Materfamilias: the association of mother’s work on children’s absolute income mobility, Southern Sweden (1947–2015)

Abstract

This article examines the association of mothers’ income with children’s economic mobility in a period of increased women’s labor market participation in Sweden. I found that whether a mother was economically independent and had an income similar to that of the father during her children’s late childhood and adolescence positively associated with upward mobility. The results show a substantial association of mother’s income position to their daughters’ mobility, but not for sons’. Among the primary mechanisms, I argue that extra resources from mothers helped human capital investment through education and that mothers influenced daughters by a gendered role model.

1. Introduction

Intergenerational associations on socioeconomic mobility have traditionally attracted social scientists’ interest, who have found evidence on how social mobility has developed since the start of industrialization until today. We now have enough knowledge to argue that higher degrees of intergenerational mobility correlates positively with creating opportunities and lower levels of inequality at the societal level (Corak 2013).

Nevertheless, recent studies on this topic have focused almost exclusively on the influence of fathers on their children while neglecting the likely impact of mothers on intergenerational mobility (Beller 2009). Researchers have argued that in the modern period, women (mothers) have become increasingly likely to earn more than men (fathers), to have a higher educational status, and to head a household (McLanahan and Percheski 2008; Blau et al. 2013). Thus, only including inputs for fathers in most social mobility analyses could lead to biases in its measurement.

Most studies that have paid more attention to the mother’s associations on intergenerational social mobility have focused only on periods after 1970 and thus have neglected earlier periods. However, the main arguments against extending studies on the mother’s influence to earlier periods are a lack of data. In most Western male breadwinner societies before the 1970s, the mother’s employment status had little impact on intergenerational mobility. Accordingly, there is little research on the mother’s influence on intergenerational mobility during historical periods when mothers had school-aged children.

This article aims to study the association of the mother’s income on her children’s social mobility in a period when women’s labor participation was still increasing. I evaluate mothers’ influence in the extensive margin (mothers’ working status) and the intensive margin (the economic independence of mothers) during their children’s late childhood and adolescence. I hypothesize that whether a mother was closer to full-time, paid work could have influenced her children’s income mobility. The main focus of the analysis is on the decades between the 1940s and the 1980s, which represented a period of transition for women’s labor market participation in Sweden. Thus, in this study, I seek to determine whether mothers’ labor market participation intensity and income correlated with their children’s social mobility. For this purpose, I use rich longitudinal data from Southern Sweden with enough income and sociodemographic information to enable me to study economic mobility consistently.

My findings show that the mother’s income level did not directly associate with her children’s social mobility. Thus, I found that whether a mother income was similar to that of the father (which is a proxy for economic autonomy), and whether the mother’s income was similar to that other mothers during her children’s late childhood and adolescence associated with the upward mobility of her children, and especially of her daughters. These results suggest that while a mother’s income level may not directly influence her children’s future income, a relatively economically independent mother can serve as a role model for her children (especially daughters), promoting social mobility and narrowing gender disparities.

2. Background

2.1. Influence of mothers on intergenerational income mobility and labor force participation

Despite the important literature in recent decades on income mobility, most studies have examined father–son relationships. In contrast, fewer studies have focused on father–daughter, and even fewer studies have analyzed mother–child. This is partly due to a lack of suitable data on women’s incomes, as assessing women’s lifetime income can be difficult because women’s labor force participation is often interrupted by factors such as childbearing, and women have higher degrees of labor career truncation after having children. However, like the other Nordic countries, Sweden has more data on both father–daughter and mother–child income mobility than most other countries because its female labor force participation rates are relatively high (Nybom and Stuhler 2016). The results of studies of income mobility across different countries have shown that income mobility between fathers and their daughters tends to be high. These mobility levels are affected by overall mobility and gender disparities in different labor markets (Pascual 2009; Heidrich 2017). These results also indicate that mother–child income mobility associations are usually much lower than father–child relationships. In some contexts, elasticities between mothers and daughters are higher than those between mothers and sons (Pascual 2009).

Therefore, the available literature is even scarcer if we only focus on the associations on intergenerational income mobility of the mother’s employment during her children’s childhood. To my knowledge, the only study that has examined the influence the mother’s paid employment on intergenerational income mobility is a chapter in a book by Stinson and Gottschalk (2016). To assess the economic mobility of children born between 1978 and 1982, the authors used American survey data on labor force participation linked to administrative records on family, demographic, and earnings information. In general, they found no evidence that the mobility levels of children differed depending on whether their mother was or was not in paid employment. Similar findings for intergenerational mobility in economic or academic performance have also been reported (Dunifon et al. 2013). Finally, Delabastita and Buyst’s (2021) study based on occupational mobility in nineteenth-century West Flanders found daughters to be more mobile than sons, but with minor translations into upward socioeconomic status with a significant social gradient.

In their study, Stinson and Gottschalk (2016) observed that daughters whose mothers were working when they were in high school were more likely to be in the labor market themselves. Such findings relate to a much more studied topic, which refers to the intergenerational mechanisms involving children’s labor force participation, primarily from daughters. Accordingly, Fernández (2013) contributes with a comprehensive analytical approach that advocates for a net gender role of mothers and other women relatives, increasing the labor participation of younger women within families in explicit “emulation” behavior. Similarly, recent studies that have examined the effects of the mother’s employment on her children when they are older have found substantial evidence of such effects, especially on the economic performance of her daughters.

However, other studies using exogenous shocks as the world wars have also pointed to different mechanisms explaining the higher probability of engaging in the labor force for daughters. Fernández et al. (2004) pointed that rather than a direct maternal effect (mother–daughter), women in the post–World War II in the USA would have their probability of staying in the labor market longer if they were married to men whose mothers also worked. In other words, the authors argued for a more substantial mother-in-law effect. Similarly, Gay (2019) studied post–World War I France to get to the findings of geographic-based influence, expanding the interpretations to a more varied social interaction for women to start gainfully working.

Nevertheless, in the period covered in this paper, the labor participation of daughters (adults between the 1960s and the 2010s) was widespread in Sweden (Stanfors and Goldscheider 2017). Therefore, this paper follows the hypothetical approach of a positive gendered role transmission from mothers to daughters in terms of upward income mobility, in line with what has been shown in many industrialized countries (McGinn et al. 2019; Olivetti et al. 2020).

The children observed in this study grew up in the period between World War II and the first years of the 1980s. During this period, Sweden was experiencing rapid economic growth and deep societal transformations (Schön 2010). In addition, during this period, economic inequality was decreasing monotonically, following a trend that started at the beginning of the century (Roine and Waldenström 2008).

In the first two decades after World War II, high school and higher education enrollment rates increased rapidly in Sweden due to the equalization of access to education across gender and class. The welfare state continued to expand during the 1970s and 1980s in strategic areas such as income security, education, child care, and health care. Although Sweden, along with the other Nordic countries, is today considered a global leader in promoting gender equality and social policies aimed at supporting working mothers, this was not always the case. The first Swedish policies on parental leave were adopted as early as 1974, and more egalitarian leave policies, including policies designed to promote gender parity, date back to the early 1990s (Ekberg et al. 2013). Hence, although women’s labor force participation rates increased continuously after World War II (Gustafsson and Jacobsson 1985; Stanfors 2003), before the late 1970s, the Swedish state offered women few incentives to work, especially after having children.

3. Data and analytical strategy

3.1. The Scanian Economic-Demographic Database

I use data from the Scanian Economic-Demographic Database (SEDD) linked to the national register data from Statistics Sweden (SCB). SEDD is a unique database of individual-level longitudinal data from five rural and semi-urban parishes and a port town in Southern Sweden spanning 1650–1968 (Bengtsson et al. 2021). The database can be used to follow the demographic and socioeconomic outcomes of individuals across multiple generations from preindustrial times up until the present. For the analysis, I also use information from continuous population registers (a household-based register continuously updated) on demographic events, including information on migration to and from households for all individuals in the selected parishes.

Moreover, from 1968 onward, individual-level information covering the entire country is available from various administrative registers at SCB. These data were linked to the pre-1968 historical sample. First, the linkage makes it possible to follow until 2015 any individual who ever lived in the area before 1968 and was still alive in that year, regardless of his/her geographic location in Sweden. Second, various kin, such as parents, grandparents, children, and siblings of individuals who belonged to the original population in Scania were aggregated to the database if they were alive and residing in Sweden sometime after 1967 and until 2015. The most important economic variable used in this study is the total income from sources related to labor (including self-employment but excluding income from capital and real estate). All income data were adjusted for constant prices using the consumer price index (SCB 2020).

For this study, separate income information should be available for the mothers in the sample, most of whom would have been married. Until 1947, married women were usually not included in the income and taxation registers, but their earnings were included in the tax returns of their husbands. However, between 1947 and 1953, the earnings of married women were included separately from those of their spouses in the registers, and between 1954 and 1971, both their income and their taxes were reported separately. Thus, although the Swedish tax reform implementing separate taxation for married men and women dates back only to 1971 (Gustafsson 1992), different economic information for mothers is available in the registers at least from 1947 onward. Therefore, this analysis focuses on the information for men and women between 1947 and 2014.

3.2. Assessing lifetime income

For computing measures of intergenerational income mobility, I follow children born between 1940 and 1964 who were living in the study area (the five rural and semi-urban parishes and the port town) during their late childhood and adolescence (ages 7–18 years). This strategy was chosen for two different reasons. First, it enables me to include information on the income reported by mothers, which became available starting in 1947. Thus, the earliest birth year for the children in the analysis is 1940. Second, it allows me to collect enough income information on the most productive working ages in adulthood to get consistent lifetime income measures, which are, in practice, until age 50 years. Thus, for the youngest birth cohorts in the analysis (born in 1964), 2014 is the last year with available income information.

Lifetime income reports the mean income for children and their parents between ages 23 and 50 years¹. It should be noted that these data allow me to use income in an almost identical lifecycle for parents and children with a considerable number of income observations for both generations (G1 and G2). Therefore, I am able to greatly reduce issues of attenuation bias (Solon 1992; Björklund and Jäntti 1997; Mazumder 2005) and of lifecycle bias (Bhuller et al. 2011; Nybom and Stuhler 2016).

Having a lifetime income measure as the mean income over adulthood allows me to set different quantiles of income distribution based on the birth years and the sex of the children. A similar approach was used for the parents, with the only difference being that they were divided into groups based on the birth years of their children. A similar procedure has also been adopted in other studies (Chetty et al. 2014, 2017; Heidrich 2017).

The different quantile measures were created based on comparisons to a much bigger sample from 1968 onward when the modern registers were linked to SEDD. The full sample of individuals with yearly positive income information is displayed in the following way. From 1947 to 1967, only information from the study area was available, which turned the sample of individuals with yearly valid income information into around 20,000 people per year. Then, from 1968 onward, the total sample increased to roughly 175,000 individuals per year, and continued to rise, reaching 400,000 cases per year in the 2000s. An advantage of having a bigger sample was that at least from 1968 onward, different lifetime incomes and quantiles could be assessed based on an income scenario that was much closer to the national level.

3.3. Measuring a proxy of mothers’ working status

Another important variable used in this study was a proxy to the working status of mothers during childhood. On an annual basis, the most accurate information is generally provided by the Labor Force Surveys (Arbetskraftsundersökningen). However, these surveys were first started in 1963 (Gustafsson and Jacobsson 1985; Stanfors 2003).

On the other hand, occupational information from decadal or quinquennial censuses and other administrative and parish data can reach further back in time. This kind of occupational data is present in the SEDD data, yearly for the historical setting (in our case, from 1947 to 1967), and every 5 years from 1968 until the 2000s. In this regard, all occupations, when available, have been coded using the historical classification of occupations and classified into social classes with HISCLASS (Van Leeuwen et al. 2002; Van Leeuwen and Maas 2011). However, using occupational data is associated with considerable methodological challenges, as women’s occupational information tends to be less reliable than men’s, especially in historical periods. This problem has been reported both in the international context (Humphries and Sarasúa 2012) and in Sweden (Nyberg 1989; Stanfors 2014).

Thus, I used a different approach to proxy whether an individual was or was not working. It was based on the continuous annual income information present in the database. In applying this approach, I was taking advantage of previous methodologies used by the OECD in research about taxation. These publications introduced the concept of the average production worker, which can be defined as, in essence, the average income of blue-collar workers. It can approach the median income of a given population in which blue-collar workers were still a considerable fraction of the population, as would have been the case in many parts of Sweden between the end of World War II and the 1980s (Schön 2010; Dribe et al. 2015). Hence, as a proxy for this methodology, for each year that I wanted to approach whether a woman was working at least working part-time based on her reported income, I first computed the median income for the year. Next, I set the working status threshold by simply multiplying the median income by 0.33. Finally, in the last step, I created a dummy variable for not working (0) and working (1) for all individuals below or above the threshold (see the appendix for more details).

3.4. Intergenerational data and variables

The complete analysis was carried out only for individuals who were living in the study area during their late childhood and adolescence (ages 7–18 years). However, for the sake of robustness and to allow for better comparisons with studies carried out at the national level, I also computed measures of intergenerational income persistence for the full sample.

Table 1 displays the main sociodemographic variables for both the study sample and the full sample of both children and parents (G2 and G1). Overall, the values across the samples are similar, with the only difference being that the mean age (computed from all observations individuals have) was slightly higher for parents in the study sample than in the full sample. This could be due to some left-censoring in the full data, which does not include a large number of observations until 1968 onward.

Table 1

Descriptive statistics of the main sociodemographic variables by generations (G1 and G2)

Variables	Study sample					Full sample
	μ	SD	Min	Max	N	μ	SD	Min	Max	N
Age at observation
Children	36.26	—	23	50	—	35.66	—	23	50	—
Fathers	39.19	—	23	50	—	35.18	—	23	50	—
Mothers	38.77	—	23	50	—	35.31	—	23	50	—
Birth year
Children	1953	7.31	1940	1965	9002	1952	7.59	1940	1965	108744
Fathers	1924	10.01	1897	1948	3558	1927	8.47	1897	1948	29216
Mothers	1927	9.90	1897	1949	5695	1928	8.33	1897	1950	32837
Age at birth
Fathers	29.96	—	—	—	—	28.79	—	—	—	—
Mothers	27.15	—	—	—	—	26.07	—	—	—	—
Childhood observation (G2 aged 7–18 years)
Years	1966	—	1947	1983	—	—	—	—	—	—
Age of children	12.5	—	7	18	—	—	—	—	—	—
Age of fathers	43	—	24	82	—	—	—	—	—	—
Age of mothers	39.8	—	22	68	—	—	—	—	—	—
Sibship size	1.73	0.95	1	9	—	—	—	—	—	—
Father working (sum of years)	8.52	4.29	0	12	—	—	—	—	—	—
Mother working (sum of years)	5.88	4.44	0	12	—	—	—	—	—	—
Household SES (highest) (%)
Higher whither-collar	17%	—	—	—	—	—	—	—	—	—
Lower white-collar	40%	—	—	—	—	—	—	—	—	—
Medium-skilled workers	23%	—	—	—	—	—	—	—	—	—
Farmers	3%	—	—	—	—	—	—	—	—	—
Lower-skilled workers	16%	—	—	—	—	—	—	—	—	—
Unskilled	2%	—	—	—	—	—	—	—	—	—
Children’s observations in adulthood
Residing out of the study area (SEDD)	55%	—	—	—	—	—	—	—	—	—
Highest SES (%)
Higher whither-collar	21%	—	—	—	—	—	—	—	—	—
Lower white-collar	51%	—	—	—	—	—	—	—	—	—
Medium-skilled workers	13%	—	—	—	—	—	—	—	—	—
Farmers	1%	—	—	—	—	—	—	—	—	—
Lower-skilled workers	13%	—	—	—	—	—	—	—	—	—
Unskilled	1%	—	—	—	—	—	—	—	—	—
Main independent variables (quantiles)
	Q1		Q2		Q3		Q4		Q5
Maternal variable during childhood	μ	SD	μ	SD	μ	SD	μ	SD	μ	SD
Mother–time spent working	0	0	0.121	0.042	0.421	0.12	0.882	0.123	—	—
Income’s parental ratio	0.025	0.03	0.196	0.059	0.415	0.072	0.9	0.418	—	—
Mother’s income quintile	1195	1511	10514	3413	22343	3446	34793	3952	58114	23954

Source: SEDD

Moreover, in the lower part of table 1, we can see late childhood and adolescence information for the G2 study sample. It spans 12 years of observations for each individual over the 1947–1983 period. Interestingly, from the annual working status dummy variable, we could obtain the sum of the years spent in the working proxy during the childhood of G2 for both fathers and mothers. The results show that, on average, mothers would have been working only about half (close to 6 years) of the period when their children were aged 7–18 years. However, it has to be taken into account that as the proxy for considering parents to be working had an income over 33% of the median income, individuals below this threshold would be considered non-working; the reason for the average working years of fathers was about nine.

Finally, from the lifetime income for parents and children, I have created different income variables to be tested within the income persistence measures. This approach is in line with the most recent literature on stratification, which has tried to disentangle which income combinations better track the intergenerational influence on children. Thus, another two variables were derived in addition to the variables for the lifetime income of fathers and mothers. The first measure captured the average lifetime income of both parents over time by summing the income of the father and the mother who shared the same children and dividing it by two. The second measure picked up the highest income present in each year. While this was usually the father’s income, in some cases in which the father’s income was low or absent, the mother’s income was observed.

4. Methods

First, I compute the intergenerational rank–rank associations developed in previous studies by Chetty et al. 2014. In rank–rank models, the dependent variable regarding G2 income is replaced by the children’s percentile of lifetime mean income, assessed by their birth year (separately by men and women). The dependent variable, in turn, takes the G1 percentile in the distribution of all fathers or mothers who had children born in the same year. The lifetime income used in rank–rank estimates is defined with a focus on the children, which means that the percentiles were derived among all parents sharing children who were born in the same year.

$$ \begin{equation} y{(children)}_i=\alpha +{\beta}_1\, parent{s}_i+{\varepsilon}_i \end{equation}$$

(1)

After measuring the income persistence across generations, I focus only on the study sample for whom we have information on the G2 late childhood and adolescence. This analysis investigates absolute income mobility or, in other words, the rank mobility of each individual and having a lifetime income higher than that of the father. Here, to reflect the association of mothers, the models have as dependent variables the rank position of individuals and the likelihood of having absolute mobility compared with the father.

The income mobility determinants are assessed in two ways. First, I run OLS regressions on the percentile rank of sons and daughters to test whether maternal income could associate with better performance in the economic position of their children. Second, I apply logistic regression models with the dependent variable of earning more than the father did. The specified models test four different types of main independent variables. First, I include a variable regarding mothers’ working status during their children’s childhood in the extensive margin. Accordingly, following the proxy of mothers’ work set by the yearly threshold of 33% of the median income, I categorize whether mothers would have worked at least 1 year against those who would never work when their children were aged 7–18 years. Second, in the intensive margin, I test three different outcomes regarding mothers’ work intensity and their relative economic independence during their children’s childhood.

The first variable in this block relates to the mother’s time-intensity working when her children were aged 7–18 years, which implies summing the years they worked and categorizing the results by quartiles (see table 1). The second variable considers both the father’s and the mother’s mean income during G2 childhood to assess the mother’s contribution to the parental economy. Thus, I compute a ratio of childhood income, by dividing the mother’s income/father’s income. The ratio gives a result closer to zero if, for instance, the mother did not have any income in that period, and a value close to one if the mother’s mean income during G2 childhood was similar to that of the father. This variable is also categorized by quartiles². The third variable in this block refers to the quintile of the mothers’ income distribution during G2 childhood, which simply places the mothers in comparison to each other. Besides, in all models, I set controls on the father’s income and the sibship size to control the number of children per family.

$$ \begin{equation} {\displaystyle \begin{array}{l}{Y}_i={\beta}_1{\left( mother\ working\ status\ proxy\ and\ income\ during\ G2\ childhood\right)}_i\ \\ {}+{\beta}_2{\left( socioeconomic\ and\ demographic\ controls\ during\ G2\ childhood\right)}_i+{\varepsilon}_i\end{array}} \end{equation}$$

(2)

5. Results

5.1. Did mothers directly associate with their children’s income? Insights on intergenerational income persistence

In the first approach, I attempt to disentangle how the mother’s economic status might have influenced her children by assessing her direct influence. To do so, I first look at estimates of income persistence measured with rank-rank, which indicates to what extent the income positions of sons and daughters correlated with the income of the father, the income of the mother, the averaged parental income, or the higher of the two parental incomes.

Table 2 presents estimates of rank–rank for sons and daughters for all of the birth cohorts studied (1940–1965) for both the study and the full samples. The associations for the father’s income show similar values for sons (around 0.2) and daughters (between 0.18 and 0.14) in all sample specifications. These results are in line with most of the existing empirical evidence for Sweden; I find a little variation in the coefficients, ranging between 0.18 and 0.27.

Table 2

Intergenerational rank–rank associations for sons and daughters by different income measures

	Study sample		Full sample
	Son	Daughter	Son	Daughter
Father	0.21	0.18	0.21	0.16
	(0.02)	(0.02)	(0.007)	(0.008)
N	4,090	4,005	14,864	14,349
Mother	−0.02	0.04	0.032	0.060
	(0.02)	(0.02)	(0.007)	(0.007)
N	4,379	4,316	16,685	16,148
Average	0.22	0.18	0.17	0.14
	(0.02)	(0.02)	(0.006)	(0.007)
N	4,322	4,277	19,407	18,753
Highest	0.21	0.18	0.18	0.14
	(0.02)	(0.02)	(0.007)	(0.007)
N	4,368	4,320	19,742	19,087

Note: The table reports rank–rank associations (slopes) for the lifetime income percentile of sons and daughters and their parents (separately and together). The study sample refers to all children who spent their childhood and adolescence in the SEDD area regardless of where they lived in Sweden as adults. In contrast, the full sample includes individuals from any place in Sweden who might not have ever lived in the area. The coefficients are shown with standard errors between parentheses below.

Source: See table 1.

My estimates show associations of 0.21 for the study area and 0.18 for the full sample, while previous studies found estimates of 0.27 (Nybom and Stuhler 2016), 0.21 and 0.23. Overall, my measures of income persistence are in line with those of other studies and confirm previous findings showing that in Sweden, income persistence is lower than in other countries such as the USA, where estimations are usually above 0.4 (Corak 2013; Mazumder 2016), or are around 0.35 (Chetty et al. 2014). Father–daughter income associations are much less common in the literature, but in comparison with the findings of studies at the international level as well as in the Swedish context, the estimates shown in table 3 are similar. Overall, income persistence between fathers and daughters is found to be lower than that between fathers and sons (Pascual 2009; Heidrich 2017).

Table 3

OLS regressions on children’s average lifetime income rank attained by proxied working status and income of mothers when children were aged 7–18 years

A	All	Sons	Daughters
Mother never worked (ref)
Mother worked	0.17	−0.88	1.13
	(0.68)	(0.97)	(0.92)
N	8095	4090	4005
adj. R-sq	0.257	0.045	0.064
B	All	Sons	Daughters
Mother work intensity (Q1 ref)
Q2	0.48	−0.15	1.21
	(1.08)	(1.51)	(1.51)
Q3	−1.15	−2.81*	0.35
	(0.90)	(1.27)	(1.21)
Q4	1.68	0.71	2.76*
	(0.88)	(1.24)	(1.19)
N	8095	4090	4005
adj. R-sq	0.259	0.048	0.065
C	All	Sons	Daughters
Childhood income ratio (mother/father)(Q1 ref)
Q2	−0.20	−0.85	0.85
	(0.75)	(1.09)	(1.02)
Q3	1.56	1.45	1.99
	(0.83)	(1.22)	(1.09)
Q4	1.51	−0.61	3.91**
	(0.96)	(1.39)	(1.22)
N	8095	4090	4005
adj. R-sq	0.258	0.046	0.066
D	All	Sons	Daughters
Mother income quintile (Q1 ref)
Q2	−1.00	−2.20	0.40
	(0.83)	(1.17)	(1.12)
Q3	−0.07	−0.36	0.43
	(0.87)	(1.25)	(1.17)
Q4	0.41	−2.26	3.12*
	(0.94)	(1.37)	(1.24)
Q5	1.12	−0.88	3.47*
	(1.08)	(1.54)	(1.39)
N	8095	4090	4005
adj. R-sq	0.258	0.046	0.066

Controls: sex, father’s income percentile, and sibship size.

Source: See table 1.

^*p < 0.1,

^**p < 0.05,

^***p < 0.01.

Nevertheless, the estimates for mothers and their children, which are of primary interest in this study, show that the mothers’ economic positions had almost no influence on their children. The rank–rank had the highest values, at around 0.04 and 0.06, respectively; they were not statistically significant in many cases. Overall, I found higher associations between mothers and daughters than between mothers and sons; a gradient that is the opposite of that found for the association between fathers and their children and that is in line with previous studies (Pascual 2009; Heidrich 2017). Interestingly, the other two income measures used, parental average income and the highest parental income observation, do not show higher coefficients for sons and daughters for lower standard errors or higher R² values (not reported in the table). In contradiction to some of the recent literature that has argued that the averaged income measure (both parents) has a more direct association with income persistence, the results in this study indicate that only the father’s income had more influence. This could be because the children’s cohorts included in this study were older than those used in most other studies on the same topic, as I observed G2 born in the 1940s, whereas other studies have used cohorts born in the late 1950s or the 1960s. A simple look at the evolving patterns of women’s labor market participation rates in recent Swedish history would show us that mothers of individuals born in late 1940s or 1950s had lower labor market participation rates, especially after childbearing. Thus, models of income persistence that consider the mother’s income could be underestimated (Stanfors 2003; Bhuller et al. 2011; Heidrich 2017).

The results for intergenerational income persistence showed that the mother’s income per se had a small association with her children’s economic outcomes in adulthood. However, having a mother who was participating in the labor market during the child’s childhood and adolescence could be associated with decreased income persistence between the father and the child. In figure 1, I present the binned scatterplots of the non-parametric association between children’s income ranks and their fathers. I compare children whose mothers were categorized as never gainfully working (following the income proxy) or as gainfully working at least 1 year when children were aged 7–18 years. The figure shows two panels, A for sons, and B for daughters, displaying twenty bins, which average the income rank of children on the y-axis by their fathers’ income rank on the x-axis. Besides, the figures also show a regression line for each rank association for each specified category.

Figure 1

Binned scatterplot of the mean average income rank of children (sons and daughters), by the father’s income percentile and mother’s proxy of working status when children were aged 7–18 years.

Overall, it can be seen that a linear regression testing the extensive margin, whether mothers ever gainfully worked or not when their children were aged 7–18 years is not statistically significant. Still, the figure presents some interesting patterns to consider in the following analysis presented. In this regard, we can see that for sons (figure A), there were virtually no differences between having or not having a gainfully working mother during their childhood. Conversely, figure B shows that daughters of working mothers, on average, attained higher income rank than daughters of non-working mothers. These descriptive results suggest that mothers’ income and relative economic position could primarily act as mediators to contribute to their daughters’ income mobility. Finally, these results are not statistically significant because they pool all working mothers. However, in the next section, I disaggregate mothers’ employment income showing that daughters to mothers at the top of the income distribution are associated positively, both statistically and substantially, with their daughter’s upward mobility.

5.2. Other pathways: mothers’ economic status and their association with their children’s absolute mobility

As suggested in figure 1, the lack of direct association in relative mobility found for mothers does not mean they did not influence their children’s social mobility in other ways. This influence could have operated in a less tangible manner, i.e., a working mother’s influence may have been based on her value as a role model due to her employment status and relative economic independence (Fernández 2013). Thus, in the next sections, I apply models for measuring absolute income mobility in order to investigate how maternal influence correlated with that mobility.

First, I estimate OLS models with children’s income rank as the dependent variable. Similar to Chetty et al. (2014), these models inform us whether children would attain a higher income rank depending on the proxy of mothers’ work and their relative economic position. Table 3 is divided into four different panels, testing dependent variables ranging from simply knowing if mothers were considered to work or not during their children’s childhood (extensive margin) to variables capturing their work intensity (number of years worked by mothers), and their relative economic position compared with their husbands (fathers) and other mothers. The models are estimated for sons and daughters together and separately by sex. Finally, as controls, I include the income decile of fathers, the sibship size, and sex (when including sons and daughters).

Panel A (see table 3) includes a binary variable for the proxy of mothers’ employment when their children were aged 7–18 years, with mothers who were considered not to work as the reference category. Overall, any model specification showed statistically significant coefficients. For both sons and daughters together, and comparing only sons, suggested a negative relationship between mothers’ work and children’s rank attainment. However, the coefficients for daughters indicated a positive association. In panel B, the model includes a variable categorizing mothers’ work intensity when their children were aged 7–18 years by counting the number of years they worked, split into quartiles. In short, Q1 (reference) indicates that mothers never worked, while Q4 group mothers who would have worked always. Again, similar to panel A, the results are not statistically significant for all (sons and daughters) and only sons and point to a negative correlation. Conversely, among daughters, having a mother with the highest work intensity (Q4) would, on average, lead to almost three percentile positions upward (see panel B, table 3).

In panels C and D, the models estimate the association between the average rank positions of children with the relative economic status of mothers. First, in panel C, we can see the correlation with the childhood income ratio by dividing mother’s income by that of father’s, categorized in quartile. The values range from Q1 (reference group), where mothers almost did not have an income, to Q4, where the mothers’ income was similar to fathers during their children’s childhood. Overall, the results are not statistically significant for all and only sons. However, when comparing daughters, we see that those with relatively economically independent mothers (in comparison to fathers) attained almost four upward percentiles higher on average. Finally, in panel D, I included a variable measuring mothers’ income position relative to other mothers when their children were 7–18 years. This variable was characterized in quintiles. Overall, the same non-conclusive pattern for all and sons is observed, while daughters from mothers in the highest income positions (Q3 and Q4) would attain higher income percentiles than other daughters.

Additionally, I have conducted family fixed-effects models (table A4 in the appendix) as a sensitivity to control for unobserved underlying characteristics in the family and to isolate the positive association between mothers’ economic independence and daughters’ upward mobility. The models (as in table 3) have the income rank of children as the dependent variable. Although the results are not statistically significant (because of the sample size), a clear gradient shows that children (especially daughters) from mothers with more independent income attained much better rank positions than others on average.

In any exercise involving men and women in economic mobility studies, it is always important to bear in mind that the existing gender gap in economic outcomes works as an additional barrier to the upward mobility of daughters. A simple absolute income measure, such as observing which children earned more than their fathers, allowed me to capture the extent to which gender disparities might have jeopardized the daughters’ upward economic mobility.

Figure 2 displays the fraction of sons and daughters who earned more than their fathers over time, by birth year. Across cohorts, we see that the oldest sons had higher mobility than the younger sons (declining from 90% to 70%). However, for the daughters, the shares who were earning more than their fathers were constantly lower than they were for the sons, as across all of the birth cohorts, less than 50% of the women were earning more than their fathers. The general level of absolute mobility found for sons and daughters in the study area is in line with that reported in previous studies using different sources (Berman 2022) or similar sources (Liss et al. 2019) in the national context. The absolute mobility differed by sex, as the share of sons who had a higher lifetime income than their fathers was almost double that of the share of daughters. Interestingly, there appears to be a declining pattern of absolute mobility for the sons, from around 90% in the oldest cohorts (1940s) to levels oscillating between 70% and 65% for those men born between the late 1950s and 1960s. Such declining patterns have been found in other European countries and Japan, Australia, and the USA (Chetty et al. 2017).

Figure 2

Fraction of children earning more than their fathers by birth year.

These disparities by gender were also clear in a parametric way by controlling for fathers’ income deciles. Figure 4³ decomposes the absolute mobility by the income deciles of fathers. As expected, the children of poorer fathers (e.g., deciles 1, 2, and 3) were more likely to experience absolute upward mobility than the children of richer parents. A similar pattern has been observed in the USA (Chetty et al. 2017). However, the gender disparity was considerable: whereas the sons of fathers up to decile 5 of the income distribution had a probability higher than 0.8 of experiencing absolute upward mobility, only the daughters of fathers in first decile overcame this threshold. Indeed, the trend for daughters declined almost linearly. For instance, the daughters of fathers in deciles 7 had a probability of 0.2 of experiencing absolute upward mobility, while for the sons, the probability was of 0.6 (see figure 3).

Figure 3

Probability of earning more than their father by father income decile.

In table 4, all models categorize earning more than their fathers or not (binary) as the dependent variable. The model specifications in panels are the same as in table 3, dealing with mothers’ gainfully work proxy and income. The results from these logistic regressions are displayed in average marginal effects to ease its interpretation and allow the sub-samples comparisons. First, in panel A, we see again that the mere division between mothers who work or did not, had no substantial or statistically significant associations in any specification. Similarly, panel B dealing with mothers’ work intensity in quartiles, did not reveal any clear gradient or differences.

Table 4

Average marginal effects of the logistic regression on the probability of children earning a higher lifetime income than their fathers by proxied working status and income of mothers when children were aged 7–18 years

A	All	Sons	Daughters
Mother never worked (ref)
Mother worked	0.02	0.02	0.01
	(0.01)	(0.02)	(0.02)
N	8095	4090	4005
Pseudo R-sq	0.28	0.19	0.25
B	All	Sons	Daughters
Mother work intensity (Q1 ref)
Q2	−0.01	−0.01	−0.01
	(0.02)	(0.02)	(0.03)
Q3	−0.02	−0.01	−0.02
	(0.01)	(0.02)	(0.02)
Q4	0.04*	0.05*	0.03
	(0.01)	(0.02)	(0.02)
N	8095	4090	4005
Pseudo R-sq	0.28	0.2	0.26
C	All	Sons	Daughters
Childhood income ratio (mother/father)(Q1 ref)
Q2	0.01	0.02	0.01
	(0.01)	(0.02)	(0.02)
Q3	0.03	0.05*	0.01
	(0.01)	(0.02)	(0.02)
Q4	0.05**	0.02	0.07***
	(0.02)	(0.02)	(0.02)
N	8095	4090	4005
Pseudo R-sq	0.28	0.19	0.25
D	All	Sons	Daughters
Mother income quintile (Q1 ref)
Q2	−0.00	−0.00	0.00
	(0.01)	(0.02)	(0.02)
Q3	0.01	0.02	0.01
	(0.01)	(0.02)	(0.02)
Q4	0.04*	0.04	0.05*
	(0.02)	(0.02)	(0.02)
Q5	0.06***	0.03	0.12***
	(0.02)	(0.02)	(0.03)
N	8095	4090	4005
Pseudo R-sq	0.28	0.19	0.25

Controls: sex, father’s income percentile, and sibship size.

Source: See table 1.

^*p < 0.1,

^**p < 0.05,

^***p < 0.01.

Conversely, when testing for the association between upward mobility and the relative economic status of mothers (panels C and D), the results showed a strong positive association for daughters. Therefore, daughters of mothers with incomes similar to fathers during childhood were seven percentage points more likely to earn more than their fathers (see panel C, table 4). Besides, the differences were even higher when testing for mothers’ income position in relation to other mothers. Accordingly, daughters of mothers in Q5 had a 0.12 higher probability of upward absolute income mobility.

All results displayed above have shown opposite estimates for sons and daughters, in which the latter tend to be beneficially influenced by mothers’ income. Such differences might suggest gender differences in how maternal work and income could mediate upward mobility. In figures 4 and 5⁴, I estimate models similar to the ones in table 4 interacting the variables on mother income relative position with sex to see how different sons and daughters were. The differences between sons and daughters are clear in figure 4, which predicts the probabilities of children earning more than their fathers by the quartiles derived from the ratio of the mother’s income to the father’s income during childhood. For the children of mothers in the fourth quartile, who had, on average, a mean income similar to that of fathers (when the children were aged 7–18 years), the daughters had probabilities of almost 0.5, while the other daughters had probabilities of between 0.35–0.39. Conversely, for the sons, there were no differences based on the parental income ratio.

Figure 4

Probability of earning more than their father by parental income ratio (mother/father) during children ages 7–18 years (quartiles).

Figure 5

Probability of earning more than their father by mother income during children ages 7–18 years (quintiles).

Finally, figure 5 shows another measure of economic independence, i.e., the probability of earning more than the father by all of the mothers’ income quintiles during the children’s childhood. This measure indicates how economically independent the mothers were compared with other mothers. We can see that the daughters whose mothers worked and had more earnings were three times more likely to earn more than their fathers than the daughters of mothers who did not have an income during their daughters’ childhood.

5.3. Potential mechanisms explaining the beneficial influence of mothers’ income on daughters’ upward mobility

All analyses above suggested a substantial gender difference in how mothers’ income when children were aged 7–18 years could correlate with upward mobility, but which mechanisms could explain such an association? Overall, we know that the general welfare state development in the post–World War II Sweden worked well for social mobility, especially by equalizing opportunities (Erikson and Jonsson 1993). Besides, it has also been argued that women in work were mainly educated women during these years. Thus, once the access to education became widespread, children from educated mothers would also have higher chances of becoming educated and therefore attaining better income positions (Stanfors and Goldscheider 2017). However, this would explain a general increase in mobility for both sons and daughters, instead of only among daughters.

Therefore, I test different mechanisms mediating mothers’ income influence on daughters’ upward mobility in this section. Overall, mechanisms can relate to two potential explanations of how mothers could help on improving their daughters’ income mobility. First, mothers’ income added to that of fathers would prompt more resources to be invested in human capital, which could be strategically allocated to daughters for compensating for gender disparities at the societal level. Second, having the example of a relatively economically independent mother during childhood could act as a gendered positive influence on daughters (Duncan et al. 2005).

The models test three different mechanisms as dependent variables using similar model specifications as the previous tables. First, I estimate the influence of mothers’ income during their daughters’ childhood on the probability of migrating permanently in adulthood by separating daughters who spent their adulthood in the study area or another geographic point of Sweden. Second, a model includes the outcome of having or not having the same social class (HISCLASS) as mothers or not. When occupational information was available for both mothers and daughters, I selected their highest social class in adulthood. Finally, the third mechanism deals with a binary outcome categorizing daughters who achieved tertiary education or not. Table 5 is divided into two parts. Panel A has as the main independent variable the quartile of the childhood income ratio (mother/father), while panel B includes the income quintile of mothers during their daughters’ childhood.

Table 5

Average marginal effects of the logistic regression on the probability of three different mechanisms channeling the influence of mothers’ income when their daughters were aged 7–18 years

A	Migration	Same class	Tertiary
Childhood income ratio (mother/father)(Q1 ref)
Q2	−0.02	−0.03	−0.02
	(0.02)	(0.05)	(0.02)
Q3	−0.05*	0.08	0.02
	(0.02)	(0.05)	(0.02)
Q4	0.02	0.05	0.05*
	(0.03)	(0.06)	(0.03)
N	4005	1001	3690
Pseudo R-sq	0.01	0.03	0.08
B	Migration	Same class	Tertiary
Mother income quintile (Q1 ref)
Q2	0.01	0.07	0.00
	(0.02)	(0.06)	(0.02)
Q3	−0.04	0.10	0.01
	(0.03)	(0.06)	(0.03)
Q4	−0.05	0.11	0.05
	(0.03)	(0.06)	(0.03)
Q5	0.04	0.08	0.12***
	(0.03)	(0.06)	(0.03)
N	4005	1001	3690
Pseudo R-sq	0.01	0.03	0.08

Controls: father’s income percentile and sibship size.

Source: See table 1.

^*p < 0.1,

^**p < 0.05,

^***p < 0.01.

The results show no clear differences between the income statuses of mothers for adult migration. When testing the outcome of having or not having the same occupational class as mothers, it can be seen some positive correlation for mothers’ income (both related to fathers or other mothers) in comparison to Q1 (not having income). However, given the sample size shrinks remarkably because fewer individuals have occupational information, there are no statistically significant coefficients. Finally, the last outcome is having or not having tertiary education, and in this regard, we can see a substantial positive association between mothers’ income and daughters’ higher education. Specifically, daughters with mothers earning similar to their fathers (Q4 at panel A) would have five percentage points higher probability of having tertiary education. Besides, having a mother at the top of the income distribution among other mothers (Q5 at panel B) would place daughters with a 0.12 increase in the probability of attaining higher education than the reference group.

Thus, among the potential mechanisms tested, education stands as the most likely channel through which mothers’ income could benefit daughters. It is clear that educational expansion in Sweden was generally the main channel toward equalization of opportunities; however, it seems that specifically for daughters, mothers’ income and relative economic independence could act as an essential extra.

6. Conclusions

In this article, I studied the association of mothers’ income and a proxy of their labor participation on the economic mobility of their children between the 1940s and 1980s. This was a period of transition for women’s labor market participation in Sweden. My findings on intergenerational income associations were largely in line with the results of the literature on the same topic in Sweden, as I found low income persistence comparable to the levels reported at the national level, even though my study sample covered a smaller area of the country. Additionally, my rank–rank estimations showed that the mother’s income did not associate directly with their children’s income, which has also been observed in most other studies. Most importantly, when analyzing the trends in upward mobility, I found that the mother’s proxied working status and her economic independence during her children’s late childhood and adolescence had a hugely positive association with upward economic mobility, especially for her daughters.

However, my analysis also has some important limitations that should be taken into account. Even if my results are similar in many respects to those of other studies at the national level, they obviously cannot be representative for Sweden as a whole. Finally, it is important to keep in mind that basing employment status is no substitute for having other more consistent occupational and socioeconomic measures for individuals. Despite these limitations, the results represent a valid contribution to understanding maternal roles in economic mobility. Accordingly, the study estimated labor rate participation for periods prior to the Labor Force Surveys using income data, which can be a valid complementary method for addressing the well-known problem of underreported information in historical census data.

Most importantly, the results suggest that while mothers may not directly associate with the incomes of their children, mothers who serve as active role models by participating in the labor market could help promote social mobility and narrow gender disparities. Mothers’ relative economic autonomy during their children’s late childhood and adolescence could be correlated with their daughters’ future labor supply decisions and economic growth. Therefore, as seen in the tested mechanisms, the potential channels determining the results could vary. On the one hand, when the mother served as a double role model (by performing both paid labor and household labor), children’s main exposure was to a working mother. Thus, being exposed to a paid working mother in a crucial age range, such as between 7 and 18 years old, may have helped to generate opportunities and furthered the development of the welfare state in Sweden (1950s–1980s). Education was obviously well established as the main channel for fostering social mobility in this context. Hence, mothers could add more economic resources at home.

On the other hand, and more specifically, the mothers’ paid labor and relative economic independence may have acted as a booster of human capital development. For instance, the father’s income might have mainly provided for the family’s basic needs. In contrast, the mother spending more time with her children would have known where to strategically allocate resources to meet her children’s needs to improve their human capital and, consequently, their adult outcomes (especially for daughters).

Thus, the main mechanism was likely to be the existence of an imitation process, as daughters who saw their mothers in active employment and economic roles would have been more likely to evolve economically (Duncan et al. 2005; Stinson and Gottschalk 2016). This mechanism would have been essential for individuals with less disadvantaged backgrounds.

Funding

This research was part of the Landskrona Population Study, funded by The Swedish Foundation for Humanities and Social Sciences.

Acknowledgements

I am grateful for the valuable comments and revisions from Martin Dribe, Joana-Maria Pujadas-Mora, Ingrid van Dijk, Elien van Dongen, and Erik Bengtsson.

Supplementary material

Supplementary material is available at European Review of Economic History online.

Footnotes

1

Setting age 23 years as the first income observation is in line with the approaches used in other studies working with similar cohorts and lifetime income measurement in southern Sweden, which uses the same age as the first observation (Lindahl et al. 2015). A primary reasoning for this choice was that the age should be set at a point when most of the children had already have entered the labor market, including those who enrolled in tertiary education.

2

The ratio mean value of quartile 1 was zero, which shows that a considerable proportion of the mothers did not have any income when the children were aged 7–18 years. In addition, the ratio mean value of quartile 4 was 0.99, which shows that most of the mothers could have had a similar income to that of the fathers (see table 1).

3

This figure is based on a model similar to the ones estimated in table 4, controlling from income decile of fathers and children’s sex, with all coefficients significant at 99% (p ≤ 0.01). The specific regression tables are not included.

4

Both figures 4 and 5 are based in similar models to the one presented for sons and daughters together, in table 4. The only difference is that sex is interacted with either the quartiles of childhood income ratio (mother/father) or the quintiles of mother income during childhood (compared with other mothers). The regression estimates are not displayed, but all interactions were statistically significant at least at 95% (p ≤ 0.05).