Data Directory UMM :Data Elmu:jurnal:J-a:Journal of Economics and Business:Vol52.Issue6.2000:

utility from self-employment and the expected utility from wage-and-salary employment, I , is a stochastic function of observable personal characteristics X. That is, I 5 Xb 1 e 1 where e is normally distributed with a mean of 0. Relevant X variables are those that may affect the individual’s taste for self-employment versus wage-and-salary employment, indicating greater less happiness in self-employment, as X j increases, when b j is positive negative. Relevant X variables may also include those that may affect the individual’s likelihood of a satisfying work situation in self-employment vs. wage-and-salary employ- ment, indicating greater less likelihood in self-employment, as X k increases, when b k is positive negative. An individual chooses self-employment if I .0, but chooses wage- and-salary employment if I ,0. Probit estimation of this relationship is important in its own right. Estimation of this relationship for men and women separately can shed light on the sources of the difference in the self-employment rates of men and women. An understanding of the employment choice through probit estimation is also vital for an accurate understanding of the determinants of earnings in the two sectors. It is assumed that the natural logarithm of earnings in each type of employment is a stochastic function of observable personal characteristics Z. That is, ln Y se 5 Z se g se 1 n se 2 and ln Y ws 5 Z ws g ws 1 n ws 3 where n se and n ws are normally distributed with means of 0 in the population. However, Equations 2 and 3 are estimated on subsamples of the population, consisting of workers within each sector i.e., self-employed and employees, respectively. Steps to avoid sample-selection bias in the estimation of g require as input the results of probit estimation of the employment choice. 6 Estimation of 2 and 3 for men and women separately, corrected for self-selection, provides a basis for studying the sources of gender differences in the determination of earnings in each type of employment. 7 It also allows one to address gender differences in the earnings gap between self-employment and wage-and-salary employment. Details concerning the data set used in the exploration of these issues are given in the following section.

IV. Data

The data used in this study come from a 11000 Public Use Microdata Sample PUMS of the 1990 Census. Only individuals who were heads of households or partners married 6 In addition to the exogenous variables, Z se and Z ws, one must use estimates of selectivity variables l se 5fc1-Fc, and l ws 52fcFc, respectively, where F is the cumulative distribution of a standard normal random variable, f is its density function, and c is -Xbs e . A positive negative coefficient on the constructed selectivity variable l se indicates that observed conditional mean earnings among the self-employed are greater less than their population means. A positive negative coefficient on the constructed selectivity variable l ws indicates that the observed conditional mean earnings among wage-and-salary workers are less greater than their population means. See Lee 1978 and Maddala 1983 for more details. 7 Oaxaca 1973 demonstrated one approach for wage-and-salary workers. Because the model specified in this paper uses predominantly dichotomous variables to reflect personal characteristics, an alternative approach is employed in the analysis that follows. 502 S. H. Clain or unmarried of household heads, 65 years or younger in age, and working at least 35 hr per week and at least 40 weeks per year, with a nonagricultural, nonmilitary occupation, are included in this analysis. 8,9 Observations with missing data, inconsistencies in data, or negative income were omitted, leaving 38,015 men and 26,667 women. 10 Of these, 4,025 10.6 men and 1,377 5.2 women were self-employed. 11 Besides information on income and type of employment, the records for these indi- viduals also include information on age, race, marital status, education, occupation, geographical location, health, fluency in the English language, and presence of one’s own children in the household. 12 The income of a spouse can be gleaned from the data record of the spouse, if any. 13 As a group, these factors can influence the individual’s choice of self-employment, through their influence on tastes and opportunities. 14,15 These factors can also influence earnings, through their influence on productivity and local market conditions. 8 The analysis was restricted to heads of households, or partners thereof, to focus on those who have had primary or joint responsibility for a household. Individuals who are working full-time, but are not household heads or partners thereof, may be in transition andor subject to undue uncertainty or instability in their life circumstances. Being viewed as an inherently different population, these individuals are omitted from the analysis. 9 Modeling the labor force participation decision or the decision to work part-time andor part-year is beyond the scope of this paper. Full-time andor full-year restrictions are also made in Rees and Shah 1986, Fairlie and Meyer 1996, Moore 1983a, 1983b and Yuengert 1994; however, the exact definition of “full” varies from paper to paper. Here, the restrictions eliminate 627 self-employed female workers and 375 self-employed male workers. These restrictions are imposed so that annual earnings can be compared without excessive variation associated with annual level of commitment to work. In effect, Equation 1 measures the difference between the utility of full-time, full-year employment in the two sectors, and Equations 2 and 3 are full-time, full-year annual earnings equations. 10 Workers who did not claim to be self-employed, but reported self-employment income, are excluded. Workers who reported both wage-and-salary income and self-employment income were either moonlighting at two or more jobs, or moving into or out of self-employment during the year. They are also excluded, because the reported hours and weeks are not separated by job, and therefore not necessarily indicating a full-time, full-year commitment to self-employment. The latter restriction eliminates 236 full-time, full-year female workers and 740 full-time, full-year male workers. As a consequence of these exclusions, the results based on this sample may not apply to individuals of these types in transition andor moonlighting in the population of self-employed. Note that wage-and-salary income was defined as total money earnings received for work performed as an employee during the calendar year 1989. That is, it included wages, salary, commissions, tips, piece-rate payments, and cash bonuses earned before deductions were made for taxes, pensions, etc. Self- employment income was defined as net money income gross receipts minus expenses from one’s own business, professional enterprise, or partnership. 11 Of the self-employed men, 1487 37 are incorporated; of the self-employed women, 385 28 are incorporated. This study does not attempt to explain gender differences in the propensity to incorporate one’s business. If incorporation systematically influences self-employment income, it would contribute to an expla- nation of the overall gender differences in self-employment income. In this study, the return to incorporation may manifest itself as a return to personal characteristics that raise the likelihood of incorporation e.g., education. 12 The possibility that occupation is jointly determined with sector is not considered here. Although the census classification consists of 500 specific occupational categories, broad classes are used in this analysis. Finer categorization would force the estimation of occupational effects to rely on smaller numbers of individuals within each occupational category. Moreover, it would make sector comparisons holding occupational category fixed more objectionable. 13 Co-ownership of family businesses by married couples cannot be inferred 14 For example, higher education may raise the value that an individual places on being one’s own boss. On the other hand, the presence of small children may reduce the opportunities for wage-and-salary employment moreso than the opportunities for self-employment. 15 Noticeably missing is a measure of assets. However, as Fairlie and Meyer 1996 warned, using assets as an independent variable in a cross-sectional analysis of self-employment could lead to faulty results, because high assets could be a consequence rather than a cause of self-employment. Gender Differences in Self-Employment 503 The variables constructed from this information are summarized in Table 1. 16 Descrip- tive statistics for men and women are presented in Tables 2 and 3, respectively. It can be seen from these tables that self-employed workers tend to be older than wage-and-salary employees. Self-employed workers are more likely to be White than are wage-and-salary employees. They are also more likely to be married. Compared to wage-and-salary workers, a smaller proportion of the self-employed workers live in a central city location, MSA or PMSA. Gender differences in the comparisons of self-employed workers and wage-and-salary employees do exist. For women, self-employment earnings fall below wage-and-salary income, on average. For men, the reverse is true. 17 Among men, self-employed workers tend to be more college-educated than wage-and-salary employees. For men, there is a greater concentration of white-collar workers among the self-employed. For women, the concentration of service workers is greater among the self-employed than among the wage-and-salary employees. As outlined in Section II, probit and regression analyses are applied to sort out the effects of these influences, ceteris paribus, on the choice of type of employment and the subsequent level of earnings, by gender. The results of this estimation are presented and discussed in the following section. 16 The omitted category in the measurement of educational attainment is the high-school dropout. Given the definitions of the educational dummy variables, b HS represents the effect of the high-school diploma, compared to the high-school dropout. b HS 1 b COLL represents the effect of the college diploma, compared to the high-school dropout, whereas b COLL represents the effect of the college diploma, compared to the high-school graduate. 17 Because the incorporated self-employed may have reported earnings as either wage or salary income or self-employment income, their earnings are defined in this analysis as the sum of those two types of income. Given the exclusions noted earlier in footnote 10, only one of these would be nonzero for any self-employed individual included in the study. Table 1. Definitions of Variables Variable Name Definition LNINC Natural logarithm of 1989 earnings MARRIED 1 if individual is married; 0 otherwise SPINCOM Spouse’s total 1989 income WHT 1 if race of individual is white; 0 otherwise HS 1 if individual completed high school; 0 otherwise COLL 1 if individual completed college; 0 otherwise AGE Years of age AGESQ Square of age CITY 1 if individual resides in central city location, MSA or PMSA; 0 otherwise HLTHLIM 1 if individual is limited in kind or amount of work, has a mobility limitation, or has a personal care limitation; 0 otherwise FLUENT 1 if individual is fluent in English; 0 otherwise DKIDS 1 if individual is female and living with her own minor children; 0 otherwise NEAST 1 if individual lives in Northeast; 0 otherwise MWEST 1 if individual lives in Midwest; 0 otherwise WEST 1 if individual lives in West; 0 otherwise WHTCLL 1 if individual’s occupation is among managerial or professional specialties, or the individual works in a technical, sales or administrative support position; 0 otherwise SERV 1 if individual is in a service occupation; 0 otherwise l se , l ws Selectivity variables, estimated from probit equation results reported in Table 4 504 S. H. Clain

V. Estimation Results