1 The French labour market has changed radically since the 1970s, as widely documented in the literature. Unemployment has remained stubbornly high, at between 7% and 12%, and the gender gap in unemployment rates has gradually disappeared (Bodier et al., 2019). The female labour force participation (LFP) rate has increased considerably since 1975, rising from 53% to more than 68% for the 15–64 age group (Maruani, 2004; Collet and Rioux, 2017). It has moved much closer to that of men, which has fallen slightly over the period. The gender wage gap has narrowed, although more slowly since the late 1990s (Meurs and Pora, 2019), and the proportion of dual-earner couples has risen substantially (Stancanelli, 2006). In parallel, part-time workers, women especially (27% of working women vs. 8% of men), account for an increasing share of wage earners: 18% in 2019 compared with just 8% in 1982 (Bodier et al., 2019). The share of short-term employment contracts—be it fixed-term contracts [1] for less than 3 months or temporary contracts [2]—has also increased, rising from 1% of total employment in 1982 to 4.5% in 2017, with a concomitant rise in the frequency of periods of non-employment (Barlet et al., 2014; Jauneau and Vidalenc, 2019).
2 These labour market transformations reflect important societal shifts. In a context of increasing relationship instability, it is important for women to have financial independence (Olivetti and Rotz, 2016) and to acquire own-pension entitlements so they do not depend on a survivor’s pension if their partner dies, or on welfare benefits if they separate.
3 But the men and women of succeeding generations are affected differently by these changes. Women’s LFP has been increasing since the 1970s, but not uniformly across cohorts. LFP rates by age have risen progressively from the 1935 birth cohort up to that of 1975 (Afsa Essafi and Buffeteau, 2005, 2006). The proportion of part-time workers (at a given age) has increased likewise, among women especially (Briard and Calavrezo, 2016). A cohort approach, alongside the successive annual snapshots published by INSEE, can thus provide further insights into these labour market transformations.
4 With this objective in mind, Afsa Essafi and Buffeteau (2006) analysed data from the INSEE Labour Force surveys (LFS) over the 30 years between 1982 and 2002 to track the increase in female LFP across successive cohorts. Using econometric modelling, the authors showed that if the trends observed at the time had continued, the mean LFP rate over women’s life cycle would have almost caught up with that of men for the cohort born in 1970, with a narrowing of the gender gap to just 5 points. Employment and full-time- equivalent (FTE) employment rates would have remained much lower, however (with a gender gap of 10 and 15 points, respectively). Using more recent data from the same source (the 1985 and 2010 editions of the LFS, with a 5-year time step [3]), Périvier and Verdugo (2018) obtained a relatively stable gender differential in employment rates (around 20 percentage points). The difference between the two sets of findings is explained by the trend change observed during the additional years studied by Périvier and Verdugo; over the period 2003–2010, the increase in female LFP across cohorts slowed considerably and even stalled.
5 Following on from these studies, and with greater historical hindsight, the present paper uses more detailed data to describe changes across cohorts in male and female LFP, employment and FTE employment rates. This analysis covers a period of 44 years, beginning in 1975 and ending in 2018. Changes over the period before 2002 are already well documented by Afsa Essafi and Buffeteau (2006), so this article will focus on more recent trends, without seeking to extrapolate them. It aims to verify whether the gender gap in employment and LFP rates is indeed narrowing more slowly, as detected by Périvier and Verdugo. The analysis will also include an approach based on level of qualification and number of dependent children, two key determinants in the construction of gender inequality on the labour market.
I – Method
1 – Definition of indicators and scope
6 Three indicators are used in this study: LFP rate, employment rate, and FTE employment rate. The LFP rate at age a is defined as the ratio between the number of active individuals at age a (working or unemployed) and the total population of that age. The employment rate at age a represents the share of individuals in employment at age a among the total population of that age. Last, the FTE rate is obtained by weighting each job by its percentage of full-time working hours. To calculate the FTE rate, a person in a full-time job is counted in the numerator with a weight of 1, while a person working 80% of full-time hours, for example, is counted in the numerator with a weight of 0.8. Everyone in the denominator has a weight of 1. While LFS respondents are asked directly to state their average weekly working hours, these data are then recoded by INSEE into time ranges, used here to calculate the percentage of full-time working hours. Five categories are defined: 0% for unemployed people, 25% for people reporting habitual weekly working hours of 1–14 hours, 50% for 15–19 hours, 75% for 30–34 hours, and 100% for 35 hours or more (full-time working). [4] The criteria for identifying working and unemployed individuals are detailed in Box 1, describing the LFS methodology.
7 The LFP, employment, and FTE employment rates are calculated using data on individuals aged 30–59 drawn from successive LFS over the period 1975–2018. A minimum age of 30 was chosen because a large share of the population below this age is still in education or starting out on the labour market. People still in initial education at the time of the survey are also excluded. Their numbers are very small beyond age 30, although the proportion increases with successive generations, rising from around 0.5% for the 1945 birth cohort (the first observed at age 30) to slightly above 1% for the 1985 cohort. The maximum age was set at 59 years as a large share of individuals beyond this age are retired (the legal retirement age was, for many years, age 60). The LFS was expanded in 2013 to include the French overseas departments, but these populations cannot be included as the relevant data do not cover a long enough period. The geographical scope is therefore limited to metropolitan France (mainland France and Corsica). These methodological choices explain why the calculated data series differ from those published by INSEE.
8 The cohorts followed in this study were born between 1925 and 1985. These cut-off points allow us to exclude certain cohorts which are not observed over a long enough period. The 1925 cohort, the oldest in the sample, is followed for 10 years (between ages 50 and 59) and the 1985 cohort, the youngest in the sample, over 4 years (between ages 30 and 33). In all, nearly 5.3 million individual observations are included. Some respondents are observed on several different dates, because starting with the 2003 edition of the LFS (Box 1), the sample of households is surveyed over six successive quarters.
2 – Subpopulations
9 To flesh out the analysis, two other variables were used: level of qualification (educational attainment) and number of children. Earlier studies have shown that female LFP rates increase more markedly across cohorts for the most highly qualified women (Afsa Essafi and Buffeteau, 2006) and for women with two or three dependent children (Minni and Moshion, 2010). The number of dependent children is defined not as the respondent’s total number of children, but as the number of children aged 18 or below who are inactive (those in employment are excluded) and living in the same household as the respondent (including children in alternating custody and stepchildren). The underlying assumption is that the presence of a dependent child is a stronger determinant of LFP than the number of children a person may have had over their lifetime, even if the birth of a child—including those no longer dependent—can also be an important factor (in cases of long-term absence from the labour force, for example). This choice is also justified in practical terms. Information on the respondents’ number of children, recorded in the LFS since 2003, has not been available for as long as the number of dependent children. For level of qualification, 4 quartiles were defined for each birth cohort: the 25% most qualified, the 25% with a higher-intermediate qualification, the 25% with a lower-intermediate qualification, and the 25% least qualified. To take account of the general increase in levels of qualification across cohorts when constructing these quartiles, the members of each cohort were ordered first by level of qualification, [5] then by their age at completing education, based on the underlying assumption that for a given qualification, a higher age at completion signifies a longer period of initial education. [6] For the 1925 birth cohort, having a high-school diploma (baccalauréat) is sufficient to belong to the most qualified quartile, while in the 1985 birth cohort, the 25% most qualified have at least 2 years of post-secondary education.
10 The results presented here focus mainly on four ages in the life course:
- Age 30 corresponds approximately to the start of a stable working career after a phase of labour market entry. It is also close to the mean age at childbirth. [7]
- Age 40 corresponds to the halfway point of working life, between labour market entry (at around age 22 for the most recent observed cohorts) and retirement (at around age 62 for the most recent observed cohorts).
- Age 50 corresponds to the age above which employment and LFP rates start to decline in France (Martin-Houssart and Roth, 2002) and at which children start to leave the parental home.
- Age 59 corresponds roughly to the end of working life, since almost 70% of people in the most recent observed cohort (persons born in 1950) have already claimed at least one retirement pension.
Box 1. The INSEE Labour Force survey (LFS)
The LFS series on which this study is based covers the period 1975–2018 in metropolitan France. Surveys before 1975, corresponding to the period 1968–1974, were not used because the definitions of employment and labour force participation were changed substantially for the 1975 edition; 2018 was the most recent available survey year when this study was conducted. Over the period covered, only minor changes were made to the various definitions. While the ILO definition of unemployment was revised at the 1982 Conference, the aim was to improve clarity rather than modify the concept, so the impact on French unemployment figures was marginal (Goux, 2003). For some information, however, historical depth is more limited, especially for level of qualification. Until 1981, fewer than half of all respondents provided this information. Likewise for the percentage of full-time working: from 1975 to 1981, respondents were asked to give their percentage of full-time working hours in the reference week, but from 1982 this changed to the habitual weekly percentage. These breaks in survey data series mean that breakdowns by level of qualification and analyses of FTE employment rates are limited to the period 1982–2018. The LFS was redesigned in 2003 and became quarterly to take account of new European regulations. In French, it was renamed enquête Emploi en continu. The principles for identifying unemployed individuals or those outside the labour force were modified slightly. Respondents’ own assessments of their situation are no longer taken into account. It is now the entire set of answers to factual questions that determine whether an individual is employed, unemployed, or economically inactive (Goux, 2003). In practice, the series breaks resulting from this change were very limited (Givord, 2003).
(a) Between 1977 and 1981, surveys were for some periods conducted every 6 months (Goux, 2003).
II – Recent stagnation of female labour force participation and employment at ages 30–55
11 There is little change in the main long-term trends relative to individuals aged 25–50 documented by Afsa Essafi and Buffeteau in 2006. Women’s LFP rates have increased substantially across cohorts, but owing to higher unemployment rates and the development of part-time working, this is not fully reflected in FTE employment rates. These changes are very pronounced at age 50 for the cohorts of women born between 1935 and 1965, and at age 40 for those born between 1945 and 1975 (Table 1; Figure 1).
12 The overall trends conceal major disparities across levels of qualification, however. For the least qualified women, increasing LFP rates correspond mainly to part-time working, since FTE employment rates have risen very little. At age 40, for example, between the 1945 and 1975 birth cohorts, the LFP rate increased by 10.4 percentage points, while the FTE employment rate grew by just 4.6 points (Table 1). For the most qualified, the increase in LFP rates corresponds almost entirely to the take-up of full-time jobs. Again at age 40 and for the same birth cohorts, the LFP rate increased by 12.8 percentage points and the FTE employment rate by 15.4 points (Table 1). The increase in the LFP rate of the most qualified women also began earlier and ended sooner (with practically no further progress for the cohorts born after 1955 who had already reached a rate of over 90% at ages 40 and 50) than was observed for the least qualified women.
Table 1. Female labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates across cohorts at different ages
Table 1. Female labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates across cohorts at different ages
Note: The indicators are presented for the cohorts born in 1925, 1935, 1945, 1955, 1965, and 1975. n/a = not available, when sample numbers represent less than 3% of the total number in the two cohorts.
Sample: Women born between 1925 and 1989 who have completed their initial education and live in metropolitan France.
13 Women’s LFP is also highly dependent on the number of children aged under 18 in the household. The increases in LFP and employment were small for women with no dependent children, moderate for those with just one dependent child, and larger for those with at least two dependent children, although part of the increase in this case corresponds to part-time employment (Table 1).
Figure 1. Labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates of men and women across cohorts at:
Figure 1. Labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates of men and women across cohorts at:
14 The 15 additional years of coverage in this study (2003–2018) provide two new insights. The first concerns changes in LFP rates. In 2006, Afsa Essafi and Buffeteau observed that progression of the female employment rate slowed down among the cohorts born after 1970. This slowdown now also concerns the LFP rate. To analyse this development, statistical significance tests were run on birth cohorts grouped over 5-year periods to obtain more robust sample sizes. For each 5-year age group, the statistical hypothesis of an increase in the various indicators with respect to the previous group was tested. The tests revealed a stagnation of employment rates at age 50 for the female cohorts born after 1960 (the 1960–1964 cohorts do not have a significantly higher employment rate than the 1955–1959 cohorts), even though the LFP rate continued to increase slightly (Table 2). At age 40, employment and LFP rates stalled for the cohorts of women born after 1975, and at age 30 for those born after 1980.
15 The second new insight concerns the age of 59, when many people are nearing retirement. The drop-off in LFP and employment rates is totally reversed at this age for the cohorts born after 1940 (Figure 1). These cohorts have been increasingly affected by the lengthening of the contribution period for a full pension entitlement (reforms of 1993 and 2003) and by the increase in the legal retirement age from 60 to 62 (reform of 2010 that concerns contributors born after 1951). Only the results at age 59 are presented here, but the same pattern is observed at all ages over 56. This trend reversal in employment and LFP at these ages is observed for both men and women alike.
III – A decrease in male labour force participation and employment at ages 30–55
16 Among men, apart from the changes at age 59 described above, the trends already identified by Afsa Essafi and Buffeteau in 2006 are unchanged. LFP rates have continued to decrease slightly across cohorts, and with the rise in unemployment, employment rates have fallen even faster. As part-time employment is much less frequent among men than women, FTE employment rates have remained close to employment rates (Figure 1). At age 40, for example, between the 1945 and 1975 cohorts, the LFP rate fell by 4.5 percentage points, the employment rate dropped by 8.3 points due to the rise in unemployment, and the FTE employment rate dropped by a similar amount (9.9 points), indicating that part-time employment remains quite rare for men (Table 3).
Table 2. Female labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates of 5-year cohorts at different ages and statistical tests of rate changes with respect to previous 5-year cohort
Table 2. Female labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates of 5-year cohorts at different ages and statistical tests of rate changes with respect to previous 5-year cohort
Interpretation: The LFP rate at age 30 of women born between 1945 and 1949 is 66.7%. It is significantly higher than that of the cohorts born between 1940 and 1944.Note: n/o (not observed) means the cohorts were not observed at these ages.
Statistical significance: *** significant at 1% level; ** at 5% level; * at 10% level. The statistical test is a unilateral Student’s test comparing the rate with that of the group of five preceding cohorts.
Sample: Women born between 1925 and 1985 who have completed their initial education and live in metropolitan France.
17 These trends conceal large disparities by level of qualification. For the least qualified men, LFP and employment rates have fallen sharply for the most recent cohorts (by 10.6 points at age 30 between the 1955 and 1985 cohorts, for example), but the decrease is much more limited for the most qualified men (Table 3). At age 40, men with no dependent children are also more affected by the drop in LFP and employment rates than men with children.
IV – Slower progress in closing the gender gap
18 The gender gaps in LFP and employment rates have decreased across all cohorts (Table 4; Figure 1). However, while these gaps narrowed very quickly for the cohorts born between 1925 and 1955, progress has slowed markedly for cohorts born after 1965. For the cohorts born before 1970, this convergence between men and women was due mainly to higher female rates. For younger cohorts, convergence is entirely attributable to declining rates among men.
Table 3. Male labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates across cohorts at different ages (percentage points)
Table 3. Male labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates across cohorts at different ages (percentage points)
Notes: In this table, the indicators are presented for the cohorts born in 1925, 1935, 1945, 1955, 1965 and 1975. n/a = not available, when sample numbers represent less than 3% of the total number in the two cohorts.
Sample: Men born between 1925 and 1985 who have completed their initial education and live in metropolitan France.
Table 4. Difference between male and female labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates across cohorts at different ages (percentage points)
Table 4. Difference between male and female labour force participation (LFP), employment, and full-time-equivalent (FTE) employment rates across cohorts at different ages (percentage points)
Interpretation: At age 30, the gender differential in LFP rates is 35.3 percentage points for the 1945 cohort.Note: n/a means the information is not available in the sources.
Sample: Persons born between 1925 and 1985 who have completed their initial education and live in metropolitan France.
Conclusion
19 The increase in female LFP rates across cohorts is a stylized fact well documented in the literature. This increase, boosted by changes in the behaviour of women with at least two dependent children, is driven mainly by growth in full-time employment for the most qualified women and in part-time employment for the least qualified. That said, the most recent LFS data reveal a flattening of the curve for the cohorts born after 1975. For men, on the other hand, the slow but steady downtrend across cohorts has continued due to lower LFP among the least qualified men and among childless men.
20 These analyses over a long period reveal deep-seated social and economic trends in France. They remain imprecise, however, and are sensitive to revisions of INSEE’s LFS which, while always introduced with the aim of better capturing the French labour market and adapting to its evolution, are liable to produce breaks in data series. The age limit for defining children is 18 years because surveys before 1990 did not record the presence of children in the household over this age and did not necessarily note the existence of respondents’ children living in a different household. An approach based on full-time-equivalent working hours is not possible before 1982, as a major series break occurred when the question used to record respondents’ working hours was reformulated.
21 While the trends brought to light in this paper occur over the long term, it would also be useful to study the short- and medium-term impacts of shocks, such as the 2008 economic crisis or the COVID-19 pandemic, on male and female LFP across different birth cohorts. Previous studies suggest these shocks may have asymmetrical effects on male and female employment, with men appearing to fare slightly better than women (Périvier, 2018).
Notes
-
[1]
Under French labour law, a fixed-term contract (contrat à durée déterminée [CDD]) is a private-law employment contract in which an employee is recruited by an employer for a predetermined period. The contract expiry date is defined at the time of signature. CDDs must not be used to fill jobs corresponding to a company’s normal routine activity, but only for specific, short-term assignments. Use of a CDD is justified by law on three main grounds: temporary employee replacement, a temporary increase in activity, or seasonal activity.
-
[2]
Under French labour law, a temporary contract is a private contract between a client company and a temporary work company in which the client company recruits an employee for a predetermined period. The employee and the client company are not directly bound by an employment contract. The employment contract is between the employee and the temporary work company. A temporary contract can be used by an employer primarily on the following grounds: exceptional employee replacement, a temporary increase in activity, or seasonal activity. The Labour Code also provides for other less frequent grounds.
-
[3]
In the study by Périvier and Verdugo, which covered all European Union countries, only 1 in 5 editions of the LFS are analysed, and only 1 in 5 cohorts.
-
[4]
These ranges apply to the most recent surveys. They are adjusted to take account of changes in working time legislation. In particular, the 35-hour threshold for full-time working is raised to 39 hours for surveys prior to 2002.
-
[5]
Levels of qualification are ranked from top to bottom as follows: qualification requiring more than 2 years in higher education, qualification requiring 2 years in higher education, high-school diploma (baccalauréat) or equivalent, lower secondary qualification, primary school certificate.
-
[6]
This assumption is questionable given that an additional year of education does not necessarily correspond to a year of new learning, but as individuals are ordered first by level of qualification, its importance is secondary.
-
[7]
Mean age at childbirth was 27.8 years for women and 30.5 years for men in 1925. It then fell gradually up to the mid-1940s cohorts before rising again to 26.8 for women and 30.1 years for men in the 1953 cohort, the last one for which practically all births have been observed (Robert-Bobée, 2015).
