Earnings, labor market dynamics, and inequality in Sweden

• Published date: 01 July 2024
• Author: Johan Holmberg
• Date: 01 July 2024
• DOI: http://doi.org/10.1111/sjoe.12553

Scand. J. of Economics 126(3), 561–599, 2024

DOI: 10.1111/sjoe.12553

Earnings, labor market dynamics, and

inequality in Sweden

Johan Holmberg

Ume˚

a University, SE-901 87 Ume˚

a, Sweden

johan.holmberg@umu.se

Abstract

In this paper, we develop a comprehensive model of earnings and labor market dynamics,

where employment and job change are endogenous. The model is estimated by applying the

method of indirect inference on Swedish register data, and then used to carry out some policy

experiments. There are three key conclusions from these experiments. First, employment shocks

early in life can, to a larger extent, be mitigated before retirement compared with employment

shocks occurring later. Second, we f‌ind that idiosyncratic productivity shocks, unobserved

heterogeneity, and education contribute substantially to life-cycle earnings inequality. Finally,

we f‌ind that transitory shocks to employment risk have negative effects on earnings and

employment in the short run but may increase labor market f‌luidity in the medium run.

Keywords: Earnings dynamics; employment; inequality; social insurance

JEL classif‌ication:D63; E27; H55; J24; J64

1. Introduction

Earnings dynamics and labor market transitions are intertwined, and it

is essential to model these processes jointly to capture the effects of

interdependence. Indeed, unemployment shocks have large and long-lasting

effects on future wages and earnings (e.g., Topel, 1990; Jacobson et al., 1993;

Davis and Von Wachter, 2011), and job changes are essential for career

development and wage formation (e.g., Topel and Ward, 1992; Bagger

et al., 2014). In this paper, we develop a model of earnings dynamics with

labor market transitions, unobserved heterogeneity, and state dependence,

which builds on and extends the model of Altonji et al. (2013) by using register

data, and importantly adding further emphasis on the roles of within-period

unemployment, parental leave, and migration. For example, we f‌ind that short

unemployment spells propagate job changes. This suggests that the reason for

changing employers is a vital mechanism for understanding the links between

labor market f‌luidity and career development.

We conduct the analysis using Swedish register data including

socioeconomic and geographical information for the Swedish population born

1936–1997 and observed 2002–2015. Register data have many advantages

over survey data, the information does not rely on the recollection of the

c

2023 The Authors. The Scandinavian Journal of Economics published by John Wiley & Sons Ltd on behalf of F¨

oreningen

f¨

or utgivande av the SJE.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License,

which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial

and no modif‌ications or adaptations are made.

562 Earnings, labor market dynamics, and inequality in Sweden

subjects, and there is no potentially systematic attrition from the data, apart

from emigration or death. We focus on individual labor market outcomes

as many aspects of the labor market may be conf‌lated in an analysis of the

aggregates alone.1

The estimated model is used to analyze several key questions related to

inequality and employment. First, we use the model to identify the underlying

sources of lifetime earnings inequality. The results show that a large fraction

of the observed life-cycle earnings inequality can be attributed to idiosyncratic

productivity shocks, unobserved heterogeneity, and education. We estimate

that for men, idiosyncratic productivity shocks can explain 22.0 percent of

overall earning inequality while unobserved ability contributes to 21.6 percent

of the lifetime earnings differences for women. Second, we examine the

aggregate effects of temporary shocks to employment on future earnings,

employment, and job change rates. We f‌ind that most of the effects of the

shocks decay after two years of recovery, but some adverse effects on labor

market outcomes linger for more than ten years. The shocks also increase the

job change rates in the medium run as displaced workers climb back up the

job ladder (which is in line with Burdett and Mortensen, 1998).

Finally, we examine how non-employment at different ages affects the

accumulation of pension entitlements. We f‌ind that non-employment early in

the working life reduces life-cycle earnings and pension accumulations to a

lesser extent than non-employment spells during the middle ages.

Altonji et al. (2013) constructed a semi-structural2model of earnings

dynamics3with endogenous job mobility and employment, which is estimated

1For example, if transition probabilities vary across the earnings distribution, this will affect

the distributional effects of macroeconomic shocks. If one considers an aggregate shock to

employment, the pool of unemployed is over time increasingly represented by individuals

who are more prone to stay in unemployment, even within groups with the same observable

characteristics (Ahn and Hamilton, 2020). In addition, heterogeneity in terms of marginal

propensities to consume affects the extent to which income and wealth shocks translate into

changes in expenditure. Heterogeneity is important for understanding how income shocks (e.g.,

Alan et al., 2018) and wealth destruction (e.g., Kaplan and Violante, 2018) affect the aggregate

demand. Even when individual heterogeneity is constant over time, it still affects the current

state of the system, making the responses in terms of aggregated statistics sensitive to the timing

of a shock.

2A semi-structural model can be viewed as a model that identif‌ies some of the mechanisms

and parameters of an underlying structural model without including fully specif‌ied behavioral

components.

3There is a large body of literature on earnings dynamics based on the permanent income

hypothesis of Friedman and Kuznets (1954) and Friedman (1957) where earnings inequality is

decomposed into permanent differences and transitory shocks. Important contributions to this

literature include Lillard and Willis (1978,1979), Hause (1980), MaCurdy (1982), Moff‌itt and

Gottschalk (1995), and Guvenen (2007). Gustavsson (2008), Gustavsson and ¨

Osterholm (2014),

and Gustafsson and Holmberg (2019) provide some recent examples of earnings dynamics

c

2023 The Authors. The Scandinavian Journal of Economics published by John Wiley & Sons Ltd on behalf of F¨

oreningen

f¨

or utgivande av the SJE.

J. Holmberg 563

on PSID data using indirect inference. Their modeling framework builds on

that of Low et al. (2010), who analyzed wage dynamics by combining an

earnings dynamics model with a model of endogenous job mobility and

employment based on a search-theoretical framework.4These studies go

beyond the traditional earnings dynamics literature by modeling some of

the underlying sources driving the observed earnings inequality. Altonji

et al. (2013) used their model to decompose the sources of life-cycle earnings

inequality to measure the contributions of the f‌irm match, unemployment

shocks, and job changes. They found that job-specif‌ic heterogeneity in

wages contributes to 34 percent of the variance in life-cycle earnings. Low

et al. (2010) found that the size of permanent shocks becomes smaller when

allowing for endogenous transitions, and that wage shocks are, to a large

extent, a result of individuals changing jobs to f‌irms offering better matches.

Other contributions to this emerging literature can be found in Friedrich

et al. (2019), Sauer and Taber (2021), and Card and Hyslop (2021). Sauer and

Taber (2021) present a model of earnings dynamics including endogenous

fertility decisions and human capital formation. They estimate the model

using US data for female workers and f‌ind that women’s wages would be

3 percent higher if there were no relationship between fertility and labor

market participation. Card and Hyslop (2021) analyze how women’s earnings

inequality has changed across cohorts active in the US labor market between

1967 and 1997. They f‌ind that around 87 percent of the observed fall in earnings

inequality for the younger cohorts can be explained by a reduced effect from

family-related variables. Friedrich et al. (2019) estimate an earnings dynamics

model with endogenous labor market transitions using Swedish register data.

They study how f‌irm-level productivity shocks affect earnings dynamics

when participation is endogenous and f‌ind that f‌irm heterogeneity contributes

substantially to wage inequality among college-educated workers. Compared

with Friedrich et al. (2019), the model developed in this paper has a more

comprehensive specif‌ication of the labor market transitions that considers

state dependence, unobserved heterogeneity, family formation, and migration.

We also use our model to address key questions such as the labor market

impacts of employment shocks and how non-employment at different stages

during the working life affects the accumulation of pension entitlements.

The present paper contributes to the literature in several dimensions. First,

with few exceptions, earlier comparable studies have focused on earnings

literature using Swedish data. Yet, none of these studies models earnings dynamics and labor

market transitions simultaneously, which is the key issue.

4For search-theoretical models, see among others Mortensen (1977), Pissarides (1985),

Mortensen and Pissarides (1994), and Burdett and Mortensen (1998). Lise et al. (2016),

Jung and Kuhn (2019), and Merkurieva (2019) provide examples of recent applications of

structural search-theoretical models.

c

2023 The Authors. The Scandinavian Journal of Economics published by John Wiley & Sons Ltd on behalf of F¨

oreningen

f¨

or utgivande av the SJE.