Social Justice issues

Greedy Jobs, Labour Market Institutions, and the Gender Pay Gap*

KRISTEN SOBECK

Tax and Transfer Policy Institute, Crawford School of Public Policy, Australian National University, Canberra, Australian Capital Territory, Australia

Previous research argues that occupational gender pay gaps arise from greedy jobs within occupations. Greedy jobs involve working long and unpredictable hours in jobs where individuals are not easily substitutable. They engender compensating differentials resulting in an earnings-to-hours elasticity that often exceeds 1. This paper shows that greedy jobs also exist in Australia, where labour market institutions differ substantially from those in the United States. It shows that occupational gender earnings gaps are highest in occupations where greedy jobs proliferate. Wage-setting institutions engen- der heterogeneous effects on occupational gender earn- ings gaps. Relative to the United States, occupational gender earnings gaps are smaller in Australia, consistent with evidence that labour market institutions compress the earnings distribution. Within occupations, the use of collective agreements attenuates the size of occupational gender earnings gaps, while the use of individual agreements increases them. Not surprisingly, individuals employed in greedy occupations predominantly use individual agreements to negotiate pay.

I Introduction The early literature on the gender pay gap

focused on the observable and unobservable characteristics of men and women (Altonji & Blank, 1999), attributing differences in pay to human capital accumulation and

discrimination. Over time, however, as women’s educational achievements have surpassed those of men – largely eliminating the explanatory power of human capital as a source of variation – gender pay gaps have persisted. As a result, new streams of

*I am grateful for the the feedback provided by colleagues from the Research School of Economics applied micro seminar series, the Australian Gender Economics Workshop, and the Australian Labour Market Research Workshop. I also appreciate feedback received from Robert Breunig, Nathan Deutscher, Claudia Goldin, Jennifer Graves, Guyonne Kalb, Maria Racionero, Leonora Risse, Yixiao Zhou and two anonymous referees. A special thank-you goes to Nathan, in particular, for the extra encouragement to publish this paper. This research was supported through an Australian Government Research Training Program Scholarship.

JEL classifications: J16, J31, J32 Correspondence: Kristen Sobeck, Tax and Transfer Policy Institute, Crawford School of Public Policy,

Australian National University, JG Crawford Building 132, Lennox Crossing, Canberra, ACT 2601, Australia. Email: [email protected]

462

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. doi: 10.1111/1475-4932.12824

ECONOMIC RECORD, VOL. 100, NO. 331, DECEMBER, 2024, 462–490

literature have emerged that point to occu- pation as an important factor underlying the remaining differences in pay observed between men and women (Bertrand, 2011, 2020; Goldin, 2014, 2021; Pan, 2015; Blau & Kahn, 2017; Cortes & Pan, 2018). One area of this occupationally focused

research centres on gender differences in preferences for workplace flexibility. In particular, some occupations reward long hours and hours worked at specific times of the day; these occupations are ‘greedy’ (Goldin, 2014, 2021). Greedy occupations are mostly or entirely comprised of greedy jobs. By contrast, other, flexible occupations do not reward long or particular hours worked. These occupations are entirely or mostly comprised of ‘not so needy’, flexible jobs. Workers employed in flexible occupa- tions are largely substitutable, either by design (i.e., obstetricians and gynaecolo- gists) and/or as a result of technological improvements over time (i.e., pharmacists) (Goldin, 2014, 2021; Goldin & Katz, 2016). Large gender pay gaps emerge in greedy occupations as a result of differences in male and female preferences (Pertold-Gebicka et al., 2016; Lundborg et al., 2017; Wiswall & Zafar, 2018; Mas & Pallais, 2020; Le Barbanchon et al., 2021; Meekes & Has- sink, 2022; Wasserman, 2022) for greedy and flexible employment, partially shaped by social norms. Greedy jobs also proliferate in the United

States due to features specific to the Amer- ican labour market. In particular, in the United States, wages are predominantly negotiated individually between a worker and her employer. Arguably, greedy jobs exist because employers have the ability to compensate individuals to undertake long and unpredictable hours. Yet, in many countries across the globe, labour market institutions (i.e., trade unions, collective bargaining, minimum wages) play a much more important role in wage-setting. These institutions put greater emphasis on collec- tive negotiation of wages for groups of workers, rather than placing the onus of negotiation on the individual. For example, in Australia, 61.2 per cent of employees had their pay set through collective bargaining in 2018, relative to 11.7 per cent of employees in the United States (ILOSTAT). Across 19

countries in northern, western and southern Europe,1 the median country’s collective bargaining coverage rate was 76.7 per cent of employees in 2018 (ILOSTAT). In these countries, one might expect greedy jobs to be less greedy to the extent that labour market institutions constrain employers’ discretion to compensate individuals per- forming similar work differently. In addition to potentially reducing jobs’

greediness, high rates of collective bargain- ing should also reduce occupational gender pay gaps. A large literature shows that labour market institutions’ compress coun- tries’ wage distributions (for a review, see Hayter, 2011; Daehoon et al., 2022). Free- man and Medoff (1984) argue that unions take into account the preferences of the median union member when they negotiate. Since the median union worker is often lower skilled and remunerated less than the average worker, union negotiations raise the wage floor higher than it would otherwise be, compressing the wage distribution (‘wage floor effect’). The authors also argue that since unions negotiate wage levels and wage increases for groups of union members, they reduce the role of managerial discretion in wage-setting and, as a consequence, varia- tion in pay across individuals undertaking the same work (‘transparency effect’). Importantly, this second effect ensures transparency as part of the negotiation process (i.e., unions negotiate the same pay for men and women), but also as part of the outcome (i.e., any employee can consult the final collective agreement and view the negotiated wage levels and increases). Both the ‘wage floor’ and ‘transparency’

effects engendered by such labour market institutions can help reduce the gender pay gap. If women’s wages are disproportion- ately at the bottom of the wage distribution, the ‘wage floor’ effect will exert upward pressure on female wages in general, but also in occupations where they are concen- trated. Meanwhile, the ‘transparency’ effect can help reduce the gender pay gap in

1 Albania, Austria, Belgium, Switzerland, Ger- many, Denmark, Spain, Estonia, Finland, France, Greece, Iceland, Italy, Lithuania, Latvia, the Netherlands, Portugal, Sweden and the United Kingdom. The median country is the Netherlands.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 463

circumstances where women are paid less than men for the same or similar work. A growing body of research points to pay transparency – achieved through a collective agreement or other mechanism, such as legislation – as a successful means to reduce gender pay gaps (Cullen & Pakzad- Hurson, 2021; Bennedsen et al., 2022; Baker et al., 2023). Transparency can help redress pay inequity because recent research shows that women are reluctant to negotiate their salary or are more likely to propose a lower asking salary for a job for which they have applied (Card et al., 2015; Leibbrandt & List, 2015; Biasi & Sarsons, 2021; Rous- sille, 2022). Trade unions and collective bargaining can contribute to reductions in the gender pay gap in other ways too, such as through gender-neutral job classifications used in Belgium and in selected sectors in the United Kingdom, Germany and Canada (OECD, 2020). Given the broad coverage and decentralised

design of the industrial relations system in Australia, it is an appropriate context to examine the role of labour market institutions on greedy jobs and occupational gender pay gaps. Historically, throughout most of the twentieth century, wages in Australia were determined centrally by awards, which set wages for particular industries or occupations. The wage levels and wage increases stipulated in awards were determined through compul- sory conciliation and arbitration (for a succinct summary, see Daehoon et al., 2022). Wage increases applied to all workers (union and non-union) covered by a particular award, nationally or within a particular jurisdiction. Since the 1990s, however, the industrial relations system has decentralised to an enterprise-based bargaining system, with mod- ern awards2 and a national minimum wage acting as the wage floor. Decentralised enterprise-based systems have the potential to reduce the flow-on effects of negotiated wage increases, since they only apply to (union and non-union) workers in a particular enterprise, rather than to all workers (in a particular sector or occupation) across all enterprises. As a

result, unlike countries with more centralised industrial relations systems – where workers in a particular occupation are likely to be subject to the same national collective agreement – the Australian system has greater heterogeneity in wage-setting mechanisms for workers in the same occupation. This variation is exploited in this paper in order to determine if wage-setting mechanisms have an impact on occupational gender pay gaps. This paper makes two contributions to the

literature. First, given the different institu- tional settings in place in Australia relative to the United States, it replicates analyses conducted in Goldin (2014, 2021) with Australian data in order to confirm whether greedy jobs exist in Australia and assess how they differ from the United States. Second, it extends the analyses for Australia to under- stand the extent to which labour market institutions affect occupational gender pay gaps. The results confirm that greedy occu- pations exist in Australia and occupational pay gaps tend to be smaller in Australia, consistent with wage compression effects associated with strong labour market insti- tutions (such as high rates of collective bargaining). When occupational pay gaps are calculated exclusively for workers in Australia whose pay is set by individual agreement, pay gaps are larger and more similar to those observed in the United States. Across occupations, the greediest – defined as those occupations with an earnings-to-hours elasticity that exceeds one – disproportionately employ men, pre- dominantly use individual agreements to set pay, and have the largest occupational gender pay gaps. In addition, while the use of individual agreements among men and women within the same occupation widens the gender earnings gap, the use of collective agreements reduces it. The next section discusses the theoretical

framework that underpins the greedy jobs hypothesis and summarises the related empirical research. This paper does not develop a new theoretical framework, but instead applies and briefly describes the one provided by Goldin (2014). Section III describes the data used for the paper. Section IV discusses descriptive trends on pay gaps and measures of occupational segregation in Australia. Section V confirms

2 https://www.fairwork.gov.au/tools-and-resources/ fact-sheets/minimum-workplace-entitlements/modern- awards.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

464 ECONOMIC RECORD DECEMBER

that greedy occupations exist in Australia and discusses some of the sources of difference in pay gaps observed in the two countries, which includes the role of labour market institutions. Section VI concludes.

II Theoretical Background and Literature Review

In Goldin (2014), the author presents her theoretical model and argues that differences between the earnings of men and women employed in the same occupation arise due to compensating differentials (Rosen, 1987). Her theoretical framework is applied in this paper. Compensating differentials emerge from matching workers’ preferences with firms’ costs. ‘Temporal flexibility’ is one type of firm cost and is defined as the firm’s cost of allowing a worker to choose when, as well as the number of hours, she works. The cost of temporal flexibility can vary by job within an occupation. The cost of temporal flexibility is high (low) for greedy (flexible) jobs. Among workers, some employees greatly value temporal flexibility. Worker preferences for temporal flexibility can also change over time. The cost of providing temporal flexibility

decreases as worker substitutability increases. For example, as a result of a shift in management away from self-employed pharmacists to corporate ownership, the extensive use of IT systems which enhanced the ability to share information across pharmacists and pharmacies, and the stan- dardisation of pharmacy products and ser- vices (i.e., medications are produced by pharmaceutical companies rather than com- pounded by individual pharmacists on-site), pharmacists have become more readily substitutable. Pharmacists represent a flexi- ble occupation. They are also one of the most egalitarian professions in terms of pay (Goldin & Katz, 2011, 2016; Goldin, 2014). Goldin (2021) describes similar ongoing changes among veterinarians enabling vets to become more substitutable. For example, veterinary clinic ownership is shifting from individuals to the corporate sector. She also provides evidence of worker substitution among selected medical doctor specialties (e.g., obstetricians, gynaecologists, anaes- thetists, paediatricians), partially as a result of the development of group practices. By

contrast, in greedy employment, the cost of worker substitution is quite high. Lawyers and MBA graduates are often employed in greedy jobs (Goldin & Katz, 2011; Goldin, 2014), since client-specific knowl- edge, accumulated over time, can be quite costly for firms to transfer or delegate to another employee. Since greedy jobs require specific individ-

uals to work long hours and/or at potentially unpredictable times of the day, workers receive a compensating differential, reflected by nonlinear pay structures and an earnings- to-hours elasticity that often exceeds 1. The elasticity of an occupation will depend on the share of greedy and flexible jobs within it. For example, not all lawyers are employed in greedy jobs. Some lawyers work in less prestigious firms with more flexibility, fewer time pressures and a more linear pay structure. The average elasticity for the occupation will, however, include all lawyers in both greedy and flexible jobs. The greater the elasticity, the greedier the jobs, and/or the greater the share of greedy employment in an occupation. Gender pay gaps will arise in greedy occupa- tions if women disproportionately select jobs, within an occupation, associated with greater temporal flexibility. For example, if men aim to become law partners at tier-one firms and women disproportionately choose to work as lawyers in small local firms with fixed hours, a gender wage gap will emerge as a result of the compensating differential disproportionately received by men. Empirically, research suggests that

women do express a greater preference for flexible jobs. While both sexes are willing to pay for non-wage attributes, the amount varies across studies and by gender (Mas & Pallais, 2020). Wiswall and Zafar (2018) estimate preferences for job amenities using a sample of undergraduates attending New York University. The authors surveyed students during university and several years later when the same students were employed in their mid-twenties. They found that women have a much higher willingness to pay for workplace hours flexibility than men. Differences in workplace preferences by gender emerge through two channels: first, selection of field of study (choice of major); and second, selection of occupations within a particular field. The authors find that the

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 465

gender wage gap among early-career gradu- ates is mostly operating through this second channel and is a result of the associated costs of ‘purchasing’ workplace attributes, such as workplace hours flexibility. Within households, women also tend to

prioritise flexible jobs, especially following the arrival of children. This suggestion is consistent with literature showing changes in female labour force decisions when women become mothers. Pertold-Gebicka et al. (2016) use Danish data to show that, following the birth of their first child, mothers tend to switch from the private to the public sector. The authors explain this through mothers’ preferences for particular job characteristics, operating through com- pensating differentials. They argue that mothers choose employers that accommo- date their preferences for flexibility. Wasser- man (2022) shows that a reform that capped the average working week for medical residents at 80 hours induced women to enter specialties that had exceedingly high hours prior to the reform. Lundborg et al. (2017) evaluate the impact of success- ful in-vitro fertilisation (IVF) on the labour market outcomes of mothers (in comparison to unsuccessful women who attempted IVF). They find that mothers reduce labour supply on the extensive and intensive margins. If mothers continue to work, they tend to work fewer hours in the short term, and in the longer term they change to lower-paid jobs that are closer to home. Le Barbanchon et al. (2021) reach similar conclusions with respect to commuting distances in France. They estimate that around 10 per cent of the gender wage gap can be attributed to gender differences in the willingness to pay for a shorter commute and that this effect is supply-side driven. The preference for a shorter commute is also observed for all women, not just mothers. Meekes and Hassink (2022) evaluate gender differences in employment, hours worked, unemploy- ment duration and commuting distance following job displacement as a result of firm bankruptcy. The authors show that displaced women take longer to find a job than displaced men. However, once women find a new job, their commute time is shorter and they work fewer hours. Interestingly, displaced women experience similar hourly

wage losses to men, suggesting that longer job search enables women to find employ- ment that corresponds to their commuting time and working time preferences. Changes over time also alter preferences

and willingness to pay for non-wage ameni- ties. Recent empirical work suggests that greedy jobs have grown greedier. The returns from working long hours have increased over time (Kuhn & Lozano, 2008; Cha & Wee- den, 2014; Cortes & Pan, 2016; Denning et al., 2022), influencing women’s labour force decisions. Cortes and Pan (2017) use cross-country data to show that the labour force participation rates of tertiary educated women aged 23–57 decline as the share of full-time skilled males working more than 50 h per week increases. They argue that the prevalence of overwork discourages ever-married women from working in the labour market. They also find that the percentage of tertiary educated ever-married women working in a specific occupation declines as the prevalence of overwork among men (working 50+ weekly hours) increases. Among women who continue to work in greedy jobs, some find ways to relieve time pressures. Cortés and Pan (2019) show that low-skilled immigration can help to reduce the gender wage gap among highly skilled women; low-skilled migrants, employed to assist with household production responsibil- ities, enable mothers to work long hours.

III Data The data used are from the Household

Income and Labour Dynamics Survey (HILDA), an annual panel data set of Austra- lian households conducted from the 2000–01 to 2020–21 financial years (e.g., 1 July 2000– 30 June 2001). HILDA collects data about economic and personal well-being, labour market dynamics and family life through interviews with people aged 15 and older in a household. In the latest year of the survey, data were collected for about 17,000 individuals. Analyses are based on the unbalanced panel

in order to draw on the largest sample size possible. The HILDA manual (Summerfield et al., 2020) finds that attritors are more likely to be ‘living in Sydney and Melbourne; aged 15–24; single or living in a de facto marriage; born in a non-English speaking country; Aboriginal or Torres Strait Islander; living in

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

466 ECONOMIC RECORD DECEMBER

a flat, unit or apartment; of relatively low levels of education; unemployed; or working in blue collar or low-skilled occupations’. By focusing on 25–65-year-old, university-educated indi- viduals (further described below), the impact of attrition is reduced. Earnings data are available at the weekly and annual level (for the financial year). Weekly earnings data are separated by the ‘main job’ and ‘other jobs’. By contrast, annual earnings data are only avail- able for ‘all jobs’. Hours worked are only available at the weekly level and are separated by ‘main job’ and ‘other jobs’. Weekly earnings and the corresponding weekly hours data are used for most analyses since consis- tency between the two measures allows for an approximation of hourly wages. Wages are deflated by the national CPI and expressed in 2020 dollars. Consistent with the approach applied in

Goldin (2014), data are restricted to the 25–65- year-old employee population. Different sam- ples of workers are constructed: all employees, full-time employees, university-educated employees, and full-time university-educated employees. Full-time employees refer to employees who worked at least 35 hours per week in their main job. Thirty-five hours per week is the threshold used to define full-time employment by the Australian Bureau of Statistics (ABS). Employees without earnings data and employees who report zero earnings have been excluded.3

Data on occupations are available at the four-digit Australian and New Zealand Standard Classification of Occupations (ANZSCO). Summary statistics are pro- vided using the full sample of these workers across all occupations. By contrast, regression analyses, which focus on occu- pational differences, exclude occupations with fewer than 25 men and 25 women; this exclusion follows the approach used by Goldin (2014) and helps ensure compara- bility with her results. The sample size restriction imposed (25 men and 25 women) is also relaxed later in the paper and the modification reinforces the main results. A summary of the sample selection for the full-time employee sample is provided in Table 1. Data are also drawn from the Occupa-

tional Information Network (O*NET), the successor to the United States Department of Labor’s Dictionary of Occupational Titles. O*NET collects data, at the occupa- tional level (1,016 occupations in total) in the United States, about: knowledge, skills and abilities; education, experience and training; interests, work values, work styles; tasks; technology, skills and tools; work activities; and work context. Importantly, O*NET provides detailed information on job characteristics aggregated at the occupa- tional, as opposed to worker, level. Hence, heterogeneity across workers within an

TABLE 1 Sample Size Selection for All Full-Time

Employees

Number of non-missing values

Full sample 410,658 Dropping <25 and >64 200,340 Dropping <35 usual hours worked

102,936

Dropping missing occupations 102,867 Dropping missing or zero earnings

92,890

Dropping individuals without a university degree

31,885

Source: Author’s calculations using HILDA.

3 Employees who report a wage equal to zero are excluded since their occupational composition (disproportionately farmers and tradespersons) is more consistent with the use of trusts and tax planning. Trusts have historically been used by farmers for asset protection and as means to distribute income to beneficiaries. Beneficiaries, often family members who will later inherit the farm, work as employees with minimal or zero wages in anticipation of future ownership (see Evans, 2019, for a discussion of the use of trusts in the Australian context). Zero wages are also consistent with tradespersons, who can create their own companies and combine the use of a company structure with trusts to distribute and retain income in ways that differ from employees in a standard employment relationship (Sainsbury & Breunig, 2020). In summary, workers who do not report a wage are excluded because their wage most likely does not reflect their labour productivity.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 467

occupation cannot be observed using this survey.4

IV Descriptive Statistics Table 2 shows summary statistics from the

HILDA sample of full-time employees, pooled across all years, by sex and by whether employees received a bachelor’s degree. The table shows that the different subsets of employees are quite similar in terms of mean age. Men tend to work slightly more hours than women. Not surprisingly, while the earnings levels are higher for university graduates, men tend to earn more than women across both education groups. In terms of education, significantly more women are university grad- uates than men. As a result, the higher proportion of women with a university educa- tion results in a smaller gender earnings gap for all full-time employees than for full-time university graduates. Education is included in the subsequent regression analyses to control for this important difference. Among univer- sity graduates, a similar share of men and women hold a postgraduate degree. Figure 1 shows changes in the weekly

gender earnings gap over time by different samples of employees and data sources. The figure shows that while the magnitude of the gap differs by employee subgroup and data source, the declining trend remains the same. The gap for all employees is larger than for full-time employees since the measure used is weekly earnings (for HILDA and ABS data), which captures differences in hours worked as well as wage rates. Consistent with Table 2, the gap for all full-time employees is smaller than that for full-time university graduates because of the

aforementioned composition effect. Data from the Workplace Gender Equality Agency (WGEA) include all employees, but adjust for working hours. As a result, WGEA estimates lie in between the gender earnings gaps estimates for full-time and all employees using HILDA and ABS data5

Figures 2 and 3 show how occupational segregation has evolved alongside the decline in the gender earnings gap. The Duncan index captures the magnitude of occupational segregation in the labour mar- ket by comparing employment shares of men and women by occupation. It ranges from 0 to 100 and equals 100 when men and women are completely segregated (no men or women work in the same occupation) and 0 when equal shares of both sexes work in all occupations (complete integration) Indices were calculated using two-, three-

and four-digit occupational codes. Figure 2 considers different samples of employees using the two-digit occupational codes. Figure 3 calculates the index for full-time university graduate employees at the two-, three- and four-digit levels. The figures show three trends. First, similar to the gender earnings gap, segregation has declined over time. This trend applies irrespective of the subgroup of employees considered or the occupational code level used. It is also consistent with findings from Borland and Coelli (2016) who show that the Duncan index has declined in Australia since the 1960s and up to 2011. Second, the level of segregation is highest among all employees, followed by full-time employees, all univer- sity graduate employees and full-time uni- versity graduate employees. This trend suggests that both working hours and edu- cation level influence the degree of segrega- tion among professions where individuals work. Finally, the bottom panel shows that

4 Since O*NET is collected in the United States, there is also an underlying assumption that occupations are similar across the two countries and that workers would have responded in similar ways to the questions posed. In March 2021, the Australian Skills Commission mapped the O*NET and ANZSCO occupations in order to use O*NET data to create a skills classification structure. The application of O*NET data to the Australian context by the Australian government suggests some congruence between occupations in the two countries. The use of O*NET also allows for a direct comparison with the results from Goldin (2014).

5 Variation in survey coverage could also underlie the difference in gender earnings gaps produced using ABS and HILDA data, relative to WGEA data. Specifically, WGEA data, derived from its employer census, only include data from non-public sector employers with 100+ employees. By contrast, ABS and HILDA include employees in the public sector and those who work in firms with fewer than 100 employees.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

468 ECONOMIC RECORD DECEMBER

T A B L E 2

S u m m a ry

S ta ti st ic s in

H IL D A , P o o le d A cr o ss

A ll T im

e P er io d s

F u ll -t im

e em

p lo y ee s

N o n -u n iv er si ty

g ra d u at e

U n iv er si ty

g ra d u at e

A ll

M al e

F em

al e

T o ta l

M al e

F em

al e

T o ta l

M al e

F em

al e

T o ta l

A g e (m

ea n /S D )

4 1 .7 5

4 2 .7 6

4 2 .0 8

4 1 .3 0

4 0 .1 3

4 0 .7 6

4 1 .6 2

4 1 .6 4

4 1 .6 3

(1 0 .5 7 )

(1 0 .8 4 )

(1 0 .6 7 )

(1 0 .0 7 )

(1 0 .6 2 )

(1 0 .3 4 )

(1 0 .4 3 )

(1 0 .8 2 )

(1 0 .5 8 )

H o u rs

u su al ly

w o rk ed

(m ea n /S D )

4 5 .8 1

4 1 .2 8

4 4 .3 5

4 5 .5 5

4 3 .3 2

4 4 .5 3

4 5 .7 3

4 2 .1 5

4 4 .4 1

(9 .8 5 )

(6 .9 6 )

(9 .2 6 )

(8 .3 4 )

(7 .5 7 )

(8 .0 8 )

(9 .4 3 )

(7 .3 0 )

(8 .8 8 )

W ee k ly

ea rn in g s m ai n jo b , 2 0 2 0

$ (m

ea n /S D )

1 5 7 3 .3 8

1 2 0 1 .1 8

1 4 5 3 .6 6

2 2 9 1 .7 7

1 7 4 5 .2 3

2 0 4 0 .2 8

1 7 8 4 .4 1

1 4 3 3 .9 4

1 6 5 5 .0 2

(8 3 0 .8 9 )

(5 4 3 .2 2 )

(7 7 0 .3 6 )

(1 4 4 8 .4 4 )

(8 0 6 .4 2 )

(1 2 2 7 .1 9 )

(1 1 0 0 .4 1 )

(7 2 0 .7 7 )

(9 9 2 .0 9 )

H ig h es t ed

u ca ti o n le v el

N o u n iv er si ty

d eg

re e (%

) 1 0 0 .0 0

1 0 0 .0 0

1 0 0 .0 0

0 .0 0

0 .0 0

0 .0 0

7 0 .6 2

5 7 .2 2

6 5 .6 7

B ac h el o r’ s d eg

re e (%

) 0 .0 0

0 .0 0

0 .0 0

5 6 .7 8

5 5 .7 5

5 6 .3 1

1 6 .6 8

2 3 .8 5

1 9 .3 3

P o st g ra d u at e d eg

re e (%

) 0 .0 0

0 .0 0

0 .0 0

4 3 .2 2

4 4 .2 5

4 3 .6 9

1 2 .7 0

1 8 .9 3

1 5 .0 0

N 4 1 ,3 8 3

1 9 ,6 2 2

6 1 ,0 0 5

1 7 ,2 1 3

1 4 ,6 7 2

3 1 ,8 8 5

5 8 ,5 9 6

3 4 ,2 9 4

9 2 ,8 9 0

S o u rc e:

A u th o r’ s ca lc u la ti o n s u si n g H IL

D A .

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 469

the level of segregation increases with each additional occupational digit. This suggests that while men and women work in similar professions broadly defined (i.e., at the two-digit level), segregation increases as occupations are more narrowly defined. Collectively, these findings imply that while full-time university graduate employees tend to work in professions that are more integrated, these professions remain segre- gated when analysed at a very detailed four-digit occupational level. In other words, even among women who ‘look’ the most like men, occupational segregation persists, particularly when very narrow definitions of occupation are considered (i.e., at the four-digit level). Bearing this in mind, is the gender

earnings gap bigger due to differences in pay between men and women in the same occupation or is it largely attributable to

differences in pay across different occupa- tions? To calculate the importance of gender earnings gaps within occupations, all female employees were assigned the average earn- ings of all male employees in their occupa- tion in 2020. Then the average gender earnings gap for 2020 was recalculated. Figure 1 shows that in 2020, the gender earnings gap for all employees in HILDA amounted to about 28 per cent. When women are assigned the same earnings as men in their occupation, the mean gender earnings gap declines by 18 percentage points, to 10 per cent (the ‘within’ occupa- tion effect). To calculate the importance of differences in pay across occupations, women retain their earnings, but those earnings are weighted by men’s employment structure (the ‘between’ occupations effect). When the between occupations effect is calculated, the gender earnings gap declines

FIGURE 1 The Gender Earnings Gap (Weekly Earnings) by Employee Type and Data Source, 2001–02 to 2020–21

(Colour is Available Online)

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

470 ECONOMIC RECORD DECEMBER

to about 17 per cent.6 These findings suggest that, similar to the United States (Goldin, 2014, 2021), occupation plays an important role in explaining the gender earnings gap, but differ- ences within occupations are especially impor- tant. This finding is also consistent with other studies for Australia (Rimmer, 1991; Kidd, 1993; Miller & Lee, 2004; Kee, 2006; Baron & Cobb-Clark, 2008; Cobb-Clark & Tan, 2011), which apply more sophisticated methods to disentangle the between and within effects, and show that the within occupations effect is particularly important. The impor- tance of occupation is also confirmed by econometric analyses in the next section.

V Empirical Analyses In order to quantify the importance of

occupation to the gender earnings gap, Goldin (2014) undertakes four analyses which allow her to draw the following conclusions:

1. Occupation is an important explanatory variable that contributes to the gender earnings gap.

2. The gender earnings gap varies by occupation.

3. The gender earnings gap tends to be higher in occupations where ‘temporal flexibility’ is costly.

4. High costs, associated with ‘temporal flexibility’, give rise to ‘greedy’ jobs with an earnings-to-hours elasticity that exceeds 1. The gender earnings gap is highest in occupations where greedy jobs proliferate.

FIGURE 2 The Duncan Index by Type of Employee at the Two-Digit Occupational Level, 2001–02 to 2020–21

(Colour is Available Online)

6 The importance of the within effect, relative to the between effect, also emerges if female wages and the female employment structure are applied to men (as opposed to vice versa).

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 471

This section replicates each of these analyses using HILDA.

(i) Analysis 1: Occupation is an Important Explanatory Variable that Contributes to the Gender Earnings Gap In order to determine the importance of

occupation in explaining the gender earnings gap, five ordinary least squares regressions are estimated using four samples of employees: all employees, full-time employees, university graduate employees, and full-time university graduate employees. The basic specification is given by

ln Yð Þit = β0 þ β1ageit þ β2age 2 it þ β3sexit

þ ∑ T

t= 2

β4tyeart þ εit:

(1)

The hours worked specification is written as

ln Yð Þit=β0þβ1ageitþβ2age 2 it

þβ3sexitþ ∑ T

t=2

β4tyeart

þβ5ln weeklyhoursworkedð Þitþεit:

(2)

The education specification is

ln Yð Þit = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit þ εit: (3)

FIGURE 3 The Duncan Index for Full-Time University Graduates, by Occupational Code Detail, 2001–02 to 2020–

21 (Colour is Available Online)

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

472 ECONOMIC RECORD DECEMBER

The occupation specification is

ln Yð Þitk = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit

þ ∑ K

k= 2

β7koccupationitk

þεitk: (4)

Finally, experience specification is written as

ln Yð Þitk = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit

þ ∑ K

k= 2

β7koccupationitk

þβ8occtenureit þβ9emptenureit þ εitk: (5)

The dependent variable, ln Yð Þitk, refers to the natural log of weekly earnings for individual, i, at time, t, in occupation, k. Sex is a dummy variable that equals 0 for men and 1 for women. University is a dummy variable that equals 1 if an individual completed university and 0 otherwise. In the specifica- tions estimated across university graduates, university is excluded and replaced with aboveBA which is a dummy variable equal to 1 if an individual completed a postgraduate degree and 0 if she only has a bachelor’s degree. Yeart and occupationitk are dummy variables for each financial year, t, and occupation, k, at the four-digit level. The variables occtenure and emptenure refer to the number of years of experience an individual has working in their current occupation and with their current employer, respectively.7

A summary of the results is presented in Table 3 (the complete specifications are provided in Tables S1–S4). The inclusion of all variables results in a gender earnings gap that is about equal across all samples (between 11 and 13 per cent). Irrespective of the sample of employees, the gender earn- ings gap is smallest for the specification that includes occupation (Eqn 4). In addition, the adjusted R-squared increases substantially when occupational controls are added. For example, focusing on all full-time employees, the adjusted R-squared increases by 6 percentage points between the basic and hours worked specifications, an additional 10 percentage points between the hours worked and education specifications, and an additional 19 percentage points between the education and occupation specifications. Both the reduction in the coefficient on female and the substantial increase in the adjusted R-squared confirm the importance that occupation plays in explaining the gender earnings gap. The fifth specification exploits the rich-

ness of the HILDA data set by including information on employer and occupational tenure (variables not available in Goldin, 2014) and shows that they do not explain much (or any) additional variation in male and female earnings, depending on the subgroup of employees. To better understand the statistically small contribution of employer and occupational tenure, Equations (1–4) were estimated separately by sex. In addition, occupational and employer tenure were added to all of the regressions estimated from Equations (1–4). The results (available in Table S8) show that across all specifications (basic, hours worked, education and occupation), occupa- tional and employer tenure remain statisti- cally significant, but economically insignificant (in terms of magnitude) for women and women. The return to general labour market experience (proxied by age and age squared), however, remains both statistically and economically significant. The returns to general labour market expe- rience are also bigger for men relative to women and are smallest for both sexes once occupation is included as an explanatory variable. Focusing on full-time university graduates in particular, the return on

7 These specifications were also estimated for the 26–59 year old population – a population segment more commonly used in Australian labour studies – and the results do not substan- tively change.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 473

experience (proxied by the coefficient on age) declines from 0.09 to 0.07 with the addition of occupation, relative to 0.05–0.04 among women. The greater decrease among full-time university men with the addition of occupational dummy variables suggests that men may select into occupations that have higher returns to experience than women. Another possibility, which cannot be con- firmed with the data available, is that men select specific jobs (within an occupation) that value experience more, relative to the jobs selected by women in the same occupation.

(ii) Analysis 2: The Gender Earnings Gap Varies by Occupation While the preceding analysis confirmed

the importance of occupation in explaining the gender earnings gap, this second analysis identifies if and by what magnitude the wages of men and women differ within a specific occupation. For example, this sec- ond analysis shows how much more (or less)

female pharmacists earn relative to their male pharmacist counterparts. To estimate gender earnings gaps within occupations, sex and occupation are interacted in the following specification:

ln Yð Þitk = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit

þ ∑ K

k= 2

β7koccupationitk

þ ∑ K

k= 2

β8koccupationitk

$sexit þ εitk: (6)

The results from this regression are pre- sented in Table S5. The sum of β3 and β8, which captures the gender earnings gap within a specific occupation, is graphed for

TABLE 3 Gender Differences in Earnings and the Role of Occupation

Sex coefficient

Sex std. error

Adjusted R-squared

Number of observations

Full-time all employees, basic %0.18*** (0.00) 0.07 92,890 Full-time all employees, hours worked %0.13*** (0.00) 0.13 92,890 Full-time all employees, education %0.18*** (0.00) 0.23 92,890 Full-time all employees, occupation %0.13*** (0.00) 0.42 92,890 Full-time all employees, experience %0.13*** (0.00) 0.42 92,890 All employees, basic %0.45*** (0.00) 0.13 125,816 All employees, hours worked %0.13*** (0.00) 0.50 125,816 All employees, education %0.17*** (0.00) 0.55 125,816 All employees, occupation %0.12*** (0.00) 0.65 125,816 All employees, experience %0.12*** (0.00) 0.65 125,816 Full-time uni. employees, basic %0.20*** (0.01) 0.12 31,885 Full-time uni. employees, hours worked %0.17*** (0.01) 0.18 31,885 Full-time uni. employees, education %0.17*** (0.01) 0.18 31,885 Full-time uni. employees, occupation %0.11*** (0.01) 0.38 31,885 Full-time uni. employees, experience %0.11*** (0.01) 0.39 31,885 All uni. employees, basic %0.40*** (0.01) 0.12 41,980 All uni. employees, hours worked %0.16*** (0.00) 0.50 41,980 All uni. employees, education %0.17*** (0.00) 0.50 41,980 All uni. employees, occupation %0.12*** (0.00) 0.62 41,980 All uni. employees, experience %0.11*** (0.00) 0.63 41,980

Note: ***P < 0.01, **P < 0.05, *P < 0.1. Source: Author’s calculations using HILDA.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

474 ECONOMIC RECORD DECEMBER

each occupation against the natural log of the mean weekly earnings of men by occupation. Occupations were also grouped into five broader categories: business, sci- ence, technology (tech), health and other. Occupations are grouped according to the classification used by Goldin (2014). In some cases, there was a one-to-one occupa- tional match between the Australian ANZSCO occupational categories, and the occupational list provided in Goldin (2014). In the absence of a one-to-one match, a category was selected based on the job description. The ‘broad’ categorisation cap- tures individuals who only reported an occupation which could be identified at the one- or two-digit level (this grouping was not provided in Goldin).8 The results are presented in Figure 4 for full-time university graduates and show that a gender earnings gap exists across many occupations and tends to increase with male earnings. It also shows that the largest gender earnings gaps tend to appear primarily among business occupations, consistent with the findings in the United States presented by Goldin (2014). This is represented by the concentration of red squares below %0:2 on the y-axis.

(iii) Analysis 3: The Gender Earnings Gap Tends to be Higher in Occupations where ‘Temporal Flexibility’ is Costly In order to better understand the reasons

why business professions tend to have a larger gender earnings gap, Goldin (2014) exploits the O*NET database. She focuses on seven specific questions posed in the O*NET survey that characterise the degree of temporal flexibility in a given occupation.9 The first six questions aim to capture time demands on the job, whereas the seventh captures the degree of competi- tion. The questions are as follows:

1. Time pressure: How often does your current job require you to meet strict deadlines? A low value refers to workers who never or rarely have strict deadlines. Lower pressure means the worker does not have to be around at particular times.

2. Contact with others: How much contact with others is required to perform your current job (face-to-face, by telephone, or otherwise)? A low value means a worker has minimal to no contact with others. Less contact means greater flexibility.

3. Establishing and maintaining interper- sonal relationships: How important is establishing and maintaining interper- sonal relationships to the performance of your current job? A low value indi- cates interpersonal relationships are unimportant for job performance and an employee does not require much contact with other workers or clients.

4. Structured versus unstructured work: How much freedom do you have to determine the tasks, priorities or goals of your current job? A low value implies a job with little autonomy and that is very structured. If a job is highly structured, there is a higher chance the worker has close substitutes.

5. Freedom to make decisions: How much freedom do you have to make decisions without supervision? A low value indi- cates a worker has little autonomy to make decisions without supervision. If the worker determines what each client should receive, rather than being given a specific task, greater worker autonomy suggests poorer substitution.

6. Frequency of decision-making: How often do your decisions affect other people, or the image, or reputation, or financial resources of your employer? A low value suggests a worker’s decisions minimally impact others.

7. Level of competition: How competitive is your current job? A low value implies a job is not competitive.10 An in-depth discussion of competition is provided in the text below.

8 This grouping includes individuals who indi- cated they were ‘professionals’ (one-digit occu- pation level), ‘ICT professionals’ (two-digit occupation level), or ‘specialist managers’ (two-digit occupation level). These occupations represent <2 per cent of employment.

9 Questions 6 and 7 were added in Goldin (2021).

10 O*NET does not provide an additional definition of ‘competitive’; it is for the worker responding to the survey to self-assess.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 475

The O*NET survey allows respondents to rank their responses, generally from 1 to 5 or 1 to 7. For example, in response to question 1 on ‘time pressure’, respondents can rank their response as never (i), once a year or more but not every month (ii), once a month or more but not every week (iii), once a week or more but not every day (iv), or every day (v). O*NET provides an average of the responses of respondents to the questions, by occupation, in the database (i.e., they do not provide worker-specific responses). The responses to each of the questions were normalised to have a mean of 0 and a standard deviation of 1. In order to understand how these charac-

teristics relate to the gender earnings gap by occupation, the O*NET occupations were mapped to the ANZSCO occupations (for occupations with a minimum of 25 men and 25 women). A detailed description of the

O*NET and the O*NET–ANZSCO mapping is given in Appendix S1. Once the occupa- tional codes were mapped, a mean ANZSCO occupational O*NET score was calculated by averaging the values associated with the seven questions for each occupation, gener- ating an ANZSCO occupation-specific O*NET score (e.g., pharmacist, registered nurse, etc.). Mean ANZSCO occupational O*NET scores were then calculated across four occupational groupings (technology and science, business, health, and law), defined by Goldin (2014); these were calculated by averaging the ANZSCO occupational O*NET score across occupations included within each grouping and weighting the individual occupations by employment. For example, if pharmacists and registered nurses were the only two health occupations, the ‘health grouping’ O*NET score would be the mean of the O*NET scores of registered

FIGURE 4 Gender Earnings Gaps at the Four-Digit Occupational Level, Full-Time University Graduates (Colour is

Available Online)

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

476 ECONOMIC RECORD DECEMBER

nurses and pharmacists, weighted by the employment in each occupation. Results are presented in Table 4.11

Consistent with Goldin (2014), technol- ogy and science occupations have the lowest score on all questions, except the level of competition. This implies that individuals who work in these sectors tend to have few strict deadlines, little contact with others, and less need to establish and maintain relationships. They also undertake struc- tured (substitutable) work, have less free- dom to make their own decisions without supervision, and the decisions they do make minimally impact others. The combination of these characteristics suggests that employees in these occupations have few time demands and limited personal interac- tion. As illustrated by the green and yellow triangles in Figure 5, these occupations also have among the smallest gender earnings gaps. By contrast, while legal occupations have greater autonomy to make decisions, those decisions impact others more fre- quently and must be made in competitive environments under substantial time pres- sure. Similarly, business professions have fewer time pressures, but high levels of competition and emphasis placed on

establishing interpersonal relationships and contact with others. Since legal and business professions impose time demands and require substantial interaction with others, their occupational pay gaps are also larger. Health professions do not align with this

framework. They have among the highest values for the six questions associated with time demands on the job, but gender earnings gaps that fall between the extremes of those in business and science and technology occupations. To account for this theoretical inconsistency (i.e., high average time demands and low average gender earnings gaps), Goldin extends her 2014 framework in her 2021 research (Goldin, 2021) and includes the seventh question noted above (about competition) from the O*NET questionnaire. Competition represents another type of time demand. In health professions, time demands are high due to the time-sensitive nature of the work (i.e., there is little health professionals can do to change when someone has a heart attack or when a baby is born). However, in business professions, time demands are high due to the nature of the work (i.e., deadlines imposed by clients) and also due to the amount of time required to secure additional business. Business professionals need to allocate time to prepare tenders (i.e., bids for contracts they may not win), as well as time to build and maintain professional relationships to secure additional business. The amount of time individuals spend securing additional business is proxied by

TABLE 4 Mean O*NET Score (Normalised) by Occupational Grouping

Business Health Law Sci. and tech.

Establishing and maintaining interpersonal relationships 0.89 1.05 0.76 %0.07 Contact with others 0.63 0.89 0.52 %0.46 Structured versus unstructured work 0.71 0.80 0.19 %0.21 Freedom to make decisions 0.37 0.83 0.79 %0.43 Time pressure 0.24 0.59 1.26 0.04 Frequency of decision-making 0.43 1.06 1.30 %1.04 Level of competition 0.62 %0.38 0.66 0.45 Number of occupations 33 7 2 12

Note: The category ‘other’ occupations are excluded, consistent with the approach used by Goldin (2014). Source: Author’s calculations using HILDA and O*NET.

11 Consistent with the approach used by Goldin (2014), the ‘other’ category has been excluded from this table. As a result, the total number of occupations does not sum to the 69 occupations for which O*NET data are available. See Section S1 in Appendix S1 for further details.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 477

levels of competition in an occupation. The fiercer the competition, the more time is invested in securing the business and main- taining the relationships. The O*NET data reveal that while there are high levels of competition in business occupations, there is little competition reported in health pro- fessions. Reduced competition in the labour market for health professionals is largely associated with barriers to entry (i.e., occu- pational regulations and licensing, inability to match increased demand for health pro- fessionals quickly due to the extensive time requirements for training, etc.). The differences summarised in Table 4 are

presented by occupation in Figure 5. The average O*NET scores by occupation are based on the seven questions presented above. Consistent with Goldin (2014), a statistically significant negative relationship emerges between the size of the gender earnings gap

and the average O*NET score. This relation- ship suggests that the greater time demands, the levels of personal interaction, and/or degree of competition in an occupation, the larger the gender earnings gap. By contrast, the more substitutable workers and/or the services and products they provide, the smaller the gender earnings gap.

(iv) Analysis 4: High Costs, Associated with ‘Temporal Flexibility’, Give Rise to ‘Greedy’ Jobs with an Earnings-to-Hours Elasticity that Exceeds 1. The Gender Earnings Gap is Highest in Occupations where Greedy Jobs Proliferate To assess the extent to which the cost of

temporal flexibility is reflected by an earnings-to-hours elasticity >1, an interac- tion term between weekly hours worked and occupation is included in the following specification:

FIGURE 5 Gender Earnings Gaps at the Four-Digit Occupational Level and the Average O*NET Score

(Goldin, 2021), Full-Time University Graduates (Colour is Available Online)

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

478 ECONOMIC RECORD DECEMBER

ln Yð Þitk = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit

þ ∑ K

k= 2

β7koccupationitk

þ ∑ K

k= 2

β8koccupationitk

$sexit þ ∑ K

k= 2

β9koccupationitk

$ln weeklyhoursworkedð Þit þεitk: (7)

The results from this regression are pre- sented in Table S5. The elasticity of weekly

earnings to weekly hours worked, equal to the sum of β5 and β9, is presented in Figure 6. Consistent with Goldin’s results, a statistically significant and negative relationship emerges between the size of the occupation-specific gender earnings gap and the earnings-to-hours elasticity. The negative relationship suggests that the greater the cost of temporal flexibility, represented by the earnings-to-hours elasticity, the greater the gender earnings gap. The results presented in Figure 6 are also consis- tent with differences in hours worked by men and women within occupations. Figure 7 shows that the gender earnings gap is largest in occupations where the difference in mean hours worked between men and women is largest. In other words, both figures suggest that full-time university graduate women disproportionately select flexible jobs in greedy occupations or work fewer hours than men in greedy jobs.

FIGURE 6 Gender Earnings Gaps at the Four-Digit Occupational Level and the Elasticity of Weekly Earnings to

Hours Worked, Full-Time University Graduates (Colour is Available Online)

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 479

The results presented thus far are consis- tent with the findings in Goldin (2014) and remain valid when both countries’ data are limited to the same occupations (see Fig. S1 in Appendix S1). At the individual occupa- tion level, however, differences emerge. Figure 8 limits gender earnings gaps to those occupations for which estimates are available in both countries and that are statistically significant in Australia. Confi- dence intervals are also provided for the Australian estimates (they are not available for the USA). For occupations where data are available in both countries, occupational gender earnings gaps are slightly higher in the United States in most occupations, compared to their Australian occupational counterparts. This is consistent with wage compression effects observed in countries where collective bargaining coverage is large. Cross-country differences in

occupational gender earnings gaps could, however, arise for several reasons. Some of these, including differences in labour market institutions, are further discussed below.

(v) Differences in Occupational Definitions in the United States and Australia Since occupational classifications differ

between the USA and Australia, this could explain some of the variation in occupational cross-country gender earnings gaps. For example, the classification used by Goldin (2014) has three occupational codes for the legal profession: lawyers, and judges, magistrates, and other judicial workers; miscellaneous legal support workers; and paralegals and legal assistants. There are also three, slightly different, occupational codes that encompass this group in Austra- lia: barristers, solicitors, and judicial and

FIGURE 7 Gender Earnings Gaps and the Difference in Male and Female Weekly Hours Worked, Full-Time

University Graduates (Colour is Available Online)

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

480 ECONOMIC RECORD DECEMBER

other legal professionals. In this instance, cross-country differences in occupational classifications prove problematic since the barrister profession – an occupation that specialises in defending clients in court – is extremely male-dominated, which precludes calculation of a gender earnings gap and an earnings-to-hours elasticity using the sample size restrictions imposed by Goldin. As a result, a direct comparison of the earnings for lawyers is impossible across the two countries unless the occupational codes for the three groups are aggregated in the Australian classification. To assess the extent to which

cross-country differences in occupational classifications contribute to occupational pay gaps across both countries, sample size restrictions are relaxed and occupations with any number of men and women are included in the estimates. This broadens the sample to include occupations that may be highly sex-segregated or have small sample sizes in Australia, but not the United States. Figure 9 shows that including these

occupations does not alter the negative relationship observed between the occupa- tional gender earnings gap and the earnings- to-hours elasticity. Disaggregation of the unrestricted sample by occupational group- ing is presented in Figure S2.12 When the earnings-to-hours elasticity is graphed against the share of women in an occupation, a statistically significant and negative rela- tionship also emerges. Combined, these results imply that in addition to greater pay inequity in occupations with a high earnings- to-hours elasticity, these occupations tend to be male-dominated. The gender earnings gaps for legal occu-

pations and physicians are also recalculated below to improve comparison across the two countries. The total for each occupation is then weighted by the number of individuals employed in each sub-occupational grouping for each country. Table 5 shows that

FIGURE 8 Occupational Gender Earnings Gaps for Full-Time University Graduates in Australia and the USA

(Colour is Available Online)

12 The regression results produced using the unrestricted sample in Equations (6) and (7) are also available in Table S7.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 481

aggregating the Australian occupational codes in the legal profession, to better approximate the occupations included in the American classification, does not elim- inate gender earnings gaps between Austra- lia and the USA. The gender earnings gap in the American legal profession remains more than double the gap in Australia. By contrast, aggregating the Australian health professions that are encompassed by the American ‘physicians and surgeons’ classi- fication substantially reduces the gap observed between the two countries (Table 6). A gap of 10 percentage points in earnings does, however, remain across the two countries. These two sectors suggest that differences in occupational classifications can explain some cross-country heterogene- ity in occupational gender earnings gaps, but not all.

(vi) Differences in Institutional Settings Another source of cross-country variation

could emerge from differences in labour market institutions, namely the prevalence of collective bargaining. In comparison to the United States, the Australian labour market is more highly regulated. Figure 10 focuses on the pay-setting mechanisms used by full-time university graduate employees in 2020.13 The figure illustrates that most employees in this subgroup have their pay set by collective or individual agreement. There are, however, differences by sex; men more frequently have their pay set through indi- vidual agreement, while women rely more

FIGURE 9 Gender Earnings Gaps at the Four-Digit Occupational Level and the Elasticity of Weekly Earnings to Hours Worked for Full-Time University Graduates, Sample Size Restricted and Unrestricted (Colour is

Available Online)

13 While 2020 is presented, trends in pay- setting mechanisms are relatively constant since the question was initially posed in HILDA in 2008.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

482 ECONOMIC RECORD DECEMBER

heavily on collective agreement. Substantial heterogeneity in wage-setting mechanism also emerges among full-time university graduates within occupations. Figure S3 illustrates how very few occupations have a dominant form of wage-setting. This hetero- geneity at least partially reflects that collec- tive bargaining occurs at the enterprise level in Australia. As a result, individuals can

work in the same occupation, sector, and geographic area, for example, but be subject to a different form of wage-setting. To better understand the impact of wage-

setting institutions on the gender earnings gap, five wage-setting dummy variables were added to Equation (5). The wage-setting dummy variables, denoted by wagesettingitka, equal 0 or 1, for wage-setting arrangement a applica- ble to individual i at time t in occupation k. Wages are set by: collective agreement; an individual agreement; a combination of a collective agreement and individual agree- ment; by a minimum wage (award); or through other mechanisms. In order to retain the same sample size, an additional unknown dummy variable is also included where individuals do not indicate the type of wage-setting arrange- ment in place. A summary of the results from these specifications (directly comparable to full-time university graduates in Table 3) on the gender earnings gap is presented in Table 7. Table 7 suggests that controlling for wage- setting mechanism does not reduce the gender earnings gap, but its inclusion does slightly increase the explanatory power (the adjusted R-squared) of the model. One explanation, however, could be that wage-setting mecha- nisms are more important in some occupations than others. To better understand the impact of wage-

setting institutions on specific occupational gender earnings gaps, the five (aforemen- tioned) wage-setting dummy variables are included in the following equation, captured by β9 (a vector of coefficients, denoted by k):

ln Yð Þitka = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit

þ ∑ K

k= 2

β7koccupationitk

þ ∑ K

k= 2

β8koccupationitk $ sexit

þ ∑ A

a= 2

β9awagesettingitka

þεitka: (8)

TABLE 6 Gender Differences in Earnings Among

Physicians and Surgeons

Country of Classification

Australia USA

Anaesthetists %0.38 Generalist medical practitioners

%0.09

Other medical practitioners %0.36 Specialist physicians %0.18 Surgeons %0.64 Physicians and surgeons %0.34 Total %0.24 %0.34

Source: Author’s calculations using HILDA and estimates from Goldin (2014).

TABLE 5 Gender Differences in Earnings in the Legal

Profession

Country of Classification

Australia USA

Barristers %0.47 Judicial and other legal professionals

0.01

Solicitors %0.10 Lawyers, and judges, magistrates, and other judicial workers

%0.19

Miscellaneous legal support workers

%0.24

Paralegals and legal assistants

%0.11

Total %0.10 %0.19

Source: Author’s calculations using HILDA and estimates from Goldin (2014).

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 483

Next, the following equation adds an additional interaction term between sex and the wage-setting dummy variables. The

intent of this second equation is to permit comparison between a male and female pharmacist who both have their pay set

FIGURE 10 Pay-Setting Mechanism Among Full-Time University Graduate Employees in 2020, by Sex (Colour is

Available Online)

TABLE 7 Gender Differences in Earnings and the Role of Wage-Setting Mechanisms

Sex coefficient

Sex std. error

Adjusted R-squared

Number of observations

Full-time uni. employees, basic %0.20*** (0.01) 0.12 31,885 Full-time uni. employees, hours worked %0.17*** (0.01) 0.18 31,885 Full-time uni. employees, education %0.17*** (0.01) 0.18 31,885 Full-time uni. employees, occupation %0.11*** (0.01) 0.38 31,885 Full-time uni. employees, experience %0.11*** (0.01) 0.39 31,885 Full-time uni. employees, wage-setting %0.11*** (0.00) 0.40 31,885

Note: ***P < 0.01, **P < 0.05, *P < 0.1. Source: Author’s calculations using HILDA.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

484 ECONOMIC RECORD DECEMBER

through an individual agreement; this effect is captured through the sum of β3, β8 and β10.

ln Yð Þitka = β0 þ β1ageit þ β2age 2 it

þβ3sexit þ ∑ T

t= 2

β4tyeart

þβ5ln weeklyhoursworkedð Þit þβ6universityit

þ ∑ K

k= 2

β7koccupationitk

þ ∑ K

k= 2

β8koccupationitk $ sexit

þ ∑ A

a= 2

β9awagesettingitka

þ ∑ A

a= 2

β10awagesettingitka

$sexit þ εitka: (9)

The results from Equations (8) and (9) are illustrated in Figure 11 and also provided in Table S6.14 Occupational wage gaps, esti- mated from Equation (6) (the sum of β3 and β8), are plotted on the y-axis against those that account for wage-setting mechanisms (specified in Eqn 8). The appearance of green squares on both sides of the 45-degree line implies that controlling for wage-setting has a variable effect on occupational gender pay gaps; in some occupations the gender pay gap increases and in others it declines. By contrast, when the interaction between sex and wage-setting mechanism is included in the specification, very different results emerge. The blue diamonds (the partial derivative of Eqn 9 with respect to sex) all situate above the 45-degree line, indicating that pay disparities between men and women within occupations are smaller when both sexes have their pay set through collective agreement. Conversely, all of the red tri- angles appear below the 45-degree line, indicating that pay gaps are larger between men and women in the same occupation when both sexes have their pay set by

individual agreement. The particular mechanism(s) underlying the reduction in pay gaps that arises when collective agree- ments are in place (i.e., pay transparency, collective representation, etc.) remain(s) outside the scope of this research. Since collective bargaining coverage is

limited in the United States, a more appro- priate cross-country occupational compari- son of gender pay gaps is those estimated for individuals with an individual agreement. The results for the legal and ‘physicians and surgeons’ occupations are displayed in Tables 8 and 9. The pay gaps observed in these professions, among individuals who have their pay set by individual agreement, are higher than those estimated in Equa- tions (6) and (8). As a result, the weighted average pay gaps for these two occupations in Australia, relative to the United States, converge slightly, but still remain higher in the United States. Finally, when the share of employees’ pay set by an individual agree- ment is juxtaposed against an occupation’s earnings-to-hours elasticity, greater use of individual agreements is not surprisingly associated with larger elasticities (see Fig. S4).

VI Concluding Remarks Goldin (2014, 2021) developed a theoret-

ical framework for greedy jobs; they have nonlinear pay structures and reward long and unpredictable hours. She also shows empir- ically that greedy jobs exist in the United States and that gender earnings gaps are largest in greedy jobs. Yet, part of the reason why greedy jobs can proliferate in the United States is that the industrial relations system enables them to, through its dispro- portionate reliance on individual pay nego- tiation. This research assesses the extent to which greedy jobs proliferate in Australia, a setting where the industrial relations system substantively differs from the United States. It also examines how labour market institu- tions impact occupational gender earnings gaps in greedy and flexible occupations. The results show that, similar to the

United States, the gender earnings gap declines substantially once occupation is included in traditional Mincerian wage equations. In addition, consistent with find- ings from other Australian research

14 While Table 7 uses the full sample of full- time university graduates (to ensure comparabil- ity with Table 3), the results in Figure 11 use the unrestricted sample.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 485

(Rimmer, 1991; Kidd, 1993; Miller & Lee, 2004; Kee, 2006; Baron & Cobb- Clark, 2008; Cobb-Clark & Tan, 2011), the

results suggest that most of the gender occupational gap in earnings is attributable to differences in earnings between men and women within the same occupation, rather than across different occupations. I also investigate the role of general labour market experience, relative to occupational and employer tenure on earnings. The estimates show that while occupational and employer tenure make a positive and statistically significant contribution to earnings, the coefficients are of economically smaller magnitude than those estimated for general labour market experience (proxied by age and age squared). To better understand the lack of explana-

tory power associated with occupational and employer tenure, relative to general labour market experience, regression estimates were calculated separately for men and

FIGURE 11 Occupational Gender Earnings Gaps by Pay-Setting Mechanism, Full-Time University Graduate

Employees (Colour is Available Online)

TABLE 8 Gender Differences in Earnings in the Legal

Profession

Standard (Eqn 6)

Wage- setting (Eqn 8)

Individual agreement (Eqn 9)

Barristers %0.47 %0.39 %0.43 Judicial and other legal professionals

0.01 %0.01 %0.04

Solicitors %0.10 %0.11 %0.14 Total %0.10 %0.11 %0.13

Source: Author’s calculations using HILDA.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

486 ECONOMIC RECORD DECEMBER

women. The results for occupational and employer tenure remain unchanged. By contrast, large gender differences emerge in the returns to general labour market experi- ence, and these differences remain when occupational controls are included. One potential explanation for differing returns to general labour market experience observed across men and women relates to their job selection within occupations. If men disproportionately select more highly compensated, greedy jobs and women select flexible jobs, the returns to general labour market experience would differ with the inclusion of occupational control variables. By replicating Goldin (2014, 2021), I

show that greedy jobs exist in Australia and gender earnings gaps are largest in greedy jobs. In addition, a smaller share of women work in greedy occupations and, among those who do, they work fewer hours per week, on average, than their male counterparts. Fewer hours worked could indicate that females select flexible jobs within greedy occupations. Alternatively, it could imply women work in greedy jobs, but work fewer hours in them. Unfortunately, data from HILDA do not permit these two possibilities to be disentangled; this is an area for future research. Both factors, however, increase gender earnings gaps since, by not working long hours, women

cannot earn as much as their male counter- parts. Occupations with small gender earn- ings gaps also share similarities. These occupations are overrepresented in the tech- nology and science sectors. In these sectors, workers are more substitutable and as a result, there are fewer demands on specific workers’ time and the earnings-to-hours elasticity tends to be less than or equal to 1. Gender earnings gaps in Australian greedy

jobs also differ from those in the United States. In particular, gender earnings gaps are smaller in Australia, consistent with the effects expected by high rates of collective bargain- ing. Gender earnings gaps also vary within occupation, depending on the wage-setting mechanism in place. In occupations where men and women’s pay is set by collective agreement, the gender earnings gap narrows. By contrast, in occupations where individuals negotiate pay individually, the gender earnings gap widens. More research is required to identify the causal mechanism(s) specific to labour market institutions that underpin(s) this result. For example, if reductions in pay gaps are achieved via pay transparency – a feature of a collective agreement – other firm-level policies regarding pay transparency could potentially achieve a similar result. By con- trast, if reductions in pay gaps emerge from improvements in the bargaining position of workers, enabled through collective action, firm-level policies regarding pay transparency would be less effective. Eliminating occupational gender earnings

gaps will require changing the nature of greedy jobs. History shows various paths that can facilitate occupational transition from greedy to ‘not so needy’. Business manage- ment practices and technological change can accelerate the transition (Goldin, 2021). Good old-fashioned ingenuity, on the part of workers in an occupation, can also help achieve it (Goldin, 2021). However, changing individ- uals’ (largely male) preferences for greedy jobs also forms part of the solution. Gender norms impose different expectations on women, which shape their preferences and choices about employment, particularly fol- lowing the arrival of children. Women also bear the financial costs of these different preferences, in the form of less superannuation and reduced work experience, which can be detrimental in the event of relationship

TABLE 9 Gender Differences in Earnings Among

Physicians and Surgeons

Standard (Eqn 6)

Wage- setting (Eqn 8)

Individual agreement (Eqn 9)

Anaesthetists %0.38 %0.41 %0.46 Generalist medical practitioners

%0.09 %0.10 %0.14

Other medical practitioners

%0.36 %0.36 %0.41

Specialist physicians

%0.18 %0.19 %0.24

Surgeons %0.64 %0.64 %0.69 Total %0.24 %0.25 %0.29

Source: Author’s calculations using HILDA.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 487

breakdown. Indeed, while individual families benefit from specialisation, societies bear the economic cost due to an inefficient allocation of a highly valuable resource: women.

Conflict of Interest The author declares no conflicts of

interest.

Supporting Information Additional Supporting Information may

be found in the online version of this article:

Appendix S1. O*Net Description and Additional Analyses and Figures. Figure S1. Gender Earnings Gaps at the 4-

Digit Occupational Level and the Elasticity of Weekly Earnings to Hours Worked in Australia and the USA, Full-Time University Graduates Figure S2. Gender Earnings Gaps at the 4-

Digit Occupational Level and the Elasticity of Weekly Earnings to Hours Worked for Full-Time University Graduates by Occupa- tional Grouping, Unrestricted Sample Size Figure S3. Wage-Setting Mechanism, by

Occupation and Share of Female Employ- ment in an Occupation, Full-Time Univer- sity Graduate Employees in 2020 Figure S4. Share of Full-Time University

Graduate Employees (in 2020) with their Pay Set by Individual Agreement (IA) and the Elasticity of Earnings to Hours Table S1. Gender Earnings Gap Estimates

Among Full-Time Employees Table S2. Gender Earnings Gap Estimates

Among All Employees Table S3. Gender Earnings Gap Estimates

Among Full-Time University Graduates Table S4. Gender Earnings Gap Estimates

Among All University Graduates Table S5. Gender Earnings Gap Estimates

Among Full-Time University Graduates Table S6. Gender Earnings Gap Estimates

Among Full-Time University Graduates, with Wage-Setting Table S7. Gender Earnings Gap Estimates

Among Full-Time University Graduates, Unrestricted Sample Table S8. Returns to Labour Market

Experience Among Full-Time University Graduates, by Sex

REFERENCES

Altonji, J. and Blank, R. (1999), ‘Race and Gender in the Labor Market’, in Ashenfelter, O. and Card, D. (eds), Handbook of Labor Economics, vol. 3, Part C, Chapter 48, 1st edn. Elsevier, Amsterdam; 3143–3259.

Baker, M., Halberstam, Y., Kroft, K., Mas, A. and Messacar, D. (2023), ‘Pay Transparency and the Gender Gap’, American Economic Journal: Applied Economics, 15, 157–683.

Baron, J. and Cobb-Clark, D. (2008), ‘Occupa- tional Segregation and the Gender Wage Gap in Private- and Public-Sector Employment: A Distributional Analysis’, https://doi.org/10. 2139/ssrn.1155836

Bennedsen, M., Simintzi, E., Tsoutsoura, M. and Wolfenzon, D. (2022), ‘Do Firms Respond to Gender Pay Gap Transparency?’, The Journal of Finance, 77, 2051–91.

Bertrand, M. (2011), ‘New Perspectives on Gender’, Handbook of Labor Economics, 4B, 1543–90.

Bertrand, M. (2020), ‘Gender in the Twenty-First Century’, AEA Papers and Proceedings, 110, 1–24.

Biasi, B. and Sarsons, H. (2021), ‘Flexible Wages, Bargaining, and the Gender Gap*’, The Quar- terly Journal of Economics, 137, 215–266.

Blau, F.D. and Kahn, L.M. (2017), ‘The Gender Wage Gap: Extent, Trends, and Explanations’, Journal of Economic Literature, 55, 789–865.

Borland, J. and Coelli, M. (2016), ‘Labour Market Inequality in Australia’, The Economic Record, 92, 517–547.

Card, D., Cardoso, A.R. and Kline, P. (2015), ‘Bargaining, Sorting, and the Gender Wage Gap: Quantifying the Impact of Firms on the Relative Pay of Women*’, The Quarterly Journal of Economics, 131, 633–686.

Cha, Y. and Weeden, K.A. (2014), ‘Overwork and the Slow Convergence in the Gender Gap in Wages’.

Cobb-Clark, D.A. and Tan, M. (2011), ‘Noncog- nitive Skills, Occupational Attainment, and Relative Wages’, Labour Economics, 18, 1–13.

Cortes, P. and Pan, J. (2016), ‘Prevalence of Long Hours and Skilled Women’s Occupational Choices’, Technical Report 10225. Institute of Labor Economics (IZA).

Cortes, P. and Pan, J. (2017), ‘Cross-Country Evidence on the Relationship Between Over- work and Skilled Women’s Job Choices’, American Economic Review, 107, 105–9.

Cortes, P. and Pan, J. (2018), ‘Occupation and Gender’, in Averett, S.L., Argys, L.M. and Hoffman, S.D. (eds), The Oxford Handbook of Women and the Economy. Oxford University Press, Oxford. https://doi.org/10.1093/oxfordhb/ 9780190628963.013.12

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

488 ECONOMIC RECORD DECEMBER

Cortés, P. and Pan, J. (2019), ‘When Time Binds: Substitutes for Household Production, Returns to Working Long Hours, and the Skilled Gender Wage Gap’, Journal of Labor Economics, 37, 351–398.

Cullen, Z.B. and Pakzad-Hurson, B. (2021), ‘Equilibrium Effects of Pay Transparency’, Working Paper 28903. National Bureau of Economic Research.

Daehoon, N., Michael, D. and Craig, M. (2022), ‘Union Wage Effects in Australia in a Period of Declining Union Power: The Role of Endowments and Returns to Endowments’, Australian Journal of Labour Economics, 25, 148–170.

Denning, J.T., Jacob, B.A., Lefgren, L.J. and vom Lehn, C. (2022), ‘The Return to Hours Worked within and Across Occupations: Implications for the Gender Wage Gap’, ILR Review, 75, 1321–47.

Evans, A.C. (2019), ‘Why we Use Private Trusts in Australia: The Income Tax Dimension Explained’, The Sydney Law Review, 41, 217– 254.

Freeman, R. and Medoff, J. (1984). What Do Unions Do. N.Y.: Basic Books. Current publisher: currently Perseus Book Group Japanese translation (1986), French translation (1987), Management Association Prize Book, 1985. Reissued in Mandarin and English, by Beijing University Press, first in series, Clas- sics in Economics (Bejing: 2006). Excerpts reprinted in Summers, Dau-Schmidt and Hyde (eds) Legal Rights and Interests in the Work- place (Indiana University, 2006).

Goldin, C. (2014), ‘A Grand Gender Conver- gence: Its Last Chapter’, American Economic Review, 104, 1091–1119.

Goldin, C. (2021), Career and Family. Princeton University Press, Princeton. https://doi.org/10. 1515/9780691226736

Goldin, C. and Katz, L. (2011), ‘The Cost of Workplace Flexibility for High-Powered Pro- fessionals’, The Annals of the American Acad- emy of Political and Social Science, 638, 45–67.

Goldin, C. and Katz, L.F. (2016), ‘A Most Egalitarian Profession: Pharmacy and the Evo- lution of a Family Friendly Occupation’, Journal of Labor Economics, 34, 705–745.

Hayter, S. (2011), The Role of Collective Bar- gaining in the Global Economy: Negotiating for Social Justice. Edward Elgar, Cheltenham. https://doi.org/10.4337/9781849809832

Kee, H.J. (2006), ‘Glass Ceiling or Sticky Floor? Exploring the Australian Gender Pay Gap’, The Economic Record, 82, 408–427.

Kidd, M.P. (1993), ‘Sex Discrimination and Occupational Segregation in the Australian

Labour Market’, The Economic Record, 69, 45–55.

Kuhn, P. and Lozano, F. (2008), ‘The Expanding Workweek? Understanding Trends in Long Work Hours Among U.S. Men, 1979–2006’, Journal of Labor Economics, 26, 311–343.

Le Barbanchon, T., Rathelot, R. and Roulet, A. (2021), ‘Gender Differences in Job Search: Trading Off Commute Against Wage*’, The Quarterly Journal of Economics, 136, 381–426.

Leibbrandt, A. and List, J.A. (2015), ‘Do Women Avoid Salary Negotiations? Evidence from a Large-Scale Natural Field Experiment’, Man- agement Science, 61, 2016–24.

Lundborg, P., Plug, E. and Rasmussen, A.W. (2017), ‘Can Women have Children and a Career? IV Evidence from IVF Treatments’, American Economic Review, 107, 1611–37.

Mas, A. and Pallais, A. (2020), ‘Alternative Work Arrangements’, Annual Review of Economics, 12, 631–658.

Meekes, J. and Hassink, W.H. (2022), ‘Gender Differences in Job Flexibility: Commutes and Working Hours After Job Loss’, Journal of Urban Economics, 129, 103425.

Miller, P.W. and Lee, Y.L. (2004), ‘Occupational Segregation on the Basis of Gender: The Role of Entry-Level Jobs’, Australian Journal of Labour Economics (AJLE), 7, 355–374.

OECD (2020), ‘Can Collective Bargaining Help Close the Gender Wage Gap for Women in Non-Standard Jobs?’

Pan, J. (2015), ‘Gender Segregation in Occupations: The Role of Tipping and Social Interactions’, Journal of Labor Economics, 33, 365–408.

Pertold-Gebicka, B., Pertold, F. and Datta Gupta, N. (2016), ‘Employment Adjustments around Childbirth,’ Technical Report 9685. Institute of Labor Economics (IZA).

Rimmer, S.M. (1991), ‘Occupational Segregation, Earnings Differentials and Status Among Aus- tralian Workers’, The Economic Record, 67, 205–216.

Rosen, S. (1987), ‘The Theory of Equalizing Differences’, in Handbook of Labor Econom- ics, vol. 1. Elsevier, Amsterdam; 641–692.

Roussille, N. (2022), ‘The Central Role of the Ask Gap in Gender Pay Inequality’, Working Paper.’

Sainsbury, T. and Breunig, R. (2020), ‘Tax Planning in Australia’s Income Tax System’, Agenda: A Journal of Policy Analysis and Reform, 27, 59–83.

Summerfield, M., Garrard, B., Hahn, M., Jin, Y., Kamath, R., Macalalad, N., Watson, N., Wil- kins, R. and Wooden, M. (2020), HILDA User Manual – Release 19. Melbourne Institute: Applied Economic and Social Research, Uni- versity of Melbourne.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

2024 GREEDY JOBS AND THE GENDER PAY GAP 489

Wasserman, M. (2022), ‘Hours Constraints, Occu- pational Choice, and Gender: Evidence from Medical Residents’, Review of Economic Stud- ies, 90, 1535–68.

Wiswall, M. and Zafar, B. (2018), ‘Preference for the Workplace, Investment in Human Capital, and Gender*’, The Quarterly Journal of Eco- nomics, 133, 457–507.

! 2024 The Author(s). Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

490 ECONOMIC RECORD DECEMBER

Copyright of Economic Record is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.