FLSEVIER

SELF-EMPLOYMENT ENTRY ACROSS INDUSTRY GROUPS

TIMOTHY BATES Wayne State University

Highly educated individuals possessing a net worth exceeding $100DO0 are the most likely to enter self-employment. Recent studies have linked personal wealth holdings positively to self-employment entry, but educational attainment has been a weak and erratic determinant of self-employment status. Although barriers such as financial capital constraints clearly shape self-employment entry decisions, the nature of these barriers varies substantially across small business industry groups. The strong role often played by education and work experience in identifying self-employment entrants is often obscured by inappropriate aggregation across industries. Entry barriers shape not only the decision to enter self-employment but also the ~pes of industries that potential business owners are likely to enter. The likelihood of self-employment in skilled services increases greatly as level of education rises, whereas the opposite situation prevails in construction. The major barriers to self-employment entry in manufacturing and wholesaling are clearly financial in nature. In contrast, the primary route to self-employment in the growing skilled services industries is the attainment of advanced educational credentials; personal wealth holdings are secondary. Several factors perform consistently in identifying self-employment entry across industries. The likelihood of entry increases with age, peaking as people approach age 40 and then leveling out. African-Americans are less likely to pursue self-employment than nonminorities. Factors associated with self-employment entry among women are quite different than those that predict male entry, but this partially reflects the differing industry concentrations among men and women. Women are more likely to enter self-employment in skilled services fields. Thus, advanced education and work experience--the human capital variables---are the strongest predictors of female entrants.

EXECUTIVE SUMMARY

Address correspondence to Professor Timothy Bates, College of Urban, Labor, and Metropolitan Affairs, Wayne State University, Detroit, MI 48202. Research reported in this study was conducted at the U.S. Bureau of the Census, Center for Economic Studies. Findings expressed are those of the author and do not necessarily reflect the views of the U.S. Bureau of the Census. The author thanks Constance Dunham for her extensive contribution to this study. Journal of BusinessVenturing 10, 143-156 1995 Elsevier Science Inc. 655 Avenueof the America.s, New York, NY 10010 0883-9026/95/$9.50 SSD10883-9026(94)00018-P

144

T. BATES

Barriers to self-employment are most binding to those who possess neither the human capital nor the financial capital traits associated with entry. For others, barriers need not deny access to self-employment---cather, they may simply shape the industry choice. This study is based upon the 1984 Survey of lncome and Program Participation (SIPP), a longitudinal survey of households carried out by the U.S. Bureau of the Census. Entry into self-employment was observed over a 32-month period for a sample of 24,428adults.

INTRODUCTION Reynolds and Miller suggest that roughly one in 30 of all adults active in the U.S. labor force "appear to be involved in a new finn start-up at any point in time" (p. 405, 1992). Many of these potential finns never emerge from the gestation stage, ending up as stillborns. Public and private efforts to facilitate new firm formation could target their assistance to potential start-ups more effectively if the traits predicting actual self-employment entry were better understood (Birley 1986). Recent studies have argued that financial capital constraints act as effective barriers to self-employment entry (Acs and Audretsch 1989; Evans and Jovanovic 1989). Reynolds and Miller (1992) clearly concur with these findings: they designate obtaining outside financial support as one of the significant gestation markers related to actual emergence of firms from the start-up planning process. Owner human capital resources, in contrast, have not been emphasized as essential to the process of new finn emergence. Measures such as owner educational background and work experience have been weak and erratic determinants of self-employment status in empirical studies (Evans and Leighton 1989). This study demonstrates that the role played by education and work experience in identifying self-employment entry is, indeed, strong, but it is often obscured in empirical studies by inappropriate aggregation across industries. The practice of grouping entrepreneurs into overly broad classifications, without regard to industry-specific conditions, although quite common in empirical research, "only increases the imprecision of the research results and interpretations generated" (Carsrud, Olm, and Eddy 1986). Empirical research in this study is guided by Dunkelberg and Cooper's hypothesis (1982) that there are differing paths to entrepreneurship, depending upon owner background, education, and the industrial sector in which the owner chooses to operate. Rather than examining central tendencies of entrepreneurship, this study investigates diversity across major industry groups. Statistics describing self-employment entrants generally mask wide differences in alternative paths to business ownership. The finding that self-employment entry varies significantly across industry groups complements the Cooper, Woo, and Dunkelberg demonstration (1989) that firms of different initial size--larger versus smaller--are associated with peculiar entrepreneurial characteristics. Entry into self-employment, then, is viewed as a process shaped by the individual characteristics and resources of potential entrepreneurs, interacting with industry-specific factors; personal traits associated with entry into self-employment vary sharply across industry subgroups. This study utilizes longitudinal data from the Survey of Income and Program Participation (SIPP) to examine self-employment entry for individuals aged 21 to 60. We first explore self-employment entry generally, but the main focus of this study is upon entry into three specific industry groupings: (1) skill-intensive service fields, (2) construction, and (3) large-scale goods industries--manufacture and wholesale. Consistent with previous studies (Evans and Jovanovic 1989; Evans and Leighton 1989), we find that individuals with high net assets are--across the board--highly likely to enter self-employment. Although barriers such as capital constraints clearly shape self-

SELF-EMPLOYMENT ENTRY ACROSS INDUSTRIES 145 employment entry decisions, the nature of these barriers varies substantially across small business sectors. Characteristics associated with construction entry, in particular, have little in common with the traits that identify people choosing self-employment in skill-intensive services. Further, when traits such as net asset holdings and educational credentials are controlled for, women are much less likely to become self-employed than men, and African-Americans and Latinos are less likely than nonminorities to enter self-employment. The relevance of race and gender factors, however, varies substantially across industry subsectors. Thus, women and minorities not only exhibit lower rates of self-employment than nonminority males, but their distribution across industry subgroups also differs substantially. CHARACTERISTICS OF INDIVIDUALS ENTERING SELF-EMPLOYMENT Theoretical economists portray individuals who possess varying levels of entrepreneurial ability as choosing between employee and self-employment status. Lucas suggests that people entering self-employment possess relatively more entrepreneurial ability than those choosing to be employees; ability (business acumen) is a known parameter in his model (1978). A troubling aspect of the Lucas approach is its limited insight into causes of the huge self-employment rate differentials that exist across the different gender and racial subgroups. Do the low self-employment rates typifying women and African-Americans reflect their relative lack of business acumen? In their extension of the Lucas self-employment entry model, Evans and Jovanovic conclude that possession of financial capital is vitally important: "liquidity constraints tend to exclude those with insufficient funds at their disposal" (1989, p. 808). Presence of financial capital barriers to small business entry would certainly explain the lower self-employment rates observed among African-Americans. Acs and Audretsch (1989), in their study of small firm entry, found that a higher capital to labor ratio was associated with a lower rate of entry among 247 manufacturing four-digit SIC industries. Finally, White (1982) found that the existence of small firms in manufacturing was negatively related to capital intensity. Empirical studies of self-employment entry are rare. In their pioneering time series analysis of self-employment entry among nonminority men, Evans and Leighton (1989) found a strong, positive association between greater asset holdings and self-employment entry. Further, they observed that the percentage of nonminority men who were self-employed increased with age until the early 40s and then remained constant until retirement. Evans and Leighton (1989) investigated several human capital measures-education and work experience--but these variables behaved erratically as predictors of self-employment entry. Qualitative studies by sociologists suggest that married individuals may be more likely to enter self-employment (Waldinger 1986) because family members can serve as a source of cheap labor. Self-employment rate differentials among minority groups have been partially attributed to this factor. The higher self-employment rates among Asians, in particular, have been explained, in part, by their greater utilization of family help (Fratoe 1988). Although this study generally confirms the importance of personal wealth, age, and marital status as predictors of self-employment entry, we attribute major importance to the human capital traits of potential business entrants. The barriers to self-employment entry in specific industry sectors--manufacturing and wholesaling--are clearly financial assetspecific, but human capital traits--particularly advanced educational credentials--are key to

146

T. BATES

shaping the entry process into the large and rapidly growing skilled services area. Self-employment growth among women and minorities is disproportionately concentrated in the skilled services. The Evans and Leighton (1989) findings that human capital traits were poor predictors of self-employment entry reflect both their exclusive focus upon nonminority males and their implicit assumption that the entry process is essentially the same across industry groups. Using such a "one model fits all" for personal services entry, construction, professional services, and so forth, Evans and Leighton did not consider disaggregation of dissimilar industry groups.

ANALYSIS OF S E L F - E M P L O Y M E N T E N T R Y The data source for this analysis is the 1984 Survey of Income and Program Participation (SIPP), a longitudinal survey of households carried out by the U.S. Bureau of the Census. SIPP is one of the few household surveys that provides data on household net worth and detailed information on self-employment activity. In the 1984 SIPP panel, approximately 20,000 households were interviewed every four months between October 1983 and August 1986. The analysis presented here is based on the eight interviews that covered the panel period. The database analyzed in this article included the 24,428 individuals aged 21 through 60 who did not own a business during the first interview period, and for whom variable information was complete. The SIPP panel data are particularly well suited for examining the impact of personal wealth upon the likelihood of self-employment entry. A cross-section analysis of the factors underlying business ownership would not permit as clear an interpretation of the wealth variable. If a cross-section analysis of business entry were to show that wealth was positively related to business ownership, this could imply either that higher initial personal wealth facilitated new business entry, or that business ownership had, over time, generated greater wealth for the owner. Researchers have rejected the notion that self-employment entry is an act taking place at one point in time. A period of gestation or incubation, which may or may not culminate in continued pursuit of self-employment, occurs over a period of months or years (Birley 1984; Reynolds and Miller 1992). Quitting one's regular job and becoming self-employed is commonly n o t the first step in this incubation process. Individuals n o t defining their primary labor force status as self-employed are particularly important (and numerous) in explaining the rapid growth of small firms in skill-intensive service industries (Bates 1986). This study does not seek to identify a point of self-employment entry. Rather, traits of individuals in the gestation/entry phase of self-employment are identified and compared for selected industry groups. The criteria used to define (and identify) people in this stage of self-employment entry, furthermore, have a clear impact on the personal traits that emerge as important for identifying entrants. This study defines "entry" in several ways in order to identify entry determinants that are sensitive to definitional changes in who is/is not self-employed. The primary measure of self-employment entry used in this study includes all firms that were incorporated or that were operating above a minimum scale--specifically, with annualized net profits or losses of at least $1,000 during an interview period. An alternative measure of business entry included all firms in which the owner reported working in self-employment activity at least 10 hours per week during an interview period. The analysis of entry includes all individuals who did not own businesses in the first interview period. Within this sample, business entry took place if the individual reported business ownership (as defined here) during any of the subsequent seven interview periods (28 months).

SELF-EMPLOYMENT ENTRY ACROSS INDUSTRIES

147

The variables used in the econometric exercises to determine the characteristics that distinguish business entrants from non-entrants are defined below.

Dependent Variable
N E W B U S is equal to 1 for all those individuals who: (1) did not own a business during the first interview period and (2) reported a business during interviews 2 through 8 of the sample period. For all others (comprising all others who did not own a business in the first interview period), N E W B U S is set equal to 0. The definition of " b u s i n e s s " excludes casual businesses.

Explanatory Variables
Education
The education variables are defined as equal to 1 under the following conditions (and defined as equal to 0 otherwise): EDUC 1: those who never graduated from high school. EDUC2: those who are high school graduates. EDUC3: those with some college education. EDUC4: college graduates. EDUC5: those with at least some graduate education.

Experience
Number of years of work experience.

Household Net Worth

The wealth variables are defined as equal to 1 under the following conditions (and defined as 0 otherwise): W 1: household net worth W2: household net worth W3: household net worth W4: household net worth W5: household net worth under $10,000. of at least $10,000 and less than $25,000. of at least $25,000 and less than $50,000. of at least $50,000 and less than $100,000. of at least $100,000.

Age
To account for possible nonlinearities, the individual's age (AGE) and the square value o f the age (AGESQ) are both used as explanatory variables.

Gender
M A L E is a dichotomous variable, set equal to 1 for males and 0 for females.

Marital Status
Defined as W E D equal to 1 if the individual is married, 0 otherwise.

148

T. BATES

TABLE 1 Self-Employment Entry for Individuals Aged 21 to 60: Logistic Regression Variable Intercept EDUC2 EDUC3 EDUC4 EDUC5 Experience W2 W3 W4 W5 Wed Age Age2 Male African-American Hispanic Asian n Chi-squarc Logit Coefficient Standard Error Chi-Square Statistic

-6.292 .394 254.09 -. 129 .086 2.23 .178 .090 3,87 .062 .108 0,33 .426 .100 18,19 .011 .003 9,877 .252 .096 6,87 .221 .094 5,56 .157 .091 2,97 .911 .084 118,72 223 .068 10,76 .144 .021 48,30 -.002 .0003 57.54 .823 .062 175,22 -.727 .134 29,552 -.297 .147 4,103 .020 .166 0.01 24,428 (1520 entrants) 737.61; the hypothesis that the explanatory variables have no effect is rejected (ct - 0.01).

Race and Ethnicity

These variables are equal to 1 under the following conditions (0 otherwise):
BLACK: if black, whether or not Hispanic. ASIAN: if Asian or Pacific Islander. HISPANIC: if Hispanic (of any race).

In the econometric exercises, the education variable group excludes people having less than 12 years of formal schooling and the wealth variable group excludes people whose household net worth was under $10,000. EMPIRICAL FINDINGS Table 1 's logistic regression, applied to the sample of individuals who did not own businesses in the first interview period, explains self-employment entry where "entry" is defined as generating annualized net profits or losses of $1,000 or more during any of the subsequent seven SIPP interview periods (28 months). Entry (Table 1) is strongly associated with possession of $100,000 or more of household net worth. Years of work experience along with graduate education are associated positively with self-employment entry. Age and marital status, further, exhibit the expected relationship to self-employment entry. Controlling for these factors, as well as for education, work experience, and wealth, women and minorities (particularly African-Americans) are much less likely to enter self-employment than nonminority males. Asians, however, display the same propensity to enter self-employment as their nonminority cohorts, and Hispanic entry is higher than that of blacks. Table 2 describes the Table 1 dependent and explanatory variables and it is disaggregated to highlight the extreme differences in self-employment entry rates as well as wealth holdings and education backgrounds among the various race/ethnicity and gender groups. The much lower self-employment entry rates among blacks (2.43%) and Hispanics

SELF-EMPLOYMENT ENTRY ACROSS INDUSTRIES

149

TABLE 2 Summary Statistics: Variables Used in Econometric Analysis by Gender, Race, and Ethnicity
Variable % Newbus Education EDUC 1 EDUC2 EDUC3 EDUC4 EDUC5 Nonhispanic White Male All Minorities 9.71% 16.6% 34.2 24.0 12.0 13.1 100.0% Experience Household Net Worth Wl W2 W3 W4 W5 17.4 27.8% 13.7 16.0 21.4 21.1 100.0% 3.46% 33.5 % 33.0 20.7 6.7 6.1 100.0% l 1.7 51.5% 13.5 10.4 12.0 12.7 100.0% AfricanAmerican 2.43% 30.9% 36.3 22.0 5.9 4.9 100.0% 12.3 56.6% 14.0 14.5 10.6 4.2 100.0% Asian 7.06% 21.5 % 24.7 19.8 18.2 15.8 100.0% 10.9 41.0% 12.8 6.2 15.0 25.0 100.0% Hispanic 3.97% 44.2 % 30.2 18.3 3.4 3.8 100.0% 10.8 53.3% 12.2 10.0 15.9 8.7 100.0% All Women 3.81% 19.1% 40. I 22.7 9.6 7.9 100.0% 10.5 32.5% 12.9 15.3 19.2 20.1 100.0%

Mean value % Male % Wed Mean Age n

$69,279 100.0% 69.7 % 37.6 9386

$32,474 43.7% 55.3 % 36.7 4505

$22,232 42.3% 45.0% 37.0 2592

$73,222 46.9% 71.1% 36.6 595

$34,857 44.8% 67.1% 36.2 1336

$68,214 0 66.1% 37.8 13,074

Source: U.S. Bureau of the Census, Survey of Income Program Participation (SIPP).

(3.97%) relative to nonminority males (9.71%) clearly reflect, in part, their lower wealth holdings and weaker educational backgrounds. The lower self-employment entry rates characterizing women (3.81%), in contrast, are more difficult to explain in terms of wealth and human capital differences: 20.1% of the women (Table 2) were in the $100,000 + wealth category, nearly equal to the 21.1% of nonminority males in the top wealth grouping. Among African-Americans, in contrast, only 4.2% had a household net worth of $100,000 or more. Thus, whereas the finding that high levels of wealth and education are strongly related to new business entry explains why nonminority males and Asians are overrepresented among self-employment entrants, it inadequately explains the low entry rate typifying women. The coefficients on the two racial and ethnic group variables--African-Americans and Hispanics--do indicate significantly lower rates of new business formation, even after taking into consideration the lower levels of wealth, less education, and demographic differences. This result implies either that African-Americans and Hispanics face barriers to business entry that nonminorities do not face, or that they have weaker preferences for business ownership than do nonminorities. At least for African-Americans, there is clear-cut support for the interpretation of business barriers. The alternative possibility, that African-Americans have weaker business preferences, is undermined by the 1987 survey which found that they consistently were more highly inclined toward business ownership than were whites, Asians, or Hispanics (Development Associates, Inc. 1987).

150

T. BATES

TABLE 3 Entry into Skilled Services Self-Employment Entry for Individuals Aged 21 to 60: Logistic Regression Variable Intercept EDUC2 EDUC3 EDUC4 EDUC5 Experience W2 W3 W4 W5 Wed Age Age2 Male African-American Hispanic Asian n Chi-square Loglt Coefficient Standard Error Chi-Square Statistic 111.67 5.43 24.13 34.84 69.25 4.56 5.16 2.08 2.13 8.83 2.85 17.09 18.15 3.08 5.45 0.42 0.37 -9.690 .917 .703 .302 1.459 .297 1.827 .310 2.452 ,294 .017 .008 .465 ,205 .299 .207 .284 .195 .552 .186 -.227 .134 .194 .047 -.002 .0006 .219 .125 -.701 ,300 .196 ,302 .192 .315 24,428 (309 entrants) 270.00; the hypothesis that the explanatory variables have no effect is rejected (a - 0.01).

Entry rates among women and African-Americans are highly sensitive to the definition of self-employment entry being utilized. When "entry" is defined as pursuing selfemployment at least 10 hours per week during any interview period, entry among females and African-Americans jumps by 44.0% and 52.4%, respectively, above the percent Newbus (self-employment entry) rates reported in Table 2. When Table 1 's logit exercise is reestimated utilizing the hours per week criterion for self-employment entry, the W5 coefficient value declines by nearly 10%, suggesting that high levels of household wealth are less of a prerequisite for entry into self-employment that yields little income. Wealth levels are clearly less likely to serve as barriers to entry when the line of self-employment being entered generates very low levels of profits--less than two dollars per hour of entrant labor input in this case. Secondly, the male coefficient value declined by 12%, which is consistent with the large increase in women entering self-employment--from 498 to 7 1 7 - - w h e n the definition of noncasual self-employment is expressed in terms of labor hours rather than net profits. Female self-employment rates generally are highly sensitive to any sort of cut-off criterion (such as net income or sales revenues) that excludes very tiny operations from the self-employment universe. In the Census Bureau's comprehensive Characteristics of Business Owners (CBO) database, which attempts to be representative of the small business universe, 49.7% of the female business sample reported 1982 gross sales of under $5000 (Bates 1990a). Women entering self-employment are most heavily overrepresented in the skilled services--finance, insurance, and real estate, professional services, and business services-and most underrepresented in construction. Tables 3 and 4 explore entry into these two self-employment subfields; as in Table 1, entrants are identified as those generating net profits or losses of $1000 or more. In Table 3, "entrants" are defined solely to include people entering skilled services self-employment, whereas Table 4 identifies entrants solely to include those entering construction self-employment. The Table 3 and 4 logistic regression

SELF-EMPLOYMENT ENTRY ACROSS INDUSTRIES

151

TABLE 4 Entry into Construction Self-Employment Entry for Individuals Aged 21 to 60: Logistic Regression
Variable Intercept EDUC2 EDUC3 EDUC4 EDUC5 Experience W2 W3 W4 W5 Wed Age Age2 Male Minority" n Chi-square Logit Coefficient Standard Error Chi-Square Statistic 59.38 1.25 I 44 7.77 9.38 0.02 0.13 2.14 0.76 3.60 3 60 6.95 8.13 69.94 4 99

-9.250 1.200 -.246 .220 -.292 .243 -1.030 .369 - 1.138 .372 .0005 .012 0.91 .252 -.426 .291 .232 .266 .486 .244 .393 .207 .167 .063 .002 .0008 2.154 .261 -.595 .266 24,428 (153 entrants) 156.21; the hypothesis that the explanator3' variables have no effect is rejected (ct = 0.01 ).

"Refers to all except non-Hispanicwhites; small sample size necessitates use of "minority," defined as equal to I for African-Americans, Hispanics,and Asians, and other minorities; 0 otherwise.

equations explaining entry produced variable coefficients that are very different from each other--suggesting that construction entry is a much different process than skilled services self-employment. The role of higher education in identifying self-employment entrants-highly positive for skilled services, highly negative for construction--is particularly revealing. The logit coefficients attached to the education variables in Tables 3 and 4 are:
Skilled Services
EDUC2 EDUC3 EDUC4 EDUC5 (high school grad) (some college) (college graduate) (graduate education) .703 1.459 1.827 2.452

Construction
-.246 -.292 - 1.030 -1.138

Level of education is the most important factor in identifying entrants into skilled services self-employment: probability of entry rises substantially at each of the higher levels of college education. In construction, the opposite situation prevails: high school dropouts are much more likely to enter self-employment than college graduates. The different role that higher education plays in explaining the entry process in construction, as opposed to skilled services, helps to explain the weak role played by education variables (Table 1) in explaining self-employment entry generally. Similarly, the weak and inconsistent explanatory power of education level in predicting self-employment entry observed by Evans and Leighton (1989) obviously reflects the fact that educational background is directly related to entry in some fields and negatively related in construction. Aggregation of these diverse industry groups in studies seeking to identify the traits of self-employment entrants simply leads to a misrepresentation of the human capital characteristics that predict self-employment entry in specific industries. Table 4's logit results literally suggest that years of work experience have no relevance to construction self-employment entry, whereas college graduation severely reduces the

152

T. BATES

likelihood of self-employment. In fact, the human capital requirements for successful creation of a small business in construction are apt to be very real and specific. Construction firms are often started by experienced skilled craftsmen. The specific type of human capital possessed by a journeyman plumber or an experienced sheet metal worker, along with the specific experience most likely to translate into construction self-employment, is not commonly acquired in college, nor is it widely available to women or minorities (Bates and Grown 1992). The strong positive male coefficient value and the strong negative coefficient value for "minority" suggest that the human capital that facilitates self-employment in construction is most accessible to nonminority males. A college education is the primary route to self-employment in skilled services. Whereas greater years of work experience as well as wealth of $10,000 or more are clearly associated positively with entry, college education is a much more important factor for identifying skilled service entrants. Being male--the most important single factor in identifying construction self-employment entry--is a much weaker determinant in skilled services; 42.4% of the skilled services entrants were women, versus 30.3% in all other fields and 12.4% in construction. Similarly, entry into skilled services is easier than in other lines of self-employment for Hispanics; Table 3's "Hispanic" variable coefficient is statistically insignificant. Only among African-Americans do we find a negative, highly significant variable coefficient, suggesting limited access to skilled services self-employment. Among the women described in Table 2, the self-employment entry rate was 2.88% for those who had not attended college, versus 6.33% for the college graduates. Long-run growth in female skilled services self-employment is likely to be fueled by trends in higher educational attainment: the number of professional degrees awarded nationwide to women rose 164.3% over the 1976--1989 period, whereas professional degrees awarded to men shrank 13.9% over this time period (Carter and Wilson 1992). This growth in the educational attainment of women is highly consistent with their disproportionate entry into the area of self-employment where college degrees are an important part of the price of admission. In the theoretical writings of Lucas (1978) as well as Jovanovic and Evans (1989), business entry is determined by a parameter, business acumen. At least in skilled services, the source of this acumen appears to lie in high levels of educational attainment. The largest-scale, most capital-intensive lines of business--manufacturing and wholesaling--were entered by 87 of the 24,428 individuals described in Table 2. Table 5 identifies the traits of these 87 entrants, utilizing the methodology employed in Tables 3 and 4. In sharp contrast to both construction and skilled services, educational background has no explanatory power whatsoever in identifying manufacturing and wholesaling entrants. Years of work experience emerges as an important explanatory variable in Table 5. People entering manufacturing and wholesaling have significantly more years of work experience (19.4 on average) than entrants in other fields. The goods industries resemble construction in the sense that nonminority males are much more likely to enter self-employment than their female and minority cohorts. Manufacturing and wholesaling, finally, stand out as areas where possession of $100,000 or more of personal wealth is very strongly associated with self-employment entry. This finding is consistent with the large-scale, more capital-intensive nature of the goods industries. A further logistic regression analysis exercise entailed redefining the dependent variable, self-employment entry, so that entrants included only those people reporting annualized net profits or losses of at least $5000 during an interview period. Reestimating Tables 3, 4, and 5 using this self-employment criterion, the significance of wealth as an entry determinant--

SELF-EMPLOYMENT ENTRY ACROSS INDUSTRIES 153

TABLE 5 Entry into Manufacturing and Wholesaling Self-Employment for Individuals Aged 21 to 60:

Logistic Regression
Variable Intercept EDUC2 EDUC3 EDUC4 EDUC5 Experience W2 W3 W4 W5 Wed Age Age2 Male Minority n Chi-square Logit Coefficient Standard Error Chi-Square Statistic 38.08 0.24 0.64 0.01 0.37 3.28 0.41 0.11 0.66 10.31 0.04 4.40 5.0 l 16.56 3.53 ~s - 10.214 1.655 -.171 .352 .285 .356 -.039 .438 .242 .400 .026 .014 .263 .413 .137 .408 .100 .389 1.091 .339 -.056 .270 .178 .085 -.002 .001 1.098 .270 -.813 .433 24,428 (87 entrants) 76.03; the hypothesis that the explanatory variables have no effect rejected (o~ = 0.01 ).

the W5 coefficient, most specifically--rose substantially, particularly in the Table 5 analysis of entry into manufacturing and wholesaling. The male coefficient also increased substantially, but the racial grouping variable coefficients were not impacted.

OVERVIEW

The varying factors linked to self-employment entry in different industries tend to obscure the role played by education and work experience in identifying self-employment entrants. The likelihood of self-employment in skilled services increases greatly as educational attainment rises, whereas the opposite situation prevails in construction. The likelihood of selfemployment entry in manufacturing and wholesaling rises steadily as years of work experience accumulate; the same pattern, albeit weaker, prevails in skilled services. A single factor or two dominates the econometric models of self-employment entry for each of the industries under consideration: (1) the skilled services entrant is highly educated; (2) the construction entrant is male and decidedly not highly educated; (3) the person entering self-employment in manufacturing and wholesaling possesses household net worth exceeding $100,000 and is most likely male. The statistical findings also shed light on who may or may not emerge from the new firm gestation process, and who will pursue self-employment on a long-term basis versus who may choose to abandon dreams of small business ownership. Particularly for employees contemplating quitting their jobs to pursue self-employment full-time, the ability to make a living will influence whether or not self-employment shall become one's primary labor force status. Those commanding greater household wealth are clearly more likely to report higher self-employment earnings. Household wealth alone provides resources for business capitalization but is also very positively associated with the borrowing power of potential entrepreneurs (Bates 1990b). Thus, the person commanding significant household wealth is much more capable than the less wealthy of assembling adequate business capitalization to

154

T. BATES

TABLE 6 Self-Employment Entry for Women Aged 21 to 60: Logistic Regression

Variable Intercept EDUC2 EDUC3 EDUC4 EDUC5 Experience W2 W3 W4 W5 Wed Age Age2 African-American Hispanic Asian n Chi-square Logit Coefficient Standard Error Chi-Square Statistic

-6.863 .679 102.17 -.030 .154 .04 .398 .161 6.07 5.58 .445 .189 .625 .188 11.09 10.77 .018 .005 .494 .160 9.49 .427 .157 7.41 .111 .162 0.47 18.62 .647 .150 4.28 .233 .113 .166 .036 21.22 -.002 .0004 24.75 -1.145 .261 19.32 -.512 .278 3.38 0.07 .074 .272 13,074 (498 entrants) 183.28; the hypothesis that the explanatory variables have no effect is rejected (ct = 0.01).

overcome applicable barriers to business entry. The less wealthy person, furthermore, is particularly disadvantaged in the financial capital-intensive lines of self-employment-manufacture and wholesale--and hence less likely to make the transition from gestation to small business maturity. Several factors perform consistently in identifying self-employment entry across industries. The likelihood of entry increases with age, peaking as people approach age 40 and then leveling out. African-Americans are less likely to pursue self-employment, all other factors constant. When sample size is too small to permit use of variables identifying African-American, Hispanic, and Asian subgroups, "minority" is substituted. The minority factor is consistently a strong negative predictor of entry into construction and goods self-employment. In light of the different industry groups in which self-employed men and women concentrate, the self-employment entry process may vary substantially by gender. This is investigated in Tables 6 and 7, which explain entry separately for males and females. These logit analyses are performed for all industry groups and the entry criterion is the previously discussed $1000 annual net profits threshold. As expected, the human capital variables are important determinants of self-employment entry for women. Relative to men, women are relying much more heavily upon advanced education and work experience as their route to self-employment. Possession of graduate education, in contrast, is the only human capital variable strongly associated with self-employment entry in Table 7's all-male logit exercise. Relative to women, men entering self-employment are relying more on substantial wealth holdings and they are disproportionately attracted to male preserves such as construction. Barriers to self-employment entry are most binding to those who possess neither the human capital nor the financial capital traits that are associated with entry. Lack of wealth and advanced educational credentials are particularly pronounced barriers to black and Hispanic self-employment. Reestimating Table l's logit equation for nonminority males only and applying its coefficients to the African-American and Hispanic populations described in

S E L F - E M P L O Y M E N T ENTRY ACROSS INDUSTRIES

155

TABLE 7 Self-Employment Entry for Men Aged 21 to 60: Logistic Regression

Variable Intercept EDUC2 EDUC3 EDUC4 EDUC5 Experience W2 W3 W4 W5 Wed Age Age2 African-American Hispanic Asian n Chi-square Logit Coefficient Standard Error Chi-Square Statistic

-5.473 .482 128.72 -. 164 .105 2.43 .087 .110 0.63 -. 136 .133 1.05 .307 .119 6.66 .001 .005 0.08 .115 .120 0.91 .094 .118 0.64 .178 .110 2.61 1.048 .102 106.19 .252 .087 8.38 .147 .026 32.75 -.002 .0003 35.25 -.558 .157 12.66 -. 199 .174 1.32 .036 .210 0.29 11,354 (1022 entrants) 311.92; the hypothesis that the explanatory variables have no effect ts rejected (tx = 0.01 ).

Table 2 results in a predicted rate of entry that rises by less than 25% relative to the 2.43% and 3.97% entry rates, for African-Americans and Hispanics respectively, shown in Table 2. For all women, in contrast, the predicted rate of entry more than doubles the actual 3.81% rate (Table 2) when nonminority male logit coefficients are applied. For African-Americans and Hispanics, therefore, large gains in self-employment await advances in human and financial capital characteristics of potential entrepreneurs. For women, very large increases in self-employment are possible without corresponding increases in human and financial capital. Barriers to increased self-employment of women may exist in select fields such as construction, but barriers need not deny access to self-employment. Rather, such barriers may simply shape the industry choice of those entering self-employment.

REFERENCES
Acs, Z.J., and Audretsch, D.B. 1989. Small firm entry in manufacturing. Economica 56:255-265. Bates, T.M. 1986. Characteristics of minorities who are entering self-employment. The Review of Black Political Economy 15(2): 31-49. Bates, T.M. 1990. The characteristics of business owners data base. Journal of Human Resources 25(4):752-756. Bates, T.M. 1990. Entrepreneur human capital inputs and small business longevity. The Review of Economics and Statistics 72(4):551-559. Bates, T.M., and Grown, C. 1992. Commercial Bank Lending Practices and the Development of Black-Owned Construction Companies. Journal of Urban Affairs 14(1 ): 25-41. Birley, S. 1984. Finding the new firm. Proceedings of the Academy of Management Meetings 47:64-68. Carsrud, A.L., Olm, K.W., and Eddy, G.G. 1986. Entrepreneurship: research in quest of a paradigm. In D.L. Sexton and R.W. Smilor, eds., The Art and Science of Entrepreneurship. Cambridge, MA: Ballinger. Carter, D., and Wilson, R. 1992. Minorities in Higher Education. Washington, DC: American Council on Higher Education.

156

T. BATES

Cooper, A.C., Woo, C.Y., and Dunkelberg, W.C. 1989. Entrepreneurship and the initial size of firms. Journal of Business Venturing 4(3): 317-332. Cooper, A.C., and Dunkelberg, W.C. 1982. Entrepreneurial typologies: an empirical study. In K. Vesper, ed. Frontiers of Entrepreneurship Research. Boston, MA: Babson College. Development Associates, Inc. 1987. Attitudes and Inclinations of Minority Youth Toward Business Ownership. Washington, DC: U.S. Department of Commerce Minority Business Development Agency. Evans, D.S., and Jovanovic, B. 1989. An estimated model of entrepreneurial choice under liquidity constraints. Journal of Political Economy 97(4): 808-827. Evans, D.S., and Leighton, L. 1989. Some empirical aspects of entrepreneurship. The American Economic Review 79:519-535. Fratoe, F. 1988. Social capital and black business owners. The Review of Black Political Economy 16(4):33-50. Lucas, R.E. 1978. On the size distribution of business firms. The Bell Journal of Economics 9(3):508-523. Reynolds, P., and Miller, B. 1992. New firm gestation: conception, birth, and implications for research. Journal of Business Venturing 7(3):405--417. Waldinger, R.D. 1986. Through the Eye of the Needle. New York: New York University Press. White, L. 1982. The determinants of the relative importance of small business. The Review of Economics and Statistics 54 (1).