L. Leete J. of Economic Behavior Org. 43 2000 423–446 429
localized wage comparisons take on a particular importance. However, what is ‘local’ will be defined by the scope of the particular labor market. Thus, one might expect wage equity
to occur within or across all occupations within an organization, or within an occupation but across organizations in a particular industry. Any of these comparisons may apply only
to higher level employees — managerial and professional employees, and technical and administrative support staff — those who have the most information about relative wages
and whose workplace conduct is most dependent on self-determination. Furthermore, to the extent that wage inequity in the US has race- and gender-related components, one
might expect these to be diminished in nonprofit organizations as compared with for-profit organizations.
This emphasis on wage equity in the nonprofit sector is not meant to downplay the possible importance of such factors in for-profit settings, but rather to highlight the nonprofit sector
as a context in which these factors may be relatively more concentrated as compared with the economy at large. While motivated employees or organizational pride might be productivity
or sales enhancing in the for-profit sector, they may be the sine quibus non of the nonprofit sector.
5
Furthermore, the implication that wage equity is important to maintaining worker motivation can only be taken to suggest that wage equity may be a necessary precondition,
not that it is a sufficient one. For example, Freeman and Medoff 1984 note that while wage equity is greater among union than among non-union employees, stated job satisfaction is
not.
3. Evidence of wage equity in the nonprofit sector
If nonprofit organizations require more intrinsically motivated and organizationally ori- ented employees, and if wage equity is central to maintaining intrinsic motivation, group
cohesiveness and organizational pride, then one would expect the wages of nonprofit em- ployees to be less dispersed either within or between organizations than those of for-profit
organizations. Previous researchers have collected some evidence that is consistent with this hypothesis. Mirvis and Hackett and Mirvis 1992 examine survey data for 1977 and
1990, respectively, and find some suggestions that this is the case. In both years, they find that nonprofit employees are more likely than for-profit employees to report that they are
paid fairly “as compared to what other people doing my kind of work are paid.” Mirvis notes further that there is “usually less disparity in wages and working conditions from top
to bottom in nonprofits” p. 26.
In this Section, I use 1990 US Census data to examine whether these perceptions of greater wage equity in the nonprofit sector are supported by evidence on differences in
nonprofit and for-profit sector wage structures. Of course, if wage equity within or between organizations is to gain expression in the wage distribution of an entire sector this requires
either that there be no countervailing factors, or that those factors be properly controlled for. For example, wages could be less dispersed within nonprofit organizations than within
for-profit organizations, but nonprofit organizations themselves could be more dispersed
5
Of course, to the extent that either ideology or quality is costly, ideological employees could be a liability to a profit-maximizing organization.
430 L. Leete J. of Economic Behavior Org. 43 2000 423–446
across the wage distribution than for-profit ones. Therefore, I control for as many underlying determinants of the wage distribution as possible.
The basic measure examined here is the difference in the variance of wages between the two sectors. Raw wages are decomposed into predicted and residual components using
OLS regression. The variance differential for predicted wages is further decomposed into the portions attributable to the differences in returns and differences in characteristics. The
variance differentials are examined for all workers, and within both broad and detailed oc- cupations and industries. I also examine differences in race and gender wage discrimination
between the two sectors.
3.1. Data The dataset used here is the 5-percent Public Use Microdata Sample PUMS of the
1990 Census. This includes a sample of 4.1 million individuals employed in the private sector, of whom 8.5 percent work in the nonprofit sector. These data were self-reported
by individuals receiving the ‘long form’ survey of the census. Sector of employment was identified by asking individuals if they were an employee of ‘a private for-profit company
. . . a private not-for-profit tax-exempt, or charitable organization, local government, state government, etc.’
6
In addition to type of employment, the PUMS reports individuals’ wage and salary income for 1989, occupation and industry of employment, weekly hours of work,
number of weeks worked, and individual characteristics, such as age, education including type of degree held, gender, race, area of residence, language fluency, and disability status.
Hourly wages are imputed as annual wage and salary income divided by total hours worked in 1989. Government workers are eliminated from the sample so that the comparisons made
are strictly between for-profit and nonprofit workers. The sample was also limited to those who were not currently enrolled in school and who did not report having a disability that
limited their ability to work.
7
The sample was not restricted in any other way.
8
Sample means of the variables used are reported in Table 1 by organizational status. Nonprofit
workers average higher hourly wages, are more likely to be part-time, female, and fluent in English, and they have on average more years of potential labor market experience defined
as age minus years of education minus six. The racial mix in the nonprofit sector is slightly more white, and educational levels are higher. All of these characteristics are as expected
given the preponderance of white-collar, service sector occupations in the nonprofit sector.
6
Responses were checked for consistency with answers to questions on employer name, location, industry and occupation. As part of the consistency check, data processors could use a directory of company names to identify
the correct industry code and legal form of an organization. They could then recode the answer to the ‘class of worker’ question accordingly. For a detailed discussion of the implications of misreporting of nonprofit status, see
Leete 2001.
7
Disabled workers are eliminated because degree of disability is not sufficiently identified and severely disabled clients of ‘sheltered workshops’ in the nonprofit sector whom earn some pay will alter the distribution of nonprofit
wages.
8
The discussion in this paper relates primarily to the major classes of nonprofit organizations formed as a result of information asymmetries or on ideological grounds. These nonprofits are predominately found in the professional
services sector of the economy, while other types of nonprofits are found elsewhere. However, because this sector accounts for 82 percent of nonprofit employment, limiting the analysis here to this sector of the economy has little
effect on the results.
L. Leete J. of Economic Behavior Org. 43 2000 423–446 431
Table 1 Means of dependent and independent variables by sector
a
Variable For-profit
Nonprofit Means
Lnhourly wage 1989 2.23
2.27 Years of potential experience age-Ed-6
19.9 21.8
Percent Female
43.9 66.6
Not fluent in English 3.1
1.3 Average hours worked per week, 1989
40.6 37.9
Working 10 h per week, 1989 1.8
3.8 Working part-time 25 h per week, 1989
9.8 16.5
Average weeks worked, 1989 45.1
45.1 Working 13 weeks, 1989
5.7 5.1
Race White
81.5 83.8
African–American 9.4
9.5 Hispanic
3.4 2.5
Asian 2.6
2.4 Other race
3.2 1.9
Educational attainment No school
0.8 0.4
Nursery school 0.0
0.0 Kindergarten
0.0 0.0
1st–4th grade 0.8
0.3 5–8th grade
4.0 2.0
9th grade 2.5
1.1 10th grade
3.7 1.7
11th grade 3.6
1.7 12th grade, no diploma
3.8 2.0
High school graduate or GED 35.2
21.7 Some college, no degree
20.5 17.6
Associate degree, occupational program 3.8
5.2 Associate degree, academic program
2.9 4.2
Bachelors degree 13.5
22.6 Masters degree
3.0 12.6
Professional degree 1.5
3.7 Doctorate degree
0.4 3.1
N 3822413
323548
a
Private sectors workers, not disabled or enrolled in school.
3.2. The decomposition of variance differentials Using a human capital earnings function, actual wages can be decomposed into the por-
tions attributable to the presence of certain characteristics, the returns to those characteris- tics, and parts unexplainable. Differences in the wage distribution between two sectors can
be attributed to differences in the distribution of any of these components. Thus, I use OLS
432 L. Leete J. of Economic Behavior Org. 43 2000 423–446
regression to estimate a standard human capital earnings function separately for workers in the for-profit and nonprofit sectors. I then examine the distribution of actual, predicted and
residual wages for the two sectors. The estimated equation is
W
k
= a
k
+ b
k
Z
k
+ ε
k
. 1
where W represents the natural log of the hourly wage, ε the error term of the equation. The superscript k is an indicator for either the nonprofit or for-profit status. Control variables
summarized in Z are: dummy variables for six 1-digit occupations, 11 1-digit industries, location in an urban area, lack of fluency in speaking English, 10 race-gender categories,
and part-time work 25 hours per week. In addition, continuous variables for years of education, potential experience and potential experience-squared are included.
9
This equa- tion is estimated on samples of 323,521 nonprofit workers and 3,822,020 for-profit workers
for whom no data were missing for any of the included variables. A full set of estimated coefficients for each sector are shown in Appendix A.
Predicted log wages in a given sector are calculated as ˆ
W
k
= ˆ a
k
+ ˆ b
k
ˆ Z
k
2 where ˆa
k
and ˆ b
k
are estimated coefficients for that sector. Residual log wages are then calculated as
¯ W
k
= W
k
− ˆ W
k
3 As a basic measure of differences in wage equity between the two sectors, we examine the
variance of each of these wage measures actual, predicted and residual and compare them across sectors. Following Freeman 1980, the difference in the variance of predicted wages
can also be decomposed into the portions attributable to differences in characteristics and differences in returns to those characteristics between sectors. This decomposition presents
the classic index number problem of any such decomposition e.g. Oaxaca, 1973. The differences attributable to different characteristics can be weighted by the returns of either
the nonprofit or for-profit sector, and vice versa. The variance differential for predicted wages is
Var ˆ W
fp
− Var ˆ W
np
= Var ˆa
fp
+ ˆ b
fp
Z
fp
− Var ˆa
np
+ ˆ b
np
Z
np
= X
i,j
ˆ b
fp i
ˆ b
fp j
covZ
fp i
Z
fp j
− X
i,j
ˆ b
np i
ˆ b
np j
covZ
np i
, Z
np j
4 which in turn can be expressed either as
X
i,j
[ ˆ b
fp i
ˆ b
fp j
− ˆ b
np i
ˆ b
np j
]covZ
fp i
, Z
fp j
+ X
i,j
ˆ b
np i
ˆ b
np j
[covZ
fp i
, Z
fp j
− covZ
np i
, Z
np j
] 5a
9
Indicators for firm size and union status are two variables that would ideally be included here but are not available in this dataset.
L. Leete J. of Economic Behavior Org. 43 2000 423–446 433
Or as X
i,j
[ ˆ b
fp i
ˆ b
fp j
− ˆ b
np i
ˆ b
np j
]covZ
np i
, Z
np j
+ X
i,j
ˆ b
fp i
ˆ b
fp j
[covZ
fp i
, Z
fp j
− covZ
np i
, Z
np j
] 5b
where i, j=1 . . . n, and n is the number of independent variables in Z. In both Eqs. 5a and 5b, the first term represents the variance differential attributable to different returns to
characteristics in the nonprofit and for-profit sectors. In 5a these differences are weighted by for-profit characteristics; in 5b they are weighted by nonprofit characteristics. Similarly,
the second term of each equation is the difference attributable to differences in worker characteristics between the sectors, weighted by the returns of either sector.
3.3. Differences in wage variances between two sectors The variance of actual wages is 0.587 across the entire for-profit sector, and 0.494 across
the nonprofit sector. Of these amounts, 0.165 and 0.097, respectively, are attributable to the predicted portion of the wage. The remaining variance in each sector is a function of
residual wages.
10
The for-profitnonprofit variance differentials and their decomposition are shown in Table 2, for all workers and by broad occupation and industry categories. All
differentials shown represent the variance in the for-profit sector minus the variance in the nonprofit sector; a positive number indicates that variance in the nonprofit sector is lower
and that wages in the nonprofit sector are less dispersed than in the for-profit sector. Results are shown for all workers, for white collar and blue collar workers separately, for executive
and non-executive white collar workers, and for workers in finance, insurance and real estate, entertainment and recreation services, and professional services.
11
All differences are statistically significant at the 0.005 level or higher F-test. As shown in line 1, actual
hourly wages are more tightly clustered in the for-profit sector than in the nonprofit sector across all groups. The size of the differences is also meaningful, with variances in the
for-profit sector typically 10–20 percent higher than in the nonprofit sector.
Of course, the distributions of a multitude of characteristics underlie the distribution of actual wages. The variance differential of predicted wages is shown in line 2 of Table 2.
Again, the differences are all positive and significant. This explained portion can be decom- posed into its components. Regardless of the set of weights chosen, the largest share of the
difference is attributable to differences in returns to characteristics lines 3a and 3b rather than to differences in the characteristics themselves 4a and 4b. Finally, in line 5, I show the
difference in the variance of residual wages. If the theory discussed above implies behav- ioral differences in wage setting across nonprofit and for-profit institutions, then we would
expect this to manifest itself either in the differences in returns to particular characteristics
10
While not shown here, the magnitude of each type of variance in each sector is comparable across the different occupation and industry sub-groups analyzed.
11
White collar workers are defined as census occupational categories: ‘managerial and professional workers’ and ‘technical, sales and administrative workers’. Executives are identified as census occupation codes 0–22. The
1-digit industries shown are those in which there is the most significant level of nonprofitfor-profit competition. See Table 3 for nonprofitfor-profit percentage compositions of these industries.
434 L. Leete J. of Economic Behavior Org. 43 2000 423–446
L. Leete J. of Economic Behavior Org. 43 2000 423–446 435
436 L. Leete J. of Economic Behavior Org. 43 2000 423–446
or in differences in unexplained residual wages. These elements are captured in lines 3a, 3b and 5. Together they comprise most of the observed differences in wage distribution
between the sectors. The pattern across occupations evident in Table 2 is quite striking. The differences are
consistently largest among white collar workers, especially among white collar workers classified as executives. For blue collar workers, the differences in actual and predicted
wages are relatively small but still positive, but the differences in residual wages become negative — i.e. for blue collar workers the residual wage is more dispersed among nonprofit
workers than among for-profit workers. Interestingly, the pattern among white collar workers is strongest among executives. These findings are consistent with the perception that there are
significant differences in the way that nonprofit and for-profit executives are compensated. However, this is not the sole explanation for the differences between the sectors: differences
still persist down through the ranks of white collar workers.
The lower wage dispersion in the nonprofit sector apparent for white collar workers also holds within industry categories in which there is significant representation in both the
nonprofit and for-profit sector. In all three industries shown, actual, predicted and resid- ual wages are less dispersed in the nonprofit sector than in the for-profit sector. Only in
finance, insurance and real estate is the variance attributable to the distribution of worker characteristics higher in the nonprofit sector than the for-profit sector.
Table 2 only captures the broadest measures of differences in wage distributions across sectors, however. In order to confirm these findings at a more detailed level, I also make
similar calculations within each 3-digit occupation and within important 3-digit industries as well. Here I estimate Eq. 1 within each detailed industry and occupation once across
both nonprofit and for-profit workers and compare the variance of the residual wages for nonprofit and for-profit workers.
12
The results are summarized in Table 3 for those occupa- tions and industries in which there is significant nonprofit and for-profit representation.
13
In 48.5 percent of 262 occupations, Var ¯
W
np
was significantly lower than Var ¯ W
fp
follow- ing an F-test with P≤0.05, while it was significantly higher in only 26.3 percent. Similarly,
nonprofit wages are less dispersed in 25 out of 31 detailed industries, especially in those industries classified as professional services where nonprofits are most dominant. The pat-
tern varies, however, by broad occupational classification. Nonprofit wages are distinctly more equitable than for-profit wages in managerial and professional occupations and in
service occupations, where Var ¯
W
np
is lower in nearly 70 percent of included occupa- tions. Var ¯
W
np
is also more likely to be lower among nonprofit workers in technical, sales and administrative occupations. In contrast, there is little difference between the sectors
when comparing precision, craft and repair occupations and among operators, fabricators and laborer occupations, Var ¯
W
np
is higher in over 60 percent of the occupations. These patterns are consistent with the previous findings and with the suggestion that if equity is
used to support intrinsic motivation and organizational identification in nonprofit organiza- tions, one would most expect to see it manifested in the wages of white collar employees:
12
While not separately decomposing the differences due to returns and characteristics, this comparison of residual wages will capture the differences in returns and unexplained differences of interest here.
13
The analysis is limited to occupations in which there are at least 50 observations in each sector.
L. Leete J. of Economic Behavior Org. 43 2000 423–446 437
the employees who have may have the most information about relative wages and whose conduct is most likely to affect organizational outcomes.
3.4. Race and gender wage differences Wage inequity in the US economy has often taken the specific form of wage differences
along race and gender lines see, for example, Blau and Beller, 1992; Bound and Freeman, 1992. These differences could reflect, among other things, preference based discrimination
i.e. Becker, 1957, statistical discrimination Aigner and Cain, 1977, or occupational crowding Bergmann, 1974. The discussion above suggests that the greater apparent need
for nonprofit organizations to achieve fairness and neutrality should imply lower levels of race and gender wage discrimination in the wage structure there. I will use the 1990 US
PUMS data to investigate this aspect of wage equity in nonprofit and for-profit employment. I estimate an equation similar to Eq. 1 above:
ln W = a + bX + cForprof × Race × Gender + dNonprof × Race × Gender + ε 6
However, Eq. 6 has been augmented in a number of ways. First, race and gender wage ef- fects are estimated separately in both nonprofit and for-profit organizations. Race×Gender
and Nonprof×Race×Gender together represent a series of 19 dummy variables repre- senting all racegenderorganization groups except the comparison group, white males in
for-profit organizations. In addition, control variables summarized in X are in some cases more detailed than those included in Z in Eq. 1:
14
X includes dummy variables for 46,933 detailed occupationindustry cells,
15
367 urban and rural areas,
16
lack of fluency in speaking English, part-time work 25 hours per week, 17 categories of educational at-
tainment and type of degree earned, as well as potential experience and potential experience- squared.
17
Coefficients in the vector c c
1
–c
9
represent average wage differences in for-profit or- ganizations between white males and other demographic groups. Coefficients in vector d
d –d
9
can be normalized as d
1
−d , d
2
−d , and so on to similarly represent average
wage differences in nonprofit organizations between white males and others. The coeffi- cients c
1
–c
9
and the normalized coefficients d
1
−d through d
9
−d are presented in Table 4.
The results are quite striking. In every case, race and gender wage differences are dimin- ished in the nonprofit sector as compared with the for-profit sector. The nonprofit effects
are between 22 and 45 percent lower than the for-profit effects. All differences between
14
The specification in Eq. 1 was computationally limited by the need to calculate a decomposition of predicted values and residuals from that specification.
15
These 46,933 industryoccupation cells constitute the non-empty intersection of 490 occupation and 234 industry identifiers.
16
Dummy variables represent each separate Census MSAPMSA area as well as the non-MSAPMSA areas of each state.
17
For an analysis of how detailed occupationindustry cell controls affects the estimates of race and gender wage effects, see Leete 1998a.
438 L. Leete J. of Economic Behavior Org. 43 2000 423–446
Table 4 Estimates of race and gender wage differences within sector of employment
a
Wage differences relative to white males own sector For-profit
Nonprofit White men
Black men −0.155
−0.095
b
Hispanic men −0.138
−0.078
b
Asian men −0.143
−0.111
b
Other men −0.113
−0.062
b
White women −0.277
−0.198
b
Black women −0.288
−0.186
b
Hispanic women −0.310
−0.217
b
Asian women −0.303
−0.195
b
Other women −0.289
−0.210
b
Controls included in specification Education
Yes 17 Potential experience
Yes Potential experience
2
Yes MSAPMSA
Yes 367 Not fluent in English
Yes Part-time work status
Yes Occupationindustry cells
Yes 46933 N
4145608 R
2
0.358
a
Private sectors workers in the 1990 PUMS, not disabled or enrolled in school.
b
For-profitnonprofit difference significant at 0.05 level or higher F-test.
the sectors are statistically significant. These results are supported by similar findings by Preston 1990 and Shackett and Trapani 1987.
As was suggested above, wage equity generated by the motivational needs of the organi- zation might be expected to gain greater expression in the wages of white-collar workers.
The analysis presented in Table 4 is repeated by broad occupational classification to fur- ther investigate this contention. The results are displayed in Table 5. The relative race and
gender wage equity in nonprofits so apparent in Table 4, is in fact limited to white collar occupations.
3.5. Alternate explanations While the greater wage equity and the diminished race and gender differences apparent
in nonprofit wages are supportive of the view that nonprofits use wage equity to provide appropriate motivational conditions, there are several classes of alternate explanations that
must be considered as well. First, the greater reliance of nonprofit organizations on government funding may increase
the requirement that they pursue affirmative action in hiring and promotion. This in turn could account for the diminished race and gender wage differences within nonprofits. A
L. Leete J. of Economic Behavior Org. 43 2000 423–446 439
Table 5 Estimates of race and gender wage differences within sector of employment by occupation group
a
Wage differences relative to white males own sector Managerial and
professional Technical sales and
administrative Service
For-profit Nonprofit
For-profit Nonprofit
For-profit Nonprofit
White men Black men
−0.201 −0.047
b
−0.202 −0.121
−0.110 −0.126
Hispanic men −0.155
−0.093
b
−0.180 −0.147
−0.061 −0.057
Asian men −0.121
−0.095 −0.207
−0.119
b
−0.127 −0.165
Other men −0.177
−0.017
b
−0.154 −0.082
b
−0.065 −0.052
White women −0.296
−0.193
b
−0.273 −0.180
−0.207 −0.188
b
Black women −0.336
−0.187
b
−0.284 −0.175
b
−0.225 −0.195
b
Hispanic women −0.330
−0.217
b
−0.309 −0.188
b
−0.227 −0.261
Asian women −0.319
−0.182
b
−0.319 −0.203
b
−0.180 −0.182
Other women −0.358
−0.191
b
−0.293 −0.191
b
−0.217 −0.235
Controls included in specification Education
Yes Yes
Yes Potential experience
Yes Yes
Yes Potential experience
2
Yes Yes
Yes MSAPMSA
Yes Yes
Yes Not fluent in English
Yes Yes
Yes Part-time work status
Yes Yes
Yes Occupationindustry cells
Yes 9859 Yes 12541
Yes 3347 N
923128 1314486
473564 R
2
0.321 0.305
0.184 Farm, forestry
and fishing Precision, craft
and repair Operators, fabricators
and laborers For-profit
Nonprofit For-profit
Nonprofit For-profit
Nonprofit White Men
Black men −0.188
−0.252 −0.151
−0.165 −0.107
−0.063
b
Hispanic men −0.062
0.108 −0.117
−0.088 −0.105
−0.028
b
Asian men −0.042
−0.202 −0.102
−0.071 −0.125
−0.115 Other men
−0.046 −0.089
−0.091 −0.091
−0.082 −0.041
White women −0.244
−0.205 −0.278
−0.255 −0.269
−0.209
b
Black women −0.311
−0.313 −0.294
−0.378 −0.279
−0.203
b
Hispanic women −0.183
−0.211 −0.385
−0.306 −0.333
−0.265
b
Asian women −0.216
−0.305 −0.331
−0.219 −0.325
−0.307 Other women
−0.194 −0.292
−0.323 −0.138
b
−0.280 −0.236
Controls included in specification Education
Yes Yes
Yes Potential experience
Yes Yes
Yes Potential experience
2
Yes Yes
Yes MSAPMSA
Yes Yes
Yes Not fluent in English
Yes Yes
Yes Part-time work status
Yes Yes
Yes Occupationindustry cells
Yes 617 Yes 9392
Yes 11177 N
99180 549643
785607 R
2
0.131 0.280
0.256
a
Private sectors workers in the 1990 PUMS, not disabled or enrolled in school.
b
For-profitnonprofit difference significant at 0.05 level or higher F-test.
440 L. Leete J. of Economic Behavior Org. 43 2000 423–446
related explanation derives from the fact that nonprofit organizations often rely for revenue on long-term relationships with outside funders. Their continued viability depends in part
on their public reputation, which may include their reputation as a fair employer. Thus, reputational considerations could lead nonprofit organizations to pursue a more equitable
wage structure, both generally and along the lines of race and gender. However, one would expect either of these explanations to manifest themselves across the occupational structure.
This is inconsistent with the limitation of more equitable nonprofit wages to white-collar employees demonstrated above.
A second possible set of explanations for the findings here with regard to malefemale wage differences could stem from the occupational mix of the sectors and the effect of
occupational segregation on wages. Men in the nonprofit sector are relatively more con- centrated in ‘traditionally female’ occupations than are men in the for-profit sector. The
average male nonprofit employee works in an occupationindustry cell that is 79 percent female, while his for-profit counterpart works in an occupationindustry cell that is only
41 percent female.
18
As is well known, wages fall as the percent female in an occupation rises see Leete, 1998b, for a summary of this literature. This effect could cause the wages
of men in nonprofits to be relatively lower, and gender wage differences in the nonprofit sector to be diminished. I investigate this contention with three alternate specifications of
Eq. 6:
ln W = a + bY + cForprof × Race × Gender + dNonprof × Race × Gender + ε 6a
ln W = a + bY + cForprof × Race × Gender +dNonprof × Race × Gender + ePct Fem + ε
6b ln W = a + bY + cForprof × Race × Gender + dNonprof
×Race × Gender + ePct Fem + f Pct Fem × Female + ε 6c
In Eq. 6a, the variable vector Y is substituted for the vector X in Eq. 6. Y includes controls for 3-digit occupation and industry categories only. This equation provides a basis for
comparison for the next one. In Eq. 6b, the variable Pct Fem is added to the specification to measure the percent of employment in each occupationindustry cell that is female.
19
If the results here are in fact due to occupational mix, Eq. 6b should not generate the gender wage
differences shown in Tables 4 and 5. Finally, while nonprofit men are disproportionately employed in ‘traditionally female’ occupations, it is also possible that men earn less in those
occupations than comparable women do. This could result from either worse performance by men or discrimination against men in these occupations. To account for this possibility,
in Eq. 6c an additional interaction term, Pct Fem×Female, is added. The results of these three estimations are shown in column 1 through 6 in Table 6. In each specification, the
18
Author’s calculations from the 1990 PUMS.
19
Leete 1998b discusses how this kind of ‘percent female’ specification differs from specifications that are more traditional.
L. Leete J. of Economic Behavior Org. 43 2000 423–446 441
Table 6 Estimates of race and gender wage differences within sector of employment controlling for percent female in
occupationindustry cell
a
Wage differences relative to white males own sector 1
2 3
4 5
6 For-profit
Nonprofit For-profit
Nonprofit For-profit
Nonprofit White men
Black men −0.160
−0.095
b
−0.160 0.096
b
−0.159 −0.095
b
Hispanic men 0.144
0.082 0.144
0.083
b
−0.143 −0.084
b
Asian men 0.151
−0.130 0.149
−0.128 −0.148
0.128 Other men
−0.118 −0.064
b
−0.118 −0.066
b
−0.118 −0.067
b
White women −0.286
−0.203
b
−0.276 −0.197
b
−0.301 −0.227
Black women −0.301
−0.193
b
−0.291 0.187
b
−0.316 −0.218
b
Hispanic women −0.322
−0.226
b
−0.312 −0.220
−0.338 −0.251
b
Asian women 0.316
−0.201
b
−0.307 −0.194
b
−0.332 −0.225
b
Other women −0.301
−0.213
b
−0.290 −0.207
b
−0.315 −0.237
Controls included in specification Education
Yes 17 Yes 17
Yes 17 Potential experience
Yes Yes
Yes Potential experience
2
Yes Yes
Yes MSAPMSA
Yes 367 Yes 367
Yes 367 Not fluent in English
Yes Yes
Yes Part-time work status
Yes Yes
Yes Occupationindustry cells
Occupation Yes 490
Yes 490 Yes 490
Industry Yes 234
Yes 234 Yes 234
Percent female in occupation industry cell
Yes Yes
Female×percent female in occupationindustry cell
Yes N
4145608 4145608
4145608 R
2
0.343 0.343
0.343
a
Private sectors workers in the 1990 PUMS, not disabled or enrolled in school.
b
For-profitnonprofit difference significant at 0.05 level or higher F-test.
relative race and gender wage effects are virtually unchanged from the original findings. Thus, none of these possible explanations account for the existing pattern of race and gender
wage differences in nonprofit organizations. A third group of possible explanations hinges on the crudeness of the measures of human
capital used here. Only the most basic measures are included in Census data: e.g. age, education, and language fluency. Of course, if different demographic groups exhibit different
levels of unobservable to the researcher but productivity relevant characteristics, measures of race and gender wage differences can mistakenly reflect these differences. Here, a relevant
possibility is that women and non-whites in the nonprofit sector have higher levels of unmeasured human capital relative to white males, than do women and non-whites in
the for-profit sector. If this were the case, it could account for the observed pattern of
442 L. Leete J. of Economic Behavior Org. 43 2000 423–446
wage differences. This might result if nonprofit firms had more family friendly or less discriminatory policies, allowing women and racial minorities there to accumulate more
actual experience than comparable women in the for-profit sector. To empirically test this possibility, I estimate a final set of variants of Eq. 6 in which the returns to potential
experience, education and language fluency are allowed to vary first by gender and nonprofit status, and then by race and nonprofit status. The resulting patterns of wage differentials
are unchanged from those presented in Table 4.
20
Finally, it is possible that the observed wage patterns reflect differences between the nonprofit and for-profit sector that are not measured here. Some authors have suggested
that, as a result of the non-distribution constraint, nonprofit managers operate with greater managerial discretion than for-profit managers e.g. Feldstein, 1971. As discussed above,
many suspect that attitudes and preferences differ systematically between the sectors. If this difference in preferences included a greater value being placed on wage equity in the non-
profit sector, then the increase in managerial discretion there might be used to accomplish this. It is unclear, however, why a greater desire for equity would gain expression only in
white-collar occupations and not throughout the occupational structure. Similarly, differ- ences in the distribution of wages could result from other systematic differences between
the sectors. If nonprofit and for-profit organizations adopt and use technology differently, this could have implications for their wage structures, even after controlling for detailed
occupations and industries.
4. Summary and implications