A prelude to the dual provider family – The changing role of female labor force participation and occupational field on fertility outcomes during the baby boom in Sweden 1900–60

ABSTRACT

By investigating changes in the association between women’s socioeconomic status, labor market activity and fertility outcomes during the Swedish baby boom 1900–60 this study reaches three main conclusions. First, the results show that a convergence of fertility behavior occurred across female socioeconomic strata during the peak baby boom period in the 1940s and 1950s in terms of a strong two child norm. Second, the negative socio-economic gradient of fertility found in Sweden before the baby boom declined sharply among women who came of age during the 1940s and 1950s, as white-collar women increased their fertility more than all the other strata. Third, this was especially the case for women engaged in the so called ‘caring professions’ that exhibit the largest changes in behavior. The pattern found in contemporary Western contexts where women in healthcare and education have a substantially higher fertility was thus formed in Sweden already during the 1940s and 1950s. The empirical finding fit with the interpretation that middle-class women employed in the public sector experienced stronger reductions in constraints to family formation compared to women employed in the private sector. We propose that the pronatalist polices implemented in the 1930s and 1940s, especially the extensive improvements in employment protection implemented for women who got married or became pregnant in the late 1930s in Sweden, is one important factor to consider when we try to understand why especially women employed in the public sector in education and healthcare increased their fertility more than other groups.

KEYWORDS:

• Fertility
• baby boom
• female labor force participation
• caring professions
• Sweden

1. Introduction

The aim of this study is to investigate how female labor force participation and occupational field influenced the fertility behavior of Swedish women in the first half of the twentieth century. The issue of how increased human capital and labor market activity among women have influenced fertility outcomes has been at the center of a large body of both theoretical and empirical research. For several reasons, it can be argued that the mid-twentieth century baby boom stands out as a particularly interesting period when we try to understand how the association between economic roles of women and fertility has changed during the twentieth century. Specifically, the baby boom period between the late 1930s and early 1960s increasingly stands out as an interesting period of change, due to a growing body of research that shows that fertility differentials across educational strata actually decreased sharply during the baby boom in a number of Western countries (Reher & Requena, 2015; Van Bavel, 2014; Van Bavel et al., 2018) including Sweden (Sandström, 2014). However, at present we do not know much about the causes for the reduction of the negative socio-economic gradient of fertility for women during this period. Neither do we know if all women with more human capital investments behaved similarly, or if some white-collar women spearheaded the shift toward higher fertility. Also, previous studies have almost exclusively looked at the impact of education rather than actual labor market activity. This study addresses all of these issues by analyzing individual-level longitudinal data with detailed information on women’s labor market activity up until first childbirth for women born 1880–1939 and by disentangling fertility differentials in different occupational groups among white-collar women during the baby boom period.

From a theoretical point of view, social scientists have tended to argue that larger investments in education and increased labor force participation among women is an important shift explaining decreasing fertility in developed and developing countries since the nineteenth century. Education and employment is thought to increase the opportunity cost of childbearing for women and, in turn, cause lower nuptiality and fertility (Becker, 1991; Willis, 1973). Although the theoretical notion of a negative link between female economic activity and fertility has been highly influential in contemporary demographic research, it is in no way a new idea. The increase of female labor force participation in the first decades of the twentieth century in Sweden and opportunity costs of employed and educated women were stressed already by Swedish social scientists such as Alva Myrdal and Karl Edin in the 1930s as the most important explanation for the sub-replacement fertility experienced in Sweden at the end of the historical fertility decline in the mid 1930s (Van Bavel, 2010b).

The decline in fertility to levels below replacement had a profound impact on Swedish policy in the late 1930s and 1940s where political concerns regarding low fertility resulted in an expansion of the social security system and legal changes to facilitate the family formation of employed women. In 1934, Sweden reached the lowest period fertility recorded in the world up to that time, when the TFR dropped to 1.71 children per woman. In the same year, the so called ‘crisis in the population question’ was declared by the Social Democratic intellectuals Gunnar and Alva Myrdal in a highly influential book (Myrdal & Myrdal, 1934), that made the population question the center of political debate in Sweden in the 1930s and 1940s (Hirdman, 1989). The end result was the first breakthrough period of the typical comprehensive Scandinavian welfare state. The Scandinavian solution to the population crisis was not to make female economic activity more difficult as this would be counterproductive according to the opinion of the Population Commission that submitted its report in 1936.

Rather new policies were implemented to make marriage and childbearing more feasible for employed women through the implementation of a number of welfare state policies that aimed to reduce the alternative cost of childbearing for young couples engaged in wage labor (Frangeur, 1998; Hatje, 1974; Hirdman, 1989; Kälvemark, 1980). An example of measures directly aimed at facilitating the employment of mothers was the law adopted by the Swedish parliament in 1939 that prohibited employers from dismissing women if they married or had a child (Frangeur, 1998; Göransson, 2006; SFS 1939:171, 1939).

Today, the proposed negative link between female human capital investments and fertility is still identified as a primary explanation for the ‘lowest-low’ fertility found in many contemporary Western societies. Although a body of recent research is pointing to the conclusion that increasing levels of female human capital investment is only found to depress fertility to ‘lowest-low’ levels if they are combined with a lack of gender-egalitarian norms on the societal level as well as within the family. Low fertility has been viewed as the result of a lack of social insurance institutions that can counteract the difficulties of combining family responsibilities and female labor force participation (Anderson & Kohler, 2015; Billari & Kohler, 2004; ESHRE Capri Workshop Group, 2010; Esping-Andersen & Billari, 2015). A recent study on Sweden has also shown that the baby boom period in Sweden was characterized by a break with the late and non-universal marriage pattern typical of Scandinavia. During the baby boom in Sweden, marriage became nearly universal, and this was achieved through a sharp reduction of a negative socio-economic gradient of nuptiality among higher strata women (Sandström, 2017).

To better understand how the institutional, economic and cultural changes that occurred during the baby boom period might be related to larger net increases in the fertility of more educated women we will probe deeper into the socio-economic differentials in fertility. In this analysis we investigate differences in fertility not only between working-class women and white-collar women but in particular between different white-collar groups. Also, in this study we use the arguably more precise indicator of women’s actual labor market activity in specific occupations and sectors of the economy rather than education per se.

Differentiating between women in different sectors is important because studies of contemporary data find that the field of occupation often has a larger net effect than the level of education. For example, in many contemporary Western countries women in the so called ‘caring professions’ in healthcare and teaching have a lower chance of remaining childless and achieves higher levels of total cohort fertility than other highly educated women (Van Bavel, 2010a; Lappegård, 2002; Lappegård & Rønsen, 2005; Hoem, Neyer, & Andersson, 2006a, 2006b; Martín-García & Baizán, 2006; Neyer & Hoem, 2008; Michelmore & Musick, 2014). However, we do not know whether such differences across sectors of employment were present before the 1960s or to what extent they changed during the baby boom.

What we do know is that the right for married women to work was a frequently debated issue during the first decades of the twentieth century in Sweden, especially among women in the largest female academic professions such as teaching and nursing (Andersson, 2002; Florin, 1987; Frangeur, 1998). During the 1920s, female labor force participation increased rapidly, among unmarried young women in urban areas and the educational attainment also increased rapidly, especially among the women who came of age in the 1940s and 1950s. The baby boom period is often seen as a period of stalled change to the provider model, as overall rates of female employment tended to stagnate at approximately 30% of all women during the 1940s and 1950s. It was only in the early 1960s that participation rates once again shifted into rapid increase to levels substantially above the ones reached already before the Second World War (Åmark, 2006; Stanfors, 2003).

However, it is clear that, below the surface, a shift toward a dual provider model continued during the baby boom period. This is seen in terms of the substantial increases in the share of gainful employed women that were married. In the 1930s, only 10% of all gainfully employed women were married, but this share increased sharply throughout the baby boom in Sweden, reaching 44% in 1960 just before the second peak of the Swedish baby boom (Qvist, 1974). In this sense the composition of the female labor force changed substantially in terms of half of the women being married just before the second peak of the baby boom in Sweden and in 1965 47% of the women having a child aged 0–16 years were gainfully employed according to the labor market survey conducted by Statistics Sweden (Statistics Sweden, 1972).

Largely due to a lack of individual-level demographic data with socio-economic information for the period covering the late stages of the fertility transition and the subsequent baby boom (for most Western countries late nineteenth century – 1960s), the constraints on family formation and how they changed for the growing group of women that entered the labor market during the first half of the twentieth century are mostly known on the basis of qualitative accounts, and are primarily focused on specific groups of highly educated women (Andersson, 2002; Elgqvist-Saltzman, 1993; Florin, 1987). So, although recent studies indicate that women who achieved higher levels of education did not experience the same strong penalty on family formation during the baby boom as previous cohorts had (Reher & Requena, 2015; Sandström, 2014; Van Bavel, 2014; Van Bavel et al., 2018), we do not know whether this increase in fertility was a general trend that affected all educated women equally or whether some groups spearheaded the change.

To clarify the issues raised above for the Swedish case, this study addresses three questions. 1. How was the female labor force participation associated with fertility during the first half of the twentieth century in Sweden, and did the association change over time? 2. Were there substantial differentials in fertility among gainfully employed women depending on their social stratum during the first half of the twentieth century in Sweden, and how did this change over time? 3. How did fertility differ between women in the different occupational groups belonging to the middle and upper social strata such as nurses, teachers and office clerks that where the most numerous white-collar groups among women after the 1930s when women entered higher education in greater numbers?

The reason that we choose to focus on women’s socioeconomic position rather than that of both genders is the importance given to the changing economic roles of women in research on fertility differentials across space and time. Although the roles of both men and women have changed due to the shift from a male breadwinner to a dual provider model in countries like Sweden, scholars tend to argue that the changes in women’s roles have been much more profound and important for the understanding of changing family behavior in modern societies (Stanfors, 2003).

The focus on specific occupational groups within the middle class and elite is motivated by the fact that previous research has shown that the educational gradient was reduced for women with secondary and post-secondary education in Sweden during the baby boom (Reher & Requena, 2015: Sandström, 2014; Van Bavel, 2014; Van Bavel et al., 2018), but little is known regarding whether there were large variations between middle-class women in different sectors. Research on women born from the 1950s onwards shows that individuals in the so-called caring professions, such as nursing and teaching, have a substantially higher fertility than other women with post-secondary education. However, we do not know when these patterns formed or to what extent they were present already during the first half of the twentieth century.

It is important to note that this study focuses on one aspect of the baby boom that could be regarded as a contributing factor. Although the increased fertility of a larger share of educated women contributed to the fertility recovery, we are not saying that this is the only, or even most important driver of the fertility boom in the West during the period 1940–1960. Previous research has highlighted a number of contributing factors. An extensive review of this research is found in Van Bavel and Reher (2013) and additionally in Macunovich (2010). Rather our ambition here is to dig deeper into the issue of how socioeconomic differentials in fertility among women developed during the baby boom period and especially into the changes in fertility among white-collar women that has been put forward in previous research. These changes are of importance as they function as the historical underpinnings of the contemporary patterns of fertility in countries like Sweden. In particular the increasingly positive socioeconomic gradient of fertility that has developed in the Nordic countries since the 1970s were the employment of women is now a prerequisite rather than an impediment for motherhood (Hoem, 2000; Stanfors, 2006; Statistics Sweden, 2002), and that highly educated women have ended up having much lower levels of childlessness than the lowest educated women (Andersson et al., 2009; Jalovaara et al., 2018).

1.1. Data and method

This study uses the recently developed POPLINK database housed at the Demographic Data Base (DDB), Umeå University, Sweden. The source is based on the population registers kept by the Swedish state church that had the responsibility for collecting population statistics at the time. The POPLINK database consists of longitudinal individual-level micro-data, and has primarily been developed for medical research to be able to link the historical population registers to the modern registers at Statistics Sweden and the National Board of Health and Welfare in order to e.g. study the genetic components of morbidity and mortality. However, the source also gives unique opportunities for research on long-term demographic processes stretching from pre-transitional times up until the present. A detailed discussion on the design and methodology used to build the POPLINK-database is found in Westberg, Engberg, and Edvinsson (2016). The database contains rich longitudinal information on all vital events including: marriages, childbirths and divorces as well as moves within the region and detailed information on the labor market experience for more than 100,000 men and women living in the county of Västerbotten in Northern Sweden during the period 1900–60. The source have been used in a number of previous studies of fertility and nuptiality patterns in Sweden during this period (Junkka, 2018; Reher, Sandström, Sanz-Gimeno, & van Poppel, 2017; Sandström, 2017). In a previous study on male and female nuptiality patterns during this period, Sandström (2017) have shown that although Västerbotten experienced somewhat higher levels of fertility than the Swedish average up until the early 1940s, the trends in fertility and nuptiality are identical to the national level for the entire period under analysis. From the 1940s also differences in fertility levels are negligible. Sandström also shows that the region is representative for the Swedish case with regards to socio-economic development and change in the occupational structure during this period. For the sake of brevity, we do not repeat this discussion here and refer the reader to Sandström (2017) for an in-depth discussion of these matters.

The data source used thus covers both the fertility bust of the 1930s and the baby boom following in Sweden during the 1940s and 1950s. The parishes included in the database are distributed throughout the coastal region of Västerbotten, and include more than two-third of the population in the county during the nineteenth and twentieth centuries. The economic development presents the typical traits of most Swedish regions; being mainly rural up until the early twentieth century, and thereafter characterized by rapid industrialization and a growing public sector (Sandström, 2017; Westberg et al., 2016).

To analyze the transition to first, second and third birth for women in different socio-economic and occupational groups, we use event history analysis in terms of non-parametric Kaplan Meier estimates of the proportion having experienced the event in question at age t. For the analysis 54,871 women born 1880–1939 are followed from the time they become at risk of experiencing a childbirth, which we set to the year the woman turned 15 years of age. The simple reason for this choice is that we observe no childbirths among women younger than that in this dataset. We do not include women born after 1939 as they did not contribute in a substantial way to the fertility recovery that took place from the mid 1930s up until the mid 1960s (Sandström, 2014). We can follow women alive and present in the country in 1960 also after out-migration from the POPLINK area, as we have linked the data to the population registers at Statistics Sweden (SCB). If the woman was not present in the registers at SCB and they out-migrated from the area covered by POPLINK, we set the censoring date to the time they out-migrated from the parishes under observation in Västerbotten. To get complete records of the fertility of women born in the 1930s that had a substantial part of their reproductive period after 1960 we include the births observed for women after the end of the POPLINK database in the 1960s by using the SCB register data up until they are censored in the 1970s and 1980s due to menopause, death or outmigration from the country.

Women moving in to the region with no accompanying children who have no records of prior childbirths in the registers are assumed not to have experienced a first birth and enter the study at the age they come under observation. In-migrating women at risk of childbirth that had already experienced first childbirth would normally be accompanied by their minor children. Both low divorce rates and historically very low levels of non-marital fertility during the period will have contributed to few women moving without their biological children (Sandström, 2017, p. 1627). Additionally, any bias due to unobserved births prior to start of observation is unlikely as results are practically unaffected if we choose to drop women immigrating with no record of previous childbirths. The women are followed from the time they come under observation if they have no record of a prior childbirth or onset of risk (age 15) until a birth is observed or the woman is censored due to either out-migration, death or reaching age 50 (Demographic Database, 2014).

To analyze differences across birth cohorts, the women are divided into three different cohorts according to the year they reach age 20. For women born 1880–1939, this gives us three groups 1900–19, 1920–39 and 1940–59. These cohorts are used throughout the analysis to contrast women having most of their reproductive years prior to the baby boom (1900–19) against woman who had their peak reproductive years around the shift from baby bust to baby boom in the 1930s (1920–39) and additionally women who entered their reproductive years during the peak baby boom period in the 1940s and 1950s (cohort 1940–59).

The main independent variable in the analysis is the socio-economic position of the woman prior the birth of her first child, and the outcome is the incidence and timing of childbirths. To test the statistical significance of differences in transition rates between women in different social strata we apply Log – rank tests of the equality of the survivor functions (Hosmer & Lemeshow, 2008). The socio-economic position is determined by using the occupational information in the population registers coded according to the HISCO-classification system for historical occupations. For the analysis of different social strata rather than specific occupational groups, the HISCO-codes were coded into a social stratification indicator (SES) according to the Social Power coding scheme (SOCPO) developed by Van de Putte and Miles (2005). If more than one occupation was registered prior to the birth of the first child, the variable was set to the highest achieved SES of the individual prior to experiencing first birth or being censored. When specific occupational groups are analyzed, we have grouped the occupations according to their HISCO minor group’s e.g. Teachers having all HISCOs in the group 13.XXX and Nurses having all HISCOs in the group 07.XXX (Van Leeuwen, Maas, & Miles, 2004). The highest SOCPO-category ‘elite’, that mainly consists of higher professionals with university degrees, has been merged into the category ‘middle class’, since less than 1% of all women belonged to the ‘elite’ stratum. This is because teachers and nurses are all counted as lower professionals/middle class in the SOCPO-scheme, rather than as elite/higher professionals. To further reduce complexity, the categories of semi-skilled and skilled workers have been merged into one, as they were found to exhibit almost identical fertility patterns. Almost no women show downward social mobility prior to first birth so in practice we almost exclusively use the occupation recorded in the parish books at the time of first birth or at the time of marriage. In a few cases we have chosen to assign women to the middle-class stratum when we have information on them being in training and recorded as e.g. ‛teacher student’ rather than categorizing them as non-employed.

The church officials had the responsibility for collecting all information on population statistics in Sweden during the period in question and had strict instructions to record occupational status in detail on all individuals recorded in the parish books older than age 15. The information was required to be updated every time a status change occurred such as a move to a new address, changes in civil status, births etc. (Forkman, 1948). For a detailed discussion of how the information was recorded and the function of the church in collecting population statistics in Sweden see Wannerdt (1982). Table 1 lists the absolute and relative frequencies of the two variables used to categorize women according to their socioeconomic stratum in our analysis.

Table 1. Absolute and relative frequencies according to socio-economic and occupational strata of the woman at onset of risk for the first child

	1900–19 (N = 17,509)	1920–39 (N = 24,126)	1940–59 (N = 13,236)	Total (N = 54,871)
Socio-economic strata of woman
No occupation	6905 (39.4%)	10,545 (43.7%)	4653 (35.2%)	22,103 (40.3%)
Unskilled workers	5347 (30.5%)	4614 (19.1%)	3897 (29.4%)	13,858 (25.3%)
Skilled workers	2444 (14.0%)	5940 (24.6%)	3009 (22.7%)	11,393 (20.8%)
Farmers	1263 (7.2%)	594 (2.5%)	75 (0.6%)	1932 (3.5%)
Middle Class and Elite	1550 (8.9%)	2433 (10.1%)	1602 (12.1%)	5585 (10.2%)
Occupational groups middle-class
(07) Nurses & Midwifes	371 (2.1%)	796 (3.3%)	238 (1.8%)	1405 (2.6%)
(13) Teachers	675 (3.9%)	550 (2.3%)	146 (1.1%)	1371 (2.5%)
(−09) Professionals	37 (0.2%)	123 (0.5%)	104 (0.8%)	264 (0.5%)
(30) Office workers and secretaries	230 (1.3%)	683 (2.8%)	1034 (7.8%)	1947 (3.5%)
Other white-collar	237 (1.4%)	281 (1.2%)	80 (0.6%)	598 (1.1%)

Source: POPLINK-database, Demographic Data Base, Umeå University

1.2. Background

Figure 1 shows the total fertility rate (TFR) for the parishes included in Västerbotten (1900–63) compared to the corresponding numbers on the national level (1900–70). During the period leading up to the Second World War, fertility in Västerbotten was substantially above the average levels of the country as a whole. This pattern is an example of the regional differences in the fertility transition in Sweden, where the Northern parts of the country entered the fertility decline with some delay compared to the rest of the country and started from somewhat higher pre-transitional fertility levels (Hofsten & Lundström, 1976). Sweden and the other Scandinavian countries belong to a cluster of countries including Australia, New Zeeland and Germany where the fertility recovery started already before the Second World War. In the county of Västerbotten this recovery occurs slightly later and is similar to the majority of the countries that experienced a baby boom where fertility starts to increase substantially in the early 1940s where typical examples are Iceland, Switzerland, Ireland and England and Wales (Van Bavel & Reher, 2013). Although the break in the declining trend occurred in 1940 in Västerbotten rather than in the mid 1930s, the subsequent recovery in fertility during the 1940s and 1950s was substantial. The TFR increases from 2.2 children per woman in 1940 to 2.9 in 1946, and the TFR continued to stay above the 1940 level into the latter part of the 1950s, when the TFR for Västerbotten converged with the rate for the country as a whole. Since the early 1960s, the TFR in Västerbotten has matched the national level almost perfectly (Hofsten & Lundström, 1976; Statistics Sweden, Population Register, 2015). Additionally, nuptiality trends in Västerbotten during the baby boom period shows an almost perfect match to the trend at the national level further illustrating the similarities in trends with the national level (Sandström, 2017). To summarize, the close match to the trend in the rest of the country and the marked increase in period fertility during the baby boom show that the data from Västerbotten can be used to analyze the fertility swings during this period.

Figure 1. Total fertility rate (TFR) 1900–1963 in the POPLINK-sample from Västerbotten and in Sweden 1900–70.

Source: POPLINK-database, Demographic Data Base, Umeå University and Statistics Sweden (2002, p. 95)

In Figure 2, the relative distribution of women according to their social strata (SOCPO-stratification scheme) is shown for the three cohorts under study.

Figure 2. Relative distribution of socio-economic strata by risk cohorts (year reaching age 20) 1900–59 for Swedish women in Västerbotten county.

Source: POPLINK-database, Demographic Data Base, Umeå University

Regarding the issue of how the socio-economic composition of the female labor force developed during the first half of the century, the current literature has some limitations, as it primarily focuses on changes in the participation level rather than changes in the occupational structure among women (Åmark, 2006; Silenstam, 1970; Stanfors, 2003). However, the marked increase in participation rates seen in Figure 2 between the 1920–39 cohort and those that came of age during the 1940s and 1950s is in line with previous research for the national level in Sweden. For unmarried women aged 20–39, the employment rate on the national level increased from approximately 55% in the 1920s to around 75% of the women in this age-group in 1950 (Silenstam, 1970), which corresponds well to the figures found in Figure 2.

Teaching and nursing is by far the most common occupation among women in the middle class, as more than 50% of the middle-class category consists of these types of occupations, although their relative share decreases in the last risk cohort due to a growing number of women doing various white-collar clerical work. This illustrates that the labor market for qualified women started to become more diversified during the 1940s and 1950s in terms of expanding out of the typical female white-collar occupations in education and healthcare in the public sector. Across the cohorts, female farmers nearly vanish, from a small group (7%) to only 0.5% amount women in the 1940–59 cohort.

After covering the general development in period fertility, nuptiality and changes in the socio-economic composition of women, we now turn to the main focus of the analysis in terms of the fertility differentials according to socio-economic strata and occupational field.

2. Results

2.1. Transition rates to first, second and third child in different female socio-economic strata

Figure 3 gives the Kaplan-Meier estimates of the proportion not having experienced first, second and third birth as a function of the woman’s age by socio-economic strata and risk cohort. Looking at the general changes over time, what stands out is a strong convergence of fertility behavior among the women who entered their childbearing years during the baby boom. Over time, there is a sharp reduction in the difference between economically active women and those outside the labor market in terms of female employment being less and less negative for fertility outcomes. There is a sharp reduction in the number of childless women, as well as a strong tendency for an increasing proportion of women to stop having more children after their second child. This is in line with previous research that has stressed the increased significance of a two-child norm both in Sweden (Sandström, 2014; Statistics Sweden, 1992) and in general in the West during the baby boom (Bean, 1983; Notestein, 1950; Van Bavel & Reher, 2013).

Figure 3. Kaplan-Meier estimates of the proportion not having experienced first, second and third birth as function of the woman’s age by socio-economic strata and at-risk cohort (year reaching age 20) 1900–59.

Source: POPLINK-database, Demographic Data Base, Umeå University

Regarding the socio-economic patterns, the women exhibiting the biggest changes in behavior are women in the middle-class/elite strata where employment in general required secondary or tertiary educational qualifications. Women in skilled worker positions who also required relatively high qualifications in terms of having secondary education exhibit similar substantial changes in behavior across the cohorts. Unskilled women and especially women registered as farmers, on the other hand, appear to have experienced fewer constraints and/or a stronger preference to start a family already in the 1900–19 risk cohort that came of age during the fertility decline. Being economically active was less negative for the fertility outcomes of farmers and unskilled women than is the case for women in the skilled and middle-class strata who clearly experienced strong constraints and/or preferences working against having children prior to the baby boom. In the first two risk cohorts including women that had a substantial part of their reproductive history prior to the baby boom, this reduced fertility is apparent even among unskilled women who exhibit higher proportions of childlessness at age 40 compared to women outside the labor market. This highlights the fact that, regardless of the educational requirements, female gainful employment was clearly at odds with family formation for women prior to the baby boom, although the negative link varied along a strong negative socio-economic gradient.

Looking more closely at the women in the first risk cohort, who had their peak reproductive years during the rapid fertility decline in the first decades of the century, the differences between economically active women and those with no record of having an occupation were large. The general pattern found in Figure 3 is a strong negative SES-gradient among the women who turned 20 during the period 1900–19. Among middle-class and elite women, as many as 50% of them never experienced a first birth, and they were closely followed by women in the skilled worker category of whom about 40% remained childless at age 40. Comparing them to the group that has no record of being employed, only 15% of these women outside the labor market remained childless.

Transition rates to second and higher parities after experiencing a first birth continued to be high during the late stage of the fertility decline, reflecting the typical Northern European fertility regime as described by John Hajnal (Hajnal, 1965, 1982). In Northern Europe, marriage was on average much later than in southern Europe, and a substantial proportion of women remained unmarried and ultimately childless. However, cohort fertility rates still reached levels substantially above replacement, as married couples on average had large families. Still, there are some SES-differences also in the transition to parities above one among the women belonging to the first cohort (1900–19). Although the socio-economic differences are smaller for parity 2 and parity 3, stopping at only one or two children was more common for women in middle-class and skilled worker occupations. This shows that, even when they had a family, they tended to be more prone to stop childbearing at low parities, and in this sense, they acted as early adopters of the shift to an ‘ideal’ family size of two children typical of most post-transitional societies in Europe.

For women in the second risk cohort who entered their twenties during the late stage of the fertility decline and early baby boom years (1920–39), socio-economic patterns had only changed marginally as opposed to the prior cohort, and the general trend is one of continuity rather than a radical shift. The negative SES-gradient had to some extent weakened, especially for skilled women, but it remained significant for the more qualified women in the middle-class. Looking at the transition rates to parity 2 and parity 3, the most substantial change in this cohort of women is that the shift toward a two-child norm is well underway. This is reflected by a general drop across all strata in the transition rate to parity 3.

Changing the focus to the last risk cohort (1940–59) these women experienced a major shift in the way that female economic activity was associated with fertility. Socio-economic differences in the transition into motherhood are almost eliminated apart from some postponement of first birth among women in the middle-class/elite who on average had the longest education. Still, even though they had their first child slightly later than women in the other strata, they more or less completely close the gap by age 40, and reach the same low levels of childlessness as women in other socio-economic strata.

Much of the same changes found for parity 1 are also apparent in the transition to parity 2, where socio-economic differentials almost disappear. The proportion progressing to a second child remains high, indicating the growing significance of the two-child norm in the peak baby boom cohort. The fact that the propensity to have a second child remained virtually unchanged across all the risk cohorts among skilled and middle-class women, while stopping increasing in the other social strata, shows how the upper strata led the shift to an ‘ideal’ family size of two children. This tendency for a convergence toward a two-child norm is also a major factor contributing to reduced differences in complete family size between women with different levels of levels of human capital and labor market experience during the baby boom period. The growing proportion of couples having exactly two children is further illustrated by the rather substantial fall in the proportion progressing to parity 3 among women in the 1940–59 cohort where only 35–60%, depending on which group, continued with a third child compared to 70–90% in the first risk cohort. Again, we find that women in the middle-class act as the primary agents of change, in terms of being by far the most prone to stop childbearing at exactly two children. We know that the difference in family sizes decreased across the baby boom cohorts and reached a minimum with women born in 1945 that entered their twenties in the mid 1960s when the baby boom shifted to the baby bust of the 1970s (Statistics Sweden, 1992). Here we only include women born up until 1939, but the results show that middle-class women acted as leaders in this development toward a convergence in family sizes and more and more families choosing to have exactly two children.

2.2. Transition rates to first, second and third birth in different female middle-class occupations for women born 1880–1939

The results for the broader socio-economic groups show how middle-class and elite women experienced the most significant changes in behavior during the shift from the baby bust of the 1930s to the baby boom of the 1940s and 1950s. However, it is unclear whether these changes occurred evenly for all economically active white-collar women or whether the so called ‘caring professions’ in health care and teaching spearheaded the change. We know that these groups had higher fertility than other middle-class women in the cohorts born from the mid 1950s (Hoem et al., 2006a, 2006b; Neyer & Hoem, 2008), but fertility differentials according to occupational field for women that produced the baby boom are so far unknown. To investigate these issues, we will look at the fertility pattern of women in different occupational fields in the middle class and compare them to women who remained outside the labor market during the period 1900–59. In the analysis, we choose to differentiate between three different HISCO-minor groups classified as middle class/elite in the SOPCO-classification system: (07) Nurses and Midwifes, (13) Teachers and (30) Office Clerks and Secretaries. These three HISCO-categories together make up approximately 85% of all women classified as middle class/elite. The remaining 15% is grouped into a separate category labeled ‘other middle class’. Although this group is heterogeneous, it contains a fair share of higher professionals especially in the two later cohorts categorized as elite in the SOCPO classification. Furthermore, we use women having no record of being economically active prior to first birth as a comparison category that is contrasted against the fertility patterns found among middle-class women.

In Figure 4, the Kaplan-Meier estimate for transition to parity 1 is shown for different middle-class occupations according to the minor group HISCO-classification of the occupation. In all occupational groups within the middle class, the proportion of childless women decreases throughout the period, although there are differences in the timing and extent levels of childlessness decrease across different occupations within the middle class. Looking at the separate occupations, the same type of convergence of childbearing behavior that is evident for different social strata is also occurring within the middle-class stratum. Also, in the case of the middle class, we start out with substantial differences between different occupations, but end up with more similar behavior in the last risk cohort that turned 20 during the peak baby boom years. As seen in the legend of Figure 4 differences in transition rates are highly significant (Pr. Chi2 < .000) when all groups are included. If non-employed women are excluded and we only compare the differences between middle-class women, these differences are significant for the two first cohorts. In the 1940–59 cohort when the fertility pattern of middle-class women exhibits the smallest relative differences the significance level is Pr. Chi2 = .065. It is clear that postponement as well as ultimate childlessness was very high for women employed in middle-class occupations during the first decades of the century when fertility was declining rapidly on the aggregate level. As we saw in Figure 3, about 50% of all middle-class women in the earliest cohort remained childless. However, there are marked differences within the middle class, where two clusters are distinguishable in both the first and second risk cohort.

Figure 4. Kaplan-Meier estimates of the proportion not having experienced first birth as a function of the woman’s age in different female middle-class occupations compared to non-employed women by risk cohorts (year reaching age 20) 1900–59.

Source: POPLINK-database, Demographic Data Base, Umeå University

First, nurses and women employed in clerical occupations stand out as the ones that experienced the greatest constraints and/or preferences that discouraged family formation during the first decades of the twentieth century, where approximately 60% of these women remained childless.

Second, teachers and women in other middle-class occupations exhibit substantially higher transition rates to first birth compared to nurses and clerical workers although their levels of childlessness were more than twice as high as women outside the labor market.

Across the cohorts, there is a very strong decrease in the levels of childlessness for all middle-class women, but the most striking change is how nurses and midwifes switch from having the highest levels of childlessness to having levels on par with non-employed women. This shift must be viewed as extraordinary. Over a period of 20 years between the second and third risk cohort shown in Figure 4, the nurses and midwifes moved from a situation where 47% of them remained childless at age 50 to a situation where only 12% never had children. These results show that especially women in healthcare experienced profound changes in constraints on family formation, and/or changes in preferences where starting a family became a prioritized goal in the 1940s and 1950s.

Figure 5 shows the proportion progressing to have a second child by risk cohort and age. The tendency for middle-class women employed in teaching and healthcare to stand out by exhibiting the highest transition rates in the baby boom cohorts (1940–59) is further reinforced in the transition to parity 2. In the second risk cohort, which contributed to the first peak of the baby boom in the 1940s the most, teachers and nurses have markedly higher transition rates than other middle-class women. A log rank test of the difference between teacher and nurses compared to all other middle-class women reaching age 20 in 1920–39 is highly significant (Pr. Chi2 = .005). In the last risk cohort, nurses and teachers have the lowest levels of childlessness in the middle-class which is in line with the overall pattern of higher fertility for the women in education and healthcare. However, due to the much smaller socioeconomic differences in fertility during the period 1940–59 the difference between nurses and teachers and other middle-class women is not statistically significant in this last cohort.

Figure 5. Kaplan-Meier estimates of the proportion not having experienced second birth as a function of the woman’s age in different female middle-class occupations compared to non-employed women by risk cohorts (year reaching age 20) 1900–59.

Source: POPLINK-database, Demographic Data Base, Umeå University

One of the most notable results for parity 2 is the low transition rates of the office clerks and secretaries across the risk cohorts. Women in the clerical occupations were either constrained to, or choose to, have small families, and they stopped childbearing already after the first child to a higher degree than other middle-class women. For the first two risk cohorts the lower fertility in this group is highly significant compared to nurses and teachers.

Last, the transition to parity 3 for middle-class women is shown Figure 6. Here, women in the group ‘other middle class’ had too small numbers to be meaningfully analyzed as there were fewer than 50 women in each of the risk cohorts. Consequently, we removed them from the analysis of parity 3.

Figure 6. Kaplan-Meier estimates of the proportion not having experienced a third birth as a function of the woman’s age in different female middle-class occupations compared to non-employed women by risk cohorts (year reaching age 20) 1900–59.

Source: POPLINK-database, Demographic Data Base, Umeå University

What stands out across the risk cohorts is the strong shift toward smaller family sizes as shown by the rapidly falling proportions progressing to a family size of 3 or more children. This is the case both within the middle class as well as among women outside the labor market (see also Figure 3). In the first two cohorts, teachers deviate from the other middle-class women by having a higher probability of having a third child and this difference is significant in the first cohort 1900–09 and Pr. Chi2 = .059 in the cohort 1920–39. This is not the case for the last cohort were the contrast between middle class women and non-employed women remain highly significant as shown by the p-value in the legend but all occupations within the middle class show a similar probability of having a third child, a probability between 12% and 16% lower than unemployed women.

3. Concluding discussion

This paper set out to study how female labor force participation and socio-economic stratum influenced fertility for women born between 1880 and 1939. Additionally, special attention is given to the issue of how fertility behavior changed among women in different occupational fields within the middle class. There are some general tendencies that can be seen throughout the period that are exhibited by all women regardless of their social-economic standing and labor market attachment.

First, the results how the rebound in fertility during the baby boom resulted in a clear break with the Northern European fertility pattern where about 20–25% of all women never married and in turn remained childlessness. Second, this shift toward diminishing differences in fertility behavior also meant that differences in family sizes were reduced. During the baby boom, the proportion that had three or more children decreases by more than 30%. This combined decrease in both childlessness and in transition rates to parities above two children resulted in a strong shift toward a two-child norm as more and more women ended up having exactly two children. These two general findings are in no way new and strengthen what we already know about the changing fertility patterns during the baby boom (see e.g. Bean, 1983; Sandström, 2014; Van Bavel et al., 2018; Van Bavel & Reher, 2013). Rather, the main contribution of this analysis is that it is among the first to show how the association between female labor market activity and socio-economic stratum changed during the first half of the twentieth century. Another contribution is that our analysis does not fit with the interpretation that the baby boom period can be understood as a reaction against more equal conditions between men and women and that increased fertility during the period primarily can be understood as a result of women once again retreating from the labor market into the household sphere.

As seen in Figure 3, about 75% of women who came of age in the 1940s and 1950s were active in the labor market prior to first childbirth during the period when almost all of them remained unmarried But as more and more women got married and at a younger age during the baby boom (Sandström, 2017), the composition of the female labor force also changed during the 1940s and 1950s. As already stated in the introduction, this meant that the share of working women that were married increased from only one in 10 in the 1930s to nearly half of the female labor force toward the end of the 1950s (Qvist, 1974). The results presented here show a reduced socio-economic penalty on fertility for economically active women in general, and especially for higher strata women. To summarize, the results show here how the negative association between female human capital and fertility was sharply reduced during the baby boom period regardless of whether we measure this as educational attainment (Reher & Requena, 2015; Sandström, 2014; Van Bavel, 2014; Van Bavel et al., 2018) or, in the Swedish case, as socio-economic stratum based on the occupation of the woman prior to first childbirth. The main pattern found is a strong convergence of childbearing behavior during the baby boom across the different social-economic strata, where a two-child norm was combined with very low levels of childlessness and a weaker connection to the woman’s socio-economic standing and labor market attachment as opposed to the pre-baby boom period.

It should be noted that we do not consider whether and to what extent women continued to be gainfully employed after having their first child. Institutional support for working mothers was still weak during the 1940s and 1950s, and previous research indicate that childbearing resulted in a period out of the labor market for a majority of women born prior to the mid 1940s (Stanfors, 2003). Therefore, we can safely assume that most of the women in the cohorts under analysis had at least some years out of the labor market when their children were young. However, according to the labor market surveys conducted by Statistics Sweden since the 1960s, approximately 47% of both married women and those having a child aged 0–16 years were gainfully employed in 1965 during the second peak of the baby boom (Statistics Sweden, 1972). Women with larger investments in labor market human capital, such as those having longer educations, can be assumed to have continued to work to a higher degree than women with no record of being employed. Also, it is likely that these women returned to the labor market earlier and to a greater extent than women with lower levels of human capital investments. Therefore, it is intriguing that the results show that especially women employed in the higher social strata occupations changed their behavior by far the most in the baby boom cohorts, as these women should have experienced higher opportunity costs related to childbearing than women with lower levels of human capital investments and/or weaker labor market attachment.

As stated in the introduction our focus here is not primarily to answer the question of why the fertility recovery occurred in general, but rather to contribute to the understanding of one of it aspects, in terms of the changes in association between female human capital and labor market activity and fertility up until the 1960s. Although we do not argue that decreased disincentives to family formation for economically active women is the single most important explanation for the recovery in fertility during the baby boom, our results show that this is clearly a part of the underlying mechanism in the Swedish case. The results from comparative studies on how the educational gradient of fertility changed for women during the baby boom in the West also indicates that our results are of relevance for a number of other countries in the in Europe and Anglo-Saxon countries (Van Bavel et al., 2018).

On the matter of how these results fit with more general theoretical explanations for the mid twentieth century baby boom that have been proposed over the years, such as: crowding out of younger women from the labor market (Bellou & Cardia, 2014; Doepke, Hazan, & Maoz, 2015), improved maternal health (Albanesi & Olivetti, 2014), technological progress in the household sector (Greenwood, Seshadri, & Vandenbroucke, 2005) and more generally the sustained economic recovery and high levels of economic growth that prevailed in the West after the severe downturn during the depression in the 1930s (Caldwell, 2006; Easterlin, 1961). It is clear that our findings do not fit at all well with the mechanism of crowding out younger women from the labor market, while it is more compatible with others, such as improvements in maternal health and technological progress.

Regarding the issue that is specifically addressed in our analysis of why employed upper strata women increased their fertility in such a pronounced manner during the baby boom period, we purpose that aspects of the pronatalist polices implemented in the late 1930s and throughout 1940s in Sweden stand out as a substantial structural change that need to be accounted for. Our interpretation of the changing behavior in this group of middle-class women thus primarily point to changes in constraints that especially influenced the family formation of women with higher investments in human capital. We do not dispute that the baby boom as a general phenomenon was associated with a shift in values were small families increasingly was framed as undesirable (Bean, 1983; Blake & Gupta, 1975; Surkyn & Lesthaeghe, 2004). On the contrary we think that changes in values were a necessary precondition for the policy shifts that occurred in many countries in response to the low fertility levels before the baby boom. However, we see no immediate reason that such a value shift would influence middle-class women more than other groups in society. Therefore, changes in the labor market and in the institutional structure stand out as factors that are difficult to overlook when trying to understand the disproportionate increase in fertility among middle-class women. Theoretically, the importance of institutional structures such as pronatalist policies and social insurance directed at families and mothers have been stressed by McNicoll (1980, 2001)) and, with regards to the baby boom, the population polices implemented in a number of Western countries in response to the ‘crisis in the population questions’ in the 1930s (Myrdal & Myrdal, 1934) have been identified by a number of scholars as a contributing factor to the baby boom (Caldwell, 2006; Chesnais, 2005; Notestein, 1950; Calot & Sardon, 1998). In the Swedish case, the pronatalist reforms introduced after the report of the Population Commission in 1936 were explicitly designed to reduce the obstacles to marriage and childbearing for gainfully employed women (Hatje, 1974; Hirdman, 1989; Myrdal, 1941).

This was especially evident in the new law on female employment protection that was implemented by parliament in 1939 (Bill No. 114, 1939; Frangeur, 1998). This law made all employers with more than four employees liable if they fired or made unfavorable changes to working conditions for a woman who had got engaged, married or pregnant. Furthermore, a right to three months of maternity leave with pay that was further extended to six months in 1945. The maternity benefit was not insubstantial, and most privately employed women received about 40% of the mean monthly female salary in benefits, while government employees received even higher payments (Frangeur, 1998). The Social Democratic government stated in the bill from 1939 that the intention of the law was to increase the possibilities for women to ‘combine marriage and motherhood with gainful employment’ (Bill No. 114, 1939). In 1955 the maternity benefit was further extended to include an additional supplement with nine different levels that was directly related to the wage that was lost while the mother was on maternity leave. This reform worked to further reduce the loss of income for women with higher salaries when they were on maternity leave (Social Policy Commission, 1961, pp. 16–22). Klas Åmark argue that his reform had a profound impact in the behavior of women that now were incentivized to have stable employment and sufficient income to qualify for the income supplement in the maternity insurance before having their first child. Åmark argues that these improvements in the maternity leave benefit worked to shift Swedish families toward a dual provider model. It is clear that the incomes of women that formed a family in the 1950s increased substantially as the share of mothers who qualified for the income based benefit rapidly increased from 23% in 1955 to 42% already in 1959 (Åmark, 2005, pp. 250–251).

The timing of these reforms from the late 1930s up until the mid 1950s fits well with the interpretation that these institutional changes worked to substantially reduce the alternative cost of family formation for women with relatively large investments in labor market skills and education. What we can show is that the pattern of high fertility among women employed in so called ‘caring professions’ in the public sector found in contemporary Sweden and other countries in the West was established already during the 1940s and 1950s by women that had their peak reproductive years during the baby boom. Further, our results show that the women in education and healthcare in this cohort (born 1920–39) were less likely to remain childless than other middle-class women. In the last risk cohort that turned 20 during the peak baby boom period in the 1940s and 1950s, we can also see the formation of the high fertility pattern found among women in the caring professions as especially nurses followed by teachers reach the lowest levels of childlessness and the highest transition rates to parity 2.

Prior to 1939, a substantial share of women and especially those employed in the white-collar sector were dismissed from work if they choose to start a family (Frangeur, 1998) Previous research has shown that the acceptance of continued employment of white-collar women after marriage had its breakthrough within jobs in the national government such as teachers. Here the institutional change occurred already in 1923 when married women were given equal rights to government positions through a new law passed by the Swedish parliament in that year. In 1926, the Social Democratic parliamentarian Elof Lindberg argued that this had meant that female teachers as a rule remained in employment when they got married (Frangeur, 1998). There are indications that similar rapid changes in the labor market attachment of married teachers was going on also in other countries such as the US, where the share of schoolteachers who were married increased from 9.7% in 1920 to 22% in 1940 (Cott, 1987). For women in healthcare, who were mostly employed in the county and municipal public sector, it was only with the legal change in 1939 that the final breakthrough came. But, it was clearly among private employers in, for example, the banking sector that the resistance to giving women a legal right to continue working after marriage was most entrenched (Frangeur, 1998).

The timing of these institutional changes and differences in the acceptance of the employment of mothers and married women across different sectors employing white-collar women fit well with the tendency for teachers to be more similar to women outside the labor market compared to other middle-class women already in the pre- baby boom cohort. This was especially the case in terms of having the highest propensity to have a second and a third child of all middle-class women. However, women employed in healthcare who came of age during the peak baby boom years after the 1939 law exhibit remarkable increases in fertility where they go from having the highest to the levels of childlessness of all middle-class women to levels at least as low as non-employed women. Again, results from historical studies indicate that attitudes to the combination of family and employment for women in healthcare were changing rapidly in the 1940s. Åsa Andersson has shown that the discourse within the nurses’ union changed during the 1940s in terms of married nurses increasingly being discussed as something common which only 15 to 20 years prior would have been regarded as a fringe behavior (Andersson, 2002).

The higher alternative cost experienced by women employed in white-collar private sector jobs is also seen in the differences in fertility behavior within the middle class. Although women employed in clerical work, of whom many worked in banks and private companies, experienced large decreases in the levels of childlessness, they continued throughout the entire period covered by the analysis to be more likely to stop childbearing after having only one child. Similarly, that we find a tendency for higher levels of childlessness in the group defined as other middle class is what we would expect. This group contains a substantial proportion of higher professionals, and that these women continue to have higher levels of childlessness is plausibly explained by their on average higher levels of investment in educational capital and the higher alternative cost of childbearing in a setting where institutional support for working mothers was still rudimentary compared to contemporary Nordic welfare state regimes.

Summarizing, the results show how the fertility patterns established during the baby boom work as a hinge that connects the fertility regime that prevailed during the fertility decline that ended in the 1930s with its strongly negative SES-gradient, to the fertility patterns found in Sweden today. Since then, employment has increasingly become a prerequisite for female family formation rather than the opposite (Hoem, 2000; Ministry of Social Affairs, 2001; Statistics Sweden, 2002). In addition, recent cohorts of Swedish women with tertiary education have ended up with much lower levels of childlessness than those with only primary education (Jalovaara et al., 2018). Lastly, our results show that the high fertility observed among women employed in education and healthcare in the public sector was established already in the 1940s and 1950s in Sweden. Here, we argue that more ‘family friendly’ attitudes that were part of the pronatalist polices of the 1930s and 1940s was likely more readily accepted by the employers in the public sector. The fact that the greatest changes in behavior are seen among middle-class women employed in education and healthcare indicates that employers in the public sector were less inclined to resist the guidelines set out by the new employment protection law implemented to protect the rights of married women and mothers on the labor market in the late 1930s.

Disclosure statement

No potential conflict of interest was reported by the authors.