(AGENPARL) – gio 22 giugno 2023 Women, labour markets and economie growth
Seminari e convegni
Workshops and Conferences
number
June 2023
F. Carta, M. De Philippis, L. Rizzica and E. Viviano
Women, labour markets and economie growth
Seminari e convegni
Workshops and Conferences
F. Carta, M. De Philippis, L. Rizzica and E. Viviano
Questa pubblicazione raccoglie i risultati di un progetto a cui hanno contribuito ricercatori del
Dipartimento Economia e statistica e della rete territoriale della Banca d’Italia.
Il testo è disponibile nel sito internet: http://www.bancaditalia.it
ISSN 2281-4337 (print)
ISSN 2281-4345 (online)
Le opinioni espresse in questo libro sono personali e non impegnano la responsabilità della Banca d’Italia.
Stampa a cura della Divisione Editoria e stampa della Banca d’Italia
© 2023 Banca d’Italia
Women, labour markets and economic
growth?
Francesca Carta†
Marta De Philippis†
Eliana Viviano†
Lucia Rizzica‡
Abstract
This report summarises the findings of a research project that was carried out
by the Bank of Italy over the past three years. The report provides an overview of
the gender divides in the Italian labour market and a comparison with the other
main EU economies. It then traces the origins of such gaps considering: (i) the educational choices and school-to-work transition, (ii) motherhood and within-family
interactions, (iii) career progressions. The report frames the contributions of the
underlying research in the most recent economic literature and discusses some avenues for effective policy action.
JEL Codes: J16, J13, J11 J22, J24, J31, J71, H31
Keywords: Gender gaps, STEM, School-to-work transition, Child-penalty,
Childcare facilities, Fertility, Tax-transfer system, Gender quotas, Stereotypes,
Wage bargaining.
We would like to thank all the colleagues and external coauthors involved in this research project:
Jaime Arellano-Bover, Audinga Baltrunaite, Federico Barbiellini Amidei, Nicola Bianchi, Francesco
Billari, Giulia Bovini, Ines Buono, Mario Cannella, Alessandra Casarico, Niccol`o Cattadori, Fabrizio
Colonna, Maria De Paola, Silvia Del Prete, Sabrina Di Addario, Davide Dottori, Michela Giorcelli,
Matteo Gomellini, Patrick Kline, Salvatore Lattanzio, Salvatore Lo Bello, Agata Maida, Sauro Mocetti,
Francesca Modena, Berkay Ozcan,
Giulio Papini, Matteo Paradisi, Santiago Pereda Fern`andez, Paolo
Piselli, Annalivia Polselli, Enrico Rettore, Concetta Rondinelli, Raffele Saggio, Paolo Sestito, Mikkel
Sølvsten, Giulia Martina Tanzi, Marco Tonello. We are also grateful to Andrea Brandolini, Michele
Caivano, Federico Cingano, Nicola Curci, Francesco D’Amuri, Emanuele Dicarlo, Silvia Giacomelli,
Alessandro Notarpietro, Andrea Petrella, Massimiliano Pisani, Antonella Tomasi and Roberto Torrini
for providing valuable help for this Report and insightful comments.
Bank of Italy, Directorate General for Economics, Statistics and Research, Family and Labour Market
Division.
Bank of Italy, Directorate General for Economics, Statistics and Research, Law and Economics
Division.
Contents
1 Introduction and motivation
2 An overview of gender gaps
3 Education and school-to-work transition
3.1 Descriptive evidence on gender gaps in education . . . . . . . . . . . . .
3.2 Understanding gender gaps in education . . . . . . . . . . . . . . . . . .
3.3 School to work transition: descriptive evidence on early career gender earning gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Understanding the early career gender earning gaps: a two-step analysis .
study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Policy implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4 Maternity and labour market outcomes
4.1 The “cost” of motherhood in the labour market . . .
4.2 The role of family policies . . . . . . . . . . . . . . .
4.3 The effect of maternal employment on fertility . . . .
5 Family interactions and the role of the tax-transfer system
5.1 Household labour supply and partners interactions . . . . . . . . . . . . .
5.2 The tax-transfer system and the labour supply of married women . . . .
6 Career progressions
6.1 Descriptive evidence . . . . . . . . . . . .
6.2 The origins of the glass ceiling . . . . . . .
6.3 Policies . . . . . . . . . . . . . . . . . . . .
Firm organisation and diversity management . . . . . . . . . . . .
Female leadership and the impact of affirmative action policies . .
7 Discussion and conclusions
A Additional tables and figures
Introduction and motivation
The integration of women into the labour market goes far beyond considerations of equity
and fairness, as there is a strong and undeniable link between women’s participation in
the labour market and economic development. Indeed, the literature has established that
not only economic development is a prerequisite for improving women’s living conditions
but, at the same time, women’s participation is an important driver of economic growth
(see for example Goldin, 1995 and Duflo, 2012 among many).
GDP per capita (USD-PPP; logs)
Figure 1: Female labour force participation rate and per capita GDP in OECD countries
Women labour force participation
Italy
Notes: Estimates of the relationship between GDP per capita (in logs) and female activity rate. Population aged 15-64.
Source: own calculation on OECD data.
Figure 1 shows that the estimated relationship between GDP per capita and female
activity rates in the OECD countries in several selected years (1991, 2001, 2011 and 2021)
is positive.1 Moreover, Figure 1 indicates that Italy (represented by the red triangles)
lags behind many other OECD countries in terms of both participation and GDP per
capita. If we focus only on the European Union, the picture is even worse. In 2022 the
average participation rate of women in the 15-64 age group was 69.5% in the EU and just
56.4 in Italy, the lowest level recorded among EU countries. In quantitative terms, if the
female participation rate in Italy was at the same level as the EU average, there would
be about 2.4 million additional individuals (+10%) in the Italian labour force.
The urge of enhancing women’s labour market participation emerges even more clearly
in the light of the population drop predicted for the future decades. According to the
latest Istat projections, the total population aged 15-64 will count around 6 million
More specifically, the relationship between female labour force participation and GDP per capita is
U-shaped, both when looking at the evolution within a single country from a historical perspective and
when looking at a cross-section of countries at different stages of development. Goldin (1995), among
others, points out that in the early stages of development women work on family farms and in domestic
production. As the economy develops and work is transferred to firms, it becomes more difficult for
women to combine household production with participation in the labour market. Instead, the rise of
the service economy (e.g., Ngai and Petrongolo, 2017) makes it easier for women to participate (and this
determines the rising part of the U-shaped curve). For a recent discussion of the U-Shaped hypothesis,
see also Buono and Polselli (2022).
individuals less in 2040 relative to 2022. Such a demographic change would imply a
significant decline in the total labour force in the next 20 years. An increase in the
female participation rate could mitigate these adverse demographic trends. For example,
if the female participation rate were to converge towards the current EU average in ten
years’ time, the expected decline in the Italian labour force in 2040 would be halved.
An increase in female participation would have a positive impact on GDP. Estimates
based on the model developed in Bulligan et al. (2017) suggest that, other things being
equal, an increase of 10% in the labour force — due to the convergence of the Italian
female activity rate to the EU level — would raise GDP by roughly the same amount in
the long run. However, these figures do not take into account the potential extra gains
deriving from the consequent reallocation of talents in the economy; according to Hsieh
et al., 2019 such mechanisms accounted for a large part of the economic growth observed
in the US since the 1960s.2
For all these reasons, 11 years ago the Bank of Italy carried out several studies aimed
at analysing the determinants of the gender gaps in the Italian economy (see Bianco et al.,
2013). Despite some progress since then, the overall picture has not changed much. Hence,
starting in 2020, a new wave of research projects has been carried out by the economists
of the Bank of Italy (20 papers — marked with red title in the References). This report
summarises their main findings, highlighting how they contribute to the evidence provided
in the most recent international literature on the topic and indicates the most promising
avenues for effective policy action.
The report is structured in six chapters. Chapter 2 provides an overview of gender
gaps in the Italian labour market, also highlighting the relative position of Italy in the
European context. Besides showing that Italy stands as the country in the EU with
the lowest female participation rate and the second lowest female employment rate, it
documents that women, when employed, work fewer hours, are more likely to involuntarily
working part-time and are paid less than men.
Chapter 3 investigates the differences in education between boys and girls and discusses how they affect gender gaps at entry in the labour market. It shows that girls tend
to outperform boys at school and at university in terms of both educational attainment
and achievement. However, girls tend to systematically self-select into fields of study
that are associated with worse labour market prospects Already one year after having
completed their studies there is a large gap between men and women in terms of both
employment probability and wages, which is largely related to the choice of high school
track and — especially — university major. This result points out that, to tackle gender
gaps, it is important to act also at the very early stage, when women are at school and
choose their field of study.
Ostry et al., 2018 also highlight that a non-negligible part of the effect of increasing female labour
supply on GDP is driven by the gains in productivity that derive from a higher diversity in the workplace.
Chapter 4 discusses the relationship between female employment and fertility decisions. The traditional role of women as main providers of childcare and domestic work
has been undoubtedly one of the main obstacles to the integration of women in the formal
labour market. Despite the enormous progress observed in the last decades, motherhood
is still a critical point for women’s employment and careers. Family-friendly policies —
like parental leaves and subsidised childcare — may play a significant role in boosting
female labour supply and reducing gender gaps. Based on the vast literature on the topic
and on the characteristics of the policies currently in place in Italy, we outline the main
areas of policy interventions.
Chapter 5 highlights the role of partners’ interactions in shaping female and household
labour supply and the impact of the tax-transfer system on these joint decisions. The
labour supply of married women, as second earners in the household, may be discouraged
by joint taxation or transfers sharply decreasing with family income. We provide an
overall assessment of how the Italian tax-transfer system affects monetary incentives to
work by gender, given the current distribution of employment among Italian households.
We also look in detail at specific policy interventions that have been implemented in the
most recent years.
In Chapter 6 we document the extent and determinants of women’s vertical segregation in the labour market and their under-representation in top earnings classes and
jobs. We show that such glass ceiling is determined not only by the initial unfavourable
sorting across fields of study (Chapter 3) or by the child penalty (Chapter 4) but also
by differences across employers and in career progressions within firms, where women
tend to be assigned less remunerative tasks and hold lower paid positions. We discuss
some of the interventions that may be more useful in counteracting such disparities. In
particular, we point at the importance of stimulating corporate managerial practices that
favour work-life balance and of promoting, even through affirmative action interventions,
the presence of women also in medium-ranked management.
Finally, Chapter 7 discusses some possible policy measures that could help alleviate
the existing gender gaps, also in the light of the recent actions taken within the scope of
the National Recovery and Resilience Plan (NRRP).
An overview of gender gaps
This Chapter describes gender gaps in the Italian labour market, focusing on women’s
labour supply and employment relative to men’s, as well as on gaps in wages and work
intensity among employed individuals. Finally, we discuss the contribution of these gender
gaps to labour income inequality and in-work poverty.
In Italy the female employment rate in the 15-64 age group was 51.1% in 2022, a value
which is high if compared to the past three decades, but is still almost 14 percentage points
lower than the EU average (Figure 2 panel a). The difference between the employment
rate of men and women was in Italy 18.1 percentage points, the second largest in the
European Union and appears to be mostly driven by the lower participation of women
in the labour market: Italy indeed shows the lowest female activity rate in the EU
(in 2022 56.4% and 69.5%, respectively; Figure 2 panel b), while gender differences in
unemployment rates — although larger than the European average — are more limited
(2.3 percentage points, compared to 0.6 in the EU).
Figure 2: Female employment and participation rates
(a) Evolution of the female employment rate, (b) Female participation rate (2022), percent-
age points
Sweden
Norway
Portugal
Malta
Austria
Slovakia
Hungary
France
Ireland
Spain
Czechia
Bulgaria
Poland
Greece
Switzerland
Germany
Netherlands
Spain
Denmark
France
Germany
Italy
Belgium
Italy
Romania
percentage points
Notes: The data refer to individuals aged 15-64. The series shows a discontinuity in 2021, following the updates established
by regulation EU/2019/1700 which provides, in particular, new criteria for identifying employed individuals. Source:
Eurostat, European Labour Force Survey.
De Philippis and Lo Bello (2023) use quarterly panel data from the Italian Labour
Force Survey to estimate transition probabilities across labour market states (employment, unemployment and inactivity) by gender and use them to decompose the dynamics
observed in the male and female employment rates in the past 40 years. They document
that flows into and out of inactivity were the key driver of the convergence of the employment rate across genders (the gap shrank by approximately 18 p.p. from 1985). In
particular, the convergence operated mainly through a reduction in the labour market
exit rate of women and a decrease in the entry rate of men; differences in flows between
employment and unemployment among active individuals played instead a much smaller
role. The authors point out moreover that also nowadays the gender employment gap is
largely due to women’s lower entry into the labour market and higher exit rates — especially after the birth of their first child —, while the role of flows between employment
and unemployment status is limited.
The significant growth in educational levels among women contributed to the marked
increase in the female employment rate observed in the past decades: if in 1990 less than
7% of women aged 25-34 had a university degree, in 2022 this share increased to more
than 35% (from 7% to 23% among men). Indeed, the growth in women’s employment
rate did not take place much within but mostly between educational levels (De Philippis,
2017), as highly educated individuals tend to be more attached to the labour market,
especially women.3 Italy, however, still stands as one of the countries in Europe with the
lowest diffusion of tertiary educated individuals, also among women (in 2022 the share
of 25-34 years old women with university education in the EU was 12 p.p. higher than
in Italy); this may contribute to the low participation of Italian women into the labour
market.
The growth of female employment observed in the last decades has also been favoured
by the diffusion of part-time contracts, which in principle allow for a better reconciliation
between family and work. While at the beginning of the 1990s just over 10% of female
workers were employed part-time in Italy, the share rose to 31.7% in 2022 (from 2.4% to
7.7% for men). This is partly associated with some structural transformations, like the
rise of the service economy, where part-time contracts are more widespread and women’s
employment is concentrated (in 2022, approximately 84% of employed women against less
than 60% of employed men worked in the service sector, especially in education, health,
social and domestic services). Indeed, some papers document that the expansion of the
service economy created jobs with characteristics that better match female preferences
and household roles, and, at the same time, increased the relative demand for female work
as long as women have a comparative advantage in the production of services rather than
in manufacturing (see Olivetti and Petrongolo, 2016a and Buono and Polselli, 2022).
Overall, however, the incidence of part-time contracts among Italian female workers is
particularly high in comparison with other EU countries with similar female employment
rates (Figure 3, panel a). This is in part due to the very high share of involuntary
part-time in Italy. Indeed, Italy is the country in the EU with the highest proportion of
women for whom the choice of part-time work is determined by the lack of full-time job
opportunities (more than one in two women employed part-time compared to less than one
in five women in the European Union average in 2022; Figure 3, panel b).4 Therefore,
In 2021 almost 80% of women aged 15-64 with a tertiary degree and approximately 30% of women
without secondary education was employed; similar values were observed in 1990.
Other studies link the proliferation of part-time work to undeclared work, i.e., the growing tendency to reduce the number of completely undeclared workers, but to compensate for this by increasing
the number of undeclared hours in regular employment, in order to avoid some of the social security
besides supply-driven factors related to work-life balance considerations, demand-side
aspects are also relevant in explaining the large diffusion of part-time work among Italian
women.
Figure 3: Part-time contracts, involuntary part-time and female employment
(b) Share of involuntary part-time among
part-time, percentage points
women employed part-time, percentage points
Share of part-time over employed women
(a) Female employment rate and share of
Italy
Italy
Spain
Greece
Bulgaria
Romania
France
Portugal
Sweden
Slovakia
Czechia
Belgium
Hungary
Poland
Ireland
Austria
Malta
Norway
Slovenia
Women employment rate
Denmark
Germany
Netherlands
Notes: The data refer to individuals aged 15-64. Source: Eurostat, European Labour Force Survey.
The particularly marked expansion of part-time, together with the strong diffusion
of fixed-term contracts (among the employees, 18% of women against 16% of men were
employed on a temporary basis in 2022), translated into a decline in the average number
of hours worked per year among employees that is sharper for women than for men.
Looking at employees in the non-agricultural private sector, the gender gap in annual
full-time equivalent work units has widened, going from 7% in the early 1990s to almost
15% in 2021.
Figure 4: Time use, domestic and total (paid and unpaid) work by gender
(b) Total paid and unpaid work, hours
(a) Household and family care, hours
IT GR RO ES HU AU PL LU EE UK FR BE DE FI
women
NL NO
GR RO IT
gender gap
ES EE HU FI UK LU PL BE FR AU DE NO NL
women
gender gap
Source: Harmonised European Time Use Survey statistics, conducted in 18 countries between 2008 and 2015.
contributions due (see Tirozzi, 2018).
Gaps in hours worked in the labour market are also associated with important differences in time use among men and women (Figure 4). Italy is the country with the largest
gap between the time spent by women and by men in domestic work and care-giving
activities, compared to the countries included in the latest European-wide Harmonized
Time Use Survey (approximately 4 hours and 40 minutes per day for women and 1 hour
and 50 minutes for men). This large gap is still observed when focusing on working individuals and on younger cohorts. However, if we consider the time spent on total paid and
unpaid work, as defined by Eurostat (which includes time spent on paid work, domestic
work, care activities and travelling to and from work), Italian women work more than
men, and this gap is relatively large with respect to the European average. Indeed, total
working time is higher for women than for men in almost all the countries considered:
in Italy, women work about 1 hour and 15 minutes per day more than men (in line with
Spain, but larger than in France and Germany, where the difference is lower than half an
hour).
Figure 5: Net hiring, cumulated values from January 2018 (December 2019=1))
1/1/2019
1/1/2020
1/1/2021
1/1/2022
1/1/2023
women
Source: Ministero del Lavoro e delle politiche sociali, Banca d’Italia, ANPAL (2023), Mandatory reporting (deseasonalized)
data.
The employment and unemployment rates of men and women also differ in terms of
their cyclical behaviour (Buono and Polselli, 2022; Albanesi, 2019). In particular, the
female employment rate tends to vary less along the business cycle, both because women
usually work in sectors and occupations less affected by recessions and economic booms
(like the public sector) and because of the so-called added worker effect, which reduces the
procyclicality of female employment. Indeed, it is found that women who are married or
cohabiting, increase their own labour supply in response to a job (and therefore income)
loss for their partner (Lundberg, 1985 and Chapter 5). Unlike standard recessions, the
Covid-19 pandemic instead heavily hit female-dominated in-person services. As a consequence, net hirings for women dropped more markedly than for men at the outbreak
of the pandemic; overall, in Italy in 2020 women lost more than 70,000 jobs while male
employment increased by more than 60,000 units (Figure 5). However, starting from the
second half of 2021, net hiring increased rapidly among the female population, reaching
historically high levels. In the last year and a half, women accounted for nearly 40% of
job creation, a value 2.5 percentage points higher than in the two-year period 2018-19.
As in other European economies, in Italy women’s hourly wages are on average lower
than men’s (Leythienne and Ronkowsk, 2018): the gap among private sector employees
— although it has gradually decreased over the last three decades — was around 11%
in 2021 (Figure 6 panel a).5 The decline in the gender wage gap was mostly driven by a
reduction of the gap in the top percentiles of the wage distribution (Figure 6 panel b),
possibly also because of the marked increase in women’s average education levels observed
in the same period. Still, in 2021 the gender wage gap in unitary wages — present along
the entire wage distribution — was larger among top earners: the percentage difference
between the 9th deciles of the wage distributions of men and women was twice as large
than that between the 1st deciles of the gender-specific distribution (see also Chapter 6).
Figure 6: Gender gaps in Full-Time Equivalent (FTE) wages
(a) FTE wages and work units, percentage (b) Percentiles and mean of FTE wages, per-
centage points
points
FTE weekly wages
FTE weeks
Notes: Differences between the annual averages of Full-Time Equivalent (FTE) wages or work units of men and women,
as a percentage of the level observed among men. Non-farm private sector employees aged 15-64; FTE wage is the weekly
wage adjusted for full-time equivalent work units; the full-time equivalent work units are determined by dividing the total
number of hours worked by the full-time contractual hours. Panel b: percentiles are computed along the distribution of
women and men respectively. Source: Elaborations on INPS data.
Furthermore, since women earn less than men also at the bottom of the wage distribution, they are more likely to receive a unitary wage that stands below 60% of the median,
a threshold that usually defines low-wage workers. According to Depalo and Lattanzio
(2023) 7.1% of women against 4.6% of men are considered working poor according to this
definition. If we also take into account that employed women also work fewer hours per
year, and therefore look at the probability that annual earnings fall below 60% of the
According to INPS data, also considering public sector employees (about 20% of employees in the
total economy), the average pay gap in daily wages would be slightly lower; indeed, in the public sector
— where the wage distribution is more compressed — the gender gap is almost 14% lower than that
observed among private employees.
median, the gender gap is even larger (39% of women against 25% of men have a yearly
wage below this threshold).
According to data from the Eurostat Structure of Earnings Survey for 2018 (the latest
available year), the average pay gap between men and women in Italy is lower than that
observed in the other main European countries, but this only depends on compositional
effects. Since the participation of less qualified women is particularly low, Italian female
workers tend to be relatively more educated than the European average and therefore to
receive higher wages: net of this heterogeneity, the gender hourly wage gap in Italy is in
line with the European figure.
The low female employment rate, together with the small number of hours worked
if employed, affect the economic conditions of women and their families. Indeed the
incidence of couples in which the woman is not employed or works only a few hours in
Italy is particularly high in the European context — especially among the less well-off
households. According to data from the EU-Survey on Income and Living Conditions
referred to 2018, in Italy in 35% of the households with at least two adult members there is
only one labour income recipient (in about 80% of cases, a man); this share is much lower
in the other main EU countries (28% in Spain, 24% in France and 23% in Germany). This
boosts labour income inequality at the household level (which is higher in Italy compared
to other EU countries like for instance France and Germany) and explains a significant
part of the difference with respect to the other main EU countries. For example, if Italian
women worked the same number of hours per year as German women, the Gini index
on equivalised labour income would drop by almost 3 p.p., closing about 80% of the gap
with Germany (Bovini et al., 2023).
Education and school-to-work transition
This chapter first documents the existence of gender gaps in educational achievements
and attainments, indicating that on average girls tend to outperform boys both at school
and at university. However, women tend to sort and graduate into less remunerative fields
of study, which imply worse career prospects in the labour market, both at secondary
school and at university. The chapter then reviews the existing literature that analyses
the determinants of gender gaps in the field of study choices, indicating that they mainly
originate from gender-specific preferences that are largely shaped by cultural norms and
stereotypes. The second part of the chapter describes and explores the determinants of
wage and employment gaps between boys and girls emerging already one year after having
completed their studies and confirms that differences in fields of study play a major role,
especially among university graduates. Finally, it discusses some policy measures that
can be effective in reducing part of these gaps, for instance, the exposure of girls to
women employed in male-dominated occupations, which act as positive role models, or
interventions aimed at revealing teachers’ or parents’ implicit biases.
Descriptive evidence on gender gaps in education
In all advanced economies, girls reach higher levels of educational attainment than boys.
Across OECD countries, boys are more likely than girls to lack an upper-secondary qualification. When selecting an educational trajectory, boys are usually over-represented
in vocational paths and less likely to enter into and graduate from tertiary education
(OECD, 2022). Italy is no exception: while the share of tertiary graduates is in general
very low in international comparison, both for boys and for girls, the percentage of girls
among university graduates aged 25-34 is above 60%, a value substantially higher than
the European average (Figure 7).
In all developed countries girls outperform boys also in terms of educational achievements: on average across OECD countries with available data, boys are more likely to
repeat a grade than girls and represent 61% of repeaters in lower secondary education
and 57% in upper secondary education (OECD, 2022). Bovini et al. (2023) rely on administrative data on the universe of Italian graduates from upper secondary school and
university between 2011 and 2018 to assess gender gaps in educational performance. They
show that girls, if anything, obtain higher final grades than boys at all educational levels:
this holds true also within all high-school tracks and most university majors (Figure 8).
When inspecting the origin of this girls’ advantage in education, Bovini et al. (2023)
find that it appears larger at the bottom of the ability distribution. Girls’ final university
grade is above that of boys, especially among students who obtained relatively lower
grades at the end of secondary school. Similarly, the gender gap in university enrolment
is larger among students from less advantaged socio-economic backgrounds, with lower
Figure 7: Educational attainment by gender
(a) Share of girls among tertiary graduates aged(b) Education attainment, individuals aged 25-
25-34
Italy
Spain
Portugal
Denmark
Austria
Australia
EU22 average
Belgium
France
Canada
Greece
Germany
Females
Males
Females
EU-27
Males
Italy
Primary
Secondary
Tertiary
Notes: The data refer to the last available year (2021 for panel a, 2022 for panel b). Source: panel a, OECD Education
at a Glance, 2022; panel b, Eurostat.
Figure 8: Final grades
(a) in secondary school, by high school track
(b) at university, by major
Math, stats., phys., chem.
Biology, geology
Engineering, ICT
Architecture, design
Agricult., veterinary
Health, pharmacy
Academic track
Technical track
Legal studies
Industry and
trades
Services
Technology
Business and
economics
Language
and arts
Classical
studies
Scientific
studies
Business, economics
Soc. and pol. sciences, psych.
Education
Arts, humanities
Vocational track
Girls
Girls
Notes: Cohorts who graduated from upper secondary school or from university (2nd level degree or one-cycle degree) in
the period 2011-2018. Final grades are re-scaled to range from 0.6 (minimum) to 1.01 (maximum, corresponding to 100
cum Laude). Source: Bovini et al. (2023).
final secondary school grades (Figure 9, panel a) and who graduated from technical and
vocational tracks (as opposed to the academic track; Figure 9, panel b).6
However, in all advanced countries there exist substantial differences in the fields of
study chosen by boys and girls, leading to very different career paths (OECD, 2021).
Bovini et al. (2023) explore gender differences in secondary school tracks and university
majors’ choices among Italian students. In secondary school, girls make up a larger
proportion of graduates in all non-STEM fields, such as humanities, languages, services
and social sciences (Panel a of Figure 10). Also at university, girls tend to enrol in
This is in line with some existing literature showing that boys are particularly affected by poorer
home and school environments (see for instance Bertrand and Pan, 2013 or Chetty et al., 2016, even if
on different outcomes).
Figure 9: Probability of enrolling in tertiary education
(b) by secondary school track
.84 .85
.87 .88
Girls
Technical track
Industry and
trades
Services
Technology
Business and
economics
Language
and arts
Classical
studies
Scientific
studies
Academic track
.05 .15 .25 .35 .45 .55 .65 .75 .85
(a) by secondary school final grade
Vocational track
Girls
Notes: Cohorts who graduated from upper secondary school in the period 2011-2018. Secondary school graduates are
assumed to have enrolled in tertiary education if they are registered as students in administrative data in the year after
graduation. Final grades range between 60 and 100 and are re-scaled to take values between 0.6 and 1. Full marks with
honours (100 cum Laude) are re-scaled to take value 1. In panels (b) and (c) lines capture quadratic fits to the data.
Source: Bovini et al. (2023).
non-STEM majors.7 Panel b of Figure 10 indicates that girls represent 94% of those
graduating in education, more than 80% of those graduating in foreign languages and
psychology and over 70% of those graduating in humanities and social sciences. Girls
are instead only 40% of graduates in the STEM majors (they are 27% of students in
engineering and ICT and 46% in math, chemistry and physics).8
Girls moreover tend to enrol in lower-quality universities. First, they are less likely
to move away from home (to a different region) to study; the gap is particularly large for
those born in the South of Italy, a macro area where universities are on average of lower
quality (Mariani and Torrini, 2022) and out-migration rates are large. Approximately
48% of boys against 43% of girls born in the South of Italy attend university in a different
region.9 Second, Bovini et al. (2023) indicate that, also because girls are less likely to
change region to study, they tend to graduate from universities that are less likely to be
ranked among the top 50 departments, according to the 2022 QS Italian ranking, and
which receive a lower grade by the Italian Institute of university evaluation (Evaluation
of Research Quality, VQR score, ANVUR 2016).10
This considers a narrow definition of STEM fields (that only includes Math, Natural Sciences,
Engineering and ICT).
Similar gaps are observed on average in all OECD countries: in 2020, the latest available data,
girls represented approximately 80% of graduates in Education and 70% of graduates in the fields of
Arts and Humanities or Social Sciences. The share of girls among graduates in Business or Law was
smaller, approximately 57%. Finally, girls represented about half of those graduating from the fields of
Natural Sciences and Mathematics and only 27% and 22% of those graduating in Engineering and ICT,
respectively. Overall the share of girls graduating in a STEM field in Italy seems to be in line with the
average for OECD countries.
This pattern is in line with the existing literature, like for instance Rizzica (2013), De Angelis et al.
(2016) and De Angelis et al. (2017).
The VQR exercises provide an up-to-date assessment of the state of research in the various scientific
Figure 10: Girls’ fields of study
(a) share of girls across secondary school tracks
(b) share of girls across university majors
Math, stats., phys., chem.
Biology, geology
Engineering, ICT
Architecture, design
Agricult., veterinary
Health, pharmacy
Business, economics
Academic track
Technical track
Industry and
trades
Services
Technology
Business and
economics
Language
and arts
Classical
studies
Scientific
studies
Legal studies
Soc. and pol. sciences, psych.
Education
Arts, humanities
Vocational track
Notes: Cohorts who graduated from upper secondary school or from university (2nd level degree or one-cycle degree)
in 2011-2018. Panel (a) shows the share of girls among secondary school graduates and within each sub-track; panel (b)
reports the share of girls among university graduates and within each major (excluding majors in Defence and Strategic
Studies and in Performing Arts). Source: Bovini et al. (2023).
Overall, Bovini et al. (2023) show that these differences in field of study imply that
girls tend to self-select into secondary school tracks and university degrees that will
guarantee them lower-paying jobs.
Figure 11 refers to students who enrol in non-academic secondary school tracks (and
who are therefore more likely to stop studying after secondary school). It defines secondary school tracks’ potential career prospects depending on: i) the median annual
earnings (panel a) and ii) the average employment probability (panel b) 5 years after
graduation of male non-immigrant graduates who did not continue to study. The results
indicate that there is a large heterogeneity in high school tracks’ average potential labour
market outcomes: the average yearly earnings (employment probability) 5 years after
graduation varies from about 20,000 (87%) euro for graduates in Mechanics to approximately 13,000 euro (70%) for graduates in the Tourism track. Importantly, girls are more
likely to graduate from tracks characterised by worse average labour market prospects in
terms of both average earnings and employment probability.
Even at university women make educational choices that translate into lower expected
labour market earnings. Figure 12 plots the average labour market returns (measured as
the median yearly earnings 5 years after graduation of male non-immigrant students) of
each combination of universities and majors (referred to as a degree).11 It indicates that
girls select lower paying degrees along the entire ability distribution (panel a): female
students graduate in degrees that imply on average yearly earnings approximately 3,000
fields, in order to allocate the performance-based share of the Ordinary Financing Fund for the Italian
University system. This ranking refers to the latest data available, those of the VQR exercise that covers
the years between 2011 and 2014 and took place in 2015.
We do not perform the same analysis also looking at differences in employment probability across
majors, because — as will be shown later — gender differences in the employment rate among university
graduates are negligible.
Figure 11: Types of secondary school track attended by girls
20000
(a) Higher paying tracks
(b) High employment probabilities tracks
Chemistry and biotechnology
Electronics
Mainteinance and repair
Fashion
Graphics and design
Agricultural science
Transport and logistics
Construction and environment
Agriculture
Median empl. probability (native men)
Median labour earnings (native men)
14000
16000
18000
Mechanics
Mechanics
Electronics
Graphics
and design
Chemistry
and biotechnology
Mainteinance and repair
Fashion
Construction and environment
Transport and logistics
Production
Hospitality
Accounting, finance, marketing
Production
Accounting, finance,
marketing
Socio-sanitary services
12000
Commercial services
Tourism
Socio-sanitary services
Commercial services
Agricultural science
Agriculture
Tourism
Share of female
Hospitality
Share of female
Source: Notes: Cohorts who graduated from non-academic (i.e., technical and vocational) tracks of upper secondary
school in 2011-2018. On the x? axis one reads the share of girls among graduates by track; on the y-axis one reads
the median annual income (panel a) or employment rate (panel b) 5 years after graduation by track, computed on the
population of male native students who stop studying after secondary school. The blue line captures the linear fit to the
data. Bovini et al. (2023).
euro lower than those selected by male students. This gap, moreover, is larger for higherability students (it is about 2,000 euro at the bottom and 4,000 euro at the top of
the ability distribution). This largely reflects the low share of girls, especially among
high-achievers, who pursue STEM majors that are typically associated with high-paying
careers. Indeed, when decomposing this by major (panel b), it appears that most of the
differences comes from students’ choice of major: within major (when the difference in
degree quality only depends on the institution of enrolment) the gap is almost null (the
blue and the red lines are very close to each other).
Figure 12: Degree quality by students’ gender and ability
(b) Higher paying degrees, by major
Math, stats., phys., chem.
Biology, geology
Engineering, ICT
Architecture, design
Agricult., veterinary
Health, pharmacy
Business, economics
Legal studies
Soc. and pol. sciences, psych.
Education
Arts, humanities
30000
Median labour earnings (native men)
20000
Median labour earnings (native men)
22000
24000
26000
28000
(a) Higher paying degrees
Final HS grade
25000
20000
15000
30000
25000
20000
15000
30000
25000
20000
15000
Girls
Final HS grade
Girls
Notes: The sample consists of students who belong to the 2016-2018 cohorts of university graduates (2nd level degree
or one-cycle degree) and for whom also the final high-school grade is recorded. A degree is a university×major couple.
The quality of a degree is measured by the median labour earnings of its male, native students 5 years after graduation.
Source: Bovini et al. (2023).
Understanding gender gaps in education
From the previous section, it emerges that there exist large gender gaps in education. On
the one hand, boys are disadvantaged in terms of educational attainment and achievements: they are less likely to finish secondary school, to graduate from university and
they tend to perform worse at school, especially those at the bottom of the income and
ability distribution. This is a very important dimension that deserves further investigation (see for instance Goldin et al., 2006) but is not our focus, since the aim of this report
is to assess the origin of women’s disadvantages in the labour market. On the other hand,
women make educational choices that translate into lower future labour market returns,
and therefore lower attachment to the labour market. Understanding what drives these
differences is then crucial and we will focus our attention on this.
The role of skills
Different choices might reflect differences in skills. For example, girls may perform worse
relative to boys in maths and sciences, and hence have a comparative disadvantage in
these subjects, which are mostly needed to succeed in high-paying majors.
We can explore the relevance of this explanation by looking at boys’ and girls’ scores in
standardised tests administered by INVALSI.12 Figure 13, panel (a) shows the difference
between the average scores of male and female students in Language and Mathematics
tests taken at different stages of the education cycle. From the beginning of primary school
(grade 2) to the end of upper secondary school (grade 13), girls on average outperform
boys in the language test, but they score lower in maths. Moreover, the gender gap
in mathematics is negative along the entire distribution of grades (Figure 13, panel b).
Notably, the gap widens with the age of the students and is larger for older students at
the top of the grade distribution than at the bottom, so that it is particularly sizeable
for high-achievers.
This pattern is not specific to Italy. In all countries that participate in the Programme
for International Student Assessment (PISA) standardised testing of skills, 15 years old
females have on average higher reading skills than males; on the other hand, in 31 out
of 37 countries they have lower Mathematics skills (OECD, 2019; Appendix Figure A.1).
In Italy the positive gender gap in reading scores is slightly larger than in the average of
the OECD; worryingly, the negative gender gap in mathematics is much larger than the
OECD average and is the largest in Europe.13
INVALSI (Istituto nazionale per la valutazione del sistema di istruzione e formazione) is the Italian
agency that administers each year standardised tests to all pupils at the start and at the end of primary
school (grades 2 and 5), at the end of lower secondary school (grade 8), at the start and at the end of
upper secondary school (grades 10 and 13).
Moreover, the literature shows that, not only in Italy, but also in other countries the male advantage
is not present (or is very small) at school entry and tends to emerge when students get older, during the
first four years of school (Hyde et al., 2008; Penner and Paret, 2008; Fryer and Levitt, 2010).
Figure 13: Gender gap in scores in standardised INVALSI tests
(a) By grade and subject (girls-boys)
(b) In math, over the grade distribution
Upper secondary – Year 5
Language
Year 5
Upper secondary
Year 2
Year 3
Lower secondary
Year 5
Primary
Year 2
Primary – Year 2
percentile
Mathematics
Girls
Notes: The data refer to the academic year 2018-19. Scores are standardised to have a mean equal to 0 and a standard
deviation equal to 1 across all test takers. Source: INVALSI.
Comparative advantage is likely to be an important driver of a student’s choice of
high-school track and university major. Table 1 confirms that the share of girls with
a comparative advantage in Mathematics over language is smaller than that of boys in
grade 8, when Italian students decide which high-school track to enrol in:14 31% of male
students have a maths score that belongs to a forth higher than that of their language
score (i.e., they have a comparative advantage in maths); this figure drops to 17% for
female pupils. The gender gap in comparative advantage persists at grade 13, when
students need to decide whether to pursue tertiary education and which major to enrol
Table 1: Comparative advantage at grade 8
a. Boys
b. Girls
Mathematics forth
Mathematics forth
Language quartile
Notes: Each cell (i, j) of the table contains the share of boys (panel a) and girls (panel b) with a Language score belonging
to the i-th forth of the distribution and a Mathematics score belonging to the j-th forth of the distribution. In the main
diagonal (orange cells) students are equally skilled in Language as in Mathematics; in the lower triangle (yellow) they have
a comparative advantage in Language; in the upper triangle (blue) they have a comparative advantage in Mathematics.
Source: INVALSI, scholastic year 2018-19.
The evidence discussed so far shows that differences in skills and comparative advan14
In the Italian school system there are three main high-school tracks: academic, technical and vocational. Within the academic track, liceo scientifico (scientific studies) features the most maths-heavy
curriculum, while liceo classico (classical studies) and the other sub-tracks typically devote less instruction time to Mathematics. Within the technical tracks, istituti tecnologici (technology sub-track) have
the most maths-intensive curricula.
tages are relevant. Breda et al. (2018) indeed document that girls’ comparative advantage
in reading over maths is a key determinant of the gender gap in enrolment in STEM majors across OECD countries. However, it appears not to be the only driver of gender
differences in education paths, at least in Italy.
Figure 14: Comparative advantage at grade 8 and maths intensiveness of high-school curriculum at grade 10
Technical track
Vocational track
Weeekly math hours, grade 10
Academic track
Neither Lang.
Neither Lang.
Neither Lang.
Comparative advantage, grade 8
Girls
Notes: Comparative advantage at grade 8 and weekly Mathematics instruction hours at grade 10 are computed for the
cohort who attended grade 8 in the scholastic year 2014-15 and grade 10 in the scholastic year 2016-17. An individual has
a comparative advantage in language (maths) if her language (maths) score belongs to a fourth of the grade distribution
that is higher than that of the maths (language) score; an individual has no comparative advantage if the fourth of the
grade distribution of her language score is the same as that of the maths score. Source: INVALSI.
Figure 14 analyses the choice of high-school track. It shows the average number of
maths weekly instruction hours at grade 10 (a measure of how maths-intensive a highschool track is) by high-school track, gender and, notably, by individuals’ comparative
advantage measured at grade 8 (the last year of middle school).15 Focusing on the academic track, for both genders having a comparative advantage in math at the end of
lower secondary school is associated with choosing more math-heavy curricula in upper
secondary school. However, even among students with a math comparative advantage,
girls are much less likely to choose a math-intensive track than boys.16 Gender differences
In INVALSI microdata we cannot distinguish the various sub-tracks within the three principal
tracks (academic, technical, vocational). We, therefore, retrieve the information on weekly Mathematics
instruction hours (that vary across sub-tracks) from questions asked by INVALSI to a representative
sample of grade 10 maths teachers about their weekly teaching schedule in the surveyed class. This
information, which we can link to students’ data through a unique anonymised class ID, is therefore only
available for a sample of pupils. We focus on the cohort who attended grade 8 in the academic year
2014-15 and grade 10 in the academic year 2016-17, as it is the one for which we have all the needed
data available.
This likely reflects the fact that even girls with a comparative advantage in Mathematics are less
likely to attend liceo scientifico.
are more muted in technical and vocational tracks, also likely due to the smaller variation
of Mathematics instruction time across sub-tracks.
Concerning the choice of major, Bovini et al. (2023) show that, among university
graduates, females are less likely than males to graduate from a narrow STEM major
and, importantly, this gender gap is sizeable also among those who attended a mathsintensive upper secondary school track, who presumably have a comparative advantage
in maths and science.17 This finding is in line with the existing literature. Delaney and
Devereux (2019) show that in Ireland, even for students with identical preparation at the
end of secondary school in terms of both subjects studied and grades, there remains a 9
p.p. gender gap in the propensity to enrol in STEM courses at the tertiary level.
The role of preferences and stereotypes
Boys and girls may also make different choices because of differences in preferences. For
instance, it is common for girls to state they dislike STEM subjects. The existing literature tends to agree that preferences are the main driver of gender differences in fields of
study choices. For instance, a survey by Zafar (2013) documents that the largest determinant of college major choices for both genders is whether they enjoyed the coursework.
Wiswall and Zafar (2017) find that job attributes, and especially non-pecuniary job attributes (like work flexibility), have a sizeable impact on choices of university majors for
women. Also Ceci et al. (2014) stress the importance of preferences towards job characteristics. They show that girls, differently from boys, prefer people-oriented rather than
thing-oriented jobs and that this dichotomy explains a large part of gender differences in
secondary school track and university major choices. Finally, Wiswall and Zafar (2014)
randomly provide some students with information about the earnings and employment
of people who chose a certain major; they still find that preferences — rather than expected earnings — are the dominant factor in the choice of university major and that
they explain a large part of different choices by gender.
It, therefore, becomes important to assess what generates differences in preferences.
If preferences are innate, then there may be no need for corrective actions, as choice
differences would depend on intrinsic parameters. However, if preferences depend also on
the context and vary over time, they may be a reflection of gender norms and stereotypes,
and policy action may be needed (Bertrand, 2020). Some descriptive evidence shows that,
at least in Italy, the gender gap in preferences appears quite small at early ages and it
expands at older ages. According to a survey administered by INVALSI to Italian 5-th
grade pupils on their feelings about maths, girls are slightly less likely than boys to report
having good or very good feelings about learning Mathematics at age 11 (approximately
See Appendix Figure A.2. Notice that Bovini et al. (2023) cannot use the same definition of
comparative advantage used for the choice of secondary school tracks, since there is no information
about subject-specific grades in their data.
a 2 p.p. difference). Instead, PISA data show that at age 15, when asked about their
work expectations, even girls who are top performers in STEM subjects are much less
likely to state they expect to work in STEM jobs.18
The existing literature points out that the family and school environment, by affecting
individuals’ stereotypes and cultural values, impacts girls’ performance and willingness
to enrol in male-dominated subjects and majors (like STEM degrees). At the country
level, many studies have indicated how average maths performance among girls is strongly
correlated with gender attitudes in a country (Guiso et al., 2008; Pope and Sydnor, 2010;
Nollenberger et al., 2016). Besides cross-country evidence, Brenøe (2022) uses a quasiexperimental variation to show that girls born in a family with opposite-sex siblings
acquire less traditional gender norms, and are more likely to graduate in a STEM degree
and to work in less female-dominated occupations than girls with a same-sex sibling. Dossi
et al. (2021) identify families with a preference for boys by using fertility-stopping rules
and show that girls raised in boy-biased families have more traditional gender attitudes
and that this translates into lower performance in maths. Also teachers are an important
determinant of stereotypes and gender norms, as will be discussed in detail in Section
a STEM major at university and of becoming an inventor crucially depends on cultural
factors about the role of women in society, which are very persistent over time. In
particular, the authors show that a higher participation of women in Medieval guilds —
which according to some historians were a tool that gave women decision-making and
economic power — at the municipal level is associated with a higher incidence of female
inventors nowadays, and a higher number of patent submissions by women.
The role of the school environment: curricula, peers and teachers
The school environment can be also a relevant determinant of gender gaps in the choices
of the field of study and in the performance in scientific subjects (see for instance Kahn
and Ginther, 2018; McNally, 2020); some evidence suggests that schools are very heterogeneous in their ability to develop talents in STEM, especially among girls, and that this
heterogeneity mostly depends on unobservable factors (see Ellison and Swanson, 2010).
Peer composition at school could be important. The existing literature tends to show
that attending classes with many same-sex peers does not affect girls’ field of study choices
nor improve their performance at school both in the short and in the long run (while it
generally has a positive effect on boys, at least in the short term; see Anelli and Peri,
2019; Park et al., 2018; Doris et al., 2013). Instead, it appears that attending classes
with very high-ability same-sex peers has positive effects on educational outcomes for
See Appendix Figure A.3. In particular, in panel b appears that in Italy the gender difference in
the share of top performers who expect to work in STEM jobs is approximately 14 p.p. (26% of boys
against 12% of girls); in the OECD it is slightly smaller (12 p.p.).
both girls and boys. Modena et al. (2022) exploit individual-level administrative data
on the population of Italian university students to analyze the effects of high-performing
male and female peers on individual academic performance. They find that higher shares
of high-performing peers improve both the extensive margin of student performance (in
terms of the number of exams taken) and its intensive margin (in terms of grades). Two
findings appear clearly. First, the strongest effect comes from same-gender peers. This
result can be rationalised within the role model framework: observing similar-gender
students doing particularly well positively affects students’ own perception of themselves
and their study effort. Second, high-performing female peers have beneficial effects on
males too while high-performing male peers do not affect female students; when they do
so, their impact can even become negative. This is consistent with the results of the literature according to which men increase their self-confidence in competitive environments
with other men, while women tend to become less confident under competitive pressure
with men (Niederle and Vesterlund, 2007).
The structure of curricula is another determinant of school choice and performance.
For instance, more exposure to maths or science early in the educational path may induce
more girls into STEM. Several studies have analysed the effect of curricula reforms, which
generally increase the teaching hours in maths or science, on gender gaps in educational
outcomes. These studies suggest that the type of curriculum reforms and of students who
are affected (for instance whether higher-ability or lower-ability students) are important
to understand the effect of such policies. In general, more scientific subjects in secondary
school increase overall enrolment in STEM at the tertiary level, but — even if the evidence
is mixed — the increase appears to be concentrated mostly among boys (Joensen and
Nielsen, 2016; G¨orlitz and Gravert, 2018; De Philippis, 2023). These reforms, therefore,
tend to widen the gender gap in STEM enrolment at university.
Finally, teachers are a key input of the production of student achievement (Rockoff,
2004 and Rivkin et al., 2005); some recent papers have pointed out that their impact
persists through adulthood (Chetty et al., 2014). In general, the existing literature points
out that teachers can reduce gender gaps in education. First, they can act as a role model.
While the effect of having a female teacher for female students appears to be small
(and even slightly negative) in primary school (Antecol et al., 2015), studies looking at
secondary school and university teachers usually find that having a female teacher has
a positive effect on female student achievements and on the probability of selecting and
graduating in male-dominated degrees (e.g., Nixon and Robinson, 1999; Bettinger and
Long, 2005; Dee, 2007; Hoffmann and Oreopoulos, 2009). Carrell et al. (2010) for instance
using the random allocation of students to teachers finds that, although professor’s gender
has little impact on male students, the highest-ability girls are more likely to graduate in
a STEM major and achieve better grades in maths and sciences if assigned to a female
STEM instructor at university. Bottia et al. (2015) also show that exposure to female
maths and science teachers in high school increases the probability that female students
are willing to major in STEM in college, especially for girls with high maths ability. These
results confirm that stereotypes and cultural factors seem to play an important role in
girls’ field of study choices.
Second, teachers affect students’ performance in maths and science also through their
possible bias in favour or against female students, with long-term consequences on subject
choices. Carlana (2019) studies teachers’ stereotypes in Italy. She finds that teachers
are strongly implicit gender-stereotyped and tend to believe that girls are worse than
men in scientific fields. This has a negative and significant effect on girls’ performance
and self-confidence in maths and induces girls to self-select into less demanding high
schools, following their teachers’ (biased) track recommendations. Moreover, Lavy and
Sand (2018) measure bias using class-gender differences in scores between school exams
graded by (randomly assigned) teachers and national exams graded blindly by external
examiners. They show that in maths girls outscore boys in the “blind” exam and boys
outscore girls in the “non-blind” exam, suggesting there is significant gender bias among
teachers. Furthermore, they show that a negative bias has long-term negative effects
both on achievements and on the probability of enrolling in advanced-level maths courses
during middle and high school. Also Terrier (2020) uses a combination of blind and nonblind test scores to show that in France girls tend to benefit from positive gender bias
by middle school maths teachers and that this makes them more likely to select a science
track in secondary school.19
School to work transition: descriptive evidence on early
career gender earning gaps
This section focuses on the gender pay gap in the first few years after labour market
entry; this is a relevant object of study since it tends to persist over the life-cycle, to
widen after childbirth, and to become especially large at the top of the distribution (see
Chapters 4 and 6). Arellano-Bover et al. (2023) show that the decrease in the overall
gender hourly wage gap observed in the last forty years in Italy (Chapter 2), as well as in
other countries, was mainly due to a reduction in the gender gap in entry wages across
cohorts rather than to within-cohorts dynamics (see Figure 15). This is important to
understand the sources of the observed gender convergence, as it suggests that it was
mainly driven by factors taking place during the initial sorting of workers to firms or
by compositional changes in the characteristics of men and women entering the labour
market (like the more marked increase in women’s educational attainments relatively
to men’s). The paper shows indeed some evidence of a gradual change in the initial
Also Jansson and Tyrefors (2020) finds that bias in exam correction in favour of boys at university,
especially among male graders.
allocation of young men and women across firms.
Figure 15: Gender gap in weekly wages between and within cohorts
Birth year
Total gap
Gender gap in log weekly earnings
Notes: In each year, the data pools information about all workers who were between 25 and 64 years old, worked in the
private sector for at least 27 weeks year, earned strictly positive wages, and did not retire. Weekly wages are expressed in
full-time equivalent units. Source: Arellano-Bover et al. (2023).
Bovini et al. (2023) assess what determines gender gaps in entry wages using a unique
dataset assembled from multiple administrative registers. The dataset matches, for the
universe of students graduating in Italy from upper secondary school or from university
over the period 2011-2018, information on their educational career paths with data about
their earnings and their jobs’ characteristics in the first years since graduation (up to 7
for the oldest cohort).20 The remainder of this chapter summarises the main findings of
this study.
First, the authors show that girls’ labour market outcomes are worse than boys’
already at the start of their careers. Strikingly, this disadvantage materialises already
1 year after graduation and holds at any level of education. Figure 16 confirms that,
as shown in the previous section, girls graduating from secondary school are more likely
than boys to enrol in further education. However, among those who do not continue
studying, girls are less likely to be employed: 1 year after secondary school graduation
there is already a 10 p.p. gender gap in employment probabilities, which remains stable
for up to 5 years. Instead, among university graduates the gender employment gap is
very small (3 p.p.) and it vanishes after 5 years.
Figure 17 displays the percentage gap in earnings at the mean, focusing on graduates
who work and have completed their studies. Already 1 year after graduation, there is
Taken from: the Italian Ministry of Education and Merit (MIM); the Ministry of University and
Research (MUR), the tax records collected by the Ministry of Finance (MEF); mandatory reporting data
on job contracts collected by the Ministry of Labour and Social Policies (MLPS) and firms registers and
matched employer-employee data, maintained, collected by INPS.
Figure 16: Boys’ and girls’ condition 1 year after graduation
(a) Secondary school
(b) University
Study, not work
Study and work
Work, not study
Not study and
not work
Girls
Study, not work
Study and work
Work, not study
Not study and
not work
Girls
Notes: Cohorts who graduated from upper secondary school or university (2nd level or one-cycle degree) in the period
2011-2018. An individual is classified as studying if he/she is enrolled in a course 1 year after graduating from upper
secondary school (panel a) or from university (panel b). An individual is classified as working if he/she has a non-zero
annual labour income. Source: Bovini et al. (2023).
a significant gender gap in annual labour earnings (32% among high-school-educated
workers and 26% among university-educated). The gap in daily wages is smaller at any
education level indicating that part of the annual earnings gap stems from days worked,
especially among secondary school graduates.21 Focusing on full-time workers — hence
only considering differences in pay rates and not in work intensity —, the gap in daily
wages further declines (16% among high-school-educated and 13% among universityeducated), due to a higher prevalence of part-time among female employees. Five years
after graduation, differences remain large.22,23
Finally, Figure 18 shows that the gap is approximately constant in all the deciles of
the wage distribution.24 This is an important finding: the existing literature tends to find
that the aggregate gender wage gap widens at the top of the distribution (Arulampalam
et al., 2007; Casarico and Lattanzio, 2023b). Bovini et al. (2023) show that this is not
the case for young workers.
Since days worked are only recorded for employees, the gap in daily wages is only computed for
workers whose main employment in a given year is salaried; the gap in daily wage for full-time workers
is only computed for the subset of employees whose type of contract is recorded (approximately 80%).
The magnitude of the gaps, however, is very similar even if the sample is kept fixed.
It is interesting to notice that the gap in annual earnings among secondary school graduates shrinks,
due to a narrower gap in days worked, whereas that in daily wages remains roughly constant. For
university graduates, on the other hand, the gap in annual wages does not change while that in daily
wages increases, suggesting growing disparities in pay rates.
These earning gaps may seem large if compared to the figures for the overall population (see for
instance Chapter 2), but are broadly in line with similar evidence for the early career gender gaps in
the population of university graduates in Italy emerging from survey data (AlmaLaurea, 2022). Similar
figures have also been found for Germany (Francesconi and Parey, 2018).
The figure considers only full-time employees, as gender differences in hours and days worked would
otherwise confound the comparison.
Figure 17: Gender pay gap at the mean 1, 3 and 5 years after graduation
After
1 year
After
3 years
After
5 years
After
1 year
Secondary school
graduates
-25 -24
-26 -26
After
3 years
After
5 years
University
graduates
Annual labour income
Daily wage (employees only)
Daily wage (full-time employees only)
Notes: Values represent the differences between women and men wages, relative to men’s wages. Gaps 1 year after
graduation are computed on cohorts who graduated from upper secondary school or university (2nd level or one-cycle
all workers; the daily wage is computed only for employees (i.e. those whose prevalent source of labour income is from
private employment); the daily wage for full-time workers is computed for the subset of employees for whom the information
on their schedule is available (81% of employees 1 year after graduation and 86% of employees 5 years after graduation.
Source: Bovini et al. (2023).
Figure 18: Daily wage distribution of full-time employees, 1 year after graduation by gender
and educational level.
Deciles of the distribution
Boys – secondary school
Girls – secondary school
Boys – University
Girls – university
Notes: cohorts who graduated from upper secondary school or university (2nd level or one-cycle degree) in 2011-2018.
Only graduates who work and no longer study are considered. Annual labour income is calculated for all workers; the
daily wage for full-time workers is computed for the subset of employees for whom the information on the work schedule
is available (81% of employees). The figure plots the deciles of the wage distribution, by gender and final education level.
Source: Bovini et al. (2023).
Understanding the early career gender earning gaps: a twostep analysis
Understanding the determinants of these sizeable early career gender pay gaps is important to design policies that can tackle them. Bovini et al. (2023) model their emergence
as a two-step process. First, boys and girls choose which high-school track or university
major to pursue. Second, after graduation, they decide whether to become employees or
self-employed; if they decide to become employees, they match with an employer.
In their empirical analysis, the authors assess to what extent gender differences in
educational choices can explain aggregate gender gaps. They document that the sorting
of girls into less remunerative fields explains a large share of gender pay differences after
graduation, especially among university graduates.
Nevertheless, they show that within-field gender gaps are substantial. This motivates
the second step of the analysis, which explores within-field gender differences in jobs’ and
employers’ characteristics and assesses to what extent they account for residual withinfield-of-study gender pay disparities. The authors show that, since the beginning of their
careers, girls sort into less productive firms.
Step 1: The role of educational choices
Figure 19 shows the result of the Oaxaca-Blinder decomposition that explores the role of
educational choices.
Figure 19: Oaxaca decomposition of gaps in daily wages 1 year after graduation, educational
controls only
(b) University graduates
(a) Secondary school graduates
-0.25
Log(daily wage)
P(daily wage
in bottom 10%)
Demographics
Graduating cohort
-0.21
P(daily wage
in top 50%)
Performance
Unexplained
-0.07
P(daily wage
in top 10%)
-0.29
Log(daily wage)
Demographics
Performance
Unexplained
Field
P(daily wage
in bottom 10%)
-0.24
P(daily wage
in top 50%)
Graduating cohort
Field
-0.08
P(daily wage
in top 10%)
University away
University ID
Notes: Cohorts who graduated from non-academic upper secondary school tracks or university (2nd level or one-cycle
degree) in 2011-2018. Only graduates who work and no longer study are considered. P(daily wage) is the probability
that the daily wage falls within the bottom tenth, is below the median or exceeds the top tenth of the wage distribution.
“Demographics” includes region of birth fixed effects, marital status, and socio-economic background as captured by ventiles
of parents’ total income. “Performance” includes age and final grade at graduation, as well as their squares. “Field” contains
high-school track (panel a) or major (panel b) fixed effects. “Graduating cohorts” includes year-of-graduation fixed effects
(2011-2018). In panel b the Oaxaca decomposition further includes fixed effects for each university as well as a dummy
for whether the worker graduated from a university located in a region different from the birth one. The decomposition is
performed on the sample of private sector employees for whom these controls, as well as – for the sake of comparability –
those that will be included in later Oaxaca-decomposition (see Figures 21 and 22), are observed. Bars sum to 100. The
number at the top of each bar reports the value of the overall gender gap in each dimension. Source: Bovini et al. (2023).
For secondary school graduates, differences in high-school tracks explain 30% of the
gender gap in average (log) daily wage 1 year after graduation. The figure is slightly larger
when looking at differences at the bottom (i.e., the probability of being in the bottom
10% of the wage distribution) and slightly smaller when focusing on gaps at the top (i.e.,
the probability of being in the top 10% of the wage distribution). Gender differences
in academic achievements (final grade and age at graduation) play a very limited role
and, if anything, they narrow the pay gap, as girls on average outperform boys; also
demographic differences appear to give a very small contribution. Overall, around 70%
of the gap in daily wages remains not explained by educational and demographic factors
for this group of individuals.25
Different educational choices are more important for university graduates, likely because of the higher degree of specialisation of university majors with respect to high school
tracks. One year after graduation, majors explain almost 60% of the average gender wage
gap. Interestingly, their role is more important at the bottom than at the top of the wage
distribution. Differences in the choice of which university to attend (yellow bars) play
a very small role, which is however more visible among top earners.26 All in all, 40%
of the wage gap is not explained by education and socio-demographic characteristics on
average; this share is increasing along the wage distribution.
Step 2: The role of firms and job characteristics for given fields of
study
Nonetheless, young women earn less than men even within each field of study. At the
median, gaps in annual labour income among workers are ubiquitous and in some fields
as large as 30-40% 5 years after graduation.
Table 2 shows how graduates — who are employed and have terminated their studies
— are distributed among types of employment one year after graduation. The table
displays the average incidence among boys (column 1), the aggregate female-male gap
(column 2), as well as the 25th, 50th and 75th percentile of the distribution of the withinfield female-male gap in each considered dimension (columns 3, 4, and 5). Virtually all
those with a secondary education who are in employment work as private employees, and
gender differences in the type of employment are modest. Among those with tertiary
education, both the share of employees and of self-employed are higher. Moreover, girls
are much more likely than boys to be public employees (by 7.4 p.p.) and less likely to be
either self-employed (by 5.5 p.p.) or private employees (by 2.4 p.p.). These differences
mainly stem from the field of study rather than from within-field gender gaps (indeed,
The main sample for this analysis consists of all individuals who (i) graduated from secondary
school or university between 2011 and 2018 and (ii) 1 year after graduating were employed as private
sector employees and no longer studied. Moreover, most of the decomposition focuses on private sector
employees because only for this group detailed info on the job characteristics and employers are available.
For self-employed the only information available is their sector of activity; for public sector employees
detailed information on employer attributes (i.e., size, wage bill, workforce education) is not available.
As found for secondary graduates, disparities in academic performance would imply a smaller gap
than that observed (especially at the top), due to girls’ advantage in this dimension. Socio-demographic
controls have a very limited impact.
Table 2: Gender gap (girls-boys) in type of employment, 1 year after graduation
Within fields
A. Secondary-school graduates
Private sector employees (%)
Public sector employees (%)
Self-employed (%)
B. University graduates
Private sector employees (%)
Public sector employees (%)
Self-employed (%)
Notes: Cohorts who graduated from non-academic upper secondary school or university (2nd level or one-cycle degrees)
degrees in 2011-2018. Only graduates who work and no longer study are considered. Column (1) displays the value among
males; column (2) reports the gender gap (female-male); columns (3), (4) and (5) show the 25th, 50th and 75th percentiles
of the distribution of the within-field-of-study gap. Source: Bovini et al. (2023).
the aggregate gender gap — which encompasses both within and between field of study
differences — tend to be very different from within-major gender differences). Appendix
Figure A.4 shows that gender pay gaps are very different by type of employment (they
are for instance much smaller among public employees, see also Chapter 6); the different
allocation of workers by type of employment therefore matters for the aggregate gender
wage gap.
Table 3: Gender gap (girls-boys) in job and employer characteristics, 1 year after graduation
from secondary school
Private employees
Within fields
Part-time contract (%)
Permanent contract (%)
High-skill occ. (%)
Medium-skill occ. (%)
Low-skill occ. (%)
-17.3
-17.7
-15.0
-13.3
Firm distance from birth muni.
-16.9
Firm average age
Firm size
-13.7
Co-workers’ average educ.
Female co-workers (%)
Firm VA per worker
-10.0
Firm av. wage
-229.8
-143.6
-82.6
Notes: Cohorts who graduated from non-academic upper secondary school in 2011-2018. Only graduates who work and no
longer study are considered. Column (1) displays the value among males; column (2) reports the gender gap (female-male);
columns (3), (4) and (5) show the 25th, 50th and 75th percentiles of the distribution of the within-field-of-study gap.
Employer characteristics are computed as averages over the period 2014-2018. Source: Bovini et al. (2023).
Tables 3 and 4 look at other job characteristics for secondary school and university
graduates, respectively.27 Already one year after leaving secondary school, women are
The results are presented for employees only, as information on the type of self-employment is far
Table 4: Gender gap (girls-boys) in job and employer characteristics, 1 year after graduation
from university
Private employees
Public employees
Within fields
Within fields
Part-time contract (%)
Permanent contract (%)
-10.8
High-skill occ. (%)
Medium skill occ. (%)
Low skill occ. (%)
-12.0
-46.4
-89.7
-12.6
-522.4
-24.5
-20.4
-276.6
-144.6
-53.7
-21.4
-84.4
-12.8
Firm distance from birth muni. 246
Firm average age
Firm size
Co-workers’ average educ.
Female co-workers (%)
Firm VA per worker
Firm av. wage
Notes: cohorts who graduated from university (2nd-cycle and single degrees) in 2011-2018. Only graduates who work and
no longer study are considered. Column (1) displays the value among males; column (2) reports the gender gap (femalemale); columns (3), (4) and (5) show the 25th, 50th and 75th percentiles of the distribution of the within-field-of-study
gap. Employer characteristics are computed as averages over the period 2014-2018. Source: Bovini et al. (2023).
more likely than men to have a part-time contract and less likely to have a permanent
contract. Employees with a high-school diploma are mostly found in medium-skill occupations, especially so for girls (who are mainly employed as sales assistants or restaurant
staff). Interestingly, young women work closer to their birthplace (-7 kilometres compared to the male average of 93). The characteristics of the employer are also different
between genders. Firms employing females who just graduated from secondary school
are on average younger, smaller (22 employees; 31 for boys), and have a much higher
proportion of female employees. Notably, they pay roughly 9% lower wages on average
and have 18% lower value added (VA) per worker (used as a proxy for firm productivity).
Median within-field gaps are in most cases of similar magnitude to aggregate ones. In
line with what documented so far, this suggests that sorting into different high-school
tracks only explains a fraction of early career gender gaps and that, from the very start,
female graduates hold different, lower-quality, jobs than same-track male graduates.
Table 4 looks at university graduates. Compared to recent secondary school graduates, they have more stable jobs28, are employed in higher-skill occupations and work
in firms that are farther away from their birthplace, older, larger, higher-paying, and
more productive. Gender gaps in job characteristics are marked for this population as
scarcer. Given the very small share of public employees with just a high school attainment (less than
1%), Table 3 focuses on private sector employees.
The share of temporary workers appears nonetheless very high; this is because in Italy firms typically hire workers on a temporary basis, and then they convert their contract to a permanent one (see
for instance, Sestito and Viviano, 2018). Five years after graduation, the share of employees with a
permanent contract is equal to 75% on average (and larger for boys).
well. They go in the same direction as those documented for high-school graduates. Importantly, however, for university graduates median within-major gaps are substantially
smaller than the aggregate ones along all the considered dimensions, confirming the importance of gender differences in the choice of majors, as documented in the previous
section.
Moreover, Figure 20 shows that these initial allocative gender differences in the firms
of first employment do not fade away over time. Five years after graduation boys and
girls are equally likely to have remained in the firm of their first employment (grey bars);
among those who changed employer, however, boys are more (less) likely than girls to
have moved into a more (less or equally) productive firm than the starting one.
Figure 20: Gender gaps in employer productivity (value added per employee)
(a) Allocation 1 year after graduation
(b) Movements along the productivity ladder
(shares)
between 1 and 5 years after graduation (shares)
Notes: cohorts who graduated from upper secondary school or university (2nd level or one-cycle degree) in the period
2011-2018 (panel a) and 2011-2014 (panel b). Only graduates who work and no longer study are considered. Panel a:
boys’ and girls’ allocation into fifths of the distribution of employer productivity (value added per worker) 1 year after
graduation. Panel b: red (blue) bars capture the probability of falling down (climbing up) the productivity ladder, i.e.
moving to a firm with productivity lower or equal (higher) than that of the origin firm; grey bars represent the probability
of remaining in the same firm. Private sector employees only, as the information on productivity is not available for public
sector employees. Source: Bovini et al. (2023).
It is then important to understand how much differences in job characteristics, as
those in Tables 3 and 4, explain earnings’ within field gender gaps. Figures 21 and 22
present the results of the Oaxaca-Blinder decomposition of the average daily wage gap for
private-sector employees for whom there is no missing information on job and employer
attributes (i.e., the same sample as in Figure 19). The Oaxaca decomposition is run
separately for each high-school track/university major. To summarise the results of this
set of Oaxaca-decompositions, Figures 21 and 22 plot the 25th, 50th and 75th percentiles
of the distribution of the fraction of the gender gap within each field that explained by
each job attribute.29 The control variables included in the Oaxaca decomposition are
More formally, let J be the number of fields of study and I the number of job attributes; eji is
the fraction of the gender gap within field j explained by each job attribute i and uj is the part of
the gender gap within track j that remains unexplained after controlling for all job attributes i. The
reported in detail in the footnote of the table.
Figure 21 shows that a higher prevalence of part-time contracts and of lower-paying
occupations plays the largest role in explaining the within-field wage gender gaps for
secondary school graduates. Focusing on employer attributes, sorting of female graduates
into smaller, younger and less productive firms than male graduates matters the most.
It is also interesting to notice the non-trivial role of sorting of women into firms with a
higher share of female co-workers: this evidence could signal non-wage amenities that are
particularly valued by young women. Nevertheless, despite the availability of this rich
set of controls, the within-field unexplained component of the gender early career gap is
around 25% at the median (40% at the 75th percentile of the distribution).
Figure 21: Within field of study gap: Oaxaca-Blinder decomposition of average (log) daily
wage 1 year after graduation, secondary school graduates (percent)
Notes: cohorts who graduated from upper secondary school in the period 2011-2018. Only graduates who work and no
longer study are considered. “Part-time” and “Temporary” include dummies for whether the contract is part-time or
temporary, respectively. “Firm prod., size, age” contains the following information on the employer: age, size, productivity
(i.e. value added per worker), and their squares; “% female workforce” includes the share of the workforce in the firm who
is female, and its square; “% edu. workforce” consist of the average education (in years) of the workforce in the firm, and
its square; “Firm distance” captures the distance between the birth municipality and the work municipality, aggregated
in 10 bins; “Sector” and “Occupation” are 2-digit occupation and sector fixed effects, respectively. Finally, “Unexplained”
captures the fraction of the gap not explained by the aforementioned factors. The decomposition is performed on the
sample of private sector employees for whom these controls, as well as – for the sake of comparability – those that have
been included in later Oaxaca decomposition (see Figure 19), are observed. Source: Bovini et al. (2023).
Figure 22 displays results for university-educated private sector employees. As for
high-school-educated employees, the job attribute that matters the most is whether the
contract is part-time; the most relevant employer attributes are size, age and productivity.
However, the residual (i.e., the unexplained) portion of the within-field gender gap is far
Figures then plot the 25th, 50th and 75th percentiles of the distribution of (e1i, e2i, .., eJ?1
, eJi )i=I
i=1 and of
J?1
(u, u, .., u
, u ).
larger than for secondary school graduates: 40% at the median and around 55% at the
75th percentile of the distribution. This suggests that, while the choice of majors is a very
important determinant of the gender pay gap among university graduates, within-major
differences in firm and job characteristics play a relatively little role; the residual gender
gap is probably related to factors that cannot be observed even in the very granular data
available to Bovini et al. (2023) (like for instance, negotiation skills or availability to work
very long hours, see Chapter 6).
Figure 22: Within field of study gap: Oaxaca-Blinder decomposition of average (log) daily
wage 1 year after graduation, university graduates (percent)
Notes: cohorts who graduated from upper secondary school in the period 2011-2018. Only graduates who work and no
longer study are considered. “Part-time” and “Temporary” include dummies for whether the contract is part-time or
temporary, respectively. “Firm prod., size, age” contains the following information on the employer: age, size, productivity
(i.e. value added per worker), and their squares; “% female workforce” includes the share of the workforce in the firm who
is female, and its square; “% edu. workforce” consist of the average education (in years) of the workforce in the firm, and
its square; “Firm distance” captures the distance between the birth municipality and the work municipality, aggregated
in 10 bins; “Sector” and “Occupation” are 2-digit occupation and sector fixed effects, respectively. Finally, “Unexplained”
captures the fraction of the gap not explained by the aforementioned factors. The decomposition is performed on the
sample of private sector employees for whom these controls, as well as – for the sake of comparability – those that have
been included in later Oaxaca-decomposition (see Figure 19), are observed. Source: Bovini et al. (2023).
Summing up the evidence on early career gender wage gaps
Bovini et al. (2023) then summarise the extent to which differences in educational choices,
academic achievement and job characteristics, explain the early career gender gap. This
is done by estimating regressions that progressively add controls at a higher level of granularity than how is possible in the Oaxaca-Blinder decomposition. They find that, after
controlling for all variables that capture education choices and academic achievement,
the raw (unconditional) gap in daily wages one year after graduation shrinks by almost
30% for high-school-educated workers and by almost 60% for college-educated ones and
that differences in fields of study play by far the largest role. The larger contribution of
educational choices for university students relative to secondary school students probably depends on the higher degree of specialisation of university majors with respect to
secondary school tracks.
Moreover, they find that all controls that capture differences in job attributes and
employer characteristics, net of education effects, account for about 45% of the raw gap
for secondary school graduates and 20% for university graduates.
Finally, even with the very detailed set of available controls included in the regressions,
approximately 25% of the gender early career wage gap remains unexplained among
secondary school graduates (20% among university graduates); this share is higher at the
top of the wage distribution. This unexplained component is probably due to differences
in characteristics that are difficult to measure and observe, like negotiation skills or noncognitive traits.
Policy implications
This chapter documents four important facts. First, girls on average achieve higher
educational levels and outperform boys in education. Second, girls select very different
university majors and high school tracks from boys, and in particular they choose subjects
that involve considerably worse labour market prospects. Bovini et al. (2023) find that
the choice of the field of study explains about 60% of the wage gap 1 year after graduation
among university graduates. Among secondary school graduates high school track choices
still matter substantially (they explain 30% of the gap) but the largest contribution comes
from differences in jobs characteristics: even among men and women who graduate from
the same high school track, women are employed under lower-paying contracts (parttime) and in lower-paying and less productive firms. Finally, approximately 20 to 25% of
the gender gap remains unexplained by the very detailed set of available controls included
in the regressions.
Overall from these results it emerges how actions should be taken also at the time
when boys and girls choose their high school track and, especially, their university major.
As discussed in this chapter, the literature agrees that differences in the choice of the
field of study mostly depend on preferences, which are largely shaped not by innate
and immutable factors, but rather by the school and home environment, and by cultural
norms and stereotypes. It is therefore possible for policymakers to act in order to partially
overcome some of these cultural barriers, for example through interventions at school
that expose students to positive role models (women working in STEM jobs or occupying
high-level positions) or that raise awareness of parents’ and teachers’ (even implicit)
stereotypes.
The existing evidence indicates indeed that role models can be very effective in reduc-
ing some of these cultural barriers. As mentioned before, female students’ performance
and enrolment in STEM and other male-dominated subjects are positively affected by
the presence of a female math teacher and the interaction with high-performing female
peers. Moreover, some papers indicate that even very small interventions — that shortly
expose female students to women working in male-dominated occupations — can have
large and long-lasting effects.30
Finally, policies aimed at reducing teachers’ or parents’ biases can induce students
to modify their behaviours (see also Chapter 6). For instance, schools and universities may favour exam formats that limit possible biased behaviours (like written, blind
graded tests). Moreover, Alesina et al. (2018) show that revealing teachers’ implicit biases improves school performances of discriminated students. Importantly, the effect only
emerges if teachers are given information about their own biases: generic debiasing messages informing them of the presence of false beliefs toward certain demographic groups
are not sufficient.
Breda et al. (2018) show that in France a brief exposure to female role models working in scientific
fields largely affected high school students’ perceptions and induced them to enrol in selective and maledominated STEM programs in college, especially among high-achieving girls. Also Porter and Serra
(2020) show that female students exposed in introductory classes to successful women who majored in
economics at the same university are more likely to enrol in further economics classes and to complete
their economics degrees.
Maternity and labour market outcomes
This chapter extensively investigates the evidence on the relationship between female
employment and fertility decisions. Section 4.1 first discusses how, at the aggregate level,
the correlation between these two variables has changed across OECD countries over
time, turning from negative in the 1980s to positive in the 2000s. Second, it shows that
child penalties — the negative effects of parenthood on the careers of mothers — are
large and observed everywhere. In Italy, a sizeable share of women still exits employment
at childbirth, even if this is less so relative to past decades. Among those who remain
employed, women tend to work less after childbirth and this translates into lower earnings
in the long run. Indeed, the asymmetrical impact of parenthood on men and women is
still considered the main driver of the remaining gender gaps in many developed countries
(Angelov et al., 2016; Kleven et al., 2019).
Family-friendly policies — like parental leaves and subsidised childcare — may play a
significant role in boosting female labour supply and reducing gender gaps. Informed by
the vast literature on the topic and given the system of family-friendly policies currently
in place in Italy, Section 4.2 outlines the main areas of policy interventions, focussing in
particular on expanding childcare facilities and providing more generous paternity leaves
to increase fathers’ involvement in household responsibilities.
The last part of this chapter returns to the relationship between employment and
fertility, looking in particular at the causal impact of employment on fertility. This issue
is particularly relevant in a country like Italy, which is characterised by both very low
female employment and fertility rates. The literature shows that this relationship varies
according to the sample of women analysed (Section 4.3). If one focuses on women who,
for exogenous reasons, lose their job, the effect of employment on fertility is positive; but if
one looks at other contexts, where women are at the margin of the participation decision,
the sign may change. In general, working time flexibility and job stability positively
affects fertility.
The “cost” of motherhood in the labour market
Female employment and fertility are two highly interrelated and jointly determined decisions, also affected by common factors, including culture and social norms.
The macro literature has shown that the relationship between female employment
and fertility rates has evolved and reversed over time. Doepke et al. (2022) show that
the negative relationship between fertility and female employment rates observed in the
1980s has turned to be positive in high-income countries that feature ultra-low fertility
rates since 2000. According to the authors, in high-income countries, the compatibility of
women’s career and family goals is now a key driver of fertility decisions. New-generation
fertility models refer to a shift in women’s aspirations to explain the change in the rela38
tionship between employment and fertility. While in the past women believed that work
and family were incompatible, they now aspire to have both — likewise men. Thus, in
countries where it is easy to combine career and family, women have both; in countries
where the two are in conflict, women are forced to choose between the two, leading to
both fewer children being born and fewer women working. In this sense, family-friendly
policies which aim to help parents — especially women — to combine work and family
responsibilities seem to play a crucial role.
This evidence is consistent with the results of Barbiellini Amedei et al. (2023) that
provide a long-run estimation of the relationship between fertility dynamics and female
labour force participation in Italy after WWII. The authors show that the unconditional
correlation across all Italian regions by year is strongly negative up to the mid ’90s,
turning positive in early 2000s (see Figure 23). A more formal econometric analysis —
based on panel cointegration techniques — confirms that the negative correlation holds
until 1995, then it weakens and becomes not significant afterwards.
Figure 23: Cross-correlation: Female participation rates (FPR) and Total Fertility rates
(TFR), Italy (1951-2019)
correlation TFR-FPR
Notes: The figure plots the unconditional correlation between fertility and female employment rates across Italian regions
by year. Source: Barbiellini Amedei et al. (2023).
Barbiellini Amedei, F., S. Di Addario, M. Gomellini, and P. Piselli (2023). Female
labour force participation and fertility in Italian history. Temi di Discussione (Working
papers) forthcoming, Bank of Italy.
Moving to micro data, simple descriptive evidence based on the the European Labour
Force Survey (EULFS; Figure 24) shows that marriage and parenthood differently correlate with employment across genders. In the main European economies married women
experience lower employment rates than unmarried women especially at fertility ages,
while the opposite is observed for men (Figure A.5). In Italy, and to a lesser extent in
Spain, this gap persists over time; in France and in Germany, instead, the gap observed
at childbearing age cancels out at older ages.
Figure 24: Employment rates along the life cycle, by country and marital status – Women
Employment rate
0 .2 .4 .6 .8 1
Employment rate
0 .2 .4 .6 .8 1
married
singles
married
singles
Employment rate
0 .2 .4 .6 .8 1
Employment rate
0 .2 .4 .6 .8 1
married
singles
married
singles
Notes: 30 years old is the mean age of women at birth of first child across the selected countries in 2019. Source:
European Labour Force Survey, 2019.
Some studies aimed at estimating the causal impact of fertility on maternal labour
outcomes use preferences for a mixed sibling-sex composition (Angrist and Evans, 1998) or
success at IVF (in vitro fertilization) treatments (Lundborg et al., 2017) as instruments
for childbearing. However, the generalization and external validity of their results is
rather limited since estimates are based on small samples and refer to very particular
treatment effects (i.e., having at least a third child) and samples (i.e., large families or
those who access IVF treatments).
Other studies have analysed the trajectories of earnings of men and women, and how
they diverge during childbirth, for highly-educated professionals. Bertrand et al. (2010)
document that female, but not male, MBA graduates employed in the financial sector
experience a significant loss in earnings at childbirth. Azmat and Ferrer (2017a) observe
the same evidence among lawyers.
Thanks to the growing access to large administrative data, a rich literature that
estimates the child penalty in the overall population has recently flourished (seminal works
are Angelov et al. 2016; Kleven and Landais 2017). The child penalty is, broadly speaking,
the negative effects that childbirth — typically the first — has on labour outcomes of
mothers relative to fathers (or similar women without children). Under given identifying
assumptions this provides evidence of the causal impact of fertility choices on maternal
labour outcomes.31
Child penalties are relevant, long-lasting (up to 20 years) and widespread in all countries, even in those considered as the benchmark for gender equality — like Nordic countries (Angelov et al. 2016 for Sweden; Kleven et al. 2019 for Denmark; Sieppi and Pehkonen 2019 for Finland; Andresen and Nix 2022 for Norway).32 Goldin et al. (2022) show
that in the US the motherhood penalty (with respect to non-mothers) lowers over time
when children grow up and women increase their working hours; however, fathers expand their relative premium — with respect to non-fathers observed after childbirth —,
particularly among those with a college degree.33
De Philippis and Lo Bello (2023) and Casarico and Lattanzio (2023a) contribute to
this literature by investigating the child penalty in Italy on different maternal labour
supply outcomes employing Social Security administrative data provided by the Italian
National Social Security Institute (INPS) and data from the Italian Labour Force Survey
(ILFS). The former looks at the effect of childbirth on the employment gap between
mothers and non-mothers. The latter assesses the earnings gap relative to non-mothers
by looking at mothers who continue to work after childbirth, thus focusing on differences
along the intensive margin of labour supply.
De Philippis and Lo Bello (2023) first show that the main driver of the reduction in the
gender employment gap observed between 1984 and 2019 has been the significant decline
in women’s flows out of the labour market that is typically observed around fertility age.
Figure 25, highlights a considerable rise in employment rates of married women at fertility
and subsequent ages across different generations.
Then, the authors turn to Social Security administrative data to assess the child
penalty in flows into and out of employment and how this has evolved over time. Comparing the employment trajectories before and after childbirth of mothers and non-mothers,
they find that childbirth almost doubles the probability of quitting employment (Figure
26, Panel a).34 Despite to a smaller extent, this increase is observed even 15 years later.
The main approach is based on event study analyses around the birth of the first child. This relies
on the idea that, while fertility choices are not exogenous, the event of having a child generates sharp
changes in labour market outcomes that are arguably not correlated to other unobserved determinants,
which should evolve smoothly over time. The identification of the child penalty in the short run relies on
the assumption that absent maternity the outcome would be evolve smoothly. The identification of the
long-run penalty relies on parallel trends between men and women, conditional on controls for time and
lifecycle trends. While event-studies cannot formally address endogeneity of fertility, Kleven et al. (2019)
prove that the event-study estimates of the effect of a third birth are close to the IV-based estimates
that use twins or sex composition as an IV for a third birth.
Kleven et al. 2019 provide estimates of the child penalty in earnings for six developed economies
(Denmark; Sweden; US, UK, Austria and Germany); de Quinto et al., 2021 provide evidence for Spain;
Cort´es and Pan, 2023 for the US; Martino, 2022 for Italy.
The authors show that in the US the parental gender gap in earnings is lower for highly educated
with respect to low-educated mothers but still substantial.
De Philippis and Lo Bello (2023) and Casarico and Lattanzio (2023a) follow Kleven et al. (2019) in
assigning placebo births to non-mothers; they do so by assigning a random draw from the distribution
Figure 25: Employment probability along the life cycle and across generations in Italy
(a) Women
(b) Men
Notes: Share of employed individuals over age (15-74), by marital status. The authors use married individuals to proxy
for individuals with children, since this information is not available in the ILFS before 2004. The definition of married
individuals includes all cohabiting couples. Source: De Philippis and Lo Bello (2023).
At the same time, the probability of entry into employment is lower around childbirth and
goes back to the original level approximately 5 years after maternity (Figure 26, Panel
b). Overall, the change in employment is largely negative after childbirth. However, the
authors also find that child penalties in employment probabilities have strongly decreased
over time: the effect of childbirth on the probability of exiting and entering employment
has almost halved for younger cohorts with respect to older generations.
Figure 26: Child employment penalties
(b) Non-employment-to-employment flows
(a) Employment-to-non-employment flows
Notes: The figure shows the evolution of the probability of exiting employment (Panel a) and entry into employment
(Panel b) relative to the year before the birth of the first child for women with children compared to those who never
have children (assigning placebo births based on the observed distribution of age at first child among those who have
children; see De Philippis and Lo Bello, 2023 for details). In the figure the authors control for year/region, age and year
of first contribution fixed effects. The vertical bars indicate the 95% confidence intervals based on robust standard errors.
Individuals are considered employed if they worked at least one month in year t. Source: De Philippis and Lo Bello
(2023).
Finally, De Philippis and Lo Bello (2023) combine the information on fertility trends
of age at birth of mothers to non-mothers.
with the estimated evolution of the employment child penalty to decompose how much of
the increase in the female employment rate observed between 1984 and 2019 reflects these
determinants. Overall, declining fertility rates and reductions in the child employment
penalty explain approximately 70% of the observed growth; the reduction of the child
employment penalty alone accounts for 60% of the change. One of the main drivers
of this reduction has been the increase in women’s educational attainment, since highly
educated women are characterised by stronger attachment to the labour market and lower
exit rates. Moreover, the authors show that closing the existing child penalty among new
mothers would increase female employment rate by 6.5 p.p. by 2040 (covering 38% of the
current gender employment gap). Removing the child penalties for both new and existing
mothers, female employment would raise by 14 p.p. already by 2030, thus closing 85% of
the existing gender gap.
Casarico and Lattanzio (2023a) find that the birth of the first child entails a sizeable
and long-lasting reduction in mothers’ annual earnings with respect to non-mothers,
even if one focuses only on women who remain in the labour market after childbirth.
Fifteen years after childbirth this penalty equals 52% relative to the year before maternity.
Most of the child penalty observed on mothers is due to a reduction in weeks worked
(approximately 70%; Figure 27). These are rather large long run estimates if compared to
the evidence provided in Kleven et al. (2019),35 especially if one considers that focusing on
mothers who remain in the labour market likely leads to underestimate the total penalty
in annual earnings.
Figure 27: Child penalty in earnings
Child penalty, coefficients relative to t-1
Wage related penalty
Weeks related penalty
-0.20
Part-time related penalty
-0.40
-0.60
Total child penalty in annual earnings
-0.80
-5 -4 -3 -2 -1 0
Years from maternity
Notes: The figure reports the child penalty in log annual earnings, and its decomposition into the contribution of reduction
in weekly wages, reduction in weeks worked and switch to part-time. Source: Casarico and Lattanzio (2023a).
Kleven et al. (2019) provide estimates of the child penalty for Denmark, Sweden, US, UK, Austria,
Germany. The long-run child penalty, measured ten years after childbirth and as the difference in earnings
between mothers and fathers, is 21-26% in the Scandinavian countries, 31-44% in the English-speaking
countries, and 51-61% in the German-speaking countries.
Using the same Social Security administrative data as described before, De Paola
and Lattanzio (2023), provide evidence that the child penalty has expanded during the
economic crisis due to the COVID-19 pandemic. Figure 28 shows that working mothers
experienced a larger penalty in terms of reduced labour market earnings compared to
working fathers throughout the period March 2020-May 2021, and that this was mainly
driven by a substantial fall in the number of days worked. The penalty is larger for
mothers of younger children, for those working in non-essential activities36 and for those
living in couples where the pre-pandemic mother-father pay gap was larger, suggesting
that second earners with lower bargaining power suffered a higher penalty. This means
that both demand and supply factors played a role in explaining the gendered impact of
the economic crisis triggered by COVID-19.
Figure 28: Impact of the COVID crisis on earnings of fathers and mothers
Notes: The figure reports the estimated difference in log annual earnings for workers in 2020 and 2021 relative to 2019
in each month using February as a reference. Control variables include: labour market experience, age, dummy for whitecollar workers, the number of children, dummies for workers taking the parental leave and COVID-19 leave, dummy for
workers in short-time work compensation schemes and region dummies. Source: De Paola and Lattanzio (2023).
The debate on the determinants of the child penalty is still ongoing (see Cort´es and
Pan (2023) for an extensive review). First, there is evidence that after childbirth mothers
work in less productive firms (Casarico and Lattanzio, 2023a) preferring other amenities
(like shorter commuting time Le Barbanchon et al., 2021 or job flexibility Bang, 2021;
De Philippis and Lo Bello, 2023) to higher wages. Second, there is large evidence showing
that conservative gender norms play a significant role in determining the child penalty
Prime Minister’s decree n.64 of 11th March 2020 established the nationwide March lockdown and
specified the activities that were deemed as essential and could continue to operate, and those that
were classified as non-essential and were forced to shut down (the Prime Minister’s decrees n. 76 of
22nd March 2020 and n. 79 of 25th March 2020 further specified the definition if essential and nonessential activities). The former mainly include agriculture, some manufacturing, energy and water
supply, transports and logistics, ICT, banking and insurance, professional and scientific activities, public
administration, education, healthcare and some service activities. Shutdown sectors include most of
manufacturing activities, wholesale and retail trade, hotels, restaurants and bars, entertainment and
sport activities.
as they attribute to women the full burden of childbearing (Kleven et al., 2019; Kleven,
2022; Andresen and Nix, 2022; De Philippis and Lo Bello, 2023 provide evidence for Italy).
Finally, public policy, too, can influence the child penalty, by allowing better work-life
balance or by changing gender norms. Here the evidence is mixed, as more extensively
illustrated in the next subsection.
The role of family policies
Parental leave and childcare policies are the two main family-friendly programmes adopted
in developed countries to help parents, especially mothers, to reconcile work and family
responsibilities.
Parental leaves are job-protected leaves for working parents. Across OECD countries
there are three types of leaves. Maternity leaves are specifically reserved for mothers
and are granted around the time of childbirth; parental leaves are shareable entitlements
among parents but usually taken by mothers, as we will further explore in Subsection
leaves should increase fathers’ involvement in caring activities, rebalancing domestic work
among partners with positive effects on maternal labour supply. Finally, subsidised childcare is intended to provide families with an accessible market care alternative to maternal
one, thus freeing up mothers’ time for market work.
Because extending parental leaves or providing highly subsidised child care entail large
costs for public budgets, understanding their causal impact on female labour supply is
crucial. Despite the positive cross-country correlation between the duration of leave (or
publicly provided child care) and female participation and employment rates,37 empirical
works aimed at estimating the causal effect of similar policies on female labour supply
have found mixed and non-conclusive results. More in general, these policies are highly
debated not only because of their cost. Some scholars (see Rossin-Slater, 2017; Carta,
2019; Canaan et al., 2023 for a review) argue that long leaves may reduce mothers’ labour
market attachment and depreciate their human capital; they could increase employers’
expected labour costs at the time of hiring women of childbearing age; formal child
care may be detrimental for children’s development, especially if it replaces high-quality
maternal care. In the next subsection we discuss the empirical literature on these relevant
topics.
Ruhm (1998); Olivetti and Petrongolo (2017) find a non-monotonic relationship between the duration of parental leave and female outcomes: short or intermediate leaves are associated with higher female
employment rates and no wage effects, while longer entitlements lead to negligible effects on employment, but negative ones on wages. According to Olivetti and Petrongolo (2017) childcare expenditure is
more positively correlated with smaller gender gaps since, allowing mothers to go back to work earlier,
subsidised childcare avoids losses of human capital or of working experience.
Maternity and parental leave policies
The impact of leave-taking on female employment is theoretically ambiguous since it
may involve two different groups of women: i) those who would have otherwise remained
employed and on the job, ii) those who would have otherwise quit their jobs. For women
in the first group, the leave policy does not have any impact on short-run employment
since they would have remained employed anyway. If any, the policy increases time away
from the job with negative consequences on their careers and wages, thus feeding the child
penalty. For women in the second group, the policy may increase short-run employment
by providing a period of leave instead of quitting their jobs, which contributes to reduce
non-employment spells and the child penalty. We focus only on labour supply, but there
may also be labour demand effects: longer parental leaves, especially if mostly used
by women, may raise the costs of hiring and promoting women of childbearing age, in
anticipation of their future leave-taking behaviour, fostering what is labelled as statistical
discrimination.
In general, the impact of leave policies on maternal labour supply mainly depends on
the characteristics of the program: the length of the leave; the replacement rate paid;
the degree of job protection. Most of the empirical works that assess the causal impact
of paid leaves exploit changes in their duration. The evidence is mixed.
Lalive and Zweim¨
uller (2009) find that leave duration negatively affects maternal
earnings in the short run (during the first three years after birth) since mothers delay
return to work even after the benefits are exhausted; however, effects are small in the
long run. Similar employment effects are also detected by Sch¨onberg and Ludsteck (2014);
they also highlight the role of job protection in determining job attachment and positive
employment effects (see Zurla 2022 for evidence on Italy). According to Kluve and Tamm
(2013); Bergemann and Riphahn (2010) shortening long parental leave duration (from 24
to 12 months) in combination with an increase in the replacement rate brought positive
employment effects in Germany in the medium and long term. In Norway, Dahl et al.
(2016) find no significant impacts of a variety of extensions in paid maternity leave from 4
to 8 months on either earnings or labour force participation among mothers. Employment
effects are instead positive and stronger when the parental leave extensions are shorter
to begin with (Baker and Milligan, 2008; Baum and Ruhm, 2016). Overall, the literature
has established the existence of a concave relationship between the length of parental
leave and maternal labour market outcomes. Short parental leaves (up to approximately
6 months) improve maternal labour market outcomes. Up to 1 year these programmes
seem to have limited effect, while leave entitlements longer than 1 year seem to have
adverse effects on mothers’ wages and employment.
Carta et al. (2023) study a policy that increased the duration of the unemployment
benefit from 8 months to up to 24 months. The unemployment benefit can be seen as a
job leave right with no job protection paid at a higher replacement rate relative to the
parental leave (75 and 30%, respectively).38 The authors show that the higher economic
convenience to take the leave without job protection significantly increased mothers’
probability to quit at childbirth, resulting into a higher non-employment probability
up to at least 18 months after childbirth (see Figure 29). In response to the higher
turnover of new mothers, firms tend to hire more male workers with respect to women.
This would imply that the reform has deterred hiring childbearing age women, limiting
their employment and career opportunities. These results are informative on the role of
job protection in determining mothers’ labour market attachment and how policies can
backfire and contribute to statistical discrimination against women.
Figure 29: Quit and non-employment probability of new mothers — response to an increase
in the duration of the unemployment benefit
Notes: Each dot represents the change in the quit/non-employment probability of new mothers having their child after
the UB reform relative to the period before and exposed to a 100 days increase in their potential UB duration, between
the 6 months before until 18 months after childbirth. The confidence intervals are obtained from cluster-robust standard
errors at the individual level. Other controls are included. Source: Carta et al. (2023).
On top of understanding the effects of maternity and parental leave on maternal
employment, it is also crucial to study the determinants of the leave-taking behaviour
so as to create a more women (parent)-friendly environment. As we will further discuss
countries, and especially low in Italy. Rossin-Slater (2017) survey the main literature on
the determinants of leave-taking behaviour. On the one hand, lack of awareness of the
existing policies, low (or lack of) pay and absence of job protection may prevent workers
from taking up the leave. On the other hand, the literature has shown that cultural
norms play a crucial role: peer effects and role models may reduce stigma and foster
leave taking behaviour (Dahl et al., 2014).
This is true only for new mothers, within the first year of the child; they can access unemployment
benefits even if they voluntarily resign from their job rather than being laid off.
Dottori et al. (2023) estimate the relevance of peer effects in leave-taking behaviour
for Italian mothers. The peer effect might be relevant along two main channels. First,
colleagues who previously took parental leaves can provide useful information on their use;
second, looking at their experience, mothers can infer about possible earnings reductions
or penalties associated with their use. The paper exploits a reform, implemented in
Italy in 2015, that extended from 0-3 to 0-6 the child’s age over which parents can take
advantage of paid parental leave. They study whether post-reform mothers are more likely
to use the paid parental leave when a greater share of their peers previously used it. The
latter is instrumented with the share of peer-mothers with 4-6 y.o. children who were
exposed to the reform. The authors first show that the reform increased the leave-taking
rate by 12.4% for mothers with 4-6 years old children.39 Then, they find that peer effects
increase both the probability of using parental leave and the number of weeks of leave
taken by mothers. Moreover, the peer effects determine a reduction in the probability
of working part-time, suggesting that part-time and parental leave may be substitutes.
The latter is more flexible and avoids changes in the contractual relationship that might
result in long term earnings losses. The effects are concentrated among mothers with low
tenure in the firm, suggesting that the channel through which the peer effect operates is
that of providing information to the employees about the employer’s reaction to the use
of parental leaves by female employees.
Paternity leaves
Paternity leave is reserved for fathers, i.e. it is not transferable to mothers. Paternity
leaves should increase fathers’ involvement in caring activities, rebalancing domestic work
among partners with positive effects on maternal labour supply. OECD countries have
only recently introduced paternity leaves, in addition to maternity and shareable parental
leaves. Thus, the available empirical evidence on their effects on maternal labour supply
and the allocation of household chores is more limited. Cools et al. (2015) show that
a specific paternal quota of parental leave of 4 weeks introduced in Norway in 1993 did
not affect maternal labour supply and slightly reduced fathers’ earnings in the short run.
Rege and Solli (2013), adopting a different empirical strategy, find a penalty on fathers’
earnings associated with the same policy in the medium and long run, but again no
effects on maternal employment, resulting into an overall increase in home production.
The introduction in 2007 of 2-weeks of voluntary paternity leave in Spain increased home
production by fathers (Gonz´alez and Zoabi, 2021), fostered maternal labour supply (Farr´e
and Gonz´alez, 2019) and changed gender norms across generations: children raised in
couples where fathers took paternity leaves tend to have more progressive and egalitarian
gender views (Farr´e et al., 2022). Overall, the evidence from the Nordic countries — the
The reform did not change the economic convenience to take parental leaves for 0-3 year-old-children.
first to introduce paternity leave — show null or modest effects on paternal and maternal
labour supply, while the effects estimated in countries with low female employment rates
(like Spain or Germany) seem to go in the expected direction of a lower specialisation
of domestic chores within the household. In countries characterised by lower female
participation there should thus be more room for such adjustments.
Subsidised childcare
Even more debate surrounds the estimated effects of childcare services on female labour
supply (see Olivetti and Petrongolo 2017 for a review of the literature). The literature
looking at the causal effect of subsidised childcare on maternal labour supply is vast and
reaches different results, varying not only across countries but also across individuals
within the same country. The heterogeneity of the estimates depends on the level of
maternal employment rate before the policy intervention, the existing formal alternatives
to childcare, and the age of the targeted children.
According to most US studies, the availability of public kindergarten for 4-year-old
(Fitzpatrick, 2010; Wikle and Wilson, 2022) and 5-year-old kids (Cascio, 2009; Barua,
2014; Fitzpatrick, 2012) generates only small increases in maternal employment (typically
limited to single or less educated mothers), mainly because publicly provided care replaces
market care.40 Similar negligible or small results are found for European countries such
as Norway (Havnes and Mogstad, 2011) and France (Goux and Maurin, 2010) where,
unlike the US, services to families are generally publicly provided. These papers look at
episodes of expansion of subsidized childcare that took place in a context of already large
public provision and high female labour supply. Labour demand and general economic
conditions are also crucial to get sizeable effects of subsidised childcare on maternal
labour supply. Nollenberger and Rodriguez-Planas (2015) show that in Spain, despite
the low public childcare provision, very few private alternatives and low female labour
supply, an expansion of public full-time childcare for 3 y.o. children determined a modest
increase in maternal labour supply due to a context of extremely low labour demand and
depressed wages. Finally, Kleven et al. (2020) show that increases in the provision of
public childcare provision failed to increase maternal labour supply in either the short
or long run in Austria, mainly due to the availability of free childcare by relatives and
strong preferences for maternal care (i.e., gender norms).
On the other hand, the introduction of highly subsidised childcare for younger children (generally 0-3 year-olds) did prove successful in boosting female labour supply in
Canada, Baker et al., 2008), Germany (Bauernschuster and Schlotter, 2015; M¨
uller and
Exceptions are Gelbach (2002), who finds significant — albeit smaller — effects on married mothers
too looking a reform that took place in 1974, and Herbst (2017), who, nevertheless, focuses on a very
peculiar setting, i.e. the provision of childcare during World War II. In both cases, the baseline maternal
employment rate was low.
Wrohlich, 2020) and Argentina (Berlinski and Galiani, 2007). Also Carta and Rizzica
(2018) find that access to early kindergarten, a much cheaper option than nurseries for 2year-old children implemented in Italy, significantly increased maternal participation and
employment. Positive sizeable effects are concentrated in the Northern regions, where
labour market conditions are more favourable and less traditional gender norms prevail.
In Italy also providing care for older children seems to boost the maternal labour
supply. On this point, Bovini et al. (2023) study the short- and medium-term effect of
increasing the length of the school day in primary education on parental labour supply.
They find that attending a long school day (full-time) in primary school has a positive
effect on maternal labour force participation and employment (approximately 2 p.p., concentrated among lower-educated mothers). Moreover, the effect persists in the medium
term, even when students are no longer in a long-day schedule. No effect is found on
fathers’ employment.
Overall, the take-home message of this literature is that providing childcare services
to younger children has positive effects on maternal employment when the latter is at
very low levels, there are few alternatives (both formal and informal) to maternal care,
and the overall culture is favourable to childcare use and maternal labour supply. Finally,
these kinds of reform should be implemented when economic/labour demand conditions
are good enough to produce positive employment effects.
Parental leaves and childcare policies in Italy
The Italian parental leave system envisages three different parental leave entitlements for
employed parents (see Carta, 2019 for a more detailed review).
First, mothers have access to 5 months of compulsory maternity leave paid at 80% of
their earnings around the time of childbirth.41 Almost all collective agreements establish
that employers provide the remaining 20%. In order to compare the length of the leave
periods across OECD countries,42 we consider the Full Rate Equivalent (FRE) number
of weeks, which is the length of the paid leave in weeks if it were paid at 100 per cent
of previous earnings — a figure provided by the Family OECD Database (OECD, 2023)
that is given by the product of the number of weeks of leave and the replacement rate.
According to this measure, Italy provides a rather generous maternity leave: in 2022 the
FRE number of weeks in Italy was 17.4 (the figure considers only the leave paid by the
government and not the top-up provided by employers), vis-`a-vis 16.0 weeks in Spain,
14.6 in France and 14 in Germany.43
Mothers can choose how to use the leave: either 2 months before childbirth and 3 afterwards, or 1
month before and 4 after, or using all the leave after childbirth.
Following the ILO recommendations (International Labour Organization, 2012), at present almost
all the OECD countries — with the exception of the US — have in place paid maternity leave rights.
Not the entire maternity leave needs to be compulsory. Usually maternity leaves envisage a compulsory period very close to childbirth (like in Spain and in France).
Second, Italian working parents — both mothers and fathers — are eligible for parental
leave. Differently from the maternity leave, this is taken on a voluntary basis. The
parental leave for each parent — working as employee — is up to 26 weeks (six months);
the sum of the two periods cannot exceed ten months (eleven if the father enjoys the
leave for at least three months) and the leave expires at the child’s 12th birthday. The
total leave is paid at 30% of average earnings for the first nine months, while the rest of
the leave is unpaid. Each parent has a quota of three months of paid leave while the last
three months can be shared.44 The availability and generosity of paid parental leaves vary
considerably across countries. The average entitlement available to mothers (or fathers,
excluding periods specifically reserved to them) among OECD countries is just over 37
weeks, with most of those countries that offer at least one week providing somewhere
between 26 and 52 weeks (see Figure 30a).45 Most countries provide benefits that replace
somewhere around 30 to 60% of previous earnings. The lowest payment rates tend to be
found in countries with the longest entitlements. According to the OECD (2023) data,
the use of parental leave in Italy is rather low with respect to the OECD average46 and
recipients are mainly women (approximately 80%). On top of the role of social norms
and preferences, the large use of parental leave by mothers may be also due to the fact
that the household, when choosing which income to give up to take the leave, opts for the
lowest income that is typically the one earned by the wife. Interestingly, in the Nordic
countries, where there is a more equal distribution of household chores and a higher
female labour supply, the replacement rate of voluntary leave for fathers is significantly
higher than the one envisaged for mothers.47 While the different replacement rates paid
to mothers and fathers may appear as a source of inequity, a higher replacement rate for
leave-taking fathers may induce the family to choose fathers as the one taking the leave,
giving up a lower share of income with respect to the case in which the mother takes the
leave.
Finally, in Italy fathers working as employees are entitled to a compulsory paternity
leave of 10 days, that can be taken since two months before childbirth up to five months
after childbirth.48 The leave is paid at 100% of earnings. Indeed, as periods of paternity
The Law n. 234 of 30th December 2021 — in accordance with the EU Directive 2019/1158 which
established two months of paid leave for each parent not transferable to the other — increased from six
to nine months the length of the paid parental leave and established that three months are reserved for
each parent. The total length (the sum of paid and unpaid leaves) did not change, meaning that the
reform simply increased the replacement rate from 0 to 30% for three months of leave.
Twelve of the OECD countries offer no entitlement to paid parental leave at all.
The figure refers to the number of users of parental leave entitlements over 100 births, so the low
use rate may partly reflect lower employment and larger share of self-employment in Italy.
For fathers, the replacement rate is 100% in Norway, 77.6 in Sweden and 62.8 in Finland. For
mothers they are, respectively, 34.0, 57.2 and 18.7 (OECD, 2023).
The Italian legislation introduced the paternity leave in 2012 on an experimental basis initially for
three years; the leave was mainly symbolic since it envisaged only one day of compulsory leave and two
days of voluntary leave. The compulsory leave has been gradually increased over time, reaching ten days
in 2021 (Law n. 234 of 30th December 2021), in accordance with the EU Directive 2019/1158 which
leave are usually much shorter than periods of maternity leave, they are usually fully
paid. Figure 30a shows that Italy offers a rather short overall leave earmarked for fathers
in comparison with other OECD countries (it is the sum of FRE weeks of paternity leave
and the father’s quota of paid parental leave). According to INPS (2022), in 2021 only
approximately one out of three Italian fathers took the leave, despite being compulsory
for employees that represent 75% of the overall number of employed individuals.
The whole Italian parental leave system is significantly less generous relative to those
in place in the Nordic countries, while similar to systems implemented in other main European economies (Figure 30a). With respect to the latter, however, the Italian program
is relatively less generous for fathers, to whom only 20.5% of the FRE leave is reserved
(Figure 30b).49
Figure 30: Maternity, parental and paternity leave entitlements in selected OECD countries,
(b) Share of leave reserved to fathers over the total
leave entitlements
Earmarked for fathers
Earmarked for mothers
Shareable (most often taken by mothers)
(a) Duration of earmarked and shareable
paid parental leaves
Notes: Duration is measured by the Full Rate Equivalent (FRE) number of weeks in panel that is the length of the paid
leave in weeks if it were paid at 100 per cent of previous earnings. Figures refer to the laws in place in April 2022. The
figure for Italy has been updated at April 2023. Source: OECD (2023).
Other than parental leaves, governments can financially helps families with children.
The government’s interventions in early childhood consist mainly of childcare provision
and cash transfers. According to OECD (2023), in 2019 (the latest available year) public
expenditure per child, converted into USD PPP, in Italy for young children (0-5) was
slightly smaller than in the OECD average, 28.5 and 38.1% lower than in France and
Germany, respectively. Looking at younger children (0-2), public expenditure in Italy
was relatively even lower, meaning that the distribution of resources across younger (0-2)
and older (3-5) children was more uneven than what was observed in the OECD average
(or in France or Germany, for example).
imposed a minimum paternity leave entitlement of 10 days on member states. The voluntary paternity
leave has been reduced to one day only. From January 2022 the paternity leave became a permanent
policy which does not need yearly financing.
Shareable parental leave is mainly taken by mothers according to OECD (2023).
More than in other OECD countries, in Italy there is a marked difference in the
share of resources devoted to childcare and cash transfers across the different age groups.
Looking at expenditure for 0-2 year-olds, only 22.8% is spent on childcare (29.5% in the
OECD average, 55.4 in Germany and 66.8 in France), while the rest in cash transfers.
For children of 3-5, the figure is completely reversed: the share of resources spent on
childcare provision is 75.4%, larger than what is observed in the OECD average or in the
main European economies. A comparison of the enrolment rates in childcare for the two
age groups across countries reflects these disparities in public interventions across ages.50
Figure 31 shows that the enrolment rate in childcare services for children 0-2 in Italy
is lower (at 26.4%) than the OECD and the EU averages (36.0 and 32.6, respectively),
ranking Italy as one of the countries recording the lowest rates among the selected ones.
Enrolment rates among 3-5-year-old are much less heterogeneous across countries. In
Italy, it stood at 94.6% in 2020, above the OECD and the EU averages (87.1% and 89.5,
respectively), such that Italy is among those countries recording the highest rates.
Figure 31: Enrolment in childcare services in selected OECD countries, 0-2 year-olds
Be ce
D ium
Sw rk
te ort
ng al
Au om
O and
Notes: Data for European countries are OECD estimates for 2020 based on information from EU-SILC. Data refer to
children using center-based services (e.g. nurseries or day care centers and pre-schools, both public and private), organized
family day care, and care services provided by (paid) professional childminders, regardless of whether or not the service is
registered or ISCED-recognised. Source: OECD (2023).
According to Istat (2022), there is large geographical heterogeneity in the supply