- About NATSEM
- Author note
- General caveat
- Executive Summary
- 1 Introduction
- 2 The gender wage gap in Australia, 1990 – 2009
- 3 Potential drivers of the gender wage gap in Australia
- 4 Microeconomic Modelling of the gender wage gap in Australia
- 5 Gender inequality and macroeconomic performance
- 5.1 Background
- 5.2 Measuring economic performance
- 5.3 The Macroeconomic costs of the gender wage gap
- 6 Summary and Conclusion
- Appendix A - Decomposition Approaches
- Appendix B - Microeconomic model specification, variable descriptions and detailed decomposition results
- Appendix C - Gender inequality in education: impact on economic performance
- Appendix D - The development of growth theory and its application to studies of gender inequality
- Appendix E - Macroeconomic estimation methods and results
The National Centre for Social and Economic Modelling was established on 1 January 1993, and supports its activities through research grants, commissioned research and longer term contracts for model maintenance and development.
NATSEM aims to be a key contributor to social and economic policy debate and analysis by developing models of the highest quality, undertaking independent and impartial research, and supplying valued consultancy services.
Policy changes often have to be made without sufficient information about either the current environment or the consequences of change. NATSEM specialises in analysing data and producing models so that decision makers have the best possible quantitative information on which to base their decisions.
NATSEM has an international reputation as a centre of excellence for analysing microdata and constructing microsimulation models. Such data and models commence with the records of real (but unidentifiable) Australians. Analysis typically begins by looking at either the characteristics or the impact of a policy change on an individual household, building up to the bigger picture by looking at many individual cases through the use of large datasets.
It must be emphasised that NATSEM does not have views on policy. All opinions are the authors’ own and are not necessarily shared by NATSEM.
© Commonwealth of Australia 2009
This work is copyright. Apart from any use as permitted under the Copyright act 1968, no part may be reproduced by any process without prior written permission from the Commonwealth, available from the Commonwealth Copyright Administration, Attorney-General’s Department. Requests and inquiries concerning reproduction and rights should be addressed to the Commonwealth Copyright Administration, Attorney-General’s Department, Robert Garran Offices, National Circuit, Canberra ACT 2600 or posted at on the Attorney-General's Deartment website.
Rebecca Cassells is a Senior Research Fellow at the National Centre for Social and Economic Modelling (NATSEM) at the University of Canberra. Yogi Vidyattama is an Acting Research Fellow, Riyana Miranti is a Research Fellow and Justine McNamara is an Acting Principal Research Fellow, all at NATSEM.
National Centre for Social and Economic Modelling University of Canberra ACT 2601 Australia 170 Haydon Drive Bruce ACT 2617
Phone: + 61 2 6201 2780
Fax: + 61 2 6201 2751
This research was initiated and funded by the Office for Women, Department of Families, Community Services, Housing and Indigenous Affairs (FaHCSIA).
This research report has been produced with the assistance of many people. The details of each contribution are provided briefly below.
Macroeconomics.com.au was consulted as a specialist adviser on the project, and we would like to acknowledge the director of Macroeconomics.com.au, Dr Stephen Anthony, and thank him for the expert advice he provided in relation to our macroeconomic model.
Associate Professor Anne Daly from the University of Canberra, and an Associate Researcher at NATSEM, was also consulted as a specialist adviser on the project in relation to gender wage gaps, decomposition methodologies and Australian labour markets. We would like to acknowledge Anne’s contribution and guidance for our microeconomic model.
Previous work by Professor Stephan Klasen (Professor of Development Economics, University of Göttingen, Germany), has helped shape and inform the macroeconomic model used here. Stephan has also given helpful advice during the course of this project.
Dr Ian Watson (Senior Research Fellow, Macquarie University and Social Policy Research Centre, UNSW), provided useful assistance and advice in relation to technical aspects of wage gap decomposition methodologies.
This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Australian Government Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) and is managed by the Melbourne Institute of Applied Economic and Social Research (MIAESR). The findings and views reported in this paper, however, are those of the authors and should not be attributed to either FaHCSIA or the MIAESR.
NATSEM research findings are generally based on estimated characteristics of the population. Such estimates are usually derived from the application of microsimulation models or other economic modelling techniques to microdata based on sample surveys.
These estimates may be different from the actual characteristics of the population because of sampling and nonsampling errors in the microdata and because of the assumptions underlying the modelling techniques.
The microdata do not contain any information that enables identification of the individuals or families to which they refer.
- The persistent nature of Austalia's gender wage gap
- Key determinants of the gender wage gap in Australia
- The costs of the gender wage gap to the macro economy
- A few words of caution
The persistence of the gender wage gap in Australia and overseas has been the subject of much research. Debates about how best to measure the wage gap, how to calculate the relative contribution different factors make to the gap, and how the gap and its determinants differ across sectors and income levels abound. However, few if any studies, and none for Australia, provide a comprehensive and detailed account of the impact of the gender wage gap to economic growth. This paper seeks to provide such an account, with our research presented in the body of this paper, and our key findings summarised, as follows.
The persistent nature of Austalia's gender wage gap
Analysis of Australian Bureau of Statistics earnings data confirms that Australia has a persistent gender wage gap. Data from the Average Weekly Earnings Survey show that between 1990 and 2009, the gender wage gap remained within a narrow range of between 15 and 17 per cent. Indeed over the last four years the gap has steadily increased within this range from a low of 15.1 per cent in February 2005 to 17.0 per cent in February 2009.
Key determinants of the gender wage gap in Australia
In estimating the impact of a sustained gender wage gap on the Australian economy, it was instructive to first identify and quantify the relative contributions of the key determinants of the gap.
Utilising robust microeconomic modelling techniques, based on a comprehensive and critical evaluation of several methodologies, we found that simply being a woman is the major contributing factor to the gap in Australia, accounting for 60 per cent of the difference between women’s and men’s earnings, a finding which reflects other Australian research in this area. Indeed, using wage gap analysis from the HILDA survey, the results showed that if the effects of being a woman were removed, the average wage of an Australian woman would increase by $1.87 per hour, equating to an additional $65 per week or $3,394 annually, based on a 35 hour week.
Other key determinants of the gap that were identified and quantified as part of the microeconomic modelling component of our research were industrial segregation (25 per cent), labour force history (seven per cent), under-representation of women with vocational qualifications (five per cent) and under representation of women in large firms (three per cent).
Overall, and as detailed in the body of this paper, our finding that simply being a woman is the major contributing factor to the wage gap in Australia is significant. Consistent with results from other Australian studies it highlights the considerable impact that discrimination and other differences between men and women, including differing motivations and preferences, can have on reducing the earnings of women relative to men, irrespective of similar labour force and work-related characteristics.
The costs of the gender wage gap to the macro economy
At the microeconomic level, the negative impacts of key drivers of the gender wage gap are significant. Yet what are its effects on the Australian economy as a whole? An increasing body of literature shows that in addition to fairness and equity there are also strong economic imperatives for addressing the gender wage gap.
Using rigorous macroeconomic modelling techniques, again selected on the basis of a critical evaluation of several methodologies, we found that the gender wage gap has a substantial effect on Australia’s economic performance, measured in terms of GDP per capita, and that the value of reducing the gap is substantial. For example, we estimate that a decrease in the gender wage gap of 1 percentage point from 17 per cent to 16 per cent would increase GDP per capita by approximately $260. This equates to around $5,497 million (2007 dollars) or 0.5 per cent of total GDP, assuming that the Australian population is held constant. The results also indicate that eliminating the whole gender wage gap from 17 per cent to zero, could be worth around $93 billion or 8.5 per cent of GDP.
The impact of the wage gap on macroeconomic performance was measured both in terms of its direct impact on economic growth and through several indirect channels, including investment, fertility, hours of work and labour force participation. Diagnostic analysis of these results showed that while the relationship between the gender wage gap and economic growth was negative for each of these variables, except labour force participation, only the hours of work channel was statistically significant. This suggests that based on the assumptions of the model, the negative impact of the gender wage gap on Australia’s macroeconomic performance stems primarily from the disincentives to work associated with women’s earnings being lower than men’s.
Using our findings at a microeconomic level, we have also estimated the effect of each key determinant of the gender wage gap at the macro level. Disaggregating the estimated $93 billion cost to the economy of the gender wage gap, we estimate that removing the negative effects associated with the prime determinant of the gap, that is being a woman, could add around $56 billion or 5.1 per cent to total annual GDP.
A few words of caution
We recognise that the findings presented in this paper are subject to the limitations and assumptions of the underpinning microeconomic and macroeconomic models. As detailed in the main body and appendices of this paper, we have sought to identify and minimise, to the greatest extent possible, these limitations. It should be noted, however, that both areas of modelling are the subject of vigorous debate about the strengths of a range of methodological techniques. While our work is firmly based on best practice and has been subjected to rigorous quality assurance procedures, the use of different data sets coupled with ongoing developments and improvements in methodological techniques may potentially produce different results. Details of our methodology and its potential limitations are provided in the appendices to this report.
The gender wage gap in Australia and overseas has been the subject of much research. Debates about how best to measure the wage gap, how to calculate the relative contribution different factors make to the gap, and how the gap and its determinants differ across sectors and income levels abound. Little is known, however, about the costs to the economy of a sustained gender wage gap or indeed the value in terms of the macroeconomic benefits of reducing the gap. In June 2009, the Office for Women, Department of Families, Community Services, Housing and Indigenous Affairs (FaHCSIA), commissioned the National Centre for Social and Economic Modelling (NATSEM) to undertake research estimating the value to the economy of reducing the gender wage gap in Australia. This report presents the results of our research with the analysis structured as follows:
- The gender wage gap in Australia 1990-2009. This section presents the historical context to the report with a time series analysis of data from the Australian Bureau of Statistics’ (ABS) Average Weekly Earnings survey. It shows that between 1990 and 2009, Australia experienced a persistent gender wage gap in the range of 15.5 to 17.5 per cent.
- Potential drivers of the gender wage gap in Australia. This section reviews recent international and Australian literature that examines the complex and inter-related causes of the gender wage gap. We identify the most widely accepted determinants of the wage gap, and the most robust method for decomposing the wage gap for our purposes in this project. This analysis is then used as the basis for selecting a methodology and variables for use in modelling the gender wage gap for this project.
- Microeconomic modelling of the gender wage gap in Australia. This section discusses the important considerations to take into account when measuring wage gaps, in particular the importance of feedback effects, and measuring direct discrimination. We find an hourly wage gap using HILDA data of $3.13 (approximately 11 per cent). We also find that the key factor associated with the wage gap is simply being a woman. We estimate that 60 per cent of the wage gap is due to either direct discrimination or other factors to do with being a woman.
- Gender inequality and macroeconomic performance. In this section, we discuss the importance of measuring the effect of wage gaps on economic growth and pathways through which wage gaps can impact upon economic growth, drawing on a range of international literature. We find that there are significant benefits to addressing the wage gap in terms of economic efficiency. Our modelling shows that the Australia economy would gain approximately 0.5 percent of GDP if the wage gap decreased by one per cent, from the current AWE calculation of 17 per cent to 16 per cent. This would mean a gain per capita of GDP of approximately $260, equating to around $5,497 million (2007) dollars. We also use the findings from our microeconomic modelling of the key determinants of the wage gap to estimate the effect of each of these on the macro economy. Specifically, we estimate that removing the effects associated with being a woman (60 per cent of the wage gap), could see an improvement in GDP of around $56 billion per year.
Appendices. There are several appendices provided with this report, which provide greater detail and discussion of the literature, results and the extensive testing and modelling that has been undertaken.
2 The gender wage gap in Australia, 1990 – 2009
Over the past century, and more notably the past few decades, there have been many changes that shape the way in which Australian women participate in society and the economy. Today, women are now more likely than ever before to be participating in paid employment, are highly educated, and are delaying and having fewer children. Important legislative changes including the Matrimonial Causes Act (1961) and the Federal Sex Discrimination Act (1986), as well as social transformations, including the introduction of the contraceptive pill and the feminist movement, have all aided women’s quest for equality. In relation to pay equity, the Commonwealth Conciliation and Arbitration Commission’s 1969, 1972 and 1974 judgements in relation to the principle of equal pay for equal work and work of equal value were important milestones in relation to wage discrimination (see Pocock 1999 for a review of relevant policy, legislative and practice developments). Has equality been achieved though? In particular, are Australian women being rewarded for their efforts in the workplace fairly?
In this section we analyse the gender wage gap in Australia over the past two decades in order to see if the ground gained in terms of education, employment and social justice is reflected in women’s incomes. This provides an historical context to the gender wage gap within Australia, by calculating its movements over time using data from the Australian Bureau of Statistics (ABS) Average Weekly Earnings series.
We have considered several data sources that could be used in order to gain an accurate, consistent, un-broken time series of the gender wage gap from the 1990s to the present., including the ABS Survey of Income and Housing (SIH), the ABS Employee Earnings and Hours Survey (EEH), the Employee Earnings, Benefits and Trade Union Membership (EEBTUM) and the ABS Average Weekly Earnings (AWE) Series. While each of these data sources has advantages and limitations, we have chosen to derive a time series of the gender wage gap using the ABS Average Weekly Earnings Series. The AWE has the advantage that, as an employer-based survey, the data on earnings are considered more reliable than household-based surveys such as EEBTUM and the SIH. In addition, unlike the EEH, which is also an employer based survey, the AWE is regular, consistent and has the advantage in that it provides a long uninterrupted time series.
In order to make the raw earnings data of men and women as comparable as possible over time, we limited our sample to full-time working adults working ordinary time1 . It is important to note that a large proportion of women work part-time compared with men and including part-time employees in the wage gap estimate would increase the gap. Also, had we included overtime hours, the wage gap would also increase slightly, as men are more likely to have access to paid overtime hours than women2 . Thus, limiting our sample in these ways increases the rigour of our wage gap estimates, as it reduces the labour market differences that exist between men and women, allowing us to focus more closely on wage differences only. However, there will nevertheless be some individual and organisational differences that cannot be accounted for in these wage gap estimates, such as level of education, or number of years work experience.
Note: The gender wage gap is calculated for full time, ordinary time adult employees, using original data. The reference period for data used in this figure is February for each year. Source: ABS Average Weekly Earnings, Data cube, 2009, Cat No. 6302.0
The persistence of the gender wage gap in Australia over the past two decades is illustrated in Figure 1. Whilst the wage gap has fluctuated slightly over this period, it has consistently remained between 15 and 17 percent, with women receiving around 83 to 85 per cent of the average man’s wage. Between 1996 and 2005 the gender wage gap exhibits a downward trend falling from 16.8 per cent to 15.1 per cent but in the four years since then it has risen quite sharply, wiping out the previous gains and in effect leaving the gap slightly above the level it was almost twenty years earlier. Currently (based on data from February 2009), women who work ordinary full-time hours earn on average $1,049.40 per week in comparison to full-time men who earn on average $1,263.70 per week.
- Full-time working adults are defined as those persons aged 21 years and above. Ordinary-time covers standard or agreed hours of work, thus excluding overtime, (ABS, 2009a).
- The gender wage gap calculated based on AWE, using an alternative group of people - full time adults, total earnings, which includes overtime was 20.8 per cent in 1990 and 19.7 per cent in 2009.
3 Potential drivers of the gender wage gap in Australia
The causes of the gender wag gap while complex, inter-related and likely to vary over time can for analytical simplicity be grouped into two main factors, namely explained (human capital or labour market factors) and unexplained characteristics.
3.1 Discrimination or Characteristics
Many Australian and international studies on the gender wage gap have tended to focus on the question of whether it is discrimination that is the major driver of the gap in wages between men and women, or if the gap is driven primarily by differences in men’s and women’s human capital attainment and other characteristics. These differences have proved difficult to tease out, partly because of the ways in which discrimination may be intertwined with the attainment of other characteristics.
Discrimination can be defined as occurring ‘when participants in the marketplace take into account such factors as race and sex when making economic exchanges’ (Borjas 2010, p. 365). However, such effects may occur not just in the selection by employers of men over women into higher-paying jobs, but may also form a complex part of the decisions women and men make around study, career, family and labour market participation. These indirect effects of discrimination on men’s and women’s decisions and characteristics are also known as ‘feedback effects’ (see, for example, Grimshaw and Rubery 2002; Rummery 1992). The interplay between direct and indirect effects of discrimination means that it is very hard to isolate this from other drivers of the gap. Teasing out discrimination from other factors which drive wage differences is generally discussed in the literature in terms of endowments, and rewards for endowments.
Most studies decompose the wage gap using the Oaxaca-Blinder method or variations thereon (see Appendix A for further discussion), with the focus of this decomposition being to calculate the extent to which the difference between men’s and women’s wages can be explained by the total effect of all the variables included in the model, and what cannot be explained by these variables. The effect of the included variables is deemed to be that part of the gender wage gap which is explained by ‘endowments’ (that is, measureable differences between men and women in education, work experience, training and so on), while the portion of the wage gap that cannot be explained by these endowments is assumed to be explained by ‘rewards’ for these endowments3. These rewards represent the extent to which women are paid less than men once all other measurable characteristics are held constant, and may include discrimination as well as any other unobserved differences between men and women (which, as Olsen and Walby (2004) note, may include differing motivations and preferences). In these types of studies, therefore, the only determinants of the wage gap which are often actually quantified are the totality of ‘endowments’ and the remaining unexplained portion of the variation in wages. Attempts are not always made to separately quantify the role of the particular variables which fall within the broad category of ‘endowments’, and which will themselves be to a certain extent influenced by discrimination4.
Findings about the determinants of the Australian gender wage gap generally show that rewards for endowments are more important than endowments themselves (see, for example, Borland 1999; Cassells et al. 2008; Daly et al. 2006; Eastough and Miller 2004; Kee 2006; Preston 2000; Wooden 1999). Miller (2005, p.405), summarising these findings, notes that in many studies only around one-fifth of the wage gap between men and women can be explained by productivity-related characteristics5. Kee (2006) additionally examined the extent to which the wage gap, and the extent it can be explained by endowments versus rewards, differs across the income distribution. She found that the difference in return to gender characteristics (rather than differences in endowments) was the most important determinant of the gender wage gap with the effect of rewards most profound at the highest end of the income distribution in the private sector.
Thus overall there is substantial evidence to suggest that a combination of discrimination or other unobserved characteristics play an important role in maintaining the wage gap in Australia. However, some Australian work goes on to examine in more detail the other drivers of the wage gap. A summary of these findings is provided below.
3.2 Other wage gap determinants
The findings summarised above suggest that human capital and labour market differences between men and women are less important in understanding the gender wage gap in Australia than variations in the way these differences are rewarded (or reflected in pay). However, the fact that discrimination in some form (as Olsen and Walby 2004 note) is likely to be closely intertwined with a range of the endowment and control variables used in wage gap models, suggests that further enquiry into the role which these factors play in maintaining the wage gap is vital. This is all the more so as current methodologies (discussed further in Appendix A) make it difficult to identify what proportion of the ‘rewards for endowments’ is in fact explained by discrimination, and what may simply be due to other unobserved differences between men and women (which in themselves may also be affected by discrimination).
Australian work which provides some information about the key determinants of the gender wage gap includes studies which present detailed decomposition results, as well as those which discuss more broadly factors known to influence women’s earnings. However, the contribution of various factors to the wage gap in Australia is very uncertain.
Studies of the gender wage gap have generally divided possible determinants of the wage gap into two groups: those related to human capital, and those related to labour market-related institutions and rigidities (see, for example, Walby and Olsen 2002). There is some consensus in the literature about a number of determinants of the wage gap in Australia within both these broad groups, but not for all variables. An additional body of literature examines the effects of personality characteristics which may affect occupation choice, hours of work, promotion and so on, and thus wages. Fortin (2008) and Cobb-Clark and Tan (2009) study the effects of noncognitive traits (e.g. self-efficacy and personality traits), on wages and the gender wage gap. Fortin (2008), focusing particularly on factors which are known to differ by gender (in particular the relative importance or value put on money/work and people/family) finds a modest but significant role of these variables. Using Australian data, Cobb-Clark and Tan (2009) further refine this methodology by examining the influence of noncognitive factors on occupational attainment first, and through occupation on wages. They do this by examining the effects of the noncognitive factors (along with other factors likely to influence wage gaps) for each occupation separately, and find that such factors have little effect on the wage gap.
Human capital variables
The recognition of human capital as a predictor of wage levels (and thus the wage gap) emerges from work by Becker (1975) and Mincer (1974). Walby and Olsen (2002) define human capital as:
‘all the skills and experience that a person brings to employment that are relevant to that employment. It encompasses not only educational qualifications and training by employers, but also experience that people have as result of prior labour market experience’ (Walby and Olsen 2002, p. 22).
In studies of the gender wage gap, human capital is usually measured through formal educational attainment and years of work experience (often proxied, as discussed below). In addition, some studies include further variables to capture human capital, including the use of employer-provided training. The literature usually takes as given the human capital developed at the point of entry to the labour market and focuses on the rewards to that human capital once in the labour market.
The education component of human capital is captured in virtually all gender wage gap studies, most often through the use of a set of dummy variables capturing the highest level of education completed – for example, completed high school, some post school training, bachelor degree and above (see, for example, Cassells et al . 2008; Preston 2003)6 . Some work (e.g. Drolet 2001) also includes field of study as a variable within gender wage gap models, as wages differ by field of study, and there is considerable gender segregation in post-secondary fields of study.
Suggestive of discrimination and labour market rigidities, Australian studies have found that returns to education for women are generally lower than those for men, despite women’s somewhat higher level of educational attainment (Miller 2005; Rummery 1992)7 . As Miller (2005) notes: ‘additional schooling opens up access to better paying positions more readily for males than for females’ (Miller 2005, p. 413).
While previous literature is unanimous in acknowledging the importance of previous work experience in measuring the human capital contribution to the gender wage gap, there are considerable differences in the way in which this variable is measured. In the international literature, Walby and Olsen (2004), Grimshaw and Rubery (2002) and Drolet (2001) note that experience has generally been measured through a proxy variable designed to capture potential work experience, constructed by deducting years of schooling from age, less an approximate number of pre-school years, generally due to an absence of actual data about work experience (Australian examples of this approach include the work of Daly et al. 2006; Miller 2005; Preston 2003). However, the use of this type of proxy variable has serious flaws, especially for women, who are likely to have breaks in their labour market experience (Olsen and Walby 2004; Rummery 1992). As Drolet (2001) notes, a proxy variable for potential years of work experience based on age and years of schooling does not take into account withdrawal from the labour market or participation in part-time (rather than full-time) work.
A range of approaches to the measurement of years of work experience have been used in Australian studies8 . Despite the use of different measures to proxy experience (Eastough and Miller 2004; Miller 2005; Rummery 1992), results are consistent with findings generally confirming that returns to work are higher for men than women. Additional years of labour market experience translate into greater increases in wages for men than for women. Miller, using the potential experience variable from the 2001 Census data (which is equal to age minus school leaving age), finds that returns to labour market experience (evaluated at 10 years experience) is 1.8 per cent for men and 1.4 per cent for women. The contribution of years of experience to the gender wage gap is particularly substantial when variables which capture actual rather than potential experience are used. For example, Rummery (1992) using the 1984 National Social Science Survey, finds that the return to actual labour market experience (evaluated at ten years of work) for men is 1.45 per cent, yet for women it is only 1.32 per cent.
The effects on human capital attainment, earnings and the gender wage gap of interruptions and alterations to labour market experience (not working or working part-time) due to child bearing and rearing and other caring duties are widely acknowledged in the literature (for a detailed discussion see Baum 2002; Budig and England 2001; Walby and Olsen 2002). These issues are complex and multifaceted, and are by no means limited to the simple fact that not working, or working part-time, eliminates or reduces wages for a particular period.
The possible repercussions of interruptions to work on lifetime levels of pay follow several pathways. Drolet (2001, p. 7) summarises the effects of labour force withdrawals as follows. First, non-continuous work is associated with shorter periods of job tenure, which in turn is associated with lower pay. Second, the value of human capital may deteriorate while women are out of the workforce, and thus when they return these effects may result in a lower likelihood of promotion or lower wages. Third, women facing interruptions to their career may choose not to participate in training, or may decide to accept low-wage jobs. Finally, labour market withdrawals may often coincide with the beginning of women’s careers, a time at which the acquisition of job skills (and therefore job advancement and wages growth) is particularly strong for non-withdrawers. Olsen and Walby (2004) also note that withdrawals from the labour force can have a negative impact on earnings through discrimination.
In Australian studies, labour force withdrawals have been captured through variables which show how many children women have (as the presence of children is likely to have resulted in a period of withdrawal), and is also captured by years of work experience when this is included in models. As noted above, the latter variable is generally shown to be an important determinant of the wage gap, and the presence of children also tends to contribute to the wage gap. The presence of children, particularly young children, causes women to participate less in the labour market and work fewer hours than women without children or men (Lundberg and Rose 2000; Sigle-Rushton and Waldfogel 2006). This also contributes to women with children earning less wages than women without children or men (Lundberg and Rose 2000; Sigle-Rushton and Waldfogel 2006). Eastough and Miller (2004) note that in Australia, among full-time wage and salary earners, women with dependent children earn 7.5 per cent less than women without dependent children, whilst men with dependent children have slightly higher earnings than men who do not have dependent children. Further, the presence of children has been found to influence men’s and women’s lifetime earnings. Cassells et al. (2009) find that the lifetime earnings in Australia for a man with children is almost double those of his female counterpart ($2.5 million compared to $1.3 million), whereas in contrast the lifetime earnings for a man without children is hardly any different from a woman without children ($2 million compared to $1.9 million).
In contrast to international findings, in Australia current part-time work status has not been found to be a significant driver of the gender wage gap (see, for example, Booth and Wood 2006; Rodgers 2004). However, whilst current part-time work status in Australia does not appear to drive the wage gap, a prolonged history of part-time work may be associated with lower pay, due to factors associated with long-term part-time employment such as lower on-the-job training being offered and taken up. Olsen and Walby (2004), in a UK context, note some of the complexities associated with the effect of a history of part-time work on women’s wages. They point out that part-time work in itself may be associated with lower rates of human capital attainment, that is, it cannot be assumed that years of experience in part-time work equates to the same level of skills acquisition (and therefore pay rate) as years of experience in full-time work.
Labour market factors
Not all possible determinants of the gender wage gap are related to individual characteristics such as education and work experience. Walby and Olsen (2002, p.22), in discussing the implications of the gender wage gap for productivity, note the role that ‘failures in the market for labour’ play in generating and maintaining the wage gap, and point particularly to labour market rigidities associated with occupation and industrial segregation, insufficient flexibility in the labour market to allow women to combine work with child-rearing, and discrimination. In addition, they note a series of other labour market factors which are broadly associated with wage determination including occupational segregation, unionisation, public versus private sector employment (which can reflect a centralised or decentralised bargaining system); industrial sector and firm size. Many of these appear to play some role in the persistence of the gender wage gap in Australia, but once again findings are mixed.
Occupational segregation can be defined as the extent to which ‘women and men are differently distributed across occupations than is consistent with their overall shares of employment’ (Watts 2003, p. 631). Occupational segregation is important to include in analyses of gender wage gaps, as it can often be an indicator of the undervaluing of women’s work. That is, traditionally ‘female’ jobs tend to be less well paid than traditionally ‘male’ jobs. Occupational segregation is a complex area of research, with debates about measurement of this concept driven by a range of theoretical and empirical approaches (Watts 2003). Occupation, and particularly the gender segregation of occupations, plays an important role in predicting wages, although the ways in which occupation and occupational segregation are included in models can mean varying results9.
In an Australian context, the effects of occupational segregation on the gender wage gap are not clear. While some studies have found that occupational segregation contributes to the gender wage gap in Australia (see, for example, Miller, 1994, Preston and Whitehouse 2004; Robinson, 1998; Wooden 1999), other work suggests that in Australia occupational segregation has the opposite effect - that is, if occupations were desegregated and no longer had unequal representations of men and women, women’s pay would be lower, not higher (Barón and Cobb-Clark 2008; Preston and Crockett 1999; Watts 2003). Cobb-Clark and Tan (2009), examining the gender wage gap separately for 18 different occupational classifications, conclude that, while this methodology leads to a greater portion of the gender wage gap being explained by men’s and women’s different characteristics, it is not occupational segregation that contributes to this, but rather women getting paid less than men within the same occupational groups.
Industrial segregation has been a less strong focus of gender wage gap literature. However the effect of segmentation of industries along gender lines is included in many wage gap models (see, for example, Cassells et al. 2008; Daly et al. 2006; Eastough and Miller 2004; Preston 2000; Preston and Crockett 1999; Wooden 1999), and Australian studies have generally shown that industrial segregation tends to widen the wage gap (see, for example, Cassells et al. 2008; Miller, 1994; Preston and Crocket 1999). These findings concur with those of a number of international studies which show that industrial segregation is associated with larger wage gaps (Grimshaw and Rubery 2002). Drolet (2001) found that industry of employment accounted for around eleven percent of the wage gap in Canada. Preston and Crockett (1999), who explicitly quantified the contribution that industry makes to the wage gap in Australia, found that this variable accounted for around 45 per cent of the explained portion of the gap. Their study decomposed the wage gap separately for each state, and they found that the effect of industry was particularly strong in Western Australia and Queensland.
Sector of Employment
Public-private sector effects on the gender wage gap in Australia have been examined and studies generally find that the wage gap is larger in the private than the public sector (Barón and Cobb-Clarke 2008; Kee 2006; Preston 2000), although findings about the possibly differential effect of public sector employment on men’s and women’s wages differ somewhat across studies (Miller 2005; Wooden 1999).
Unionisation may have some effects on reducing the gender wage gap in Australia, particularly for lower wage workers (Barón and Cobb-Clarke 2008; Miller 2005) but findings about this relationship have been mixed. Wooden (1999), for example, found insignificant or weak negatively significant effects of union membership on wages, and Cai and Liu (2008) found that unions have a larger effect on men’s wages than on women’s.
Firm size is also associated in the international literature with both income levels and income gaps – large firms pay more than small firms on average, and women are more likely to work in small firms than men (Drolet 2001). In Australia, Daly et al. (2006) use a set of six variables measuring firms with less than 30 employees, to firms with more than 1000 employees and found that for both men and women, hourly rates of pay were higher in larger firms. However, although larger firms may pay their employees on average higher wages, this may not necessarily be translated into a lower gender wage gap. Mitra (2003) finds that the gender wage gap between professional men and women in the US is the highest for firms with the smallest establishment size (1-25 employees), and interestingly second highest for firms with a very large establishment size (501 and more employees). Controlling for human capital variables and other characteristics, Mitra found that the gender wage gap for the ‘very large’ establishment size was 24 per cent, whereas for the ‘large’ establishment size (101-500 employees) it was 17 per cent. Further, recent findings from the 2008 EEH survey, show that as the firm size increases, the raw wage gap also increases (ABS 2008).
There have been several studies that have investigated the wage gap along the income distribution, rather than focusing on average values. Barón and Cobb-Clarke (2008), Kee (2006) and Miller (2005) note that the wage gap increases at the top end of the income distribution, although both Barón and Cobb-Clarke (2008) and Kee (2006) note that this effect is only evident in the private not the public sector. Authors also emphasise, however, that a substantial wage gap exists at all points of the income distribution, and that efforts to address the gap therefore need to be targeted at all income levels, not just high earners (Kee 2006; Miller 2005). In addition, Barón and Cobb-Clarke (2008) also find that for low-paid workers, the proportion of the wage gap explained by workers’ characteristics (endowments) is much larger than for higher paid workers. Further, Austin et al. (2008), in their analysis of low-paid Australian workers, find that the minimum wage helped to reduce the gender wage gap by approximately 1.2 per cent between 1995 and 2005.
Drawing on these Australian findings, as well as overseas studies, allows us to identify a set of variables to include in the decomposition of the gender wage gap being undertaken here. A large body of literature exists regarding theoretical and empirical considerations in selecting variables to predict the gender wage gap, and Grimshaw and Rubery (2002) note that as wage gap decomposition studies have developed, they have tended to include increasing numbers of variables (Grimshaw and Rubery 2002, p. 26). Drolet (2001, p. 11) notes that there is no ‘universally accepted’ set of variables that should be included in equations designed to assess gender earnings differentials. As noted above, Australian findings about many possible determinants have been mixed, and it is not possible from the literature to draw any definitive quantitative conclusions about the role of particular factors in maintaining the gender wage gap. In addition, many factors that influence wages and wage gaps overlap, and there is often the added difficulty of unmeasurable pre-labour market discrimination not being included in the model, and consequently under-estimating the overall effect of discrimination on gender wage gaps. Given this, it is important that the decomposition approach taken allows to some extent for these feedback effects to be included, and for direct discrimination to be measured.
- Other authors give these concepts different names. For example, Drolet (2001) refers to the explained portion of the wage gap (or endowments) as ‘productivity-related characteristics’ and the ‘rewards for endowments’ portion as compensations received for productivity-related characteristics (p. 13).
- Some studies additionally examine the differing contribution which rewards or endowments make in different sectors, or at different points of the income distribution (see, for example, Barón and Cobb-Clarke 2008; Kee 2006).
- Barón and Cobb-Clarke (2008) did find, however, that endowments, or wage-related characteristics, explained most of the wage gap at the very bottom end of the income distribution.
- Olsen and Walby (2004) and Walby and Olsen (2002) introduce an additional refinement to this method of measurement by using known educational attainment to develop a numeric index of education level. While this method has the advantage of being able to attempt a precise quantification of the contribution of education levels to the gender wage gap (e.g. for every extra year of education, so many dollars are gained), it does involve having to make substantial assumptions about the additive value of extra educational qualifications
- Miller (2005) notes that at the very bottom end of the income distribution, only the returns to education for women are actually slightly higher than for men.
- There are various approaches in the literature to attempting to better capture work experience. Preston (2003) uses the additional controls of marital status and numbers of children in an attempt to overcome the limitations of her otherwise standard labour market experience proxy variable. Grimshaw and Rubery (2002) note, however, that one of the problems with the inclusion of both numbers of children and proxied experience is the likelihood of correlation between these two variables. Australian studies have sometimes used tenure with current employer to capture work experience (Kee 2006), along with age and number/ages of children, or both tenure with current employer and tenure in current occupation combined with age and number of children (Cassells et al . 2008). Drolet (2001) notes that tenure with current employer can be considered a measure of on-the-job training, as employers are more likely to invest in longer-tenured employees. Both Kee (2006) and Cassells et al. (2008) use a set of dummy variables that capture whether the individual is employed on a full-time, part-time, contract or casual basis. Rummery (1992), with access to data about labour market history, uses a measure of years worked. Olsen and Walby (2004), in the context of their argument about the long-term effects on human capital attainment and wages of life-time working patterns, include three measures of work history in their gender wage gap models. They include a measure of years spent in part-time work and years spent in full-time work, and also capture additional information about women’s labour force withdrawals by including a variable which measures years out of employment due to caring for children or other family members.
- Grimshaw and Rubery (2002) note that the use of overly detailed occupational characteristics, while improving the explanatory power of models, can tend to obscure the role of discrimination in the wage gap (also noted by Drolet (2001), and thus relatively broad occupational categories are preferred.
4. Microeconomic Modelling of the gender wage gap in Australia
- 4.1 How to quantify the wage gap: decomposition approach
- 4.2 Key determinants of the wage gap in Australia: our findings
4.1 How to quantify the wage gap: decomposition approach10
As discussed in Section 3, feedback effects are an important factor to consider when attempting to estimate the impact of various factors on the gender wage gap. Pre-labour market discrimination is not taken into account when measuring wage gaps using traditional methodologies. These feedback effects result in differences in human capital levels, and can feed into the labour market through a variety of ways, including employer discriminatory practices resulting in lower human capital (for example, an employer may choose to invest less in training for particular people). Individual decisions to invest less in human capital due to lower returns, which are influenced by current discrimination practices, are also another form of feedback effects. These feedback effects can cause an under-estimation of labour market discrimination when using traditional decomposition methodologies.
In an effort to address these shortcomings, as well as to provide policy relevance, we have chosen to use the recent simulation approach pioneered by Olsen and Walby to identify and quantify the key determinants of the gender wage gap. Olsen and Walby’s approach involves ‘simulating the hypothetical changes needed to bring women’s levels….into line with those of men’ (Olsen and Walby 2004, p.24). In a hypothetical example, if the mean years of formal education for women were eight years and for men this was ten years, an increase of two years would be required in order to bring women’s years of education in line with the level of men. This extra two years of education is then multiplied by the corresponding coefficient (reward) for every extra year of education, which is, say 0.07 (seven per cent). This gives a simulated effect of 0.14 (0.07*2). This means that if women had the equivalent average level of formal education of men, their wage rate would increase by 0.14 (14 per cent). The Olsen-Walby technique is one of an ‘integration’ rather than ‘separation’ approach. To that effect, the simulation approach does not calculate separate ‘reward’ and ‘endowment’ components, but instead focuses on ‘hypothetically moving the market in ways that equalise men’s and women’s experiences’ (Olsen and Walby 2004, p.69). For a detailed description of this approach please refer to Appendix A.
The Olsen-Walby technique utilises the combined male and female prices for labour in their wage estimation in an effort to achieve a true non-discriminatory wage structure. Whilst there are limitations with this choice of wage structure, in that it may not be relevant in a ‘real world’ setting, and in that individual characteristics may be rewarded differently for men and women, (the effects of which may be neutralised within a combined labour market estimation), wage prices estimated for the entire pooled labour market offer a sensible middle ground11. One of the key advantages of the Olsen-Walby decomposition technique is that the gender component is illuminated and is not lost, or spread across factors, as is the case using other methodologies. This allows the direct discrimination component of being a woman to be measured or proxied. Limitations of the Olsen-Walby technique include those which are inherent in all regression estimation methodologies, including measurement error associated with chosen variables, and omitted variable bias.
Olsen and Walby also choose to exclude from their decomposition the effects of factors which are ’female-advantaging’, (those that help to decrease the wage gap), factors that do not change (hold the same value for men and women), and those that they consider to be controls only and not relevant to gender wage gaps, however relevant in estimating wages – for example, geographical variables (Olsen and Walby, 2004, p.63). Whilst removal of these components could be considered to produce biased estimates, or in fact an over-estimation of the wage gap, their justification is based on the argument that the components of the wage gap considered are ones with practical policy relevance to reduce gender wage gaps. We concede that within the Australian context, and for the purposes of this analysis, removal of such variables is warranted and have also taken this path12.
In summation, we see the advantages of the Olsen-Walby simulation technique, for our purposes in this study, being:
- The gender component is visible, enabling the effect of direct discrimination, or other aspects related to being a woman to be clearly measured.
- There is the option to bring all of the 'policy relevant' variables into the forefront, and to treat all other variables as controls or irrelevant.
- Offsetting 'female advantaging' aspects are removed
- The tug-of-war about what component is due to 'rewards' and what is due to 'endowments' is removed.
- Feedback effects (pre-labour market discrimination) are to some extent addressed by giving women the ‘best average situation among men’ (Olsen and Walby 2004, p.8).
4.2 Key determinants of the wage gap in Australia: our findings
The average characteristics of male and female wage earners in Australia, using data from HILDA, are shown in Table 1. The unadjusted wage gap shows men earning on average $3.13 more than women, with men receiving around $28 per hour on average and women around $25. This equates to an approximate 11 per cent hourly wage gap, which is different to the 17 per cent wage gap measured using AWE data14 . Many of the characteristics of male and female wage earners are very similar, including whether they are partnered, whether they have a long term health condition, their work schedule, union membership status and areas of residence; however there are some notable differences. Men have on average around two years more time in paid work than women, and 1.21 years more time in their current occupation, giving them an advantage in the attainment of human capital in the form of work experience. Female wage earners are more likely to have a bachelor degree or higher, when compared with males – 33 per cent and 27 per cent respectively. However, women are under-represented in vocational qualifications and over-represented in having a qualification of year 12 or below, with 39 per cent of men with a vocational qualification compared to 29 per cent of women and 38 per cent of women with Yr 12 qualification level or below, compared with 34 per cent of men.
There are considerable differences in the hours worked per week for men and women, with higher proportions of women working part-time hours (41 per cent), and higher proportions of males working 41-49 and 50 or more hours per week (49 per cent of men compared with 21 per cent of women). Similar proportions of men and women work between 35-40 hours per week - 40 and 38 per cent respectively. Whilst these differences are quite large, our findings indicate that working long hours does not necessarily mean that there will be higher rewards compared with those working fewer hours, further, working part-time hours does not feature as a liability compared to those working more hours per week (see Table B2 in Appendix B).
Our occupational segregation variable is created by calculating the proportion of men within a particular occupation (using data from the ABS Labour Force Survey) and multiplying this by ten in order to produce an index, showing the extent to which individuals are employed in male dominated occupations. It shows that on average, men are working in occupations which are dominated by men (61 per cent male), and that on average women are working in occupations which are female-dominated (44 per cent male) The industry segregation variable is calculated in a similar fashion and has similar results, showing that on average men are employed in industries that are 60 per cent male, and on average women work within industries that are 44 per cent male.
|Variable||Men (M)||Women (W)||Difference (M-W)|
|Hourly wage ($)||27.71||24.58||3.13|
|Time in paid work (years)||19.13||17.09||2.04|
|1-34 hours per week (%)||10||41||-0.31|
|35-40 hours per week (%)||40||38||2|
|41-49 hours per week (%)||21||11||10|
|50+ hours per week (%)||28||10||18|
|Children 0-4 (%)||16||11||5|
|Children 5-14 (%)||21||25||-4|
|Has long term health condition (%)||15||15||0|
|Bachelor or higher qualification (%)||27||33||-6|
|Vocational qualification (%)||39||29||10|
|Year 12 or lower qualification (%)||34||38||-4|
|Occupational segregation (male proportion x 10)||6.11||4.35||1.76|
|Industry segregation (male proportion x 10)||6.01||4.39||1.62|
|Tenure in current occupation (years)||9.91||8.7||1.21|
|Tenure with current employer (years)||7.25||6.41||0.84|
|Regular work schedule (%)||83||83||0|
|Firm size: Less than 20 employed (%)||37||34||3|
|Firm size: 20-100 employed (%)||28||33||-5|
|Firm size: 100-500 employed (%)||21||18||3|
|Firm size: 500+ employed (%)||15||14||0.01|
|In a union (%)||28||27||0.01|
|Private sector employee (%)||75||60||0.15|
|Public sector employee (%)||14||25||-11|
|Other sector employee (%)||10||15||-5|
|Lives in major urban area (%)||69||71||-2|
|Lives in other urban area (%)||2||18||2|
|Lives in rural area (%)||11||11||0|
Source: Authors’ calculations from HILDA, 2007, Wave 7 unit record data.
Wage gap determinants
Using HILDA data, (which, as noted earlier, includes part-time workers), Table 2 illustrates the simulation effect of moving Australian women to the average situation of Australian men. The largest effect is that of being a woman, followed by industry segregation and labour force experience. If the negative effects of being a woman, (which includes direct discrimination) were removed, the gender wage gap would decrease by 60 per cent, or an equivalent $1.87 per hour. For a woman who worked an average of 35 hours per week, this would equate to an additional $65 per week, and in a full year, this would mean an extra $3,394 in earnings. These findings about the size of the effect of being a woman are consistent with those from other Australian studies which, as noted earlier, have generally found that discrimination or other unexplained differences between men and women dominate explanations of the wage gap in this country.
If women were represented equally (to men) within industries, we would see a 25 per cent reduction in the gender wage gap, which for a 35 hour week would mean around an extra $27 for the average Australian woman. The hourly wage for women would increase by 22 cents per hour (seven per cent of the wage gap) if their labour market history was equivalent to that of men – this includes increasing their average time in paid work, time with current employer and time in current occupation.
Smaller effects exist for vocational qualifications and firm size, however these effects do combine to make up eight per cent of the differences in wages between men and women. If the representation of women with vocational qualifications increased to the same proportion as men, we would see the wage gap decrease by five per cent, or 15 cents per hour. Representation of women in larger firms to similar proportions as those of men would also have a positive impact on women’s wages, decreasing the wage gap by three per cent, which would mean an increase in hourly wages of 11 cents on average.
|Proportion of overall wage gap||Cents/hour equivalent||$/per 35 hour week||$/per year|
|Labour force history||7||22||7.78||404|
Note: The proportion of the overall wage gap is measured as the simulated change in the characteristics of women to that of the average situation of men multiplied by the reward or coefficient for that particular characteristic. Figures may not add to total due to rounding. The total gap has been derived using selected variables only. The wage gap of $3.13 per hour has been derived from the difference between the average wage of men and women for all wage earners.
Source: Authors’ calculations from HILDA, Wave 7 unit record data.
A graphical representation of the Olsen-Walby decomposition is provided below in Figure 2, where the effects of being a woman clearly dominate. If the negative effects associated with being a woman were removed, and equal representation of women within industries were to occur, this would see a decrease in the wage gap of 85 per cent. As mentioned earlier, these large negative effects associated with being a woman are likely to signify direct discrimination towards women, but can also include personal attributes that are unable to be observed with our dataset, such as motivation and drive. The effects of firm size are also evident in the chart, with larger firm sizes (measured by number of employees) having large associated returns (see Table B2 in Appendix B). Firm size may be acting as a proxy for a range of characteristics, and these are likely to include a number of features of larger organisations, including economies of scale, the likelihood of being a government organisation, and greater rates of union membership and centralised bargaining systems.
Industry segregation (the predominance of men working with men and women with women) has been shown in a number of Australian studies to be a key driver of gender wage gaps (Miller, 1994; Preston and Crockett, 1999, Wooden, 1999). Work undertaken in female dominated industries is often undervalued within society compared to male dominated industries (Pocock 1999). Our findings here suggest that if there were equal proportions of men and women within industries in Australia, the gender wage gap would reduce by around a quarter. This finding (along with all others) is dependent upon everything else remaining unchanged. That is, our quantification assumes that if we were to see equal representation of men and women throughout all industries, the reward associated with this would not change. However, studies have shown that this may not be the case, and as women move into particular industries and occupations, the monetary value that society places on these sectors may decline (Kongar 2008).
Note: The proportion of the overall wage gap is measured as the simulated change in the characteristics of women to that of the average situation of men multiplied by the reward or coefficient for that particular characteristic. Figures may not add to total due to rounding. The total gap has been derived using selected variables only. The wage gap of $3.13 per hour has been derived from the difference between the average wage of men and women for all wage earners.
Source: Authors’ calculations from HILDA, Wave 7 unit record data.
The policy implications of these findings are substantial but complex, and involve questions of equity and justice, as well as practical issues related to how, where and with what purpose policy interventions could take place. Using our findings about industrial segregation as an example, one question that would arise in addressing this issue at a policy level would be the extent to which improving women’s wages by aiming towards a more equal distribution of men and women across industries would be an appropriate response to these findings. Should the focus be on re-distributing genders between industries, or targetting industrial relations policy towards fairer outcomes for sectors that are female-dominated? Issues such as these are beyond the scope of this paper, but our findings should be considered and interpreted in the light of such complexities.
- There are several issues to take into account, when measuring gender wage gaps, including the choice of decomposition methodology, and choice of an appropriate non-discriminatory wage structure for which to compare wages. These issues are discussed further in Appendix A.
- See Appendix A for further discussion on the choice of a non-discriminatory wage structure.
- We have also estimated the effects of different components of the wage gap including all factors that influence or predict wages (both negative and positive) and find similar results as to the main drivers of the wage gap –these results are also shown in Appendix B.
- Details about the data, sample and model specification used to produce the results in this section are provided in Appendix B, along with detailed decomposition results.
- The wage gap calculated from the HILDA survey is different to that calculated using AWE for several reasons. Firstly, the HILDA wage gap is calculated at an hourly wage rate, and the AWE gap at a weekly wage rate. There are also differences in the samples used in the AWE and HILDA wage gap calculations. The AWE sample includes only adult, full-time ordinary time workers. The HILDA sample excludes people still at school, self-employed, those aged below 21 years, and those aged 65 years and above. We also excluded people with unusually low or high hourly wages.
- This table has been calculated using a sample of wage earners from HILDA, 2007. We excluded people who were self-employed people, people aged below 21 years, and those aged 65 years and above. We also excluded people with unusually low or high hourly wages.
5. Gender inequality and macroeconomic performance
- 5.1 Background
- 5.2 Measuring economic performance
- 5.3 The macroeconomic costs of the gender wage gap
The gender wage gap does not just impose a cost on individuals in terms of loss of income, but it also inhibits economic performance. Earlier sections of this paper have focused on the existence of a gender wage gap, and the possible drivers of this gap. In this section, we move on to considering the impact of the gender wage gap on economic performance. There is an increasing body of international research which focuses on the potential impacts of the gender wage gap (and gender inequality generally) on economic growth. Such work extends arguments about wage equality beyond issues of gender equity, and instead considers the implications of gender wage gaps for a society’s economic well-being.
Since the mid-1980s, feminist economists have been increasingly arguing for recognition of the importance of gender issues for the achievement of macroeconomic objectives (Cagatay et al. 1995). One key issue identified by such researchers has been the likelihood that gender differences in earnings (that is, the gender wage gap) can hinder an economy from achieving its macroeconomic objectives, and can be considered a cost to the economy (Cavalcanti and Tavares 2007).
Literature examining the impact of gender inequality on macroeconomic outcomes has more frequently focused on gender equality in education (e.g. Barro and Lee 1994; Barro and Sala-i-Martin 1995; Hill and King 1995; Klasen 1999; 2002; Dollar and Gatti 1999; Lagerlof 2003), and less often on the gender wage gap (see for example Seguino 2000; Cavalcanti and Tavares 2007; Caro 2008; Walby and Olsen 2002). The former body of literature contains important insights for developing methodology to measure macroeconomic impacts, and we have drawn on this in devising our modelling techniques for this study. The substantive findings from this work are also summarised in Appendix C.
Before presenting our results, we will discuss some issues related to concepts, measurement and definitions which are relevant to any study of gender inequality and its macroeconomic effects.
5.2 Measuring economic performance
While there are ongoing debates about how to measure economic performance, most research has used gross domestic product as the best measure of economic performance (Mankiw 1997). While GDP is widely accepted, it should nevertheless be noted that such a measure excludes the value of women’s activities at home – that is, unpaid work such as looking after a household and family (Klasen 1999; 2002), as these activities are not included in national accounts data. The exclusion of this unpaid work means that we are undervaluing women’s total economic contribution (Klasen 1999; 2002).
In this study we use GDP per capita as our primary measure of economic performance. GDP per capita is a better measure of the economic well-being of a country than total GDP, as it takes into account population size – very populous countries may have very large overall GDP, but when divided by population size the resultant GDP per capita figure will give a much clearer indication of the country’s comparable wealth.
Conceptualising the pathways toward economic growth
Various theoretical frameworks exist for explaining the ways in which economic growth comes about. There are different channels or ‘pathways’ that explain economic growth. The Australian Treasury has divided these channels into three main components: population, participation and productivity, otherwise known as the 3Ps. Treasury has focused on labour as the main input; and therefore productivity within the framework of the Treasury 3Ps can be defined as similar to labour productivity. To some extent, this understanding of productivity can be incorporated in the type of growth model we are using for our estimation in this study. Our modelling also includes measures of participation (the size of the labour force – using total number employed, and the number of hours worked) and changes in population (measured by fertility).
Walby and Olsen (2002) note the importance of the impact of the wage gap on productivity and hence on growth. Walby and Olsen (2002) used wages as a proxy measure of productivity in order to estimate the impact of gender discrimination on the wage gap (which, as wages were being used as a proxy for productivity, was equal to the productivity gap). Using a decomposition technique, they found that gender discrimination (being female woman) accounted for 29 per cent of the wage and productivity gap which in turn would affect the size of GDP and economic growth. However, Walby and Olsen (2002) did not quantify the impact of the productivity gap on the size of GDP or economic growth16 .
The relationship between gender inequality and economic performance will depend on the characteristics of the country or countries included in the sample; therefore it is important when comparing the literature and interpreting different results to also take into account differences in terms of a country’s stage of development and the way in which its economy operates. In particular, differences in labour market structures are important to consider. For example, Seguino (2000), found that the gender wage gap was a stimulus (rather than a hindrance) to GDP growth in export-oriented economies, in her study of semi-industrialised countries. However, this finding may only be relevant for semi-industrialised or developing countries that rely on manufacturing exports as a significant component of output. For semi-industrialised economies, her findings emphasise the importance of investment, and the productivity of investment for growth. She argues that investment is a function of profitability as well as output. Profitability is signalled as the share of income going to employees. Thus, a lower share of income going to employees indicates a higher profit and stimulus for investment. In the economies studied by Seguino, distinguished by highly gender-segregated export sectors where price elasticity of demand and profitability effects are large, this translated into the stimulation of investment (and therefore economic growth) by a widening wage gap.
On the other hand, another stream of literature, including Cavalcanti and Tavares (2007) and Caro (2008) posits that a higher gender wage gap (measured as female to male earning ratios) leads to lower output per capita indirectly through two channels – female labour force participation and fertility. The female labour force participation channel shows that as the gender wage gap decreases, women work more, due to the added wage incentive, and thus GDP per capita increases, (assuming that the population remains constant – and other things ceteris paribus).
In general, literature related to developed countries shows a positive relationship between wages and female labour force participation. Birch (2005) presented a comprehensive literature review on the labour supply of women in Australia and found that among other factors, increases in women’s wages would increase both female labour force participation and the number of hours worked. The positive association between lower wage gaps and female labour supply was also supported by the study of female labour participation for all OECD countries as discussed in Jaumotte (2003). Jaumotte (2003) found that the existence of a gender wage gap reduced female labour supply.
However, this positive relationship between higher wages and greater working hours may not always be the case. Interesting evidence about differences between single and married women’s behavioural responses to changes in the gender wage gap may refute the generalisation that changes in the gender wage gap will always have a positive relationship with female labour force participation. Using data for the United States, Jones et al. (2003) examined the phenomenon of why hours of work of married women increased so much when the gender wage gap reduced. They found that married women were more sensitive to changes in the gender wage gap than single women. They found that changes in the wage gap did not have large effects on single women, however for married women a small reduction in the gender wage gap generated a large labour supply response, which meant that their unpaid work at home was substituted for paid work. Recent evidence within the Australian context (Fitzpatrick and Lester 2009), has found the opposite effect for partnered women, where an increase in wages leads to a decrease in labour supply, however the authors do note that this effect is weak (Fitzpatrick and Lester 2009, p.43).
In relation to the effect of the gender wage gap on fertility, Cavalcanti and Tavares (2007) and Caro (2008) found that as the gender wage gap decreased, the opportunity cost of having children increased, therefore fertility is reduced, and consequently women’s hours of work are increased, which in turn increases GDP per capita. The size of this impact varies across studies. Cavalcanti and Tavares (2007) using a calibration cross-country analysis of both developed and developing countries, found that a 50 per cent increase in the gender wage gap caused a decrease of income per capita of a quarter of the original income per capita and an increase in hours worked at home by 65 per cent. It should be noted that this analysis did not take issues such as childcare availability and affordability into account. On the other hand, Caro (2008) in his Spanish study, also using calibration methodology, found that if the gender wage gap were completely eliminated, women’s hours of work would increase by 20 percentage points and GDP per capita would increase by 17.6 per cent.
5.3 The macroeconomic costs of the gender wage gap
5.3.1 The impact of the wage gap on economic growth
As noted above, we are measuring economic growth in this study as gross domestic product (GDP), focusing on GDP per capita. Detailed descriptions of a range of relevant modelling approaches, and their strengths and weaknesses, are provided in Appendix D. The model we have developed to estimate the effect of the gender wage gap on GDP per capita uses a type of regression analysis is described in detail in Appendix E, along with a discussion of our data sources, testing procedures and the model’s limitations. This model allows us to estimate the impact of the gender wage gap on growth, both directly and indirectly through the channels of investment, participation and fertility17.
Our estimations of the impact of the gender wage gap on GDP per capita are presented in Table 3 and show that, holding other variables constant (ceteris paribus), and including the effect of all the channels in our model, the average impact of the gender wage gap on GDP per capita is -0.507. This means that on average, a one percentage point increase in the gender wage gap would decrease economic growth by 0.507 per cent. This effect, however, should be treated as indicative only. It is an average figure, and is calculated (as shown in Table 3) by adding up all the effects of each of the modelled pathways. Not all these modelled pathways had a statistically significant effect on economic growth in our model18 , which is likely due to data limitations inherent within the model19 .
However, an examination of the individual pathways through which the gender wage gap affects economic growth does provide us with additional evidence about the substantial negative effects of the wage gap on growth. In particular, the impact on growth of average hours of work (the only statistically significant pathway in our model) is negative and substantial. If the gender wage gap increased by one percentage point, average hours of work would decrease, and thus GDP per capita would fall by 0.318 per cent (Table 3). This finding makes intuitive sense, as an increased gender wage gap (thus lower female wages compared to male wages) is likely to act as a disincentive for women to work more hours. This effect is likely to be particularly noticeable as so many Australian women work part-time20 . Reduced working hours in turn lead to lower economic growth. Conversely, these findings suggest that a decrease in the gender wage gap would be significantly associated with an increase in women’s hours of work. It should be noted, however, that women’s ability to work more hours is likely to be affected by a range of factors, including unemployment rates, the availability of appropriate and affordable child care and access to flexible working hours. Given the importance in our model of hours of work, these factors may be critical to consider if a narrowing of the gender wage gap is to be translated into greater economic growth.
While the other pathways in our model do not have statistically significant effects, each of the factors except labour participation has an effect on economic growth which is in the same direction as the (significant) effect of average hours of work, providing further support for the hypothesis that the wage gap affects growth negatively. However, the non-significance of these relationships means that we cannot draw any conclusions about underlying processes. However, we do note that the direction of the relationship between the gender wage gap and investment is the opposite to that found by Seguino (2000) in relation to semi-industrialised countries, suggesting that wage gap effects may operate on investment in different ways in developed compared with developing countries.
|Wage gap à economic growth||-0.250|
|Wage gap → investment → economic growth||-0.021|
|Wage gap → fertility → economic growth||-0.181|
|Wage gap → average hours of work → economic growth **||-0.318|
|Wage gap → labour participation → economic growth||0.263|
Note: ** significant at a confidence interval of 95 per cent. Economic growth refers to economic growth of GDP per capita. The results should be read as, for example, a one percentage point increase in the gender wage gap reduces economic growth of GDP per capita by -0.507 per cent.
Source: Authors’ calculations, using a variety of data sources. Please refer to Appendix E for more detail about the data sources and variables used in the macro model.
5.3.2 Quantifying the cost of the gender wage gap to the Australian economy
After examining the impacts above, we can translate these effects21 into dollar amounts to quantify how much the gender wage gap costs the Australian economy22 . All dollar amounts are recorded in 2007 dollars.
Table 4 shows the estimated cost of the gender wage gap to the Australian economy if we simulate an increase in the gender wage gap from its current position of 17 per cent by one percentage point to 18 per cent. Two conditions are shown, before (1) and after the wage gap increase (2).
|Current situation||Increase in the wage gap from 17 to 18 per cent||Cost to the economy|
|(1)||(2)||(1) – (2)|
|Gender wage gap||17%||18%||Na|
|GDP per capita||$51,114||$50,854||$260 per capita|
|GDP (millions)||$1,084,146||$1,078,649||$5,497 million|
Source: Authors’ calculations, using a variety of data sources. Please refer to Appendix E for more detail about the data sources and variables used in the macro model.
Table 4 shows that an increase in the gender wage gap of one percentage point is estimated to decrease GDP per capita by around $260, which equates to around $5,497 million of GDP in total, assuming the population is held constant at 21.21 million. Conversely, if the gender wage gap were to decrease by one percentage point, this would translate into an increase in GDP per capita of $260. This also means that if the whole gender wage gap were to be eliminated (that is, reduced from 17 per cent to zero), such a change could be worth around $93 billion of GDP.
5.3.3 Quantifying the key determinants of the wage gap at a macro level
Our findings at a microeconomic level (shown in Section 4) estimated that the key determinants of the gender wage gap in Australia stem from being a woman, industry segregation, women having less labour force experience than men, women having less vocational qualifications than men, and the differences in the size of organisations where women work compared to men. In our final piece of analysis presented here, we apply these determinants to the findings from our macroeconomic model in order to estimate the approximate contribution of each of these determinants to the total amount of GDP per capita lost through the wage gap. It is important to note that the findings presented in this section do have limitations, particularly in regards to the differences in data sources and variables being used to capture the wage gap in our microeconomic and macroeconomic models, as well as time periods for each calculation23 . However, bearing these limitations in mind, it is still possible to use our decomposition results as proxies for the key determinants of the gender wage gap in our macroeconomic model, and thus estimate the effect that each of these may have at a macroeconomic level.
Table 5 and Figure 3 show the disaggregation of the estimated $93.4 billion cost to the economy of the gender wag gap into the key determinants of the gender wage gap. We estimate that the effect of being a woman (direct discrimination and unobserved characteristics) is costing the economy over $55 billion per year in GDP. Industry segregation is contributing to around a quarter of the gender wage gap, which at an aggregate level, equates to over $23 billion of GDP. The labour force history component shows that if women had on average the same amount of time in paid work, tenure with their current employer and tenure in their current occupation that men do, we would see a reduction in the wage gap of seven per cent, which equates to a saving of $6.6 billion in GDP. If the representation of women with vocational qualifications increased to the same proportion as that of men, we would see a reduction in the wage gap of five per cent, which means an estimated gain of $4.5 billion of GDP. If women increased their representation in larger firm sizes to match that of the proportion of men, we would see a reduction in the gender wag gap of three per cent, which equates to around $3.2 billion in GDP.
6 Summary and Conclusion
This study has examined the gender wage gap in Australia from a number of perspectives, and one of its key contributions has been to estimate the dollar impact of the gender wage gap on Australia’s economic performance.
The major findings of this report can be summarised as follows:
- The gender wage gap in Australia has persisted over the past twenty years. The size of the gap depends on the data being used to measure the ratio of female to male wages, but our analysis of the ABS Average Weekly Earnings Survey shows a wage gap of 17 per cent in 2009. Over the period 1989 to 2009 the size of the wage gap has remained fairly constant, fluctuating slightly between 15 and 17 percent.
- Previous research on the wage gap in Australia has been quite extensive, but drawing firm conclusions about the key determinants of the wage gap in Australia from the literature is difficult due to the range of findings, and the wide variation in samples, methods and focus in earlier studies.
- It is clear from previous studies that a large proportion of the wage gap in Australia cannot be explained simply by the differing characteristics or endowments of men and women. In most studies, a very large majority of the wage gap is due to differences between the wages of men and women that cannot be explained by the sorts of variables which are included in gender wage gap models. These unexplained differences in wages may be due to direct discrimination, or to other unmeasured differences between men and women
- In order to quantify the contribution of particular variables to the wage gap in Australia, as well as provide policy relevance, we use a simulation methodology developed by Olsen and Walby. This approach moves the average characteristics of women to those of men, and quantifies how much the wage gap would be reduced and how much women’s wages would improve if this were to occur. This process is undertaken for variables considered to have relevance to policy and gender wage differentials, and all other variables are treated as controls.
- Using this simulation methodology we find that the key determinants of the gender wage gap in Australia include industrial segregation (25 per cent), labour force history (seven per cent), under-representation of women with vocational qualifications (five per cent) and under-representation of women in large firms (three per cent). However, being a woman accounts for a very large 60 per cent of the wage gap. This finding is in line with other literature and represents that part of the wage gap which is due to discrimination or other differences between men and women not captured by differences in their measured characteristics.
- We find that the gender wage gap has a substantial effect on Australia’s economic performance. If the wage gap were closed entirely (that is, if women earned the same as men) we estimate that GDP would grow by around $93 billion. Much of this effect on economic growth is brought about through the channel of hours of work, with a narrowing of the gender wage gap leading to significantly more hours of work, and thus to greater economic growth. As mentioned above, a woman’s ability to work more hours is determined by a number of factors including the availability of appropriate and affordable child care. Further, increased hours of work for women has other implications, including the ever increasing struggle to balance work and ‘life’, and the division of household labour and care arrangements between men and women.
- Using our decomposition findings, we can also tentatively estimate the effect of each key determinant of the gender wage gap at a macro level. In particular, removing the negative effects associated with being a woman (direct discrimination and unobserved characteristics), might see an improvement in GDP of around $56 billion per year.
It is important to note that these findings do have limitations. Both the microeconomic and macroeconomic modelling included in this report incorporate a range of assumptions, and have various limitations in relation to data availability (discussed in the appendices). Both areas of modelling are the subject of vigorous debate about the strengths of a range of methodological approaches. While our work is strongly informed by best practice, and has been subjected to rigorous quality assurance procedures, it should be noted that the use of different data and alternative methodological approaches could potentially produce different results.
The findings in this research report are of particular relevance to policy makers. There are several areas highlighted where policy may contribute meaningfully in helping to reduce the gender wage gap in Australia, and in turn enhance economic performance and national well-being. Additional research that focuses on the contribution of each industry separately to the gender wage gap would also help inform and direct policy efficiently.
Austin, S., Jefferson, T., Preston, A., and Seymour, R, (2008), ‘Gender Pay Differentials in Low-Paid Employment’, Report commissioned by the Australian Fair Pay Commission.
Australian Bureau of Statistics (ABS 2008), Employee Earnings and Hours, August, ABS Cat No. 6306.
Australian Bureau of Statistics (ABS 2009), Australian National Accounts: National Income, Expenditure and Product, ABS Cat No. 5206.0, March.
Barón, J.D. and Cobb-Clarke, D. A. (2008) Occupational Segregation and the gender wage gap in private- and public-sector employment: a distributional analysis. IZA DP No. 3562.
Barro, R., and Sala-I-Martin, X. (1995), Economic Growth. New York: McGraw-Hill.
Barro, R. J. (1990), ‘Government Spending in a Simple Model of Endogenous Growth.’, Journal of Political Economy, 98(5), pp. S103-125
Barro, R. J. (1991), ‘Economic Growth in a Cross Section of Countries.’ The Quarterly Journal of Economics, 106(2), pp. 407-443.
Barro, R. J. (1999), ‘Notes on Growth Accounting.’ Journal of Economic Growth, 4, pp. 119–137 (June 1999)
Barro, R. J. and J. W. Lee., (1994), ‘Sources of Economic Growth (with comments from Nancy Stokey),’ Carnegie-Rochester Conference Series on Public Policy, 40, pp. 1-57
Baum, C.L, (2002), ‘The Effect of Work Interruptions on Women’s Wages’, Labour 16(1): 1-36
Beblo, M., Beninger, D., Heinze, A. and Laisney, F., (2003), ‘Methodological Issues Related to the Analysis of Gender Gaps in Employment, Earnings and Career Progression’, Mannheim: European Commission, Employment and Social Affairs DG
Becker, G. (1975), Human Capital, 2nd ed. University of Chicago Press, Chicago, IL.
Becker, G.S., Murphy, K.M., and Tamura, R., (1990), ‘Human Capital, Fertility and Economic Growth.’ Journal of Political Economy, 98(5), pp. S12-37
Birch, E. (2005), ‘Studies of the Labour Supply of Australian Women: What Have We Learned?’. The Economic Record, vol.81, no. 252, pp. 65-84.
Blau, F.D. and Kahn, L.M. (1997) `Swimming Upstream: Trends in the Gender Wage Differential in the 1980s', Journal of Labour Economics, vol. 15, no.1, pp.1-42.
Blinder, A. S. (1973), ‘Wage Discrimination: Reduced Form and Structural Estimates’, Journal of Human Resources 18(4), pp. 436–455.
Blomstrom, M., R. E. Lipsey and M. Zejan. (1996), ‘Is Fixed Investment the Key to Economic Growth?’ Quarterly Journal of Economics, 111(1), pp. 269-276
Booth, A. and Wood, M., (2006), Back-to-front Down-under? Estimating the Part-time/Full-time Wage Differential over the Period 2001-2003, The Australian National University Centre for Economic Policy Research Discussion paper No.525
Borjas, G., (2010), Labor Economics 5th ed. McGraw Hill Irwin.
Borland, J. (1999), ‘The Equal Pay Case—Thirty Years On’, Australian Economic Review 32(3), pp. 265–72.
Budig, M.J. and England, P. (2001), The wage penalty for motherhood, American Sociological Review, Vol. 66, No. 2. (Apr., 2001), pp. 204-225.
Cagatay, N., Elson, D. and Grown, C. (1995). Introduction. World Development, Vol. 23, No. 11, pp. 1827-1836.
Cai, L. and Liu, A.Y.C (2008) Union wage effects in Australia: is there a variation along the distribution? The Economic Record, 84(267), 496-510
Caro, L.P. (2008), ‘A Macroeconomic Assessment of the Gender Wage Gap Effect: The Case of Spain’, Masters thesis in Economics, Univesidad del Pais Vasco
Caselli, F., Esquivel, G., and Lefort, F., (1996), ‘Reopening the Convergence Debate: A New Look at Cross-Country Growth Empirics.’ Journal of Economic Growth, 1(3), pp. 363-389
Cass, D. (1965), ‘Optimum Growth in an Aggregative Model of Capital Accumulation.’ Review of Economic Studies, 32, pp. 233–240
Cassells, R., Daly, A., and McNamara, J., (2008),’The Gender Wage Gap for Generation Y in Australia’, Paper presented at the Annual Conference of Economists, Gold Coast 2008
Cassells, R., Miranti, R., Nepal, B., and Tanton, B., (2009),’She works hard for the money: Australian women and the gender divide’, AMP NATSEM Income and Wealth Report Issue 22, April
Cavalcanti, T.V.d.V., and J. Tavares, (2007), ‘The Output Cost of Gender Discrimination: A Model-Based Macroeconomic Estimate’. Centre for Economic Policy Research, 6477, London.
Cobb-Clark, D.A., and Tan, M., (2009), ‘Noncognitive Skills, Occupational Attainment, and Relative Wages’, paper presented at the HILDA conference, 2009
Cotton, J. (1988), ‘On the Decomposition of Wage Differentials’, The Review of Economics and Statistics 70, pp. 236–243
Daly, A., Kawaguchi, A., Meng, X., and Mumford, K., (2006),’The Gender Wage Gap in Four Countries: an update of Bob Gregory’s contribution’, Economic Record, vol. 82, Issue 257, pp.165-176
Denison, E. F., (1962), ‘The Sources of Economic Growth in the United States and the Alternatives Before Us’, Washington, DC: Committee for Economic Development.
Dickey, D. A. and Fuller, W. A. (1979), ‘Distribution of the Estimators for Autoregressive Time Series With a Unit Root’, Journal of the American Statistical Association 74(366), pp. 427-431
Dollar, D. and Gatti, R., (1999), ‘Gender Inequality, Income and Growth: Are Good Times Good for Women?’. Gender and Development Working Paper Series, no. 1, The World Bank, Development Research Group/Poverty Reduction and Economic Management Network.
Drolet, M., (2001), ‘The persistent gap: New evidence on the Canadian gender wage gap’, Business and Labour market analysis division, Statistics Canada
Durlauf, S. N., Johnson, P.A., and Temple, J.R.W., (2004). ‘Growth Econometrics’ Vassar College Economic Working Paper No. 61
Durlauf, S. N. and Quah, D.T., (1999), ‘The New Empirics of Economic Growth.’ In J. B. Taylor and M. Woodford, eds., Handbook of Macroeconomics (Volume 1A), Amsterdam: Elsevier, pp. 235–308
Easterly, W. and Levine, R., (2001), ‘It's Not Factor Accumulation: Stylized Facts and Growth Models.’ World Bank Economic Review, 15(2), pp. 177-219
Eastough, K. and Miller, P. W. (2004), `The Gender Wage Gap in Paid- and Self-Employment in Australia', Australian Economic Papers, 43, pp 257-76
Enders, W., (1995), ‘Applied Econometric Time Series’, Wiley, New York.
Engle, R.F., and Granger, C.W.J., (1987), ‘Cointegration and error-correction: representation, estimation, and testing.’ Econometrica 55, pp. 251-276
Fitzpatrick, D. and Lester, L., (2009), ‘Dynamics of Female Labour Supply: Evidence of Differences in Preference between Single and Couple Females’, Paper presented at the 2009 HILDA Survey Research Conference, 16th and 17th July 2009, the University of Melbourne.
Folster, S. and Henrekson, M. (1999), ‘Growth and the Public Sector: a Critique of the Critics.’ European Journal of Political Economy, 15(2), pp. 337-358
Fortin, N., (2008), ‘The gender wage gap among young adults in the United States’, The Journal of Human Resources, Vol 43, No.4, 884-918.
Galor, O., and Weil, D.N., (1996), ‘The Gender Gap, Fertility and Growth’. The American Economic Review, vol.86, no. 3, pp. 374-387.
NJ: Princeton University Press)
Ginther, D. K. and Hayes., (2003), Gender Differences in Salary and Promotion for Faculty in the Humanities, 1977-1995. Journal of Human Resources, Vol 38, No. 1, 34 – 73.
Grimshaw, D. and Rubery, J. (2002),’The adjusted gender pay gap: a critical appraisal of standard decomposition techniques’, EWERC Manchester School of Management.
Healy, J., Kidd, M., Richardson, S. (2009), ‘Gender pay differentials in the low-paid labour market’, National Institute of Labour Studies, Flinders University, Adelaide, Report commissioned research by the Australian Fair Pay Commission
Heckman, J. J. (1979), ‘Sample Selection Bias as a Specification Error’, Econometrica 47(1), pp. 153–161.
Hill, M.A. and King, E.M., (1995), ‘Women's Education and Economic Well-Being’, Feminist Economics, vol.1, no. 2, pp. 21-46.
Jaumotte, F. (2003), ‘Labour Force participation of Women: Empirical Evidence on the Role of Policy and Other Determinants in OECD Countries’. OECD Economic Studies, No. 37
Jones, C., (1995), ‘Time Series Tests of Endogenous Growth Models,’ Quarterly Journal of Economics, 110(2), pp. 495-525.
Jones, L.E., Manuelli, R.E. and McGrattan, E.R. (2003), Why Are Married Women Working So Much? Federal Reserve Bank of Minneapolis Research Department Staff Report 317.
Studies 34, pp. 249–280.
Juhn, C., Murphy, K.M, and Pierce, B. (1993),’’Wage inequality and the rise in returns to skill’, Journal of Political Economy, vol 101, no. 3, pp. 410-442.
Workers and their wages: changing patterns in the United States, edited by Marvin H. Kosters. Washington: University Press of America, Inc.
Kee, H. J., (2006), ‘Glass Ceiling or Sticky Floor? Exploring the Australian Gender Pay Gap’, The Economic Record 82(259), pp. 408–427
Kidd, M.P. and Shannon, M. (2001), ‘Convergence in the gender wage gap in Australia over the 1980s: identifying the role of counteracting forces via the Juhn, Murphy and Pierce decomposition’, Applied Economics, vol.33, pp. 929-936.
Klasen, S. (1999), ‘Does Gender Inequality Reduce Growth and Development? Evidence from Cross-Country Regressions’. Policy and Research Report on Gender and Development Working Paper, No. 7.
Education on Economic Development. The World Bank Economic Review, 16(3), pp. 345-373.
Kongar, E, (2008), ‘Is Deindustrialization good for women? Evidence from the United States’, Feminist Economics, 14(1) January, 73-92
Koopmans, T. C. (1965), ‘On the Concept of Optimal Economic Growth.’ In Study Week on the Econometric Approach to Development Planning, chap. 4, Amsterdam: North-Holland Publishing Co., pp. 225–287.
Kunze, A. (2008), ‘Gender wage gap studies: consistency and decomposition’, Empir Econ, no.35, pp.63-76.
Kwiatkowski, D., Phillips, P. C. B., Schmidt, P. and Shin, Y. (1992). `Testing for the Null of Stationarity Against the Alternative of a Unit Root', Journal of Econometrics 54, 159-178
Lagerlof, N. (2003), ‘Gender Equality and Long-Run Growth’. Journal of Economic Growth, vol.8, pp. 403-426.
Levine, R. and D. Renelt. (1992), ‘A Sensitivity Analysis of Cross-Country Growth Regressions.’ American Economic Review, 82(4), pp. 942-963.
Lucas, R. E. (1988), ‘On the Mechanics of Economic Development.’ Journal of Monetary Economics, 22, pp. 2-42.
Lundberg, S. and Rose, E. (2000), Parenthood and the earnings of married men and women, Labour Economics, vol.7, no.6, pp. 689-710.
Mankiw, N. G. (1997), Macroeconomics, 3rd ed., New York: Worth Publisher, Inc.
Mankiw, N. G., D. Romer, and D. Weil. (1992), ‘A Contribution to the Empirics of Economic Growth.’ Quarterly Journal of Economics, 107(2), pp. 407-437
Mikkola, A. (2005), ‘Role of Gender Equality in Development - A Literature Review’. Helsinki Center of Economic Research Discussion Papers, 84
Miller, P. W. (1994), ‘Occupational segregation and wages in Australia’, Economics Letters, 45, pp. 367–371.
Miller, P.W. (2005), ‘The Role of Gender among Low-Paid and High-Paid Workers’, The Australian Economic Review 38(4), pp. 405–17.
Mitra, A. (2003), Establishment size, employment and the gender wage gap, Journal of Socio-Economics, vol. 32, 317-330.
Neumark, D. (1988), ‘Employers’ Discriminatory Behaviour and the Estimation of Wage Discrimination’, Journal of Human Resources, 23, pp. 279–295.
Oaxaca, R. (1973), ‘Male-Female Wage Differentials in Urban Labor Markets’, International Economic Review 14(3), pp. 693–709.
Oaxaca, R.L. and Ransom, M.R. (1999),”Identification in detailed wage gap decompositions’, Review of Economics and Statistics, vol. 81, No.1, pp. 154-57
Oaxaca, R. L. and Ransom, M. R. (1994), ‘On discrimination and the decomposition of wage differentials’, Journal of Econometrics, 61, pp. 5–21
Oaxaca, R. L. and Ransom, M. R. (1988), ‘Searching for the Effect of Unionism on the Wages of Union and Nonunion Workers’, University of Arizona, Tuscon.
Phillips, A.W. (1957), ‘Stabilization policy and the time forms of lagged responses,’ Economic Journal, 67, pp. 265—277.
Pocock, B., (1999), ‘Equal Pay Thirty Years On: The policy and practice’, The Australian Economic Review, Vol.32, No.3, pp 279-85.
Preston, A.C. and Crockett, G., (1999), ‘The Gender Earnings Gap in Australia: Learning from State Comparisons’, Discussion Paper Series 99/4, School of Economics and Finance, Curtin University Women's Economic Policy Analysis Unit (WEPAU)
Preston, A.C. (2000) Female earnings in Australia: an analysis of 1991 Census data, Australian Bulletin of Labour, 26(1), 38-58.
Preston, A.C. (2003) Gender earnings and part-time pay in Australia, 1990-1998. British Journal of Industrial Relations, 41(3), 417-433.
Preston, A.C. and Burgess, J., (2003), ‘Women's Work in Australia: Trends, Issues and Prospects', Australian Journal of Labour Economics, Vol. 6, No.4, December2003, pp 497-518
Preston, A.C. and Whitehouse G., (2004), ‘Gender Differences in Occupation of Employment within Australia’, Australian Journal of Labour Economics, Vol.7, No.3, pp 309-327
Quisumbing, A. (1996).’Male-Female Differences in Agricultural Productivity: Methodological Issues and Empirical Evidence’. World Development, vol.24, pp.1579-95/
Ramsey, F. (1928), ‘A Mathematical Theory of Saving.’ Economic Journal, 38(152), pp. 543–559.
Reimers, C. W. (1983), ‘Labor Market Discrimination Against Hispanic and Black Men’, The Review of Economics and Statistics, 65, pp. 570–579.
Working Paper Series No. 8.
Rodgers, J.R. (2004), Hourly Wages of Full-Time and Part-Time Employees in Australia, Australian Journal of Labour Economics, 7(2), 231-354.
Romer, D. (2001), Advanced Macroeconomics, New York: MacGraw-Hill
Romer, P. M. (1986), ‘Increasing Returns and Long-Run Growth.’ The Journal of Political Economy, 94(5), pp. 1002-1037.
Romer, P. M. (1990), ‘Endogenous Technological Change.’ Journal of Political Economy, 98(5), pp. S71-102.
Rummery, S, (1992), ‘The contribution of intermittent labour force participation to the gender wage differential’, The Economic Record, Vol, 68. No.202, pp 351-364
Sala-I-Martin, X. (1997), ‘I Just Ran Two Million Regressions.’ AEA Papers and Proceedings, 87, pp. 178-183.
Seguino, S. (2000), ‘Gender Inequality and Economic Growth: A Cross-Country Analysis’. World Development, vol.28, no. 7, pp. 1211-1230.
Silber, J., and Weber, M., (1999), ‘Labour market discrimination: are there significant differences between the various decomposition procedures?’ Applied Economics, 31, pp. 359-365
Sigle-Rushton, W. and Waldfogel, J. (2006), Motherhood and women’s earnings in Anglo-American, Continental European, and Nordic Countries, Luxembourg Income Study Working paper Series, no. 454.
Solow, R. M. (1956), ‘A Contribution to the Theory of Economic Growth.’ Quarterly Journal of Economics, 70(1), pp. 65-94.
Solow, R. M. (1957), ‘Technical Change and the Aggregate Production Function.’ The Review of Economics and Statistics, 39(3), pp. 312-320
Solow, R. M. (2001), ‘What have we learned from a decade of empirical research on growth? Applying Growth Theory across Country.’ The World Bank Economic Review, 15(2), pp. 283-288
Swan, T. W. (1956), ‘Economic Growth and Capital Accumulation.’ Economic Record, 32(2), pp. 334-61
Walby, S., and Olsen, W., (2002),’The impact of women’s position in the labour market on pay and implications for UK productivity’, Report to Women and Equality Unit, November 2002.
Watson, I, (2009), ‘The gender wage gap within the managerial workforce: an investigation using Australian panel data’, paper presented at the 2009 HILDA Survey Research Conference, The University of Melbourne, 17 July.
Watts, M., (2003), ‘The Evolution of Occupational Gender Segregation in Australia: Measurement and Interpretation’, Australian Journal of Labour Economics, Vol. 6. No.4, pp 631-655
Wooden, M. (1999), ‘Gender Pay Equity and Comparable Worth in Australia: A Reassessment’, The Australian Economic Review 32(2), pp. 157–71.
World Bank (2001), Engendering Development: Through Gender Equality in Rights, Resources and Voice.
Yun, M-S. (2009), ‘Wage differentials, discrimination and inequality: A cautionary note on the Juhn, Murphy and Pierce Decomposition Method’, Scottish Journal of Political Economy, vol. 56, No. 1, pp. 114-122
Listen to this pageAppendices A-E: The impact of a sustained gender wage gap on the Australian economy
- Appendix A
- Appendix B
- Appendix C
- Appendix D
- Appendix E
- Extensions of Oaxaca-Blinder: Choice of weighting matrix
- Simulation technique or ‘Gross Decomposition’ approach
- Juhn-Murphy-Pierce Technique
Decomposing the wage gap adjusts for the overall level of differences between men and women in the labour market. (Walby and Olsen, 2002, p.104) Theoretically, it enables the wage gap to be compartmentalised into that portion which can be attributed to differences in characteristics of individuals, and a portion that is unexplained by these differences and could indicate discrimination1. Identification of components that are affecting the wage gap considerably can direct policy appropriately.
As noted in the main body of the report, there are several standard methodologies that have been developed in order to decompose the underlying components of wage gaps, each with benefits and limitations. As with any empirical analysis, the estimation technique used depends upon assumptions made about particular relationships. In this case, these assumptions are about how labour markets within Australia operate and how individual and organisational characteristics affect wages.
In this appendix, we discuss three widely used decomposition techniques – the Oaxaca-Blinder, Juhn-Murphy-Pierce and similar extensions of the Oaxaca-Blinder technique. We also discuss the relatively new simulation approach used by Walby and Olsen (2002, 2004) that is used in this report to gauge the underlying components of the gender wage gap.
1) Oaxaca-Blinder technique
Ronald Oaxaca and Alan Blinder produced two very similar estimation techniques to decompose the gender wage gap, almost simultaneously in 1973 (Oaxaca 1973; Blinder 1973). These two autonomous studies remain as the foundation of wage gap studies today, and they represent one of the most common decomposition techniques used in studies of wage inequality. The earnings regressions used for these decompositions are based on earlier work by Mincer (1974). Mincer developed a model of earnings differentials based on endowments of human capital, where individual earnings principally depended on endowments of formal education and work experience.
The purpose of Oaxaca’s well-known 1973 study of the U.S. labour market was to separate or ‘decompose’ the differences in men and women’s wages into those parts that can be explained by differences in human capital and those parts that are due to discrimination. There are several key assumptions that underlie Oaxaca’s decomposition; the most notable is the assumption that individual characteristics are a measure of marginal productivity, and marginal productivity equates to wages (Oaxaca 1973, p.695).
Under the Oaxaca-Blinder methodology, earnings are estimated for separately for each gender using the following equation:
where is the natural log of the wage for individual i; is an intercept term or constant; is a vector of regressors capturing the individual characteristics expected to impact on wages, such as education and experience; and is a residual term.
Following Oaxaca (1973), the gender wage differential can then be decomposed into that part attributable to differences in human capital endowments (the difference in the mean values or Xs) and that part attributable to differences in rewards for those endowments (the βs):
The first term on the right hand side of the equation measures the differences in the mean characteristics or endowments between men and women, and here is weighted by the estimated price for male endowments. These differences are often thought to be legitimate or ‘explained’ sources of the wage gap, however as expanded on below, these differences could also reflect pre-labour market choices or discrimination. The second part of the equation is the ‘unexplained’ or unjustifiable component of the wage gap, and is weighted by the mean of X. If, on average, the price of male labour exceeds the price of female labour, then there exists a male advantage.
The allocation of the gap into ‘explained’, and ‘unexplained’ components will depend on the choice of coefficients (or weights) used for the decomposition (βm or βf, where m and f indicate male and female respectively). Oaxaca assumes that rewards to endowments are interchangeable, that is, without discrimination, men could be paid as women are and women could be paid as men are, however he implies that the true wage structure lies somewhere between the female and male wage structures, and applies both in his calculations (Oaxaca 1973). The wage structure is defined as a set of prices for various labour market skills for both measured and unmeasured characteristics and the rents received for employment in particular sectors of the economy (Blau and Kahn 1997). The majority of applications of the Oaxaca-Blinder decomposition technique use the male coefficients as the weighting mechanism, with the assumption that the male wage structure is the current non-discriminatory price. Arguments for maintaining the male wage structure as the non-discriminatory price of labour are ones with tones of legality, where it is reasoned that in legal proceedings related to gender discrimination, it is the male wage structure that would be used as the non-discriminatory wage when determining compensation (Oaxaca and Ransom 1994; Ginther and Hayes 2003; Healy et al. 2009).
The use of either the female or male weight in determining the amount of the wage gap that is attributed to differences in endowments is one of the major criticisms of the Oaxaca-Blinder technique and there have been several extensions of this work in order to achieve a more appropriate non-discriminatory wage structure. These extensions are reviewed further below.
One of the benefits of the Oaxaca-Blinder technique is that it is relatively easy to understand, has its foundations based in a solid neo-classical human capital framework, and enables identification and quantification of the components of the wage gap that can be explained by differences in endowments and those that relate to differences in prices for these endowments. However, one of the key disadvantages of the Oaxaca-Blinder technique is that the unexplained component of the gender wage gap cannot be separated out into factors related to direct discrimination (that is, being a woman), and factors that are related to other attributes such as motivation, drive and initiative, that are difficult to quantify and include. Estimating separate wage equations for men and women is also a shortcoming, in that a direct comparison between men and women cannot be made. Also, as the Oaxaca-Blinder technique is based upon the neo-classical assumption that wages are equal to productivity, if all characteristics related to productivity (e.g. education, experience, motivation, initiative) cannot be included in the model due to data limitations, then these ‘left-over’ elements of productivity are included in the ‘unexplained’ component of the wage gap, along with a discrimination component, which may lead to an over-estimation of the unexplained component. Over the years, more and more explanatory variables have been included in wage gap analysis in order to try and overcome this shortcoming, and in the event that particular important variables, such as work experience, were not available, proxy indicators would be substituted in order to achieve a more fully specified model. Further, the model can also be limited by measurement error within the variables chosen to represent productivity. For example, not having the level of data needed to ascertain a true endowment effect may mean that the results are inaccurate.
Another critical assumptions that is inherent in the Oaxaca-Blinder methodology, and one that again stems from neo-classical economic theory, is the assumption that there exists individual choice (or freedoms), associated with gaining human capital factors, such as education, work experience and on the job training (Grimshaw et al. 2002 p. 4; Olsen and Walby 2002). Oaxaca notes the limitation of his model in that it ‘does not take into account the effects of the feedback from labour market discrimination on the male-female differences in the selected individual characteristics’, (Oaxaca 1973, p.708). The differences in human capital levels could be due to employer discriminatory practices, or the individual choice to invest less in human capital due to lower returns, which is itself influenced by current discrimination (Grimshaw et al. 2002, p.27). Grimshaw argues that these feedback effects result in an under-estimation of labour market discrimination (Grimshaw et al. 2002, p.27).
In addition, there is also an argument that because the wage gap is estimated around the mean value of the total population this is not giving a true estimation, and the Oaxaca-Blinder methodology has been extended to quantile wage regression analysis, where the wage distribution is separated into income groups. This has also been used in order to examine the ‘glass ceiling’ and ‘sticky floor’ effects that may exist within the labour market (see, for example, Kee 2006).
Overall, the Oaxaca-Blinder decomposition technique has four major disadvantages:
- Unsatisfactory choice of a true non-discriminatory wage structure.
- Feedback effects will mean that discrimination is under-estimated.
- Women and men cannot be compared directly due to separate wage estimations.
- Discrimination may be blurred with omitted variable bias, making it impossible to truly separate discrimination from other factors.
The advantages of the Oaxaca-Blinder methodology are that:
- Separate effects of endowments and prices can be calculated and quantified.
- Separate coefficients for men and women can be measured for each endowment.
Extensions of Oaxaca-Blinder: Choice of weighting matrix
There have been several variations of the Oaxaca-Blinder technique that seek to overcome the problem of gaining an appropriate non-discriminatory wage structure or weighting matrix (see for example Oaxaca and Ransom 1988, 1994; Neumark 1988; Cotton 1988; Reimers 1983). Each of these involves calculating a weighting matrix that is more reflective of the non-discriminatory price for wages.
Oaxaca and Ransom (1994) extend the original equation (A2), by introducing a third term which includes the estimation of the non-discriminatory wage structure for a pooled labour market. The Oaxaca-Ransom equation is as follows:
Where β* is the estimated non-discriminatory wage structure, and the other notations are the same as the values given above in equation (A2). The first term on the right hand side of the equation estimates male wage advantage, the second term is female wage disadvantage, and the third term is an estimate of the productivity differential (Oaxaca and Ransom 1994, p.8). One of the underlying concepts of this equation is that, in wage discrimination, there not only exists a group that is disadvantaged (receiving below the non-discriminatory price), but there also exists a group that is advantaged (receiving above the non-discriminatory price). In the Oaxaca-Blinder technique, there is only ever one or the other, consequently, the result will either be under or over estimated.
The main difference between the Oaxaca-Ransom, Neumark, Cotton and Reimers approaches is in the way that they derive their weighting matrix. Oaxaca-Ransom and Neumark have both used the coefficients derived from the entire population in order to estimate their matrix, Cotton uses labour force weights (the share of the majority group), and Reimers chooses a weighting mechanism that lies directly between the two groups (see Silber and Weber (1999), Oaxaca and Ransom (1994) and Beblo et al. (2003) for comparisons).
Each of these weighting matrixes has its benefits and limitations. Cotton’s approach gives a greater weight to the dominant labour group in the working population, with the assumption that the ‘nondiscriminatory wage structure will be closer to the current white wage structure than to the current black wage structure’ (Cotton 1988, p.239). Cotton’s estimation of the non-discriminatory wage rate is given by:
Where and are the proportions of males and females (males and females have been substituted for white and blacks in this equation for relevance to the issue at hand).
Reimers estimates the non-discriminatory wage structure more simplistically, using a division of the two estimated male and female wage structures and is denoted by:
Oaxaca and Ransom and Neumark use the pooled labour market in order to ascertain the non-discriminatory wage structure, and is shown by:
Where = is a weighting matrix and is a diagonal unit matrix (identity matrix as explained by Oaxaca and Ransom 1994, p. 12).
All three approaches have similarities; in fact Oaxaca and Ransom argue that Cotton’s weighting technique can give the same result as the one they have proposed, and Reimer’s is a special case of Cotton’s methodology (Oaxaca and Ransom 1994). One of the disadvantages of both the Reimers and Cotton approaches is that the non-discriminatory wage is bounded or ‘bracketed’ by the separate male and female estimations, whereas the Oaxaca-Ransom methodology is not (Oaxaca and Ransom 1994, p.18). In the Oaxaca and Ransom (1994) comparison of the four alternative weighting matrices, they conclude that using a single gender wage structure is “too extreme”, and that the pooled methodology gave the lowest standard errors. (Oaxaca and Ransom 1994, p.18). However, they do note that despite the empirical superiority of the decomposition technique, in a real world situation, it would be the male wage that would be taken as the non-discriminatory wage in wage discrimination litigation proceedings. However, within the Australian context, legislative changes in Queensland and New South Wales have moved away from using the male wage rate as the benchmark in industrial disputes related to wages, focusing more on the value of the work.
Silber and Weber (1999) compare five decomposition procedures (Oaxaca-Blinder, Neumark, Reimers, Cotton and Oaxaca-Ransom) using bootstrap techniques and find no significant differences between each technique. They do note, however, that the Neumark technique places a higher weight to the endowment effect and a lower weight to the remuneration effect when compared with the other four approaches (Silber and Weber 1999, p.365).
Overall, regardless of the weighting matrix used, combining the wage structures of men and women in an effort to achieve a true non-discriminatory wage structure offers a sensible middle ground in overcoming the shortcomings of using a single sex wage structure.
An alternative technique to the Oaxaca-Blinder (1973) technique is the Juhn-Murphy-Pierce (1991) decomposition (for simplicity we refer to this as JMP), which is often used for comparisons over time or between countries. The main contribution of this technique is that it includes the whole wage distribution, rather than just the mean, in order to allow for changes in the overall gender wage gap. Juhn et al. (1991) analyse trends in the slowdown in black and white wage convergence to examine the effects of wage inequality on the black and white wage gap. The core advantage of the JMP decomposition over Oaxaca-Blinder (1973) is that it allows the decomposition of changes in the residual of the gender wage gap into “price” and “quantity or gap” effects (this is explained further below). We will start from the Oaxaca-Blinder (1973) decomposition as discussed in Yun (2009). Suppose we compare the wages of men and women and decompose their (mean) wage differentials, using this standard decomposition equation (Yun 2009, p.16):
where m and f refer to males and females respectively, is the log of wages, is the vector of explanatory variables, are vectors of coefficients and is the residual. The ’over bar’ is the value of the sample average.
Similar to equation (A2), the first term on the right hand side of equation (A7) measures the “characteristic effect”, the second term measures the coefficient effect, and the third term represents the residual effect. The coefficient effect is often interpreted as the discrimination effect, whilst the residual effect measures differences in the unobserved characteristics distribution. When equation (7) is estimated using the OLS (Ordinary Least Squares) technique, the residual term collapses to 0 since both are equal to 0 (Yun 2009). Juhn et al. (1991) argue that including the residual effect is important as it represents unobserved characteristics such as motivation and drive. However, if OLS estimation is used, these unobserved characteristics cannot be estimated. Therefore, Juhn et al. (1991) propose a decomposition equation which enables the examination of the effect of unobserved skills to be measured when using OLS to estimate wages. Following Blau and Kahn (1997 pp. 6-7), who decompose the gender wage gap using the JMP methodology, the male wage equation for worker m in year t can be expressed as:
where is the natural log of wages for males, is the vector of explanatory variables for males, are vectors of coefficients for males, is the standardised residual that captures unobserved skills of males (Kunze 2008), with a mean of zero and variance of one for each year, while is the standard deviations of the residual for male wages for that year (to measure the inequality amongst the residual of the male wage).
Blau and Kahn (1997) assume that returns to observed characteristics are similar for males (m) and females (f) and thus the non-discriminatory wage structure is given by.
So, under the assumption that prices from the male sample wage regression are equivalent to competitive prices and therefore discrimination to observed characteristics is ignored (Kunze 2008). Given this, the gender wage gap for year is written as follows:
Equation (A9) shows that the gender wage gap can be decomposed into a component due to gender differences in observable characteristics weighted by the male coefficients, and a component due to gender differences in the standardised residual of the male equation (unobserved variables such as motivation) multiplied by the money value per unit difference in the standardised residuals. The second term on the right hand side of equation A9 ( ), is identical to the unexplained or discrimination effect in the standard Oaxaca-Blinder (1973) decomposition.
Blau and Kahn (1997) use the JMP decomposition to examine changes in the gender wage gap in the USA over time. Therefore, the difference in the gender wage gap between two years (between years t and t+1) can be decomposed using equation (A10) below. For simplicity, we drop the prefix m, therefore in the equations below, refers to, whilst refers to.
The discussion about each component of the JMP decomposition below closely follows Blau and Kahn (1997, p.7).
The first term, is the endowment effect which measures the impact of the change in differences in endowments such as the level of educational attainment and years of work experience.
The second term measures the effect of changing prices or rewards for the observed labour market characteristics of males. For example, if women have lower mean work experience levels in comparison to men, and there is an increasing return to male work experience, this will put some weight on women’s lower experience levels and raise the gender wage gap.
The third term measures the gap/quantity effect or changes in the levels of the unobservable characteristics (Kunze 2008, p.71). This will reflect the contribution to the change in the gender wage gap, if the percentile rankings of the residuals of the female estimated wage have changed, while the distribution/inequality amongst the residuals of the male wage remain constant.
Finally, the fourth term measures the unobserved price effect or changes in the returns to unobservable characteristics (Kunze 2008, p.71) and measures the impact of a change in inequality on the change in the male-female wage differential, assuming that women maintain the same position in the male residual wage distribution. This will measure the unobserved deficits in relative characteristics or discrimination that lower women’s position in the male residual wage distribution. For example, Kidd and Shannon (2001) seek to calculate the unobserved price effect of the gender wage gap in Australia for two different periods of time: 1981-82 and 1989-90. They hold the percentile position of the mean female residual fixed at its 1981-82 position and the unobserved price effect is calculated as the difference in the value of the male residual at this percentile between 1981-82 and 1989-90.
Grimshaw and Rubery (2002, pp. 10) note that ’the gender-specific factors are reflected in the first and the third terms’, which are the effect of gender differences in endowments and gender differences in wage rankings at a given observed characteristic. In contrast, the effects of a country’s labour market structure are reflected in the second and fourth terms, which measure the effect of changes in returns to observed and unobserved characteristics (Blau and Kahn 1997).
Overall, in comparison to the Oaxaca-Blinder decomposition, the JMP decomposition has several advantages, however, there are also drawbacks related to the suitability of JMP as it relies on several strong assumptions which are considered difficult to verify (Yun 2009).
The advantages of the JMP decomposition methodology are as follows:
- Enables accurate estimates of the wage gap over-time or between countries.
- Includes and allows for the decomposition of changes in the residual into price and quantity effects, thus allowing for the relative importance of gender-specific-factors and the wage structure in the narrowing/widening of the gender wage gap (Blau and Kahn 1997).
- Minimises the problem of sample selection bias, which is strongly present in the Oaxaca-Blinder technique, by avoiding the need to make separate estimates of wage equations for women (Juhn et al. 1991). Sample selection bias usually occurs because female labour force participation changes over time and women who participate (or do not participate) in the labour market have specific characteristics. For example, women with lower levels of human capital often choose not to participate in the labour market (Grimshaw and Rubery 2002). In this case, women who participate in the labour market may not be representative of the female population (Grimshaw and Rubery 2002).
The drawbacks of the JMP decomposition are as follows:
- The assumption that there are changes in the distribution of male wage residuals may not necessarily be always the case. Changes in the distribution of male wage residuals also capture changes in sample composition and measurement error (Blau and Kahn 1997).
- The use of the male equation as the benchmark implies that the set of prices that influence women are similar to those that influence men, which then implicitly leads to the assumption that there are similar factors that raise wage inequality of both men and women (Blau and Kahn 1997).
In addition, the use of the JMP decomposition is not likely to be the best choice if the aim is to examine the impact of discrimination on the gender wage gap. Although, within a traditional framework of wage decomposition, the sum of the third and fourth terms represents changes in the unexplained differential which are commonly used as an estimate of discrimination Blau and Kahn (1997) acknowledge the complexity of the interpretation if the focus is on discrimination. In the Oaxaca-Blinder methodology discrimination is embedded in only one component of the equation, whilst in the JMP decomposition, discrimination is divided between the third and fourth components. Juhn et al. (1991) note that changes in the residual gap are a combination of changes in the distribution of unobserved characteristics and changes in discrimination. Although the second term of the Oaxaca-Blinder decomposition is also not perfect in explaining discrimination, Yun (2009) claims that researchers may need to choose between the use of the Oaxaca-Blinder decomposition and JMP decomposition depending on the needs of the researchers and whether the focus of the analysis is on discrimination or the price of unobserved skills. Yun (2009) suggests that the Oaxaca-Blinder decomposition is better to use for capturing discrimination in comparison to the JMP decomposition.
Further, the JMP decomposition would be more relevant if researchers are attempting to examine changes in the gender wage gap over time (for example, Juhn et al. 1991 and 1993; Blau and Kahn 1997; Kidd and Shannon 2001) or across countries (for example, Rice 1999; Blau and Kahn 2001). Kidd and Shannon (2001) examine the gender wage gap in Australia during the 1980s and find that around 44 to 80 per cent of the gender wage gap (depending on the specification) is explained by observed characteristics (human capital endowment of experience and education). The gap/quantity effect shows the improvement of women’s position within the male wage residual distribution which is related to a combination of women’s unobservable characteristics or a decline in the level of discrimination due to policy changes. In contrast, Kidd and Shannon found that the unobserved price effect had little effect on the gender wage gap over this period.
Simulation technique or ‘Gross Decomposition’ approach
Olsen and Walby (Olsen and Walby 2004; Walby and Olsen 2002) propose a different approach in order to gain a more indicative measure of the underlying components affecting the gender wage gap, which involves ‘simulating the hypothetical changes needed to bring women’s levels of these components into line with those of men’ (Walby 2004, p.24). Walby and Olsen (2002, 2004) refer to their approach as a ‘gross decomposition’, as the slope coefficients are derived from an overall regression (both men and women are included), and label the Oaxaca-Blinder methodology as a ‘net decomposition’ as it uses ‘net components’, or male and female slope coefficients estimated separately (Walby and Olsen 2002, p.104).
This technique is relatively new and not widely used in wage gap decompositions. Recently, Watson (2009) has applied the Olsen-Walby technique in his decomposition of the gender wage gap for full-time managers in Australia.
From Walby and Olsen (2002), the decomposition is calculated using the formula:
Where, ∆Xί is the hypothetical change in female mean values required to make them equal with the male mean values (or the difference in the means), and is the slope coefficient estimated for men and women jointly2. In a hypothetical example, if the mean years of formal education for women is eight years and for men it is ten years, an increase of two years is required in order to bring women’s years of education in line with men’s. This extra two years of education is then multiplied by the corresponding coefficient (reward) for every extra year of education, which is, say 0.07 (seven per cent). This gives a simulated effect of 0.14 = 0.07*2. This means, that if women had the equivalent average level of formal education, their wage rate would increase by 0.14 (14 per cent). This formula is carried out for every component or variable that is used to predict wages, and the simulated effects are then summed to find the overall wage gap, or, the proportion that would shift women’s wages to be equal to men’s wages. One of the key differences between the Olsen-Walby technique and the Oaxaca versions and the JMP approach is that it is one of an ‘integration’ rather than ‘separation’. The simulation approach does not calculate separate ‘reward’ and ‘endowment’ components, but instead focuses on ‘hypothetically moving the market in ways that equalise men’s and women’s experiences’ (Olsen and Walby 2004, p. 69). Further, the gender component is illuminated with the Olsen-Walby approach, and is not lost, or spread across factors as in the other methodologies (Walby and Olsen 2002, p.104). This allows a direct discrimination component to be measured or proxied.
Olsen and Walby also choose to exclude from their decomposition the effects of factors which are ’female-advantaging’ (those that help to decrease the wage gap), factors that do not change, and those that they consider to be controls only and not relevant to gender wage gaps, however relevant in estimating wages – for example, geographical variables. (Olsen and Walby 2004, p.63). Whilst removal of these components could be considered to produce biased estimates, or in fact an over-estimation of the wage gap, their justification for this approach is that the variables considered are ones with policy relevance to the gender wage gap.
In summation, the advantages of the Olsen-Walby simulation technique are as follows:
- The gender component is visible enabling the effect of direct discrimination or other aspects related to being a woman to be measured.
- There is the option to bring all of the 'policy relevant' variables into the forefront, and to treat all other variables as controls or irrelevant.
- Offsetting 'female advantaging' aspects are removed
- The tug-of-war about what component is due to 'rewards' and what is due to 'endowments' is removed.
- Feedback effects (pre-labour market discrimination) are to some extent addressed by giving women the ‘best average situation among men’ (Olsen and Walby 2004, p.8).
Some of the disadvantages of the Olsen-Walby technique include those discussed above in the Oaxaca-Blinder section, including measurement error associated with chosen variables, and omitted variable bias. Further, the removal of particular factors related to the wage gap if they are deemed ‘controls’ or policy irrelevant, could be considered as a shortcoming of this approach.
- Throughout the literature, a number of terms are used to describe differences in the characteristics of individuals and differences in the rewards for these characteristics. Individual characteristics are often referred to as ‘endowments’, ‘levels’, ‘characteristics’, ‘human capital’ or ‘productivity differentials’. Rewards for these characteristics are also referred to as ‘prices’, ‘remuneration’, ‘wage structure’,’ non-discriminatory wage structure’, ‘coefficients’ or ‘discrimination’.
- Walby and Olsen (2002) argue that using the slope coefficient for the whole population is effectively using a weighted average somewhere between βm and βf, and therefore superior to choosing either men’s or women’s separately. (Walby and Olsen 2002, p.106)
Microeconomic model specification, variable descriptions and detailed decomposition results
- Model specification
- Table B1 Olsen-Walby detailed simulated decomposition of the gender wage gap
- Table B2 OLS regression for log hourly earnings, 2007
- Table B3 Olsen-Walby detailed simulated decomposition of the gender wage gap – all variables
We have chosen to use the Olsen-Walby decomposition technique in order to develop the most appropriate estimations of the key determinants of the gender wage gap. In doing so, we must first measure the effects of factors that are known to best predict wages, and which will include the main drivers of the gender wage gap. In doing this, we apply a standard statistical technique – regression analysis, which allows the impact of each factor that affects wages to be quantified. All variables are included in our regression analysis, however we then go on to decompose the wage gap using only variables that are relevant to wage differentials. Following Olsen and Walby (2004) we have included only those variables which meet the following criteria:
- they have a non-trivial, positive and statistically significant impact on the gender wage gap; and
- they are policy-relevant (although many factors would require substantial further research to inform the development of suitable policy responses).
Selecting these variables provides a practical way of quantifying the effects of the key determinants of the wage gap. It should be noted that this approach means that variables which are female-advantaging are not included in this figure, nor are a range of variables which have very small effects on the wage gap and which do not have clear policy relevance. Consider, for example, our set of hours of work variables. Only one of these variables – working 41 to 49 hours per week, has a positive effect on the wage gap, and this effect is very small. If we decompose the effect of this variable, we find that if as many women as men were to work 41-49 hours per week, two per cent of the gender wage gap would be explained or eliminated. However, this is not an area that is likely to be the subject of policy intervention. While some additional hours of work per week might have a modest effect on the ratio of women’s wages to men’s, the intensification of work in this way is widely regarded as being associated with a range of other possibly negative outcomes, for individuals and societies, and is unlikely to be the subject of policy intervention.
Data source and sample
We have used the most recent wave of the Household, Income and Labour Dynamics in Australia (HILDA) survey - 2007, in order to gain the best estimates of factors affecting the gender wage gap. The HILDA survey is a longitudinal survey, and has been conducted annually since 2001. The sample is designed in a way to best represent the entire Australian population, and survey weights are calculated in order to facilitate this. In 2007, there were 12,789 persons and 7,063 households in the sample.
For our sample of wage earners, we decided to exclude several groups of people in order to gain the most accurate measure of the wage gap – this is a common practice in gender wage gap studies, although decisions about which groups to include or exclude differ across studies. We excluded people still at school, self-employed people, people aged below 21 years, and those aged 65 years and above. We also excluded people with unusually low or high hourly wages3 . Our final sample of wage earners is 6,137 people, of whom 3,073 were women and 3,064 men. When this sample is weighted, it represents 7,951,654 people, or 3,764,055 women and 4,187,599 men.
One of the advantages of using the HILDA data set is that it has a wide range of variables appropriate for estimating wages, including important work history variables such as the number of years in paid work, that are not available in other surveys such as the EEH or EEBTUM.
In many empirical analyses it is important that samples which are used to examine relationships between variables are unbiased, meaning that those included in the sample should be there randomly, rather than due to any other influencing factor. In our situation, using a sample of wage earners only, it is likely that what is called a ‘selection effect’ is present. This means that sample membership is not random, but rather is influenced by a range of factors, some of which will also be related to our outcome variable (which here is wages). This is particularly the case for women wage earners – previous research has shown that women with lower human capital will be more likely to opt out of the labour force (and thus not be in our sample), thereby introducing a bias in our estimations (Heckman 1979). This situation has been rectified by the use of a selection model, which predicts the likelihood of persons being employed and ‘corrects’ for this within our estimation, thereby eliminating selectivity bias.
As well as including variables capturing the most likely determinants of the wage gap, our model also includes a range of control variables, in line with those generally used in other Australian research. Additional control variables generally consist of variables that are likely to affect wages, but are not believed to be major drivers of the wage gap. These include marital status (found to have a positive effect on men’s wages, but a negative effect on women’s (Miller 2005)), and region (Kee 2006; Miller 2004; Miller 2005; Preston 2003; Wooden 1999). In Australian studies, regional control variables used include metropolitan versus non-metropolitan (Preston 2003; Wooden 1999), state and remoteness region (Kee 2006). We have also added control variables taking into account a person’s health status and whether they have children aged four and under and/or children aged between five and 14.
Industry segregation has been measured following Olsen and Walby (Olsen and Walby, 2004). We have used the Australian and New Zealand Standard Industrial Classification (ANZSIC), 1-digit industry classification (the broadest level) in order to measure the level of industry segregation a person is experiencing (see ABS Cat No. 1292.0 for further details of the classification). This has been measured by calculating the proportion of males within a major industry grouping and multiplying this proportion by ten in order to gain an index of industry segregation. Data from the ABS Labour Force survey, at a similar time period to when HILDA 2007 was conducted (August, 2007), has been used in order to gain more accurate results, as the labour force survey has a larger sample size and specialises in these variables (see ABS Cat No. 6291.0.55.001 for further details). The level of industry segregation an individual was experiencing, measured by the index, was then applied to the base dataset.
Occupational segregation has also been measured following Olsen and Walby (Olsen and Walby 2004). We have used the Australian and New Zealand Standard Classification of Occupations (ANZSCO), 1-digit occupation classification (the broadest level) in order to measure the level of occupational segregation a person is experiencing (see ABS Cat No. 122.0 for further details of the classification). This has been measured by calculating the proportion of males within a major occupation grouping and multiplying this proportion by ten in order to gain an index of occupational segregation. Data from the ABS Labour Force survey, at a similar time period to when HILDA 2007 was conducted (August, 2007), has been used in order to gain more accurate results, as the labour force survey has a larger sample size and specialises in these variables (see ABS Cat No. 6291.0.55.001 for further details). The level of occupational segregation an individual was experiencing, measured by the index, was then applied to the base dataset.
The number of employees in a firm has been disaggregated into four categories – less than 20 employees, 20-100 employees, 100-500 employees and 500+ employees. We have used those employed in firms with less than 20 employees as the reference group in our regression estimation.
Persons with children aged four or under have been included in our regression analysis as a dichotomous variable, as well as persons with children aged five to 14 years.
A person is considered to be partnered if they identify as being legally married or in a defacto relationship. This variable is dichotomous.
We have divided the qualification field into three categories, those with a Bachelor or above (including graduate diploma, masters and doctorate), those with vocational training (including those with certificates, diplomas and associate diplomas) and those with year 12 or below. We have chosen those with year 12 or below as the reference group in our regression estimation.
Long term health condition
A person is considered to have a long term health condition if they identify as having a condition that has or is likely to last six months or more, restricts everyday activity and cannot be corrected by medication or medical aids (HILDA, Living in Australia Showcard set, Wave 7). Such conditions include sight problems not corrected by glasses or contact lenses, hearing problems, speech problems, blackouts, fits or loss of consciousness and any other long-term condition such as arthritis, asthma, heart disease, Alzheimer’s disease and dementia (please refer to HILDA documentation for a full list).
Hours of work
We have chosen to create categorical variables from the continuous hours of work variable available in HILDA. There is discussion in the literature about the possible effects of variations in hours of work (especially unpaid overtime hours) on observed wages (Preston 2003), and some studies control for this with a set of dummy variables (Preston 2003; Wooden 1999). Through evaluation of the distribution of hours worked, and following standard definitions of part-time and full-time work in Australia, we have chosen four categories of comparison – working 1-34 hours per week, 35-40 hours per week, 41-49 hours per week and 50+ hours per week. We have chosen those that work 35-40 hours per week as the reference group.
Tenure in occupation and with current employer
Tenure in occupation and with current employer are continuous variables that integrate weeks and years into a year unit of measurement.
Union membership is a dichotomous variable.
Sector of employment
Sector of employment has been divided into three groups – public, private and other. The other grouping consists of other government business enterprise or commercial statutory authority, other commercial, private sector not for profit organisation, and other non-commercial organisations.
The type of work schedule an employee is working has been divided into ‘regular’ and ‘irregular’, with the regular classification including those working a regular daytime schedule, evening shift or night shift. Irregular work schedules include a rotating shift (changes from days to evenings to nights), a split shift (two distinct periods each day) and being on call.
HILDA allocates the geographic residence of persons using the Australian Standard Geographic Classification (ASGC, ABS Cat. No. 1216.0). The classification we have chosen to use is that of ‘Section of State’, which divides locations into ‘major urban’, ‘other urban’, ‘bounded locality’ and ‘rural balance’. We have combined those persons living in bounded localities and the rural balance to form a rural group.
|Men||Women||Change factor||Pooled coefficient||Simulation effect||Simulated change as a % of the wage gap||Cents/hour equivalent||$/35 hour week|
|Time in paid work (years)||19.13||17.09||2.04||0.0024242||0.004945368||0.0288024574||9.02||3.16|
|Tenure in occupation (years)||9.91||8.7||1.21||0.0055275||0.006688275||0.0389533713||12.19||4.27|
|Tenure in current employment (years)||7.25||6.41||0.84||0.000669||0.00056196||0.0032729271||1.02||0.36|
|Vocational qualification (%)||0.39||0.29||0.1||0.0823698||0.00823698||0.0479732278||15.02||5.26|
|Industry segregation (male proportion x 10)||6.01||4.39||1.62||0.0266268||0.043135416||0.2512261945||78.63||27.52|
|Firm size: 100-500 employed||0.21||0.18||0.03||0.1302226||0.003906678||0.0227529937||7.12||2.49|
|Firm size: 500+ employed||0.15||0.14||0.01||0.1935139||0.001935139||0.0112704977||3.53||1.23|
Note: Pooled coefficient refers to the estimated return for the entire group of wage earners for each additional unit of a particular variable, or in comparison to the reference group in the model. The reference group for firm size are those employed in a firm with less than 20 employees, and for educational qualifications, it is those with year 12 or below. All other variables are continuous or dichotomous.
Source: Authors’ calculations from HILDA, Wave 7unit record data.
|Coefficient||Std. Err.||t||P>t||[95% Confidence Interval]|
|Time in paid work (years)||0.0024242||0.0004856||4.99||0||0.0014723||0.003376|
|1-34 hours per week (reference: 41-49 hours)||0.0282771||0.0136763||2.07||0.039||0.0014667||0.0550874|
|35-40 hours per week (reference: 41-49 hours)||0.0337767||0.0147377||2.29||0.022||0.0048856||0.0626678|
|41-49 hours per week (reference: 41-49 hours)||-0.0626823||0.0140833||-4.45||0||-0.0902907||-0.035074|
|Has long term health condition||-0.0478118||0.0140783||-3.4||0.001||-0.0754102||-0.0202134|
|Bachelor and above (reference: yr12 or below)||0.3092289||0.0133484||23.17||0||0.2830613||0.3353966|
|Vocational qualification (reference: yr12 or below)||0.0823698||0.0120043||6.86||0||0.0588371||0.1059025|
|Tenure in current occupation||0.0055275||0.0006759||8.18||0||0.0042026||0.0068524|
|Tenure with current employer||0.000669||0.0008444||0.79||0.428||-0.0009863||0.0023243|
|Regular work schedule||-0.0116553||0.0135952||-0.86||0.391||-0.0383066||0.0149961|
|Firm size: 20-100 (reference: <20 employed)||0.0559333||0.0126326||4.43||0||0.031169||0.0806977|
|Firm size: 100-500 (reference: <20 employed)||0.1302226||0.0145037||8.98||0||0.1017903||0.158655|
|Firm size: 500+ (reference: <20 employed)||0.1935139||0.0166336||11.63||0||0.1609062||0.2261216|
|In a union||0.0288412||0.01253||2.3||0.021||0.0042779||0.0534045|
|Public sector employee (reference: private)||0.0510677||0.0154618||3.3||0.001||0.020757||0.0813783|
|Other organisation employee (reference: private)||0.003912||0.0162786||0.24||0.81||-0.0279998||0.0358237|
|Lives in other urban area (reference: urban)||-0.0613304||0.0130876||-4.69||0||-0.0869866||-0.0356741|
|Lives in rural area (reference: urban)||-0.061574||0.0164713||-3.74||0||-0.0938636||-0.0292844|
Dependent var = lnhrlywage, N=6116, R-squared=0.2234, Adjusted R-squared=0.2203
Source: Authors’ calculations from HILDA, Wave 7 unit record data.
|Men||Women||Change factor||Pooled coefficient||Simulation effect||Simulated change as a % of the wage gap||Cents/hour equivalent|
|Time in paid work – years||19.13||17.09||2.04||0.0024242||0.004945368||0.05||15.72|
|1-34 hours per week||0.1||0.41||-0.31||0.0282771||-0.008765901||-0.09||-27.87|
|41-49 hours per week||0.21||0.11||0.1||0.0337767||0.00337767||0.03||10.74|
|50+ hours per week||0.28||0.1||0.18||-0.0626823||-0.011282814||-0.11||-35.87|
|Has children under 4||0.16||0.11||0.05||0.0694577||0.003472885||0.04||11.04|
|Has children 5-14||0.21||0.25||-0.04||0.0075368||-0.000301472||0.00||-0.96|
|Has long term health condition||0.15||0.15||0||-0.0478118||0||0.00||0.00|
|Occupation segregation (male proportion x 10)||6.11||4.35||0.65||-0.0084968||-0.00552292||-0.06||-17.56|
|Industry segregation (male proportion x 10)||6.01||4.39||0.61||0.0266268||0.016242348||0.16||51.63|
|Tenure in occupation (years)||9.91||8.7||1.21||0.0055275||0.006688275||0.07||21.26|
|Tenure in current employment (years)||7.25||6.41||0.84||0.000669||0.00056196||0.01||1.79|
|Regular work schedule||0.83||0.83||0||-0.0116553||0||0.00||0.00|
|Firm size: 20-100 employed||0.28||0.33||-0.05||0.0559333||-0.002796665||-0.03||-8.89|
|Firm size: 100-500 employed||0.21||0.18||0.03||0.1302226||0.003906678||0.04||12.42|
|Firm size: 500+ employed||0.15||0.14||0.01||0.1935139||0.001935139||0.02||6.15|
|Public sector employee||0.14||0.25||-0.11||0.0510677||-0.005617447||-0.06||-17.86|
|Other sector employee||0.1||0.15||-0.05||0.003912||-0.0001956||0.00||-0.62|
|Other urban area||0.2||0.18||0.02||-0.0613304||-0.001226608||-0.01||-3.90|
|Sum of components||0.098463553||1.00||313.00|
Source: Authors’ calculations from HILDA Wave 7 unit record data.
- A small amount of people appeared to be receiving unusual hourly wages of under $5 per hour and over $260 per hour. We have chosen to remove these people from our sample, as these hourly rates are believed to be inaccurate.
Gender inequality in education: impact on economic performance
The World Bank (2001) undertook a comprehensive literature review on the relationship between gender and development, with a particular focus on gender inequality in education. The study claimed that conditions such as female illiteracy and low female education can inhibit productivity and earnings not only for women but also for whole economies. For example, in rural economies in Kenya, the lower yields of female farmers reflect lower levels of inputs or education relative to male farmers, thus increasing the education and input levels of female relative to male farmers can increase farming output as a whole (Quisumbing 1996 as discussed in World Bank 2001).
A number of other authors (Klasen 1999; 2002; Hill and King 1995; Dollar and Gatti 1999) have recognised the direct influences of gender inequality in education on economic performance (i.e. higher female-male ratios of total years of schooling are associated with higher economic growth).
Hill and King (1995) explored the impact of gender inequality in education (measured by female/male enrolment ratio) on Gross National Product (GNP). Their results showed that large gender disparities in educational attainment were associated with lower levels of GNP. These finding illustrated that in those countries where the ratio of female to male enrolments was less than 0.75, GNP was 25 per cent lower than those countries in which the ratio of female to male enrolments was close to one (greater than 0.95).
Dollar and Gatti (1999) applied a similar analysis when analysing the effect of education on growth of income per capita in developed countries, using female and male secondary educational achievement as the explanatory variable. They found that gender inequality in education lowered economic growth in developed countries: a one percentage point increase in the share of adult women with secondary school education increased income per capita by 0.3 percentage points in developed countries. However, their findings in relation to the effects of education for men and women on economic growth were mixed. When looking at secondary educational attainment, they found that female secondary education attainment is positively and significantly associated with growth in developed countries only, while male secondary education attainment is not significantly associated with growth in either developed or developing countries. Klasen (1999; 2002) however, suggests that these findings were possibly due to high multicollinearity between female and male educational attainment.
Klasen (2002) found that gender inequality in education has impacts on economic growth both in developing and industrialised countries. Differences between Klasen’s (2002) and Dollar and Gatti’s (1999) findings, Klasen argues, may be due to the shorter time period used by Dollar and Gatti (1999) (1975-90) in comparison to Klasen (2002) who used 1960-1992 and also to their use of a different education variable, with Dollar and Gatti (1999) using a secondary educational attainment measure in contrast to Klasen (2002) who used total years of schooling, a more comprehensive measure of education.
In addition to the direct channels through which gender inequality in education may affect economic growth, Klasen (1999; 2002) measured other channels through which gender inequality might affect economic growth. He found that higher investment rates are related to higher productivity of the labour force and economic growth of income per capita. Reduced gender inequality in education is also related to higher investment rates. Klasen (2002) argued that a combination between lowering gender inequality and wage discrimination would increase investment as it would be profitable to invest in female-intensive industries.
Lagerlof (2003) further adds to our understanding of the channels through which gender inequality may affect economic growth by noting the possible impacts of human capital investment on the next generation. Using an overlapping generation framework applied to European data, Lagerlof (2003) found that a decrease in gender educational inequality induced a quantity-quality of children substitution which then created a fall in fertility, thus increasing human capital investment in children and increasing incomes per capita for the next generation. The results also showed that at the same time, increasing levels of human capital reduced mortality, inducing population growth to rise simultaneously with per capita income growth (GDP per capita). However, as the mortality level stabilises, and fertility continues to fall, population growth starts to fall, and per capita income also continues to rise and stabilise on a balanced growth path. Thus in the long run, Lagerlof finds that reducing gender inequality in education increases economic growth.
The development of growth theory and its application to studies of gender inequality
- Solow Growth Model
- Growth accounting
- The endogeneity of growth
- Methodology used in previous empirical studies on the impact of gender inequality on economic performance
This appendix provides a history of the development of growth theory (originating from work by Solow and Swan in 1956), explaining the underlying concepts of economic growth which are embedded in this approach, and particular insights provided by endogenous growth theory. We discuss the complexities of these types of models and provide details of the ways in which previous studies of the macroeconomic impact of gender inequality have utilised growth theory.
The basic growth model, developed by Solow in 1956, is based upon how the flow of economic output, represented by GDP, is produced by two stocks of inputs - capital and labour in a production function. This basic model has been developed over the decades to include the impact of other aspects of the economy, such as human capital and financial development, on economic output. This impact upon economic output can be considered as the change in the two inputs (capital and labour), or the change in productivity, also known as total factor productivity (TFP).
The analytical framework of growth theory went through a boom after Solow and Swan launched their theories of growth in 1956. These neoclassical theories of growth assume one capital input, assuming a closed and single production function. Both the Solow and Swan models conclude that in the long run the growth of an economy will not be determined by aspects of the economy, with factors such as government policy having no long run impact. This type of system is referred to as an exogenous growth process. Starting with Romer (1986), Lucas (1988) and Barro (1990), the study of growth continued to expand, and frameworks were modified by including additional, previously un-measured types of capital input and developing several production functions for different types of output. This expansion and modification incorporated an understanding of economic growth as an endogenous process, and this understanding of the endogeneity of growth transformed the nature of the analytical framework. The endogeneity of growth reflects the theory that economic growth is determined by a range of variables that represent aspects of the economy. This increased interest in more complex economic relationships in predicting growth triggered a body of empirical work in search of the “real” sources of economic growth.
Initially, endogenous growth theory produced a very complicated growth model that could not be tested empirically (Sala-I-Martin 1997). Barro (1991) pioneered growth regression developments in order to produce an endogenous growth model that could be applied empirically, abandoning the structure of the theoretical model and using one linear model to ascertain which variables actually determined growth according to the data. Barro’s (1991) work was followed by other empirical studies, examining a larger number of variables to seek the “real” determinants of growth (Mankiw et al. 1992; Levine and Renelt 1992; Sala-I-Martin 1997).
In the next section we describe and discuss the two most common frameworks that are used in growth analysis and how these are used to measure the impact of economic factors on economic growth. These two frameworks are (i) the Solow growth model4 and (ii) growth accounting. We discuss the merits and limitations of each of these growth frameworks, provide additional detail about the development of methodologies based on growth theory and discuss the application of these techniques to studies of the impact of gender inequality on economic performance. As noted in the main report, we used a modified version of the Solow growth model in our calculations of the impact of the gender wage gap on economic growth.
Solow Growth Model
The Solow growth model uses a Cobb-Douglas production function that assumes constant returns to scale. The Cobb-Douglas production function can be described as
where Y is output, K is physical capital, L is labour and A is an index of labour augmenting technology, while (AL) is defined as the number of effective units of labour (Mankiw et al. 1992)5. The assumption of constant returns to scale means if all production factors are increased in the same proportion, output will also increase by that proportion. This implies diminishing marginal returns for factors since the increase in output reduces if only one factor of production is increased.
In the capital accumulation process, labour and technology are assumed to grow at exogenous constant rates (that is, n and g in equation D2 below). Meanwhile, the accumulation of physical capital comes from the fraction of output invested in physical capital, which is constant and known as savings, s. In calculating the growth of physical capital, the model has also taken the rate of depreciation (d) into account, to be subtracted from output. Therefore, the accumulation of capital per effective unit of labour can be described as:
where k=K/AL, which is the stock of capital per effective unit of labour and y=Y/AL, whichis output per effective unit of labour.
The capital accumulation process combined with the production function that assumes constant return to scale results in a steady state condition. This means that the return to output as a result of additional capital has been diminished and consequently, so has the ability to accumulate capital from the diminished output. This steady state condition has become the basis of the Solow model and predicts that increased savings and population growth are the main determinants of long run economic output per capita in the form of:
The Solow neoclassical growth model has several critics. Most have focused on the inconsistency between the prediction in the model and the stylized facts. For example, the importance of the capital accumulation process as the only growth determinant (reflecting the exogenous growth theory that underlies this model) has attracted much criticism. Romer (2001) and Easterly and Levine (2001) point out that factors other than capital accumulation are responsible for income growth (i.e., endogeneity of growth), and point particularly to the role of technological progress.
An alternative theoretical approach to modelling economic growth is known as growth accounting, which originated in Solow’s (1957) work and was further developed by Kendrick (1961) and Denison (1962).
Growth accounting is a technique used to decompose or ‘break down’ economic growth into components that are contributed by growth in population (labour), capital and other aspects of total factor productivity (Barro 1999). The main difference between the growth accounting framework and Solow’s growth model is that growth accounting moves away from the assumptions in the Solow model related to the capital accumulation process and development of the production function.
Growth accounting’s foundations are based upon a neoclassical production function as shown by:
where Y is output, K is physical capital, L is labour and A is an index of technological stage or TFP. Growth accounting differentiates this production function into
where g is technological improvement or TFP growth. As noted earlier, TFP can incorporate any aspect of the economy that can affect productivity, and thus it is possible to estimate g by replacing TFP with a set of determinant variables such as in equation D3 (Durlauf et al. 2004).
The main advantages of growth accounting are that the model is not based on the assumption that there will be convergence in the long run, and it allows the effect of many variables on economic growth to be estimated. On the other hand, the main problem with growth accounting is its assumption that marginal or incremental productivity is constant and represented by factor prices such as wages and rents (Barro 1999). This assumption has made the approach unable to cope with various developments in growth theory. For example, Romer (1986), and Lucas (1988) found that there may be increasing returns on some factors and spillover effects in the economy, which will be likely to increase the marginal productivity of the two inputs (i.e., not only change g but also or ). Another problem occurs when there are more inputs that cannot be distinguished clearly by the available data (Jorgenson and Griliches 1967). In this case, the marginal productivity of input cannot be calculated correctly and as consequence, TFP and its determinants cannot be estimated correctly either.
The endogeneity of growth
Solow’s growth model has been continuously developed to acknowledge more complex aspects and relationships within the economy, emerging out of theories that acknowledge the endogeneity of growth. One of the most well known developments is the addition by Mankiw et al. (1992) of the stock of human capital as another input in the production model. This new capital input is assumed to have a similar accumulation process to physical capital in the Solow-Swan model. The Mankiw et al. (1992) version of the model can be expressed as:
where h is the level of human capital and β is its input share to output growth. Empirically, Mankiw et al. (1992) claim the inclusion of additional capital in the growth regression reveals the real magnitude of the impact of the physical capital accumulation process on growth and argue that its impact is not as large as the Solow-Swan model predicted. Instead, the accumulation of human capital could account for a high percentage of the explanation for cross-country variations in income.
Another example of developments of the Solow-Swan model relates to the work done by Cass (1965) and Koopmans (1965) who investigated the endogeneity of the Solow-Swan physical capital saving based on Ramsey’s (1928) notion of individual behaviour toward consumption and capital accumulation. They found that an individual’s propensity to save depends on the return to capital, its time discount rate, and also the diminishing value of consumption utility. In other words, individuals will consider how much will be gained from saving, compared to the implicit loss of not being able to immediately consume. Becker et al. (1990) observe the endogeneity of population growth or, to be more exact, the fertility rate. They find that high fertility rates inhibit growth, especially because the production and rearing of children is very time-intensive. As a result, there will be an opportunity cost of having a child. It follows that the higher the individual income the higher this opportunity cost will be. This means higher incomes will have a negative impact on population growth.
In recent years we have seen a growing interest in theoretical and empirical determinants of economic growth in the context of understanding the endogeneity of growth. Durlauf et al. (2004) have argued that despite the complex structure of growth models, most factors, apart from capital and labour, can be classified as part of total factor productivity and consequently have a direct impact on economic performance. However, these factors can also affect growth indirectly through limiting or advancing the composition of capital and labour as input factors. Durlauf and Quah (1999) list 87 specific variables in 36 different categories that have been examined as growth engines or growth determinants, including gender issues. Within this framework, we can see that socio-economic factors including gender inequality can have an impact on economic growth through several channels, as well as directly.
In the next section, we discuss the approaches that have been taken to these issues by authors specifically examining the impact of gender inequality on economic growth.
Methodology used in previous empirical studies on the impact of gender inequality on economic performance
Previous studies examining the impact of gender inequality (either inequality in education or wage inequality) on economic growth have used two main empirical methods.
(a) Econometric estimation
Simplistic exogenous gender inequality model
The empirical estimation used by Seguino (2000 p. 1216), is derived from the neoclassical production function as follows:
where is output, is technological change, is capital stock, are female and male human capital, is the country index and is time. Technological change is defined as where is the gender wage gap and
is the effect of gender wage differentials on growth.
Transforming the production function, Seguino (2000) then built several regressions using cross-country data from 1975-1995, taking into account a country’s fixed effects to estimate the impact of the gender wage gap on economic growth, with this main equation exploring the direct channel and given by:
whereis the difference operator (representing economic growth), is the growth rate of technological change when variables are measured at the mean, represents the fixed effects, is the gender wage gap, are female and male human capital, and is the error term.
Seguino (2000), in contrast to the methodology we use in our study, does not directly examine the ways in which the gender wage gap may be operating through indirect channels. Instead, she runs a separate estimation by regressing investment as a share of GDP on the gender wage gap and finds this relationship to be positive. Seguino (2000) also highlighted the importance of human capital variables as control variables.
In Seguino’s regression, the results also show that while the coefficient of female human capital as a control variable is positive and significant on growth, the coefficient of male human capital is either negative and significant or completely insignificant. Klasen (1999; 2002) noted that these results were possibly due to a multicollinearity problem between female and male human capital. Klasen (1999; 2002) overcame this problem by instead including the ratio of female to male human capital along with a control variable which measured the overall state of human capital.
The next section will discuss an endogenous gender inequality model which acknowledges two-way causality between gender inequality and economic growth. That is, gender inequality influences growth, and growth influences gender inequality.
Endogenous gender inequality model
There is an additional body of econometric literature which acknowledges the endogeneity of gender inequality, and uses endogenous growth theory to estimate the relationship between gender inequality and macroeconomic performance. In relation to gender inequality studies, Klasen (2002) based his work on a modified version of the growth model, incorporating the Mankiw human capital refinements. He estimated the impact of gender inequality in human capital attainment on economic performance in a cross-country study. He did this by building up four equations as follows (Klasen 2002, p.355):
Growth equation to estimate the direct channel of gender inequality on growth (D10)
Investment equation (D11)
Population growth (D12)
Labour force growth (D13)
: average annual (compounded) rate of growth of GDP per capita 1960-1992
average investment rate 1960-1992
average annual (compounded) rate of population growth 1960-1992
average annual (compounded) rate of labour force growth (15-64 years), 1960-1992
total years of schooling in 1960
female-male ratio of total schooling 1960
annual (absolute) growth in total years of schooling 1960-1990
female-male ratio of growth in total years of schooling 1960-1990
other independent variables typically included in cross-country growth regressions (openness (average ratio of exports plus imports to GDP), initial income per capita in 1960 and regional dummy variables)
Using a path analysis method, Klasen (2002) was able to calculate the total effect of gender inequality upon growth, which is a summation of the direct and indirect effects.
Total effect = Direct effect + Indirect Effects
Examining the four main equations above, the total effect from gender inequality in education (i.e.) can be calculated as follows:
where is the direct effect of gender inequality in education,
is the indirect effect of gender inequality in education via investment, is the indirect effect through population growth,
is the indirect effect through population growth and investment,
is the indirect effect through labour force growth; and
is the indirect effect through labour force growth and investment
Klasen (2002) also then estimates a reduced form of the regression which omits the indirect channel to measure the total effect of gender bias directly on economic growth.
Thus mathematically, should replace equation (D14).
However, since equations (D11 to D13) do not contain all dependent variables from equation (D10), we may expect that will not to be identical to above equation.
Since in equation (D10) above, economic growth depends not only on the total years of schooling and female-male ratio of total schooling in the initial period of 1960, but also on the growth of these variables (GED and RGED), there is a possibility that a reverse causality exists from economic growth to both GED and RGED . Therefore, GED and RGED are endogenous variables. Klasen (2002) trialled several specifications, including the use of instrumental variables, to address this simultaneity problem. Interestingly, Klasen (2002) found that the causality runs from gender inequality in education to economic growth and not vice versa.
Although Klasen (1999; 2002) concentrated on the impact of gender inequality in education, intrinsically, his “path analysis” methodology should be able to be applied to other types of gender inequality, including the gender wage gap.
Dollar and Gatti (1999, p. 6) also acknowledged the endogeneity issues of gender inequality in a much simpler framework, however they only measure the direct effect of gender inequality to growth, and do not examine indirect channels.
(b) Calibration or simulation models
While the econometric estimations discussed above focus more on the empirical aspects of the impact of gender inequality to growth, another stream of literature builds a more complex theoretical model, by setting up several functions in the economy, then carrying out a calibration or simulation on the baseline economy with relevant parameters.
Researchers who have applied calibration methodology to studying the impact of gender inequality on economic performance (Calvacanti and Tavares 2007 and Caro 2008) developed an overlapping generations model to study the cost of the gender wage gap, particularly following Galor and Weil (1996). Galor and Weil (1996) linked fertility and economic growth by showing that increasing women’s relative wages reduces fertility and increases labour force participation, by raising the opportunity cost of having children.
In this “overlapping generations model”, women and men are assumed to live for three periods. During the first period, women and men are children and consume a fraction of their parent’s time endowment (parental care). During the second period of life, these children grow up to become adult men and women, organise as couples and make decisions together (divisions of household labour) about time at work, time raising children, and savings for the future. In the third period, each couple consumes what they saved in the second period. This overlapping generations model is based on very strong assumptions and is highly structured: for example, it is assumed that only men will be engaging in production involving physical work.
Calvacanti and Tavares (2007) and Caro (2008) used a simple growth model with endogenous female labour participation and fertility to estimate the cost of gender discrimination. They set up a production function in which gender wage discrimination is introduced where women only receive a fraction of men’s salary, a utility function of the household and a budget constraint to set up equilibrium in the labour market, capital market and goods market.
After equilibrium was reached, calibration was conducted. Calvacanti and Tavares (2007) calibrated the performance across countries using the United States economy as the baseline economy to estimate whether wage discrimination explained the whole output difference in comparison to US output. On the other hand, Caro (2008) calibrated the performance of the Spanish economy by simulating different discrimination levels.
- The Swan (1956) model was developed at the same time as the Solow (1956) model, but separately.
- α is known as the input share coefficient since it can represent the contribution of a particular input growth to the growth of output.
Macroeconomic estimation methods and results
- Data and Variables
- Statistical Summary
- The estimation model and procedure
- 1.2 Estimation Methodology
- Estimating the Impact
- Model Limitations
One of this project’s key outputs is an estimation of the impact of the gender wage gap on economic performance or GDP as a proxy of economic output. In Appendix D, we discussed the range of approaches that have been taken in previous studies to conduct this type of estimation. We have opted to conduct the estimation using a regression technique based on the expansion of the Solow growth model, similar to that used by Klasen (1999; 2002) and Dollar and Gatti (1999). We consider this endogenous econometrics approach to be the most robust modelling technique for our purposes because it reflects the actual nature of the relationships that exist within the broad macroeconomic framework (based on a widely accepted model) without relying on a very strong assumption about the growth processes such as in an exogenous growth model or the strict structure of the calibration model.
As we are studying the impact of the gender wage gap in a particular country - Australia – we need to use time series data for our estimation (as opposed to cross-sectional data which can be used when a number of countries are being studied together). While both cross-sectional and time series data can pose problems for the type of regression analysis we are undertaking here, a single country, time series analysis has a number of advantages. First, the Solow model was originally built to examine the dynamics of one economy (Solow 2001) and although it can be adapted for cross-country analysis, was not explicitly intended for this purpose. Second, some technical problems common to cross-sectional, multi-country studies are less likely to be experienced in a one country, time series studies such as ours6.
We use data from a variety of Australian Bureau of Statistics sources. The variables we include in the model (along with a measure of the gender wage gap) are those common to macroeconomic analysis of the kind being undertaken here. They are GDP, investment, human capital, fertility, labour participation and hours worked. Issues related to data availability (common in single country studies where data points may be limited) have affected both our modelling approach and our results. These issues are discussed in more detail below. In macroeconomic modelling, the longer the time period of data available, the more likely that such a time span smooths out the ups and downs of the business cycle. As one of our key variables – human capital – was only available from 1989 onwards, we were able to include only 20 data points in our analysis. However, we were able to test and adapt our model to address the issue of a relatively short span of data. Nevertheless, these data limitations should be kept in mind when interpreting our results.
The use of time series data raises the issue of the stability of the data over time – referred to as stationarity (Durlauf et al. 2004). In essence, unless we can be reasonably confident that the variables we use have a consistent average over the long run, we cannot be sure that the relationships we see from a regression analysis are accurate, or if in fact they simply (or partly) capture fluctuations over time in the data. We ran a series of tests related to stationarity (described below) and found that we could not be sufficiently confident of the stationarity of our data without making an additional adaptation to our model
In order to adapt the model to overcome possible problems with stationarity, we used first difference equations, rather than equations which use the data in what is known as a level form. This approach resolves problems of data stationarity by subtracting the equation at time t from the equation at time t-1. While this method means that we lose one observation point, the increased confidence that we have in our results due to ‘first differencing’ the model is very important.
Below we describe the database used in our modelling, including details of our rationale for choosing particular variables and data (including issues of data availability), and the characteristics of the data. We then describe our estimation technique, and present detailed results. Finally we discuss how we come up with an estimation of the total impact to the Australian economy of the gender wage gap, and provide details about the limitations of our modelling.
Data and Variables
In this section we describe the ways in which we have operationalised variables for use within our growth model. We then summarise the chosen variables and the data used at the end of each variable description. Finally, the variables we have used and their data sources are provided in Table E1.
As discussed earlier in this report, economic performance is the main variable of interest in this study and it is often measured by economic output, in particular gross domestic product (GDP). GDP is a suitable measure for examining the performance of developed countries, while for developing countries other measures such as the Human Capital Index and the achievement of Millennium Development Goals have been used to measure economic performance. GDP measures the output produced by activities in the economy. However, when using GDP as a measure of economic performance, a set of decisions needs to be made in relation to definitions (and thus in turn about data).
There are three common approaches used to calculate GDP – value added, expenditure and income (Mankiw 1997). The value added approach measures GDP by summing up all goods and services in the economy subtracted by the intermediate products used to produce them. By doing so, the calculation will not “double count” the products that are part of other products. The second approach is based on income. The income approach to GDP is calculated by summing the income of capital, labour, and entrepreneurship having first deducted the cost of tools and materials. Finally, the expenditure approach uses an assumption that for every seller there is a buyer. Therefore, with the expenditure approach, GDP can be calculated by summing up all consumption, investment, government expenditure and net exports.
Figures of GDP which are based on these three approaches are available in the Australian Bureau of Statistics (ABS) publication Australian National Accounts: National Income, Expenditure and Product (Cat. no. 5206.0) in which quarterly data are available from June 1960 and in the publication Australian System of National Accounts (Cat. no. 5204.0) which provides annual data. Conceptually, the income, expenditure and production approaches to measuring GDP should achieve the same results; however this is dependent upon how the data are derived. In Australia, prior to 1994-95, these estimates were produced independently, and consequently there are statistical discrepancies between each estimate. However, over the period in which these estimates were calculated separately, the ABS also provided a compilation number, representing the average of these three approaches. Since 1994-95, there are no discrepancies between any of the estimates, with the exception of the quarterly estimates for the June 1994 quarter (ABS 2009).
Another issue in relation to the use of GDP as a proxy of economic output is whether to use GDP in constant or current prices. Current price GDP means that the currency amount of products in one economy is calculated based on the price of the product at that year or quarter. However, if our interest in GDP is to conduct economic growth analysis, and as this is based on the production (supply) side of the economy, we need the representation of quantity produced in the economy where the change in the value of output is not merely caused by the current prices of those outputs. The constant price GDP provides this representation since although it is reported in currency units it nevertheless is based on estimated fixed prices in the base year. Therefore, the changes in GDP value will represent the changes in the quantity of goods and services produced, avoiding the effects of price changes on the output calculation. The ABS now uses a chain volume measure of GDP in the Australian System of National Accounts as their measure of constant price GDP. The ABS uses the chain volume measure because it takes into account different growth rates of unit prices of goods and services as well as the growth rate of their quantities, which results in changes in price relativities. The chain index used in chain volume GDP is calculated based on annual changes so that price relativities remain the same (5216.0 - Australian National Accounts: Concepts, Sources and Methods, 2000).
Variable:Gross domestic product (GDP)
Data used: Chain volume measure of GDP, annual series, ABS Australian System of National Accounts (Cat. no. 5204.0)
Capital and Investment
Capital stock is one of the two major inputs that produce economic output (the other is labour). The addition to capital stock is known as ‘investment’. Data measuring capital stock are available annually in the publication Australian System of National Accounts (ABS Cat. no. 5204.0). Investment can be calculated by an increase or decrease in capital stock. However, a more direct measure of investment is gross fixed capital formation, available in the publication Australian National Accounts: National Income, Expenditure and Product (ABS Cat. no. 5206.0) or total capital accumulation and net lending, available in the publication Australian System of National Accounts (ABS Cat. no. 5204.0). We opted to use total capital accumulation and net lending in the Australian System of National Accounts since this takes into account changes in financial capital as well in physical capital. We use this variable measured as a ratio of GDP (i.e. the rate of investment).
In most of the growth models discussed earlier, investment predicts economic performance. This theory is supported by empirical results that estimate the robustness of this variable compared with other growth determinants (Barro 1991, Levine and Renelt 1992, Mankiw et al. 1992, Caselli et al. 1996). Nevertheless, studies that compare the effects of investment and income on economic growth have shown that the superiority of investment as a predictor of economic growth is not as strong as earlier predictions would suggest (Mankiw et al. 1992; Easterly and Levine 2001). Moreover, the direction of causality could be from growth to investment rather than vice versa (Blomstrom et al. 1996). Barro (1996) shares this view by claiming that high growth prospects are one of the main drivers of high investment7.
Variables: Gross fixed capital accumulation (Total investment)
Data used: Chain volume measure of gross fixed capital accumulation, Australian National Accounts: National Income, Expenditure and Product (Cat. No. 5206.0)
Labour and Fertility
In addition to capital, labour is the other major input to economic growth. There are several ways to measure labour input, including the number of people participating in the labour force, hours of work and their cost (i.e. wages). In the publication - Labour Force, Australia, Detailed, Quarterly (Cat. no. 6291.0.55.003), the ABS provides two of the three measures of labour input, which are the number of employed persons by full-time and part-time status and their hours of work. We use these measures rather than a wages measure, as the former are more commonly used in Australian macroeconomic modelling, and in order to better separate our labour variable from our gender wage gap variable (which is, of course, based on wages).
It is also important to include fertility in growth models such as the one we are using in this study. The Solow growth model (on which our estimation technique is based) has a strong assumption of diminishing marginal returns of labour. This means that although an increase in labour is likely to increase economic output, theoretically it will decrease labour productivity if there is no change in other inputs or aspects of the economy. In addition, the model also assumes that the labour force is growing in the rate of population growth. As a result, an increase in labour can also decrease the output or income per capita of an economy given that it is solely boosted by population growth. Most cross-country studies have supported this hypothesis. The strongest result to confirm this negative impact of population growth on economic growth is from Mankiw et al. (1992) while Levine and Renelt (1992) produce a less significant relationship. However, if the fertility rate is controlled for in economic growth models, the impact of population growth on income per capita may also have a positive sign. Barro and Lee (1994) argue that this positive result is due to either positive net migration (reflecting an increase in skilled workers which has a positive impact on growth) or a low mortality rate that indicates a better health system.
Variables: Labour: Number of employed persons and hours of work; Fertility: Total Fertility Rate
Data used: Labour: Labour Force, Australia, Detailed, Quarterly (Cat. no. 6291.0.55.003)
Fertility: Fertility data in Australia is available annually in Births, Australia publication (ABS cat. no. 3301.0) or in Australian Historical Population Statistics (ABS cat. no. 3105.0.65.001).
Human capital is the most common additional variable used in growth estimations (Durlauf et al. 2004). Lucas (1988) defined human capital as a general skill level, indicating that human capital contributes to production by increasing worker productivity as well as directly increasing output through its contribution to technological improvement. Research has shown that the impact of human capital accumulation contributes to greater technical progress or Total Factor Productivity (TFP) growth by affecting knowledge accumulation (Romer 1990). The indirect impact of human capital accumulation mostly occurs through lower fertility or population growth (Becker et al. 1990; Galor and Weil 2000).
There are various types of human capital measurement, which are based on (i) health – mostly using life expectancy as a proxy; and (ii) education, mostly using school attainment as a proxy. Previous studies that have discussed gender inequality in human capital have mostly focused on education (Galor and Weil 2000; Klasen 1999; Seguino 2000), and that is the measure we use in this study. Annual data for this educational proxy of human capital, is available in ABS data on educational attainment in the publication Education and Work, Australia (ABS cat. no. 6227.0). Data begins in 1989, with details for the number of persons and workers by their highest educational attainment (in terms of non-school qualifications).
Variables: Proportion of persons with non school qualification
Data used: Education and Work, Australia (ABS cat. no. 6227.0)
Gender Wage Gap
The gender wage gap is the main input variable of interest in this study as our primary aim is to estimate the impact of the gender wage gap on economic growth. The connectivity between the gender wage gap and economic growth has been discussed in other parts of this report. The ABS conducts several surveys which have data and information about the gender wage gap. In this estimation we will use Average Weekly Earnings, Australia (Cat. no. 6302.0), which provides quarterly estimates of average gross weekly earnings of workers in Australia. Average Weekly Earnings data was chosen because it provides the longest time span of consistent data on wages.
Variable: Ratio of wage difference between male and female to male wage
Data used: Average Weekly Earnings, Australia (Cat. no. 6302.0) (Full time adults, ordinary time earnings)
|Economic Output||GDP||Australian System of National Accounts (Cat. No. 5204.0)||Annually|
|Investment||Gross Fixed Capital Formation||Australian National Accounts: National Income, Expenditure and Product (Cat. No. 5206.0)||Annually and quarterly|
|Labour||Labour participation (Number of employed persons full time and part time)||Labour Force, Australia, Detailed, Quarterly (Cat. No. 6291.0.55.003)||Quarterly|
|Hours of work||Labour Force, Australia, Detailed, Quarterly (Cat. No. 6291.0.55.003)||Quarterly|
|Fertility||Total fertility rate||Births, Australian Publication (Cat. No. 3301.0)||Annually|
|Australian Historical Population Statistics (Cat. No. 3105.065.001)||Annually|
|Human capital (Education)||Proportion of persons with non school qualifications||Education and Work, Australia (Cat. No. 6227.0)||Annually|
|Gender wage gap||Ratio of wage difference between male and female to male wage||Average Weekly Earnings, Australia (Cat. No. 6302.0)||Annually and quarterly|
During the past two decades Australia’s GDP based on the chain volume measure grew by 3.4 per cent annually from $585 billion in 1989 to $1,084 billion in 2008. Meanwhile GDP per capita grew by 2.1 per cent annually from $35,000 in 1989 to $51,000 in 2008. Although both GDP and GDP per capita showed an increasing trend during this period, there was slight negative economic growth (especially in terms of GDP per capita) in the period 1991-1992. The overall increasing growth trend was also experienced by capital stocks, which increased from $1,750 billion in 1989 to $3,178 billion in 2008. This included an exponential trend in capital stocks between 2000 and 2008. However, there was a considerable drop in the flow of investment during 1989-1992, before experiencing a steady increase since then.
In the same period of 1989-2008, labour grew by 2.0 per cent annually, from 7.7 million people participating in the labour force in 1989, to 10.8 million people by 2008. In line with the fall in investment flow between 1990 and 1992, there was also a drop in labour force numbers over that period, but this was less marked than the drop in investment. The amount of hours of worked grew at a somewhat lower pace than overall labour force participation, at 1.8 per cent annually: from 267,000 hours a week in 1989 to 355,000 hours a week in 2008. Putting together the data on persons in the labour force with the information about hours worked, we find that the average hours of work of those in the labour force dropped by 0.2 per cent annually over the period, falling from 34.6 hours per week in 1989 to 32.9 hours per week in 2008. In terms of total hours of work, there was also negative growth in 2001 and 2006, as well as negative growth during the 1991-1992 downturns.
Overall population numbers grew slower than the number of labour force participants during the 1989-2008 period. At 1.3 per cent annually, the population grew from 16.7 million people in 1989 to 21.2 million people in 2008. The growth rate actually slumped from 1.8 per cent in 1989 to below 1.0 per cent in 1999. However, the growth rate then increased steadily to 1.2 per cent in 2005 and more gradually to 1.7 per cent in 2008. The latest increase in population growth is driven in part by the a jump in the Total Fertility Rate from 1.7 live births per 1000 women in the 2001 calendar year to 1.9 live births per 1000 women in 2007. This happened after fertility gradually decreased from 1.9 live births per 1000 women in 1990.
The percentage of persons aged 15-64 who attained a non-school qualification or tertiary education also increased substantially between 1989 and 2008. In 1989, 39.2 per cent of persons aged 15-64 held a non school qualification. This had increased to 53.9 per cent in 2008. The trend shows that the percentage of persons with non-school qualifications or tertiary education generally grew from year to year except for 1993, 1994 and 1997. In terms of gender inequality, the school attainment gap has closed considerably. In 1989, 45.4 per cent of men aged 15-64 had a non school qualification, compared to 39.2 per cent of women. These figures had increased to 55.3 per cent for men and by 52.6 per cent for women by 2008.
To be able to produce a reliable and consistent estimate when conducting econometric estimation, it is necessary that all variables have a stationary data stream – that is, that they are not being unduly influenced by other factors such as the normal peaks and troughs of the business cycle. Stationarity means that the variable is likely to converge to a single value or a trend path in the long run, and thus ensures that the mean, variance and relationships we are going to estimate will be achieved in the long run (Enders 1995). Without stationarity, we risk estimating a ‘false’ relationship that is based on insufficient serial autocorrelation within each variable. In addition, it is always important that the value that represents a variable at one moment, in a certain set of conditions, shows a consistent long run value, given everything else being the same, and not a value that will be likely to change in the short term.
Thus before undertaking the sort of estimation procedure we are using in this project, it was important to find out if the variables we are using have stationarity or not. Well-established tests exist to identify this, and we used these to explore the extent of non-stationarity in our data8. These tests, and the results from our preliminary modelling, when considered overall, led us to the conclusion that there was some degree of non-stationarity in our data. We addressed this through incorporating a first differencing approach into our modelling, as described in the following section.
The estimation model and procedure
The Growth Model
Having completed our preliminary testing, and drawing lessons from our initial results, we then went on to develop our final estimation approach.
First, one of the most important findings from the stationarity tests of the potential variables was that our capital stock variable was not stationary in either the level (original) form or the first difference form. As a result, we could not use that variable to build the model. This in turns means that we cannot develop the model based on the growth accounting structure (i.e., equation D6 in Appendix D). Therefore, the model was built based on the Solow growth model, in particular, the structure that is offered by Mankiw et al. (1992) which is known as the augmented Solow growth model in the form of:
where A is Total Factor Productivity (TFP), Y is output (GDP), I is physical capital accumulation or investment, L is labour9, h is human capital level (proxied by the proportion of population with non school qualifications) and n is population growth (proxied by the fertility rate). Following Seguino (2000), the technological or TFP change is formulated in the form of:
where Ci is a country-specific variable which in our case should be constant because we only use one country, Ø is the time effect and Wgap is the gender wage gap, calculated as where wm is average gross weekly earnings of male workers and wf is average gross weekly earnings of female workers.
Substituting this TFP formulation into the equation (1), we will have
The next step is to relax the assumption that the growth of labour will be equal to population growth, which allows us to better measure labour participation. This is done because we want to explore the impact of the wage gap on workers’ productivity based on the total hours of work that they offer, rather than just whether or not they work at all. With the need to include the gender wage gap in the equation, we then adjust equation (E3) to be:
where Hw is hours of work and P is the total population so the average hours of work is calculated as . With this equation, we can examine the direct effect of the gender wage gap on economic growth.
1.2 Estimation Methodology
As discussed elsewhere, there are several possible modelling options within the overall framework described above, and we need to know more about the possible presence of indirect channels within the model in order to decide on a suitable estimation approach. Therefore, we first carried out a sensitivity analysis to examine the existence of indirect channels from the gender wage gap to economic growth, by taking in and out of the model variables that the literature suggests may be indirect channels of the gender wage gap’s impact on economic growth (for example, fertility and investment). We found that when we added in and removed variables there were changes in the magnitude of the coefficients of the gender wage gap and that the coefficients of the gender wage gap in various combinations of specifications were not robust. These findings confirm the existence of indirect channels (if the only effect of the wage gap on economic growth was direct, then we would expect to see little change in the coefficients when variables are added into or removed from the model). These findings show that our best estimation option was to conduct simultaneous equations using three-stage least squares in which the independent variables are mostly endogenous, allowing us to examine the direct and indirect channels together.
The essential component of the three-stage least squares estimation is the existence of enough instrumental variables11 to allow us to have combinations that have sufficient explanatory power in relation to the variable that is instrumented12. This should include adequate exogenous instrumental variable(s), otherwise the equations will end up in perfectly multicollinearity. In other words, we need to have variables in our model which explain much of the variation in the instrumented variable (here, rate of investment, average hours of work, fertility, human capital and labour participation), but are not determined by the outcome variable (which here is GDP per capita) which would retain the endogeneity problem. To achieve this, we use the lag of instrumented variables as the exogenous instrumental variables, given that the lag variables are likely to determine current conditions (that is, current GDP per capita) but theoretically cannot be determined by current GDP per capita. Using these lag variables means that we use the instrumented variable at time t-1, t-2 and so on until we get a reasonable estimate of the instrumented variable at time t. Unfortunately, in doing so, we are losing the ability to estimate the impact of the gender wage gap through human capital accumulation, as data for this variable is not available before 1989. Introducing a human capital lag would cut many of the available observations in the model, and as the other instrumented variables are available from 1985 or earlier, we decided to take advantage of this longer period of data for the other variables, despite losing our ability to use human capital accumulation as an indirect channel. The human capital variable is still included in the model, but the extent to which the gender wage gap may operate through it to impact on GDP per capita cannot be estimated.
As our stationarity tests had given us slightly inconclusive results in regard to the stationarity of our variables (the tests suggested a possible problem with stationarity, but when analysed carefully did not provide sufficiently strong evidence to absolutely reject the possibility that our data was stationary), we needed to conduct some initial modelling in order to further inform our understanding of possible problems with stationarity. If overall we considered stationarity to be a problem, then we would need to address this by using first difference terms of our variables (subtracting the equation at time t from the equation at time t-1), rather than the original (or level) variables.
We conducted several estimations using both the level term and the first difference term of the variables, in order to see if our results could provide any further information about which of these approaches we should use. We found that, using the level terms, we came up with a very high R2 value (99 per cent)13. Very high R2 values like this are a cause for concern. In this case, such a high value may be an indication of model misspecification and also may point to the existence of spurious regression caused by variables that are not sufficiently stationary. In contrast, the estimation based on first difference variables produced an R2 value of 77 per cent. Based on these findings, and in the light of the stationarity testing, we opted to estimate the impact of gender wage gap on economic growth using a first difference equation14.
To estimate the equation using a first difference equation, we subtract equation (E4) with
The next step is to build the system to recognise the endogeneity of the growth determinants and hence the indirect effect of the wage gap through other growth determinants. As with the main growth equation, we also estimate all equations in this system using the first difference form, especially given the variables we use in these subsequent equations are simply the first difference terms of those used in the main growth equation. This system can be written as:
Tables E2 to E7 show the estimation result of equations (E6)-(E11). The system is estimated using the data described above with t from 1990 to 2008. We utilise the coefficients from these tables to generate the impact coefficient that we use to estimate the impact of the gender wage gap on economic growth as shown in Table E8.
|(GDP per capita)|
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The result of estimating equation (E6)
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The result of estimating equation (E7)
|(Average hours of work)|
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The result of estimating equation (E8)
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The result of estimating equation (E9)
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The result of estimating equation (E10)
|(Gender wage gap)|
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The result of estimating equation (E11)
Estimating the Impact
The coefficients that have been produced from the regression system above allow us to calculate the impact of the wage gap on economic growth. For example the coefficient °in the main growth equation (i.e., equation (E4)) represents the impact of the wage gap on economic growth while holding other variables constant (referred to as partial impact). This impact can be estimated as follows:
In the same way, the partial impact of other growth determinants on economic growth can be estimated.
The impact of investment: (E13),
The impact of fertility: (E14),
The impact of labour participation: (E15),
The impact of average hours of work: (E16),
The equations (E7)-(E10) in the system provide further information about how much these growth determinants will change if there is change in the wage gap. This can estimated as,
The impact on investment: (E17),
The impact on fertility: (E18),
The impact on labour participation: (E19),
The impact on average hours of work: (E20).
By substituting the equations in (E17)-(E20) into the equations in (E13)-(E16), we will have the impact estimate of the wage gap through other growth determinants as follows:
The impact through investment: (E21)
The impact through fertility: (E22)
The impact through labour participation: (E23)
The impact through average hours of work: (E24)
Therefore, the total differential of growth on the wage gap provides the total impact of the wage gap on growth as
Table E8 shows the impact estimate of the gender wage gap on economic growth
|Wage gap à economic growth||-0.25||-0.250|
|Wage gapàinvestmentàeconomic growth||-0.261||0.081||**||-0.021|
|Wage gapàfertilityàeconomic growth||0.993||-0.182||-0.181|
|Wage gapàaverage hours of workàeconomic growth||-1.432||**||0.222||***||-0.318|
|Wage gapàlabour participationàeconomic growth||0.378||0.695||***||0.263|
Note: *, **, *** is the10%, 5% and 1% significance level, respectively
Source: The coefficient from tables E2 to E7
Estimating the Total Impact on GDP
The previous section has shown how to get the total impact coefficient of the wage gap on GDP. The final step is to calculate the total volume of GDP change (in dollars) which would result from a decrease or increase of the wage gap. To do so, we should rearrange equation (25) as
and multiply both the right hand side and the left hand side of equation (26) by Pt to get
Knowing that the wage gap is currently 0.17 and the chain volume GDP is at $1,084,146.00 million we can calculate the impact of a change of one percentage point in the gender wage gap, as well as the cost of the whole 17 per cent of the gender wage gap as seen in Table E9.
|Current GDP ($ millions)||Change in GDP economic growth (%)||Change in GDP ($ millions)|
|Gender wage gap increases by one percentage point||1,084,146.00||-0.50||-5,496.65|
|Total cost of wage gap ( wage gap is eliminated by 17 percentage points)||1,084,146.00||8.50||93,443.06|
The confidence level
While the coefficient estimates of our main growth equation have relatively low standard errors (which meant that coefficients were statistically significant and thus supported the direction expected by the theoretical growth model), we did however find high standard errors when estimating the impact of the gender wage gap directly on growth and most of its components. This was a drawback of our model. The lack of observations available to us is the likely cause of this. The high standard errors mean that our impact estimate has wide confidence intervals. Nevertheless, the coefficient estimate of the main growth equation (equation (E6)) is significant and shows a relationship between the gender wage gap and GDP per capita that is supported by growth theory. Therefore, despite data availability issues and high standard errors, our result is one which makes sense theoretically.
The question is, to what extent we expect that our estimation reflects true relationships in the economy. As can be seen in table E8, the impact of the gender wage gap on economic growth through an increase in average hours of work is the only impact that we found to be statistically significant at the 5% significance level. That means that we can be confident that there is 95% probability that this will actually be the outcome of a change in the gender wage gap.
In this interpretation, we have taken into account in calculating the total gap the impact of all channels – both significant and non-significant. However, what would the gap look like if we only took into account the impact of the one significant channel (hours of work)? As shown in Table E10, the impact is still in the same direction, and still very substantial – if we take only the significant channel into account, we still see that for an increase in the gender wage gap of one percentage point, GDP falls by 0.318 per cent. Thus we find that just this single channel accounts for 62.7 per cent of the total expected impact.
Another way of considering these issues is to think about what would happen to the apparent impact of the gender wage gap on GDP if we lowered our confidence level to 80%. Would we still see only one significant channel, and would the apparent impact change? We did this (see Table E10), and found that this added another significant channel to the model. With a confidence level of 80% we found that the gender wage gap affected both fertility and hours of work significantly. If we then use the impact of these two channels only to estimate the effects of the gender wage gap on GDP, we find that this effect (at -0.499) is very similar in magnitude to the total effect when all channels are used.
These results provide further support for our overall findings – they suggest that even when we adjust various aspects of the assumptions built into our modelling, the overall story about the impact of the gender wage gap on GDP per capita remains much the same.
Gender wage gap increases by one percentage point
|Change in GDP economic growth (%)||Compared to total impact (%)||Change in GDP ($ millions)|
|Total Impact with all channels||-0.507||-5,496.65|
|Total Impact with only channels significant at95% confidence level||-0.318||62.7||-3,447.58|
|Total Impact with only those channels significant at 80% confidence level||-0.499||98.4||-5,409.89|
Source: Authors’ calculations
The discussion in this appendix has shown that our model is placed strongly within current approaches to growth modelling, has been rigorously tested, and is capable of providing an estimate of the impact of the gender wage gap on economic growth, and the indirect channels through which such an impact may be occurring. However, this model is subject to some weaknesses and limitations, and these are to some extent reflected in our results. We discuss these limitations here.
The main source of weakness in our model is the fact that it relies on a limited (20 year) data span. These data limitations affected the choices we made about model specification and also the strength of our estimation.
Because not all variables which we needed to include in the model were available on a quarterly basis, we had to use annual data, with 1990 as the starting year. Because we needed to use a first difference equation in addition to using lagged versions of our variables in order to achieve a robust estimate of the coefficient of growth determinant, we set 1990 as our base year (i.e., at time t), which means that we needed to use data from 1985 to construct the lag variables. Our preliminary data testing showed that the combination of instrumental variables can only give a good estimation of the channels if we include three lags to explain the behaviour of each channel, except for fertility which needed four lags.
Therefore, we only have 19 years of observations for the estimation given that we use 2008 data as the end of the observation period. As noted earlier, the short data span meant that our variables were less likely to be significantly stationary in the level form, and this contributed to our decisions about model specification.
One of the key decisions we had to make was to use the first difference form rather than the level form of our equations. The very high coefficient of determination (R2) of 99 per cent in the level form is likely a result of the short data span. While our tests of stationarity did not conclusively reject the level form data, this high R2, as noted earlier, raises substantial concerns about the possibility of a spurious regression. On the other hand, the coefficient of determination (R2) of 77 per cent in the first difference form is the better choice since this means that the model can still explain the variation of the dependent variable with a reasonable range of error. However, the use of first differences also means that the result may not give a picture of long term equilibrium. With only 19 years of data this is an expected result, since the average rate of convergence in growth models is around 35 years.
Another decision we had to make was not to estimate the impact of the wage gap through the indirect channel of human capital. While the short data span available to us is mostly due to the length of human capital data available from the Education and Work Survey, we nevertheless considered the inclusion of human capital in our model as being of significant theoretical importance. Our estimation result bears this out, showing that the inclusion of human capital is almost as important as the inclusion of physical capital investment (see Table E2). This can be seen from the magnitude of the human capital coefficient in the main growth equation at 0.06 compared to 0.08 for physical capital investment. The fact that human capital is so important in predicting growth, and our inability to model this indirect channel in this study, means that our model in fact may underestimate the overall effects of the gender wage gap on GDP – that is, had we been able to examine the human capital channel, this might have further contributed to the impact of the gender wage gap on economic growth. This is particularly so as some of our preliminary modelling showed that the wage gap may have a negative impact on human capital accumulation. However, given the high standard error of the human capital coefficient in our final model, we do not expect that such underestimation would be of a large magnitude.
As noted above, the impacts of the wage gap on growth components within our model have relatively high standard errors. These relatively high standard errors are another weakness of our model, making it difficult to detect statistically significant relationships between channels of impact and economic growth. The only channel for which we have statistically significant results is for the average hours of work variable, which shows that a decrease in the wage gap will increase the average hours of work and thus eventually increase economic output. However, as discussed above, our findings related to the total impact of the gender wage gap on GDP per capita appear fairly robust.
The final limitation of the model is that we cannot separate the impact of the gender wage gap on growth components based on gender. The short data span is partly the reason, but most importantly the high collinearity of the time series data between the two genders of labour force participation made this decision necessary – that is, male labour force participation and female labour force participation were too highly correlated with each other to be able to effectively separate out their explanatory power in the model. If variables which are highly collinear are used in a model such as the one developed here, the model will be both inefficient and liable to give misleading results.
- These include fewer problems related to omitted variable bias (see Greiner et al. 2003).
- There is also a body of literature which suggests that different types of investment (for example, government investment versus private investment) may have different impacts on growth (see, for example, Barro 1990; Folster and Henrekson 1999).
- We used both the Augmented Dickey-Fuller (ADF) unit root test (Dickey and Fuller 1979) as well as an additional stationarity test based on Kwiatkowski et al. (1992). Results of this testing are available on request.
- We have chosen the ratio of labour (measured by total employed) to total population instead of the labour force participation rate to represent labour input for two reasons :
- mathematically, to be consistent with the dependent variable which is GDP per capita
- It is more suitable in gaining an accurate measure of the impact of the gender wage gap on productivity, which covers the population who are employed only.
- Given the fact that
- An instrumental variable (IV) is a variable which is uncorrelated with the error term in the main (growth) equation and is correlated with the endogenous explanatory variable.
- We need to make sure that the impact of the wage gap and other instrumental variables included in estimating each indirect channel will have a good representation of that channel. If not, this system will perform poorly.
- The R2 value indicates the total amount of variation in the dependent variable (GDP per capita in the equation above) which is explained by the variables being used in the model.
- One possible disadvantage of the estimation using a first difference equation is, as discussed in Engle and Granger (1987) following Phillips (1957), that the coefficient estimated in this equation may not hold in the long run especially when there is a temporary relationship that causes a temporary disequilibrium in the model. However, we also conducted the estimation at the level form and in many respects, the two sets of results are similar, with the magnitude of the total effect of the gender wage gap on growth being closely comparable using both estimation methods (-0.51 against -0.57). An exception to this overall similarity between the two approaches relates to an analysis of the indirect channels of impact. The level estimations indicate that the negative effect of the wage gap on economic growth is much more strongly associated with a drop in investment than in average hours of work, while our first difference findings show that hours of work is a much more important channel. The latter result is much more commonly found in previous studies than the former, which provides further support for our choice of first differencing.