The gender wage gap in Italy : Study on the changes in the wage gap during the period of financial crisis

The gender wage

gap in Italy

BACHELOR THESIS WITHIN: Economics NUMBER OF CREDITS: 15hp

PROGRAMME OF STUDY: International Economics AUTHOR: Raneem Jisri & Boguslawa Aleksandra Stec JÖNKÖPING May 2020

Study on the changes in the wage gap during the

period of financial crisis

Bachelor Thesis in Economics

Title: The gender wage gap in Italy: study on the changes in the gap during the period of financial crisis

Authors: Boguslawa Aleksandra Stec & Raneem Jisri Tutor: Andrea Schneider

Date: May 2020

Key terms: gender wage gap, financial crisis, pay gap, Oaxaca-Blinder decomposition, Italy

Abstract

Everywhere around the world, whether in developing or developed countries, women earn less

than men. This phenomenon is in no way new and it has been investigated for many years. Still,

in today’s modern society, the wage gap does not appear to be closing. In times of economic

instability, such as the economic crisis, the progress towards equality may be pushed back, since

specific groups, sectors, and occupations may be affected differently. Therefore, the purpose of

this study is to investigate the Italian gender wage gap with a closer look at the fluctuations

during the period of the financial crisis. In order to analyse and understand the fluctuations of

the pay gap, the three main theories used in the research are the human capital theory,

occupational segregation, and theories regarding the labour market structure. By applying the

Oaxaca-Blinder decomposition method, this study analyses to what extent the gap could be

explained by differences in observable characteristics, such as level of education or age, and

how much remains unexplained. The empirical model is applied to the Italian Survey of

Household Income and Wealth (SHIW) microdata between the period of 2002 and 2016. The

main findings show that the Italian gender wage gap, for the most part, remains unexplained.

This indicates that the differentials in pay cannot be accounted for by differences in observable

characteristics, such as education, age, contract type. The results of this research show that the

Italian wage gap was, to some extent, negatively affected by the financial crisis. Furthermore,

implemented austerity measures were found not to have significant negative impacts on the

gap, which only increased in the initial phase of the crisis.

Table of Contents

1. INTRODUCTION ... 1

2. LITERATURE REVIEW ... 3

2.1

H

UMAN

C

APITAL

T

HEORY

... 3

2.2

L

ABOUR MARKET STRUCTURE

... 4

2.3

O

CCUPATIONAL

S

EGREGATION

... 5

3. THE INSTITUTIONAL BACKGROUND OF THE FINANCIAL CRISIS ... 7

4. HYPOTHESES ... 10

5. EMPIRICAL MODEL AND DATA ... 11

5.1

D

ATA

... 11

5.2

O

AXACA

-B

LINDER DECOMPOSITION

... 11

5.3

V

ARIABLES INCLUDED IN THE REGRESSION

... 13

6. EMPIRICAL RESULTS ... 16

6.1

D

ESCRIPTIVE

S

TATISTICS

... 16

6.2

W

AGE

D

ETERMINANTS

... 17

6.3

T

HE

G

ENDER WAGE GAP IN

I

TALY

... 18

6.3.1 General decomposition output ... 19

6.3.2 Distinction between public and private sectors ... 22

7. CONCLUSION ... 26

REFERENCE LIST ... 28

1. Introduction

“If I had a nickel for every time someone told me” the gender pay gap is a myth,” I may have made back the income I’ve lost over the years for being a woman” (Salam, 2019).

It is not a myth, and every day around the world, women suffer from discrimination and inequality in various shapes or forms. The inequalities in pay between men and women are just one example. Not only does it bring multiple disadvantages to women, but it is also a persistent phenomenon, affecting all countries, whether developing or developed (Blau & Kahn, 2003). The issue of the gender wage gap is not in any way new. It has been investigated for years with the aim to understand the problem and eventually eliminate it. Additionally, it has been an important aspect of European legislation and policy (Plantenga & Remery, 2006). Despite increasing research and attention towards the subject, the issue is still persistent. The unadjusted gender pay gap is the difference between the average earnings of female and male workers, which is usually expressed in percentage (European Commission, 2018). In Italy, over the years, the wage gap has been decreasing. As of 2017, the unadjusted wage gap in Italy stood at 5%, compared to the European average of 16% (EIGE, 2019). However, this trend reversed with the beginning of the Great Recession in 2008. Unlike in many other European countries, it led to an increase in the Italian gender wage gap (Eurostat, 2020).

In order to analyse the possible factors that can contribute to the inequalities in pay between men and women, multiple theories have been developed. Firstly, the human capital theory - explaining the wage gap by differences in accumulated human capital. These differences are, in turn, explained by the individual characteristics and productivity differences such as experience and education (Plantenga & Remery, 2006). Secondly, there seem to be differences in the vulnerability of men and women due to the influence of occupational segregation, which results in women clustering into occupations often perceived as lower-paying (Boll et al., 2016). In addition, increased commitment to parenting is associated with higher part-time rates among women. Therefore, the over-representation of women in part-time work could worsen the wage gap since part-time work is not perceived as equally efficient as full-time work (Boll et al., 2016).

In times of economic downturn, countries go through many changes. Whether this is in the labour market, such as increasing unemployment rates, or in public finances. Economic shocks, as such, can also play an essential role in fighting the inequalities, and may even slow down the progress. The Great Recession that began in 2008 had affected numerous countries worldwide. However, Europe was one of the greatly affected regions, including Italy (Smith & Villa, 2014). This period of recession in Italy has been termed as a “double-dip recession” as it recorded two phases (Figari & Fiorio, 2015). The first phase began in 2008, with the beginning of the Great Recession. While the second phase, the sovereign debt crisis, began in 2011 and lasted until 2014 (Figari & Fiorio, 2015). The impact of the recession on Italy can be viewed from multiple different aspects. Firstly, the negative effect on unemployment rates. Just in the first year, these rates increased from being 6.7% to 7.8% (Istat, 2010). Furthermore, during the period of double-dip recession, the real GDP declined by almost 13.5%, while the GDP per capita decreased by approximately 17% (Figari & Fiorino, 2015). Similarly, to many European countries, the Italian public debt had drastically increased to 120% of GDP in just the first year (OECD, 2011). This shows that the effects of this crisis were very severe, and Italy is recovering to this day. In the first phase of the crisis, in multiple European countries, the wage gap has declined, however, in the case of Italy, the gap has increased (Eurostat, 2020). In order to combat the crisis, the Italian government introduced several policies. The majority aimed at reducing the size of the public sector and were not implemented until the second phase of the crisis (Bordogna and Neri, 2014). Nonetheless, in the first phase, even though moderate, the government introduced fiscal stimulus packages (OECD, 2011). In contrast, during the second phase, the policies included public sector wage cuts, wage freezes, workers cuts, and pension reforms. However, these policies may affect men and women differently, and therefore affect the wage gap (Bettio et al., 2012). This aspect is especially important with regard to sectors that are considered more dominated by women. A prime example of which is the public sector, which has been affected differently by the austerity measures (Unicri, 2015). Moreover, the government did not conduct

any assessments to measure the different potential impacts of the policies, which, as a consequence, could instead lead to more inequalities (Kushi & McManus, 2018).

There have been many studies on the subject of pay inequalities in Italy. There has been notable attention towards the impact of educational attainment (Addabbo & Favaro, 2011; Castagnetti et al., 2018; Mussida & Picchio, 2014a). Some studies focused on different major choices among college students and their potential impact on the wage gap (Piazzalunga, 2018). Additionally, there are also studies which aim to look at the gender wage gap over time (Mussida & Picchio, 2014b), or the occupational choices (Gaiaschi, 2019; Cutillo & Centra, 2017). Some studies aim to compare the situation in Italy to other countries (Arulampalam, 2007; Christofides et al., 2013). Labour market mobility was also considered and studied by Del Bono & Vuri (2011). Piazzalunga (2015) shed light on the ethnic impact on the wage gap. While Cutillo & Centra (2017) studied differences in parental commitment. However, there are not many studies focusing on the impact of an economic crisis on gender inequalities in pay. Therefore, the thesis aims to contribute to the existing literature by shedding light on the importance of understanding the potential impact a financial downturn can have on the crucial topic, such as inequalities in pay. The literature on the issue of the gender wage gap is extensive. A few studies are analysing the effect of economic downturns on the labour market, however, not many pay attention to how an economic shock may affect the wage gap, and how it may set back progress towards equality. Understanding the potential impacts, and how it may affect a country through various channels is crucial as it may be necessary for policymakers, in order to aid adequate policies as a response.

The purpose of this study is to analyse the trend of the gender wage gap in Italy, in addition to potential changes during the period of the financial crisis. More specifically, to look into the initial increase, during the first phase of the crisis, and how the gender wage gap changed between the public and private sectors. Therefore, with the use of Italian data on wages and important determinants, this paper will estimate and analyse the gender wage gap in Italy, and how it changed in the period of the financial crisis. For the purpose of measuring the gender wage gap in Italy, microdata provided by the Bank of Italy is used. The Survey on Household Income and Wealth (SHIW) is a large dataset providing multiple essential aspects of household income, as well as savings (Banca D’Italia, n.d.). To measure the gender differences with regard to wage, the information provided in SHIW is used to generate the necessary wage determinants later used in the empirical study. The method followed Oaxaca-Blinder decomposition. This methodology estimates two separate regressions for men and women that, upon subtracting, give the adjusted wage gap. Furthermore, a separate decomposition is constructed for both public and private sectors in order to measure differences between the sectors, as it is expected to differ significantly.

The main findings of this research are that the observable characteristics, such as education or age, cannot explain the gender wage gap in Italy, and for the most part, it remains unexplained for all the years. These results are consistent in the estimation of the general gap, as well as for both the private and the public sectors. Furthermore, it is found that the wage gap is much more significant in the private sector, which did not seem to fluctuate greatly over the years. Moreover, the financial crisis appears, to some extent, negatively affect the gender wage gap. Primarily in the first phase of the crisis, as the gender wage gap increases. The austerity measures implemented in the second phase of the crisis to counteract the effects of the financial crisis, do not appear to significantly affect the gender wage gap since it only increased in the initial phase of the crisis.

The remaining thesis is structured as follows: section 2 provides the theoretical background by reviewing the existing literature. Section 3 includes a short background on the financial crisis in Italy, and the response of the Italian government. The hypotheses will be introduced in section 4. Section 5 will present the empirical model, as well as the data used, and the variables included in the study. The main findings of this paper will be presented in section 6, followed by an analysis. Section 7 concludes the paper and provides implications for future studies.

2. Literature Review

The purpose of this section is to provide some background on the subject of the gender wage gap. It covers relevant theories, past literature, and some explanations as to why the gender wage gap exists.

2.1 Human Capital Theory

There are multiple theories and explanations for the existence of differences in pay. One of the most popular theories is the human capital theory. It is a comprehensive theory that could be considered as one of the primary explanations as to why the gaps exist. According to this theory, every individual accumulates different levels of capital throughout their lives, through education, training, or experience (Grybaite, 2006). The importance of this theory lies in its implication that the wages are reflected in productivity, which stems from the accumulated human capital (Beblo et al., 2003). It thus enables understanding to what extent can the wage gaps be explained by the contrast between men and women with regards to their accumulated human capital. One of the first studies on the gender wage gap using the human capital theory was conducted by Jacob Mincer (1995), considered one of the founders of the theory. In his study, Mincer found that human capital is not only important in terms of labour economics, but also the growth theory. His method went on to be one of the most used methods in estimating the gender wage inequalities.1 _{After Mincer, there have been many studies such as (Polachek 1981, 2004;}

Polachek & Xiang 2014; Addabbo et al., 2012), studying gender inequalities in pay with the help of human capital theory. One of the often proposed explanations comes from the gender differences in commitment to parenting. It is often perceived that women are not only more likely to interrupt their employment, but also not participate in the labour force due to family reasons (Mussida & Picchio 2014b; Signorelli et al., 2012).

As a consequence, women accumulate less work experience than men and eventually, loss of human capital (Beblo et al., 2003). For Italy, this can be observed in Table 1 in the Appendix, where the descriptive statistics are presented.2 _{Women accumulate, on average less work-related experience than}

men, and in this case, it is a difference of two years. This can commonly be seen in Italy, as it is among the European countries in which men are perceived as the “breadwinners” and women are perceived as the “primary caretakers” (Signorelli et al., 2012). As a consequence, many women interrupt or change their employment status (Polachek & Xiang, 2014). Moreover, a bigger commitment to parenting by women is also often associated with women choosing jobs that are described as more flexible and with fewer working hours (Mussida & Picchio, 2014b). Thus, higher part-time rates among women (Boll et al., 2016). “As long as women are trapped into traditional roles by social norms,” even the flexible contracts contributing to gender equality will not be enough (Mussida & Picchio, 2014b, p.1104). A large percentage of women employed part-time brings several disadvantages, further widening the gender wage gap. One of the disadvantages of working part-time is that, from the employer’s perspective, it can be viewed as less efficient than full-time work (Boll et al., 2016). As a consequence, the part-time workers get paid less, even in case of similar human capital factors such as education or experience (Boll et al., 2016). The share of females enrolled in part-time work is higher than that of men. In Italy, on average, the share of females working part-time was 23 percent higher than that of male share in the last 20 years (OECD, 2019). Furthermore, as part-time employees have lower wages and the higher rates of women working part-time, it therefore, contributes to the existence of the gender wage gap (Boll et al., 2017).

One of the most important aspects covered by the human capital theory regards the differences in educational attainment. According to the European Commission, “education is the single most important

1 _{His regression based on human capital theory is also a basis for the Oaxaca-Blinder methodology.}

2 _{The experience variable is only an approximation of the potential experience rather than the actual experience. It}

was estimated through taking an individual's age and subtracting the age at which they got their first job. This does not include any potential breaks in the employment, or parental leave. Therefore, was not included as an independent variable in the regressions later on.

observable characteristic explaining the current level of inequality” (European Commission, 2005, p. 185). When both men and women experience an increase in the education level, they end up in more similar jobs, and occupations than it would be expected, and the gender gap is also decreasing (Addabbo & Favaro, 2011). However, similarly to other aspects of the human capital theory, women and men face different opportunities regarding their education. The decision made by women to pursue higher education is, to some extent, affected by social norms and expectations. Additionally, if women anticipate an interruption in their careers in the future, it decreases their returns and also their motivation for investing in their human capital, such as education (Grybaite, 2006; Goldin & Polachek, 1987). Furthermore, while having higher levels of education is often associated with higher wages, it may not always be the case. Many studies attempted to understand this phenomenon, especially in Italy. These studies found that it is usually women with lower education levels that suffer from more significant pay penalties, than the higher educated women (Mussida & Picchio, 2014a; Addabbo & Favaro, 2011; Piazzalunga, 2018). Therefore, the wage gap may be more significant in occupations with lower educational attainment levels. On the other hand, the number of women with a higher level of education, that is post-secondary, is greater than that of men (Mussida & Picchio 2014b). While women, on average, have higher rates of educational attainment, at the upper secondary and university level, they also accumulate less experience than men, possibly contributing to lower wages among women (Blau & Kahn, 2000). However, this can differ drastically between sectors and occupations.

The human capital theory is one of the most important theories as there are significant differences between men and women with regard to their accumulated human capital. Understanding the underlying elements enables understanding part of the wage gap that arises due to differences in the observable characteristics such as educational attainment level. Numerous studies attempt to understand and describe the underlying factors causing income inequalities between men and women by using the human capital theory. However, this theory alone is not enough, as it does not account for all the reasons and possible fluctuations behind the pay gap (Grybaite, 2006; Blau & Kahn, 2000). Therefore, when attempting to account for changes in the gender wage gap, with a closer look at the impact of an economic downturn, additional theories should be introduced.

2.2 Labour market structure

Labour force participation among women is especially important when considering the case of Italy. Despite the number growing in the past years, women’s labour force participation in Italy is still low - currently standing at approximately 42% (Istat, 2020). In comparison with the European average of approximately 51%, it is quite low (World Bank, 2020). This is commonly attributed to the perception of males being the “breadwinners” in a household. (Signorelli et al., 2012; Mussida & Picchio 2014b). However, if males are the sole income earners in a house, it might also make Italian households more vulnerable in times of economic recession.

During the period of a financial crisis, unemployment, as well as employment rates, tend to fluctuate more than usual. It is, therefore, important to take into account how these changes occur and what they implement as they might affect the gender wage gap. To account for various changes in both the unemployment and employment, terms like “added worker effect” and “discouraged worker effect” are used. Situations in which women’s labour supply increases as a response to their husband’s loss of job in order to counteract the potential financial losses that would occur, refer to as the “added worker effect” (Bredtman et al., 2018). The “discouraged worker effect” has the opposite effect. It refers to situations in which an individual, due to their beliefs about the lower availability of jobs, becomes discouraged and therefore leaves the labour force, decreasing the labour supply (Brodolini & Lyberaki, 2011). While the “added worker effect” can contribute to the increase of labour supply, the “discouraged worker effect” can have the opposite effect. Through the “added worker effect” the women labour supply increases, corresponding to an increase of the women’s labour force participation. However, this also means that women agree to work for lower pay, and possibly part-time, since it has been used as a “source of job growth in previous recessions” (Brodolini & Lyberaki, 2011, p. 14). This can, therefore, lead to an increase in the gender wage gap because of lower wages (Boll et al., 2017). Both of these

effects could be accounted for by performing correction into self-selection. However, due to some restraints of the data, it will not be performed in this thesis.3

Moreover, there is a significant difference between men and women in the labour market, as the indicators do not always behave the same way (Smith & Villa, 2014). According to the data from the World Bank, the employment rate for men in Italy decreased by approximately 2.4 percentage points between 2008 and 2010. It decreased again by approximately 3 percentage points between 2010 and 2014 (World Bank, 2019b). The employment rate for women decreased by 0.97 percentage points between 2008 and 2010, while it increased by approximately 0.4 percentage points between 2010 and 2012. It decreased again by 0.568 percentage points between 2012 and 2014 (World Bank, 2019a). The unemployment rate has fluctuated with similar regard. For women, it has increased by 1.1 percentage points between 2008 and 2010 and decreased by 4.2 percent between 2010 and 2014 (World Bank, 2019d). The unemployment rate for men increased by two percentage points between 2008 and 2010 and increased by approximately 4.4 percentage points between 2010 and 2014 (World Bank, 2019c). From the statistics above, it can be noticed that the overall employment between the period of 2010 and 2014 decreased by 5.6 percentage points for men, while for women, it has only decreased by 1.166 percentage points. This was mostly due to changes in working hours that decreased for both men and women, and due to fixed contracts not being prolonged (Brandolini et al., 2011). In times of an economic recession, the labour market tends to fluctuate greatly. However, some employees may be considered as a risk group and being more likely to be laid off first. Among the most significant factors for what is considered a risk-group are age, contract type, and performance (Wagenaar et al., 2015). It is most likely that people who are young, on temporary contracts and with low work performance, that are at risk of being dismissed (Wagenaar et al., 2015). The fluctuations in the labour market indicators are important in our study as they may also affect fluctuations in the estimated components of the wage gap.

2.3 Occupational Segregation

Segregation has been considered an important factor affecting not only the inequalities between men and women, but has also been portrayed as responsible for constraints on women's careers (Grybaite, 2006). Occupational segregation refers to an “observation” that women are prone to cluster into specific occupations and sectors (Boll et al., 2016). Furthermore, these occupations are often characterized as having lower pay than those jobs that are described as typical for men, which then contributes to gender inequalities in pay (Boll et al., 2016). As a result of occupational segregation, female workers are more likely to work with other women, while male workers with other men (Olsen & Walby, 2004).

Moreover, one of the sectors that are often referred to as heavily employed by women is the public sector (Olsen & Walby, 2004). This is also the case in Italy. While the public sector is often small, it is also more female-dominated. In Italy, the size of the public sector amounted to 17% of the total employment in 2013 (OECD, 2015). In our data, on average, the public sector accounted for 20% of the total employment. Furthermore, what can be observed in Table 1 in the Appendix is that, on average, 30% of employed women work in the public sector. In comparison, the percentage of employed men stands at around 20%.

In most of the studies on occupational segregation, it is pointed out how it might work against women, as it most often contributes to a larger wage gap (Grybaite, 2006; Boll et al., 2016; Boll et al., 2017). On the other hand, some claim that it might instead be benefiting women, as they may be shielded from job loss since they are protected from competing with men, and placed in the so-called “protected sectors'' (Rubery & Rafferty, 2013). These protected sectors may also be more shielded in times of economic downturns. This phenomenon is sometimes referred to as the “silver lining effect”, which means that, because women’s’ jobs are portrayed as “protected” and occupations as “insulated”, women

3 _{However, in order to account for these effects, we will rely on analysis performed by Piazzalunga and Di}

will experience lighter unemployment effects (Brodolini & Lyberaki, 2011). However, the case of “silver lining effect” might have different outcomes when undertaken austerity measures have a rather negative impact on what would be believed as a “protected sector”. This might be the case in Italy, as the main austerity measures undertaken by the government during the financial crisis were aimed towards reducing the size of the public sector. Working in the public sector is not only associated with higher wages (see Figure 1), but it is also highly employed by women. Therefore, in our estimation of the general gender wage gap, we will include a separate variable accounting for the public sector. The empirical method is also applied separately between the public and private sector in order to see the hypothesized differences.

3. The Institutional background of the financial crisis

The financial crisis that began in 2008 was a global crisis that affected many economies, especially in the European region. Italy suffered from what is termed as a double-dip recession (Figari & Fiorio, 2015). The first dip occurred in the period of the financial crisis, starting in 2008, while the second dip refers to the Sovereign debt crisis that began in 2011 (Figari & Fiorio, 2015). As a response, various actions have been taken by the Italian government in order to minimize the effects of the crisis on the different sectors. In the first phase, measures implemented regarded financial stimulus packages (OECD, 2011). While in the second phase, the Italian government began implementing austerity measures, for which the key target was the public sector (Bordogna & Neri, 2014). Additionally, introducing some improvements in the tax collection systems as well as spending cuts (Figari & Fiorio, 2015; OECD, 2011). However, this led to a decline of 13% in the household real incomes during the period of 2007 to 2013, resulting in higher levels of poverty (Brandolini, 2014). It is also important to realize that the situation of the per-capita GDP was even worse, that by the end of 2014, had decreased by 16% (Figari & Fiorio, 2015).

However, the financial crisis did not only affect the GDP and the public debt but also the employment rates. This is due to several reasons: the wage supplementation scheme, reduction of working time, and the increase of the numbers of firms shutting down after the second phase of the financial crisis (Figari & Fiorio, 2015). The employment rate in Italy decreased by around 1.5%. However, approximately half of this decrease was due to changes in fixed-term contracts that, upon termination, were not renewed (Brandolini et al., 2011). The total unemployment rate, on the other hand, rose from around 6.7% in 2008 to 12.7% in 2014 (ILOSTAT, n.d).

Moreover, there is a notable difference in the wages of the public and private sectors (see Figure 1). This could be due to the policies that were implemented during the period of recession, such as the “wage freeze” and the ending of the public sector contract renewals (Figari & Fiorio, 2015). While the wages in the private sector have been increasing for both men and women, in the public sector, the trend was slightly different. The average hourly wages for men in the public sector have been increasing, while for women, the wages decreased between 2008 and 2010. From 2012, after the implementation of the wage freeze, they remained stable without either decreasing or increasing (see Figure 1).

5 6 7 8 9 10 11 12 13 2002 2004 2006 2008 2010 2012 2014 2016 Av er gae Ho ur ly wag e

public/men public/women private/men private/women

Combating the crisis

During the first phase of the crisis, i.e., in 2009, the Italian government response was considered to be moderate (Figari & Fiorio, 2015). In order to counteract the effects of the crisis, similarly to other European countries, the Italian government introduced fiscal stimulus packages that were considered the lowest among other OECD countries (OECD, 2011). During the first phase, the fiscal deficit in Italy amounted to 5.2% of GDP, and the public debt increased to 120% of GDP (OECD, 2011). The effect of stimulus packages can have a different impact on men and women. During the period of the financial crisis, the government did not conduct any assessment to measure the policies impact on the genders (Kushi & McManus, 2018). As a consequence, the packages turned out to be more beneficial for sectors deemed as male-dominated, and hence the male workers, while less for females (Kushi & McManus, 2018).

During the period of the financial crisis, the public sector became the primary target for the Italian government that introduced three significant austerity measures (Bordogna, 2013). Firstly, policies aimed towards reducing the number of employees in the public sector. Secondly, the “wage freeze” aimed at public sector employees. Lastly, a pension reform system (Bordogna, 2013). However, as noted by Bordogna (2013), most of the measures have been adopted without considerations or negotiations with the union. To reduce the number of employees in the public sector, the government increased the replacement ratios. In 2009 the ratio stood at 10%, while in 2012, it increased to 50% of the number of people retired in the previous years (Bordogna & Neri, 2014). Furthermore, measures affecting temporary workers were introduced that meant a decrease in resources for hiring employees on such contracts (Bordogna & Neri, 2014).

The second policy concerned wages and salaries. In the period between 2011 and 2012, the Italian government introduced the “public sector wage freeze”, which meant that the salaries and wages of public employees were frozen at the level of 2010, that could not be exceeded (Bordogna, 2013). This period was later extended to 2014. As a result, public sector employees could not bargain for their wages. Additionally, it prevented any increase in salaries, even in cases of a promotion or seniority (Bordogna et al., 2013). According to Bordogna and Neri, “ No possibility of recovery at the end of the period exists with the consequences for career promotions and future pension payments” (Bordogna & Neri, 2014, p. 363). These measures could imply discrimination between the employees of the private and public sectors since no such measures were introduced in the private sector (Bordogna et al., 2013). However, the Constitutional Court replaced these measures in July of 2015, as it was considered “illegitimate” (Piazzalunga & Di Tommaso, 2019). There were also five percent cuts for the gross salaries between 90 and 150 thousand Euros a year, and cuts of 10 percent for the gross salaries that exceed 150 thousand Euros4_{. The effects of the policies were noticeable as early as in 2010, where the}

number of public employees decreased by four percent, compared to 2008 (Bordogna, 2013). Furthermore, the austerity measures undertaken to decrease public employment, as well as salaries and wages, may instead lead to increased inequalities between genders (Kushi & McManus, 2018). Based on these observations, our study puts a specific focus on the differences between the public and private sectors.

4 _{In our data, the majority of men earned less than 90 000 Euros per year, and even less women. Furthermore,}

barely any men or women earned above the 150 000 Euros mark. Therefore, these cuts will not have a major impact on our estimation of the gap.

Another sector affected was the education sector. According to UNESCO, in Italy, the percentage of teacher employees who are female is approximately 74% (UNESCO, n.d). In the period of the financial crisis, the education sector lost approximately 129 thousand jobs (Périvier, 2014). Figure 2 represents the teacher employees between 2002 and 2016, where approximately 80% of teacher employees are women. Furthermore, since 2010 the female employees’ rates have been decreasing, while for male employees, it is more stable throughout the years.

Figure 2: Teacher employees, 2002-2016; Source: SHIW, own calculations

50 100 150 200 250 300 350 400 450 500 2002 2004 2006 2008 2010 2012 2014 2016 N umb er o f te ach er e mp lo ye es

4. Hypotheses

Based on the literature review in section 2 and the institutional background on the financial crisis in section 3, we raise the following three hypotheses that will be tested in the empirical analysis in section 6.

Human capital theory suggests that individuals’ wages are reflected in their productivity, which arises from accumulated human capital. Based on the theory, education and the accumulated experience are expected to affect individuals’ wages. However, the effects are expected to differ between genders. Firstly, education will be directly controlled for, and it is expected to have positive effects on wages. Additionally, women are expected to have higher educational attainment than men. Secondly, based on theory, the accumulated experience should also have a positive effect on wages. However, due to a higher commitment to parenting, and higher part-time participation rates among women, it is expected that women will accumulate less experience, and hence have lower wages. Therefore, based on this theory, the first hypotheses are:

Hypothesis 1a: Education has a positive effect on wages.

Hypothesis 1b: Since we cannot control for experience level, that is expected to have a positive impact on wages and is, on average lower for women, it is expected that women will have lower wages than men.

According to occupational segregation theory, women tend to cluster into specific occupations and sectors. These are often described as having lower pay than occupations deemed as male-dominated. This is expected to result in women having lower wages than men. Furthermore, one of the sectors which is considered to be mostly employed by women is the public sector. However, working in this sector is also connected to higher wages. Therefore, based on the public sector being highly employed by women, and higher wages, the wage gap is assumed to be lower than in the private sector. Therefore, the second hypothesis is:

Hypothesis 2: The gender wage gap is expected to be higher in private than in the public sector.

With the start of the second phase of the financial crisis, between 2011 and 2014, the Italian government began implementing austerity measures aimed at reducing the size of the public sector. Including wage freeze, employee and wage cuts, as well as the pensions reform system. Accordingly, with theory, the public sector is deemed as a female-dominated sector. Furthermore, the implemented austerity measures are expected to negatively affect wages of the public sector, since the government did not assess the potential impact of said measures. Hence, it is expected to have a negative impact on the gender wage gap since women are expected to be more negatively affected. Therefore, the third hypothesis is:

Hypothesis 3: As a result of the undertaken austerity measures, the gender wage gap is expected to increase in the second phase of the crisis.

5. Empirical model and data

In this section, the empirical model used is presented, as well as data and variables that were chosen for this study.

5.1 Data

In order to study the inequalities in pay between Italian men and women, the microdata provided by the bank of Italy is used. The Survey on household Income and Wealth (SHIW) is a large dataset carried out by the Bank of Italy. The data gathered in SHIW contains information about Italian households in many aspects, including both savings of the households and the incomes (Banca D’Italia, n.d.). This is a broad dataset, and in the most recent surveys, the sample consists of around 8000 households, which corresponds to roughly 20 000 individuals.

The primary advantage of using this data is that it offers information on many essential wage determinants, such as income, age, job sector, and the type of contract. Moreover, it follows the same households over the years, with some exceptions when some households decide to withdraw from the study. This allows accounting for changes in salaries, or the wage determinant variables through time. However, there are also some limitations to the dataset provided by the Bank of Italy. One of the limitations of the data is that it does not provide individuals experience with respect to work. Nonetheless, it provides the year at which a person began working, which can allow for estimating an approximate of experience.

To fully understand what, if any, impact the financial crisis and implemented policies had on the gender wage gap in Italy, the timespan chosen for the studies is between 2002 and 2016. The data is performed every two years, which gives multiple reference points to compare the gap and look for trends, if any, and further deepen the analysis. The sample chosen for the study consists of 15 to 65-year-old employed men and women. In this study, self-employed, unemployed, or retired people are not included. In the end, it translated to approximately 5000 observations per year.

5.2 Oaxaca-Blinder decomposition

The method applied in this thesis will follow the Oaxaca-Blinder decomposition, which was developed by Ronald Oaxaca and Alan Blinder (Blinder, 1973; Oaxaca, 1973). It was one of the first methods attempting to study the differences in the labour market by groups. Since then, it has become a prevalent and widely used method to study the wage differentials between men and women (Jann, 2008). Following the Oaxaca-Blinder method, the decomposition is divided into two parts, the “explained '' and “unexplained” components (Jann, 2008). The “explained component” describes the differentials arising due to differences in observable characteristics such as level of education, which stem from the human capital (Jann, 2008). The “unexplained” component measures how much of the wage differential remains unexplained. That is, how much cannot be explained by the differences in the observable characteristics and the returns. The unexplained part of the decomposition was termed by Oaxaca (1973) as the discrimination component. However, this can also include other factors that do not only stem from discrimination (Jann, 2008). Components such as austerity measures, or shifts in the labour market structure, which may affect the wage gap in a way that the model cannot account for, therefore contributing to a more significant unexplained component.

The Oaxaca-Blinder decomposition contains three main elements. Firstly, the variable of interest, which is the natural log of hourly wages. Secondly, various predictors such as age or educational levels, and lastly, the group defining the decomposition, which in this case is gender. The decomposition method estimates two separate regression equations for both men and women, the equation, as described by Blinder (1973) looks as follows:

where Yi corresponds to the natural log of wages, Xji corresponds to X1i,...,Xni which corresponds to n

observable characteristic used to explain the wage differentials, and νi is the error term.

The above equation can be rewritten as follows:

𝑙𝑛𝑊_' = 𝛽_"'+ ∑ 𝛽_#'𝑋_#'+ 𝜀_!' (2) 𝑙𝑛𝑊( = 𝛽"(+ ∑ 𝛽#(𝑋#(+ 𝜀!( (3)

where there are two separate wage equations for both males (m) and females (f). In equation (2) and (3), the variable X is a vector for different observable characteristics, such as age, education, or contract type, and ɛ is the error term. The adjusted gender wage gap is obtained after taking the difference between these two equations.

In order to study the Italian gender wage gap, and its changes during the period of the financial crisis as well as the difference between specific sectors, the decomposition will be performed on two slightly different models. Firstly, Oaxaca-Blinder decomposition will be applied in order to generate the adjusted wage gap over the years. In this general decomposition, there will be no distinction between the public and private sectors. Additionally, two other decompositions will be performed; one for the private, and one for the public sector. The purpose of separating the decomposition is to observe the differences with regards to the size of the gap, as well as the fluctuations over the years. For the general Oaxaca-Blinder decomposition, the estimation of the linear wage equations will be similar to the interpretation of this method by Piazzalunga and Di Tommaso (2019). It will be performed for both males (m) and females (f) separately using the following equation:

𝑙𝑛𝑊_)* = 𝛽_)*𝑋_)*+ 𝜀_)* (4)

where the variable g = (m,f) and t= 2002, 2004,..., 2016. The estimation of equation (4) will be performed separately for each year. In the equation the outcome variable is the natural log of hourly wage (Wg).

The variable Xgt is a vector for the variables selected (age, level of education, contract type, marital

status, area of residence, part-time and public), ßgt corresponds to the coefficients that are obtained from

the regressions, and εgt is the error term. Despite the decomposition being divided between the sectors

in later parts, a dummy for working in the public sector will be included in the general Oaxaca-Blinder decomposition. This variable is included, as working in the public sector is often associated with higher wages. Furthermore, as mentioned previously there are more women than men working in the public sector, therefore the potential effect was also accounted for. This will also increase the accuracy of the explained component as important variables that can be accounted for will not be omitted. If the dummy for working in the public sector was not included, it may lead to overestimatimation of the unexplained component, as part of the wage gap that can be explained by individuals working in the public sector, will not be accounted for. Therefore, underestimating the explained component.

The resulting wage equations will be used to estimate the gender wage gap, which corresponds to the following:

𝐺𝑊𝐺_* = 𝑙𝑛𝑊111111111 − 𝑙𝑛𝑊_'* 11111111_(*

Where lnWmt and lnWft correspond to the estimated mean log hourly wages for males and females

respectively, Xmt and Xft correspond to the average characteristics of male and female workers, ßmt and

ßft are the obtained coefficients estimated from the separate regressions. The subscript t corresponds to

2002, 2004, ..., 2016. The first term in the equations corresponds to the explained component. It refers to the difference in predictors chosen for specification, such as age or education level. It indicates how much of the wage gap results due to differences between males and females with regards to observed characteristics such as age or education. The second part of the equation represents the unexplained component. It captures how much of the wage is attributable due to the difference in returns. The unexplained component captures how much of the gap remains unexplained, meaning not accounted for by the chosen predictors such as education, or contract type.

The second part of the empirical study concerns separate decomposition for the private and public sectors. It is hypothesized that as a result of the crisis, the gap changes differently in the private and public sectors. Additionally, it is expected that the wage gap in the respective sectors will differ in size. In order to see how these changes occurred, the Oaxaca-Blinder decomposition will be applied to both sectors separately. Therefore, the variables used are similar as in case of equation (4), with a small change to the outcome variable. For these decompositions, separate hourly wages are constructed for both the public and private sector employees. To estimate the wage differentials in two sectors, the following equations will be used:

𝑙𝑛𝑃𝑢𝑏𝑊)*= 𝛽)*𝑋)*+ 𝜈)* (6)

𝑙𝑛𝑃𝑟𝑖𝑣𝑊_)* = 𝛽_)*𝑋_)*+ 𝜈_)* (7)

Where lnPubWgt and lnPrivWgt are the outcome variables that correspond to the natural log of the hourly

wage in the public and private sectors, respectively. The subscript g=(m,f) corresponds to males and females, respectively. The subscript t=2002, 2004,...,2016, and νgt is the error term. The variable Xgt is

a vector of the variables chosen (age, level of education, are of residence, marital status, contract type, and part-time). In the decomposition for the separate sectors, the dummy for the public is no longer used as an explanatory variable. However, the remaining variables are the same as in the case of equation (4). The main distinction from the general decomposition is the hourly wages that are now divided by sectors, and there is no public dummy as an explanatory variable. The gender wage gap for the respective sectors is obtained similarly to the general composition, that is, following equation (5).

5.3 Variables included in the regression

Outcome variable

The outcome variable in the regressions is the natural log of hourly wages. It is self-defined based on reported wages in the SHIW data. The hourly wage is obtained by dividing the monthly wage by hours worked per week and multiplying it by 4.3. The multiplier of 4.3 is obtained after dividing the number of weeks in a year by the number of months. Later on, for the decomposition between sectors, the hourly wage is also computed for the public and private sectors. These wages are obtained by creating variables for employees of public and private sectors, respectively, and assigning corresponding monthly wages. Afterward, the monthly wages are also divided by the hours worked per week and multiplied by 4.3 in order to obtain the hourly wages. Lastly, in order to obtain the outcome variable, the natural log is taken of the hourly wages.

Public

This variable is only included in the regression when estimating the general gender wage gap. It is not part of the second decomposition when a distinction between the sectors is made. It is a dummy variable that takes a value of 1 if a person works in the public sector and 0 otherwise. It is self-defined based on the answers by the survey participators to the question about the sector in which they work. The reason for this variable to be included is that, as observed in Figure 1, wages are higher in the public sector, which is expected to affect the gender wage gap. Furthermore, as noted previously, and as can be observed in Table 1 in the Appendix, more women than men tend to work in the public sector, which

together with higher wages associated with the public sector can be attributed to having some effect on the wage gap.

Age

This variable is included as it is part of the observable characteristics that also might have an impact on wages. It is self-defined based on the variable “Anasc” that gave the individuals’ date of birth and subtracting the year of the survey. It is expected to have a positive effect on wages, as wages tend to increase with increased gathered experience, that rises with age. The focus of this paper lies on the age group of 15 to 65-year olds. The reason behind this age group starting at a young age of 15 is due to it being the legal age at which a person can begin working in Italy. While it may be that in some cases, people who report to be working at age below 20 may be doing so alongside studying, contributing to lower hours and wages, it is not expected to affect wages on a large scale.5

Education level

Education is often thought of as one of the most important variables impacting wages. It is also in line with human capital theory, which assumes that there might be more differences in educational attainment between genders. To account for the level of educational attainment, it is divided into four separate categories for which dummy variables are created. These categories are defined based on provided data, under variable “Studio”, which gave the educational level of an individual. Firstly, “primary” takes the value of 1 if a person has no education or primary education level and 0 otherwise. Secondly, “lowersec” takes the value of 1 if a person has a lower secondary education and 0 otherwise. Thirdly, “uppersec” that takes the value of 1 if a person has post-secondary education or non-tertiary education and 0 otherwise. Lastly, “univ” takes the value of 1 if a person has a university degree and 0 otherwise.

Marital Status

This variable is included in line with theories, and since it is also expected to have an impact on wages. In order to account for marital status, three distinctive dummy variables are created, based on the variable “Staciv” from the available data. Firstly, the dummy “married” that takes the value of 1 if a person is married and 0 otherwise. Secondly, a dummy “single” that takes the value of 1 if a person is single and 0 otherwise. Lastly, the dummy “div” that corresponds to 1 if a person is divorced and 0 otherwise.

Region of Residence

The region in which individuals live may also affect wages, as some regions can have higher average wages than others. In order to account for the area in which an individual lives, it is divided into three separate categories for which dummy variables are created. The categories for dummies are defined based on data under variable “Area3”. Firstly, “North” that takes the value of 1 if a person lives in northern Italy and 0 otherwise. Secondly, the variable “Centre” which takes the value of 1 if a person lives in central Italy and 0 otherwise. Dummy “south” takes the value of 1 if a person lives in the south and islands, and 0 otherwise.

Contract type

The contract type is self-defined by the use of variable “Contratt” in the data set. This variable is not only chosen due to possibly having an impact on wages, but also because of each contract type being associated with having different effects on the wages. What is expected is that a person with a temporary or a fixed contract will have lower wages than a person with a permanent contract. In order to represent the contract type, three dummy variables are created. Firstly, the “perm” dummy, which takes the value of 1 if a person has a permanent contract and 0 otherwise. Secondly, “fixed” dummy that takes the value of 1 if a person has a fixed contract and 0 otherwise. Lastly, the “temp” dummy, which takes the value of 1 if a person has a temporary contract and 0 otherwise.

5_{In our data, the amount of people who were below 20 and reported to be employed was very small in all the years}

Part-time

It is another important variable that is assumed to have a negative impact on wages. As suggested by theory, working part-time is usually associated not only with fewer hours worked, but also lower pay. It is self-defined based on the answers by survey participants on the type of contract they have. This variable is constructed as a dummy variable, which takes the value of 1 if a person works part-time, and 0 otherwise.

Female:

This variable is included in the regression outputs to see whether being a female is associated with lower wages. This variable is constructed based on the variable “sex” in the dataset. It is a dummy variable that takes the value of 1 if a person is a female and 0 otherwise.

6. Empirical Results

This section presents the main findings of this thesis. In 6.1, the descriptive statistics and a figure illustrating the Italian gender wage gap are presented. In 6.2, the output for the wage determinants regression is presented. In 6.3, the outputs for the Oaxaca-Blinder decompositions are presented. Section 6.3 is followed by an analysis of the findings in this paper.

6.1 Descriptive Statistics

The descriptive statistics for the variables in the thesis are presented in Tables 1 and 2. Based on the results, it can be noted that educational attainment is higher among women. According to the data, on average, 20.5% of women have a university level of education, while for men, this value stands at approximately 12.7% (see Tables 1 and 2). Referring to part-time rates among women, it is on average, 17% higher than that of men. Furthermore, women work less than men, as the approximate for accumulated experience is lower for all years. Additionally, in all the years, the percentage of women working in the public sector is higher than that of men.

Table 2: Descriptive statistics employed men (15-65 years old)

2002 2004 2006 2008 2010 2012 2014 2016

Monthly wage 1018.495 1076.434 1152.838 1183.384 1209.1 1192.511 1230.575 1269.156

Hourly wage 7.29 7.57 8.43 8.67 8.69 8.94 9.06 9.07

Log Wage 1.82 1.88 1.97 2.01 2.01 2.04 2.07 2.08

Hours per week 34.55 34.48 34.17 34.25 33.94 32.56 32.62 33.75

Age 39.57 40.43 41.28 41.81 42.91 43.75 45.25 46.26 Experience 18.34 18.91 19.56 20.08 20.91 21.92 23.16 23.77 Public sector 0.312 0.308 0.286 0.335 0.338 0.316 0.251 0.205 Part-time 0.18 0.18 0.19 0.21 0.23 0.29 0.29 0.26 North 0.54 0.54 0.55 0.55 0.51 0.50 0.54 0.51 Centre 0.22 0.23 0.21 0.20 0.22 0.23 0.21 0.24 South 0.24 0.23 0.24 0.25 0.27 0.27 0.25 0.25 Married 0.59 0.58 0.59 0.60 0.60 0.59 0.58 0.53 Single 0.32 0.29 0.29 0.27 0.27 0.27 0.28 0.31 Divorced 0.07 0.09 0.10 0.10 0.10 0.10 0.14 0.16 Primary 0.07 0.06 0.04 0.05 0.04 0.03 0.03 0.02 Lower secondary 0.25 0.26 0.27 0.25 0.22 0.22 0.22 0.23 Upper secondary 0.51 0.52 0.51 0.52 0.52 0.51 0.51 0.52 University 0.16 0.17 0.18 0.19 0.22 0.24 0.25 0.23 Observations 2,477 2,470 2,494 2,562 2,499 2,398 2,286 1,975 2002 2004 2006 2008 2010 2012 2014 2016 Monthly wage 1273.80 1375.09 1449.479 1496.975 1517.78 1516.72 1547.705 1560.044 Hourly wage 7.77 8.26 8.93 9.17 9.51 9.50 9.49 9.51 Log Wage 1.90 1.97 2.05 2.08 2.09 2.09 2.13 2.13

Hours per week 39.70 39.96 39.59 39.8 39.01 38.43 38.41 38.71

Age 40.30 40.70 41.18 41.60 42.90 43.75 44.86 44.99 Experience 20.70 20.90 21.26 21.78 22.8 23.80 24.49 24.33 Public sector 0.230 0.202 0.202 0.178 0.228 0.205 0.161 0.124 Part-time 0.04 0.03 0.04 0.04 0.05 0.09 0.09 0.08 North 0.46 0.45 0.48 0.46 0.43 0.43 0.47 0.44 Centre 0.20 0.22 0.19 0.20 0.21 0.21 0.20 0.21 South 0.34 0.33 0.33 0.34 0.36 0.36 0.33 0.35 Married 0.64 0.63 0.63 0.64 0.65 0.66 0.62 0.58 Single 0.30 0.32 0.33 0.32 0.31 0.29 0.32 0.36 Divorced 0.06 0.04 0.04 0.04 0.04 0.04 0.05 0.06 Primary 0.11 0.09 0.07 0.07 0.05 0.05 0.05 0.04 Lower secondary 0.37 0.37 0.37 0.37 0.35 0.35 0.34 0.33 Upper secondary 0.42 0.43 0.45 0.44 0.45 0.46 0.46 0.48 University 0.097 0.11 0.11 0.12 0.14 0.14 0.15 0.15 Observations 3,443 3,452 3,370 3,320 3,130 3,057 2,797 2,418

Figure 2:Gender wage gap, 2002-2016; Source SHIW, own calculations.

Figure 3 illustrates how the gender wage gap looks over the years and how it fluctuates. Based on the results, there is a significant difference between gaps in the public and private sectors. Furthermore, an interesting observation is that in 2008 the gap in the public sector has increased by almost four percent, while for the private sector, it has increased by approximately half. Furthermore, after 2012, the gap in the private sector has been decreasing, while for the public sector, it fluctuated between decreasing and increasing.

6.2 Wage Determinants

In order to observe what impact the chosen explanatory variables have on wages, a separate linear regression is performed. This is performed in order to see the hypothesized results of the wage determinants, which are not visible in the output of the Oaxaca-Blinder decompositions. Once again, the outcome variable is the natural log of hourly wages. The reason behind wages being in the log form is to interpret the percentage change from the chosen variables on the hourly wages. The explanatory variables are age, region of residence, marital status, educational attainment level, contract type, part-time, female, and public. This regression is performed to test the effect of the chosen predictors on the wages. Additionally, the female variable is included to see whether it will have a predicted negative impact on wages.

The results of the regression can be seen in Table 3. Based on the results, it can be observed that the wage determinants such as contract type, part-time and female affect the wages negatively in all the years. Furthermore, as predicted, the level of educational attainment has a positive effect on wages. The age variable is seen to have a positive effect on wages, however, to a lesser degree. The dummies associated with marital status and the area of residence are also associated with having a positive effect on wages in all years.

0,0% 2,0% 4,0% 6,0% 8,0% 10,0% 12,0% 14,0% 16,0% 18,0% 2002 2004 2006 2008 2010 2012 2014 2016

Table 3: Wage determinants for Oaxaca-Blinder decomposition

*p<0.1; **p<0.05; ***p<0.01. In the parenthesis are the robust standard errors. The regression was performed on the natural logs of wages. The base categories for the dummy variables are: single for the marital status, south for the region of residence, permanent for type of contract, and primary for educational attainment level.

Based on the output in Table 3, the variables for the fixed and temporary contracts are significant at 1% level for all years, and both are associated negatively with the wages. For the year 2010, a fixed contract is associated with wages being lower by 51.4%. The variable part-time is significant for all years at a 1% level, except for 2008 and 2014, where it is found insignificant (see Table 3). For 2010 working part-time is associated with wages being 3.4% lower. The variable for female is significant at 1% for all years of the analysis. For 2010 it can be observed that being a female is associated with wages being lower by approximately 13%. The educational attainment level has a positive effect on wages, and the higher the educational level, the greater the effects on wages. All the dummy variables for educational level are significant at 1%, except for lowersec in 2016, which is significant at 5% (see Table 3). Interpreting the results for 2010, having a university degree is associated with wages being approximately 61% higher. Regarding the marital status, the variable married is significant at 1% for all the years, while the variable divorced is significant at 1% for majority of years, except 2012, 2014, and 2016 where it is significant at 5% and 10% respectively. Interpreting the results for 2010, being married corresponds to wages being 9.9% higher. Based on the output in Table 3, both of the variables for north and centre are significant for all the years, at 1%, with an exception for years 2002 and 2012 where centre is significant at 5% and 10%, respectively. In 2010 living in the north means on average wages being 9% higher. Despite being significant at 1% for all years of interest, the age variable affects wages to a small extent, as it is only associated with around 1% change in all the years.

6.3 The Gender wage gap in Italy

The following section begins with a presentation of the results for the general Oaxaca-Blinder decomposition, and the interpretation of its elements. Later on, it will continue onto the main results regarding the decomposition for the separate sectors.

2002 2004 2006 2008 2010 2012 2014 2016 Age 0.012*** (0.00) 0.009*** (0.00) 0.010*** (0.00) 0.010*** (0.00) 0.009*** (0.00) 0.011*** (0.00) 0.08*** (0.00) 0.008*** (0.00) North 0.146*** (0.02) 0.127*** (0.01) 0.097*** (0.01) 0.110*** (0.01) 0.091*** (0.01) 0.079*** (0.01) 0.099*** (0.01) 0.092*** (0.02) Centre 0.046** (0.02) 0.079*** (0.02) 0.075*** (0.02) 0.130*** (0.02) 0.065*** (0.02) 0.034* (0.02) 0.062*** (0.01) 0.083*** (0.02) Married 0.100*** (0.02) 0.139*** (0.02) 0.111*** (0.02) 0.110*** (0.01) 0.099*** (0.02) 0.090*** (0.02) 0.133*** (0.02) 0.119*** (0.02) Divorced 0.108*** (0.03) 0.176*** (0.03) 0.110*** (0.03) 0.102*** (0.03) 0.119*** (0.03) 0.062** (0.03) 0.055** (0.03) 0.0484* (0.03) Lowersec 0.182*** (0.02) 0.132*** (0.02) 0.149*** (0.03) 0.146*** (0.03) 0.181*** (0.03) 0.135*** (0.03) 0.148*** (0.03) 0.087** (0.04) Uppersec 0.365*** (0.02) 0.299*** (0.02) 0.337*** (0.03) 0.319*** (0.03) 0.351*** (0.03) 0.317*** (0.03) 0.291*** (0.03) 0.237*** (0.04) Univ 0.651*** (0.03) 0.613*** (0.03) 0.602*** (0.03) 0.584*** (0.03) 0.615*** (0.03) 0.580*** (0.04) 0.551*** (0.04) 0.524*** (0.04) Parttime -0.076*** (0.02) -0.117*** (0.02) -0.067*** (0.02) -0.012 (0.02) -0.033* (0.02) -0.069*** (0.02) -0.028 (0.02) -0.057** (0.02) fixed -0.518*** (0.02) -0.504*** (0.02) -0.483*** (0.02) -0.397*** (0.02) -0.514*** (0.02) -0.433*** (0.0185) -0.406*** (0.02) -0.449*** (0.430) temp -0.296*** (0.06) -0.428*** (0.07) -0.686*** (0.05) -0.503*** (0.05) -0.599*** (0.06) -0.406*** (0.01) -0.615*** (0.04) -0.499*** (0.04) Female -0.127*** (0.01) -0.128*** (0.01) -0.129*** (0.01) -0.128*** (0.01) -0.130*** (0.01) -0.102*** (0.01) -0.112*** (0.01) -0.108*** (0.01) Public 0.144***_(0.02) 0.174***_(0.02) 0.143***_(0.02) 0.161***_(0.01) 0.169***_(0.02) 0.159***_(0.02) 0.166***_(0.02) 0.131***_(0.02)

6.3.1 General decomposition output

*p < 0.1; **p <0.05; ***p<0.01. In the parenthesis are the robust standard errors. The variables controlled for in the wage equation were age, education level, region of residence, marital status, contract type, public and part-time. The coefficients that were used as benchmarks are the male coefficients.

The primary decomposition in this thesis is the estimation of the general wage gap in Italy, which can be observed in Table 4. It can be divided into two main components, the difference, and the decomposition part. The difference component contains the main regression results and the gender wage gap. The gender wage gap is interpreted in percentage; hence, in 2016, the gender wage gap is equal to 0.048, which corresponds to a gap of 4.8% (see Table 4).

The decomposition part can be divided into five separate components. The endowments component represents the average change in female wages if they would possess the same characteristics as males (Jann, 2008). As can be observed in Table 4, the endowments component is negative for all the years of interest. This means that if women possessed the same characteristics, such as education level, as men, their wages would be lower. For 2016 the endowments is equal to -0.073, which means that if females had the same characteristics as males, their wages would on average be lower by 7.3%. The coefficients measure the change to female wages if applying men’s coefficients to women’s characteristics (Jann, 2008). This term is positive for all the years, indicating that male returns to the observable characteristics are higher than that of females. Therefore, applying male returns to the female characteristics would lead to female wages increase. For 2016 the coefficients equals to 0.078, which implies an increase in female wages of 7.8%. The interaction accounts for both the coefficients and endowments existing simultaneously, therefore the effects of both the endowments and coefficients effect at the same time (Jann, 2008). In a few years of the analysis, such as 2002, 2006, and 2014 the interaction effect has a higher p-value, which is above any significance level. However, this indicates that the differences between the two groups in coefficients and endowments simultaneously do not affect the difference in the outcome, or affect it to a small degree. This can be observed in Table 4, as for those years, the interaction effect is small.

However, the main focus of the analysis in the thesis will lay on the two last components, the explained and unexplained components. The explained coefficients measures how much of the wage gap can be explained by the differences between the two groups in the predictors that were presented in the methodology section. This translates to how much of the gap can be explained by differences in

2002 2004 2006 2008 2010 2012 2014 2016 Predicted hourly wages for men 1.902*** (0.009) 1.972*** (0.009) 2.050*** (0.009) 2.079*** (0.009) 2.099*** (0.010) 2.09*** (0.010) 2.125*** (0.010) 2.129*** (0.011) Predicted hourly wages for Women 1.819*** (0.012) 1.884*** (0.012) 1.971*** (0.012) 2.014*** (0.011) 2.014*** (0.012) 2.04*** (0.012) 2.074*** (0.011) 2.081*** (0.012) GWG 0.083***_(0.016) 0.088***_(0.015) 0.079***_(0.015) 0.065***_(0.013) 0.085***_(0.015) 0.057***_(0.016) 0.051***_(0.015) 0.048***_(0.016) Endowments -0.058*** (0.011) -0.064*** (0.011) -0.057*** (0.011) -0.084*** (0.010) -0.058*** (0.012) -0.051*** (0.011) -0.066*** (0.011) -0.073*** (0.012) Coefficients 0.135*** (0.014) 0.115*** (0.014) 0.144*** (0.015) 0.102*** (0.014) 0.119*** (0.015) 0.079*** (0.015) 0.099*** (0.015) 0.078*** (0.016) Interaction 0.006 (0.011) 0.035*** (0.010) -0.009 (0.011) 0.047*** (0.011) 0.023** (0.011) 0.030* (0.012) 0.017 (0.011) 0.042*** (0.013) Explained -0.053*** (0.010) -0.053*** (0.011) -0.059*** (0.010) -0.071*** (0.010) -0.050*** (0.011) -0.044*** (0.011) -0.068*** (0.011) -0.062*** (0.012) Unexplained 0.136*** (0.013) 0.141*** (0.013) 0.138*** (0.014) 0.136*** (0.012) 0.135*** (0.014) 0.101*** (0.014) 0.119*** (0.013) 0.110*** (0.015) Number of obs 5920 5922 5864 5882 5629 5455 5083 4402