the rate of female lawyers’ self-employment.

4. Results

4.1. The Effects of Pull and Push Factors on Self-Employment of Female Lawyers

Table 2. Cox analysis of the hazard rate of female lawyers’ self-employment.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
Race (Hispanic)	0.409	0.455	0.448	0.411	0.487	0.474	0.468
	(0.316)	(0.303)	(0.314)	(0.319)	(0.305)	(0.310)	(0.299)
Race (Native)	1.271 *	1.178 *	1.274 *	1.250 *	1.147 *	1.068 †	1.021 †
	(0.533)	(0.523)	(0.525)	(0.546)	(0.541)	(0.562)	(0.549)
Race (Asian)	−0.530	−0.435	−0.467	−0.523	−0.394	−0.510	−0.399
	(0.372)	(0.393)	(0.376)	(0.374)	(0.396)	(0.437)	(0.386)
Race (White)	0.0945	0.0731	0.150	0.0970	0.114	0.0793	0.0810
	(0.237)	(0.234)	(0.239)	(0.238)	(0.235)	(0.238)	(0.226)
Race (Others)	−0.0473	−0.101	0.0180	−0.0519	−0.0331	−0.0437	−0.369
	(0.744)	(0.771)	(0.760)	(0.735)	(0.773)	(0.768)	(0.887)
ln No. Turnovers	0.0888	0.134 *	0.0831	0.0872	0.129 †	0.135 *	0.137 *
	(0.0634)	(0.0681)	(0.0647)	(0.0643)	(0.0695)	(0.0677)	(0.0684)
ln Age	2.621 ***	2.567 ***	2.599 ***	2.628 ***	2.542 ***	2.495 ***	2.385 ***
	(0.455)	(0.417)	(0.443)	(0.455)	(0.409)	(0.419)	(0.418)
Marital Status (Married)	0.192	0.231	0.182	0.175	0.220	0.223	0.205
	(0.161)	(0.170)	(0.160)	(0.169)	(0.178)	(0.178)	(0.182)
Family Influence	−0.283	−0.223	−0.270	−0.289	−0.211	−0.208	−0.196
	(0.254)	(0.251)	(0.253)	(0.247)	(0.242)	(0.241)	(0.239)
Legal Practice Experience	0.171	0.199	0.149	0.170	0.175	0.162	0.150
	(0.183)	(0.180)	(0.184)	(0.183)	(0.181)	(0.187)	(0.170)
ln Prior Salary		−0.546 ***			−0.552 ***	−0.506 ***	−0.457 ***
		(0.126)			(0.126)	(0.136)	(0.125)
Attribute-based Discrimination			0.154 *		0.149 *	2.130 †	0.140 *
			(0.0665)		(0.0652)	(1.106)	(0.0690)
Family-based Discrimination				0.0200	0.00353	0.00315	2.592 ***
				(0.0469)	(0.0497)	(0.0504)	(0.590)
ln Prior Salary *						−0.180 †
Attribute-based Discrimination						(0.102)
ln Prior Salary *							−0.236 ***
Family-based Discrimination							(0.0549)
Employment by Law Firms	0.315 †	0.307 †	0.329†	0.320 †	0.320 †	0.291	0.270
	(0.177)	(0.181)	(0.176)	(0.175)	(0.179)	(0.177)	(0.176)
Log-likelihood	−820.883	−810.440	−819.135	−820.800	−808.753	−807.597	−803.988
AIC	1663.765	1644.879	1662.27	1665.599	1645.505	1645.195	1637.977
Deviance (χ2)	-	21.7 ***	3.74 †	0.18	25.4 ***	30.9 ***	36.86 ***

No. of female lawyers: 1356, No. of events: 117, State-based clustered standard errors in parentheses. ^† p < 0.1 * p < 0.01 *** p < 0.001; two-tailed tests.

shows the results of the Cox proportional hazard models estimating the hazard rate of female lawyers’ self-employment with respect to pull and push factors (i.e., prior salary, attribute-based discrimination, and family-based discrimination) and control variables (i.e., race, turnovers, age, marital status, gender consciousness, family influence, legal professional experience, and employment by law firms).

Model 1 of Table 2 included only control variables. In Models 2 through 4, the effects of prior salary (as a pull factor) and attribute-based and family-based discrimination on the rate of self-employment were respectively tested. In Model 5 as a comprehensive model we found two things. From the pull perspective, the higher the prior salary, the less likely the self-employment participation (β = −0.559; p < 0.001). From the push perspective, the more likely the female lawyer has experienced attribute discrimination, the greater the rate of self-employment (β = 0.156; p < 0.05). Meanwhile, for the family-based discrimination, there was no significant effect on self-employment.

In addition, we considered the effects of pull and push factors on self-employment simultaneously. Models 6 and 7 present the interaction effects of pull and push factors on the rate of becoming self-employed. In Model 6, we found that the interaction effect between prior salary and attribute-based discrimination was negatively related to the rate of self-employment (β = −0.198; p < 0.10). Likewise, as seen in Model 7, there was a negative interaction effect between prior salary and family-based discrimination on the rate of self-employment (β = −0.269; p < 0.001). Those results revealed that if a given female lawyer with a high salary (indicating high skills or competence) experienced any attribute-based or family-based discrimination, she would likely decide to stay with the law firm and be less likely to become self-employed.

Figure 2 shows how female lawyers become self-employed, depending on the interaction between prior salary and attribute-based and family-based discrimination. In the lower range of prior salary, the female lawyers who experienced attribute-based or family-based discrimination were more likely than others to consider self-employment.

4.2. Robustness Checks

To further understand the situations where women lawyers encountered both push and pull factors, we split the sample into two groups at the median of prior salary. Then we ran the estimation models with these split samples. Since the pull factor was used for the sample split, the estimation models included only the main effects of the push factors (i.e., attribute-based and family-base discrimination). The estimation results, as seen in Models 4 and 8 in

Table 3. Pull-based sample split models.

	High Prior Salary				Low Prior Salary
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Race (Hispanic)	0.0786	0.103	0.0839	0.128	0.595	0.602	0.614	0.608
	(0.718)	(0.693)	(0.723)	(0.688)	(0.491)	(0.501)	(0.521)	(0.522)
Race (Native)	1.061	1.083	1.108	1.154	1.308 †	1.259 †	1.190	1.156
	(1.059)	(1.050)	(0.973)	(0.966)	(0.769)	(0.731)	(0.776)	(0.752)
Race (Asian)	−0.964	−0.931	−0.987	−0.931	−0.112	−0.0983	−0.0732	−0.0736
	(0.609)	(0.604)	(0.625)	(0.625)	(0.554)	(0.543)	(0.565)	(0.551)
Race (White)	−0.460	−0.427	−0.489	−0.432	0.330	0.344	0.333	0.341
	(0.489)	(0.476)	(0.487)	(0.474)	(0.363)	(0.362)	(0.379)	(0.374)
Race (Others)	0.167	0.188	0.147	0.177	−0.235	−0.166	−0.284	−0.205
	(1.141)	(1.127)	(1.162)	(1.152)	(1.141)	(1.148)	(1.096)	(1.111)
ln No. Turnovers	0.147	0.145	0.153	0.152	0.106	0.0990	0.101	0.0969
	(0.101)	(0.0983)	(0.103)	(0.102)	(0.0692)	(0.0718)	(0.0696)	(0.0712)
ln Age	1.754	1.738	1.756	1.720	2.714 ***	2.695 ***	2.743 ***	2.720 ***
	(1.325)	(1.303)	(1.309)	(1.276)	(0.467)	(0.465)	(0.467)	(0.464)
Marital Status (Married)	0.399	0.397	0.529	0.538	0.124	0.120	0.0789	0.0843
	(0.339)	(0.339)	(0.325)	(0.327)	(0.185)	(0.184)	(0.192)	(0.192)
Family Influence	−0.364	−0.357	−0.317	−0.301	−0.162	−0.156	−0.171	−0.163
	(0.617)	(0.624)	(0.596)	(0.604)	(0.280)	(0.278)	(0.272)	(0.272)
Legal Practice Experience	0.247	0.241	0.215	0.203	0.177	0.145	0.163	0.140
	(0.326)	(0.321)	(0.337)	(0.329)	(0.185)	(0.187)	(0.184)	(0.186)
Attribute-based Discrimination		0.0498		0.0959		0.152 *		0.135 *
		(0.149)		(0.154)		(0.0627)		(0.0686)
Family-based Discrimination			−0.134	−0.145			0.0621	0.0460
			(0.106)	(0.102)			(0.0484)	(0.0515)
Employment by Law Firms	0.350	0.350	0.341	0.341	0.195	0.218	0.209	0.227
	(0.311)	(0.315)	(0.306)	(0.312)	(0.238)	(0.238)	(0.232)	(0.233)
Log-likelihood	−174.640	−179.934	−174.022	−173.889	−569.484	−568.134	−568.787	−567.769
AIC	371.280	373.203	372.043	373.778	1160.967	1160.269	1161.575	1161.537
Deviance (χ2)	-	21.7 ***	3.74 †	0.18	-	30.9 ***	36.86 ***
No. Female lawyers	628	628	628	628	728	728	728	728
No. Events	28	28	28	28	89	89	89	89

State-based clustered standard errors in parentheses. ^† p < 0.1 * p < 0.01 *** p < 0.001; two-tailed tests.

, indicated that women lawyers who had a relatively low prior salary were likely to be self-employed when they had experienced attribute-based discrimination (β = 0.135; p < 0.05) while those who had a relatively high salary were not likely to be influenced by the push factors. This suggests that push factors may not always affect the choice of women lawyers between salary employment and self-employment when their abilities are acknowledged by the organization.

To check if these findings were common in male lawyers, we sampled a population of male lawyers from the After the JD database. We collected data for 1590 male lawyers, showing an overall 9.4% self-employment rate. With the sample, we performed the same Cox analyses. As seen in

Table 4. Cox analysis of the hazard rate of male lawyers’ self-employment.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
Race (Hispanic)	−0.0873	−0.0692	−0.0599	−0.0833	−0.0562	−0.0384	−0.0616
	(0.460)	(0.470)	(0.492)	(0.451)	(0.469)	(0.492)	(0.475)
Race (Native)	0.226	0.0793	0.242	0.0503	−0.406	−0.260	−0.693
	(0.719)	(0.620)	(0.733)	(0.680)	(0.672)	(0.838)	(1.118)
Race (Asian)	−0.902 *	−0.811 †	−0.874 *	−0.887 *	−0.798 †	−0.785 †	−0.809 †
	(0.418)	(0.433)	(0.442)	(0.424)	(0.453)	(0.474)	(0.455)
Race (White)	−0.298	−0.316	−0.256	−0.312	−0.305	−0.287	−0.310
	(0.303)	(0.309)	(0.339)	(0.296)	(0.320)	(0.341)	(0.322)
Race (Others)	0.257	0.263	0.298	0.247	0.290	0.304	0.288
	(0.402)	(0.374)	(0.412)	(0.407)	(0.389)	(0.387)	(0.389)
ln No. Turnovers	0.00522	0.0254	0.00477	−0.00948	0.0111	0.0123	0.0103
	(0.0495)	(0.0475)	(0.0491)	(0.0510)	(0.0483)	(0.0491)	(0.0483)
ln Age	2.215 ***	1.906 ***	2.180 ***	2.314 ***	2.025 ***	2.001 ***	2.061 ***
	(0.463)	(0.477)	(0.468)	(0.459)	(0.462)	(0.449)	(0.487)
Marital Status (Married)	−0.316	−0.284	−0.311	−0.394 *	−0.355 †	−0.356 †	−0.356 †
	(0.198)	(0.198)	(0.199)	(0.182)	(0.188)	(0.190)	(0.190)
Family Influence	0.0194	0.0307	0.0242	0.0212	0.0423	0.0404	0.0449
	(0.180)	(0.173)	(0.180)	(0.174)	(0.167)	(0.166)	(0.166)
Legal Practice Experience	0.131	0.166	0.126	0.130	0.149	0.153	0.140
	(0.111)	(0.107)	(0.108)	(0.110)	(0.107)	(0.109)	(0.107)
ln Prior Salary		−0.710 ***			−0.671 ***	−0.705 ***	−0.629 ***
		(0.130)			(0.124)	(0.137)	(0.152)
Attribute-based Discrimination			0.0677		0.0513	−0.930	0.0444
			(0.0990)		(0.0793)	(2.145)	(0.0814)
Family-based Discrimination				0.229 ***	0.199 ***	0.206 ***	0.646
				(0.0533)	(0.0532)	(0.0506)	(0.963)
ln Prior Salary *						0.0873
Attribute−based Discrimination						(0.193)
ln Prior Salary *							−0.0408
Family−based Discrimination							(0.0862)
Employment by Law Firms	−0.0390	−0.274	−0.0386	−0.0216	−0.244	−0.248	−0.232
	(0.196)	(0.215)	(0.195)	(0.195)	(0.211)	(0.209)	(0.206)
Log-likelihood	−1087.857	−1076.242	−1087.678	−1083.476	−1072.781	−1072.66	−1072.608
AIC	2197.714	2176.485	2199.357	2190.952	2173.561	2175.32	2175.216
Deviance (χ2)	−	21.7 ***	3.74 †	0.18	25.4 ***	30.9 ***	36.86 ***

No. of male lawyers: 1590, No. of events: 150, State-based clustered standard errors in parentheses. ^† p < 0.1 * p < 0.01 *** p < 0.001; two-tailed tests.

, we found a negative effect of prior salary on self-employment (β = −0.734; p < 0.001), which was consistent with the result from the female lawyer sample. Yet, there was no significant effect of attribute-based discrimination but rather a significant positive effect of family-based discrimination on self-employment (β = −0.233; p < 0.001). Furthermore, there were no significant interaction effects of prior salary and any types of discrimination for male lawyers. These findings revealed that the pull factor can commonly affect self-employment between male and female lawyers, but the attribute-based discrimination is a stronger factor for female lawyers than male lawyers.

5. Discussion

5.1. Push, Pull, and Self-Employment

When professionals choose self-employment, their core jobs are not different from what they do during their salary employment period. The major difference between the two is that when professionals are attached to traditional organizations, their work is assigned to them at the enterprise level, but this is not true for professional entrepreneurs. In terms of securing new clients, professional entrepreneurs try to acquire clients for themselves. However, if professionals are already high potential individuals who can be assigned to desirable clients, they do not feel too much pressure about acquiring clients by choosing self-employment. Consequently, pull factors are antecedents to self-employment for non-professionals, but it is highly possible that they are not motives to pursue entrepreneurship for professionals.

Meanwhile, professionals have professional qualifications. Thus, we predict that push factors barely exist in professional organizations compared to non-professionals. However, because push factors and negative situational factors can be generated by race, career, personality, etc., they still exist in professional organizations. In light of the above discussion, we analyzed how pull and push factors in professional organizations induce individuals to choose self-employment.

Even though the pull and push factors have distinct dimensions, individuals make a choice of self-employment based on both factors [27]. For example, a professional woman who is married feels discrimination such as exclusion from promising projects because it is surmised that unmarried workers have more time to invest their work [12]. In this sense, we illustrated the interaction effects of push factors and pull factors on the rate of self-employment. Acknowledging that push factors tend to make minorities choose self-employment while pull factors are likely to motivate the majority group to become entrepreneurs, we noted the fact that female professionals can simultaneously experience both push and pull factors.

5.2. Self-Employment under the Glass Ceiling

One of our findings, which revealed that attribute-based discrimination demotivates the self-employment of female lawyers, is further understood with the term of glass ceiling. A glass ceiling is understood as a “systemic discrimination in which organizational policies and practices disproportionately and negatively impact women and people of color” [43] (p. 463). This indicates that discrimination is structural in nature [44,45,46], rather than based on deficiencies on the parts of certain genders or a lack of qualities necessary for leadership [45,46]. Even in a field where specific job-relevant skills should be deemed most important, such as in the case of a medical or legal professional, one’s professional identity may be less salient in advancement decisions than one’s gender identity.

In a similar vein, drawing from institutional theory, we acknowledged that differential treatment that results from glass ceiling effects is not based on overt discrimination by the majority groups [45,47], but instead is based on institutionalized rules that maintain organizational stability over time [48,49]. This stability is maintained by bylaws, policies, and management hierarchies that are undertaken by the majority groups in positions of power and dominance [43]. As such, maintaining organizational stability within an organization implicitly can be a structural barrier to mobility for women.

When female professionals, including female lawyers, experience such systemic discrimination, they may feel helpless even though the accreditation, such as the JD, can secure their career (e.g., [50]). Since the glass ceiling is institutionalized throughout the social structure, entrepreneurship (or self-employment) cannot help individuals escape from the “hidden” system of discrimination [43]. This helplessness will lead them to stay at traditional organizations even though they are aware of the presence of the glass ceiling.

6. Conclusions

This study shows how female lawyers can be motivated to engage in self-employment. Overall, our findings reveal that for female lawyers, prior salary is negatively related to the rate of the self-employment, but that attribute-based discrimination accelerates the decision to be self-employed. In addition, we found that gender consciousness negatively influences the rate of self-employment, driven by either prior salary or attribute-based discrimination. The negative moderation of gender consciousness for self-employment indicates that if female lawyers take their gender identity seriously into account for their career choices, they may be reluctant to exit traditional law firms and enter entrepreneurship in the form of solo practitioners.

This study provides two theoretical implications. First, we consider the glass ceiling effects in self-employment. That is, excessive strain through discrimination and gender consciousness causes helplessness, and those who feel helpless are less likely to search for alternatives such as entrepreneurship. Second, through this study, we can understand how individuals manage their multiple identities. This study illuminates that individuals may activate one social identity over another. This creates an interesting dynamic for female professionals who may face glass ceiling effects in their fields, where female lawyers must manage and integrate their ambivalent identities based on a number of different social and individual factors. In this sense, gender consciousness becomes one of the most salient factors.

For the future research, we can further elaborate how to manage the ambivalence and multiple identities in different contexts. This suggests that what the pull and push factors mean especially for female professionals is important to whether professional women consider the attribution of factors the result of gender or not. For example, there is a professional woman who is a key person in professional firms, but she attributes push factors she experienced in firms to gender. Then, does she choose self-employment or not? It implies two possibilities. First, professional women who attribute push factors to gender recognize that they have a low social status. Second, such recognition may affect the pull or push factors. In terms of pull factors, if a professional woman attributes her capability to gender, she may think her capability is specially differentiated from that of men. Thereupon, we need to specify how strongly the recognition of gender affects the pull factors and self-employment. In contrast, in terms of push factors, professional women judge negative situational factors generated by their gender. Because gender is an innate property, female professionals may recognize that they cannot completely escape from those factors even though they choose to be self-employed. Thus, it is important to examine how individuals define themselves in terms of their professional and gender identity associations and how individuals balance and integrate multiple identities in order to navigate barriers that may prevent them from professional advancement into higher levels of an organization. Specifically, the consequences of identity integration can be explored.