4 Equal Pay Act was enforced that prohibited differentiating wages based on gender. Australia is also a striking example in terms of promoting equal pay. The Com- monwealth Conciliation and Arbitration Tribunal took several measures to prevent discrimination on the gender basis. In 1969 the principle of equal pay was intro- duced, which aimed to prohibit the differentiation of pay for the same work. This principle was extended in 1972 and covered work of equal value followed but set- ting a single minimum wage in 1974. The law prohibiting the gender discrimination was enforced by Workplace Relations Act in 1996. (Daly et al. (2006)). Since wages are the largest and most accessible component of income, the gender wage gap has become the most widely studied topic in terms of gender inequality. The economics of discrimination equipped labour economists with necessary tools for studying gender wage gap that has resulted in innumerable research papers try- ing to quantify variables that contribute to the difference. Starting from the 1970s, a myriad of studies tried to explain the factors that contribute to the wage differentials between men and women. To put it simpler, all these studies sought to divide the gap into two parts: one that could be explained by the differences in the observable characteristics of men and women, and the other one that could not be explained, so- called discrimination. The standard gender wage gap decomposition tool emerged from a seminal study of Oaxaca (1973) and Blinder (1973). The main idea of this principle is to write the gap as the sum of two parts: structure (unexplained) and composition (explained) parts. Over time several modifications and extensions of this decomposition method have been developed: Juhn et al. (1991, 1993) extended the method to study changes over time in the unexplained gap; Albrecht et al. (2003) and Machado and Mata (2005) integrated quantile analysis; Fairlie (2005) extended the model to treat dichotomous outcomes; Bauer and Sinning (2008) modified the model for censored outcomes, and Ñopo, (2008, a,b) developed the model for non- parametric setups. Throughout the time the model framework advanced by includ- ing other distributional characteristics, even more, some methods of studying the entire distribution have been developed (e.g. see Chernozhukov et al. (2013), Firpo et al. (200), Fortin and Lemieux (2000), DiNardo et al. (1996)). Weichselbaumer and Winter-Ebmer (2005) conduct a meta-analysis of 263 pub- lished papers and they showed that the estimated size of the gap largely depends on the type of the dataset used rather than on the decomposition method. Further- more, they found approximately 35 percentage points decline in the gender pay gap from the 1960s to 1990s. The decline was mostly due to equalization of productive characteristics, so-called explained part of wage differential if we use the language of the Oaxaca-Blinder decomposition. Convergence of gender pay gap is not sur- prising at all because, as noted above, the convergence of economic status between men and women is mostly due to increased trends in women’s participation in the labour market and their education levels. Blau and Kahn (2006a,b) show that in the majority of OECD countries wage gap has been narrowed down recently but the rate of convergence is very slow. De- spite the prohibition of gender discrimination, women still do receive much lower wages than men. As Ponthieux and Meurs (2015) reported, at the end of 2010 the average gender wage gap in the OECD countries was approximately 16%, but sig- nificant variation had been observed across the countries (Table 1, from Ponthieux and Meurs (2015), p 1010).
- Related Literature 5 One possible explanation for such variation across countries is provided by Blau and Kahn (1992, 2013). In the wage distribution, women are observed to be concen- trated at the lower tail. This unfavourable ranking in the male wage distribution results in less wage differential if the distribution is more compressed. To show how this mechanism works, Blau (2012) compares hourly adjusted gender earning ratios, which are 77.3% and 65.4% in Sweden and the United States, respectively. Women’s mean ranking in men’s wage distribution is lower for Sweden than for the United States, resulting in lower gender pay gap in Sweden than in the United States that is due to more compressed wage distribution1. Important determinants of the wage distribution compression are wage-setting institutions. Low-paid workers, who are mostly women, benefit from highly centralized, unionized wage settings because it reduces wage dispersion (Ponthieux and Meurs(2015), Salverda and Checchi (2014)). Another possible explanation but at a lesser extent is the gender gap in employ- ment. Olivetti and Petrongolo (2008) introduce the effects of selection into employ- ment, which implies that when the fewer women are employed, they are more likely to be selected and the higher their relative wage is. This statement could be trans- lated into a negative correlation between the gender pay gap and the gender gap in employment. Similar effect is reported by Hunt (2002), who found that after 4 years of reunification of former East Germany the employment rate for women had fallen by 6% more than for men, which could be used for explanation the half of the rela- tive wage gain (10% point drop in gender wage gap) of women. In addition to differences in the gender pay gap among countries, a lot of inter- est has been drawn to within-country gender wage gap and its determinants. Pon- thieux and Meurs (2015) summarize the key findings of Weichselbaumer and Winter- Ebmer’s (2005) study and highlight the fact that basic human capital variables can explain only a very small portion of the gender pay gap. Similar results are re- ported by Manning and Swaffield (2008), who studied British Household Panel Sur- vey data. Becker (1993) and Mincer (1974) proposed the human capital model, which attempts explaining the gender pay gap in three dimensions: first, since women are more likely to have interruptions in their careers, it is thought that they will accu- mulate less work experience than men; second, given the fact that women expect interruptions in their careers, it may affect their investment in human capital, for example, education; and third, as Becker (1985) explains after so much time spent on childcare and housework, women have less time left for job and, therefore, they choose less demanding and well-paid jobs. In contrast to the human capital model, Manning and Swaffield (2008) found that human capital hypothesis can explain a significant portion of the gap in the early stage of a career. However, more than a half of the gap that exists 10 years after entering the labour market cannot be ex- plained by this approach. The fact that human capital hypothesis has a significant impact on early career wages is further strengthened by various studies, conducted in the United States, trying to quantify the effects of college major on wages (e.g. see Black et al. (2008), Brown and Corcoran (1997), Loury (1997)). A few women have been observed to choose science or technology as their majors, which leads to a higher degree of occupational segregation (also known as horizontal segregation). 1The basic idea of this mechanism is assigning a rank to females according to their wages in the male wage distribution. Then, the average of females’ rankings gives mean percentile ranking of females in the male distribution. Had men and women had the same distribution, the average of these ranking would have been 50. Hence, in the wage hierarchy the lower mean ranking of females implies their less favourable position.
6 This itself brings up a question why women do not choose those career paths if they promise higher wages? Polacheck (1981) claimed that women tend to choose pro- fessions that do not require high career interruption costs. However, Ponthieux and Meurs (2015) argued that this explanation does not work in the modern societies as nowadays women are more attached to their jobs and their careers are often contin- uous and pointed to psychological factors, which are discussed below. The trend of occupational segregation has not been linear over time. Blau and Hendricks (1979) observed a sluggish decline in the 1960s, which was followed by a faster decline in 1970s (Bianchi and Rytina (1986)). Since the 1990s the decline slowed down significantly (Blau et al. (1998); Hegewisch et al., (2010), observed stagnation during that period). Blau et al. (2013) showed that the occupational segregation had declined among those with college degrees, however, almost no change had been observed among school dropouts. Akerlof and Kranton (2000, 2010) proposed a model that helps to understand occupational segregation. They assumed that each individual should follow the social norms, which are associated with a certain social category: either man or woman. Once an individual deviates from these prescribed behaviours, this deviation results in disutility and also negative externalities for their coworkers. The disutility is a consequence of the fact that not following the norms makes coworkers uncomfortable and they may react and not cooperate with them (Ponthieux and Meurs, 2015). Bergmann (1974) introduced an overcrowding model, which summarizes the im- plications of horizontal segregation. Bergmann argued that traditional views on "roles" of men and women lead to the division of the labour market into males’ and females’ labour markets. When labour market experiences discrimination, demand for female workers decreases, resulting in supply surplus. Consequently, due to the laws of supply and demand, they experience depressed wages for a comparable oc- cupation. Baker and Fortin (1999) showed that horizontal segregation does not have the same impact everywhere. They made a cross-country study between the United States and Canada and claimed that occupational segregation did not have a statis- tically significant effect on women’s wages in Canada. Numerous studies tried to explain how the gender pay gap differs across the sec- tors. It has been observed that the gender wage gap is smaller in the public sector compared to private (e.g. see Arulampalam et al. (2007), Chatterji et al. (2011)). This difference can be attributed to more regulated wages in the public sector. Based on the study of Depalo et al. (2011), Ponthieux and Meurs (2015) propose a stylized fact that “the public-private sector pay gap is positive, particularly in the lower part of the wage distribution, but may be insignificant or negative at the top” (p. 1020). They argue that since women are concentrated in the lower end of the wage distri- bution, they are better off in the public sector, which contributed to the decline in the pay gap. However, de Castro et al. (2013) claimed that the budget crises do have a negative impact on this effect and recedes with high rates of privatization. An interesting approach has been introduced by Goldin (2014), who argued that the gender wage gap is mostly due to within rather than between occupation seg- regation. She showed that for some occupations there is a non-linearity between worked hours and remuneration, which lead to higher gender gap compared to the case when earnings are linear with worked hours. There are some occupations,
- Related Literature 7 where time-adjusted earnings are largely affected by the time spent out of the la- bor market and number of hours worked. Goldin (2014) provides convincing ev- idence that such non-linearity exists when employees are not perfect substitutes, which causes transaction costs to rise. Therefore, employees, who do not have per- fect substitutes, receive wage penalty form reduced working hours. The elimination of this asymmetric pay scheme may significantly reduce or even vanish the wage differential. Throughout the time there emerged a concept of vertical segregation, which in- corporates notions of “glass ceiling” and “sticky floor” effects. Vertical segregation itself describes a set of obstacles women face while climbing up a professional ca- reer. Along with horizontal segregation, it is thought to contribute to the largest part of the gender wage gap. The term “glass ceiling” was introduced by Albrecht et al. (2003) and they referred to the phenomenon when women face limited ca- reer prospects after the certain point. Using Swedish microdata from 1998, they showed that the gender pay gap was increasing throughout the wage distribution, however, the distribution was characterized by a drastic increase in the upper tail. On the other hand, Booth et al. (2003) studied British Household Panel Survey for 1991-1995 and observed that women are as likely to be promoted as men but after promotion, they may receive a smaller increase in wages compared to men. This phenomenon has been labelled as a "sticky floor" effect. Unlike from the “glass ceil- ing” effect, which is generally observed in the upper tail of wage distribution, the “sticky floor” effect is evident if the gender wage gap increases at the lower tail of the wage distribution. To study how career and outside opportunities are related to each other, Lazear and Rosen (1990) developed a model, which assumed that the differences in the pro- motion opportunities at jobs that require specific training can be ultimately blamed for the gender pay differential. Even though women and men might have the same labour market ability, women are more likely to stay away from the labour market due to their higher ability in domestic work. Therefore, employers are not willing to train and promote women as much as they are in the case of men. Arrow (1973) and Phelps (1972) proposed a concept of statistical discrimination, which reflects the consequences of imperfect information about productive charac- teristics of economic agents. The fact that employers are not willing to hire and promote women because they tend to have higher career interruption rates can be used as an example of statistical discrimination. However, it is not always that easy to distinguish statistical discrimination from the human capital model. This difficulty is easily explained by feedback effects: since women expect fewer career promotions, which is due to their employer’s misconceptions about labour supply behaviour, they are less motivated to invest in careers. Goldin (2013) proposed a complementary model of statistical discrimination, taking into account the working environment and employees’ preferences on gender composition. The model im- plies that men will be against women joining male-dominated occupation because it devaluates the occupation and makes it less prestigious. In addition to discrimi- natory factors that explain the gap in the promotion, Cannings and Montmarquette (1990) found that men are more likely to use informal connections for promotion, while women tend to follow formal framework and therefore they must wait longer for promotion.