UBC Faculty Research and Publications

Measuring gender when you don’t have a gender measure: constructing a gender index using survey data Smith, Peter M; Koehoorn, Mieke May 28, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12939_2016_Article_370.pdf [ 634.73kB ]
JSON: 52383-1.0308591.json
JSON-LD: 52383-1.0308591-ld.json
RDF/XML (Pretty): 52383-1.0308591-rdf.xml
RDF/JSON: 52383-1.0308591-rdf.json
Turtle: 52383-1.0308591-turtle.txt
N-Triples: 52383-1.0308591-rdf-ntriples.txt
Original Record: 52383-1.0308591-source.json
Full Text

Full Text

RESEARCH Open AccessMeasuring gender when you don’t have agender measure: constructing a genderindex using survey dataPeter M. Smith1,2,3* and Mieke Koehoorn1,4*AbstractBackground: Disentangling the impacts of sex and gender in understanding male and female differences isincreasingly recognised as an important aspect for advancing research and addressing knowledge gaps in thefield of work-health. However, achieving this goal in secondary data analyses where direct measures of genderhave not been collected is challenging. This study outlines the development of a gender index, focused ongender roles and institutionalised gender, using secondary survey data from the Canadian Labour Force survey.Using this index we then examined the distribution of gender index scores among men and women, andchanges in gender roles among male and female labour force participants between 1997 and 2014.Methods: We created our Labour Force Gender Index (LFGI) using information in four areas: responsibility forcaring for children; occupation segregation; hours of work; and level of education. LFGI scores ranged from 0to 10, with higher scores indicating more feminine gender roles. We examined correlations between each componentin our measure and our total LFGI score. Using multivariable linear regression we examined change in LFGI score formale and female labour force participants between 1997 and 2014.Results: Although women had higher LFGI scores, indicating greater feminine gender roles, men and women wererepresented across the range of LFGI scores in both 1997 and 2014. Correlations indicated no redundancy betweenmeasures used to calculate LFGI scores. Between 1997 and 2014 LFGI scores increased marginally for men anddecreased marginally for women. However, LFGI scores among women were still more than 1.5 points higher onaverage than for men in 2014.Conclusions: We have described and applied a method to create a measure of gender roles using survey data,where no direct measure of gender (masculinity/femininity) was available. This measure showed good variationamong both men and women, and was responsive to change over time. The article concludes by outlining anapproach to use this measure to examine the relative contribution of gender and sex on differences in healthstatus (or other outcomes) between men and women.Keywords: Gender, Sex, Labour force, Gender roles, Measurement, Survey data* Correspondence: psmith@iwh.on.ca; mieke.koehoorn@ubc.ca1Institute for Work & Health, 481 University Avenue, Suite 800, Toronto, ONM5G 2E9, CanadaFull list of author information is available at the end of the article© 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Smith and Koehoorn International Journal for Equity in Health  (2016) 15:82 DOI 10.1186/s12939-016-0370-4BackgroundBetter understanding and accounting for male and femaledifferences has been gaining attention in many health-related research areas [1–3]. In the area of work andhealth, men and women differ in their work exposuresand work-related health conditions. In addition, the rela-tionships between work exposures and health outcomesmay also differ for men and women. These male/femaledifferences can be due to sex – referring to biological dif-ferences between men and women – or gender – refer-ring to social differences between men and women [4].It is increasingly recognised that both sex and gendermatter in understanding the relationships between work-ing conditions and health outcomes, and that researchthat fails to take sex and gender into account is limitedin both quality and applicability [4, 5]. Stratifying ana-lyses to examine the relationships between work andhealth separately for men and women has been proposedas one approach to better account for sex and gender[6]. However, it is recognised that this approach does abetter job of understanding “sex” differences than itdoes in understanding “gender” differences [7, 8]. Fur-thermore, sex and gender often interact, suggesting thatdifferences between men and women might be due to acombination of both biological (sex) and social/cultural(gender) factors. To develop a better understanding ofthe relative contribution of each of these aspects re-quires measures of both sex and gender to be includedin analyses [5, 8].Measuring gender and sex can take different formsdepending on the way data are being collected. Whenconducting primary data collection for quantitativestudies researchers have the options to include mea-sures of gender, such as the Bem-Sex-Role-Inventory(BSRI) [9] which asks participants to self-identify withpersonal traits, or the Masculine Gender Role Stressscale [10]. However, are there options for researchersto measure gender if such scales are not present inexisting data?The concept of gender diagnosticity was first introducedby Lippa and Connelly [11] to estimate the probability ofbeing male or female, based on some gender-relateddiagnostic indicator. In their original study, Lippa andConnelly used occupational preference ratings as a meas-ure of gender-diagnosticity, finding that these preferenceswere distinct from responses to the Personal AttributesQuestionnaire (PAQ) [12] and the BSRI [9]. They alsoreported that occupational preference was more predictiveof being a man or a woman than either the BRSI or PAQindices [11]. This suggests that a gender diagnostic ap-proach offers an alternative method to measure socialdifferences between men and women based on their rolesand preferences, compared to indices such as the PAQand BRSI that are based on gender stereotypes [11].A decade later Lippa and colleagues used this sameapproach (occupational preferences) to examine the roleof sex and gender on mortality [13]. In this study theyfound masculinity (as measured by occupational prefer-ences) was predictive of mortality among both men andwomen, resulting in the highest mortality rate beingobserved among the most masculine men, and the lowestmortality rate observed among the most feminine women[13]. The objective in a gender-diagnostic approach is toidentify indicators that best differentiate between differentgender-based groups. As Lippa and Connelly noted intheir original paper, multiple indicators would ideally beused to form a gender index, resulting in a more reliablescale [11].Most recently a variation on this approach has beenused to examine the impact of sex and gender on cardio-vascular risk factors among individuals with prematureacute coronary syndrome [14]. In this study the genderindex was comprised of information on whether the re-spondent was the primary earner in their household;their personal income; the number of hours and respon-sibility for housework; and level of stress at home –along with measures of masculinity and femininity fromthe BSRI [14]. Similar to the previous gender diagnosticstudies, this paper found that both sex and gender wereimportant in predicting many cardiovascular risk factors,but that the gender score was generally more importantthan sex (male/female) in predicting risk in multivariablemodels [14].The preceding studies relied on primary data collec-tion where the concept of a gender index or gender-diagnostic approach was part of the study design. This isnot always feasible in population-health or health ser-vices research studies that rely on existing surveys andadministrative health records, despite the growing bodyof literature that indicated that gender differences matterto primary prevention and health care practices [15–17].In this paper the aim is to develop a gender-indexusing the Canadian Labour Force Survey (LFS). The LFSwas selected because it has a number of questions thatare commonly available in other data sources, and agender-index using this data may be readily applied (andmodified or expanded upon) to other secondary datasources. We use data from the 1997 and 2014 LabourForce Surveys to accomplish three objectives: to developa gender-index using existing population health surveydata; to examine the distribution of our gender indexacross males and females (i.e. to ensure that it measureda separate concept to sex); and to examine if there havebeen changes in gender roles (as measured by the index)among male and female labour market participants be-tween 1997 and 2014. We then discuss how this index(or a similarly constructed index) could be used in re-search that exploits secondary data to better understandSmith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 2 of 9the relative contribution of aspects of gender and sex inmale/female differences in health outcomes.MethodsData sourceThis paper uses secondary data from Statistics Canada’sLFS. The LFS is a monthly survey carried out by Statis-tics Canada with the objective of providing informationon trends in labour market participation and hours ofwork across major occupational and industrial sectors inCanada [18]. The LFS surveys approximately 56,000Canadian households per month. Households remain inthe sample for six consecutive months, with one sixth ofthe sample rotated out, and replaced by a new group ofhouseholds representing one sixth of the sample eachmonth. The target population for the LFS is the civilian,non-institutionalised population 15 years of age and overresiding in all of Canada’s provinces and territories. Per-sons living on Aboriginal reserves, full-time members ofthe Canadian Armed Forces, and the institutionalisedpopulation are excluded from coverage, as are householdsin extremely remote areas. Statistics Canada estimatesthese groups represent less than 2 % of the Canadianpopulation aged 15 and over, and that the LFS is represen-tative of its target population [18]. For the purpose of thisanalysis, the Public Use files from the 1997 and 2014Labour Force Surveys were used through StatisticsCanada’s Data Liberation Initiative [19]. The year 1997was chosen as the start point for the analysis, as thequestions asked in the LFS changed in this survey year.For each survey cycle, the analysis was restricted torespondents who were currently working for pay orprofit in the past month, excluding unpaid familyworkers, regardless of the number of hours worked.Labour Force Gender Index (LFGI)Gender is a multidimensional construct that includesfour dimensions: gender roles (behavioural norms ap-plied to men and women); gender identity (how anindividual sees themselves on the male/female con-tinuum); gender relationships (how individuals are treatedby others based on their ascribed gender); and institution-alized gender (how power and influence are distributeddifferently among men and women) [4]. The LFGI con-structed from the LFS focused primarily on the dimen-sions of gender roles and institutionalised gender amonglabour force participants. Given the data available, theLFGI was comprised of four main measures: responsibilityfor caring for children; occupational segregation; hours ofwork relative to partner/spouse; and education relative topartner/spouse. Differences in male and female participa-tion rates in education in Canada and other developedcountries have changed considerably since the early1970’s,with women outnumbering men in university andpost-secondary education completions [20, 21]. However,education was used in the construction of the LFGI as it isa measure of educational attainment relative to one’spartner/spouse, not a measure of absolute educationalattainment. Each measure in the index is described indetail below.Responsibility for caring for childrenIn each cycle of the LFS respondents are asked if theywere away from work (either completely or partially) inthe last week, and the reason for this absence, with oneoption being personal or family responsibilities. Re-spondents working less than 30 h per week are alsoasked the main reason they are not working morehours per week, with one option being caring for chil-dren and another being other personal or family re-sponsibilities. Using responses to these questions, thefollowing three category variable was created: 0 = no re-duction in labour market participation due to personalor family responsibilities; 1 = part or full week absencedue to personal or family responsibilities; 2 = workingpart-time due to personal or family responsibilities.Occupational segregationSelf-reported occupation is coded into 47 major groupsbased on the National Occupational Classification sys-tem [22]. For the LFGI responses to the 1997 LFS wereused to classify each of these 47 occupations into one offour groups: 0 = occupations where less than 26 % ofworkers were women; 1 = occupations where 26 to 50 %of workers were women; 2 = occupations where 51 to74 % of workers were women; and 3 = occupations where75 % or more of workers were women. Occupations withthe lowest participation of women are conceived as themost masculine occupations, while occupations with thehighest participation of women are conceived as the mostfeminine occupations.Hours of work relative to partner/spouseRespondents are asked the usual number of hours theyusually work each week. For respondents who are liv-ing with a spouse they are also asked the number ofhours their spouse usually works per week. Using boththese sources of hours worked each respondent wasgrouped into one of the following four categories: 0 =respondent working, but spouse not in the labourforce; 1 = respondent working more hours than theirspouse; 2 = respondent working the same number ofhours as their spouse; and 3 = respondent working lesshours than their spouse. If respondents did not have aspouse they were grouped with respondents workingmore hours than their spouse.Smith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 3 of 9Education level relative to partner/spouseRespondent’s and spouse’s highest level of education arereported in the following six categories: 0 to 8 years ofeducation; some secondary education; graduated fromhigh school; some post-secondary education; post-secondary certificate or diploma; and university degree.Using this information respondents were grouped intoone of the following three categories: 0 = respondentswith a higher level of education than their spouse; 1 = re-spondents with the same level of education as theirspouse; 2 = respondents with a lower level of educationthan their spouse. Similar to work hours, respondentswithout a spouse were grouped with respondents with ahigher level of education than their spouse.To create the LFGI the values for the above four mea-sures (caring for children, occupational segregation,hours or work and education level) were summed foreach respondent providing a score ranging from 0 to 10,with higher scores indicating more traditionally femininegender labour market roles of respondents and lowerscores indicating more traditionally masculine genderlabour market roles.AnalysisCorrelations between the four measures of the LFGIwere examined, and between each component and thefinal LFGI score. LFGI scores were then comparedfor men and women, and for the 1997 and 2014 LFS.Linear regression analyses then examined if the rela-tionship between sex (male versus female) and LFGIscores changed between 1997 and 2014, after adjust-ment for differences in age, province, and month ofsurvey participation between the 1997 and 2014 sur-veys. To examine if gender scores had changed formen and women between 1997 and 2014 a multi-plicative interaction term between sex (male/female)and survey year was included in the model. The re-gression analysis was based on a 10 % random sampleto avoid the possibility of a Type I error given thesize of the LFS samples. All analyses were weightedto account for the initial probability of selection foreach household, non-response and coverage errors, asspecified by Statistics Canada [18]. Analyses wereconducted using SAS Version 9.3 [23].ResultsTable 1 presents the distribution of each of the LFGImeasures for men and women in the 1997 and 2014Labour Force Surveys. Women were more likely to havetaken time off and be working part time due to house-hold responsibilities in both 1997 and 2014, and theywere also more likely to be working fewer hours thanTable 1 Distribution of gender index components for Canadian men and women in 1997 and 20141997 LFS (N = 696,350) 2014 LFS (N = 729,132)Men Women p-valuefor diffMen Women p-valuefor diffResponsibility for caring for childrenNo absence from work due to family or household responsibilities 98.8 % 91.1 % < 0.001 98.0 % 91.3 % < 0.001Part or full-week absence due to family or household responsibilities 1.0 % 2.7 % 1.7 % 4.5 %Works part-time due to family or household responsibilities 0.2 % 6.3 % 0.3 % 4.2 %Occupation (based on 1997 LFS only)Less than 26 % women 45.3 % 7.5 % < 0.001 46.0 % 7.7 % < 0.00126 to 50 % women 29.8 % 22.7 % 27.7 % 21.8 %51 % to 74 % women 22.0 % 43.4 % 22.4 % 45.5 %75 % women 2.9 % 26.5 % 3.9 % 25.0 %Hours of workRespondent works spouse does not 18.9 % 7.9 % < 0.001 15.1 % 8.6 % < 0.001Respondent works more than spouse/respondent does not have a spouse 62.6 % 41.8 % 63.4 % 46.3 %Respondent works same amount as spouse 13.5 % 16.2 % 15.0 % 15.5 %Respondent works less than spouse 5.0 % 34.2 % 6.6 % 28.6 %EducationRespondent higher level of education than spouse/respondent doesnot have a spouse20.2 % 18.9 % < 0.001 14.5 % 18.6 % < 0.001Respondent same education as spouse 62.2 % 63.3 % 68.1 % 69.0 %Respondent lower education than spouse 17.6 % 17.9 % 17.5 % 12.4 %Respondents to Statistics Canada’s Labour Force SurveySmith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 4 of 9their spouse in both time periods. As expected, giventhat occupation categories were based on 1997 labourmarket participation, we observed women were morelikely to be working in occupations with a greater pro-portion of women, and men in occupations with agreater proportion of men. Distribution across occupa-tional segregation groups for men and women onlychanged to a small extent between 1997 and 2014. Dif-ferences in education were also noted for men andwomen, although these were smaller in magnitude thanobserved for other measures. In 1997 an almost identicalproportion of men and women had lower levels of edu-cation than their spouse (conceptualised as being themost feminine category), but by 2014 men were morelikely than women to have lower education than theirpartner/spouse.Figures 1a and b present the distribution of LFGIscores for men and women in 1997 (Fig. 1a) and 2014(Fig. 1b). The distribution of LFGI scores was relativelysimilar for men and women in 1997 and 2014, withwomen scoring higher (more feminine) on the LFGIthan men. It is important to note, in each year malesand females were represented across the range of LFGIscores from 0 to 10, highlighting the distinction betweengender as measured by the LFGI and biological sex.Table 2 presents the polychoric correlations betweenthe LFGI and its four component measures. Correla-tions for respondents in 1997 are presented below the0%5%10%15%20%25%30%35%0 1 2 3 4 5 6 7 8 9 10Men WomenFeminity gender role index score0%5%10%15%20%25%30%35%0 1 2 3 4 5 6 7 8 9 10Men WomenFeminity gender role index scoreabFig. 1 a Distribution of gender index score (higher scores = greater feminine gender roles) for Canadian men and women. 1997 Labour ForceSurvey. b Distribution of gender index score (higher scores = greater feminine gender roles) for Canadian men and women. 2014 LabourForce SurveySmith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 5 of 9diagonal and correlations for respondents in 2014 arepresented above the diagonal. The relationship betweenthe LFGI and its component measures were similar atboth time points. The LFGI was most strongly corre-lated with occupation segregation and hours of work,and weakly correlated with education. Focusing on themeasures included in the LFGI the highest correlationwas observed between caring for children and hours ofwork in each survey year. Correlations indicated no re-dundancy between measures.Table 3 presents the results of the linear regressionmodel examining the interaction between sex and surveyyear on LFGI index scores after adjustment for age,province of residence and survey month. A statisticallysignificant interaction was observed between sex andsurvey year. Although women had higher gender scoresthan men and gender scores increased between 1997and 2014, this increase was not the same for men andwomen. To examine this interaction further, separatemodels were constructed for men and women. Thesemodels demonstrated that LFGI scores increased (indi-cating higher feminine gender roles) for men between1997 and 2014, but decreased for women during thesame time period (results not shown but available on re-quest). The adjusted mean scores for the gender indexfor men increased from 2.81 in 1997 to 3.01 in 2014. Forwomen the adjusted mean scores for the gender indexdecreased from 4.78 in 1997 to 4.64 in 2014.DiscussionDisentangling the impacts of sex and gender in under-standing male and female differences is increasingly recog-nised as an important aspect for advancing research andaddressing knowledge gaps in the field of work-health [5].However, achieving this goal in secondary data analyseswhere direct measures of gender, such as the BRSI orPAQ, have not been collected is challenging. The objectiveof this paper was to demonstrate how a gender index –based primarily on gender roles – could be developedusing routinely collected information from the CanadianLabour Force Survey. A second objective was to examinehow gender scores were distributed among men andwomen (i.e. sex) and if there had been changes in genderroles among working Canadian men and women over the17 year period between 1997 and 2014. Differences wereobserved between men and women in each component ofthe LFGI, with women generally having higher LFGIscores (indicating greater feminine gender roles) com-pared to men. While we found that women had higherLFGI scores in both 1997 and 2014 small increases inLFGI scores were observed for men between 1997 and2014, and small decreases in LFGI scores for women overthe same time period.These study results should be interpreted taking the fol-lowing strengths and limitations into account. Thehousehold-based sampling strategy employed by StatisticsCanada in conducting the LFS resulted in a truly represen-tative sample of the Canadian labour market, and findingscan be generalised to Canadian labour market participantsover the study time period. However, the large sample alsoincreases the possibility of a Type I error and the inferenceof a meaningful difference that has no practical or mean-ingful importance. This may be the case for the observeddifferences in the LFGI score over time and the interpret-ation that men are taking on greater feminine gender roleswhile women are taking on greater masculine genderroles. In each of these cases the differences over timeperiods were less than 0.5 on an index that ranges from 0to 10. To put this into context, if LFGI scores continue toincrease among men and decrease among women at thesame rate as observed over the 19-year study period (1997to 2014), it will take until 2097 for men and women tohave similar LFGI scores, indicating gender-equity inrelation to labour market roles.Three of the four measures that comprised the LFGIdistinguished between social and occupational roles ofmen and women in the expected direction. However, asimilar number of men and women had lower educationTable 2 Polychoric correlations between gender index and itscomponents1 2 3 4 51. Gender Index 1.00 0.64 0.79 0.75 0.352. Responsibility for caring for children 0.72 1.00 0.21 0.35 -0.053. Occupation(based on 1997 LFS only)0.79 0.27 1.00 0.20 -0.084. Hours of work 0.78 0.46 0.26 1.00 0.005. Education 0.43 0.03 -0.01 0.06 1.00Correlations below diagonal are for 1997 LFS. Correlations above diagonal arefor 2014 LFSTable 3 Adjusted ordinary least squared (OLS) estimates for sex,survey year and their interaction in gender index scoreEst se p-valueSexMale refFemale 1.57 0.01 < 0.001Survey Year2014 ref1997 -0.21 0.01 < 0.001InteractionSurvey year/sex multiplicative interaction term 0.37 0.02 < 0.001Respondents to the 1997 and 2014 LFS (N = 142,558; 10 % random sample)Est OLS regression estimate, se standard errorEstimates additionally adjusted for age, age2, province/territory of residenceand survey monthSmith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 6 of 9than their partner/spouse in 1997, while men were morelikely than women to have lower education than theirpartner/spouse in 2014. To some extent this result re-flects the changing nature of characteristics previouslythought of as masculine or feminine [24]. For example,the BRSI has “ambitious” and “analytical” as masculinetraits [9] while the PAQ has “likes math and science”and “intellectual” as masculine traits [12]. Given changesin educational participation between men and women,along with the information presented in this paper, fu-ture work that creates indexes/measures that reflectgender roles and institutionalised gender may choose toexclude education (as both an absolute measure and inrelation to the respondent’s partner/spouse) as a compo-nent of such measures.Finally, the construction of the LFGI represents thesum of scores for the relative components. While thisapproach has the advantage of simplicity, making theapproach easy to replicate, it does make assumptionsabout the relative contribution of each of the compo-nents of the index in relation to overall labour marketgender roles, which may not be valid. Alternative ap-proaches to constructing the index (e.g. factor analysesor cluster analyses) may be warranted, and researchersshould weigh the advantages and disadvantages toeach analytic approach if they choose to replicate thework in this paper.How could the LFGI be used to better understandprocesses that create male/female differences inhealth?While the LFS provided us with the most representativeannual estimates for the Canadian labour market, it doesnot contain information on health indicators. This infor-mation, if available could have been used to furtherdemonstrate how the LFGI might be applied to researchto examine male/female differences in health status. Toaddress this gap, a conceptual overview of how the LFGImight be included and interpreted in analyses using sec-ondary data is provided below (Fig. 2a-c). We do thisusing directed acyclic graphs (DAGs), which provide aabcFig. 2 a A simple DAG linking male/female to a health outcome of interest. b An extended DAG to include gendered labour market factors(as measured by the LFGI). c A complete DAG to examine the factors that contribute to male/female differences in a given health outcomeSmith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 7 of 9useful approach to understanding the causal relation-ships between variables and interpretation of effects inepidemiological analyses [25–27].Figure 2a presents a simple DAG where there is adifference in a health outcome for men compared towomen (note this difference could be in either direc-tion – i.e. more prevalent among men compared towomen, or more prevalent among women comparedto men). If a difference in the health outcome ispresent among men and women, path “A” in this DAGwill be not equal to zero and will be statistically sig-nificant. For theoretical purposes, the relationship be-tween male/female and the health outcome (path A) isassumed to be adjusted for all confounders, and thatmale/female and the health outcome (along with theadditional variables included in Fig. 2b and c below)have been measured without error.In order to understand why the risk of the health outcomeis greater for men (or women), the DAG is extended to in-clude intermediate or mediating variables, such as the LFGI,to understand the impact of “sex” and “gender” in male/fe-male differences in the health outcome. This extendedmodel is presented in Fig. 2b. Again, for theoretical pur-poses, we make no confounding for all paths and no meas-urement error assumptions for all variables. We now have adirect path (path A′) and an indirect path (paths B and C)linking male/female to the health outcome. The magnitudeof the indirect path will be determined by the strength ofthe relationship between male/female at the LFGI (path B)and the relationship between the LFGI and the health out-come (path C). The estimate for path B is equivalent to theregression estimate for female (relative to male) presented inTable 3 previously. It is important to note that the estimatefrom Table 3 indicates that the LFGI score was stronglyinfluenced, but not completely explained by whether therespondent was male or female. The estimate for the directeffect (path A′) can be interpreted as the difference in thehealth outcome for men compared to women that wouldremain if men and women had similar roles in relation tolabour market status (i.e. if men and women had similarscores on the LFGI). The difference between path A andpath A′ (which in an ordinary least-squared model will beequivalent for the product term of path B and C) [28], canbe interpreted as the amount of the originally observeddifference in the health outcome for men and women thatcan be explained by differences in labour market roles (asassessed by the LFGI) between men and women [27].It is important to note that if path A′ is still associatedwith the health outcome then this indicates that male/female differences in the health outcome are not com-pletely explained by differences in labour market rolesonly. The remaining differences between men andwomen (path A′) will likely be a combination of otherbiological (sex) differences between men and women thatare relevant to the outcome, and other social (gender)differences between men and women that are both rele-vant to the outcome, and not captured in the LFGI. Thishas been explicitly described in Fig. 2c. If data was avail-able to construct – either individually or as part of anindex – all other sex and gender related factors that arerelevant to the outcome of interest, then path A″ in Fig. 2cwould approach zero, and one could examine the relativecontribution of biological factors (paths F and G) and gen-der factors related to labour market roles (paths B and C)and non-labour market roles (paths D and E). The caveatfor Fig. 2c is that each of the three pathways can bemeasured and estimated as distinct from each other. Thishighlights the need for the integration of “sex” and“gender” into the study design and data collection phaseas part of a comprehensive research process [4], so that amore thorough examination of sex and gender into male/female differences in health status can be routinely under-taken using the approach outlined above.ConclusionsIn this paper we developed a measure of feminine andmasculine gender roles, using self-reported survey data,where no direct measure of gender (masculinity/femin-inity) was available. This measure had face validity interms of being related to, but distinct from sex (male/female), and was also responsive to change over time.Future research should examine the relative importanceof including additional measures to an index such as theLFGI. For example, in the study by Pelletier and col-leagues [14] primary earner status was the measure moststrongly related to masculine BRSI scores, while thenumber of hours spent doing housework and responsi-bility for doing housework were the measures moststrongly associated with high feminine BRSI scores. TheLFGI in this paper included some indication of primaryearner status and some indication of household respon-sibilities for respondents. However, a more detailed ordirect measure of primary earner status and householdresponsibilities (in particular housework) may haveallowed further refinement or distinction between mascu-line and feminine roles in the LFGI. In addition, it wouldbe interesting to examine how an index with a reducednumber of measures would perform in differentiatinggender roles for men and women. As mentioned in theintroduction to this paper, the early work by Lippa andcolleagues focused only on differences in occupationalpreferences between men and women [11, 13]. Interest-ingly, occupational segregation, as well as hours worked,was the measure most strongly correlated with the LFGIin our study. We also suggest that research examiningwork-related health outcomes should (and in many casescan) integrate measures of sex and gender, using an ap-proach similar to the one outlined in this paper.Smith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 8 of 9AbbreviationsBSRI, Bem-Sex-Role-Inventory; LFGI, Labour Force Gender Index; LFS, CanadianLabour Force Survey; PAQ, personal attributes questionnaire.AcknowledgementsPeter Smith and Mieke Koehoorn are both supported by Research Chairs inGender, Work & Health from the Canadian Institutes of Health Research.Authors’ contributionsPS and MK were both involved in the conceptual development of this paper.PS performed all data analyses and wrote the first draft of the manuscript.MK provided feedback on the manuscript. Both authors read and approvedthe final manuscript.Competing interestsThe authors declare that they have no competing interests.Author details1Institute for Work & Health, 481 University Avenue, Suite 800, Toronto, ONM5G 2E9, Canada. 2School of Public Health and Preventive Medicine, MonashUniversity, Melbourne, Australia. 3Dalla Lana School of Public Health,University of Toronto, Toronto, ON, Canada. 4School of Population and PublicHealth, Faculty of Medicine, University of British Columbia, 2206 East Mall,Vancouver, BC V6T 1Z3, Canada.Received: 7 October 2015 Accepted: 19 May 2016References1. Ristvedt SL. The evolution of gender. JAMA Psychiatry. 2014;71(1):13–4.2. Phillips SP. Including gender in public health research. Public Health Rep.2011;126(Supp 3):16–21.3. Schiebinger L. Scientific research must take gender into account. Nature.2014;507:9.4. Johnson JL, Greaves L, Repta R. Better science with sex and gender:Facilitating the use of a sex and gender-based analysis in health research.Int J Equity Health. 2009;8:14.5. Doyal L. Sex and gender: the challenges for epidemiologists. Int J HealthServ. 2003;33(3):569–79.6. Messing K, Punnett L, Bond M, Alexanderson K, Pyle J, Zahm S, et al.Be the fairest of them all: Challenges and recommendations for the treatmentof gender in occupational health research. Am J Ind Med. 2003;43:648–29.7. Nowatzki N, Grant KR. Sex is not enough: the need for gender-basedanalysis in health research. Health Care Women Int. 2011;32(4):263–77.8. Springer KW, Mager Stellman J, Jordan-Young RM. Beyond a catalogue ofdifferences: A theoretical frame and good practice guidelines forresearching sex/gender in human health. Soc Sci Med. 2012;74(11):1817–24.9. Bem SL. The measurement of psychological androgyny. J Consult ClinPsychol. 1974;42(2):155–62.10. Eisler RM, Skidmore JR. Masculine gender role stress: Scale developmentand component factors in the appraisal of stressful situations. Behav Modif.1987;11(2):123–36.11. Lippa R, Connelly S. Gender Diagnosticity: a new Bayesian approach togender-related individual differences. J Personal Soc Psychol. 1990;59(5):1051–65.12. Spence JT, Helmreich R, Stapp J. The Personal Attributes Questionnaire:a measure of sex role stereotypes and musculinity-feminity. CatalogeSelected Doc Psychol. 1974;4(43):Ms No 617.13. Lippa RA, Martin LR, Friedman HS. Gender-related individual differences andmortality in the Terman Longitudinal Study: Is masculinity hazardous toyour health? Pers Soc Psychol Bull. 2000;26(12):1560–70.14. Pelletier R, Ditto B, Pilote L. A composite measure of gender and itsassociation with risk factors in patients with premature acute coronarysyndrome. Psychosom Med. 2015;77(5):517–26.15. Borkhoff CM, Hawker GA, Kreder HJ, Glazier RH, Mahomed NN, Wright JG.The effect of patients’ sex on physicians’ recommendations for total kneearthroplasty. CMAJ. 2008;178(6):681–7.16. Leung Yinko SS, Pelletier R, Behlouli H, Norris CM, Humphries KH, Pilote L,et al. Health-related quality of life in premature acute coronary syndrome:does patient sex or gender really matter? J Am Heart Assoc. 2014;3(4):e000901.17. Mosca L, Barrett-Conner E, Wegner NK. Sex/gender differences incardiovascular disease prevention: what a difference a decade makes.Circulation. 2011;124(19):2145–54.18. Statistics Canada. A guide to the Labour Force Survey. Ottawa: StatisticsCanada; 2011. Report No.: 71-543-G.19. Statistics Canada. Labour Force Survey Public Use Microdata Files. 2013.Available through Statistics Canada’s Data Liberation Initiative.20. Frenette M, Zeman K. Why Are Most University Students Women? EvidenceBased on Academic Performance, Study Habits and Parental Influences.Ottawa: Statistics Canada; 2007. Report No.: Catalogue no. 11F0019MIE, no.303.21. Statistics Canada. College graduates by program level, Classification ofInstructional Programs, Primary Grouping (CIP_PG) and sex, annual(number). Postsecondary Student Information System (PSIS), CANSIM Table477-0016. 2011.22. Human Resources & Skills Development Canada. National OccupationalClassification Career Handbook. 2nd ed. Ottawa: Government of Canada; 2011.23. The SAS Institute. The SAS System for Windows, Release 9.3. Cary: SAS; 2011.24. Hoffman RM, Borders L. Twenty-five years after the Bem Sex-Role Inventory:a reassessment and new issues regarding classification variability. Meas EvalCouns Dev. 2001;34(1):39–55.25. Glymour MM. Using Causal Diagrams To Understand Common Problems inSocial Epidemiology. In: Oakes JM, Kaufman JS, editors. Methods in SocialEpidemiology. San Francisco: Jossey-Bass; 2006. p. 387-422.26. Schisterman EF, Cole SR, Plat RW. Overadjustment bias and unnecessaryadjustment in epidemiolgical studies. Epidemiology. 2009;20(4):488–95.27. VanderWeele TJ, Robinson WR. On the causal interpretation of race inregressions adjusting for confounding and mediating variables.Epidemiology. 2014;25(4):474–84.28. MacKinnon DP, Fairchild AJ, Fritz MS. Mediation analysis. Annu Rev Psychol.2007;58:593–614.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Smith and Koehoorn International Journal for Equity in Health  (2016) 15:82 Page 9 of 9


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items