age discrimination w against natives - queen's...

21
CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV, NO. 1 1999 Wage Discrimination Against Natives ARNOLD DE SILVA University of Ottawa Ottawa, Ontario Cet article utilise les données du recensement de 1991 pour examiner la discrimination salariale chez les salariés autochtones travaillant à temps partiel ou à temps plein et vivant ou non sur des réserves. L’analyse montre que l’écart salarial entre blancs et autochtones est principalement attribuable à des différences de dotations, ce qui implique une discrimination potentielle relativement mineure. Elle montre également que, contrairement à nos attentes, les travailleurs autochtones dont les origines sont exclusivement aborigènes ne connaissent pas toujours moins de discrimination potentielle que les travailleurs autochtones ayant des origines mixtes. This paper examines wage discrimination facing native paid employees working both full-time and part- time, living on and off reserves, using the 1991 census data. It finds that the native-white wage differential is mainly attributable to endowment differences, which implies that potential discrimination is relatively minor. It also finds that, contrary to a priori expectations, native workers of exclusively aboriginal origin do not always experience less potential discrimination than native workers of mixed ethnicity. INTRODUCTION T he 1995 Employment Equity Act has designated Aboriginal peoples as one of the four economi- cally disadvantaged groups in Canadian society. The other three groups are women, visible minorities, and persons with disabilities. A preliminary glance at the wage and employment profiles contained in the annual reports of the Employment Equity Act provides considerable support for the designation of natives as a disadvantaged group. 1 According to the raw data, Aboriginal people have a lower labour- force participation rate, a higher rate of unemploy- ment, less representation in higher paying occupa- tions, and not surprisingly, lower average wage rates than other workers. Although wages are only one aspect of labour- market performance, comparisons based on wage rates are widely used to describe the labour-market disadvantages of paid employees in the designated groups. There are already several studies document- ing the wage disadvantage experienced by many of the designated groups. In particular, gender-based wage differentials have received a great deal of at- tention, for example, Gunderson (1989); Shapiro and Stecner (1997); Miller (1987). Similarly, the wage disadvantage experienced by visible minorities and disabled persons has also been discussed in several recent studies, for example, De Silva (1992); Christofides and Swidinsky (1994); Pendakur and Pendakur (1995); Hum and Simpson (1996); Harkness (1993). In addition, a few studies have looked at wage gap changes affecting all designated groups, such as Leck et al. (1995). By contrast, there has been relatively little re- search dealing specifically with the labour-market difficulties of natives. The most rigorous treatment of wage differentials between natives and whites is a recent study by George and Kuhn (1994). 2 That study, which was based on the 1986 census data, concentrated mainly on those natives working full- time and full year, living off reserves and outside

Upload: others

Post on 28-Jul-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives65

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

Wage Discrimination Against NativesARNOLD DE SILVA

University of OttawaOttawa, Ontario

Cet article utilise les données du recensement de 1991 pour examiner la discrimination salariale chez les salariésautochtones travaillant à temps partiel ou à temps plein et vivant ou non sur des réserves. L’analyse montre quel’écart salarial entre blancs et autochtones est principalement attribuable à des différences de dotations, ce quiimplique une discrimination potentielle relativement mineure. Elle montre également que, contrairement à nosattentes, les travailleurs autochtones dont les origines sont exclusivement aborigènes ne connaissent pas toujoursmoins de discrimination potentielle que les travailleurs autochtones ayant des origines mixtes.

This paper examines wage discrimination facing native paid employees working both full-time and part-time, living on and off reserves, using the 1991 census data. It finds that the native-white wage differentialis mainly attributable to endowment differences, which implies that potential discrimination is relativelyminor. It also finds that, contrary to a priori expectations, native workers of exclusively aboriginal origin donot always experience less potential discrimination than native workers of mixed ethnicity.

INTRODUCTION

The 1995 Employment Equity Act has designatedAboriginal peoples as one of the four economi-

cally disadvantaged groups in Canadian society. Theother three groups are women, visible minorities,and persons with disabilities. A preliminary glanceat the wage and employment profiles contained inthe annual reports of the Employment Equity Actprovides considerable support for the designationof natives as a disadvantaged group.1 According tothe raw data, Aboriginal people have a lower labour-force participation rate, a higher rate of unemploy-ment, less representation in higher paying occupa-tions, and not surprisingly, lower average wage ratesthan other workers.

Although wages are only one aspect of labour-market performance, comparisons based on wagerates are widely used to describe the labour-marketdisadvantages of paid employees in the designatedgroups. There are already several studies document-

ing the wage disadvantage experienced by many ofthe designated groups. In particular, gender-basedwage differentials have received a great deal of at-tention, for example, Gunderson (1989); Shapiro andStecner (1997); Miller (1987). Similarly, the wagedisadvantage experienced by visible minorities anddisabled persons has also been discussed in severalrecent studies, for example, De Silva (1992);Christofides and Swidinsky (1994); Pendakur andPendakur (1995); Hum and Simpson (1996);Harkness (1993). In addition, a few studies havelooked at wage gap changes affecting all designatedgroups, such as Leck et al. (1995).

By contrast, there has been relatively little re-search dealing specifically with the labour-marketdifficulties of natives. The most rigorous treatmentof wage differentials between natives and whites isa recent study by George and Kuhn (1994).2 Thatstudy, which was based on the 1986 census data,concentrated mainly on those natives working full-time and full year, living off reserves and outside

Page 2: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

66 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

the Yukon and the Territories. For this group, thestudy reported several interesting results. First, theraw wage differential between whites and those re-porting any aboriginal origins is only about 11 per-cent, which is quite low compared with, for exam-ple, a white male-female wage differential of 35percent. Second, women with any aboriginal ori-gins face a lower wage gap relative to the whitesthan their male counterparts — 7 percent versus 12percent. Third, those who are exclusively aborigi-nal in origin have a much larger wage gap relativeto whites than those reporting any aboriginal ori-gins — 18 percent versus 11 percent. Fourth, en-dowment differences account for a large portion ofthe native-white wage differential in the case ofthose who are exclusively aboriginal in origin —ranging from 42 percent to 59 percent in the case ofmen, and from 46 percent to 53 percent in the caseof women. Thus it would appear that potential dis-crimination plays a relatively minor role in theexplanation of native-white differences at least forthis group of Aboriginal people.3 Finally, in someexploratory research, the authors found that nativesliving on reserves tend to earn considerably less thanthose off-reserves — about 14 percent in the caseof men and 8 percent in the case of women. Thewage gap in this case was found to be partly due tothe lack of any detectable return to education onreserves.

The main purpose of the present study is to verifythe above findings on potential discrimination fac-ing Aboriginals, using the 1991 census data and byfocusing on a larger group which includes, in addi-tion to those employed full-time and full year, oth-ers working either part-time or for part of the year.Following George and Kuhn, the sample is restrictedto those in the 15 to 64 age bracket, working as paidemployees, reporting positive earnings, and livingoutside of the Yukon and the Northwest territories.

The native sample used in this study does notdistinguish between those on reserves and thoseoutside, since it is not possible to make the distinc-tion using the Public Use Sample Tape (PUST).4

However, some indication of the relative importanceof these two groups can be obtained from a surveyof Aboriginal people undertaken by StatisticsCanada, using census data (Statistics Canada 1995).This survey is based on a 20 percent sample unlikethe PUST which relies on a 3 percent sample of cen-sus data. According to this survey, natives living off-reserves represent the majority, accounting for about81 percent of the native population.5

THE ABORIGINAL POPULATION LIVING ON OR

OFF RESERVES AND COMPARISON WITH

WHITES

In 1991, there were 621,535 working-age natives inCanada,6 representing 3.4 percent of the overallworking-age population. Although the present studyis unable to distinguish between natives living onand off reserves, it is still important to provide somebackground to show how the on-reserve populationdiffers from those living outside the reserves andthe Canadian population at large. Table 1, which isbased on the 20 percent sample of census data, showsthat while natives generally tend to fare poorly inthe labour market relative to the Canadian popula-tion, those living on reserves are considerably worseoff than those living outside the reserves. Forexample, whereas the unemployment rate among theoff-reserve population is about one and a half timesthe national rate, it is three times the national rateon the reserves. Similarly, while the labour-forceparticipation rate of those living outside the reservesis almost on a par with the national rate, those livingon the reserves report a much lower rate. About two-fifths of those living off reserves work full-time andfull year, compared with a little over one-halfnationally. By contrast, in the case of those on re-serves, less than one-third are employed on a full-time, full-year basis. The earnings pattern is alsoconsistent with the above picture. For example,whereas the earnings gap between full-time, full yearemployees outside the reserves and the Canadianpopulation is about 12.5 percent, their on-reservecounterparts face a much larger earnings gap — 35.4

Page 3: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives67

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

TABLE 1Characteristics of Aboriginals Living On and Off Reserves and Comparison with the Canadian Population, 1991

Aboriginals Canadian Population

Total On Reserves Off Reserves

Population (millions) 1.016 0.189 0.827 26.994

(percent of population)

Working-age Pop (15-64) 61.1 57.4 62.0 68.1

Age 32.5 32.7 32.5 37.0

GenderMales 48.6 51.5 47.9 49.4

Education< Grade 9 18.4 36.5 14.4 13.9Grades 9-13 42.8 37.3 44.0 39.0Trade Cert & Dipl 26.5 20.8 27.8 26.3Some University 7.5 4.2 8.3 9.4Degree 4.7 1.2 5.5 11.4

L.Force ActivityFull-time, Full-yr 40.9 30.3 42.4 53.3Part-time, Part-yr 59.1 69.7 57.6 46.7Unempl Rate 19.4 30.7 17.7 10.2Particip Rate 64.3 46.6 68.1 67.9

EarningsFull-time, Full-yr($) 28,755 21,787 29,485 33,714Part-time, Part-yr($) 10,769 8084 11,247 14,430

Govt TransfersAs % of Total Inc. 17.9 37.8 15.4 11.4

OccupationManagers 6.7 6.5 6.8 9.8Professional 8.9 12.7 14.0 17.7Manual Labour 28.3 35.4 27.1 21.6

Source: Statistics Canada(1995, pp. 8-29).

Page 4: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

68 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

percent. Similarly, in the case of those workingeither part-time or for part of the year, those livingoff reserves report an earnings gap (relative to theCanadian population) of 22.1 percent, comparedwith 44 percent for those on reserves. Evidence onthe occupational structure and the dependence ongovernment transfers reinforces the view that thoseliving on reserves are worse off than those offreserves.

A major explanation for the dismal labour-marketperformance of Aboriginal people living on reservescan be found in their relatively low educational at-tainments. While natives in general are less quali-fied than the rest of the population, it is much worsein the case of those residing on reserves. For exam-ple, more than one-third of the on-reserve popula-tion reports an education of less than grade nine,compared with a little over a tenth for those off re-serves. Likewise, whereas about 6 percent of thoseliving off reserves have a university degree, onlyabout 1 percent of those on reserves seem to havesuch qualifications.

We now turn to the mean characteristics of theindividuals used in the present sample, which isdrawn from a 3-percent sample of census data. Asmentioned before, this subsample consists of thosein the 15 to 64 age bracket, working in paid em-ployment, reporting positive earnings, and livingoutside the Yukon and Northwest Territories. Thoseliving in institutions, temporary residents of Canada,and those outside the country are excluded. Table 2presents information on natives and whites. Whitesare defined here as those who are neither aboriginalnor visible minority. George and Kuhn have alsoused a similar definition. The first two columns dealwith all whites and aboriginals in the sample,whereas columns 3 and 4 pertain to aboriginals withsingle and multiple origins respectively. By singleorigins is meant people who are exclusively of abo-riginal ethnicity, whereas multiple origins refer tothose with mixed ethnicity. The last two columnsdeal with aboriginals and whites who had workedfull-time and full year.

An examination of columns 1 and 2 reveals thatboth aboriginal men and women are somewhatyounger than whites and tend to work fewer weeksper year relative to whites. Aboriginal men report aslightly lower proportion of full-time employeesthan whites, but in the case of aboriginal women,there is hardly any difference between themselvesand their white counterparts. The evidence indicatesthe existence of significant native-white earningsgaps — 41 percent and 32 percent for men andwomen respectively.

Other interesting differences to note are that Abo-riginal people are more likely to be single or nevermarried, than whites, and tend to be considerablyless fluent in French and less proficient in Canada’stwo official languages than whites. They also pre-fer to live outside the census metropolitan areas(CMAs) much more than whites. Geographically,whereas three-quarters of whites live in Ontario,Quebec, and British Columbia, a somewhat lowerproportion — about three-fifths — of natives areconcentrated in those three provinces. Almost one-third of the Aboriginal people are located in the Prai-rie provinces, compared with only 16 percent ofwhites.

A comparison of aboriginals with single originswith those reporting multiple origins shows that thelatter tends to be more educated, more proficient inEnglish and French, and is more likely to live inCMAs than the former. Ontario is the preferred prov-ince of residence for one-third of those with multi-ple origins, with Quebec and British Columbia ac-counting for another one-third or so. About one-quarter live in the Prairie provinces. By contrast,the geographical distribution of those with singleorigins is somewhat more even. Ontario, BritishColumbia, and Quebec account for 56 percent, whilethe Prairie provinces account for almost 40 percent.Natives with multiple origins report a somewhathigher proportion of married persons than those withsingle origins. They are also more likely to workfor a greater number of weeks compared with thosewith single origins.

Page 5: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives69

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

TABLE 2Mean Sample Characteristics of Whites and Aboriginals, 1991

Xw XA XA,S XA,M XA,FT,FY XW,FT,FY

(1) (2) (3) (4) (5) (6)

Males

Age (yrs) 37 33 33 32 36 39Earnings ($) 21247 12526 8828 13863 27197 31920Earnings (logs) 9.996 9.436 9.086 9.537 10.211 10.371

Full-time/Part-timeFull-time (%) 88.8 85.8 85.4 86.0 100.0 100.0

Weeks Worked<50wks (%) 38.5 54.7 65.9 47.5 NA NANo: obs (‘000) 180.3 5.6 0.8 1.8 2.6 114.1

Females

Age (yrs) 36 32 33 32 35 38Earnings ($) 11833 8069 6943 9195 19231 22281Earnings (logs) 9.379 8.996 8.845 9.126 9.864 10.011

Full-time/Part-timeFull-time (%) 70.5 71.0 73.7 69.5 100.0 100.0

Weeks Worked<50wks (%) 49.7 58.6 62.4 56.7 NA NANo: obs (‘000) 159.3 5.2 1.8 3.4 2.1 77.8

Both Sexes(as a percent of the population)

ProvinceNewfoundland 2.2 1.6 2.0 1.4 1.1 1.6PEI+NewBruns 3.3 1.4 0.8 1.8 1.3 2.8Nova Scotia 3.4 2.4 1.5 2.8 2.5 3.2Quebec 25.8 16.2 19.3 14.8 18.3 25.9Ontario 37.9 28.8 18.4 35.2 32.9 39.6Manitoba 3.7 10.0 13.5 7.7 9.3 3.7Saskatchewan 3.1 6.8 10.1 4.4 5.3 2.9Alberta 9.4 14.6 15.8 14.4 13.8 9.5British Columbia 11.2 18.2 18.6 17.6 15.5 10.8

Urban LivingNon-CMAs 39.0 53.3 65.9 44.5 46.2 35.1

continued

Page 6: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

70 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

TABLE 2(continued)

Xw XA XA,S XA,M XA,FT,FY XW,FT,FY

(1) (2) (3) (4) (5) (6)

Both Sexes(as a percent of the population)

Marital StatusMarried 54.9 42.5 39.4 44.0 53.3 62.8Divorced 6.7 7.9 8.0 8.8 10.2 7.9Separated 2.9 4.4 5.0 4.4 4.7 3.2Widowed 1.3 1.1 1.8 1.0 1.3 1.3Single 34.2 44.1 45.8 41.8 30.5 24.8

Official LanguageEnglish 65.6 75.8 79.9 72.6 72.6 66.4French 13.1 6.0 8.5 4.8 5.9 12.1Bilingual 21.0 17.7 10.6 22.5 21.4 21.3Allophone 0.3 0.5 1.0 0.1 0.1 0.2

Education<grade 9 6.3 8.7 13.8 4.1 6.9 5.6Grades 9-13 20.9 28.9 31.2 26.2 21.8 17.3Secondary Cert 16.6 13.0 11.4 14.6 14.2 16.8Elem Trade Cert 4.7 4.3 4.6 3.5 5.1 5.3Some Non-university 7.3 9.2 9.6 9.2 8.4 6.7Higher Trade Cert 7.7 8.4 8.0 7.7 10.3 8.6Non-univ Diploma 11.9 10.6 8.7 13.6 12.6 13.3Some University 11.2 10.0 9.2 11.6 10.8 10.2B.A. 10.9 5.9 3.0 8.1 8.2 12.8M.A.+Ph.D 2.5 1.0 0.5 1.4 1.7 3.4

Ethnic OriginSingle Origin NA 37.0 100.0 NA 31.1 NAMultiple Origin NA 63.0 NA 100.0 68.9 NA

Notes: XW, XA refer to the mean characteristics of whites and aboriginals respectively; XA,S and XA,M denote the meancharacteristics of those with single and multiple origins respectively. The last two columns refer to the attributes ofaboriginals and whites working full-time, full year respectively.Source: Public Sample Tape (1991).

Page 7: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives71

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

In terms of earnings, natives with single originsfare poorly relative to those with multiple origins.Native men with single origins earn about 64 per-cent of the earnings of native men with multipleorigins, whereas native women with single originsearn about 75 percent of the earnings of aboriginalwomen with multiple origins. As a result, the earn-ings gap between aboriginals with single origins andwhites is considerably larger than the earnings gapwhich exists between aboriginals with multiple ori-gins and whites. For men, the native-white earningsgap is 58 percent (for those with single origins),compared with 35 percent (for those with multipleorigins). The corresponding estimates on the earn-ings gaps for women are 41 percent and 22 percentfor those with single and multiple origins respec-tively. Thus, aboriginal women seem to experiencea smaller earnings gap relative to whites, irrespec-tive of whether their ethnic origin is single or mul-tiple, which is consistent with the earlier discussion.

When we look at natives working full-time andfull year and compare them with similar white work-ers (columns 5 and 6), we find roughly the samedifferences as those noted earlier. One importantdifference, however, is that the native-white earn-ings gaps facing those who are employed on a full-time and full-year basis are considerably lower thanthose mentioned earlier for the larger group whichincludes, in addition to full-time, full-year employ-ees, those working either part-time or part year. Foraboriginal men, the earnings gap relative to whitesis 15 percent for those working full-time and fullyear, compared with 41 percent for those in thelarger group. Similarly, the native-white earningsgap for women working full-time and full year is14 percent versus 32 percent for those in the largergroup. Thus, the native-white earnings gap facingboth sexes is roughly the same when only full-time,full-year employees are compared, whereas, whenpart-time and part-year employees are also includedin the sample, the previous discussion showed abo-riginal women experiencing a lower earnings gapthan their male counterparts.

CONCEPTUAL FRAMEWORK

The methodology used in this study is based on theOaxaca (1973) and Blinder (1973)7 decompositiontechnique which has also been used by George andKuhn. According to this approach, the observedearnings difference between whites and aboriginalscan be decomposed into the explained and unex-plained components as follows:

1. (lnWw -lnWa )= (Xw-Xa)ßw + Xa(ßw-ßa),

the ßs are the estimated coefficients and a bar overa variable represents a sample mean. The first com-ponent on the right-hand side of the equation — i.e.,( Xw - Xa) represents the contribution of endowmentdifferences while the remaining component on theright-hand side is the unexplained component.8

There are several aspects of this technique thatshould be considered. First, the technique assumesthat the proxies of productivity differences (the Xs)are adequate and exhaustive. If the proxiesmismeasure true productivity, then the problem oferrors in variables occurs. The literature on reverseregression (Kamalich and Polachek 1982;Goldberger 1984) illustrates the gravity of the prob-lem. If the X vector does not exhaust all possibleproductivity-related factors, then there may be omit-ted variables bias. For example, we know very lit-tle, if anything, about possible differences in thequality of education and experience, absenteeism,labour-market turnover, or motivation and abilitybetween natives and whites, yet, these factors maybe important in explaining the earnings differencesbetween these two groups. As a partial solution tothe problem of omitted variable bias some studies,such as that of Daymont and Andrisani (1984), havetried to augment the earnings function with morerefined productivity measures not usually availablein census data. Unfortunately, we are not aware ofany such data we could use in this study. Second, afurther limitation of the decomposition techniqueis that it ignores questions concerning the cause of

ˆ ˆˆ

ˆ

Page 8: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

72 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

any underlying differences in characteristics. Thus,the feedback effects of labour-market discrimina-tion on the acquisition of human capital are oftenoverlooked (Welch 1975; Sloane 1985). Similarly,streaming aboriginals into low-wage, dead-end oc-cupations may be a form of discrimination. Neitherof these aspects is discussed in the present study.However, if as a result of such feedback effects, thecharacteristics (the Xs) turn out to be endogenous,then conventional regression methods may producebiased estimates of potential discrimination(Killingsworth 1993). It is conceivable that, onceall of these adjustments have been made, we mighteven find that a large portion of the potential dis-crimination actually occurs outside the labour mar-ket.9 Because of these caveats, the decompositiontechnique outlined above should be viewed as pro-viding only an upper bound, albeit imperfect, on theamount of the earnings gap attributable to potentialdiscrimination.

Up to now, we have discussed only differencesin the average market wages of whites and aborigi-nals. These are wages observed in the labour mar-ket. However, as has been widely recognized sincethe pioneering work by Gronau (1974), the analysisof discrimination should focus on the entire rangeof wages offered to an individual by employers, andnot only on the wages observed.

The wage offer function can be expressed in semi-logarithmic form as:

2. lnWi = Xiß + εi,

where ε is an error term with a normal distribution.The superscripts for whites and aboriginals havebeen left out for convenience. The expected valueof wages observed in paid employment is given by:

3. E(lnWi |Xi, Sample Selection Rule) =Xiß + E( εi | Sample Selection Rule)

If ( εi | Sample Selection Rule) = 0, then the OLSestimates based on the subsamples of employed

individuals will be unbiased. If the sample of labour-force participants is not random, however, thenE(εi | Sample Selection Rule) is not equal to zero,and the sample of observed wages will not be anappropriate basis to make generalizations about thelevel of potential discrimination. Under such circum-stances, there is selectivity bias which needs to becorrected for.

The risk of selectivity bias is generally inherentin studies that are based on earnings equations ap-plied to a sample that is essentially a subset of alarger parent sample and where the selection crite-rion is not purely random. The present study is noexception because it relies on a sample derived fromthe 1991 census. Although the total number of per-sons in the 15-64 age group is 12,191 natives (i.e.,6,392 males plus 5,799 females) and 383,158 whites(i.e., 208,103 males plus 175,055 females), the OLSsamples are smaller — 10,755 aboriginal observa-tions (5,574 males and 5,181 females) and 339,627white observations (180,349 males and 159,278 fe-males). Thus, a fair number of observations found inthe 1991 census tape have been excluded from the OLSsamples. The main exclusions are those who were self-employed, unemployed persons or those who were outof the labour force for the full year, full-time studentsduring the entire year, those living in institutions, andthose for whom the information was incomplete onsome of the variables of interest. Although these ex-clusions are fairly standard practice, we cannot com-pletely rule out the possibility of self-selection.

Heckman (1979) has outlined a two-step estima-tion technique to correct for selectivity bias. Thefirst stage involves the estimation of a probit equa-tion to determine the process of selecting individu-als into the white and aboriginal samples. The sec-ond stage involves the estimation of a wage equa-tion of the following form:

4. lnWi = Xiß + θλi + εi

where Xß is the expected log wage that may be of-fered to a randomly selected individual in the

Page 9: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives73

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

subgroup, Xß +θλ is the expected observed wagefor the group actually in the sample, λ (referred toas the inverse Mill’s ratio) is the estimate from theprobit equation of the ratio of the density to the dis-tribution function, ε is the remaining portion of theerror term, and θ is the coefficient of λ. If θ turnsout to be zero, then the previous specification with-out the selectivity bias correction applies, and theobserved and offered wages are the same.

When the non-random nature of the samples istaken into account using the Heckman procedure,the earnings differential between whites and abo-riginals can be decomposed as follows:

5. lnWw-lnWa=(Xw-Xa)ßw+Xa(ßw-ßa)+λw θw-λa θa

According to equation (5), the observed earningsdifferential (lnWw-lnWa) is the sum of three com-ponents: the contribution of endowment differences(the first term on the right-hand side); potential dis-crimination (the second term on the right-hand side);and the contribution of selectivity bias differences(the third term on the right).

Equation (5) can be rewritten as:

6. lnWw-lnWa-λwθw-λaθa = (Xw-Xa)ßw+(ßw-ßa)Xw

One can immediately see that the only differencebetween equations (5) and (6) is that in the latter,the portion of the observed earnings differential thatis due to selectivity bias differences has been addedto the observed wage term to provide a measure ofthe difference in the wage offers of whites andAboriginal people.

One other point that needs to be mentioned is that,although the focus of the present analysis is on wagediscrimination, the dependent variable used is an-nual earnings. This is not necessarily a problemsince, as shown later, we have controlled for weeksworked in the regression. However, there could besome ambiguity and confusion in the interpretationof the results since we are comparing a wage gap

derived from a regression (holding weeks workedconstant) with the raw earnings differential.10 Con-sistency, therefore, requires that we do one of twothings. The first option would be to have weeklyearnings as the dependent variable. The other is toinclude weeks worked as an explanatory variable butleave out its contribution from the denominator inassessing the importance of potential discriminationin explaining the wage gap. In this study, we de-cided to follow the latter approach.

The list of variables used in OLS analysis is givenin the Appendix (Table A1). These are fairly stand-ard and do not need any explanation. Two aspectsnot accounted for in the analysis are occupation andindustry. These had to be left out because of smallcell sizes. As a result of these omissions, this studyis unable to isolate the contribution of occupationalsegregation to the native-white wage differential toarrive at a more accurate measure of wage discrimi-nation. This could be a serious limitation, since someobservers have demonstrated that occupational seg-regation is a more important channel through whichpotential discrimination manifests itself than wagediscrimination per se, especially in the case of gen-der discrimination.11

The probit analysis leaves out the weeks of workvariables plus the full-time/part-time variable butincludes two new variables not considered in theOLS regressions. They are whether the individualreceived UI benefits, and whether or not the personwas a recipient of social assistance and other gov-ernment transfers. The probit samples for aborigi-nals consist of 6,092 observations for males and5,419 observations for females. The correspondingsamples for whites are 205,905 and 169,304 obser-vations for males and females respectively. Thesesamples were created by adding to the OLS sam-ples those who were self-employed and those withzero earnings. Thus, only students and those withmissing information are left out of the probitsamples.

Page 10: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

74 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

REGRESSION RESULTS

As mentioned before, regressions were run with andwithout adjustment for selectivity bias. We beginwith the regression results pertaining to men,unadjusted for selectivity bias (Table 3). The earn-ings profile with respect to age has the familiar in-verted U shape. Both aboriginals and whites maxi-mize their earnings when they are about 47 years ofage which, incidentally, is in accordance with theresults found by George and Kuhn. With regard to

education, compared with those with less than gradenine, those with higher levels tend to earn signifi-cantly more, with the premiums being highest forthose with BAs and MAs and PhDs. The only ex-ception is the group that has completed grades 9 to13, whose earnings are not significantly differentfrom those with less than grade nine education.

Native men with multiple origins earn signifi-cantly more than those with single origins. Full-timeemployees also earn significantly more than part-

TABLE 3Regression Results of Earnings of Whites and Natives, Full-time and Part-time (unadjusted for selectivity bias)

Males Females

Aboriginals Whites Aboriginals Whites

AGE 0.119(16.2)* 0.101(92.8)* 0.078(10.1)* 0.087(71.3)*AGE2 -0.001(13.9)* -0.001(81.3)* -0.001(8.48)* -0.001(62.5)*NFL 0.146((1.50) -0.087(7.09)* -0.077(0.78) -0.057(4.01)*PNB -0.040(0.39) -0.098(9.49)* -0.122(1.20) -0.107(9.05)*NSC -0.019(0.24) -0.137(13.9)* -0.183(2.22)* -0.165(14.7)*QUE -0.021(0.41) -0.063(8.54)* -0.051(1.02) -0.035(4.24)*MAN -0.162(3.72)* -0.127(13.2)* -0.173(3.79)* -0.127(12.1)*SAS -0.147(2.86)* -0.143(13.3)* -0.139(2.57)* -0.156(13.9)*ALB -0.052(1.34) -0.030(4.68)* -0.079(2.04)* -0.055(7.75)*BC 0.105(2.89)* 0.017(2.87)* -0.052(1.40) -0.015(2.17)*NCM -0.046(1.75) -0.061(15.7)* -0.111(4.29)* -0.132(30.0)*MUL 0.140(5.26)* 0.027(0.99)FTE 0.747(19.9)* 0.764(18.5)* 0.697(24.7)* 0.693(15.2)*GR9 0.076(1.70) 0.121(15.7)* 0.049(0.86) 0.170(16.4)*SEC 0.250(4.72)* 0.240(29.3)* 0.165(2.69)* 0.311(29.9)*ETC 0.201(3.18)* 0.268(27.7)* 0.196(2.28)* 0.367(25.5)*SNU 0.128(2.28)* 0.250(25.7)* 0.147(2.25)* 0.358(30.5)*HTC 0.171(3.20)* 0.297(33.8)* 0.276(3.95)* 0.342(27.1)*NUD 0.215(3.63)* 0.358(40.0)* 0.286(4.63)* 0.533(49.8)*SOM 0.208(3.60)* 0.333(37.7)* 0.350(5.56)* 0.548(49.6)*BA 0.406(5.96)* 0.519(58.9)* 0.588(8.32)* 0.779(69.5)*MAP 0.568(4.54)* 0.620(51.4)* 0.738(5.73)* 0.921(53.9)*

R2(ad) 0.604 0.582 0.589 0.562

Obs 5574 180349 5181 159278

Notes: Other control variables included in the regression are those pertaining to marital status, language proficiency,and weeks of work. These are not shown here for space considerations but are available from the author upon request.t-statistics are given within brackets. * indicates significance at the 0.05 level.

Page 11: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives75

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

time workers, and the earnings differential seemsto be roughly similar whether we are looking ataboriginals or whites. As far as natives are con-cerned, it makes no difference to their earningswhether they live in CMAs or elsewhere. However,in the case of whites, those who live outside theCMAs tend to earn significantly less than their ur-ban counterparts. At the provincial level, BritishColumbia residents earn more than those in Ontario,and this applies to both aboriginals and whites.However, in the case of the other provinces, thereare some differences between Aboriginal people andwhites. Natives living in Quebec and the Atlanticregion report no significant difference between theirearnings and those resident in Ontario, whereaswhites living in those provinces tend to earn rela-tively less. Similarly, while whites living in thePrairie provinces earn less than those in Ontario, inthe case of the natives, only those in Manitoba andSaskatchewan seem to have a relative earnings dis-advantage.

Turning next to the regression results for women,the age variables are significant and have the ex-pected signs. The earnings of aboriginals and whitesreach their peak at 45 and 46 years respectively. Theestimate for whites is the same as that found inGeorge and Kuhn, but in the case of aboriginals,their estimate is slightly lower: 41 years. The otherresults are broadly similar to those pertaining tomen. One important difference, however, is that,unlike aboriginal men, aboriginal women with mul-tiple origins do not have an earnings advantage rela-tive to those with single origins.

We are now in a position to decompose the earn-ings differential to estimate the degree of potentialdiscrimination against aboriginals. In the case ofmen, the earnings differential adjusted for weeksworked is 34 percent (Table 4, column 1), of whichabout three-fifths can be attributed to observableendowment differences. The remaining two-fifthsrepresents the unexplained residual, which we referto as potential discrimination. Of the endowments,the most important components are age and educa-

tion, both of which work in favour of the whites.They account for almost 70 percent of the total con-tribution of the identifiable endowments. Givingaboriginal men the same level of education as thewhites alone would reduce the wage differential byabout 13 percent. In the case of women, endowmentdifferences account for seven-tenths of the adjustedwage differential (column 2 of Table 4), which sug-gests that they face a lower degree of potential dis-crimination than men. However, the main messagecoming out of the table is that, whether we look ataboriginal men or women, the most important fac-tor explaining native-white wage differences are theendowment differences, with potential discrimina-tion playing a relatively minor role. As in the caseof men, the two most important endowments are ageand education. Raising the educational attainmentsof aboriginal women to the level of white womenwould wipe out the wage differential by about 16 to18 percent, depending on whether we use native orwhite weights.

The regression results corrected for selectivitybias are given in the Appendix (Table A2). The de-composition of the native-white earnings differen-tial reveals that the difference in wage offers isslightly higher than the observed wage differentialbecause of the difference in selectivity bias (Table 4,columns 3 and 4). For both sexes, observable en-dowment differences explain anywhere from aboutthree-fifths to three-quarters of the differential inwage offers adjusted for weeks worked, dependingon the weights used. This is consistent with the pre-vious finding that differences in endowments are themost important explanation for native-white earn-ings differentials. However, one important differencefrom the previous results is that the new evidence isnot strong enough to support the view that nativewomen face less potential discrimination than na-tive men.

The preceding analysis lumps together nativeswith single origins with those with multiple originsand makes no attempt to examine whether endow-ment differences between these two groups have any

Page 12: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

76 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

TABLE 4Decomposition of the White-Native Wage Differential, Full-time and Part-time

(1) (2) (3) (4)

Unadjusted for Selectivity Bias Adjusted for Selectivity Bias

Males Females Males Females

Observed Earnings Differential 0.5284 0.3829 0.5284 0.3829

Selectivity bias difference n.a. n.a. -0.0110 -0.0079

Offered wage differential n.a. n.a. 0.5394 0.3908

Adjusted Earnings Differential 0.3441 0.2265 0.3457 0.2488

(Using White Weights)

Contribution of Endowments

Age 0.0872 0.0882 0.1038 0.1028

Province 0.0046 0.0087 0.0076 0.0072

Non-CMA living 0.0098 0.0165 0.0096 0.0159

Marital status 0.0271 0.0041 0.0260 0.0025

Official language -0.0005 -0.0019 -0.0001 -0.0011

Full-time Employment 0.0229 -0.0035 0.0233 -0.0034

Education 0.0459 0.0409 0.0434 0.0349

Total endowments 0.1970 0.1530 0.2136 0.1588(57.25%) (67.55%) (61.79%) (63.83%)

Discrimination 0.1471 0.0735 0.1321 0.0900(42.75%) (32.45%) (38.21%) (36.17%)

(Using Native Weights)

Total endowments 0.1979 0.1465 0.2565 0.1630(62.25%) (69.37%) (75.82%) (74.46%)

Discrimination 0.1200 0.0647 0.0818 0.0559(37.75%) (30.63%) (24.18%) (25.54%)

Source: The estimates reported in columns (1) and (2) were derived from the coefficients given in Table 3, whereas theestimates given in columns (3) and (4) are based on the coefficients presented in Table A2 in the Appendix. A positiveestimate means an earnings advantage in favour of whites whereas a negative estimate signifies an advantage in favourof the natives. By “adjusted earnings differential” is meant the earnings differential adjusted for weeks worked. n.a.means not applicable.

Page 13: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives77

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

effect on the native-white wage differential. To theextent that such differences are significant, theiromission would affect the magnitude of the esti-mated degree of potential discrimination. To explorethis issue further, we ran separate regressions fornatives with single and multiple origins. Since theinverse Mill’s ratio is again insignificant, only theresults uncorrected for selectivity bias are reportedin Table 5. Note that in the case of those with mul-tiple origins, there is no estimate for allophones.This is because the data on that variable in the Pub-lic Use Sample Tape are so few that they have beensuppressed for reasons of confidentiality. We first

consider the results pertaining to males. The evi-dence shows a few similarities between those withsingle and multiple origins. In terms of the timetaken to maximize their earnings, both groups comeout even: 44.5 years for those with multiple origins,compared with 44 years for those with single ori-gins. Those who work full-time tend to earn morethan part-time workers, and this is true for bothgroups. The province of residence does not mattervery much to how much a person earns, whether heor she is of single or multiple origins. In the case ofthose with single origins, only those in Manitobaand Saskatchewan earn less than those in Ontario,

TABLE 5Regression Results: Aboriginals, Full-time and Part-time, Single and Multiple Origins (unadjusted for selectivity bias)

Single Origins Multiple Origins

Males Females Males Females

AGE 0.107(8.69)* 0.049(3.63)* 0.126(13.6)* 0.098(10.4)*AGE2 -0.001(7.55)* -0.000(2.45)* -0.001(11.7)* -0.001(9.23)*NFL 0.029(1.10) 0.051(0.30) -0.018(0.14) -0.158(1.30)PNB -0.224(1.04) -0.308(1.27) 0.019(0.17) -0.084(0.77)NSC 0.051(0.30) -0.071(0.39) -0.047(0.55) -0.216(2.39)*QUE 0.008(0.09) 0.103(1.05) -0.040(0.67) -0.144(2.47)*MAN -0.274(3.71)* -0.168(2.16)* -0.058(1.07) -0.178(3.06)*SAS -0.189(2.34)* -0.217(2.52)* -0.114(1.63) -0.061(0.83)ALB -0.031(0.42) -0.125(1.68)* -0.089(1.93) -0.055(1.22)BC 0.059(0.86) -0.165(2.29)* 0.130(3.07)* 0.010(0.23)NCM -0.116(2.41)* -0.209(4.35)* -0.011(0.35) -0.058(1.98)*FTE 0.582(9.34)* 0.731(14.2)* 0.837(17.8)* 0.672(20.1)*GR9 0.066((1.03) 0.138(1.77) 0.102(1.55) -0.004(0.04)SEC 0.197(2.26)* 0.174(1.99)* 0.311(4.34)* 0.149(1.58)ETC 0.173(1.66) 0.162(1.99)* 0.233(2.80)* 0.213(1.67)SNU -0.070(0.83) 0.042(1.43) 0.303(3.88)* 0.202(2.05)*HTC 0.056(0.66) 0.368(3.53)* 0.258(3.50)* 0.221(2.13)*NUD 0.066(0.61) 0.237(2.46)* 0.305(3.96)* 0.293(3.12)*SOM 0.285(2.89)* 0.396(4.17)* 0.209(2.74)* 0.319(3.34)*BA 0.410(2.59)* 0.717(5.26)* 0.458(5.55)* 0.548(5.50)*MAP 0.618(1.42) 0.804(2.84)* 0.625(4.78)* 0.701(4.58)*

R2(ad) 0.538 0.558 0.618 0.607

Obs. 2176 1804 3398 3377

Notes: Please see notes at the bottom of Table 3.

Page 14: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

78 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

whereas in the case of those resident elsewhere, theredoes not seem to be an earnings disparity betweenthemselves and those living in Ontario. Likewise,in the case of those with multiple origins, exceptfor individuals living in British Columbia, none ofthe provincial variables is significant.

Apart from these similarities, there are also anumber of differences between aboriginal men withsingle and multiple origins. While in the case ofthose with multiple origins, higher levels of educa-tion are associated with higher earnings, except inthe case of those who had completed only grades 9to 13, the picture is not so clear in the case of thosereporting single origins. For the latter group, exceptin the case of those who had either completed theirsecondary education, had some university education,or had completed a BA, higher education does notseem to produce an increase in earnings over andabove those who have had less than grade nineeducation.

In the case of those with single origins, there is asignificant earnings differential between those liv-ing in CMAs and those outside, but this does notapply to those with multiple origins.

A comparison of females with single and multi-ple origins reveals a number of similarities and dif-ferences. For example, full-time employment sta-tus and education are positively associated withearnings, while non-CMA living is negatively re-lated to earnings, and these relationships hold re-gardless of whether the person’s ethnic origin is sin-gle or multiple. Age is also a significant factor inboth regressions, but the time taken for earnings topeak differs for the two groups. While those withmultiple origins take only 42 years to maximize theirearnings, those with single origins take 55 years todo so. In addition to this, there are a few other mi-nor differences with respect to the province of resi-dence. In the case of those with single origins, thoseresident in British Columbia, Manitoba, Saskatch-ewan, and Alberta have an earnings disadvantagerelative to those in Ontario, whereas those resident

in other provinces find no earnings disparity betweenthemselves and those in Ontario. On the other hand,in the case of those with multiple origins, only thoseresiding in Quebec, Manitoba, and Nova Scotia earnless than those in Ontario, whereas residents in theother provinces report earnings similar to those ofOntario residents.

The results of the decomposition of the native-white earnings differential given in Table 6 arebroadly similar to the evidence presented so far thatendowment differences play a dominant role in theexplanation for native-white wage differentials.Except in one instance, endowment differences ac-count for anywhere from 55 to 82 percent of thenative-white wage differential. Furthermore, theevidence reveals that natives with single origins arenot always subject to a greater degree of potentialdiscrimination than those with multiple origins.While it is true that men with multiple origins faceless potential discrimination than men with singleorigins, a similar pattern is not observable in thecase of women. The latter finding with respect tonative women is not consistent with a priori expec-tations but we really have no explanation to offer asto why this is so.

Although the preceding analysis has failed todetect evidence of significant potential discrimina-tion against natives, we cannot rule out the possi-bility that some employers may discriminate againstnative workers by channelling them into part-time,part-year employment. This problem would not ariseif we leave out the part-time, part-year employeesand focus solely on those employed full-time andfull year. To conserve space, only the results of thedecomposition of the native-white wage differen-tial pertaining to full-time, full-year workers arediscussed here. These are based on regressionsunadjusted for selectivity bias, since the selectivitybias indicator was consistently insignificant. Asshown in Table 7, endowment differences explainover one-half to four-fifths of the native-white earn-ings differential. By implication, then, the portionof the wage differential attributable to potential

Page 15: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives79

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

discrimination ranges from about 20 percent to alittle over 40 percent, depending on the gender andweights used. Thus, the results of the decomposi-tion exercise pertaining to full-time, full-year em-ployees provide further evidence to the effect thatthe key to explaining native-white wage differen-

tials lies in endowment differences, with potentialdiscrimination playing only a minor role. Inciden-tally, the findings reported here are also broadly con-sistent with those presented by George and Kuhn, whofound that, in the case of full-time, full-year employ-ees, potential discrimination was relatively minor.

TABLE 6Decomposition of the Native-White Earnings Differential, Full-time and Part-time, Single and Multiple Origins

(1) (2) (3) (4)

Single Origins Multiple Origins

Males Females Males Females

(Using White Weights)

Wage Differential 0.8783 0.5332 0.4270 0.2522

Adjusted Wage

Differential 0.5574 0.2814 0.3298 0.1484

Endowments

Age 0.1249 0.0566 0.1649 0.0587

Province 0.0141 0.0193 -0.0002 0.0030

Non-CMAs 0.0187 0.0340 0.0042 0.0070

Marital Status 0.0368 0.0049 0.0210 0.0038

Language 0.0018 -0.0003 -0.0015 -0.0026

Full-time Empl 0.0260 -0.0222 0.0214 0.0069

Education 0.0853 0.0920 0.0217 0.0118

Total Endowments 0.3076 0.1843 0.2315 0.0886(55.18%) (65.49%) (70.19%) (59.70%)

Discrimination 0.2498 0.0971 0.0983 0.0598(44.82%) (34.51%) (29.81%) (40.30%)

(Using Native Weights)

Total Endowments 0.3328 0.2292 0.2436 0.0648(65.50%) (81.77%) (77.24%) (48.54%)

Discrimination 0.1753 0.0511 0.0718 0.0687(34.50%) (18.23%) (22.76%) (51.46%)

Note: These estimates are based on Table 5. For other explanations, please see the notes at the bottom of Table 4.

Page 16: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

80 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

CONCLUSION

In an earlier study, George and Kuhn (1994) exam-ined native-white wage differentials mainly for full-time, full-year paid employees living off reserves,relying on the 1986 census data. They found thatendowment differences account for a fairly largeportion: about 50 percent in the case of those re-porting to be of exclusively aboriginal origin. Thepresent paper is an attempt to extend and update theiranalysis by looking at a larger sample that includes,in addition to those working full-time and full year,those employed either part-time or for part of theyear, and relying on the 1991 census data. However,this study is unable to make a distinction betweennatives living on and off reserves because of the

nature of the data. Like George and Kuhn, the sam-ple is restricted to those living outside the Yukonand the Northwest Territories, in the 15 to 64 agebracket, working in paid employment, and report-ing positive earnings. Those living in institutionsand as temporary residents in the country, and stu-dents are left out of the analysis.

The main findings of the study can be stated asfollows. First, when all employees (part-time, partyear, full-time, full year) are considered, observableendowment differences account for anywhere fromabout three-fifths to seven-tenths of the native-whitewage differential, depending on the gender andweights used. Thus, even after the inclusion of thoseworking either part-time or for part of the year in

TABLE 7Decomposition of the Earnings Differential, Full-Time and Full-Year Paid Employees

(1) (2) (3) (4)

Males Females Males Females

(Using White Weights) (Using Native Weights)

Wage Differential 0.1601 0.1472 0.1601 0.1472

Endowments

Age 0.0265 0.0543 0.0226 0.0599

Province 0.0035 0.0084 0.0062 0.0050

Non-CMAs 0.0103 0.0162 0.0036 0.0142

Marital Status 0.0174 0.0005 0.0158 0.0052

Language -0.0032 -0.0030 0.0013 0.0040

Education 0.0369 0.0353 0.0375 0.0295

Total Endowments 0.0914 0.1117 0.0870 0.1178(57.09%) (75.88%) (54.34%) (80.03%)

Discrimination 0.0687 0.0355 0.0731 0.0294(42.91%) (24.12%) (45.66%) (19.97%)

Notes: The regression coefficients used to generate the above estimates are not given in this paper, but will be madeavailable upon request. For the meaning of the signs of the estimates, see the notes at the bottom of Table 4.

Page 17: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives81

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

the sample, potential discrimination turns out to beonly a relatively minor factor in the explanation ofthe native-white wage differential, thereby reinforc-ing the previous findings reported by George andKuhn. This, however, does not deny that potentialdiscrimination may operate through time spentworking in the labour market.

Second, of the endowments, age and educationare the most important components, accounting for68 to 84 percent of the overall contribution of en-dowments. Raising the level of education of nativesto be on a par with the whites would wipe out thenative-white wage differential by 13 percent in thecase of men, and by 16 to 18 percent for women.

Third, the evidence indicates that while men withmultiple origins face less potential discriminationthan men with single origins, which is consistentwith a priori expectations, the picture is somewhatmixed in the case of native women.

The above findings should be treated as tentativebecause the analysis is subject to a number of prob-lems. First, several factors such as occupation, in-dustry, turnover, absenteeism, unionization, familybackground, motivation, and ability had to be leftout due to data problems. Second, in the case ofsome variables such as work experience as well asin the selectivity bias adjustment, a number of prox-ies had to be used, which in turn could have resultedin measurement error. Third, although we havetreated endowments as pre-determined, it is conceiv-able that there may be instances such as in educa-tion and training, where discrimination could bepartly responsible for some of the endowment dif-ferences that exist between natives and whites. Manyof these issues cannot be addressed at this time dueto the unavailability of data. To facilitate such re-search, Statistics Canada should consider includingthe relevant data available in the Public Use Sam-ple Tape.

The main policy implication of the analysis is thatprobably the most fruitful approach to raising the

earnings of natives to be on a par with the whites isthrough skill development. It was shown earlier thatnatives generally have significantly lower levels ofeducation and training than whites. This is particu-larly true when we compare the percentage of uni-versity degree holders between natives and whites.The discrepancy becomes even more glaring whenwe shift our attention to natives with single originsand those living on reserves. Helping to improve andfurther their education would not only enhance theiremployment and income prospects, it may also helpto alleviate some of the other social evils usuallyassociated with poverty.

NOTES

The research reported in this study was undertaken whenthe author was associated with the Applied ResearchBranch, Human Resources Development Canada, Ottawa.I am indebted to the editor and four anonymous refereesfor their helpful comments and suggestions. The compe-tent research assistance provided by Steve Toporek andEric Khaiat is also gratefully acknowledged. I am, ofcourse, solely responsible for the views and opinions ex-pressed in the paper.

1These are based on employers’ reports submitted eachyear to the government as required under the Employ-ment Equity Act. For example, see Canada (1997).

2Other references include Alam and De Civita (1990);McDonald (1991); and Patrinos and Sekellariou (1992).

3For those with any aboriginal origins, the contribu-tion of endowment differences to the native-white wagedifferential is somewhat lower — ranging from 22 per-cent to 39 percent in the case of men, and from 26 per-cent to 37 percent in the case of women.

4The only solution would be to rely on special tabula-tions of census data. Unfortunately, resource constraintsprevented us from using this option.

5The sample used in the present study also suffers fromthe problem of underreporting due to a number of rea-sons, including the non-participation of some natives liv-ing on reserves in the census and the interruption of theenumeration before it was completed. According to somecrude estimates made by Statistics Canada, based on

Page 18: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

82 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

unpublished material, roughly 38,000 people resident onreserves fall into this category. They account for about20 percent of the on-reserve native population (based onthe 20-percent sample). I am indebted to Heather Tate,Housing, Family, and Social Statistics, Statistics Canada,for providing this information.

6This excludes those who did not take part in the cen-sus. One other point to note is that we cannot say withcertainty whether there has been a change in the aborigi-nal population between 1986 and 1991, because of achange in the definition of an aboriginal. In 1991, thecensus questionnaire asked “To which ethnic or culturalgroup(s) did your ancestors belong?” The 1986 censusquestion was slightly different. It asked “To which eth-nic or cultural group(s) did you or your ancestors belong?”For more details, see Statistics Canada (1995, pp. 4-6).

7This is the standard approach to discrimination. See,for example, the surveys by Cain (1986); and Gunderson(1989).

8In equation (1), the endowment differences areweighted by the ß coefficients from the regression forwhites. This assumes that, in the absence of potential dis-crimination, the aboriginal wage structure would resem-ble the wage structure for whites. However, not every-body agrees that this is so. The main contributors to thisdiscussion are mentioned in Gunderson (1989). One so-lution proposed by Oaxaca (1973) is to use both aborigi-nal and white weights and provide a range for the dis-crimination coefficient. Another suggestion made by Cot-ton (1988) and Reimers (1983) is to use a weighted aver-age of the wage structures of the two groups. Yet anothersuggestion made by Neumark (1988) and Langford (1995)is to use the coefficients of a pooled regression. Since weare not quite sure as to which method is conceptually su-perior, we decided to use Oaxaca’s method, and reportestimates using both sets of weights.

9This, indeed, is the conclusion of a recent study byNeal and Johnson (1996) which tried to correct forendogeneity and imprecise measurement of skills.

10I am indebted to a reviewer for pointing this out tome.

11See, for example, Dolton and Kidd (1990); andGunderson and Riddell (1988, p. 455).

REFERENCES

Alam, J. and P. De Civita (1990), “Wage Gap BetweenNatives and Non-Natives: An Empirical Analysis,”Ottawa: Indian and Northern Affairs Canada, unpub-lished paper.

Blinder, A.S. (1973), “Wage Discrimination: ReducedForm and Structural Estimates,” Journal of HumanResources 8(4):436-55.

Cain, G.G. (1986), “The Economic Analysis of LabourMarket Discrimination: A Survey,” in Handbook ofLabour Economics, vol.1, ed. O. Ashenfelter and R.Layard (Amsterdam: North Holland), ch. 13.

Canada (1997), Annual Report, Employment Equity Act(Ottawa: Supply and Services Canada).

Christofides, L.N. and R. Swidinsky (1994),“Wage De-termination by Gender and Visible Minority Status:Evidence from the 1989 LMAS,” Canadian PublicPolicy/Analyse de Politiques 20(1):34-51.

Cotton, J. (1988), “On the Decomposition of Wage Dif-ferentials,” Review of Economics and Statistics70(2):236-43.

Daymont, T.N. and P.J. Andrisani (1984), “Job Prefer-ences, College Major and the Gender Gap in Earnings,”Journal of Human Resources 19:408-28.

De Silva, A. (1992), Earnings of Immigrants A Compara-tive Analysis, a study prepared for the Economic Coun-cil of Canada (Ottawa: Supply and Services Canada).

Dolton, P.J. and M. P. Kidd (1990), “Occupational Ac-cess and Wage Discrimination,” Discussion Paper 803(Kingston: Queen’s University).

Follett, R.S., M.P. Ward and F. Welch (1993), “Problemsin Assessing Employment Discrimination,” AmericanEconomic Review Papers and Proceedings 83(2):73-78.

George, P. and P. Kuhn (1994), “The Size and Structureof Native-White Wage Differentials in Canada,” Ca-nadian Journal of Economics 27(1):20-42.

Goldberger, A.S. (1984), “Reverse Regression and Sal-ary Discrimination,” Journal of Human Resources19:293-318.

Gronau, R. (1974), “Wage Comparisons — SelectivityBias,” Journal of Political Economy, 82:1119-43.

Gunderson, M. (1989), “Male-Female Wage Differentialsand Policy Responses,” Journal of Economic Litera-ture 27(1):46-117.

Gunderson, M. and W.C. Riddell (1988), Labour MarketEconomics Theory, Evidence and Policy in Canada(Toronto: McGraw-Hill Ryerson Limited).

Page 19: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives83

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

Harkness, J. (1993), “Labour Force Participation by Disa-bled Males in Canada,” Canadian Journal of Econom-ics 26:878-89.

Heckman, J. (1979), “Sample Selection Bias as a Speci-fication Error,” Econometrica 47:153-61.

Hum, D. and W. Simpson (1996), “Canadians with Dis-abilities and the Labour Market,”Canadian PublicPolicy/Analyse de Politiques 22(3):285-99.

Kamalich, R.F. and S.W. Polachek (1982), “Discrimina-tion: Fact or Fiction? An Examination Using an Al-ternative Approach,” Southern Economic Journal49:450-61.

Killingsworth, M.R. (1993), “Analyzing EmploymentDiscrimination: From the Seminar Room to the Court-room,” American Economic Review. Papers and Pro-ceedings 83(2):67-72.

Langford, M.S. (1995), “The Gender Gap in the 1990s,”Australian Economic Papers 34(64): 62-85.

Leck, J.D., S. St. Onge and I. Lalancette (1995), “WageGap Changes Among Organizations Subject to theEmployment Equity Act,” Canadian Public Policy/Analyse de Politiques 21(4):387-400.

McDonald, R.J. (1991), “Canada’s Off-Reserve Aborigi-nal Population,” Canadian Social Trends 23:2-7.

Miller, P.W. (1987), “Gender Differences in Observed andOffered Wages in Canada, 1980,” Canadian Journalof Economics 20(2):225-44.

Neal, D.A. and W.R. Johnson (1996), “The Role ofPremarket Factors in Black-White Wage Differences,”Journal of Political Economy 5(104):869-95.

Neumark, D. (1988), “Employers’ Discriminatory Behav-iour and the Estimation of Wage Discrimination,” Jour-nal of Human Resources 23(3):279-95.

Oaxaca, R. (1973), “Male-Female Wage Differentials inUrban Labour Markets,” International Economic Re-view 14:693-709.

Patrinos, H.A. and C.N. Sekellariou (1992), “NorthAmerican Indians in the Canadian Labour Market: ADecomposition of Wage Differentials,” Economics ofEducation Review, 11:257-66.

Pendakur, K. and R. Pendakur (1995), “The Colour ofMoney: Earnings Differentials Among Ethnic Groupsin Canada,” Strategic Research and Analysis, Cana-dian Heritage, Hull, Quebec.

Reimers, C.W. (1983), “Labour Market DiscriminationAgainst Hispanic and Black Men,” Review of Econom-ics and Statistics 65:570-79.

Shapiro, D. and M. Stelcner (1997), “Language and Earn-ings in Quebec,” Canadian Public Policy/Analyse dePolitiques 13(2):115- 40.

Sloane, P. J. (1985), “Discrimination in the Labour Mar-ket,” in Surveys in Economics: Labour Economics, ed.D. Carline, C.A. Pissarides et al. (London: Longman).

Statistics Canada (1995), Profile of Canada’s AboriginalPopulation, Catalogue No. 94-325 (Ottawa: StatisticsCanada).

Welch, F. (1975), “Human Capital Theory: Education,Discrimination and Life Cycles,” American EconomicReview 65:63-73.

Page 20: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

84 Arnold De Silva

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

APPENDIX

TABLE A1Variables Used in OLS Regressions

AGE a proxy for experience;AGE2 a quadratic term to account for diminishing returns to experience;λ Inverse Mill’s Ratio

The following are 0-1 dummy variables:Province of ResidenceNFL Newfoundland (reference group: Ontario);PNB Prince Edward Island plus New Brunswick;NSC Nova Scotia;QUE Quebec;MAN Manitoba;SAS Saskatchewan;ALB Alberta;BC British Columbia.Non-Urban LivingNCM Residence in a non-Census Metropolitan Area.Marital StatusMAR Married (reference group: single or never married);DIV Divorced;SEP Separated;WID Widowed.Official LanguageBIL Proficiency in English and French (reference group: English);ALO Neither English nor French (allophones);FRE French.EthnicityMUL a person who has multiple aboriginal origins (as distinct from single or exclusively aboriginal origins).Full -Time/Part -Time EmploymentFTE a person with full- time employment;EducationGR9 Completion of grades 9 to 13 (reference group: those with less than grade 9);SEC Secondary School Certificate;ETC Elementary Trade Certificate;SNU Some Non-university education;HTC Higher Trade Certificate;NUD Non-University Diploma;SOM Some University Education;BA Bachelor’s Degree;MAP Master’s and Doctoral Degrees.Weeks of WorkTen weeks of work variables starting with 6WK (worked from 6 to 10 weeks) and ending with 51WK( worked for 51 and52 weeks). The reference group includes those who worked for less than six weeks.

Page 21: age Discrimination W Against Natives - Queen's Universityqed.econ.queensu.ca/pub/cpp/mar1999/DESILVAchap.pdf · 2001-05-15 · Wage Discrimination Against Natives 65 CANADIAN PUBLIC

Wage Discrimination Against Natives85

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES, VOL. XXV , NO. 1 1999

TABLE A2Regression Results of Earnings of Whites and Natives, Full-time and Part-time (adjusted for selectivity bias)

Males Females

Aboriginals Whites Aboriginals Whites

AGE 0.119(13.3)* 0.102(57.9)* 0.076(8.74)* 0.084(67.5)*AGE2 -0.001(9.13)* -0.001(61.2)* -0.001(5.66)* -0.001(62.5)*NFL 0.047(1.79) -0.074(5.87)* -0.065(1.33) -0.044(3.45)*PNB -0.038(1.64) -0.090(5.94)* -0.107(1.58) -0.099(4.97)*NSC -0.017(1.02) -0.114(6.96)* -0.184(1.97)* -0.107(5.08)*QUE -0.021(0.76) -0.047(5.73)* -0.055(1.23) -0.025(2.85)*MAN -0.174(5.71)* -0.126(8.51)* -0.157(2.00)* -0.106(4.67)*SAS -0.175(5.81)* -0.141(8.13)* -0.108((2.21)* -0.123(5.74)*ALB -0.024(0.65) -0.007(1.95) -0.055(1.78) -0.034(6.73)*BC 0.095(2.35)* 0.017(1.97)* -0.014(1.00) -0.006(0.87)NCM -0.045(1.95) -0.059(12.6)* -0.106(3.96)* -0.127(24.8)*MUL 0.130(5.88)* 0.015(1.06)FTE 0.744(14.8)* 0.776(16.9)* 0.689(23.0)* 0.684(12.6)*GR9 0.075(1.06) 0.121(5.74)* 0.044(0.81) 0.088(12.0)*SEC 0.236(2.31)* 0.224(6.54)* 0.126(2.00)* 0.174(19.1)*ETC 0.217(1.98)* 0.245(6.93)* 0.146(2.12)* 0.215(23.1)*SNU 0.148(2.14)* 0.258(7.95)* 0.178(3.18)* 0.305(34.7)*HTC 0.153(2.00)* 0.294(7.96)* 0.254(3.26)* 0.324(35.7)*NUD 0.206(2.07)* 0.349(8.34)* 0.280(4.18)* 0.485(43.8)*SOM 0.208(4.57)* 0.354(8.55)* 0.306(4.78)* 0.504(48.9)*BA 0.424(6.95)* 0.518(13.9)* 0.593(5.41)* 0.687(65.7)*MAP 0.557(7.86)* 0.614(23.8)* 0.716(7.38)* 0.797(77.0)*λ -0.041(1.34) -0.122(1.68) -0.026(1.08) 0.086(1.69)

R2(adj) 0.623 0.616 0.621 0.578

Obs 5574 180349 5181 159278

Notes: Coefficient standard errors and hence the t-ratios (shown in parentheses) are based upon a whiteheteroscedasticity-consistent covariance matrix. For additional notes, see the bottom of Table 3.

The regression results corrected for selectivity bias are presented in Table A2. The inverse Mill’s ratio isconsistently insignificant, thus indicating that there is no evidence of selectivity bias. Note, however, thatthis result may be partly due to the poor proxies used to correct for selectivity bias. George and Kuhn alsomade the same comment after getting results similar to those reported here. One other point to note is that acomparison of Tables 3 and A2 reveals a remarkable degree of similarity in the regression coefficientsbetween the results uncorrected for selectivity bias and the present version. The probit results are not dis-cussed here, but are available from the author upon request.