1
Online supplement for McClintock, E. A. (2017). Changing jobs and changing chores: Gender occupations and housework performance. Sex Roles.Elizabeth Aura McClintock, University of Notre Dame. Email:
ONLINE SUPPLEMENT A: ROBUSTNESS
In this online supplement, I present results of several robustness checks. The measures and the result of the analyses are described below and presented in Tables A1 and A2.
Measures
I follow England and colleagues (England, Herbert, Kilbourne, Reid, & Megdal, 1994) in developing indicators for authoritative and nurturing jobs. Although they devised these measures for the 1980 Census codes, it is easily adapted to later census coding systems (see McClintock 2017). Occupational titles with the word manager, supervisor, or administration are authoritative. Occupations in which incumbents spend a major share of working hours directly serving clients are nurturing. These measures were employed in earlier studies of occupational sex composition and housework (McClintock, 2017; Schneider, 2012).
I also consider broad occupational groups, following US Census 1990 classifications: professional, sales, service, farm/forestry/fish, craft, operatives/laborers. The 1990 Census has pre-categorized occupations into these groups and it is easy to adapt the schema for subsequent Census years. In addition to varying in average sex composition, these occupational groups also vary in other relevant characteristics such as income, social status, opportunities for career advancement, and educational requirements. Most importantly, they arguably vary in cultural meanings, such as “blue collar” masculinity, which are difficult to assess directly. I use sales as the reference category because it is approximately gender-balanced and employs a substantial proportion of female and male PSID respondents.
Finally, I create measures of occupational autonomy, supervisory responsibility, opportunities for advancement, and performance-based pay from General Social Survey (GSS, 2017) data. In addition to core GSS data, these measures draw on several topical modules, including those on work organizations, work compensation and shared capitalism. Occupational autonomy is in index capturing respondents’ control over their work, participation in workplace decision-making and evaluations, and opportunity to provide input (Cronbach’s alphas of 0.83). Supervisory responsibility is an indicator that the respondent supervises other employees. Opportunities for advancement captures respondents’ perception that their job offers the possibility of future promotion. Performance-based pay indicates that the respondent is paid in part by commission, bonuses, or tips. I code these measures at the individual-level and then aggregate responses to the occupational level so that they represent the mean score among all GSS respondents in a given occupation. Career is a scale aggregating these four measures and indicates the degree to which an occupation qualifies as a career, as opposed to being merely a job.
Results
Model 1, Removing spouses’ characteristics: Although I include spouses’ housework and work hours in the models presented in the main text, these measures are somewhat endogenous to respondent’s housework hours. In Model 1 I demonstrate that results are robust to removing spouses’ housework, occupation, income, and work hours from the model. This little changes the association between respondents’ occupation and respondents’ housework hours.
Model 2, Full-time workers: Another possible concern might be that results are driven by part-time workers. However, in Model 2 I demonstrate that results remain consistent when the sample is limited to full-time workers (both spouses work thirty-five or more hours).
Model 3, Prime working age: Yet another concern might be that occupational sex composition is serving as a proxy for career or life stage. In Model 3 I demonstrate that results remain robust when the sample is limited to respondents of prime working age, which I define as 30-50 years. I selected this age range because such workers are likely to have completed their education and settled into lasting occupations but are not yet near retirement age.
Model 4 & Model 5, Occupational characteristics: It is possible that occupational sex composition is serving as a proxy for some other occupational characteristic that itself influences housework. However, in Model 4 I show that results remain robust when I include a scale capturing the degree to which an occupation qualifies as a career or a mere job. Although women decrease their housework when moving into a more career-typed occupation, including the career scale does not alter the within-individual association of occupational sex composition and housework. Model 5 presents results removing occupational sex composition and including measures of occupational autonomy, supervisory responsibility, opportunities for advancement, and performance-based pay (the individual components of the career scale).
Model 6, Occupational groups: As alternative measures of gendered work, I substitute occupational groups for occupational sex composition. Men who move from sales (gender-mixed) into professional occupations (mostly male) decrease their housework; otherwise, there is little within-individual association. Differences are more evident between individuals and suggest that men in female-typed occupation groups (like service) may do more housework than men in gender-mixed jobs (the reference, sales), whereas men in male-typed occupational groups (farm/forestry/fish) do less housework. However, these classifications are problematic for women because very few women work in the most strongly-male occupational groups and also because they generally work in female-gendered occupations within these male-typed occupational groups. For example, for those few women classified as working in “craft” occupations, their average proportion female is 69%, whereas for men it is only 7%. In other words, such women usually work in predominately female occupations despite these jobs falling into an occupational group that is predominately male. Thus, occupational sex composition is a much cleaner measure of gender-typed employment.
Model 7, Authority and Nurturance: As yet another measure of gendered work, in Model 5 I use classifications of occupations as nurturing (female-typed behavior) or authoritative (male-typed). Men who enter authoritative jobs reduce their housework, reflecting the within-individual reduction in housework for men who enter professional occupations. Similarly, women who enter authoritative jobs reduce their housework whereas women who enter nurturing jobs increase their housework. These within-individual changes provide some support for acclimation—the experience, context, or social associations of gendered paid work may have implications for housework. Considering between-individual variation, men and women who work in authoritative occupations do less housework; women also do less housework when their husbands work in nurturing occupations. Thus, results are largely robust to this alternative measure of occupational gender-typing.
Model 8, Multiple Imputation: In the main text I present models estimate dropping cases with missing data. In Model 8, I show that results are robust to using multiple imputation. Using multiple imputation increases the sample size by about 10% but does not change findings.
References
England, P., Herbert, M., Kilbourne, B., Reid, L. L., & Megdal, L. M. (1994). The gendered valuation of occupations and skills: Earnings in 1980 census occupations. Social Forces, 73(1), 65–100. doi: 10.2307/2579918
General Social Survey (GSS). (2017, May 31). The General Social Survey. Chicago, IL: NORC at the Univerisity of Chicago. Retrieved from
McClintock, E. A. (2017). Occupational sex composition and gendered housework performance: Compensation or Ccnventionality? Journal of Marriage & Family, 79(2), 475–510. doi: 10.1111/jomf.12381
Schneider, D. (2012). Gender deviance and household work: The role of occupation. American Journal of Sociology, 117(4), 1029–1072. doi: 10.1086/662649
1
Table A1. Coefficients from mixed-effects models predicting changes in men’s weekly hours of housework. Married dual-earner couples. Panel Study of Income Dynamics 1981-2013.
Model # / 1 / 2 / 3 / 4 / 5 / 6 / 7 / 8Model description / No spouse / Full-time, both 35+ / Age 30-50 / Occ. career / Occ. traits / Occ. Groups / Auth. & nurture / Impute (MIM)
Within-individual effects
Occ. sex composition
His occupation (0-1) / -.10 / .26 / .27 / .03 / .07
Her occupation (0-1) / -.70* / -.58* / -.42* / -.49*
Career-type occupation
He has career / .15
She has career / 1.05
Occ. groups (% female)
Him, professional (38%) / -.43*
Him, sales (48%) / REF
Him, service (42%) / -.44
Him, farm/forestry (16%) / .05
Him, craft (07%) / -.20
Him, operative/labor (21%) / -.24
Him, miscellaneous (14%) / .61
Her, professional (66%) / -.27*
Her, sales (67%) / REF
Her, service (69%) / -.44*
Her, farm/forestry (66%) / -.77
Her, craft (69%) / .28
Her, operative/labor (67%) / -.06
Her, miscellaneous (64%) / .46
Gendered work behavior
His occ. Authority / -.28*
His occ. Nurturance / .01
Her occ. Authority / .02
Her occ. Nurturance / -.19
Occupational characteristics
His occupational autonomy / .15
His supervisory role / -.19
His job opportunity / -.35
His performance pay / -.39
Her occupational autonomy / 1.09
Her supervisory role / .09
Her job opportunity / .44
Her performance pay / .78*
Income (2015 $10,000s)
His income (logged) / -.14 / .36* / -.06 / -.20 / -.28 / -.21 / -.18 / -.26
Her income (logged) / -.19 / .14 / .20* / .27* / .22* / .20* / .20*
Relative income
Relative income (0-1) / -2.19* / -3.12* / -1.81 / -1.47 / -1.25 / -1.36* / -1.67 / -.71
Slope change at equality / 1.29 / -1.42 / -.36 / .21 / .44 / .16 / .51 / -.41
Between-individual effects
Occ. sex composition
His occupation (0-1) / -.32 / .01 / -.09 / .36 / .28
Her occupation (0-1) / -1.03* / -.52 / -.28 / -.35
Career-type occupation
He has career / 1.48
She has career / .79
Occ. groups (% female)
Him, professional (38%) / -.16
Him, sales (48%) / REF
Him, service (42%) / 1.93*
Him, farm/forestry (16%) / -2.07*
Him, craft (07%) / .04
Him, operative/labor (21%) / .33
Him, miscellaneous (14%) / .49
Her, professional (66%) / -.04
Her, sales (67%) / REF
Her, service (69%) / -.50*
Her, farm/forestry (66%) / .82
Her, craft (69%) / .65
Her, operative/labor (67%) / -.79*
Her, miscellaneous (64%) / 4.07
Gendered work behavior
His occ. Authority / -.56*
His occ. Nurturance / -.00
Her occ. Authority / .41
Her occ. Nurturance / .41*
Occupational characteristics
His occupational autonomy / 6.11*
His supervisory role / .57
His job opportunity / 1.04
His performance pay / -1.70*
Her occupational autonomy / -1.45
Her supervisory role / .12
Her job opportunity / .08
Her performance pay / .64
Income (2015 $10,000s)
His income (logged) / -.32* / -.19 / -.71* / -.74* / -.87* / -.79* / -.69* / -.58*
Her income (logged) / .47 / 1.20* / .95* / 1.16* / .92* / .99* / .91*
Relative income
Relative income (0-1) / -3.01* / -1.57 / -1.66 / -2.78* / -1.55 / -3.08* / -2.80* / -2.46
Slope change at equality / 1.23 / -.31 / 4.41* / 5.11* / 4.44* / 5.51* / 4.89* / 4.49*
N (person-years) / 48,721 / 24,022 / 29,892 / 44,902 / 36,913 / 44,921 / 46,142 / 50,848
*p.05, NA=Not applicable, REF=Reference. Models include all controls in Table 4 (work hours, spouse’s housework (except Model 1), personal/family traits, survey year, respondent, race).
Table A2. Coefficients from fixed-effects models predicting changes in women’s weekly hours of housework. Married dual-earner couples. Panel Study of Income Dynamics 1981-2013.
Model # / 1 / 2 / 3 / 4 / 5 / 6 / 7 / 8Model description / No spouse / Full-time, both 35+ / Age 30-50 / Occ. career / Occ. traits / Occ. groups / Auth. & nurture / Impute (MIM)
Within-individual effects
Occ. sex composition
His occupation (0-1) / -.47 / -.51 / -.46 / -.44
Her occupation (0-1) / .87* / .88* / .79* / .99* / .92*
Career-type occupation
He has career / .96
She has career / -1.92*
Occ. groups (% female)
Him, professional (38%) / .16
Him, sales (48%) / REF
Him, service (42%) / .35
Him, farm/forestry (16%) / .85
Him, craft (07%) / .32
Him, operative/labor (21%) / .08
Him, miscellaneous (14%) / 1.60*
Her, professional (66%) / -.21
Her, sales (67%) / REF
Her, service (69%) / 1.46*
Her, farm/forestry (66%) / .14
Her, craft (69%) / .44
Her, operative/labor (67%) / .63
Her, miscellaneous (64%) / -3.82
Gendered work behavior
His occ. Authority / .08
His occ. Nurturance / -.21
Her occ. Authority / -.50*
Her occ. Nurturance / .66*
Occupational characteristics
His occupational autonomy / -.70
His supervisory role / .14
His job opportunity / .28
His performance pay / .27
Her occupational autonomy / .14
Her supervisory role / -.16
Her job opportunity / -.77
Her performance pay / -.86*
Income (2015 $10,000s)
His income (logged) / -.70* / .21 / .20 / .35 / .26 / .17 / .08
Her income (logged) / -.96* / -.33 / -.96* / -1.11* / -1.24* / -1.11* / -1.06* / -.88*
Relative income
Relative income (0-1) / -.29 / 2.93 / .64 / -.43 / -.98 / -.75 / .01 / .23
Slope change at equality / 2.50* / 3.95* / 1.04 / 1.93 / 1.85 / 2.07 / 1.75 / 2.19
Between-individual effects
Occ. sex composition
His occupation (0-1) / -2.87* / -2.93* / -2.87* / -2.91*
Her occupation (0-1) / .50 / 1.51* / .90 / .53 / .65
Career-type occupation
He has career / -1.55
She has career / -1.05
Occ. groups (% female)
Him, professional (38%) / .35
Him, sales (48%) / REF
Him, service (42%) / .50
Him, farm/forestry (16%) / 2.59*
Him, craft (07%) / 1.80*
Him, operative/labor (21%) / 1.44*
Him, miscellaneous (14%) / .63
Her, professional (66%) / -.31
Her, sales (67%) / REF
Her, service (69%) / 1.10*
Her, farm/forestry (66%) / 3.77*
Her, craft (69%) / -.99
Her, operative/labor (67%) / 1.56*
Her, miscellaneous (64%) / -1.02
Gendered work behavior
His occ. Authority / -.13
His occ. Nurturance / -1.32*
Her occ. Authority / -1.32*
Her occ. Nurturance / .20
Occupational characteristics
His occupational autonomy / .22
His supervisory role / -.02
His job opportunity / -.82
His performance pay / .13
Her occupational autonomy / 3.22
Her supervisory role / -.38
Her job opportunity / -2.24*
Her performance pay / .08
Income (2015 $10,000s)
His income (logged) / -2.06* / -1.08* / -.89* / -.83* / -.36 / -.91* / -.98*
Her income (logged) / -2.41* / -1.05* / -1.97* / -1.95* / -2.20* / -1.79* / -1.98* / -1.88*
Relative income
Relative income (0-1) / 4.15* / 10.33* / 6.81* / 6.79* / 5.57* / 4.52* / 6.53* / 7.01*
Slope change at equality / -2.02 / -3.17 / -1.48 / -2.72 / -1.44 / -.78 / -1.94 / -2.78
N (person-years) / 47,845 / 24,022 / 29,662 / 44,902 / 36,913 / 44,921 / 46,142 / 50,848
*p.05, NA=Not applicable, REF=Reference. Models include all controls in Table 4 (work hours, spouse’s housework (except Model 1), personal/family traits, survey year, respondent, race).
1
Online supplement for McClintock, E. A. (2017). Changing jobs and changing chores: Gender occupations and housework performance. Sex Roles.Elizabeth Aura McClintock, University of Notre Dame. Email:
ONLINE SUPPLEMENT B: OCCUPATIONAL SEX COMPOSITION
In this online supplement, I present further details on my calculations of occupational sex composition and the extent to which occupational sex composition is an accurate measure of gendered employment.
Calculating Occupational Sex Composition
Percent occupation female is calculated from U.S. Decennial Census data and American Community Survey (ACS) data. The PSID provides 1970 Census 3-digit codes from 1981-2001 and 2000 Census 3-digit codes from 2003-2013. If I were to use US Census 1970 data to calculate occupational sex composition for 1981-2001 and US Census 2000 data to calculate occupational sex composition for 2003-2013, this would create two sources of inaccuracy. First, occupational sex composition changes gradually over time, as occupations feminize (or, rarely, masculinize). Using US Census 1970 data for PSID years 1981-2001 and US Census 2000 data for PSID years 2003-2013 would ignore this gradual change and would create a disjoint between PSID years 2001 and 2003. That is, 2001 values for occupational sex composition would reflect the realities of 1970, in sharp contrast with 2003 values which would leap forward 30 years (from 1970 to 2000). Second, occupational codes themselves change overtime to capture development/decline of new/old occupations. For workers whose occupations are reclassified, this would create a further disjoint between 2001 (using 1970 codes) and 2003 (using 2000 codes).
To address these two concerns, I standardize occupational codes over time and linearly interpolate gradual changes in occupational sex composition. To do this, I employ the crosswalk developed by Integrated Public Use Microdata Series (Ruggles et al., 2010) which links 1970, 1980, and 2000 codes to standardized 1990 codes. Thus, all years of PSID data use the same occupational codes (standardized 1990 codes) even though the PSID initially provides occupational codes using two different coding systems (1970 and 2000). In addition, I use 1980, 1990, 2000 US Census data and 2001-2013 ACS data to linearly interpolate occupational sex composition, distributing changes equally across each decade. That is, I assume that any change in a given occupation’s sex composition from 1980 to 1990 is distributed linearly across those ten years, change from 1990 to 2000 is distributed linearly across those ten years, and so on. This provides a more accurate and temporally-continuous measure of occupational sex composition over time. It consistently captures the two processes whereby individuals’ occupational sex compositions change: (1) gradually over time, as occupations feminize (or masculinize), and (2) discretely, when individuals change jobs.
Table B1 illustrates the results of this process. It presents annual values for occupational sex composition for every year of PSID data used in this analysis (1981-2013), separately by gender. It also provides the sample size of married men and women for each of these years. As is to be expected, regardless of the year, women work in mostly-female (65-71% female) occupations and men work in mostly-male occupations (26-32% female). There is a slow trend toward occupational desegregation over time, with men’s occupations becoming slightly more-female and women’s occupations becoming slightly less-female.
Occupational Sex Composition as a Measure of Gendered Employment
Occupational sex composition, calculated as described above, is based on national trends and may not capture individual experiences of gendered employment—that is, a given occupation’s composition may vary at the local or establishment level, altering individual experiences of employment in that occupation. However, as I will explain below, national trends may actually have greater theoretical and practical salience than individual experiences. In addition, national trends and individual experiences are in fact highly correlated—national-level measures of occupational sex composition are a good proxy for individual experiences, especially in strongly-gendered occupations. Finally, any measurement error whereby individual experiences differ from national level trends is more likely to create random “noise” in the data, increasing standard errors, than to systematically bias coefficient estimates.
Theoretical Salience of National Trends and Individual Experience
I consider three theoretical perspectives: gender-deviance neutralization (GDN), self-selection into occupations, and occupational acclimation. For both GDN and self-selection into occupations, it is the popular perception of occupations as male or female-typed that is most salient, regardless of an individual’s actual workplace sex composition. That is, both of these theories address the gendered stereotypes labeling occupations as “normal” or “suitable” for a given gender, and as requiring gender-typed behavior. Occupational sex composition is a very good proxy for this aspect of occupational gender-typing. The degree of sex-segregation in an occupation is directly related the expectation that gender-stereotypical traits are required for success in that occupation (Cejka & Eagly, 1999). Also, as discussed in more detail below, there is very little regional variation in the sex composition of occupations that are skewed male or female (gender-mixed occupations exhibit greater regional variation: Perales and Vidal, 2015).
In contrast, occupational acclimation proposes that gendered work experiences might alter housework performance over time. This might occur through contact with co-workers, the experience of performing male or female-typed tasks on the job, or an adjustment of gender attitudes resulting from the social image of employment in a gender-(a)typical occupation. The first of these process, interactions with co-workers, depends on the sex composition of the workplace—I consider the degree to which occupational sex composition reflects workplace sex composition below. But the latter two processes depend on the sex composition of the occupation, not that of the workplace. Thus, overall, it is actually the sex composition of the occupation that is theoretically relevant, not the sex composition of the workplace. Moreover, it is largely the perception of an occupation as male or female-typed that is relevant, and local or workplace variations in occupational sex composition may not impact these widely-shared perceptions.