Regional differences in initial labour market conditions and dynamics in lifetime income trajectories

View author details View Less
  • 1 Umeå University, , Sweden
Full Access
Get eTOC alerts
Rights and permissions Cite this article

We use longitudinal register data from Sweden to study patterns and dynamics in lifetime income trajectories. We examine divergences in these income trajectories by local economic conditions at labour market entry, in combination with other factors such as gender, education level and socio-economic background. We cannot assume that these relationships are constant over the course of individuals’ working lives. Therefore, we use methods from functional data analysis, allowing for a time-varying relationship between income and the explanatory variables. Our results show a large degree of heterogeneity in how lifetime income trajectories develop for different subgroups. We find that, for men, entering the labour market in an urban area is associated with higher cumulative lifetime income, especially later in life. The exception is men with only primary education, for whom those starting their working lives in a large city have lower incomes on average. This divergence increases in size over time. Women who enter into a large urban labour market receive higher lifetime income at all education levels. This relationship is strongest for women with primary education but decreases in strength over time for these women.

Abstract

We use longitudinal register data from Sweden to study patterns and dynamics in lifetime income trajectories. We examine divergences in these income trajectories by local economic conditions at labour market entry, in combination with other factors such as gender, education level and socio-economic background. We cannot assume that these relationships are constant over the course of individuals’ working lives. Therefore, we use methods from functional data analysis, allowing for a time-varying relationship between income and the explanatory variables. Our results show a large degree of heterogeneity in how lifetime income trajectories develop for different subgroups. We find that, for men, entering the labour market in an urban area is associated with higher cumulative lifetime income, especially later in life. The exception is men with only primary education, for whom those starting their working lives in a large city have lower incomes on average. This divergence increases in size over time. Women who enter into a large urban labour market receive higher lifetime income at all education levels. This relationship is strongest for women with primary education but decreases in strength over time for these women.

Introduction

The purpose of this paper is to study lifetime income trajectories in relation to local economic conditions at labour market entry. Several studies in different contexts have suggested substantial associations between economic conditions at labour market entry and subsequent career trajectories, as well as wage levels, at later stages in the working life (Oyer, 2006; Raaum and Røed, 2006; Åslund and Rooth, 2007; Kahn, 2010; Kwon et al, 2010; Oreopoulos et al, 2012; Altonji et al, 2016; Schwandt and von Wachter, 2019). These analyses have focused on the timing of labour market entry, comparing workers who entered into the labour market during periods of recession and high unemployment to those who started out in more favourable conditions.

However, the economic conditions that individuals face on entering the labour market do not only vary over time, but also geographically. Within-country regional income inequality presents a persistent challenge for most developed economies (OECD, 2018; Iammarino et al, 2019). Within the agglomeration economies literature, differences in labour market characteristics such as workforce composition, diffusion of technologies, skill formation and matching efficiency have been suggested as explanations for differences in economic productivity between urban and rural areas (see, for example, Duranton and Puga, 2004 for an overview). Generally speaking, larger and denser urban labour markets are not only characterised by lower overall levels of unemployment than their rural counterparts, but also tend to offer more diversified options for employment, due to higher job matching efficiency (Wheeler, 2001; Wasmer and Zenou, 2002; Abel and Deitz, 2015). There is evidence to suggest that urban labour markets offer greater possibilities for workers to accumulate human capital (Glaeser, 1999; Glaeser and Maré, 2001; De la Roca and Puga, 2017).

Regional differences in labour market characteristics have primarily been used to explain associations between current labour market conditions and wages, such as urban–rural differences in current wages. However, some of the characteristics assumed to differ between urban and rural labour markets – for instance, diverse employment options, job matching efficiency and human capital accumulation – have also been suggested as explanations for the persistent associations between timing of labour market entry and later income development (Kahn, 2010; Oreopoulos et al, 2012). That is to say, there are similarities in the mechanisms proposed to explain associations between geographic location, in terms of labour market size, and current incomes on one hand; and between initial labour market conditions at the time of workforce entry and lifetime income trajectories on the other.

This raises the question whether regional differences in initial labour market conditions could be associated with divergent lifetime income trajectories as well. We explore this question by comparing the cumulative income trajectories of individuals who enter into large urban labour markets to those who enter into medium-sized and rural labour markets, for the 1954 Swedish birth cohort. We thus compare individuals who enter the labour market at roughly the same time, but may face different conditions due to their geographic location and the size of their initial labour market, rather than their timing of labour market entry. The use of large-scale longitudinal data allows us to explore temporal dynamics in their subsequent lifetime income trajectories in detail – that is to say, how patterns of heterogeneity change or persist over different stages of life. Previous research has, for example, found time-varying associations between longitudinal income/​wage trajectories and factors such as gender (Manning and Swaffield, 2008; Boll et al, 2017; Albrecht et al, 2018), education (Tamborini et al, 2015; Bhuller et al, 2017) and parental background (Mayer, 2010). Therefore, we study heterogeneity in these relationships by gender, educational attainment and parents’ socio-economic status. More specifically, we investigate the following questions:

  1. How do lifetime income trajectories develop for individuals entering into differently sized initial labour markets?

  2. In which ways are the above differences in income trajectories mediated by (individual) socio-economic background factors?

  3. Are the relationships in (1) and (2) the same for men and women?

In answering these questions, we contribute to earlier studies in answering these questions by applying a longitudinal perspective on income trajectories spanning a large part of individuals’ life course. We moreover contribute by applying novel methods to large-scale longitudinal register data on annual incomes for the entire Swedish population, which makes it possible to follow individuals’ income trajectories for a large part of their working lives (more than 40 years). We exploit the nature and scale of this data using methods from the field of functional data analysis (FDA) (for example, Ramsay and Silverman, 1997). This choice of method allows us to study the relationships between lifetime income trajectories and covariates dynamically over time and to identify the time periods for which these relationships are statistically significant, while at the same time providing a necessary correction for multiple testing (Abramowicz et al, 2018). To the best of our knowledge, this is also the first study exploring gender differences in lifetime income profiles in a regional context.

Our findings indicate that, overall, entering the labour market in a larger urban, as opposed to rural, area is associated with higher lifetime incomes. However, these differences can be explained in large part by differences in other factors, especially education level. In addition, we find dissimilar patterns in how regional income trajectories develop over time for men and women, as well as at different education levels.

The rest of this paper is structured as follows. In the next section, we present central findings from previous studies as well as an overview of some of the theoretical mechanisms proposed to explain them. We also give a short overview of the macroeconomic developments in Sweden that are relevant for the 1954 birth cohort we study. The subsequent section describes the data and methods in more detail. We then present our results, and conclude.

Background

Previous research

Previous research has consistently found associations between initial labour market conditions, specifically the timing of labour market entry with respect to (national) unemployment rates and economic recessions, and individuals’ wage levels in later career stages. Entering the labour market in times of higher unemployment is generally associated with lower wages not only initially, but also for several years after entry (Åslund and Rooth, 2007; Kwon et al, 2010; Oreopoulos et al, 2012; Altonji et al, 2016; Schwandt and von Wachter, 2019). However, the size and persistence of these associations vary between different studies, as well as for different subgroups. In many cases, the study population consists only of men. However, the studies that do make comparisons by gender generally find no qualitative differences in associations for men and women (Altonji et al, 2016; Schwandt and von Wachter, 2019). These results tend to be stronger for entrants with lower education and, among college graduates, for those with lower predicted earning potential based on field of study, college quality and similar factors (Oreopoulos et al, 2012; Altonji et al, 2016; Schwandt and von Wachter, 2019).

In a Swedish context, Kwon et al (2010) see faster promotion rates to higher ranks, as well as higher wage levels, for workers entering the labour market in periods of economic growth and low unemployment. In contrast to the previously mentioned studies, Åslund and Rooth (2007) focus on immigrants entering the Swedish labour market as adults and employ a regional perspective as well. For this group, too, migrating to Sweden and entering the labour market during an economic crisis is associated with lower wages. These differences are largest during the first four years after labour market entry but persist to a smaller degree for at least for ten years. Åslund and Rooth (2007) find similar patterns, although more persistent, when using local rather than national unemployment levels as an explanatory variable. In a Norwegian context, Raaum and Røed (2006) find that worse economic conditions at the time of (potential) labour market entry are associated with a higher propensity of future unemployment.

The underlying mechanisms proposed by the literature highlight the importance of recession effects on job matching efficiency and the availability of high-quality job opportunities, rather than on short-term wage levels alone. Thus, labour market entrants facing adverse initial conditions might find themselves unemployed or having to accept a low-paying job for a short period of time. They might also have fewer opportunities to switch between jobs and to accumulate human capital in early career stages, leading to persistent negative effects on future career and wage development (Kahn, 2010; Oreopoulos et al, 2012).

As mentioned in the Introduction, much of the literature on regional income differences is concerned with estimating and explaining associations between current labour market size and short-term wage levels. When it comes to such short-term ‘urban wage premiums’ there is, for example, evidence of larger premiums for more highly educated workers (Venables, 2011; Andersson et al, 2014; Carlsen et al, 2016). This literature also finds a large degree of residential self-selection, in that individuals with higher productivity are more likely to locate in larger and more dense labour markets (for example, Combes et al, 2008; Venables, 2011; Ahlin et al, 2018). Bacolod (2017) concludes that even though the gender pay gap is smaller in urban areas in the US, men still benefit more, financially, from working in larger labour markets, compared to women with similar skill sets. Hirsch et al (2013) find a large and persistent urban–rural difference in the gender pay gap in West Germany. The gender gap is smaller in urban labour markets in this study as well. In Norway, the gap between public and private sector wages is larger in cities than in rural areas (Rattsø and Stokke, 2020). Using a longitudinal approach, Bosquet and Overman (2019) find that a large part of the association between current city size and wages can be explained by differences in birthplace size, even more so when taking into account accumulated previous city size. This suggests that the size of early-life locations as well as of the size of earlier labour markets play a role in explaining regional wage differences. While Bosquet and Overman (2019) make use of panel data over 18 years, their approach does not allow for a time-variant relationship between wages and the covariates.

When it comes to associations between incomes and gender, as well as other socio-economic background factors, there is a larger body of research that explicitly investigates how the relationships in question change over time. Albrecht et al (2018) find that the gender wage gap for a subgroup of Swedish university graduates is negligible at labour market entry but increases substantially over time. While the gender differences in wages increase drastically at the time of family formation for individuals who become parents, a large part of the wage gap cannot be accounted for by parenthood effects. Studying a more heterogeneous demographic in the UK, Manning and Swaffield (2008) come to qualitatively similar conclusions on how much of the gender wage gap develops in early career stages. For a subset of West German workers, Kunze (2005) finds a noticeable gender gap in hourly wages already present at labour market entry, and this essentially remains constant over 15 years. In contrast, the results of Boll et al (2017) point to even larger gender differences in accumulated lifetime earnings, which increase over the life course of several German cohorts. Explicitly taking into account divergent patterns of labour market attachment, this cumulative ‘gender lifetime earnings gap’ is up to twice as high as current cross-sectional estimates. When it comes to education, Tamborini et al (2015) find that the economic returns to higher education, in a US context, change over time and are largest in later career stages for both men and women. Bhuller et al (2017) present a similar pattern in the returns to education for Norwegian men. For this group, higher education is associated with higher incomes as well, especially in prime income ages and later career stages. The estimated returns to education with respect to lifetime incomes differ greatly from estimates based on cross-sectional data. Finally, the findings of Mayer (2010) suggest that parental background plays a greater role in determining wages in later stages of the working life.

National labour market context

Our analysis is based on Swedish income data recorded between 1968 and 2010, and we study the cohort born in 1954, as described in more detail in the next section, ‘Data and methods’. We therefore give a brief summary of how macroeconomic conditions and labour market structures have developed over this time period.

Structural change between the 1970s and 2000s included an expansion of private and public services, parallel to a reduction in employment in industrial production and, to a much smaller extent, agriculture. The largest decrease in agricultural employment, as well as the main migration from rural to urban areas had already occurred by the second half of the 20th century in Sweden. Between 1970 and 2010, the population in rural areas decreased much less than in preceding decades, while population numbers in urban areas and cities continued to rise (Statistics Sweden, 1993; 2013; Edvinsson, 2005).

Swedish women have historically had high labour-force participation rates, relative to other industrialised countries. Female labour-force participation rates rose from around 60% to 83% between 1970 and 2011. However, a much larger proportion of women than men work part-time. Swedish women also work in the public sector to a much larger extent, specifically in municipal services, which include healthcare, eldercare and education (Statistics Sweden, 1990; 2012).

In the early 1990s, Sweden experienced an economic crisis that led to drastic increases in unemployment. Between 1990 and 1993, unemployment increased from under 2% to 8% and remained at similar levels until 1997. Employment in all sectors was affected by the crisis, albeit at different stages. The private sector saw a sudden and drastic decrease in employment early in the crisis, but recovered during the second half of the decade, whereas public sector employment decreased more slowly over the entire decade. The effects of this economic crisis were felt by workers of all educational levels and ages, to different degrees and in all parts of the country (Lundborg, 2000).

Data and methods

Data and study population

The data used in this article are made available through the Umeå SIMSAM Lab (Lindgren et al, 2016). Combining information from different registers and databases, this infrastructure contains longitudinal administrative data covering the entire Swedish population over more than four decades. These data consist of a large number of socio-economic, demographic and other types of variables at an individual level, as well as intergenerational links to parents and other family members. The main data sources for our analysis are the income and taxation register (IoT), the integrated database for labour market research (LISA) and the 1970 population and housing census (FoB). These registers are administrated by Statistics Sweden (SCB).

The time period for which different measures are available varies widely, and the income variable covers the years 1968 to 2010 on an annual basis. For this reason, the 1954 birth cohort was chosen, as they were legally allowed to work in 1968 and are expected to still be in the labour market as of 2010, thus providing the longest time period of available data. Our analysis is limited to individuals born in Sweden, since the availability of data on education level and parental background is low for those born outside Sweden. Many immigrants in this age group also migrated much later than 1968, meaning that we cannot accurately compare their cumulative incomes to those of natives, since they received income in Sweden for a shorter time.

The birth cohort studied consists initially of 101,050 individuals. We exclude any individuals for which data on any of the covariates are missing, leaving a total of 79,130 observations. This is mainly due to missing information on parental data and, to a lesser extent, on education level.

Variables

The dependent variable is total earned income (LCI), measured in Swedish kronor (SEK). This measure is available annually for the period 1968–2010. Total earned income is defined as all taxable income except from capital. This covers wages and income from business/self-employment; unemployment, parental and sick leave benefits; and pensions. These different income and benefit components, including pensions, are all based on current or previous income from employment, so that benefits are generally proportional to individual income from employment, but somewhat redistributive because of income caps. Pure transfers and non-taxable benefits such as child allowances, welfare benefits and student loans are not included in our measure of income. Our primary interest lies with income directly related to employment and productivity. The income variable is adjusted for inflation to match the monetary value in 2010 and discounted with a factor of 0.03 (see, for example, Johansson and Kriström, 2016). Discounting is done from the year of income back to the present value at the time of labour force entry. We then calculate the cumulative income over the period 1968–2010. Lastly, this cumulative income is logarithmised. In the remainder of the paper, we denote this logarithmised cumulative income variable LCI.

The main explanatory variable of interest, initial labour market size, is a three-level measure of the size of the local labour market in which an individual received their first income. Local labour markets (LLMs) are functional units defined by patterns of employment and commuting at municipality level (Karlsson and Olsson, 2006; Statistics Sweden, 2010). Each LLM consists of either one municipality that is considered to be a ‘self-sufficient’ labour market, or a group of municipalities with one being considered the economic centre of that LLM. Data on the classification of municipalities into LLMs is only available from 1985 onwards and that data from 1985 is used as a proxy for earlier years. The variable initial labour market size has three levels, adapted from the classification scheme of the Swedish Association of Local Authorities and Regions (Swedish Association of Local Authorities and Regions, 2016). The LLMs of Stockholm, Gothenburg and Malmö are categorised as large cities. LLMs whose central municipality had more than 50,000 residents in 1985 are coded as smaller cities. The remaining LLMs are rural areas.

Education level has three values: primary (and lower secondary) education of seven to nine years; (upper) secondary education of two or three years; and tertiary education of any length, including postgraduate. This variable is available from the LISA database from 1990 to 2010. We use the highest educational attainment in 2010, except for individuals for whom this information is missing in 2010, in which case the latest available entry on educational attainment is used instead.

The variable years of income (before the age of 26) is the number of years an individual had (registered) income by the year 1980. It ranges from 0 (for those who are 26 or older at the time of their first income) to 12 years (for those who start working at the age of 14). This measure can be viewed as a proxy for accumulated work experience at this time point.

The highest educational attainment of parents (HEAP) is defined as the highest education level of the individual’s mother or father. This variable has the same definition of values as education level.

The variable mother’s income is calculated as the average of the mother’s total earned income (in SEK) over the four years 1968–71. This value is then logarithmised. Father’s income is calculated in the same manner.

Gender is not used as an explanatory variable in the regression models. Since we want to investigate if and how the relationships between income trajectories, initial labour market size and the other covariates differ for men and women, we instead perform separate regression analyses by gender.

While income data is available from 1968 onward, and the LCI is calculated starting from that time, the regression analyses described below focus only on the period 1980-2010. During the first years, few individuals actually receive income, and one annual income measurement constitutes a much greater part of a person’s total accumulated income in the beginning. Therefore, coefficient estimates based on this first period would be very volatile, with drastic changes in the estimate from one year to another. These unstable estimates based on a small number of observations do not provide a good representation of the systematic patterns in LCI trajectories we aim to analyze. By 1980, almost all individuals have received their first incomes and LCI increases at a more stable pace. We therefore base the estimation of regression coefficients on the period from 1980 onward.

Descriptive statistics

Table 1 shows descriptive statistics for the covariates, grouped by levels of the variable initial labour market size. There is a slightly larger proportion of men than women among those who entered the labour market in a rural area. Among those who started their careers in large cities, there is a larger proportion with higher education levels, as well as higher HEAP and parental income. We also present the average LCI, grouped by initial labour market size, for a few selected years. Those who begin working in a large city on average receive higher LCI than those who begin in a smaller city or rural area, and these differences increase slightly over time.

Table 1:

Descriptive statistics for the covariates, grouped by levels of initial labour market size

VariableRural area*Smaller cityLarge city
Mean (SD) LCI 198012.80 (0.60)12.80 (0.58)12.81 (0.56)
Mean (SD) LCI 199515.38 (0.48)15.39 (0.50)15.45 (0.53)
Mean (SD) LCI 201016.43 (0.47)16.45 (0.50)16.50 (0.57)
Men52%51%49%
Women48%49%51%
Primary education*22%20%18%
Secondary education48%46%42%
Tertiary education30%34%40%
HEAP: primary*70%62%49%
HEAP: secondary24%29%36%
HEAP: tertiary6%9%15%
Mean (SD) mother’s income, log6.5 (3.8)6.9 (3.7)7.6 (3.5)
Mean (SD) father’s income, log10.1 (0.9)10.2 (0.9)10.4 (0.9)
Mean (SD) years of income before 267.7 (1.7)7.6 (1.7)7.6 (1.8)
n19,51033,55426,106
% of sample25%42%33%

Note: *reference category in regression analyses.

Figure 1 shows the average LCI over time, grouped by gender and initial labour market size, and compared to the grand mean for all individuals. Overall, we can see that men on average have higher LCIs, and women lower, compared to the average for both male and female subjects. These gender differences are larger than the differences between categories of initial labour market size within each group. Men who start working in large and smaller cities initially receive lower average LCI than men who start in a rural area, but this relationship is reversed after around 1985. A similar pattern cannot be seen for women. For both genders, the differences between large and smaller cities are greater than those between smaller cities and rural areas. An additional figure presenting the same grouped means without comparing them to the grand mean can be found in Appendix C.

In the illustration, there are several textboxes listed as follows. Parenting beliefs at 0 years and parent-child relationship at 3 years are at the left. Arrows labeled A2 and A1 from parenting beliefs at 0 years and parenting beliefs at 3 years, respectively point at child vocabulary at three years. Arrows labeled C2 and C1 from parenting beliefs at 0 years and parenting beliefs at 3 years, respectively point at child socio-emotional difficulties at 11 years. An arrow labeled B from child vocabulary at three years points at child socio-emotional difficulties at 11 years.
Figure 1:

Ratios of mean LCI, grouped by gender and initial labour market size, divided by the grand mean for all subjects

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Methods

Even though LCI is observed annually, we can view them as discrete realisations of continuous underlying income functions. Indeed, an individual earns income not only on an annual or monthly basis, but also weekly and hourly. Treating the observed LCIs as observations from a continuous function suggests the use of FDA methods (Ramsay and Silverman, 1997). We here consider a functional-on-scalar linear model, where the response, LCI, is a function of time and the covariates are scalars that are constant over time.

Let , denote the income for individual i at time point and the value of the -th scalar covariate for individual . Then the functional-on-scalar linear regression model can be expressed as:
M1
where are the functional regression parameters to be estimated, and are independent and identically distributed random functions with mean zero and finite total variance. The regression parameters are estimated using ordinary least squares (OLS) pointwise each year by solving:
M2

While this parameter estimation is fairly straightforward, performing repeated inference on these parameters at each observed point in time poses a multiple testing problem. In other words, the family-wise error rate (FWER) increases drastically with the number of time points for which inference on the regression parameters is performed. For many common methods of correcting for multiple testing, however, the power of tests decreases with an increasing number of comparisons (Abramowicz et al, 2018). A solution to this problem is given by the interval-wise testing (IWT) procedure proposed by Pini and Vantini (2017) and adapted to the context of functional-on-scalar linear models by Abramowicz et al (2018). This procedure controls the FWER on intervals, and we describe it here briefly, letting the interested reader refer to the original reference.

For each regression parameter , and each closed interval ℐ є [a, b] on the time domain, the following hypotheses are tested:
M3

The goal is to identify those time periods, that is, those intervals ℐ, for which the regression coefficient is significantly different from zero, at a given significance level α. The hypotheses in (1) are tested using functional permutation tests (Freedman and Lane, 1983; Abramowicz et al, 2018), resulting in a p-value denoted by for each covariate and each interval ℐ. Let be the supremum of over all intervals containing . The intervals in which is significantly different from zero are identified by thresholding at a given level α, for example, 5%. In other words, the covariate is considered to have a significant association with the LCI variable at time point if < 0.05.

Within this setting, separate analyses are performed for men and women. For each separate analysis, the regression models are built by stepwise addition of covariates. In the first step, the model contains the variables initial labour market size, education level and years of income before the age of 26. In step 2, HEAP and mother’s as well as father’s income are added. The full model additionally contains interaction terms between initial labour market size and each of the other covariates.

All analyses were performed in R version 3.5.3 (R Core Team, 2018), using the fdatest package (Pini and Vantini, 2015).

Results

As seen earlier in Figure 1, those who start their careers in larger cities have higher average LCI. For women, these differences are fairly stable over time; for men, starting a career in larger cities yields lower LCI to start with but turns rapidly into a cumulative advantage over time. In the main analysis, we employ stepwise modelling of the functional-on-scalar regression to study how much of these dynamic regional differences can be explained by individual characteristics.

Step 1: individual characteristics

In step 1, LCI is regressed on the variables initial labour market size, education level and years of income. The estimated regression coefficients from this first step are shown in Figure 2 and Figure 3. The solid lines represent the estimated regression coefficients over time, where a grey background indicates that the coefficient is statistically different from zero at a 5% significance level during that time span. The coefficient for smaller city, compared to the reference category rural area, is not statistically significant for men, as seen by the white background of the top left panel in Figure 2. This means that there is no significant difference in LCI between men who enter the labour market in rural areas and those who do so in smaller cities. For women, this coefficient is significant only between 1980 and 1993 (see Figure 2, top right panel) and decreases from around 0.04 towards zero. For men, the estimated coefficient of the category large city, shown in the bottom left panel of Figure 2, is negative at first but then increases over time to around 0.05. It is significant for most of the time span, except around the time when the coefficient changes signs. The corresponding coefficient for women is significant and positive but decreases over time from around 0.1 to 0.04 (see Figure 2, bottom right panel). Thus, after controlling for individual characteristics, there are indications of some income benefits to starting a career in a large city compared to a rural area for both men and women, but for women these benefits seem to be more prominent in the beginning of their working lives.

In the graph, the horizontal axis is on the top and is scaled from 0.1 to 0.9 in increments of 0.1 unit and vertical axis is scaled from negative 0.180 to 0.000 in increments of 0.020 unit. The curves for linear paths C1 and C1’ are straight lines parallel to the vertical axis at regression coefficient B=negative 0.086 and B=0.084, respectively. Data shown by the quantile path C1 is as follows: (0.1, negative 0.010), (0.2, negative 0.022), (0.3, negative 0.040), (0.4, negative 0.045), (0.5, negative 0.65), (0.6, negative 0.090), (0.7, negative 0.120), (0.8, negative 0.140), (0.9, negative 160). Data shown by the quantile path C1’ is as follows: (0.1, negative 0.010), (0.2, negative 0.022), (0.3, negative 0.040), (0.4, negative 0.044), (0.5, negative 0.66), (0.6, negative 0.088), (0.7, negative 0.118), (0.8, negative 0.140), (0.9, negative 160). The quantile curves C1 and C1’ intersect the linear paths C1 and C1’ at (0.55, negative 0.086) and (0.56, negative 0.084), respectively. All values are estimated.
Figure 2:

Estimated regression coefficients (over time) from step 1

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.
In the graph, the horizontal axis is at the center and is scaled from 0.1 to 0.9 in increments of 0.1 unit and vertical axis is scaled from negative 0.060 to 0.080 in increments of 0.020 unit. The graph shows the following data. The curves for linear paths C2 and C2’ are straight lines parallel to the vertical axis at regression coefficient B=0.032 and B=0.010, respectively. Data shown by the quantile path C2 is as follows: (0.1, 0.035), (0.2, 0.037), (0.3, 0.050), (0.4, 0.070), (0.5, 0.030), (0.6, 0.039), (0.7, 0.000), (0.8, negative 0.042), (0.9, negative 0.042). Data shown by the quantile path C2’ is as follows: (0.1, 0.038), (0.2, 0.025), (0.3, 0.020), (0.4, 0.018), (0.5, negative 0.001), (0.6, 0.010), (0.7, negative 0.039), (0.8, negative 0.028), (0.9, negative 0.045). The quantile curves C2 and C2' intersect the linear paths C1 and C1’ at (0.5, 0.032) and (0.44, 0.010), respectively. All values are estimated.
Figure 3:

Estimated regression coefficients (over time) from step 1

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.

The estimated coefficient for secondary education, compared to the reference category primary education, is positive and significant over the entire time period for both men and women (top panels in Figure 3). For men, this coefficient is smaller but increases over time, from 0.04 to 0.1. For women, it decreases from 0.25 to 0.15. The coefficient for tertiary education increases from −0.15 to 0.4 for men and varies slightly around 0.45 for women (see middle panels in Figure 3). Higher educational attainment is thus associated with LCI income as well. There is, however, a notable gender difference, in that the income gains of higher education remain relatively stable over time for women, whereas they increase over the course of the working life for men. As speculation, this difference may reflect differences in sectors and occupations or differences between men and women in labour supply and participation in household production (for instance, Goldin and Katz, 2002; Joshi, 2016). Also, women with different levels of educational attainment may differ in the timing of their investment in education and having children (Rendall et al, 2010; van Rode et al, 2017); this could be a partial explanation of why the coefficient on tertiary education does not increase monotonically over time for women as it does for men. The coefficient estimate for years of income before the age of 26 decreases over time for both groups, from 0.1 to around zero for men and from 0.12 to 0.02 for women (bottom panels in Figure 3). This means that, while those who begin working at an earlier age initially accumulate higher LCI, these differences are less pronounced at later stages of individuals’ working lives.

Step 2: parents’ income and education

When the parental background variables are added to the model in the second step, the coefficient estimates for the other variables remain largely unchanged. All estimated coefficients from this step can be found in Figures A1 and A2 in Appendix A. For both men and women, higher parental incomes are associated with higher LCI. For men, the estimated coefficients of parents’ secondary education and parents’ tertiary education are negative but increase towards zero and are significant only in the first few years. For women, the coefficient of parents’ secondary education decreases over time and is only significant for the first ten years. The coefficient for parents’ tertiary education first increases from around −0.04 to 0.04 and then decreases again to zero and is only significant for a few years. Higher HEAP is thus associated with lower LCI for men in the early career stages, while the opposite is true for women.

Step 3: interactions

The most pronounced changes in the estimated regression coefficients occur when the interaction terms between initial labour market size and each of the other covariates (education level, years of income before the age of 26, HEAP and mother’s/father’s income) are added in step 3. Here we only present figures for the most notable results; all figures for this step can be found in Appendix B (Figures B1B4). Note that the total associations between LCI and, for example, education level are now a combination of the main effects and the interactions with initial labour market size. These total associations are evaluated further in the next subsection.

The estimated coefficients for the main effect of smaller city remain not significant for men and women. The coefficient for large city is now negative and decreasing for men, and significant from year 2000 onward. For women, this coefficient is not significant. For both groups, the main effects of education are still positive and significant, meaning that higher education levels in general are associated with higher LCI (see Appendix B). While the interactions between the category smaller city and the education variables are not significant, the corresponding interactions with large city differ notably for men and women. For men, the interaction between large city and secondary education is significant from 1992 onward and the coefficient increases from 0 to 0.08 (see Figure 4, top left panel). The corresponding coefficient for the interaction with tertiary education increases from −0.02 to 0.1 and is significant from 1994 onward (see Figure 4, bottom left panel). This means that for the reference category (primary education only), entering into a large urban labour market is actually associated with lower LCI in later life. For men with higher education levels, this is negated by the positive interactions between educational background and larger initial labour market size.

Figure 4:
Figure 4:

Estimated regression coefficients (over time) from step 3

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.

For women, only the interaction between large city and tertiary education is significant (between 1980 and 1995) and its coefficient increases from around −0.15 to −0.03 (see Figure 4, bottom right panel). This pattern is opposite from that observed for men.

Divergences in predicted income profiles

To ease interpretation of the results from the models containing interaction coefficients, we visualise the total associations between initial labour market size and income trajectories, for different education levels, in Figure 5. Here, we present the ratios of average predicted cumulative incomes for individuals who enter into an urban, compared to rural, labour market, all other things being equal.

Figure 5:
Figure 5:

Ratios of predicted cumulative income large city/rural area, by education level. Left: men; right: women. Note difference in scale.

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Recall that incomes are discounted to present value at the time of the individuals labour force entry. Using the estimated regression coefficients from the third model specification, let be the predicted LCI at time , for a man who started working in a large city, has tertiary education and tertiary HEAP, and for whom the values of the continuous covariates, years of income before the age of 26, mother’s income and father’s income, are equal to the mean values for men. Let be the predicted LCI at time for a man who started working in a rural area but otherwise has the same covariate values as above. Then, the ratio of predicted cumulative incomes is calculated for each time point as:
M4

The corresponding ratio for women is calculated analogously. We repeat this for the predicted values for individuals with secondary education and primary education, and corresponding HEAP. It should be noted that all coefficient estimates are included regardless of statistical significance.

For men with secondary and tertiary education, entering the labour market in a large city as opposed to a rural area is associated with 3–4% lower cumulative incomes in early years (see Figure 5, left panel). Over time, however, this ratio grows larger than 1, indicating that these groups may benefit later in life from starting out in a larger labour market. This amounts to around 2% higher cumulative income in 2010 for men with secondary education, and 4% higher for men with tertiary education. For men with only primary education, the pattern is reversed. This group initially seems to benefit from entering a large urban labour market, but in later years receives around 3% lower cumulative income than men with primary education who started out in rural labour markets. The ratios for men with primary and tertiary education exhibit a distinct break in their trends between 1990 and 1991. The latter ratio decreases drastically after 1990, from just below 1.02 to around 0.97 in 2010. For men with tertiary education, the urban–rural ratio shows a strong upward trend in the first ten years that ends abruptly in 1990, after which the ratio remains fairly constant around 1.04. The ratio for men with secondary education shows a similar but less pronounced pattern. The timing of this change in trends coincides with the start of the financial and unemployment crisis of the 1990s.

The relationships for women are very different (see Figure 5, right panel). It should be noted that the two different sub-graphs are not on the same scale. Women with primary and secondary education have large initial benefits of entering into an urban, compared to a rural, labour market. These benefits decrease over time, from 20% to 5% for women with primary education and from 14% to 1% for women with secondary education. For women with tertiary education, starting their working lives in a large city is associated with 14% lower cumulative income at first, but the ratio increases over time to a value of around 1.02 in 2010.

In later career stages, women with only primary education thus benefit the most, relatively speaking, from starting out in a large, urban labour market, but the ratios are larger than 1 for women at all education levels in later years. The ratio for women with tertiary education displays a similar break point as the one for men. It increases linearly between 1980 and 1990 and then essentially remains constant slightly above 1 for the remaining years.

Concluding remarks

This study is, to the best of our knowledge, the first to examine the dynamics of lifetime income trajectories in relation to regional differences in initial labour market conditions. This contribution is made possible thanks to large-scale longitudinal register data allowing us to follow one cohort’s annual income through more than 40 years of working life, and the use of state-of-the art methods for functional data. We have seen that, among both men and women, those starting their working lives in a large city generally tend to accumulate higher LCI than those starting in smaller cities or rural areas. However, the observed regional differences can in large part be explained by differences in other covariates, especially education level. This is consistent with the large incidence of residential self-selection of highly educated individuals into urban labour markets observed in the literature on agglomeration economies and urban wage premiums (Combes et al, 2008; Venables, 2011). When the association between LCI and initial labour market size is allowed to interact with the other covariates, we see a large amount of heterogeneity in how income trajectories develop over time, both by gender and education level.

For the women with primary and secondary education in our analyses, starting out in an urban labour market is associated with higher cumulative incomes, especially in earlier years, while the opposite is true for women with tertiary education. These differences in income trajectories converge over time. To the extent that they can be compared, the patterns we observe for women are similar to previous findings on the associations between initial labour market conditions (at a macroeconomic scale) and subsequent career development. Such associations have generally been found to decrease over time and to be stronger for less-educated or more-disadvantaged individuals (Oreopoulos et al, 2012; Altonji et al, 2016; Schwandt and von Wachter, 2019). It should be noted however, that our methodological approach differs greatly from these previous studies.

Our results for the male subset, on the other hand, are more consistent with previous findings regarding urban wage premiums (Venables, 2011; Andersson et al, 2014; Carlsen et al, 2016), although the estimates cannot be directly compared here, either. We find that the associations between initial location and lifetime income is strongest for those with tertiary education. Here, the differences in income trajectories between education levels diverge over time; for men with only primary education, entering an urban labour market (relative to a rural one) is actually associated with lower cumulative income levels later in life. These dynamic relationships could not have been identified from cross-sectional comparisons. For instance, a similar analysis based only on data from 1980 would have vastly underestimated the income benefits from starting out in a large urban labour market for men with secondary and tertiary education. Similarly, an analysis based only on total cumulative incomes in the year 2010 could not have portrayed the strong and divergent temporal dynamics in lifetime income trajectories.

Our results thus indicate considerable differences between men and women, not only in overall income levels but also in longitudinal patterns of lifetime income trajectories, which warrant investigation in future research. Considering the strong long-term trends of urbanisation and increased labour supply of women, the geographical and regional dimensions of gender differences in lifetime income is an under-researched topic. Further research can, for example, provide knowledge on the mechanisms underlying gender differences in the dynamics of income over time, in urban and rural labour markets. The large amount of heterogeneity in the observed patterns, with respect to both gender and education level, means that it is not possible to draw broader conclusions that would be simultaneously valid for all studied subgroups.

In Figure 6 in the previous section, we saw distinct break points in the ratios of predicted cumulative incomes between individuals who start out in urban and rural labour markets. These break points in upward trajectories occurred for women with tertiary education and men of all education levels in the early to mid 1990s. The pattern would suggest that the economic crisis that occurred during this period (Lundborg, 2000) may have had a mitigating effect on regional differences in income trajectories. This finding also deserves the attention of further research.

Our study provides a unique description of the time-varying associations between initial labour market size and lifetime incomes trajectories. As mentioned in Section 1, previous research has mainly focused on the associations between either macro-economic and temporal initial labour market conditions and subsequent income trajectories, or between current labour market size and current wage levels. The drawback of combining these two areas in a novel approach is that there is no pre-existing theoretical work that would be directly applicable to explaining the underlying mechanisms behind our specific results. Therefore, we only provide tentative suggestion for possible explanations for our findings.

We do not include variables measured after labour market entry in our analyses, because doing so would distort the associations between initial labour market conditions and income trajectories. Because we employ function-on-scalar regression, we also do not allow for time-varying covariates in our model. This is not to say that there are no other factors that potentially could be associated with initial labour market size and that affect income trajectories, but which occur after labour market entry and/or change over time (e.g. related to childrearing or later employment).

Individuals in the study may have chosen to migrate to a different labour market region just before or after entering the labour force, and this has possible implications for how to interpret the associations observed. However, we note that fewer than 20% of our sample moved to a labour market region of a different size between labour market entry and age 30. This proportion is smaller at lower education levels. Nevertheless, individuals migrating to, and earning most of their subsequent income in, a differently sized labour market than they started out in would, at worst, lead to an underestimation of the associations between initial labour market size and lifetime income trajectories. When interpreting the results, one should also take into consideration that the costs of living may vary between different regions and labour market sizes.

Conflict of interest

The authors declare that there is no conflict of interest.

References

  • Abel, J.R. and Deitz, R. (2015) Agglomeration and job matching among college graduates, Regional Science and Urban Economics, 51(C): 1420. doi: 10.1016/j.regsciurbeco.2014.12.001

    • Search Google Scholar
    • Export Citation
  • Abramowicz, K., Hager, C.K., Pini, A., Schelin, L., Sjöstedt de Luna, S. and Vantini, S. (2018) Nonparametric inference for Functional-on scalar linear models applied to knee kinematic hop data after injury of the anterior cruciate ligament, Scandinavian Journal of Statistics, 45(4): 103661. doi: 10.1111/sjos.12333

    • Search Google Scholar
    • Export Citation
  • Ahlin, L., Andersson, M. and Thulin, P. (2018) Human capital sorting: the ‘when’ and ‘who’ of the sorting of educated workers to urban regions, Journal of Regional Science, 58(3): 581610. doi: 10.1111/jors.12366

    • Search Google Scholar
    • Export Citation
  • Albrecht, J., Bronson, M., Skogman Thoursie, P. and Vroman, S. (2018) The career dynamics of high-skilled women and men: evidence from Sweden, European Economic Review, 109(09): 83102. doi: 10.1016/j.euroecorev.2018.03.012

    • Search Google Scholar
    • Export Citation
  • Altonji, J.G., Kahn, L.B. and Speer, J.D. (2016) Cashier or consultant? Entry labor market conditions, field of study, and career success, Journal of Labor Economics, 34(S1/Pt2): S361S401. doi: 10.1086/682938

    • Search Google Scholar
    • Export Citation
  • Andersson, M., Klaesson, J. and Larsson, J.P. (2014) The sources of the urban wage premium by worker skills: spatial sorting or agglomeration economies?, Papers in Regional Science, 93(4): 72747. doi: 10.1111/pirs.12025

    • Search Google Scholar
    • Export Citation
  • Åslund, O. and Rooth, D. (2007) Do when and where matter? Initial labour market conditions and immigrant earnings, The Economic Journal, 117(518): 42248.

    • Search Google Scholar
    • Export Citation
  • Bacolod, M. (2017) Skills, the gender wage gap, and cities, Journal of Regional Science, 57(2): 290318. doi: 10.1111/jors.12285

  • Bhuller, M., Mogstad, M. and Salvanes, K. (2017) Life-cycle earnings, education premiums, and internal rates of return, Journal of Labor Economics, 35(4): 9931030. doi: 10.1086/692509

    • Search Google Scholar
    • Export Citation
  • Boll, C., Jahn, M. and Lagemann, A. (2017) The gender lifetime earnings gap: exploring gendered pay from the life course perspective, Journal of Income Distribution, 26(1): 153, https://jid.journals.yorku.ca/index.php/jid/article/view/40355.

    • Search Google Scholar
    • Export Citation
  • Bosquet, C. and Overman, H.G. (2019) Why does birthplace matter so much?, Journal of Urban Economics, 110(0): 2634. doi: 10.1016/j.jue.2019.01.003

    • Search Google Scholar
    • Export Citation
  • Carlsen, F., Rattsø, J. and Stokke, H.E. (2016) Education, experience, and urban wage premium, Regional Science and Urban Economics, 60(0): 3949. doi: 10.1016/j.regsciurbeco.2016.06.006

    • Search Google Scholar
    • Export Citation
  • Combes, P.P., Duranton, G. and Gobillon, L. (2008) Spatial wage disparities: sorting matters!, Journal of Urban Economics, 63(2): 72342. doi: 10.1016/j.jue.2007.04.004

    • Search Google Scholar
    • Export Citation
  • De la Roca, J. and Puga, D. (2017) Learning by working in big cities, Review of Economic Studies, 84(1): 10642. doi: 10.1093/restud/rdw031

    • Search Google Scholar
    • Export Citation
  • Duranton, G. and Puga, D. (2004) Micro-foundations of urban agglomeration economies, in V. Henderson and J.F. Thisse (eds) Handbook of Regional and Urban Economics, Vol 4: Cities and Geography, Amsterdam: Elsevier, pp 2063117.

    • Search Google Scholar
    • Export Citation
  • Edvinsson, R. (2005) Growth, Accumulation, Crisis – with New Macroeconomic Data for Sweden 1800–2000, Stockholm: Almqvist & Wiksell International.

    • Search Google Scholar
    • Export Citation
  • Freedman, D. and Lane, D. (1983) A nonstochastic interpretation of reported significance levels, Journal of Business and Economic Statistics, 1(4): 2928.

    • Search Google Scholar
    • Export Citation
  • Glaeser, E.L. (1999) Learning in cities, Journal of Urban Economics, 46(2): 25477. doi: 10.1006/juec.1998.2121

  • Glaeser, E.L. and Maré, D.C. (2001) Cities and skills, Journal of Labour Economics, 19(2): 31642. doi: 10.1086/319563

  • Goldin, C. and Katz, L.F. (2002) The power of the pill: oral contraceptives and women’s career and marriage decisions, Journal of Political Economy, 110(4): 73070. doi: 10.1086/340778

    • Search Google Scholar
    • Export Citation
  • Hirsch, B., König, M. and Möller, J. (2013) Is there a gap in the gap? Regional differences in the gender pay gap, Scottish Journal of Political Economy, 60(4): 41239. doi: 10.1111/sjpe.12017

    • Search Google Scholar
    • Export Citation
  • Iammarino, S., Rodriguez-Pose, A. and Storper, M. (2019) Regional inequality in Europe: evidence, theory and policy implications, Journal of Economic Geography, 19(2): 27398. doi: 10.1093/jeg/lby021

    • Search Google Scholar
    • Export Citation
  • Johansson, P.O. and Kriström, B. (2016) Cost–Benefit Analysis for Project Appraisal, Cambridge: Cambridge University Press.

  • Joshi, H. (2016) Why do we need longitudinal survey data?, IZA World of Labor, art 308, doi: 10.15185/izawol.308.

  • Kahn, L.B. (2010) The long-term labor market consequences of graduating from college in a bad economy, Labour Economics, 17(2): 30316. doi: 10.1016/j.labeco.2009.09.002

    • Search Google Scholar
    • Export Citation
  • Karlsson, C. and Olsson, M. (2006) The identification of functional regions: theory, methods, and applications, The Annals of Regional Science, 40(1): 118. doi: 10.1007/s00168-005-0019-5

    • Search Google Scholar
    • Export Citation
  • Kunze, A. (2005) The evolution of the gender wage gap, Labour Economics, 12(1): 7397. doi: 10.1016/j.labeco.2004.02.012

  • Kwon, I., Milgrom, E.M. and Hwang, S. (2010) Cohort effects in promotions and wages: evidence from Sweden and the United States, Journal of Human Resources, 45(3): 772808. doi: 10.1353/jhr.2010.0017

    • Search Google Scholar
    • Export Citation
  • Lindgren, U., Nilsson, K., de Luna, X. and Ivarsson, A. (2016) Data resource profile: Swedish microdata research from childhood into lifelong health and welfare (Umeå SIMSAM Lab), International Journal of Epidemiology, 45(4): 10751075g.

    • Search Google Scholar
    • Export Citation
  • Lundborg, P. (2000) Vilka förlorade jobbet under 1990-talet?, in J. Fritzell (ed) Välfärdens Förutsättningar: Arbetsmarknad, Demografi Och Segregation, Vol SOU 2000:37, Stockholm: Fritzes, pp 1150.

    • Search Google Scholar
    • Export Citation
  • Manning, A. and Swaffield, J. (2008) The gender gap in Early-career wage growth, The Economic Journal, 118(530): 9831024. doi: 10.1111/j.1468-0297.2008.02158.x

    • Search Google Scholar
    • Export Citation
  • Mayer, A. (2010) The evolution of wages over the lifecycle: insights from intergenerational connections, Applied Economics, 42(22): 281733. doi: 10.1080/00036840801964682

    • Search Google Scholar
    • Export Citation
  • OECD (2018) Productivity and Jobs in a Globalised World: (How) Can All Regions Benefit?, Paris: OECD Publishing.

  • Oreopoulos, P., von Wachter, T. and Heisz, A. (2012) The short and Long-term career effects of graduating in a recession, American Economic Journal: Applied Economics, 4(1): 129. doi: 10.1257/app.4.1.1

    • Search Google Scholar
    • Export Citation
  • Oyer, P. (2006) Initial labor market conditions and Long-term outcomes for economists, Journal of Economic Perspectives, 20(3): 14360. doi: 10.1257/jep.20.3.143

    • Search Google Scholar
    • Export Citation
  • Pini, A. and Vantini, S. (2015) fdatest: interval testing procedure for functional data, R package version 2.1, https://CRAN.R-project.org/package=fdatest.

    • Search Google Scholar
    • Export Citation
  • Pini, A. and Vantini, S. (2017) Interval-wise testing for functional data, Journal of Nonparametric Statistics, 29(2): 40724. doi: 10.1080/10485252.2017.1306627

    • Search Google Scholar
    • Export Citation
  • R Core Team (2018) R: A Language and Environment for Statistical Computing, Vienna: R Foundation for Statistical Computing, https://www.R-project.org.

    • Search Google Scholar
    • Export Citation
  • Raaum, O. and Røed, K. (2006) Do business cycle conditions at the time of labor market entry affect future employment prospects?, Review of Economics and Statistics, 88(2): 193210. doi: 10.1162/rest.88.2.193

    • Search Google Scholar
    • Export Citation
  • Ramsay, J. and Silverman, B. (1997) Functional Data Analysis, Springer Series in Statistics, New York: Springer.

  • Rattsø, J. and Stokke, H.E. (2020) Private-public wage gap and return to experience: role of geography, gender and education, Regional Science and Urban Economics, 84: art 103571, doi: 10.1016/j.regsciurbeco.2020.103571.

    • Search Google Scholar
    • Export Citation
  • Rendall, M., Aracil, E., Bagavos, C., Couet, C., DeRose, A., DiGiulio, P., Lappegard, T., Robert-Bobée, I., Rønsen, M., Smallwood, S. and Verropoulou, G. (2010) Increasingly heterogeneous ages at first birth by education in Southern European and Anglo-American family-policy regimes: A seven-country comparison by birth cohort, Population Studies, 64(3): 20927. doi: 10.1080/00324728.2010.512392

    • Search Google Scholar
    • Export Citation
  • Schwandt, H. and von Wachter, T. (2019) Unlucky cohorts: estimating the Long-term effects of entering the labor market in a recession in large Cross-sectional data sets, Journal of Labor Economics, 37(S1): S161S198. doi: 10.1086/701046

    • Search Google Scholar
    • Export Citation
  • Statistics Sweden (1990) Women and Men in Sweden: Equality of the Sexes 1990, Women and Men in Sweden, Örebro: Statistics Sweden, https://www.scb.se/hitta-statistik/aldre-statistik/innehall/annan-historiskstatistik/women-and-men-in-sweden-19851998/.

    • Search Google Scholar
    • Export Citation
  • Statistics Sweden (1993) Markanvändningen i Sverige. Andra Utgåvan, Land Use in Sweden, Örebro: Statistics Sweden.

  • Statistics Sweden (2010) Lokala Arbetsmarknader: Egenskaper, Utveckling Och Funktion, Örebro: Statistics Sweden.

  • Statistics Sweden (2012) Women and Men in Sweden 2012: Facts and Figures, Women and Men in Sweden, Örebro: Statistics Sweden, https://www.scb.se/hitta-statistik/aldre-statistik/innehall/annan-historiskstatistik/women-and-men-in-sweden-19851998/.

    • Search Google Scholar
    • Export Citation
  • Statistics Sweden (2013) Land Use in Sweden, 6th edn, Land Use in Sweden, Örebro: Statistics Sweden.

  • Swedish Association of Local Authorities and Regions (2016) Classification of Swedish municipalities 2017, https://skr.se/download/18.4d3d64e3177db55b16631b96/1615474478946/Classification%20of%20Swedish%20Municipalities%202017.pdf

    • Search Google Scholar
    • Export Citation
  • Tamborini, C., Kim, C. and Sakamoto, A. (2015) Education and lifetime earnings in the United States, Demography, 52(4): 1383406. doi: 10.1007/s13524-015-0407-0

    • Search Google Scholar
    • Export Citation
  • van Roode, T., Sharples, K., Dickson, N., Paul, C. (2017) Life-course relationship between socioeconomic circumstances and timing of first birth in a birth cohort, PLoS ONE 12(1). doi: 10.1371/journal.pone.0170170

    • Search Google Scholar
    • Export Citation
  • Venables, A. (2011) Productivity in cities: Self-selection and sorting, Journal of Economic Geography, 11(2): 24151. doi: 10.1093/jeg/lbq040

    • Search Google Scholar
    • Export Citation
  • Wasmer, E. and Zenou, Y. (2002) Does city structure affect job search and welfare?, Journal of Urban Economics, 51(3): 51541. doi: 10.1006/juec.2001.2256

    • Search Google Scholar
    • Export Citation
  • Wheeler, C.H. (2001) Search, sorting, and urban agglomeration, Journal of Labor Economics, 19(4): 87999. doi: 10.1086/322823

Appendix

A. Coefficient estimates from modelling step 2

Figure A1:
Figure A1:

Estimated regression coefficients (over time) from step 2 for men

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.
Figure A2:
Figure A2:

Estimated regression coefficients (over time) from step 2 for women

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.

B. Coefficient estimates from modelling step 3

Figure B1:
Figure B1:

Estimated regression coefficients (over time) from step 3 for men

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.
Figure B2:
Figure B2:

Estimated regression coefficients (over time) from step 3 for men

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.
Figure B3:
Figure B3:

Estimated regression coefficients (over time) from step 3 for women

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.
Figure B4:
Figure B4:

Estimated regression coefficients (over time) from step 3 for women

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

Note: Areas with grey background represent time periods for which the coefficient is significantly different from 0.

C. Descriptives of the mean LCI over time

Figure C1:
Figure C1:

Mean LCI, grouped by gender and initial labour market size

Citation: Longitudinal and Life Course Studies 2022; 10.1332/175795921X16427665823284

  • View in gallery

    Ratios of mean LCI, grouped by gender and initial labour market size, divided by the grand mean for all subjects

  • View in gallery

    Estimated regression coefficients (over time) from step 1

  • View in gallery

    Estimated regression coefficients (over time) from step 1

  • View in gallery

    Estimated regression coefficients (over time) from step 3

  • View in gallery

    Ratios of predicted cumulative income large city/rural area, by education level. Left: men; right: women. Note difference in scale.

  • View in gallery

    Estimated regression coefficients (over time) from step 2 for men

  • View in gallery

    Estimated regression coefficients (over time) from step 2 for women

  • View in gallery

    Estimated regression coefficients (over time) from step 3 for men

  • View in gallery

    Estimated regression coefficients (over time) from step 3 for men

  • View in gallery

    Estimated regression coefficients (over time) from step 3 for women

  • View in gallery

    Estimated regression coefficients (over time) from step 3 for women

  • View in gallery

    Mean LCI, grouped by gender and initial labour market size

  • Abel, J.R. and Deitz, R. (2015) Agglomeration and job matching among college graduates, Regional Science and Urban Economics, 51(C): 1420. doi: 10.1016/j.regsciurbeco.2014.12.001

    • Search Google Scholar
    • Export Citation
  • Abramowicz, K., Hager, C.K., Pini, A., Schelin, L., Sjöstedt de Luna, S. and Vantini, S. (2018) Nonparametric inference for Functional-on scalar linear models applied to knee kinematic hop data after injury of the anterior cruciate ligament, Scandinavian Journal of Statistics, 45(4): 103661. doi: 10.1111/sjos.12333

    • Search Google Scholar
    • Export Citation
  • Ahlin, L., Andersson, M. and Thulin, P. (2018) Human capital sorting: the ‘when’ and ‘who’ of the sorting of educated workers to urban regions, Journal of Regional Science, 58(3): 581610. doi: 10.1111/jors.12366

    • Search Google Scholar
    • Export Citation
  • Albrecht, J., Bronson, M., Skogman Thoursie, P. and Vroman, S. (2018) The career dynamics of high-skilled women and men: evidence from Sweden, European Economic Review, 109(09): 83102. doi: 10.1016/j.euroecorev.2018.03.012

    • Search Google Scholar
    • Export Citation
  • Altonji, J.G., Kahn, L.B. and Speer, J.D. (2016) Cashier or consultant? Entry labor market conditions, field of study, and career success, Journal of Labor Economics, 34(S1/Pt2): S361S401. doi: 10.1086/682938

    • Search Google Scholar
    • Export Citation
  • Andersson, M., Klaesson, J. and Larsson, J.P. (2014) The sources of the urban wage premium by worker skills: spatial sorting or agglomeration economies?, Papers in Regional Science, 93(4): 72747. doi: 10.1111/pirs.12025

    • Search Google Scholar
    • Export Citation
  • Åslund, O. and Rooth, D. (2007) Do when and where matter? Initial labour market conditions and immigrant earnings, The Economic Journal, 117(518): 42248.

    • Search Google Scholar
    • Export Citation
  • Bacolod, M. (2017) Skills, the gender wage gap, and cities, Journal of Regional Science, 57(2): 290318. doi: 10.1111/jors.12285

  • Bhuller, M., Mogstad, M. and Salvanes, K. (2017) Life-cycle earnings, education premiums, and internal rates of return, Journal of Labor Economics, 35(4): 9931030. doi: 10.1086/692509

    • Search Google Scholar
    • Export Citation
  • Boll, C., Jahn, M. and Lagemann, A. (2017) The gender lifetime earnings gap: exploring gendered pay from the life course perspective, Journal of Income Distribution, 26(1): 153, https://jid.journals.yorku.ca/index.php/jid/article/view/40355.

    • Search Google Scholar
    • Export Citation
  • Bosquet, C. and Overman, H.G. (2019) Why does birthplace matter so much?, Journal of Urban Economics, 110(0): 2634. doi: 10.1016/j.jue.2019.01.003

    • Search Google Scholar
    • Export Citation
  • Carlsen, F., Rattsø, J. and Stokke, H.E. (2016) Education, experience, and urban wage premium, Regional Science and Urban Economics, 60(0): 3949. doi: 10.1016/j.regsciurbeco.2016.06.006

    • Search Google Scholar
    • Export Citation
  • Combes, P.P., Duranton, G. and Gobillon, L. (2008) Spatial wage disparities: sorting matters!, Journal of Urban Economics, 63(2): 72342. doi: 10.1016/j.jue.2007.04.004

    • Search Google Scholar
    • Export Citation
  • De la Roca, J. and Puga, D. (2017) Learning by working in big cities, Review of Economic Studies, 84(1): 10642. doi: 10.1093/restud/rdw031

    • Search Google Scholar
    • Export Citation
  • Duranton, G. and Puga, D. (2004) Micro-foundations of urban agglomeration economies, in V. Henderson and J.F. Thisse (eds) Handbook of Regional and Urban Economics, Vol 4: Cities and Geography, Amsterdam: Elsevier, pp 2063117.

    • Search Google Scholar
    • Export Citation
  • Edvinsson, R. (2005) Growth, Accumulation, Crisis – with New Macroeconomic Data for Sweden 1800–2000, Stockholm: Almqvist & Wiksell International.

    • Search Google Scholar
    • Export Citation
  • Freedman, D. and Lane, D. (1983) A nonstochastic interpretation of reported significance levels, Journal of Business and Economic Statistics, 1(4): 2928.

    • Search Google Scholar
    • Export Citation
  • Glaeser, E.L. (1999) Learning in cities, Journal of Urban Economics, 46(2): 25477. doi: 10.1006/juec.1998.2121

  • Glaeser, E.L. and Maré, D.C. (2001) Cities and skills, Journal of Labour Economics, 19(2): 31642. doi: 10.1086/319563

  • Goldin, C. and Katz, L.F. (2002) The power of the pill: oral contraceptives and women’s career and marriage decisions, Journal of Political Economy, 110(4): 73070. doi: 10.1086/340778

    • Search Google Scholar
    • Export Citation
  • Hirsch, B., König, M. and Möller, J. (2013) Is there a gap in the gap? Regional differences in the gender pay gap, Scottish Journal of Political Economy, 60(4): 41239. doi: 10.1111/sjpe.12017

    • Search Google Scholar
    • Export Citation
  • Iammarino, S., Rodriguez-Pose, A. and Storper, M. (2019) Regional inequality in Europe: evidence, theory and policy implications, Journal of Economic Geography, 19(2): 27398. doi: 10.1093/jeg/lby021

    • Search Google Scholar
    • Export Citation
  • Johansson, P.O. and Kriström, B. (2016) Cost–Benefit Analysis for Project Appraisal, Cambridge: Cambridge University Press.

  • Joshi, H. (2016) Why do we need longitudinal survey data?, IZA World of Labor, art 308, doi: 10.15185/izawol.308.

  • Kahn, L.B. (2010) The long-term labor market consequences of graduating from college in a bad economy, Labour Economics, 17(2): 30316. doi: 10.1016/j.labeco.2009.09.002

    • Search Google Scholar
    • Export Citation
  • Karlsson, C. and Olsson, M. (2006) The identification of functional regions: theory, methods, and applications, The Annals of Regional Science, 40(1): 118. doi: 10.1007/s00168-005-0019-5

    • Search Google Scholar
    • Export Citation
  • Kunze, A. (2005) The evolution of the gender wage gap, Labour Economics, 12(1): 7397. doi: 10.1016/j.labeco.2004.02.012

  • Kwon, I., Milgrom, E.M. and Hwang, S. (2010) Cohort effects in promotions and wages: evidence from Sweden and the United States, Journal of Human Resources, 45(3): 772808. doi: 10.1353/jhr.2010.0017

    • Search Google Scholar
    • Export Citation
  • Lindgren, U., Nilsson, K., de Luna, X. and Ivarsson, A. (2016) Data resource profile: Swedish microdata research from childhood into lifelong health and welfare (Umeå SIMSAM Lab), International Journal of Epidemiology, 45(4): 10751075g.

    • Search Google Scholar
    • Export Citation
  • Lundborg, P. (2000) Vilka förlorade jobbet under 1990-talet?, in J. Fritzell (ed) Välfärdens Förutsättningar: Arbetsmarknad, Demografi Och Segregation, Vol SOU 2000:37, Stockholm: Fritzes, pp 1150.

    • Search Google Scholar
    • Export Citation
  • Manning, A. and Swaffield, J. (2008) The gender gap in Early-career wage growth, The Economic Journal, 118(530): 9831024. doi: 10.1111/j.1468-0297.2008.02158.x

    • Search Google Scholar
    • Export Citation
  • Mayer, A. (2010) The evolution of wages over the lifecycle: insights from intergenerational connections, Applied Economics, 42(22): 281733. doi: 10.1080/00036840801964682

    • Search Google Scholar
    • Export Citation
  • OECD (2018) Productivity and Jobs in a Globalised World: (How) Can All Regions Benefit?, Paris: OECD Publishing.

  • Oreopoulos, P., von Wachter, T. and Heisz, A. (2012) The short and Long-term career effects of graduating in a recession, American Economic Journal: Applied Economics, 4(1): 129. doi: 10.1257/app.4.1.1

    • Search Google Scholar
    • Export Citation
  • Oyer, P. (2006) Initial labor market conditions and Long-term outcomes for economists, Journal of Economic Perspectives, 20(3): 14360. doi: 10.1257/jep.20.3.143

    • Search Google Scholar
    • Export Citation
  • Pini, A. and Vantini, S. (2015) fdatest: interval testing procedure for functional data, R package version 2.1, https://CRAN.R-project.org/package=fdatest.

    • Search Google Scholar
    • Export Citation
  • Pini, A. and Vantini, S. (2017) Interval-wise testing for functional data, Journal of Nonparametric Statistics, 29(2): 40724. doi: 10.1080/10485252.2017.1306627

    • Search Google Scholar
    • Export Citation
  • R Core Team (2018) R: A Language and Environment for Statistical Computing, Vienna: R Foundation for Statistical Computing, https://www.R-project.org.

    • Search Google Scholar
    • Export Citation
  • Raaum, O. and Røed, K. (2006) Do business cycle conditions at the time of labor market entry affect future employment prospects?, Review of Economics and Statistics, 88(2): 193210. doi: 10.1162/rest.88.2.193

    • Search Google Scholar
    • Export Citation
  • Ramsay, J. and Silverman, B. (1997) Functional Data Analysis, Springer Series in Statistics, New York: Springer.

  • Rattsø, J. and Stokke, H.E. (2020) Private-public wage gap and return to experience: role of geography, gender and education, Regional Science and Urban Economics, 84: art 103571, doi: 10.1016/j.regsciurbeco.2020.103571.

    • Search Google Scholar
    • Export Citation
  • Rendall, M., Aracil, E., Bagavos, C., Couet, C., DeRose, A., DiGiulio, P., Lappegard, T., Robert-Bobée, I., Rønsen, M., Smallwood, S. and Verropoulou, G. (2010) Increasingly heterogeneous ages at first birth by education in Southern European and Anglo-American family-policy regimes: A seven-country comparison by birth cohort, Population Studies, 64(3): 20927. doi: 10.1080/00324728.2010.512392

    • Search Google Scholar
    • Export Citation
  • Schwandt, H. and von Wachter, T. (2019) Unlucky cohorts: estimating the Long-term effects of entering the labor market in a recession in large Cross-sectional data sets, Journal of Labor Economics, 37(S1): S161S198. doi: 10.1086/701046

    • Search Google Scholar
    • Export Citation
  • Statistics Sweden (1990) Women and Men in Sweden: Equality of the Sexes 1990, Women and Men in Sweden, Örebro: Statistics Sweden, https://www.scb.se/hitta-statistik/aldre-statistik/innehall/annan-historiskstatistik/women-and-men-in-sweden-19851998/.

    • Search Google Scholar
    • Export Citation
  • Statistics Sweden (1993) Markanvändningen i Sverige. Andra Utgåvan, Land Use in Sweden, Örebro: Statistics Sweden.

  • Statistics Sweden (2010) Lokala Arbetsmarknader: Egenskaper, Utveckling Och Funktion, Örebro: Statistics Sweden.

  • Statistics Sweden (2012) Women and Men in Sweden 2012: Facts and Figures, Women and Men in Sweden, Örebro: Statistics Sweden, https://www.scb.se/hitta-statistik/aldre-statistik/innehall/annan-historiskstatistik/women-and-men-in-sweden-19851998/.

    • Search Google Scholar
    • Export Citation
  • Statistics Sweden (2013) Land Use in Sweden, 6th edn, Land Use in Sweden, Örebro: Statistics Sweden.

  • Swedish Association of Local Authorities and Regions (2016) Classification of Swedish municipalities 2017, https://skr.se/download/18.4d3d64e3177db55b16631b96/1615474478946/Classification%20of%20Swedish%20Municipalities%202017.pdf

    • Search Google Scholar
    • Export Citation
  • Tamborini, C., Kim, C. and Sakamoto, A. (2015) Education and lifetime earnings in the United States, Demography, 52(4): 1383406. doi: 10.1007/s13524-015-0407-0

    • Search Google Scholar
    • Export Citation
  • van Roode, T., Sharples, K., Dickson, N., Paul, C. (2017) Life-course relationship between socioeconomic circumstances and timing of first birth in a birth cohort, PLoS ONE 12(1). doi: 10.1371/journal.pone.0170170

    • Search Google Scholar
    • Export Citation
  • Venables, A. (2011) Productivity in cities: Self-selection and sorting, Journal of Economic Geography, 11(2): 24151. doi: 10.1093/jeg/lbq040

    • Search Google Scholar
    • Export Citation
  • Wasmer, E. and Zenou, Y. (2002) Does city structure affect job search and welfare?, Journal of Urban Economics, 51(3): 51541. doi: 10.1006/juec.2001.2256

    • Search Google Scholar
    • Export Citation
  • Wheeler, C.H. (2001) Search, sorting, and urban agglomeration, Journal of Labor Economics, 19(4): 87999. doi: 10.1086/322823

Content Metrics

May 2022 onwards Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 27 27 27
PDF Downloads 11 11 11

Altmetrics

Dimensions