male-female earnings differentials in canada

10
The Board of Regents of the University of Wisconsin System Male-Female Earnings Differentials in Canada Author(s): R. A. Holmes Source: The Journal of Human Resources, Vol. 11, No. 1 (Winter, 1976), pp. 109-117 Published by: University of Wisconsin Press Stable URL: http://www.jstor.org/stable/145077 . Accessed: 08/05/2014 12:07 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . University of Wisconsin Press and The Board of Regents of the University of Wisconsin System are collaborating with JSTOR to digitize, preserve and extend access to The Journal of Human Resources. http://www.jstor.org This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PM All use subject to JSTOR Terms and Conditions

Upload: r-a-holmes

Post on 09-Jan-2017

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Male-Female Earnings Differentials in Canada

The Board of Regents of the University of Wisconsin System

Male-Female Earnings Differentials in CanadaAuthor(s): R. A. HolmesSource: The Journal of Human Resources, Vol. 11, No. 1 (Winter, 1976), pp. 109-117Published by: University of Wisconsin PressStable URL: http://www.jstor.org/stable/145077 .

Accessed: 08/05/2014 12:07

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

University of Wisconsin Press and The Board of Regents of the University of Wisconsin System arecollaborating with JSTOR to digitize, preserve and extend access to The Journal of Human Resources.

http://www.jstor.org

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 2: Male-Female Earnings Differentials in Canada

Communications 1 109

MALE-FEMALE EARNINGS

DIFFERENTIALS IN CANADA*

I. INTRODUCTION

Male-female earnings differentials in Canada are substantial. These differentials may reflect productivity differences; they may also provide evidence of discrimi- nation that is quite unrelated to productivity differences, either real or per- ceived. Previous studies [1, 3, 8] have recognized the possibility of both wage and employment discrimination against females. Wage discrimination occurs when male-female wage differentials exceed productivity differentials in given occupations, while employment discrimination involves denial to females of access to higher paying occupations. Recent estimates suggest that after allowing for productivity differentials, U.S. females, because of wage and particularly employment discrimination, earn only three-fifths as much as U.S. males [8].

The purpose of this study is to make comparable estimates for Canada. The data employed are from the Survey of Consumer Finances (which is based on the Labour Force Survey) for 19671 and includes all persons surveyed who worked either full time or part time. Our total sample size is over 30,000 persons-19,652 males and 10,367 females. The modal education class for both sexes is "some high school," and the education range "some elementary" through "high school" includes about 88 percent of both males and females.

II. GROSS SALARY DIFFERENTIALS

A regression model is used to measure the gross male-female earnings differen- tials. This model is fitted separately for males and females in each of seven education classes (no education, some elementary, elementary, some high school, high school, some university, and university). Thus we obtain 14 regres- sion equations with earnings as the dependent variable and in which we employ the following independent variables: (1) age; (2) weeks worked; (3) region (Ontario, Maritimes, Quebec, Prairies, British Columbia); (4) residence (metro-

* The author gratefully acknowledges the assistance of the Canada Council, Statistics Canada, and Simon Fraser University in the preparation of this study. Ms. Ruth Simonton and Dr. Nicholas Skoulas of Statistics Canada provided essential assistance in the analysis of the data. Helpful comments were also obtained from the reviewers and from the Journal editor. [Manuscript received November 1974; accepted April 1975.]

1 The Survey of Consumer Finances is conducted by the Consumer Finance Research Division of Statistics Canada and is based on the same sample as the Labour Force Survey.

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 3: Male-Female Earnings Differentials in Canada

110 I THE JOURNAL OF HUMAN RESOURCES

politan, other city, urban, rural); (5) immigration status (unknown, landed before 1945, landed 1946-64, landed 1965 or later, born in Canada); (6) class of worker (paid worker, self-employed nonfarm, self-employed farm, unpaid family worker); (7) marital status (married, single, other); (8) occupation (service and recreation, professional and technical, clerical, sales, managerial, transportaion and communication, farmers and farm workers, loggers and fishermen, miners, craftsmen and related, laborers); (9) nature of work (full time, part time).

These variables are transformed in our model to obtain constant terms that are equal to overall average earnings for the given education-sex class, and dummy variable coefficients that are earnings differentials from the overall average. Our dependent variable "earnings" includes wages and salaries before deductions and net income (that is, gross income less operating expenses) from nonfarm and farm self-employment. The dependent variable in each of the 14 education-sex classes is taken to be a quadratic function of age, a linear function of weeks worked, and step functions of the dummy variables reflecting the various other characteristics previously described (region, residence, immigration status, class of worker, marital status, occupation, and nature of work).

Average age-earnings profiles are obtained from these regression equations. In order to obtain from each nonlinear profile one figure to be used for compara- tive purposes, we estimate potential lifetime earnings in each education-sex subgroup by first adjusting average earnings in each year for a 4 percent real discount rate and a 2.5 percent real growth rate, and then summing these adjusted earnings over potential working years up to retirement.2 These poten- tial lifetime earnings are based not on actual working life, but on the working life if employed full time from the completion of the education level to retirement at age 65.

Because of the relatively small sample sizes and the questionable validity of some of the data for those with "no education," little significance is found in their regressions, and we subsequently ignore this education category. In all other education categories for both sexes, the age, age squared, and weeks worked variables are highly significant; significant differences are also found within almost all the dummy variable sets of classes.3

Table 1 shows gross differentials in potential lifetime earnings. Female earnings are about 35 to 40 percent of male earnings for education levels below high school graduation and about 45 to 50 percent for high school graduation or higher levels of education. The potential lifetime earnings of the female univer-

2 The 4 percent real discount rate corresponds to an 8 percent monetary rate if inflation accounts for four percentage points, or, in our model, if the estimated 2.5 percent real rate of growth in age-income profiles converts to a monetary rate of about 6.4 percent. No adjustment has been made for mortality rates since they are not available by education level.

3 With the dummy variables, the tests of significance are on the differences from the first variable shown for the class.

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 4: Male-Female Earnings Differentials in Canada

Communications I 111

TABLE 1 PRESENT VALUE OF POTENTIAL LIFETIME EARNINGS,

FULL-TIME AND PART-TIME WORKERS, 1967

(4 Percent Real Discount Rate and 2.5 Percent Real Growth Rate)

Potential Lifetime Earnings

Education Level Males Females Females/Males

Some elementary $122,348 $ 45,092 .37 Elementary 157,597 52,450 .33 Some high school 173,464 66,873 .39 High school 212,545 96,759 .46 Some university 234,524 120,357 .51 University 351,635 169,327 .48

TABLE 2 PRESENT VALUE OF POTENTIAL LIFETIME EARNINGS,

FULL-TIME WORKERS, 1967

(4 Percent Real Discount Rate, 2.5 Percent Real Growth Rate)

Potential Lifetime Earnings

Education Level Males Females Females/Males

Some elementary $128,174 $ 57,810 .45 Elementary 161,610 67,875 .42 Some high school 182,130 85,738 .47 High school 216,082 112,675 .52 Some university 240,348 141,535 .59 University 353,836 190,334 .54

sity graduate is slightly less than those of the average male high school dropout. Overall, female potential lifetime earnings are about 41 percent of those for males.

III. NET SALARY DIFFERENTIALS

Gross salary differentials are taken by some to provide convincing evidence of the rankest form of discrimination against females.4 This conclusion may be

correct, but some part of the gross differential is probably due to "productivity"

4 See, for example, Bergmann and Adelman [3].

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 5: Male-Female Earnings Differentials in Canada

112 | THE JOURNAL OF HUMAN RESOURCES

TABLE 3 ADJUSTMENT OF FEMALE ANNUAL EARNINGS BY CHARACTERISTIC, 1967

Some Some High High Some

Characteristic Elementary Elementary School School University University

Weeks worked $-211 $-497 $-483 $ -451 $-346 $ -840 Region 74 -9 23 -10 -54 -50 Residence 156 84 51 -7 -59 -73 Immigration -43 -32 -5 3 -4 25

status Class of 21 23 -18 57 -28 -488

worker Marital -27 -61 -148 -236 -128 -736

status Occupation -141 -282 -302 -501 -298 -245 Total -171 -774 -882 -1145 -917 -2407

differences; we attempt to identify and measure the effect of some of these factors.

One such factor is the greater preference of women, particularly married women, for part-time work. In order to estimate male-female salary differentials net of this difference in part-time work, the regressions were run again but including only persons who were employed full time. The effect of this adjust- ment is to raise female potential lifetime earnings from a range of 33 to 51 percent (Table 1) to a range of 42 to 59 percent (Table 2) of male potential lifetime earnings. Overall, allowing for the greater extent of part-time employ- ment of women raises female lifetime earning potential as a percentage of the corresponding male figure by eight percentage points-from 41 to 49 percent.5

In addition to the greater frequency of part-time work, females differ from males with respect to other explanatory variables in the model. For example, full-time female employees tend to work fewer weeks per year, and they tend to be employed in lower paying occupations. In order to adjust for the effect of these factors, we estimate the change in the male age-income profiles result- ing from substitution in the male regression of the female for the male characteristics on weeks worked and the various dummy variables. This substitu- tion enables us to determine the potential lifetime earnings of females if they had been paid according to male age-income profiles, but with adjustments made for female-male differences in average weeks worked as well as occupation, marital status, class of worker, and the other dummy variables included in the model.6 These adjustments are shown in Table 3.

5 These averages are weighted by the number of females in the various education classes. 6 This approach is similar to that employed in Malkiel and Malkiel [5] and in Oaxaca [7].

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 6: Male-Female Earnings Differentials in Canada

Communications 1 113

The distribution of (full-time employed) females by region, residence, and immigration status does not differ significantly from that for males, so that the impact of those characteristics on expected annual earnings of females is negligi- ble. The class-of-worker variable also has a small effect overall ($-55), but for females with university degrees, the distribution by class of worker does substan- tially reduce potential earnings. This reflects the much smaller proportion of female than male university graduates found in the highly paid self-employed classes (doctors, dentists, lawyers, architects, etc.). Almost all female university graduates are paid employees and, as a result, their expected annual earnings are reduced below those of males by some $488.

Larger negative effects on potential female earnings are observed from the marital status, occupation, and weeks-worked variables. Marital status is prob- ably considered by employers as an indicator of stability and reliability in employment.7 Married women may be expected to have higher absence and turnover rates than single men, but conversely for men. In all education cate- gories of our sample of full-time employees, a larger proportion of women than men were single, so that if they were paid according to male rates, females would, because of their distribution by marital status, expect on average to earn less than males. The annual differential due to marital status is highest for females with university degrees ($-736); the overall effect is $-182.

Females choose and are forced into different and generally lower paying occupations than men. A smaller concentration of females than males is found in the highly paid managerial occupations, and larger proportions of females are found in the relatively low paid clerical, service, and recreation occupations. The female occupational distribution changes their potential annual earnings as compared to males by $-141 to $-501 for the various education classes and by $-343 overall.

The largest negative effect on expected annual earnings by females employed full time results from fewer weeks worked per year. This effect at male rates tends to be greater at higher education levels and ranges from $-211 to $-840. Overall, working fewer weeks per year reduces female potential annual earnings below that for males by $455.

The effect of all of these female characteristics is much the smallest for those in the lowest education class and much the largest for those at the highest education level. Females with some elementary education should expect, as a result of their differences from males with respect to class of worker, marital status, occupation, weeks worked, as well as region, residence, and immigration status, to earn annually some $171 less than males. The differential for female university graduates is $-2407, and for those with intermediate levels of educa- tion, the range is from $-882 to $-1145. The effect of these annual differentials on potential lifetime earnings is shown in Table 4.

7 Bowen and Finegan [4].

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 7: Male-Female Earnings Differentials in Canada

114 I THE JOURNAL OF HUMAN RESOURCES

TABLE 4 PRESENT VALUE OF POTENTIAL LIFETIME EARNINGS DERIVED FROM

MALE REGRESSIONS, FULL-TIME WORKERS, 1967 (4 Percent Real Discount Rate, 2.5 Percent Real Growth Rate)

Potential Lifetime Earnings

Education Level Males Females Females/Males

Some elementary $128,174 $122,629 .96 Elementary 161,610 136,108 .84 Some high school 182,130 157,277 .86 High school 216,082 189,535 .88 Some university 240,348 219,556 .91 University 353,836 316,304 .89

Table 5 relates the actual female lifetime earnings potential as derived from regressions on female data to what females would have earned had they been paid at male rates, but adjusted for differences in weeks worked, occupation, marital status, class of worker, as well as region, residence, and immigration status. The difference is substantial, ranging from 45 percent for females with some elementary education to 64 percent for those with some university education. Overall, potential female lifetime earnings are only 56 percent of those they would obtain from adjusted male rates of pay.

IV. CONCLUSION

Our analysis reveals that female earnings as a percentage of male earnings varies directly with education level. The gross female/male earnings ratio varies from about .35 to .40 for education levels up to some high school and from .45 to .50 for higher education levels (Table 1). After all adjustments have been made, these ratios range from about .45 to .55 for the lower education levels and from about .60 to .65 for the higher education levels (Table 5). These differences imply that with higher levels of education, females are better able to compete with males for higher paying jobs and are less likely to be forced into over- crowded occupations where their productivity and rate of pay is low.

Overall, the potential lifetime earnings of full- and part-time female workers in Canada have been shown to average only 41 percent of male potential earnings. With these gross differentials, which include both full- and part-time workers, the potential lifetime earnings of a female university graduate is about equal to that of a male high school dropout. If we remove the effect of greater numbers of part-time workers among females, the female-male lifetime earnings

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 8: Male-Female Earnings Differentials in Canada

Communications I 115

TABLE 5 PRESENT VALUE OF FEMALE POTENTIAL LIFETIME EARNINGS,

FULL-TIME WORKERS, 1967 (4 Percent Real Discount Rate, 2.5 Percent Real Growth Rate)

Female Adjusted Male Education Level Regression Regression Ratio

Some elementary $ 57,810 $122,629 .47 Elementary 67,875 136,108 .50 Some high school 85,738 157,277 .55 High school 112,675 189,535 .59 Some university 141,535 219,556 .64 University 190,334 316,304 .60

potential rises from 41 to 49 percent; and if we adjust for other factors included in our model (principally weeks worked, occupation, marital status, and class of worker), it rises to 56 percent.

Sawhill [8] has found that in the U.S., female/male earnings ratios increased from .46 to .57 after allowance for "productivity" differences between the sexes. If we had followed Sawhill's procedure of treating all occupational differences between males and females as employment discrimination, our ulti- mate female/male earnings ratio would have been .54 rather than .56. These results are quite similar but suggest somewhat greater discrimination against women in Canada than in the U.S.

About one-fourth (15 of 59 percentage points) of the gross differential in male-female earnings has been accounted for by the worker characteristics considered in this study. With allowance for these characteristics, the potential lifetime earnings of the female university graduate rises from the level of the male high school dropout to that of the male high school graduate. Although the amount of explanation is substantial, fully three-quarters of the gross differential remains unexplained.

This unexplained component of the gross differential might be attributed to the effects of wage and employment discrimination, which would put discrimi- nation against females in Canadian labor markets at 44 percent of male annual earnings rates. However, moving from quantitative results such as those obtained in this study to a measure of discrimination is more difficult than is sometimes assumed, because discrimination means different things to different people and and thoughtful definition includes factors whose effects cannot be measured objectively.

Of course, if markets are perfect, employers are good profit maximizers, and one takes female attitudes as given, then discrimination cannot exist, since under these conditions all wage differences will reflect productivity differences. Few

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 9: Male-Female Earnings Differentials in Canada

116 I THE JOURNAL OF HUMAN RESOURCES

would argue that this model is realistic, but many would recognize a variety of factors that tend to justify some male-female wage differentials and that, if taken into account in our model, would tend to reduce the discrimination estimate below the 44 percent figure obtained. For example, some employers are reluctant to train females for more responsible and high paying positions because of the greater expectation that single females will marry and leave the labor force and that married females will either retire early or leave the area if the husband is transferred. There is relatively high risk involved for employers who attempt to identify females with genuine long-term career interests, and for this reason some would conclude that females are justifiably denied access to better jobs involving relatively high training costs for the employer. Similarly, it can be logically argued that women who enter the labor force after 15 or more years of marriage, at a time when much of their education or training for employment is obsolete and with limited job experience and reduced years of potential service,8 cannot expect employers to see them as likely candidates for the more challeng- ing, rewarding, and responsible positions. When one adds to this the fact that most women are unsuitable for, and find unacceptable, some high paying positions (logging, mining, etc.), then one can logically conclude that full allowance for "productivity" differentials would reduce the discrimination esti- mate well below the 44 percent figure obtained in this study.

There is, however, another side to the story. To some in the women's liberation movement, almost all role differentiation both in the home and in the job market involves discrimination against women. According to their view, discrimination in Canada is even higher than the 44 percent differential left unexplained in our analysis, for "discrimination" includes a large part of the differential we have attributed to weeks worked, class of worker, and occupa- tion. The consignment of women to boring jobs with low pay and low penalties for turnover may well be a factor in the fewer weeks worked by females; in the minds of some, it is a factor of primary importance. Employment discrimination against females may also underlie a large part of male-female differences in class of worker and even broad categories of occupation, particularly if one takes the position that cultural factors that shape female job preferences are simply a disguised form of discrimination. These ideas would lead some to conclude that wage and employment discrimination against women in Canada is closer to the 59 percent gross differential than to the 44 percent figure obtained after making allowance for male-female differences in "productivity."

Consequently, whether one finds discrimination against females in Canadian labor markets higher or lower than the 44 percent of male earnings rates left unexplained by the worker characteristics included in our model depends on one's definition of discrimination. If discrimination is defined to include all

8 Sawhill [8] has found only minor differences in the salary structures of never-married and other females, which suggests that not much of male-female differentials can be attributed to these factors.

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions

Page 10: Male-Female Earnings Differentials in Canada

Communications I 117

male-female role differentiation, whatever its source, then it will be substantially higher and close to the 59 percent gross differential in male-female earnings. At the other extreme, if one takes cultural factors and female preferences as given and assumes that employers discriminate against females only the extent that their hiring and employing policies are not profit-maximizing, then discrimina- tion will be only a fraction of our 44 percent figure.

R. A. HOLMES Simon Fraser University

REFERENCES

1. Gary S. Becker. The Economics of Discrimination. Chicago: University of Chicago Press, 1957.

2. Barbara R. Bergmann. "The Effect on Whites' Income of Discrimination in Employment." Journal of Political Economy (March/April 1971).

3. Barbara R. Bergmann and Irma Adelman. "The 1973 Report of the Presi- dent's Council of Economic Advisers: The Economic Role of Women." American Economic Review 63 (September 1973): 509-14.

4. W. G. Bowen and T. A. Finegan. The Economics of Labor Force Participa- tion. Princeton, N.J.: Princeton University Press, 1969.

5. B. C. Malkiel and J. A. Malkiel. "Male-Female Pay Differentials in Profes- sional Employment." American Economic Review 63 (September 1973): 693-705.

6. James N. Morgan. A Note on the Interpretation of Multiple Regression Using Dummy Variables. Ann Arbor: University of Michigan, 1964.

7. R. L. Oaxaca. Male-Female Wage Differentials in Urban Labor Markets. Princeton, N.J.: Industrial Relations Section, Princeton University, 1971. Working Paper No. 23.

8. Isabel V. Sawhill. "The Economics of Discrimination Against Women: Some New Findings." Journal of Human Resources 8 (Summer 1973): 383-96.

9. Robert E. Sweeney and E. F. Ulveling. "A Transformation for Simplifying the Interpretation of Coefficients of Binary Variables in Regression Analy- sis." American Statistician (December 1972): 30-32.

This content downloaded from 169.229.32.137 on Thu, 8 May 2014 12:07:55 PMAll use subject to JSTOR Terms and Conditions