+ All Categories
Transcript

Firms and the Intergenerational Transmission ofLabor Market Advantage

Per EngzellUniversity of Oxford

Nathan WilmersMIT Sloan

December 2021

Abstract

Recent research finds that pay inequality stems both from firm pay-setting andfrom workers’ individual characteristics. Yet, intergenerational mobility researchremains focused on transmission of individual traits, and has failed to test howfirms shape the inheritance of inequality. We study this question using threedecades of Swedish population register data, and decompose the intergenerationalearnings correlation into firm pay premiums and stable worker effects. One quar-ter of the intergenerational earnings correlation at midlife is explained by sortingbetween firms with unequal pay. Employer or industry inheritance account for asurprisingly small share of this firm-based earnings transmission. Instead, chil-dren from high-income backgrounds benefit from matching with high-paying firmsirrespective of the sources of parents’ earnings advantage. Our analysis revealshow an imperfectly competitive labor market provides an opening for skill-basedrewards in one generation to become class-based advantages in the next.

1

1 IntroductionThe vaunted ideal of equality of opportunity often comes down to the banalities ofa job application, an interview, an offer. Opportunity is celebrated when a troubledkid gets his foot in the door at a unionized hospital (Anderson, 2000), or when adisadvantaged worker’s score on a civil service exam lands a coveted public sector job(Newman, 2009). Opportunity is undermined by a manager who recommends his ownson for a plum starting position (Staiger, 2021), or by a white-shoe investment bankthat screens candidates on experience with lacrosse and squash (Rivera, 2012). Equalopportunity, and its betrayal, seems to depend on how workers sort across employers.

Yet, a half century of quantitative research on intergenerational mobility has largelyignored the role of employers. Instead, researchers argue that social origin influencesopportunity via lasting effects on individual worker characteristics (Becker and Tomes,1986; Bowles et al., 2009; Jæger and Breen, 2016; Jencks et al., 1979; Morgan et al.,2006; Sewell et al., 1969). Parents with greater resources are able to invest in theirchildren in numerous ways, and cultivate traits in them such as skill, agreeableness,or work orientation, that are valued in the labor market (Bowles and Gintis, 1976;Farkas, 2003; Lareau, 2011). More recently, sociologists have turned their attentionto occupation-specific knowledge gained through informal training at the dinner table(Jonsson et al., 2009; Weeden and Grusky, 2012).

Indeed, these findings challenge easy causal interpretations of the apparent role ofemployers in the examples above. Perhaps screening on lacrosse is just a revealingway to learn an applicant’s underlying scholastic aptitude. Perhaps a father’s recom-mendation testifies to his son’s long, informal tutelage at the father’s expert hand.Determining whether and how firms matter for intergenerational mobility requires dis-tinguishing class differences in human capital that are rewarded by a competitive labormarket, from class differences in social capital and cultural matching that shunt privi-leged workers into high-paying jobs.

In this article, we take up this challenge in two ways. First, we decompose inter-generational earnings transmission into persistent worker earnings differences, mainlyattributable to human capital, relative to firm sorting advantages, mainly attributableto class-based social capital. To do so, we bring methods from research on labor marketinequality to the study of intergenerational mobility. This approach operationalizes theidea that employer-side pay setting is a crucial determinant of earnings (Card et al.,2013; Song et al., 2018; Tomaskovic-Devey et al., 2020). The key insight here is thatwhile worker-specific advantages like human capital are persistently valuable, the pay-off to class-based social capital is typically contextually specific. Second, we provide ageneral framework clarifying the multiple ways that class background can affect sortinginto high-paying firms.

Empirically, we draw on population-wide linked employer-employee data from Swe-den, which offer a sufficiently long time span to observe both parents and childrenat prime working age. In a first step, we distinguish the components of earnings at-tributable to two sources: (a) the firm at which one works and (b) durable individual

2

worker traits. The firm component is the premium or penalty associated with workingat a given firm, conditional on the composition of its workers. The individual earn-ings component reflects human capital and other traits that are consistently rewardedacross firms. We apply this distinction to both parents and children. We focus our mainanalysis on labor earnings—which have the clearest link to firm pay practices—and onfathers as the main breadwinner in the parental generation.

We then decompose the intergenerational rank correlation in earnings into fourpathways, the first three of which are related to firm sorting: correlation between (i)fathers’ and children’s firm components of earnings; (ii) fathers’ worker component andchildren’s firm component; (iii) fathers’ firm component and children’s worker compo-nent; and (iv) fathers’ and children’s worker components of earnings. The first twoof these pathways are of particular interest, as they capture the role of firm sortingin children’s economic attainment. The firm-to-firm path (i) subsumes mechanismssuch as the inheritance of employers (Corak and Piraino, 2011) and firm-associated,occupation-specific microclass capital (Jonsson et al., 2009). The worker-to-firm path(ii) reflects other ways that having high-earning parents can scaffold children’s la-bor market attainment—including the wider exercise of social capital (Armstrong andHamilton, 2013) or employer-side discrimination on class background (Friedman andLaurison, 2020; Rivera, 2012).

Our firm mechanism-focused approach uncovers several new facts about the role ofemployers in intergenerational earnings persistence. Firm sorting on social backgroundis evident at labor market entry and persists throughout the career. The employerpremium among children from privileged backgrounds is constant with age, but ac-counts for a shrinking fraction of total earnings transmission as individual earnings fanout over the career. Yet, even in mature earnings, firm-based mechanisms account fora full quarter of intergenerational persistence. Surprisingly, direct employer and in-dustry inheritance from a parent are largely unimportant for this firm-based earningstransmission. Instead, children from high-income backgrounds benefit from matchingwith high-paying employers irrespective of the sources of parents’ earnings advantage.Finally, we show that even among high and low social background workers who manageto sort to a high-paying employer, their employers are different. High-paying firms thatdisproportionately employ privileged workers tend to be in finance and business, and inthe capital city, while those that provide high-paying opportunities for low backgroundworkers are concentrated in manufacturing and energy extraction.

2 Intergenerational mobility and firm pay premiumsWhy do class advantages persist from one generation to the next? Canonical stratifica-tion research emphasizes the influence of socialization on individual traits (Bourdieu,1984; Bowles and Gintis, 1976; Farkas, 2003; Jæger and Breen, 2016; Jencks et al.,1979). While early models of status attainment posited the direct involvement ofparents in finding a job (Blau and Duncan, 1967), this focus was soon replaced by

3

psychological influences on ability, aspirations and, ultimately, schooling (Sewell et al.,1969; Sewell and Hauser, 1975). The resulting model faced repeated criticism for itsone-sided focus on the individual and failure to consider wider labor market dynamics(Baron, 1984; Baron and Bielby, 1980; Featherman, 1981; Spilerman, 1977; Stolzen-berg, 1978). Yet, a compelling alternative failed to emerge and the status attainmentmodel remains, implicitly or explicitly, the primary backdrop for stratification research.

By and large, then, the field today remains preoccupied with education and otherforms of human capital (Bloome et al., 2018; Bol et al., 2019; Breen and Müller, 2020;Horowitz, 2018). This is not without reason, as education is both the main conduit ofsocial reproduction and a vehicle for intergenerational mobility (Hout, 2012; Torche,2015).

However, recent research suggests that human capital differences, even broadly con-strued, are not the only sources of advantage for privileged children. Among managersand professionals, high-class children end up higher-paid, in part by sorting to largerand more centrally located firms (Laurison and Friedman, 2016). Rivera (2012) pro-vides vivid evidence that this sorting may not just be due to objective evidence ofhuman capital—like grades or ability—but instead due to cultural matching betweenhigh-class applicants and their interviewers. A more direct link comes from parentsreferring or hiring their children: evidence from the US and Canada demonstrates thatchildren often start their careers working directly for their parents’ employers (Corakand Piraino, 2011; Staiger, 2021). In all of these cases, researchers attribute social ad-vantage to some form of social capital or class preference—whether proxied by culturalfit or directly exercised as employment referral—that distorts the labor market sortingprocess.

Consistent with the importance of these mechanisms, researchers studying intergen-erational mobility increasingly argue that human capital—particularly measured nar-rowly as formal educational attainment—is insufficient to account for the persistenceof economic status across generations (Bernardi and Ballarino, 2016; Torche, 2011;Witteveen and Attewell, 2017, 2020). Indeed, while education today accounts for thebulk of occupational status transmission in the US, it only accounts for about half ofthe intergenerational persistence in family income (Torche, 2016). Even the classicalfinding that high-skilled labor markets operate more meritocratically is weaker whenincome is studied, and when accounting for selection processes into education (Fiel,2020; Zhou, 2019).

This recent work leaves several open questions. First, why, in an open labor market,does social background still matter over and above education? Prior work on socialstratification has mainly considered other, unobserved forms of human capital such ascognitive and non-cognitive skill. But, research on labor market inequality identifiesconsistent differences in pay between firms, even controlling for the composition oftheir workers (Abowd et al., 1999; Card et al., 2013; Song et al., 2018). The source ofthese premiums varies. In some cases, they reflect labor market institutions like laborunions imposing non-market pay levels (Tomaskovic-Devey et al., 2020). In others,they result from complementarities between workers that cannot be reduced to any

4

individual worker’s human capital (Song et al., 2018). Regardless of their source, socialadvantages in sorting to high-premium firms provide one explanation for non-humancapital-based earnings differences between workers. As such, we study the role of firmpay premiums in intergenerational earnings mobility.

Importantly, we do not argue that sorting to high-premium firms is the only waythat workers from high social backgrounds enjoy social capital-based advantages or ben-efit from class bias. Cultural matching, for example, may also make it more likely thata privileged worker will be promoted within a firm (Friedman and Laurison, 2020).Distinguishing persistent worker-level advantages from contextually-specific firm ad-vantages simply identifies one general means through which social background affectsmobility.

Second, if there is evidence that intergenerational earnings transmission arises fromboth persistent differences in pay-offs to worker characteristics, like human capital, andfrom firm sorting, which matters more? While research on human capital differenceshas carefully quantified the portion of overall intergenerational mobility attributable toeducation, research on social capital and class bias has focused on establishing specificmechanisms (Bills et al., 2017; Rivera, 2020). These studies provide an existence prooffor social capital and class preference mobility effects. But, they do not quantify theimportance of unequal firm sorting processes relative to other determinants of inter-generational mobility. In the analysis below, we provide a decomposition frameworkto measure this magnitude.

But first, we consider a final outstanding question: what exactly are the mecha-nisms through which firm premiums affect intergenerational mobility? As noted above,explanations range from direct parental job referrals to diffuse cultural matching orclass-based discrimination. The human capital-focused explanations have carefullydistinguished, say, the influence of cognitive traits, personality, and education creden-tials (Bowles et al., 2009; Farkas, 2003; Karlson, 2019; Zhou, 2019). Similarly, thesocial capital or firm advantages explanation needs to compare the different pathwaysof social advantage proposed by prior research. In the following section, before in-troducing our analytical approach, we conceptually distinguish different mechanismsthrough which firms affect intergenerational transmission of advantage.

3 Mechanisms of intergenerational earnings trans-mission

Here we theorize and distinguish mechanisms that contribute to intergenerational earn-ings persistence, focusing on the role of firms. The premise of our approach is to sep-arate fathers’ and children’s earnings into components attributable to firm-level paysetting and to stable worker characteristics. This leads to four potential pathways be-tween the various components: (i) father’s firm component of earnings to child’s firmcomponent, (ii) father’s individual worker component of earnings to child’s firm com-ponent, (iii) father’s firm component to child’s worker component, and (iv) father’s to

5

child’s worker components. The first two of these describe how privileged children areable match with high-paying employers. The second two capture ways that parents areable to invest in children’s human capital, in line with currently dominant theoreticalaccounts.

3.1 Family influences on children’s firm sorting

We first consider mechanisms of firm sorting. In Table 1, we distinguish several mecha-nisms that allow privileged children to match with high-paying firms. We contrast eachin turn, but focus on the key distinction between the type of parental class advantagemobilized: (i) parental employment at a high-paying firm or (ii) parents’ human capitaland other forms of individual earnings advantage.

3.1.1 Inheritance mechanisms: firm-to-firm

The most direct way that a parent can transmit a high-paying firm to their child is bysecuring them a job at the parent’s own firm. This can happen when a parent ownsa business, more common among high than low class parents, and hires a child into it(Corak and Piraino, 2011). More frequent however, is when a parent employed at ahigh-premium firm refers their child to work there. Staiger (2021) finds that in the US,7% of children find their first stable job at the employer of one of their parents. Directco-employment provides higher pay for children because they are disproportionatelyreferred to high-premium employers (Staiger, 2021). Similar results on parental referralof children at labor market entry are also found in Sweden (the context we study inthe empirical analysis below) (Kramarz and Skans, 2014).

Beyond this direct hire or referral process, parents at high-paying employers canpass along to their children employment in their high-paying industry, even if the chil-dren do not obtain a position in the same firm as the parent. Older research stressedthe role of industry in processes of stratification (Bibb and Form, 1977; Morris andMurphy, 1959; Spilerman, 1977; Stolzenberg, 1975; Tolbert, 1983). Industry inheri-tance can happen due to geographical propinquity between parents and children. Butit can also occur via informal, industry-specific training provided by parent to child,analogous to the occupation-specific informal training and lifestyles that transmit oc-cupational microclasses (Gerstl and Cohen, 1964; Jonsson et al., 2009).1 In addition tothis industry-specific training, parents employed in high-paying industries are likely toobtain network ties among their colleagues. If a child gets a job in the same high-payingindustry their parent works in, often they benefited from some combination of directparental coaching and the activation of the parent’s wider social network of colleagues.

In both of these mechanisms, parents ensconced at high-paying firms are able totransmit advantaged positions to their children. This occurs either directly, by parentalsponsorship at their own employer, or indirectly, through microclass know-how or

1For example, Jonsson et al. (2009, p. 1012) report excess mobility between “similar” microclassessuch as accountants and bookkeepers, authors and librarians, or ship officers and fishermen.

6

Tabl

e1:

Fam

ilyin

fluen

ces

onch

ildre

n’s

firm

sort

ing

Inhe

rita

nce

mec

hani

sms:

firm

-to-

firm

Scaff

oldi

ngm

echa

nism

s:w

orke

r-to

-firm

Fir

min

heri

tanc

eIn

dust

ryin

heri

tanc

ePar

enta

laid

injo

bse

arch

Soci

alne

twor

ks,c

ultu

ral

mat

chin

g

Defi

niti

onPar

ent

dire

ctly

pass

esem

ploy

men

tat

ahi

gh-p

ayin

gem

ploy

erto

child

Par

ent

pass

eshi

ghpa

ying

indu

stry

toch

ild,t

hrou

ghco

nnec

tion

sor

indu

stry

-spe

cific

trai

ning

Hig

h-sk

illpa

rent

prov

ides

dire

ctfin

anci

alsu

ppor

tor

advi

cefo

ra

child

duri

ngth

eir

earl

yjo

bse

arch

Chi

ld’s

high

-cla

ssdi

spos

itio

nsor

soci

alne

twor

kso

rtth

emto

afir

mw

ith

high

-cla

ssco

wor

kers

Exa

mpl

ePar

ent

prov

ides

refe

rral

toth

eir

empl

oyer

for

child

’sfir

stjo

b

Par

ent’

sin

dust

ryco

nnec

tion

she

lpch

ildge

tan

inte

rnsh

ipin

thei

rin

dust

ry

Par

ent

prov

ides

hous

ing

and

finan

cial

supp

ort

duri

ngch

ild’s

earl

yjo

bse

arch

Par

ent

enco

urag

eshi

gh-c

lass

hobb

ies,

cult

ural

tast

e,th

athe

lpch

ildin

inte

rvie

w

Par

ent

adva

ntag

eFir

mpr

emiu

mFir

mpr

emiu

mH

uman

capi

tal

Hum

anca

pita

l

Chi

ldad

vant

age

Fir

mpr

emiu

mFir

mpr

emiu

mFir

mpr

emiu

mFir

mpr

emiu

m

Soci

alca

pita

lPar

enta

lref

erra

lPar

ent

netw

ork

refe

rral

Par

ent

netw

ork

refe

rral

Chi

ldne

twor

k

Oth

erre

sour

ces

Geo

grap

hic

prop

inqu

ity

Info

rmal

trai

ning

Fin

anci

alsu

ppor

tC

lass

-pre

ferr

edcu

ltur

aldi

spos

itio

ns

Life

cour

sest

age

Ear

lyE

arly

Ear

lyC

onst

ant

Pri

orre

sear

chC

orak

and

Pir

aino

(201

1);

Stai

ger

(202

1)Jo

nsso

net

al.(

2009

);Tol

bert

(198

3)A

rmst

rong

and

Ham

ilton

(201

3);L

area

u(2

015)

Laur

ison

and

Frie

dman

(201

6);R

iver

a(2

012)

Cor

epr

edic

tion

Chi

ldof

pare

ntw

ith

firm

prem

ium

gets

job

atpa

rent

’sfir

m

Chi

ldof

pare

ntin

high

-pre

miu

min

dust

ryge

tsjo

bin

pare

nt’s

indu

stry

Chi

ldof

high

-ski

llpa

rent

gets

earl

yjo

bat

high

-pay

firm

Chi

ldof

high

-ski

llpa

rent

wor

ksat

high

-pay

firm

wit

hhi

gh-c

lass

cow

orke

rs

7

industry-specific parental networks. Both of these mechanisms are most viable duringa child’s labor market entry or early career, when their parents are still in their primeearning years and have not yet retired.

3.1.2 Scaffolding mechanisms: worker-to-firm

In contrast to the firm and industry inheritance mechanisms, in which parents employedat high-paying firms pass them to children, parents with high individual earnings (butnot necessarily high firm premium employment) can also help their children sort toa high-paying employer. In this case, firm premiums and social capital do not haveto undergird the original earnings advantage enjoyed by a parent. Rather, a largelyhuman capital-based parental position can be transformed into a firm-based advantagefor the next generation.

The clearest mechanism here is direct parental aid during a child’s job search.High-earning parents assist their children through their early adulthood by providingmaterial resources, housing, social connections, or strategic advice (Cooney and Uhlen-berg, 1992; Hogan et al., 1993; Swartz, 2009). Financial resources can act as a safetynet that enables riskier choices with high return (Hällsten, 2010; Pfeffer and Hällsten,2012; Toft and Friedman, 2021).2 This assistance is most likely to be relevant duringthe early period of children’s labor market entry. It is also similar to the ways thatparents help children accumulate education and human capital. However, in this case,parental helping behavior extends into the labor market itself, and thereby impactssorting, rather than children’s individual characteristics alone.

Second, and more subtly, high-paid parents can cultivate dispositions and net-work ties for their children that help them gain the confidence of gatekeepers at high-premium firms (Armstrong and Hamilton, 2013). Some class-based dispositions aredomain general and benefit workers across multiple settings (Bourdieu, 1984; Khan,2010; Lareau, 2011). Others help only in interactions with high-class alters: interest inlacrosse and squash only helps in certain firms (Rivera, 2012), while a general feelingof ease and entitlement may help across multiple work settings (Khan, 2010). It isthe former, contextually-specific cultural dispositions that we consider here: disposi-tions that specifically help privileged workers in interactions with alters from similarbackgrounds. This social advantage could either result from direct class-based discrim-ination on the part of certain employers, or from a preference in hiring for homogeneitywith incumbents of similar background. Note also that this process could be profit-maximizing for the employer: shared cultural dispositions could be valuable in certain

2In the US it has been found that, in a given year, about half of those aged 18–34 receive someform of time help from their parents or older relatives, while a third receive cash transfers (Schoeniand Ross, 2005). Data on parental help in job search is scarce but Sage and Johnson (2012, p. 260)report that nearly three quarters of a community sample in their twenties said their parents helpedthem find a job at least once. Associations with labor market outcomes have yielded mixed results:often, parental assistance appears associated with worse outcomes (Faas et al., 2013; Manzoni, 2018).However, help is likely to arise in response to existing hardship (Swartz et al., 2011, 2017), andbetter-identified studies typically find substantial benefits (Krolikowski et al., 2020).

8

high-paying firms. Regardless, this matching process will lead workers from high socialbackgrounds to be sorted into higher paying firms.

Alongside the inculcation of high-brow cultural dispositions, high-earning parentsare also more likely to place children in the social settings, from schools to neighbor-hoods, that allow them to form ties with alters of high social background (McPhersonet al., 2001; Mijs and Roe, 2021; Owens, 2016; Reardon and Bischoff, 2011). These tiesare more valuable on the labor market than are randomly formed ties, assuming thatchildren of high social background disproportionately sort to higher paying firms. Assuch, children of high-earning parents are more likely to have the kind of acquaintanceand friendship networks that can help them secure high-paying employment.

This more diffuse mechanism of class matching thus encompasses two processes.The activation of direct network ties, which much prior research on social capitaland labor market sorting has sought to measure (Granovetter, 1995; Mouw, 2006),alongside cultural matching. The latter is sometimes included in broad definitionsof social capital, as general goodwill and helping behaviors associated with class orethnic solidarity (Adler and Kwon, 2002; Portes and Sensenbrenner, 1993), but it isconceptually distinguishable from activation of direct network ties. Unlike the industryinheritance mechanism, neither of these processes are confined to a single industry nordoes it require a parent to be employed in a high-premium firm or industry. Unlikedirect parental aid and support in job search, these mechanisms of privilege can persistacross the life course. Privileged workers can reap advantage whenever their classconfederates hold gatekeeping roles in high-premium firms.

3.2 Family influences on children’s human capital

The remaining two pathways—firm-to-worker (iii) and worker-to-worker (iv)—implythat parents are able to shape their children’s human capital, either through educationor the inculcation of other consistently rewarded traits. As in the first two pathways,they differ in the source of parental advantage: parents’ strategic location in the labormarket or their individual resources. These pathways reflect the conventional focus ofstratification research.

3.2.1 Investment mechanisms: firm-to-worker

Working at a high-paying firm provides a parent with more economic resources toinvest in their children. A large literature documents how affluent parents invest intheir children by acquiring housing in an attractive neighborhood, putting them in su-perior schools, purchasing extracurricular tuition, or else providing a developmentallyconducive environment (Bianchi et al., 2004; Danziger and Waldfogel, 2000; Kaushalet al., 2011; Kornrich and Furstenberg, 2013; Schneider et al., 2018). The effects ofthese investments on children’s human capital and future earnings have been studied ina variety of ways, ranging from effects of macro-level increases in inequality (Bloome,2015; Mayer, 2001) to involuntary job loss shocks (Berger and Engzell, 2022; Brand,

9

2015) and government transfers (Cooper and Stewart, 2020). These studies find mixedresults of parental investments on children’s human capital, consistent both with sub-stantial omitted variable bias and with effect heterogeneity across types of income andnational contexts.

In our framework, the firm-to-worker pathway isolates increased children’s humancapital investments made possible by a parent specifically working at a high-payingfirm, rather than from higher earnings attributable to the parent’s individual char-acteristics. This avoids the unobserved heterogeneity in parent characteristics thatinflates estimates in some prior observational studies. Moreover, we study Sweden,where education is fully state-funded at all levels and relatively egalitarian. For ex-ample, Cesarini et al. (2016) report that in Sweden even a large cash transfer in theform of a lottery win exerts no influence on children’s cognitive and noncognitive skillsor school performance. We therefore expect that direct effects of parental firm incomeon children’s individual attainment—as captured by the firm-to-worker pathway—arelikely to be muted.

3.2.2 Endowment mechanisms: worker-to-worker

In the three pathways we have considered so far, advantage is mediated through firmpay premiums at either the parent’s or the child’s side, or both. The last pathwayis instead through direct transmission of economically valuable traits, as a parentalendowment to children. This pathway covers the dominant explanations for the per-sistence of status, which focus on ways skilled parents pass valuable characteristicsto their children. Parents endow children with ability, work orientation, and othertraits that are culturally rewarded (e.g, Bowles and Gintis, 1976; Bowles et al., 2009;Coleman et al., 1967; Jencks et al., 1972, 1979). Dispositions of entitlement and ease,taught young, are subsequently valuable across a range of social settings (Bourdieu,1984; Calarco, 2018; Khan, 2010; Lareau, 2011). More controversially, genetic advan-tages that brought high returns to parents may help ensure the success of their children(Engzell and Tropf, 2019; Jencks, 1980). All of these mechanisms involve parents pass-ing to children individual characteristics that are durably valuable in a competitivelabor market.3

As noted above, much prior research in stratification has focused on distinguish-ing these different individual characteristics. In our approach, all consistently valu-able individual characteristics, on both the parent and child side, are bundled intothis broad worker-to-worker channel. The labor economics literature we draw on at-tributes these persistent individual differences in pay to differences in human capital,

3This transmission mechanism does not presuppose that parents invest economic resources in achild. A highly educated parent buying a house in a neighborhood with better schools, or paying ex-pensive college tuition, would be covered by this mechanism. But so would non-economic endowments,like a skilled parent informally teaching their child a trade, or a parent with a genetic endowment forphysical attractiveness passing that labor market advantage to their child (Monk et al., 2021). Weintentionally define this mechanism broadly to cover both the economic and direct endowment effectsestablished in prior research.

10

like time-invariant worker ability or educational credentials. But, the transmission ofpersistently valuable worker characteristics also covers characteristics not typically de-fined as human capital. Indeed, any characteristic that affects earnings across multiplefirm contexts—including domain-general cultural dispositions like ease (Khan, 2010;Lareau, 2011), or consistently penalized characteristics like racial minority status—isincluded in the worker effect transmission mechanism. Taking this broad approach letsus identify firm effects on intergenerational transmission net of any individual workercharacteristics.

As such, the distinction we draw between firm and worker mechanisms is betweenpersistent and firm-contingent labor market advantages, not strictly between meri-tocratic and non-meritocratic processes. We also do not contrast education to non-education channels per se: higher education may either matter by fostering persistenthuman capital or by facilitating a match to a high-paying firm. Instead, we controlbroadly for both observed and unobserved worker characteristics, to isolate the roleplayed by firm pay-setting and worker sorting in intergenerational earnings transmis-sion.

4 Data and approachTo test these ideas, we incorporate firms into a standard path analytical framework.We focus on earnings and study the correlation between parents (fathers, in our mainanalysis) and children. To assess the role of firm sorting in intergenerational persistence,we decompose earnings for each generation into components attributable to differencesin individual worker characteristics and differences in pay at the firms those workerssort into. We then use the components of this decomposition to identify the specificpaths through which firms shape intergenerational mobility.

For data, we construct an earnings panel linking parents and their children inStatistics Sweden’s Longitudinal Integration Database for Health Insurance and LaborMarket Studies (LISA). The data contain annual labor earnings for the full Swedishpopulation spanning the period 1990–2019, with individual and firm identifiers that areconsistent over time. Intergenerational linkage is made possible by a unique registrationnumber that each Swedish resident is assigned at birth or immigration. As results looksimilar for fathers and mothers, we display mothers’ earnings separately in AppendixA1. We include children born 1975–1978 and biological parents born 1940–1960. Whilewe are able to link families beyond this span, these cohort restrictions allow annualearnings to be measured at prime working age for both generations within the 1990–2019 observation window of the LISA database.

Our analysis follows several steps. First, we estimate a two-way fixed effects modelon the whole working population of Sweden born 1940–1985 and aged 25–65 for theyears 1990–2019 (Section 4.1). We use this model to extract firm pay premiums,controlling for time-invariant worker characteristics, and consistent worker premiums,controlling for firm sorting. Next, we link these firm and worker earnings components

11

across generations by estimating a path model of fathers’ influence on children’s mid-career earnings at age 38–42 (Section 4.2). We then consider how the contributions offirm-related earnings transmission pathways vary across the career by studying childrenbetween ages 25 and 40 (Section 4.3). Finally, we use various covariates to clarify whichmechanisms account for earnings persistence along the paths of firm transmission andindividual traits (Section 4.4). In the following methods discussion, we provide detailson each of these steps in turn, before presenting results.

4.1 Two-way decomposition of earnings

As our main outcome, we analyze intergenerational transmission of annual labor earn-ings. In our main analysis we follow the practice, common in recent work, of study-ing the intergenerational rank correlation of incomes (Bloome et al., 2018; Dahl andDeLeire, 2008; Engzell and Mood, 2021). This parameter can be written:

ρ = corr(R[Yi], R[Y∗i ]) =

cov(R[Yi], R[Y∗i ])√

var(R[Yi])√var(R[Y ∗

i ]), (1)

where Y ∗i and Yi denote parent and child earnings and R ∈ {1, 2, . . . , 100} represents

the percentile rank transform. From here on, we do not write out the R operator butlet it be understood that Y ∗

i and Yi denote earnings ranks. To reduce the influence ofgender and life-cycle earnings profiles, as well as economy-wide shocks and inflation,we create percentile ranks within each cell defined by gender, age, and earnings year.Log earnings produce similar results and are presented in Appendix A2 below.

To assess the role of firms in intergenerational mobility, we separate labor earn-ings into parts attributable to firm-based earnings premiums and to individual workertraits. This decomposition allows us to estimate firm premiums while controlling forthe broad array of individual worker characteristics, including education, prominentin prior intergenerational mobility research. Specifically, we follow recent work whichhas fit two-way fixed effects models on linked employer–employee data (Abowd et al.,1999; Card et al., 2013; Song et al., 2018). Our model can be written:

Yit = αi +X ′itβ + ψj(i,t) + uit, (2)

where Yit is labor earnings in year t for worker i in firm j, X ′it is a vector of gender,

education, age, and full interactions between the three, β is the set of associated coeffi-cients, ψj(i,t) is a firm- specific intercept associated with the firm employing individuali at time t, αi is an individual-level intercept, and uit is an idiosyncratic error term.

The key terms for our analysis are the firm premium ψj and the individual workercomponents αi andXit. The firm-level term ψj captures the degree to which a firm paysrelatively high or low wages, after adjusting for the composition of their workers. Thisvariation is typically driven by sector, industry, product, and location, alongside moreidiosyncratic firm characteristics, like profitability, collective bargaining coverage or

12

company-wide compensation policies. The individual worker-level component αi varia-tion in unobserved, time-invariant worker heterogeneity, independent of any additionalbenefit from working in a specific firm. Combined with the controls for observableworker characteristics in the vector Xit, αi reflects individual productivity and humancapital, but it also captures less meritocratic determinants of individual workers’ paythat have consistent effects across employers, like persistent discrimination.

There are two issues with this canonical model that have received growing attention.First, limited mobility bias may increase the variance of estimated firm fixed effects atfirms that have little exposure to workers moving between firms (Andrews et al., 2008;Kline et al., 2020). This occurs because the firm fixed effects ψj are estimated usingworkers who transition between firms in the data. Reassuringly, models with similarSwedish data to what we use in the current paper show that firm effects account foralmost exactly the same share of earnings variance when this bias is corrected (Engbomet al., 2021). We also pool a very long time span of nearly 30 years of data, whichrecent work finds mitigates potential mobility bias that arises in estimating two-wayfixed effects models over short time spans (Lachowska et al., 2020).

Second, the time-invariant nature of αi assumes that an individual’s latent wageis constant conditional on observed characteristics, but this may be an unrealisticassumption (Bihagen, 2008; Cheng and Song, 2019). We deal with this fact by includinghighly flexible age-earnings trajectories in X ′

itβ that are allowed to vary with individualcharacteristics. Specifically, we include a set of age dummies interacted with genderand education in five levels. Following Engbom et al. (2021), we assume that earningsrank is roughly constant between ages 45 to 54 by grouping these ages together as ourreference category. The predicted worker component of earnings is then the sum of theindividual-specific intercept αi and the estimated age-earnings profile X ′

itβ:

Yit = αi +X ′itβ︸ ︷︷ ︸

worker (Wit)

+ ψj(i,t)︸ ︷︷ ︸firm (Fit)

. (3)

To ensure that this estimation strategy is robust, we implement a number of al-ternative solutions below. First, we extend the earnings trajectories by letting themdepend on up to 47 levels of education, as well as 118 detailed fields of study. Sec-ond, we break our estimation sample up into distinct 5-year periods which allows eachworker and firm to be associated with a number of fixed effects throughout the obser-vation window. Third, we extend the canonical specification with one that allows forindividual slopes by age as part of the worker component (Rüttenauer and Ludwig,2020). None of these alternative strategies make any major changes to our results andwe therefore display them separately in Appendix A3.

4.2 Path model

Having estimated the separate firm and individual worker contributions to earnings,we explore the pathways of transmission in a path model. Does intergenerational

13

F ∗i Fit

W ∗i Wit

Y ∗i Yit

ϵ3it

ϵ4it

ϵ1i ϵ2it

ρff,t

ρfw,t

a1 a2t

ρwf,t

ρww,t

b1 b2t

Parent Child

Figure 1. Path model.

earnings transmission mostly work through the inheritance of persistent individualtraits as traditional intergenerational mobility research would lead us to expect? Or isa significant part of intergenerational advantage instead transmitted by employment athigh-paying firms? Figure 1 displays the basic setup of our model, where the superscript∗ denotes the parental generation. We model earnings as a function of firm and workercomponents in each generation:

Y ∗i = a1F

∗i + b1W

∗i + ϵ1i, (4)

Yit = a2tFit + b2tWit + ϵ2it, (5)

where Y ∗i is the father’s long-run earnings, Yit is the child’s earnings at age t, and

the right-hand side variables are the corresponding firm (F ∗i , Fit) and worker (W ∗

i ,Wit)components. Note the absence of a subscript t for fathers: we treat social backgroundas a constant by averaging father’s earnings (as well as the father’s firm and workercomponents) over 10 years as described in Section 4.3 below. By contrast, we initiallyfix the child’s earnings by averaging over age 38-42 and then relax this to allow com-ponents to vary throughout the early career. The pathways we are interested in arethe following:

Firm-to-firm: Y ∗i ← F ∗

i → Fit → YitWorker-to-firm: Y ∗

i ← W ∗i → Fit → Yit

Firm-to-worker: Y ∗i ← F ∗

i → Wit → YitWorker-to-worker: Y ∗

i ← W ∗i → Wit → Yit

To estimate the importance of each of these paths of transmission, we take ourdeparture from Duncan’s basic theorem of path analysis (Duncan, 1966), which states

14

that the correlation between any two variables p and q, ρpq, can be decomposed into adiscrete set of paths:

ρpq =K∑k=1

ρpkdqk, (6)

where the index k iterates over all mediator variables that intervene on the pathfrom p to q, ρpk is a set of bidirectional correlations between p and mediator k, anddqk is a set of path coefficients obtained by regressing the outcome q on the joint setof mediators.

In our case, the path coefficients d correspond to the partial regression coefficientspredicting the father’s (a1, b1) and child’s (a2t, b2t) earnings in Equations 4–5 above.To trace the path through firm and worker components we further define the followingcorrelations:

ρff,t = corr(F ∗i , Fit) (7)

ρwf,t = corr(W ∗i , Fit) (8)

ρfw,t = corr(F ∗i ,Wit) (9)

ρww,t = corr(W ∗i ,Wit) (10)

Given a set of assumptions that we describe further in Appendix A4, the intergen-erational earnings correlation at child’s age t can then be decomposed as:

ρt = corr(Y ∗i , Yit) (11)

= ρff,t · a1 · a2t︸ ︷︷ ︸firm-to-firm

+ ρwf,t · b1 · a2t︸ ︷︷ ︸worker-to-firm

+ ρfw,t · a1 · b2t︸ ︷︷ ︸firm-to-worker

+ ρww,t · b1 · b2t︸ ︷︷ ︸worker-to-worker

.

Equation 11 shows that the relative contribution of each mechanism to the over-all intergenerational correlation depends not only on the correlation between father’scomponent and child’s component (the ρ terms), but also on the importance of thosecomponents for each generation’s overall earnings (the a and b terms). For example, ahigh correlation between father’s firm premium and son’s firm premium may accountfor a small share of the overall intergenerational correlation if the firm premium is asmall component of overall earnings.

4.3 Firm contributions to mobility by career stage

In the previous section, we have abstracted from the dimension of children’s age. The30-year span of our data is sufficient to let us observe fathers and children over severalyears near peak or mature earnings. Here we describe how we implement this focuson mid-career earnings, and how we extend this analysis to study earnings mobilitythroughout the early career.

In all our analyses, we treat parental social background as a constant. We do so bycollapsing father’s earnings Y ∗

i across the first 10 years in our data, 1990–1999. At this

15

point, the youngest child cohorts are aged 12–21 and the oldest 15–24. Fathers are inthe age range 30–59, which is a time when male earnings are relatively representative oflifetime income (Engzell and Mood, 2021). Because the average of a uniform variable isgenerally not uniform, we reimpose the rank transform within (child) gender–age cellsto retain the interpretation of coefficients as rank-order correlations. We create long-runmeasures of fathers’ firm-level component F ∗

i and individual worker-level componentW ∗

i in the same way, by collapsing predicted values over the years 1990–1999 andrepeating the rank transform. In additional analyses, we display results based onmothers.

For children, we focus on earnings centered on age 40 in our main analyses. Toachieve reliability, we collapse observed ranks throughout ages 38–42, excluding the1978 cohort who is only followed until age 41. Earnings at age 40 is a good proxy forlifetime income, unlike earnings earlier in the career (Engzell and Mood, 2021; Nybomand Stuhler, 2017). We average the firm and individual worker fixed effects (Fit, Wit)over the same ages, 38–42. Like for the fathers, we reimpose the rank transform aftertaking multi-year averages.

In a second step of analysis, we separate out annual earnings Yit for each year be-tween the ages 25–40. This means following children from 2000–2015 for our oldestcohort (born 1975) and 2003–2018 for our youngest (born 1978). These age-specificestimates allow us to track the importance of different earnings components to mobilityacross career stages. It also helps us begin to distinguish two theoretical mechanismssubsumed by the worker-to-firm path introduced above: direct parental support in jobsearch, which should be most important at labor market entry and early career, fromthe cultural matching and network mechanism, which should be persistently impor-tant across a career. In the next section, we discuss additional analyses that furtherdistinguish mechanisms within the broad paths discussed here.

4.4 Distinguishing mechanisms of children’s firm sorting

Our core empirical contribution is to quantify the importance of the firm-related pathspresent in Figure 1 and how they change over the course of children’s careers. Ina further set of analyses, we work to map these paths onto the specific substantivemechanisms considered in in Table 1 and proposed in prior research. These testsprovide initial evidence on the likely interpretation of the broad and rather abstractpaths.

Specifically, we work to distinguish the firm-to-firm path into mechanisms of directfirm inheritance, direct industry inheritance or other sources of firm premium correla-tion. We also build on our career stage analysis to test mechanisms of worker-to-firmtransmission: direct parental assistance in the job search, cultural matching and socialnetworks, or selection on individual worker characteristics.

To capture firm inheritance we compare results for children who find employment attheir father’s firm versus those who do not. If firm inheritance is the key source of firm-mediated transmission of earnings, we would expect that excluding these same-firm

16

parent-child pairs would substantially attenuate the role of firms in intergenerationalmobility. We code a child as overlapping if they are observed working in the same firmas the father at any time between ages 38 and 42. We also considered overlap at ayounger age for children (25-30) and found similar results.

Next, we assess industry inheritance using a similar logic, comparing children whofollow in their father’s footsteps versus those who do not. We measure industry usingthe 1-digit level of the official Swedish industry classification, SNI. This level encom-passes 10 categories—agriculture; construction; culture and services; education andresearch; energy; finance and business; health care; manufacturing; public adminis-tration; trade and communication—as well as an “unspecified” category. If industryinheritance drives firm-mediated earnings transmission, then excluding same-industry(but different firm) parent-child pairs should substantially attenuate the role of firms inintergenerational mobility. This industry inheritance channel, like the firm inheritancechannel, should mainly affect the firm-to-firm transmission pathway.

In both the firm and industry cases, we do not directly observe whether a fatherreferred a child to their firm: in some cases with large employers and large indus-tries, these apparent inheritances could occur by coincidence. So, these same industryand firm estimates should be interpreted as an upper bound on the effect of directinheritance on intergenerational transmission.

Next, to assess whether the worker-to-firm channel is mainly due to parental aidin job search, we use the assumption that to benefit from direct parental support, aparent has to be present. We ask whether the child’s current job at age 40 is locatedin the same county or municipality as the father’s main employer. Sweden consists of21 counties and 290 municipalities, the population of which averaged about 400,000(county) or 30,000 (municipality) in 2000. If direct parental involvement was criticalscaffolding for children, we would expect the intergenerational earnings association tobe weaker if the child lives in a different region from the father.

The above tests rely on subsample analyses, which we implement as the interactionbetween intergenerational earnings transmission (or specific correlation paths) and aseries of moderator variables (same firm, same industry, co-location). To assess otherdeterminants of class-based firm sorting, we take a slightly different approach of con-trolling for co-worker and worker composition while estimating the intergenerationalcorrelation. First, to assess the second scaffolding mechanism, of co-worker homophilyin which privileged workers sort into firms with workers of similar background, we takethe mean of co-workers’ parental income:

Y∗i′t =

1I−1

I−1∑i′=1

Y ∗i′ , ∀i′ : j(i′, t) = j(i, t), i′ = i. (12)

Here, Y ∗i is the long-run measure of father earnings defined above, I refers to the

total number of workers in a firm, and the procedure averages across all co-workers i′at time t, that is, leaving out the focal individual i. Consistent with our other earningsmeasures, we average this measure across ages 38–42, and reimpose the rank transform.

17

If cultural matching or network homophily was a leading cause for firm-mediated trans-mission of earnings, then the intergenerational earnings correlation should be mediatedby co-worker composition, and especially so for the worker-to-firm pathway of trans-mission.

Finally, we test whether class-based firm sorting is, alternatively, due to individualworker characteristics. High-paying firms may be particularly good at identifying andattracting high productivity workers. If this is the case, then firm sorting advantagesfor high-class workers may be due to their individual characteristics. To test this, weadd a control for worker characteristics Wit to the worker-to-firm path. This controlsfor class-based firm sorting that is mediated by individual worker characteristics.

Together, these additional tests provide suggestive evidence on the relative impor-tance of specific theoretical mechanisms among the firm-to-firm and worker-to-firmtransmission paths. However, fully establishing the role of any individual mechanismwould require its own paper. Instead, we use these tests to provide some bounding andquantification of the likely role played by each mechanism in the broader decompositionof earnings transmission.

4.5 Intergenerational mobility across types of firms

In most of the analysis, we distinguish only between high and low premium firms,consistent with our reliance on a broad earnings decomposition. In a final, descriptivestep, we distinguish among different types of high premium firms and assess which arerelatively more likely to provide high-paying employment to workers of high or lowsocial background.

Specifically, we compare among sons from the bottom and top quintile of father’searnings distribution who end up in high-premium employment. We then comparetheir employers across dimensions including industry, firm size, financial performance,and the presence of other workers of high social origin. This analysis allows us tocharacterize the kinds of firms that tend to provide high firm premiums for workers ofhigh versus low social background.

5 Results

5.1 The contribution of firms to overall intergenerational earn-ings persistence

We first assess whether parental earnings predict children’s firm sorting, and not justthe individual, time-invariant component of children’s earnings. Figure 2 displayschildren’s firm and individual worker components of earnings at each percentile offather earnings. We focus on children’s mid-career earnings centered on age 40 and weseparate results by child gender.

18

40

50

60

70

Chi

ld ra

nk

0 20 40 60 80 100Father earnings rank

Firm Worker

(a) Sons

40

50

60

70

Chi

ld ra

nk

0 20 40 60 80 100Father earnings rank

Firm Worker

(b) Daughters

Figure 2. Average child firm and worker rank by father’s earnings rank.

Figure 2 shows that fathers’ earnings have a substantial association with both theindividual worker and and firm components of child earnings. For sons, the individualworker effect association is 0.24 with fathers’ earnings, comparable to the earningscorrelation itself, at ρ = 0.24 (see Table A5). The firm-level association is smallerbut nevertheless substantial, at 0.15. For daughters, the worker and firm associationsare more similar: the individual worker effect association is 0.20 and the firm-levelassociation 0.16, compared to an overall earnings correlation of ρ = 0.20 (see Table A5).So, father’s earnings predicts not only the individual worker component of earningschildren’s earnings, emphasized in prior intergenerational mobility research, but alsothe firm sorting component.

Given these strong correlations, we next ask how much firm sorting contributes tothe overall persistence of earnings across generations. To do this, we need to estimateall of the separate terms in Equation 11 of our decomposition framework above. Webring these components together in Table 2, which details the proportional contributionto the earnings correlation of each of the four paths: firm-to-firm, worker-to-firm, firm-to-worker, worker-to-worker. Filling in the path model is critical, as, consistent withearlier work on the two-way worker-firm fixed effect model, the worker componentis substantially more important to explaining overall earnings variation (as capturedfor sons in the firm correlation a2 = 0.36 compared to the worker effect correlationb2 = 0.74). Although father’s earnings strongly predict son’s firm premium, the firmpremium, for both generations, accounts for less variation in earnings than does thestable worker component.

Taking these different correlations with overall earnings into account, we find thatthe worker-to-worker path accounts for between 71% (sons) and 72% (daughters) ofthe total intergenerational earnings correlation. Firm-related pathways account for theremaining quarter of intergenerational mobility. This decomposition shows that while

19

Table 2: Decomposition parameters at child age 38–42.

Estimate (percent)

Parameter Sons Daughtersfirm-to-firm = ρff · a1 · a2 0.016 (6.80%) 0.007 (3.91%)

worker-to-firm = ρwf · b1 · a2 0.037 (16.4%) 0.036 (19.2%)firm-to-worker = ρfw · a1 · b2 0.013 (5.66%) 0.008 (4.45%)

worker-to-worker = ρww · b1 · b2 0.162 (71.1%) 0.137 (72.4%)ff + wf + fw + ww = corr(Y ∗

i , Yi) 0.228 (100%) 0.189 (100%)ρff = corr(F ∗

i , Fi) 0.120 0.070ρwf = corr(W ∗

i , Fi) 0.126 0.150ρfw = corr(F ∗

i ,Wi) 0.048 0.031ρww = corr(W ∗

i ,Wi) 0.263 0.218a1 = corr(F ∗

i , Y∗i | W ∗

i ) 0.363 0.366b1 = corr(W ∗

i , Y∗i | F ∗

i ) 0.837 0.835a2 = corr(Fi, Yi | Wi) 0.355 0.289b2 = corr(Wi, Yi | Fi) 0.738 0.751

worker characteristics are the dominant source of intergenerational earnings transmis-sion, firm sorting plays an important role in the mobility process.

Table 2 also provides our first evidence distinguishing inheritance, scaffolding andinvestment mechanisms through which firms affect intergenerational mobility. Al-though the underlying firm-to-firm (ρff) and worker-to-firm (ρwf) correlations are sim-ilar, the worker-to-firm path is more important for the overall intergenerational corre-lation because father’s individual worker component of earnings is a more influentialpredictor of earnings than is his firm premium rank. Specifically, the firm-to-firm andfirm-to-worker paths each account for about 6–7% of the intergenerational earningscorrelation for sons, and about 4% for daughters. Markedly more important is theworker-to-firm path which alone accounts for more than half of the share explained byfirms: 16% of the overall correlation for sons and 19% for daughters.

This first set of results shows that children of high-earning parents disproportion-ately sort to high premium firms. While this process is not the main channel of inter-generational earnings persistence, it accounts for around a quarter—a quantitativelymeaningful share. We further show that the main source of this firm-related inter-generational transmission is the worker-to-firm, or scaffolding mechanisms, path. Inthe remaining analysis, we clarify the specific ways through which scaffolding, and theother mechanisms of firm-related transmission, affect intergenerational persistence.

20

0

.05

.1

.15

.2R

ank

corre

latio

n

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure 3. Decomposition of the intergenerational earnings correlation over the earlycareer, by sex and age of child.

5.2 Firm effects across career stages

We have shown how firm and individual worker earnings components contribute to theintergenerational earnings correlation at mid-career, but not how these componentschange over the early career period. As discussed above, several mechanisms of firm-related persistence—like firm inheritance and direct parental aid in job search—shouldbe strongest at the early career stage, before tapering out by mid-career.

Figure 3 tests this by decomposing the intergenerational earnings correlation intoits path components at each year of a child’s career. Firm sorting is indeed mostimportant early in the career and mediates the bulk of earnings transmission in thefirst few years of labor market entry. This fades gradually to about a quarter of thetotal correlation by age 40 (for relative numbers, see Figure A12 in Appendix A5).

This decline across the career in the firm-related share of earnings transmission islargely explained by the large increase in importance of the worker-to-worker path, asthat correlation strengthens over the career and the child’s worker component becomesa more important portion of their earnings. The firm-to-worker or investment pathwayalso strengthens slightly but remains relatively unimportant throughout the career.

At the same time, firm-to-firm transmission faces a small decline in absolute termsand a large relative decline. This is consistent with our discussion of mechanisms above:children inheriting their parents’ firms or industries is most likely early in children’scareers, when they can benefit from peak earning age parents and from finding an

21

initial foothold in the labor market. Having a parent employed at a high-earning firmmainly benefits workers when they are young.

In contrast, the worker-to-firm transmission path grows in absolute terms and re-tains an important share of intergenerational earnings persistence up through age 40.Note that this scaffolding pathway, in which children of high-earning parents end upworking at high-premium firms, could be important both at early and later careerstages. Direct parental aid should matter mainly early on, when parents are still activeand engaged with their children’s labor market attainment. Indirect class bias andsocial capital advantages, in contrast, could pay off throughout a child’s career. Thecontinued importance of the worker-to-firm pathway up through peak earnings yearssuggests that scaffolding mainly operates through bias and social capital channels,rather than through direct parental aid, channels.

Overall, these career stage results show that social background differences in firmsorting are particularly strong early in workers’ careers. Consistent with the resultsin Table 2, worker-to-firm transmission is the most persistently important firm-relatedchannel across the career. These results also provide initial evidence that rather thandirect parental aid at labor market entry, class-biased sorting processes, like culturalmatching and social networks, are the key drivers of worker-to-firm transmission. Next,we consider further tests to distinguish specific mechanisms among firm-related trans-mission paths.

5.3 Results on specific mechanisms

Our analysis so far shows that an important part of the intergenerational earningscorrelation is mediated by labor market advantages associated with firms, as opposedto individual characteristics like human capital. A simple story about the inheritanceof firm or industry would lead us to expect that the firm-to-firm component of earningstransmission accounts for the largest part of this sorting advantage. Interestingly, thisis not what we find. Instead, the most important aspect is individually high-earningfathers placing their children in high-paying firms. We now conduct some simple teststo determine which mechanisms are most plausible and important for driving the roleof firms in intergenerational transmission.

We first measure direct firm and industry inheritance by comparing children whofind employment in their father’s firm or industry with those who do not. If firmand industry inheritance are important for earnings transmission in the population asa whole, we would find much smaller intergenerational correlations when we excludechildren employed in their father’s firm or industry.

Figure 4 shows that children who work in the same firm as their father indeed havea much higher correlation between their firm premium and their father’s firm premium(around 0.6 for sons and 0.4 for daughters). This is also true, though more weakly,of children who work in the same industry as their father. However, this populationof inheritors is small enough that excluding them entirely does not substantially effectthe overall earnings correlation. Removing firm inheritors reduces the overall earnings

22

+41%

+38%

+37%

-2%

-11%

+408%

+121%

+49%

-22%

-38%

Baseline

Same firm?

Same industry?Same industry?

(Different firm)

Baseline

Same firm?

Same industry?Same industry?

(Different firm)

0 .2 .4 .6 .8

Earnings correlation: ρ

Firm-to-firm: ρff

YesNo

Coefficient

(a) Sons

+34%

+59%

+60%

-1%

-12%

+530%

+210%

+150%

-16%

-41%

Baseline

Same firm?

Same industry?Same industry?

(Different firm)

Baseline

Same firm?

Same industry?Same industry?

(Different firm)

0 .2 .4 .6 .8

Earnings correlation: ρ

Firm-to-firm: ρff

YesNo

Coefficient

(b) Daughters

Figure 4. Test of firm and industry inheritance. Coefficient change when splittingthe sample by whether child works in the same firm and/or industry as father’s mainemployer at child age 38–42.

correlation by 1 or 2% and even removing the much larger group of children who work inthe same broad industry as their father only lowers the earnings correlation by around10%.4 Overall, direct firm, and industry, inheritance does strengthen intergenerationalearnings persistence. And the larger number of industry than firm inheritors meansthat industry inheritance drives more intergenerational earnings correlation than firminheritance. However, direct inheritance of either industry or firm is rare enough thatit accounts for only a small portion of the overall intergenerational correlation.

Next, we shift from the mechanisms underlying the firm-to-firm path to those un-derling worker-to-firm transmission. First, we split out children who find employmentin the same municipality or county as their father. These restrictions get at parentalaid in the job search under the assumption that a parent has to be present to providehands-on help. However, Figure 5 shows that the difference in correlations betweenco-located and non-co-located families is modest. For example, restricting to sonswho live in an entirely different municipality from their father only reduces the overallearnings correlation by 12% and the worker-to-firm pathway by 20%. This does notmean that parental support in job search is irrelevant: living in the same municipalityas one’s father increases worker-to-firm transmission by around 40%. However, evenusing this very broad gauge of parental help, most of the worker-to-firm correlation per-sists among non-co-located children. We thus see parental aid as an interesting area

4Removing inheritors has a much stronger effect, as expected, on the firm-to-firm correlation. In theconservative specification in Figure 4, removing direct inheritors reduces the firm-to-firm correlationby 40%. When using a broader definition of co-working, in which the child ever works in the sameindustry as the father’s main employer, rather than requiring simultaneous co-working at ages 38-42,removing direct inheritors fully accounts for the firm-to-firm pathway.

23

+23%

+9%

-12%

-9%

+42%

+14%

-20%

-10%

Baseline

Same municipality?

Same county?

Baseline

Same municipality?

Same county?

0 .05 .1 .15 .2 .25 .3

Earnings correlation: ρ

Worker-to-firm: ρwf

YesNo

Coefficient

(a) Sons

+23%

+7%

-11%

-5%

+43%

+18%

-16%

+1%

Baseline

Same municipality?

Same county?

Baseline

Same municipality?

Same county?

0 .05 .1 .15 .2 .25 .3

Earnings correlation: ρ

Worker-to-firm: ρwf

YesNo

Coefficient

(b) Daughters

Figure 5. Test of parental aid in job search. Coefficient change when splitting thesample by whether child works in the same municipality and/or county as father’smain employer at child age 38–42, and whether father was alive at child age 25.

for further research, but not likely the primary source of either the overall earningscorrelation or the worker-to-firm pathway specifically.

This leaves us with our last hypothesized mechanisms of class-based sorting acrossfirms. This can occur due to privileged workers sorting into firms via cultural match-ing and explicitly class-biased hiring, or by social network homophily. It could alsosimply reflect high-premium firms selecting for high-class background workers becausethose workers have higher human capital (as suggested by the strong worker-to-workertransmission path). For the homophily interpretation, we ask whether workers sortinto firms dominated by co-workers of a similar class background. We operationalizethis by adding the average parental earnings of co-workers as a control variable. Forthe human capital selection interpretation, we simply control for the child’s workerpremium.

Figure 6 displays the results. Co-workers’ earnings background accounts for fullythree quarters of the worker-to-firm pathway. It also accounts for a third of the overallearnings correlation. In Appendix Figure A16, we show that this effect eclipses that ofa wide set of other individual and firm characteristics, including education, industry,sector, and firm size. Figure 6 also shows that even controlling for our broad measureof consistently valuable worker characteristics leaves a substantial residual social back-ground sorting advantage. So, social background affects sorting across firms. And thistest gives suggestive evidence that homophily-related processes, like cultural matchingor social capital, could account for this.

24

-35%

-85%

-74%

-61%

Baseline

Co-worker class

Worker FE

Baseline

Co-worker class

Worker FE

0 .05 .1 .15 .2 .25

Earnings correlation: ρ

Worker-to-firm: ρwf

Coefficient

(a) Sons

-34%

-83%

-75%

-30%

Baseline

Co-worker class

Worker FE

Baseline

Co-worker class

Worker FE

0 .05 .1 .15 .2 .25

Earnings correlation: ρ

Worker-to-firm: ρwf

Coefficient

(b) Daughters

Figure 6. Testing sources of class-based firm sorting. Coefficient change when control-ling for co-workers’ average father earnings rank and for individual worker component.

5.4 High and low mobility firm types

Thus far, we have bracketed the question of where firm premiums come from. This is along-standing and active area of research. We do not make strong assumptions aboutwhether firm premiums in our data arise from productivity differences, institutional ororganizational pay-setting constraints or from rent-sharing and market power (Barthet al., 2018; Groshen, 1991; Kalleberg et al., 1981; Slichter, 1950; Wilmers, 2018).However, we can use our data on characteristics of firms to shed light on the types ofhigh premium firms that tend to provide opportunities for workers from high versuslow social backgrounds.

In Figure 7 we look at where sons from high-earning compared to low-earningbackgrounds find high-pay employment. We define high-paying firms as those thatrank in the top quintile of firm pay premiums. We focus on sons in such high-payingfirms who grew up in the top quintile of father earnings, and contrast to sons whoend up in similarly high-paying firms, but who grew up in the bottom father earningsquintile. As a shorthand, we refer to these two groups as sons of “high” and “low” class.

Figure 7a shows that high-class sons that end up in high-paying firms are far morelikely than successful low-class sons to work in firms with other high-class peers. Thispattern is consistent with the results above, and with the scaffolding social capitalchannel. High-class sons are also more likely to work in the capital city. In contrast,both low- and high-class sons that sort to high-paying firms disproportionately workin large firms. This is particularly the case for low-class sons. Both low-class and high-class sons also work in relatively profitable and high-valued added firms. For bothhigh- and low-class workers, high-paying firms may thus pay more in part because ofrent-sharing or higher productivity.

25

+290%

+123%

+49%

+59%

+127%

+140%

+96%

+10%

+40%

+68%

+45%

+157%

+83%

-59%Most peers high-class?

Proportion high-class peers

Value added per worker

Operating profit per worker

Works in the capital?

Works in a large firm?

Is a top earner?

-100 0 100 200 300Relative representation (percent)

Origin: Low income High income

(a) Firms, workplaces, peers

+31%

+64%

+110%

-90%

-90%

-72%

-59%

-16%

+79%

+93%

+40%

-65%

-89%

-24%

-70%

-13%

Agriculture/other

Health, public admin, services

Construction

Education and research

Trade and communication

Manufacturing

Energy

Finance and business

-100 -50 0 50 100Relative representation (percent)

Origin: Low income High income

(b) Industry classification

Figure 7. Characteristics conditional on working in a high-paying firm, by fatherearnings (top vs bottom quintile). Sons only, daughters shown in Appendix FigureA18.

Note however, that unlike most of our results, there are meaningful gender differ-ences between sons and daughters here. Figure A18 in the Appendix shows that whilesuccessful high-class daughters also have disproportionately high-class background co-workers, they are also more likely than successful low-class background workers to workin a large firm, and particularly in a profitable firm. These gender differences showsthat while firms are important for both sons and daughters, the specific types of firmsthat benefit high- and low-class background workers can vary by gender.

Figure 7b shows comparable results by industry.5 High-class sons in high-payingfirms are overrepresented in finance and business, at a rate twice as high as the generalpopulation. For low-class sons in high-paying firms, the picture is different. Theiroverrepresentation in finance is weaker, at 40%. Instead they are more prevalent inmanufacturing, where they appear at a rate 79% higher than the general population(compared to 31% for their high-class peers), and energy extraction.

As such, even when sons from low earnings backgrounds manage to find employ-ment in high-paying firms, they are largely excluded from the kind of firms—financeand business—that draw their recruits from more elite backgrounds. Instead, theydisproportionately obtain jobs at high-paying manufacturing companies and in largefirms. These high-paying firms with many workers from low-class backgrounds maybe more likely to pay high wages due to labor union influence or organizational con-straints than the finance or business firms harboring high-class background workers.These descriptive patterns provide suggestive evidence that the role of firms in foster-

5Industry is based on the 1-digit SNI level. For chart legibility, we further group these categoriesinto 8 by collapsing health care; public administration; and culture and services, which show similarcharacteristics, and agriculture with the “unspecified” category, which are both small.

26

ing or blocking earnings mobility depends on the type of firm and the source of firmpremium. Some high-paying firms mainly provide upward mobility for low-earningsbackground workers; others cement the class position of those already from a privi-leged background.

6 ConclusionSince its origins in the 1960s, status attainment research has consistently focused onways that high-status parents endow children with education and other individualresources valued in the labor market. Although this research program has provenextraordinarily productive, it has also obscured ways that social reproduction is influ-enced by structural features of the labor market. A strong undercurrent of stratifica-tion research has questioned this focus on the individual, hypothesizing labor marketsegmentation as a driving force behind intergenerational stratification (Baron, 1984;Baron and Bielby, 1980; Bibb and Form, 1977; Stolzenberg, 1978; Tomaskovic-Deveyand Avent-Holt, 2019). However, this hypothesis has not been comprehensively tested.

We enter this old debate by noting that the status attainment account of intergen-erational mobility is increasingly in tension with research on the sources of earningsinequality. Recent inequality research emphasizes that, in addition to the persistenteffects of human capital, workers’ pay is also determined by whether they match withhigh- or low-paying firms. New empirical methods identify this important role forunequal firm premiums by controlling for firms’ worker composition. Moreover, re-cent qualitative research suggests that this sorting process can be influenced by socialbackground, whereby high-class job applicants obtain entry to elite firms.

Building on this research, we propose a comprehensive framework that clarifies howstructural inequality in the labor market amplifies intergenerational persistence. Wehighlight four mechanism—inheritance, scaffolding, investment, endowment—by whichparental advantage is passed on to children. Empirically, we find that the endowment ofindividual traits accounts for the majority of intergenerational earnings transmission,consistent with predominant theoretical perspectives. Yet, more than a quarter ofearnings transmission is channeled through the remaining three pathways, all of whichreflect workers sorting across unequal firms.

This firm-related transmission occurs mainly because parents with high individualearnings have children who are more likely to be hired to work in high-paying firms. Incontrast, direct firm or industry inheritance is less important and mainly matters earlyin the child’s career. But even by age 40, children from privileged origins continueto work in higher-paying firms than their less fortunate peers. This pattern couldreflect high-earning parents scaffolding their children’s subsequent sorting, as in recentsociological research on class and hiring: high-class background children have networkties with, and benefit from cultural matching to, their same-class brethren.

Two significant implications follow from these results. First, part of intergenera-tional earnings transmission may be a self-perpetuating cycle, in which high-paying

27

firms are stocked with high-class workers, who then homosocially reproduce and dis-proportionately hire other high-class workers. Indeed, even adjusting for our broadmeasure of persistently valuable individual worker characteristics leaves a substan-tial residual sorting advantage for privileged workers. We do not distinguish betweengeneric cultural matching, as when a high-class background worker seems to be adistinctively good fit in a job interview, from social network ties, that may operatethrough information about job openings or direct referral. Future research should con-sider strategies to isolate the specific means through which high-earning backgroundworkers are sorted to high-paying jobs.

We also expect that the importance of this process could vary across time and place,as the extent of social background sorting into high-premium firms varies. Future re-search may identify labor markets in which this correlation is weaker or interruptedaltogether. Such research could also help distinguish the kinds of firms and the sourcesof firm pay premiums that tend to provide mobility opportunities from the kinds thatlock-in advantages for already privileged workers. As initial descriptive work downthese lines, we show that finance tends to be disproportionately stocked by privilegedbackground workers, while high-paying manufacturing and energy extraction firms pro-vide opportunities for employment for workers of more humble social origins. Successfulworkers from both backgrounds tend to work at high-profit and high value-added firms.

Future research should also assess how the influence of firms on intergenerationalmobility varies with the overall importance of firms in determining workers’ pay. Swe-den is a conservative case here, as firm premiums account for a relatively small share ofearnings inequality compared to other developed countries (Criscuolo et al., 2020). But,more broadly, across most rich countries, earnings inequality is increasingly between-firms (Tomaskovic-Devey et al., 2020), due to a combination of increased positiveworker sorting across firms and persistently important firm pay premiums (Card et al.,2013; Song et al., 2018). In economies in which firms substantially determine workerearnings, we expect firms’ role in intergenerational mobility to be amplified.

The second implication is that introducing firms to intergenerational mobility re-search highlights the transformation of types of labor market advantages across gener-ations. The largest contribution firms make to intergenerational transmission is thatparents with high individual earnings, perhaps due to human capital, secure their chil-dren employment at high-paying firms. The role played by parents’ own employmentat high-paying firms is weaker and concentrated at labor market entry. This suggeststhat parents whose class position rests on human capital have children whose positionrests, in part, on social capital, broadly defined, and the access it affords to high-payingfirms.

On the one hand, this transformation of types of labor market advantage can be away that class hardens from earned to unearned advantage across generations. Parentalhuman capital fosters children’s social advantages. On the other hand, this processgives reason for optimism that firm-related intergenerational transmission could befragile across multiple generations: the advantaged child, whose social position rests

28

on employment at a high-paying firm, may not be able to pass that advantage on tohis or her own children.

ReferencesAbowd, John M., Francis Kramarz, and David N. Margolis. 1999. “High Wage Workers

and High Wage Firms.” Econometrica 67:251–333.

Adler, Paul S. and Seok-Woo Kwon. 2002. “Social Capital: Prospects for a New Con-cept.” The Academy of Management Review 27:17–40.

Anderson, Elijah. 2000. Code of the street: Decency, violence, and the moral life of theinner city . WW Norton & Company.

Andrews, Martyn J, Len Gill, Thorsten Schank, and Richard Upward. 2008. “High wageworkers and low wage firms: negative assortative matching or limited mobility bias?”Journal of the Royal Statistical Society: Series A (Statistics in Society) 171:673–697.

Armstrong, Elizabeth A and Laura T Hamilton. 2013. Paying for the Party . HarvardUniversity Press.

Baron, James N. 1984. “Organizational perspectives on stratification.” Annual Reviewof Sociology 10:37–69.

Baron, James N. and William T. Bielby. 1980. “Bringing the Firms Back in: Stratifica-tion, Segmentation, and the Organization of Work.” American Sociological Review45:737–765.

Barth, Erling, James Davis, and Richard B. Freeman. 2018. “Augmenting the HumanCapital Earnings Equation with Measures of Where People Work.” Journal of LaborEconomics 36:S71–S97.

Becker, Gary S and Nigel Tomes. 1986. “Human capital and the rise and fall of families.”Journal of Labor Economics 4:S1–S39.

Berger, Thor and Per Engzell. 2022. “Industrial automation and intergenerationalincome mobility in the United States.” Social Science Research forthcoming.

Bernardi, Fabrizio and Gabriele Ballarino. 2016. Education, Occupation and SocialOrigin: A comparative analysis of the transmission of socio-economic inequalities .Edward Elgar Publishing.

Bianchi, Suzanne, Philip N Cohen, Sara Raley, and Kei Nomaguchi. 2004. “Inequalityin parental investment in child-rearing: Expenditures, time, and health.” In SocialInequality , edited by Kathryn Neckerman, pp. 189–219. Russell Sage Foundation.

29

Bibb, Robert and William H Form. 1977. “The effects of industrial, occupational, andsex stratification on wages in blue-collar markets.” Social Forces 55:974–996.

Bihagen, Erik. 2008. “Does Class Matter Equally for Men and Women? A Study of theImpact of Class on Wage Growth in Sweden 1999—2003.” Sociology 42:522–540.

Bills, David B, Valentina Di Stasio, and Klarita Gërxhani. 2017. “The demand side ofhiring: employers in the labor market.” Annual Review of Sociology 43:291–310.

Blau, Peter M and Otis Dudley Duncan. 1967. The American occupational structure.New York: John Wiley & Sons.

Bloome, Deirdre. 2015. “Income inequality and intergenerational income mobility inthe United States.” Social Forces 93:1047–1080.

Bloome, Deirdre, Shauna Dyer, and Xiang Zhou. 2018. “Educational inequality, edu-cational expansion, and intergenerational income persistence in the United States.”American Sociological Review 83:1215–1253.

Bol, Thijs, Christina Ciocca Eller, Herman G van de Werfhorst, and Thomas ADiPrete. 2019. “School-to-work linkages, educational mismatches, and labor mar-ket outcomes.” American Sociological Review 84:275–307.

Bourdieu, Pierre. 1984. Distinction: A social critique of the judgement of taste. Har-vard University Press.

Bowles, Samuel and Herbert Gintis. 1976. Schooling in Capitalist America. New York,NY: Basic Books.

Bowles, Samuel, Herbert Gintis, and Melissa Osborne Groves. 2009. Unequal chances:Family background and economic success . Princeton University Press.

Brand, Jennie E. 2015. “The far-reaching impact of job loss and unemployment.”Annual Review of Sociology 41:359–375.

Breen, Richard and Walter Müller. 2020. Education and Intergenerational Social Mo-bility in Europe and the United States . Stanford University Press.

Calarco, Jessica McCrory. 2018. Negotiating Opportunities: How the middle classsecures advantages in school . Oxford University Press.

Card, David, Jörg Heining, and Patrick Kline. 2013. “Workplace Heterogeneity and theRise of West German Wage Inequality.” Quarterly Journal of Economics 128:967–1015.

Cesarini, David, Erik Lindqvist, Robert Östling, and Björn Wallace. 2016. “Wealth,health, and child development: Evidence from administrative data on Swedish lotteryplayers.” Quarterly Journal of Economics 131:687–738.

30

Cheng, Siwei and Xi Song. 2019. “Linked Lives, Linked Trajectories: IntergenerationalAssociation of Intragenerational Income Mobility.” American Sociological Review84:1037–1068.

Coleman, James, Ernest Campbell, Carol Hobson, James McPartland, AlexanderMood, Frederic Weinfeld, and Robert York. 1967. Equality of Educational Opportu-nity . Washington, D. C.: US Government Printing Office.

Cooney, Teresa M and Peter Uhlenberg. 1992. “Support from parents over the lifecourse: The adult child’s perspective.” Social Forces 71:63–84.

Cooper, Kerris and Kitty Stewart. 2020. “Does household income affect children’soutcomes? A systematic review of the evidence.” Child Indicators Research pp. 1–25.

Corak, Miles and Patrizio Piraino. 2011. “The Intergenerational Transmission of Em-ployers.” Journal of Labor Economics 29:37–68.

Correia, Sergio. 2019. “REGHDFE: Stata module to perform linear or instrumental-variable regression absorbing any number of high-dimensional fixed effects.” .

Criscuolo, Chiara, Alexander Hijzen, Cyrille Schwellnus, Erling Barth, Wen-Hao Chen,Richard Fabling, Priscilla Fialho, Balazs Stadler, Richard Upward, Wouter Zwysen,Katarzyna Grabska-Romagosa, Ryo Kambayashi, Timo Leidecker, Oskar NordströmSkans, Capucine Riom, and Duncan Roth. 2020. “Workforce composition, produc-tivity and pay: the role of firms in wage inequality.” .

Dahl, Molly W and Thomas DeLeire. 2008. “The association between children’s earn-ings and fathers’ lifetime earnings: estimates using administrative data.” Universityof Wisconsin-Madison, Institute for Research on Poverty.

Danziger, Sheldon and Jane Waldfogel. 2000. Securing the future: Investing in childrenfrom birth to college. Russell Sage Foundation.

Duncan, Otis Dudley. 1966. “Path analysis: Sociological examples.” American Journalof Sociology 72:1–16.

Engbom, Niklas, Christian Moser, and Jan Sauerman. 2021. “Firm Pay Dynamics.”Columbia Business School Research Paper.

Engzell, Per and Carina Mood. 2021. “How Robust are Estimates of IntergenerationalIncome Mobility?” SocArXiv, http://osf.io/preprints/socarxiv/gd2t6.

Engzell, Per and Felix C Tropf. 2019. “Heritability of education rises with intergenera-tional mobility.” Proceedings of the National Academy of Sciences 116:25386–25388.

31

Faas, Caitlin, Mark J Benson, and Christine E Kaestle. 2013. “Parent resources duringadolescence: Effects on education and careers in young adulthood.” Journal of YouthStudies 16:151–171.

Farkas, George. 2003. “Cognitive skills and noncognitive traits and behaviors in strat-ification processes.” Annual Review of Sociology 29:541–562.

Featherman, David L. 1981. “Social stratification and mobility: two decades of cumu-lative social science.” American Behavioral Scientist 24:364–385.

Fiel, Jeremy E. 2020. “Great equalizer or great selector? Reconsidering education as amoderator of intergenerational transmissions.” Sociology of Education 93:353–371.

Fox, Liana, Florencia Torche, and Jane Waldfogel. 2016. “Intergenerational mobility.”In The Oxford Handbook of the Social Science of Poverty , edited by David Bradyand Linda M Burton.

Friedman, Sam and Daniel Laurison. 2020. The Class Ceiling: Why it Pays to bePrivileged . Bristol: Policy Press.

Gerstl, Joel E and Lois K Cohen. 1964. “Dissensus, situs and egocentrism in occupa-tional ranking.” The British Journal of Sociology 15:254–261.

Granovetter, Mark. 1995. Getting a Job: A study of contacts and careers . Universityof Chicago Press, second edition.

Groshen, Erica L. 1991. “Sources of Intra-Industry Wage Dispersion: How Much DoEmployers Matter?” Quarterly Journal of Economics 106:869–884.

Hällsten, Martin. 2010. “The structure of educational decision making and consequencesfor inequality: A Swedish test case.” American Journal of Sociology 116:806–54.

Hogan, Dennis P, David J Eggebeen, and Clifford C Clogg. 1993. “The structure ofintergenerational exchanges in American families.” American Journal of Sociology98:1428–1458.

Horowitz, Jonathan. 2018. “Relative education and the advantage of a college degree.”American Sociological Review 83:771–801.

Hout, Michael. 2012. “Social and economic returns to college education in the UnitedStates.” Annual Review of Sociology 38:379–400.

Jæger, Mads Meier and Richard Breen. 2016. “A dynamic model of cultural reproduc-tion.” American Journal of Sociology 121:1079–1115.

Jencks, Christopher. 1980. “Heredity, environment, and public policy reconsidered.”American Sociological Review pp. 723–736.

32

Jencks, Christopher et al. 1972. Inequality: A reassessment of the effect of family andschooling in America. New York: Harper & Row.

Jencks, Christopher et al. 1979. Who gets ahead? The determinants of economicsuccess in America. New York: Basic Books.

Jonsson, Jan O, David B Grusky, Matthew Di Carlo, Reinhard Pollak, and Mary CBrinton. 2009. “Microclass mobility: Social reproduction in four countries.” AmericanJournal of Sociology 114:977–1036.

Kalleberg, Arne L., Michael Wallace, and Robert P. Althauser. 1981. “Economic Seg-mentation, Worker Power, and Income Inequality.” American Journal of Sociology87:651–683.

Karlson, Kristian Bernt. 2019. “College as equalizer? Testing the selectivity hypothesis.”Social Science Research 80:216–229.

Kaushal, Neeraj, Katherine Magnuson, and Jane Waldfogel. 2011. “How is familyincome related to investments in children’s learning?” In Whither Opportunity?Rising Inequality, Schools, and Children’s Life Chances , edited by Greg J Duncanand Richard Murnane, pp. 187–205. Russell Sage Foundation.

Khan, Shamus R. 2010. Privilege: The Making of an Adolescent Elite at St. Paul’sSchool . The William G. Bowen Series. Princeton University Press.

Kline, Patrick, Raffaele Saggio, and Mikkel Sølvsten. 2020. “Leave-out estimation ofvariance components.” Econometrica 88:1859–1898.

Kornrich, Sabino and Frank Furstenberg. 2013. “Investing in children: Changes inparental spending on children, 1972–2007.” Demography 50:1–23.

Kramarz, Francis and Oskar Nordström Skans. 2014. “When Strong Ties are Strong:Networks and Youth Labour Market Entry.” The Review of Economic Studies81:1164–1200.

Krolikowski, Pawel, Mike Zabek, and Patrick Coate. 2020. “Parental proximity andearnings after job displacements.” Labour Economics 65:101877.

Lachowska, Marta, Alexandre Mas, Raffaele D Saggio, and Stephen A Woodbury. 2020.“Do firm effects drift? Evidence from Washington administrative data.” Technicalreport, National Bureau of Economic Research.

Lareau, Annette. 2011. Unequal Childhoods: Class, Race, and Family Life. Universityof California Press.

Lareau, Annette. 2015. “Cultural knowledge and social inequality.” American Socio-logical Review 80:1–27.

33

Laurison, Daniel and Sam Friedman. 2016. “The Class Pay Gap in Higher Professionaland Managerial Occupations.” American Sociological Review 81:668–695.

Manzoni, Anna. 2018. “Parental support and youth occupational attainment: Help orhindrance?” Journal of Youth and Adolescence 47:1580–1594.

Mayer, Susan E. 2001. “How did the increase in economic inequality between 1970and 1990 affect children’s educational attainment?” American Journal of Sociology107:1–32.

McPherson, Miller, Lynn Smith-Lovin, and James M Cook. 2001. “Birds of a Feather:Homophily in Social Networks.” Annual Review of Sociology 27:415–444.

Mijs, Jonathan J. B. and Elizabeth L. Roe. 2021. “Is America coming apart? So-cioeconomic segregation in neighborhoods, schools, workplaces, and social networks,1970–2020.” Sociology Compass 15:e12884.

Monk, Ellis P, Michael H Esposito, and Hedwig Lee. 2021. “Beholding Inequality: Race,Gender, and Returns to Physical Attractiveness in the United States.” AmericanJournal of Sociology 127:194–241.

Morgan, Stephen L, David B Grusky, and Gary S Fields. 2006. Mobility and inequality:frontiers of research in sociology and economics . Stanford University Press.

Morris, Richard T and Raymond J Murphy. 1959. “The situs dimension in occupationalstructure.” American Sociological Review pp. 231–239.

Mouw, Ted. 2006. “Estimating the Causal Effect of Social Capital: A Review of RecentResearch.” Annual Review of Sociology 32:79–102.

Newman, Katherine S. 2009. No shame in my game: The working poor in the innercity . Vintage.

Nybom, Martin and Jan Stuhler. 2017. “Biases in standard measures of intergenera-tional income dependence.” Journal of Human Resources 52:800–825.

Owens, Ann. 2016. “Inequality in children’s contexts: Income segregation of householdswith and without children.” American Sociological Review 81:549–574.

Pfeffer, Fabian T and Martin Hällsten. 2012. “Mobility regimes and parental wealth:The United States, Germany, and Sweden in comparison.” SOEPpapers on Multi-disciplinary Panel Data Research, DIW Berlin.

Portes, Alejandro and Julia Sensenbrenner. 1993. “Embeddedness and Immigration:Notes on the Social Determinants of Economic Action.” American Journal of Soci-ology 98:1320–1350.

34

Reardon, Sean F and Kendra Bischoff. 2011. “Income inequality and income segrega-tion.” American Journal of Sociology 116:1092–1153.

Rivera, Lauren A. 2012. “Hiring as Cultural Matching: The Case of Elite ProfessionalService Firms.” American Sociological Review 77:999–1022.

Rivera, Lauren A. 2020. “Employer decision making.” Annual Review of Sociology46:215–232.

Rüttenauer, Tobias and Volker Ludwig. 2020. “Fixed effects individual slopes: Account-ing and testing for heterogeneous effects in panel data or other multilevel models.”Sociological Methods & Research .

Sage, Rayna Amber and Monica Kirkpatrick Johnson. 2012. “Extending and expandingparenthood: Parental support to young adult children.” Sociology Compass 6:256–270.

Schneider, Daniel, Orestes P. Hastings, and Joe LaBriola. 2018. “Income Inequality andClass Divides in Parental Investments.” American Sociological Review 83:475–507.

Schoeni, Robert F and Karen E Ross. 2005. “Material Assistance from Families duringthe Transition to Adulthood.” In On the frontier of adulthood: Theory, research,and public policy , edited by Robert F Schoeni, Karen E Ross, Richard A Settersten,Frank F Furstenberg, and Rubén G Rumbaut, pp. 396–416. University of ChicagoPress Chicago, IL.

Sewell, William H, Archibald O Haller, and Alejandro Portes. 1969. “The educationaland early occupational attainment process.” American Sociological Review pp. 82–92.

Sewell, William H and Robert M Hauser. 1975. Education, Occupation, and Earnings.Achievement in the Early Career. Academic Press.

Slichter, Sumner H. 1950. “Notes on the Structure of Wages.” The Review of Economicsand Statistics 32:80–91.

Song, Jae, David J Price, Fatih Guvenen, Nicholas Bloom, and Till von Wachter. 2018.“Firming Up Inequality.” Quarterly Journal of Economics 134:1–50.

Spilerman, Seymour. 1977. “Careers, labor market structure, and socioeconomicachievement.” American Journal of Sociology 83:551–593.

Staiger, Matthew. 2021. “The Intergenerational Transmission of Employers and theEarnings of Young Workers.” Equitable Growth Working Paper.

Stolzenberg, Ross M. 1975. “Occupations, labor markets and the process of wageattainment.” American Sociological Review pp. 645–665.

35

Stolzenberg, Ross M. 1978. “Bringing the boss back in: Employer size, employee school-ing, and socioeconomic achievement.” American Sociological Review pp. 813–828.

Swartz, Teresa Toguchi. 2009. “Intergenerational family relations in adulthood: Pat-terns, variations, and implications in the contemporary United States.” AnnualReview of Sociology 35:191–212.

Swartz, Teresa Toguchi, Minzee Kim, Mayumi Uno, Jeylan Mortimer, andKirsten Bengtson O’Brien. 2011. “Safety nets and scaffolds: Parental support inthe transition to adulthood.” Journal of Marriage and Family 73:414–429.

Swartz, Teresa Toguchi, Heather McLaughlin, and Jeylan T Mortimer. 2017. “Parentalassistance, negative life events, and attainment during the transition to adulthood.”The Sociological Quarterly 58:91–110.

Toft, Maren and Sam Friedman. 2021. “Family wealth and the class ceiling: the propul-sive power of the bank of Mum and Dad.” Sociology 55:90–109.

Tolbert, Charles M., II. 1983. “Industrial Segmentation and Men’s IntergenerationalMobility.” Social Forces 61:1119–1137.

Tomaskovic-Devey, Donald and Dustin Avent-Holt. 2019. Relational Inequalities: AnOrganizational Approach. Oxford University Press.

Tomaskovic-Devey, Donald, Anthony Rainey, Dustin Avent-Holt, Nina Bandelj, IstvánBoza, David Cort, Olivier Godechot, Gergely Hajdu, Martin Hällsten, Lasse FolkeHenriksen, et al. 2020. “Rising between-workplace inequalities in high-income coun-tries.” Proceedings of the National Academy of Sciences 117:9277–9283.

Torche, Florencia. 2011. “Is a college degree still the great equalizer? Intergenera-tional mobility across levels of schooling in the United States.” American Journal ofSociology 117:763–807.

Torche, Florencia. 2015. “Analyses of intergenerational mobility: An interdisciplinaryreview.” The ANNALS of the American Academy of Political and Social Science657:37–62.

Torche, Florencia. 2016. “Education and the intergenerational transmission of advan-tage in the US.” In Education, Occupation and Social Origin, edited by FabrizioBernardi and Gabriele Ballarino. Edward Elgar Publishing.

Weeden, Kim A. and David B. Grusky. 2012. “The Three Worlds of Inequality.” Amer-ican Journal of Sociology 117:1723–1785.

Wilmers, Nathan. 2018. “Wage Stagnation and Buyer Power: How Buyer-SupplierRelations Affect U.S. Workers’ Wages, 1978 to 2014.” American Sociological Review83:213–242.

36

Witteveen, Dirk and Paul Attewell. 2017. “Family background and earnings inequalityamong college graduates.” Social Forces 95:1539–1576.

Witteveen, Dirk and Paul Attewell. 2020. “Reconsidering the ‘meritocratic power of acollege degree’.” Research in Social Stratification and Mobility 66:100479.

Zhou, Xiang. 2019. “Equalization or selection? Reassessing the “meritocratic power” ofa college degree in intergenerational income mobility.” American Sociological Review84:459–485.

37

Appendix

A1 Mothers’ earnings

In our main analysis we measure parental earnings by focusing on fathers. In thissection, we replicate our results using mother’s earnings. The results for our specificargument about the role of firms are quite consistent across the choice of reference par-ent. There are other differences though, that likely reflect gender-specific transmissionmechanisms. Specifically, sons have a stronger earnings transmission from fathers thanfrom mothers, while daughters have a similar level of transmission from both parents(compare Table A3 and Table 2). However, this absolute difference does not much af-fect the share of transmission accounted for by firms, or its development across careerstages (see Figure A1). If anything, focusing on mothers and sons, rather than fathersand sons, further weakens the firm-to-firm transmission path for sons. But, our generalconclusion about the share of transmission accounted for by firm-related paths, andthe dominant role of the worker-to-firm path specifically, holds across choice of parent.

Table A1: Correlation matrix: sons. Mother’s earnings.

Variables C earnings C firm C worker F earnings F firm F workerC earnings 1.000C firm 0.573 1.000C worker 0.842 0.294 1.000M earnings 0.166 0.092 0.167 1.000M firm 0.062 0.077 0.045 0.297 1.000M worker 0.170 0.078 0.178 0.869 0.053 1.000

Table A2: Correlation matrix: daughters. Mother’s earnings.

Variables C earnings C firm C worker F earnings F firm F workerC earnings 1.000C firm 0.462 1.000C worker 0.817 0.230 1.000M earnings 0.193 0.136 0.198 1.000M firm 0.063 0.122 0.039 0.300 1.000M worker 0.204 0.124 0.219 0.869 0.058 1.000

38

Table A3: Decomposition parameters at child age 38–42. Mother’s earnings.

Estimate (percent)

Parameter Sons Daughtersfirm-to-firm = ρff · a1 · a2 0.007 (4.58%) 0.009 (4.73%)

worker-to-firm = ρwf · b1 · a2 0.024 (15.8%) 0.031 (16.3%)firm-to-worker = ρfw · a1 · b2 0.008 (5.53%) 0.007 (3.96%)

worker-to-worker = ρww · b1 · b2 0.112 (74.0%) 0.140 (75.0%)ff + wf + fw + ww = corr(Y ∗

i , Yi) 0.151 (100%) 0.187 (100%)ρff = corr(F ∗

i , Fi) 0.077 0.122ρwf = corr(W ∗

i , Fi) 0.078 0.124ρfw = corr(F ∗

i ,Wi) 0.045 0.039ρww = corr(W ∗

i ,Wi) 0.178 0.219a1 = corr(F ∗

i , Y∗i | W ∗

i ) 0.251 0.250b1 = corr(W ∗

i , Y∗i | F ∗

i ) 0.855 0.854a2 = corr(Fi, Yi | Wi) 0.357 0.289b2 = corr(Wi, Yi | Fi) 0.737 0.751

0

.05

.1

.15

.2

Ran

k co

rrela

tion

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A1. Decomposition of the intergenerational earnings correlation over the earlycareer, by sex and age of child. Mother’s earnings.

39

A2 Log earnings

Much prior work using two-way fixed effects earnings models has predicted loggedearnings as the dependent variable: the variance of logged earnings is a usefully de-composable measure of labor market earnings inequality. Because in this paper weaddress the literature on intergenerational mobility, rather than earnings inequality,we focus in the main text on studying rank transformed earnings. But here we showresults using log income to estimate the log-log correlation in earnings. This measureis closely associated with the standard intergenerational earnings elasticity, but adjustsfor potential differences in dispersion between the parent’s and child’s generation (Foxet al., 2016).

Results are similar with this alternative transformation of the dependent variable.The overall intergenerational earnings correlation is somewhat stronger for sons withthe log-log approach, and somewhat stronger for daughters with rank. Firm-relatedcomponents matter proportionately slightly more in the logs, due to weaker worker-to-worker transmission for both sons and daughters (compare Table A6 and Table 2).But the absolute correlations for the firm components are nearly identical across thesespecifications. Likewise, Figure A2 shows that trajectories over the career in the log-logapproach are very similar to those in the rank approach in the main text.

Table A4: Correlation matrix: sons. Log earnings.

Variables C earnings C firm C worker F earnings F firm F workerC earnings 1.000C firm 0.509 1.000C worker 0.765 0.090 1.000F earnings 0.209 0.116 0.196 1.000F firm 0.048 0.087 0.019 0.412 1.000F worker 0.215 0.093 0.210 0.838 -0.015 1.000

Table A5: Correlation matrix: daughters. Log earnings.

Variables C earnings C firm C worker F earnings F firm F workerC earnings 1.000C firm 0.433 1.000C worker 0.742 0.097 1.000F earnings 0.170 0.123 0.158 1.000F firm 0.029 0.048 0.016 0.417 1.000F worker 0.181 0.118 0.171 0.836 -0.015 1.000

40

Table A6: Decomposition parameters at child age 38–42. Log earnings.

Estimate (percent)

Parameter Sons Daughtersfirm-to-firm = ρff · a1 · a2 0.016 (8.89%) 0.007 (4.95%)

worker-to-firm = ρwf · b1 · a2 0.035 (18.7%) 0.036 (24.1%)firm-to-worker = ρfw · a1 · b2 0.006 (3.10%) 0.005 (3.29%)

worker-to-worker = ρww · b1 · b2 0.129 (69.3%) 0.102 (67.7%)ff + wf + fw + ww = corr(Y ∗

i , Yi) 0.185 (100%) 0.151 (100%)ρff = corr(F ∗

i , Fi) 0.087 0.048ρwf = corr(W ∗

i , Fi) 0.093 0.118ρfw = corr(F ∗

i ,Wi) 0.019 0.016ρww = corr(W ∗

i ,Wi) 0.210 0.171a1 = corr(F ∗

i , Y∗i | W ∗

i ) 0.425 0.430b1 = corr(W ∗

i , Y∗i | F ∗

i ) 0.844 0.843a2 = corr(Fi, Yi | Wi) 0.443 0.365b2 = corr(Wi, Yi | Fi) 0.725 0.706

-.05

0

.05

.1

.15

Cor

rela

tion

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A2. Decomposition of the intergenerational earnings correlation over the earlycareer, by sex and age of child. Log earnings.

41

A3 Alternative age-earnings profiles

In our main analysis, we account for population heterogeneity in age-earnings profilesby estimating flexibly specified trajectories that depend on an interaction between gen-der and four levels of education. Here, we extend this analysis by allowing even greaterflexibility in the modeling of earnings over the life course. First, we let trajectoriesdepend on up to 47 levels of education (Figure A3), as well as 116 detailed fields ofstudy (Figure A4). This makes little difference to our results.

Second, we break our estimation sample up into distinct 5-year periods. Thisallows each worker and firm to be associated with a number of fixed effects throughoutthe observation window and thus allows for more flexible trajectories workers as wellas varying firm premiums (Figure A5). Again, the picture is similar although firmsaccount for a slightly higher proportion of intergenerational earnings transmission.

Third, we extend the canonical specification with one that allows for individualslopes by age as part of the worker component (Correia, 2019; Rüttenauer and Ludwig,2020). This model can be written:

Yit = αi + γiageit +X ′itβ + ψj(i,t) + uit. (A1)

Note the individual subscript i on both the intercept αi and the age-earnings slopeγi. This specification arguably risks overfitting individual trajectories and is liable tosoak up some of the effects due to varying firm premiums over the career. Neverthe-less, even in this stringent specification, firm-based mechanisms account for a fifth ofearnings transmission at the height of the career (Figure A6).

42

0

.05

.1

.15

.2R

ank

corre

latio

n

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A3. Decomposition of the intergenerational earnings correlation by sex and ageof child, detailed levels of education.

0

.05

.1

.15

.2

Ran

k co

rrela

tion

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A4. Decomposition of the intergenerational earnings correlation by sex and ageof child, detailed fields of study.

43

0

.05

.1

.15

.2R

ank

corre

latio

n

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A5. Decomposition of the intergenerational earnings correlation by sex and ageof child, distinct 5-year periods.

0

.05

.1

.15

.2

Ran

k co

rrela

tion

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A6. Decomposition of the intergenerational earnings correlation by sex and ageof child, fixed effects individual slopes.

44

A4 Path model assumptions

In this section, we provide more detail on the assumptions of the path decomposition,and how our estimates will be affected by potential violations thereof. The first setof assumptions that we stipulate to arrive at Equation 11 in the main text is a set oforthogonality conditions:

Y ∗i = a1F

∗i + b1W

∗i + ϵ1i, ϵ1i ⊥ F ∗

i ,W∗i (A2)

Yit = a2tFit + b2tWit + ϵ2it, ϵ2it ⊥ Fit, Wit (A3)Fit = a3tF

∗i + b3tW

∗i + ϵ3it, ϵ3it ⊥ F ∗

i ,W∗i (A4)

Wit = a4tF∗i + b4tW

∗i + ϵ4it, ϵ4it ⊥ F ∗

i ,W∗i (A5)

Equations A2–A5 describe the standard least squares regression assumptions: thatthe regressors of each equation are orthogonal to the residual term of their respectiveequation. The graph in Figure 1 implies a further set of constraints:

E(Fitϵ1i) = E(Witϵ1i) = 0, (A6)E(F ∗

i ϵ2it) = E(W ∗i ϵ2it) = 0, (A7)

E(ϵ1iϵ2it) = 0. (A8)

Equations A6–A7 state that the residual in the father’s (child’s) earnings equationis orthogonal to the child’s (father’s) firm and worker components, while Equation A8states that the two residuals are orthogonal to each other. Given these conditions,it follows from Duncan (1966) that the intergenerational earnings correlation can bedecomposed into the four discrete paths that capture our mechanisms of interest.

What happens if these conditions break down? Equations A2–A5 cannot be testedand have to be accepted by fiat, as in any least squares regression. Equations A6–A7 are testable and violations are easy to incorporate by adding five terms to ourdecomposition:6

ρt = ρff,t · a1 · a2t︸ ︷︷ ︸firm-to-firm

+ ρwf,t · b1 · a2t︸ ︷︷ ︸worker-to-firm

+ ρfw,t · a1 · b2t︸ ︷︷ ︸firm-to-worker

+ ρww,t · b1 · b2t︸ ︷︷ ︸worker-to-worker

(A9)

+ corr(ϵ1i, Fit)a2t︸ ︷︷ ︸residual-to-firm

+ corr(ϵ1i,Wit)b2t︸ ︷︷ ︸residual-to-worker

+ corr(F ∗i , ϵ2it)a1︸ ︷︷ ︸

firm-to-residual

+ corr(W ∗i , ϵ2it)b1︸ ︷︷ ︸

worker-to-residual

+ corr(ϵ1i, ϵ2it)︸ ︷︷ ︸residual-to-residual

.

In Figure A7 below, we relax the constraints in Equation A6–A7 to provide a morecomplete picture of the contours of transmission. Most important of the additionalterms from Equation A9 are the worker-to-residual one, followed by residual-to-worker,and residual-to-residual. The terms for firm-to-residual and residual-to-firm turn out

6We thank Richard Breen for helping us elaborate this point.

45

to be negligible. In other words, our main model does a good job at capturing firm-based transmission, but there are age-specific patterns of individual correlations thatare lost. Specifically, this analysis reveals that, had it not been for the firm componentof child earnings, the intergenerational earnings correlation would be negative for bothsons and daughters at age 25. Nevertheless, our qualitative results remain similar:throughout most of the early career, firm-mediated components account for between afourth and a fifth of the total intergenerational earnings correlation.

46

-.1

0

.1

.2

.3R

ank

corre

latio

n

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firmFirm-to-workerWorker-to-worker

Residual-to-firmResidual-to-workerFirm-to-residualWorker-to-residual

Residual-to-residual

(a)

-100

-50

0

50

100

Perc

ent o

f tra

nsm

issi

on

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firmFirm-to-workerWorker-to-worker

Residual-to-firmResidual-to-workerFirm-to-residualWorker-to-residual

Residual-to-residual

(b)

Figure A7. Residual decomposition of the intergenerational earnings correlation by sexand age of child. Absolute numbers top, relative numbers bottom.

47

A5 Additional tables and figures

In the remainder of the Appendix we show a set of additional results using the samespecification as in the main text, that is, using father’s earnings as the right-hand sidevariable; inspecting rank correlations; and working from the age, gender, educationinteractions in the fixed effects model.

Table A7 and Table A8 show descriptive statistics for the main variables we use inthe analysis, for sons and daughters respectively. Annual arnings are in 1990 SEK anddivided by 100.

Tables A9 and A10 show the correlation matrix for sons and daughters, giving theunderlying correlations used in the path model results in Table 2.

Figure A8 plots child’s earnings rank by father’s earnings rank, without dividingearnings into firm and worker components (as in Figure 2). Figure A9 divides both thechild’s and father’s earnings into worker and firm components. It shows that while theworker-to-worker correlations are the steepest association, worker-to-firm correlationsare also quite high.

Figure A10 depicts variation in the overall intergenerational earnings correlationacross career stages, echoing the pattern framing the decomposition in Figure 3. Fig-ure A11 shows similar correlations, broken out by each generation’s firm and workercomponents. These correlations are scaled by the shifting role of firm and worker com-ponents in determining earnings, to produce the path model-based decomposition inFigure 3. Finally, Figure A12 shows a relative version of Figure 3, in which each pathsums to the total intergenerational correlation for each age year. Figure A12 showsmore clearly the decline in the relative role of firm-to-firm transmission after labor mar-ket entry. It also shows that despite the large rise in worker-to-worker transmissionover the career, the worker-to-firm path retains a fairly stable share of overall earningstransmission.

Next, we provide further details on the mechanism tests we conduct in the mainpaper. Figures A13 and A14 show, for sons and daughters respectively, how the splitsample analyses we use in the paper affect each transmission pathway. In Figure 5 and4 we focus in on the specific pathways that we try to divide into mechanisms usingthese tests.

Figures A15, A16 and A17 all provide results on alternative mediators to those wefocus on in Figure 6. Note that the role of co-worker parental earnings in mediatingthe overall intergenerational earnings correlation rivals that of education, and accountsfor substantially more of the worker-to-firm pathway.

Figure A18 shows the high-paying firm characteristics across high and low incomebackgrounds for daughters. As noted in the main text, most patterns look similaracross sons and daughters here. But the large firm advantage for low class backgroundworkers reverses for daughters and high class background successful daughters are morelikely than low class background successful daughters to work at a high profit firm.

48

Table A7: Descriptive table: sons

N Mean SDChild earnings 109,752 2190.061 1018.928Father earnings 109,752 1774.854 1015.381Same firm 109,752 0.05 0.218Same industry 109,752 0.236 0.425Same municipality 109,752 0.391 0.488Same county 109,752 0.64 0.48Education Freq. Percent Cum.Less than HS 5,625 5.13 5.13High school 54,963 50.08 55.2Some college 5,732 5.22 60.43College 41,479 37.79 98.22Above college 1,953 1.78 100Total 109,752 100IndustryUnspecified 394 0.36 0.36Agriculture 1,168 1.06 1.42Manufacturing 21,688 19.76 21.18Energy 1,813 1.65 22.84Construction 12,358 11.26 34.1Trade and communication 22,629 20.62 54.71Finance and business 24,570 22.39 77.1Education and research 6,978 6.36 83.46Health care 5,995 5.46 88.92Culture and services 5,327 4.85 93.78Public administration 6,832 6.22 100Total 109,752 100

49

Table A8: Descriptive table: daughters

N Mean SDChild earnings 103,254 1620.558 621.753Father earnings 103,254 1773.664 1017.398Same firm 103,254 0.029 0.167Same industry 103,254 0.167 0.373Same municipality 103,254 0.37 0.483Same county 103,254 0.626 0.484Education Freq. Percent Cum.Less than HS 2,976 2.88 2.88High school 36,889 35.73 38.61Some college 5,234 5.07 43.68College 56,445 54.67 98.34Above college 1,710 1.66 100Total 103,254 100Industry Freq. Percent Cum.Unspecified 490 0.47 0.47Agriculture 480 0.46 0.94Manufacturing 7,894 7.65 8.58Energy 817 0.79 9.38Construction 1,431 1.39 10.76Trade and communication 14,231 13.78 24.54Finance and business 15,948 15.45 39.99Education and research 20,798 20.14 60.13Health care 25,287 24.49 84.62Culture and services 6,529 6.32 90.95Public administration 9,349 9.05 100Total 103,254 100

50

Table A9: Correlation matrix: sons.

Variables C earnings C firm C worker F earnings F firm F workerC earnings 1.000C firm 0.572 1.000C worker 0.843 0.293 1.000F earnings 0.243 0.150 0.243 1.000F firm 0.075 0.120 0.048 0.445 1.000F worker 0.250 0.126 0.263 0.872 0.098 1.000

Table A10: Correlation matrix: daughters.

Variables C earnings C firm C worker F earnings F firm F workerC earnings 1.000C firm 0.460 1.000C worker 0.817 0.228 1.000F earnings 0.197 0.156 0.199 1.000F firm 0.044 0.070 0.031 0.449 1.000F worker 0.210 0.150 0.218 0.871 0.100 1.000

51

40

50

60

70

Chi

ld in

com

e ra

nk

0 20 40 60 80 100Father earnings rank

(a) Sons

40

50

60

70

Chi

ld in

com

e ra

nk

0 20 40 60 80 100Father earnings rank

(b) Daughters

Figure A8. Average child earnings rank by father’s earnings rank.

52

40

50

60

70

Chi

ld ra

nk

0 20 40 60 80 100Father firm rank

Firm Worker

(a) Sons

40

50

60

70

Chi

ld ra

nk

0 20 40 60 80 100Father firm rank

Firm Worker

(b) Daughters

40

50

60

70

Chi

ld ra

nk

0 20 40 60 80 100Father worker rank

Firm Worker

(c) Sons

40

50

60

70

Chi

ld ra

nk

0 20 40 60 80 100Father worker rank

Firm Worker

(d) Daughters

Figure A9. Average child firm and worker rank by father’s firm and worker rank.

53

Sons

Daughters

0

.05

.1

.15

.2

.25

Ran

k co

rrela

tion

25 30 35 40 45Child age

Figure A10. Intergenerational rank correlation in earnings by sex and age of child.Point estimates and 95% confidence intervals.

54

Sons

Daughters

0

.05

.1

.15

.2

.25

Ran

k co

rrela

tion

25 30 35 40 45Child age

(a) Firm-to-firm: ρff

Sons

Daughters

0

.05

.1

.15

.2

.25

Ran

k co

rrela

tion

25 30 35 40 45Child age

(b) Worker-to-firm: ρwf

SonsDaughters0

.05

.1

.15

.2

.25

Ran

k co

rrela

tion

25 30 35 40 45Child age

(c) Firm-to-worker: ρfw

Sons

Daughters

0

.05

.1

.15

.2

.25

Ran

k co

rrela

tion

25 30 35 40 45Child age

(d) Worker-to-worker: ρww

Figure A11. Path coefficients over the early career, by sex and age of child. Pointestimates and 95% confidence intervals.

55

0

20

40

60

80

100

Perc

ent o

f tra

nsm

issi

on

Sons Daughters25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Firm-to-firmWorker-to-firm

Firm-to-workerWorker-to-worker

Figure A12. Decomposition of the intergenerational earnings correlation by sex andage of child, main specification, relative numbers.

56

+408%

+121%

+49%

+73%

+29%

-22%

-38%

-44%

-46%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .2 .4 .6 .8Coefficient

Yes No

(a) Firm-to-firm: ρff

-44%

+39%

+58%

+18%

+8%

+4%

-10%

-9%

-6%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .1 .2 .3 .4Coefficient

Yes No

(b) Worker-to-firm: ρwf

+29%

+52%

+76%

+42%

+14%

-1%

-16%

-20%

-10%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .1 .2 .3 .4Coefficient

Yes No

(c) Firm-to-worker: ρfw

+50%

+39%

+35%

+19%

+9%

-2%

-11%

-10%

-8%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .1 .2 .3 .4Coefficient

Yes No

(d) Worker-to-worker: ρww

Figure A13. Coefficient change for path coefficients split by sample. Point estimatesand 95% confidence intervals. Sons only, daughters shown in Figure A14.

57

+530%

+210%

+150%

+79%

+32%

-16%

-41%

-43%

-42%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .2 .4 .6 .8Coefficient

Yes No

(a) Firm-to-firm: ρff

+8%

+31%

+38%

+31%

+17%

+1%

-8%

-15%

-17%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .1 .2 .3 .4Coefficient

Yes No

(b) Worker-to-firm: ρwf

-25%

+18%

+53%

+43%

+18%

+1%

-13%

-16%

+1%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .1 .2 .3 .4Coefficient

Yes No

(c) Firm-to-worker: ρfw

+26%

+53%

+54%

+15%

+4%

-1%

-9%

-7%

+1%

Baseline

Same firm?

Same industry?

Same industry?(Different firm)

Same municipality?

Same county?

0 .1 .2 .3 .4Coefficient

Yes No

(d) Worker-to-worker: ρww

Figure A14. Coefficient change for path coefficients split by sample. Point estimatesand 95% confidence intervals. Daughters only, sons shown in Figure A13.

58

-31%

-8%

-3%

-2%

-11%

-35%

-33%

-85%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(a) Sons

-36%

-12%

-7%

-2%

-12%

-34%

-34%

-83%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(b) Daughters

Figure A15. Coefficient change for the intergenerational earnings correlation addingmediators. Point estimates and 95% confidence intervals.

59

-5%

-29%

-8%

-13%

-9%

-25%

-100%

-13%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(a) Firm-to-firm: ρff

-29%

-5%

+2%

-3%

-20%

-74%

-100%

-61%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(b) Worker-to-firm: ρwf

-22%

-20%

-9%

-3%

-14%

-52%

-75%

-100%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(c) Firm-to-worker: ρfw

-33%

-6%

-3%

+0%

-8%

-26%

-14%

-100%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(d) Worker-to-worker: ρww

Figure A16. Coefficient change for path coefficients adding mediators. Point estimatesand 95% confidence intervals. Sons only, daughters shown in Figure A17.

60

-3%

-34%

-11%

-4%

-19%

-42%

-100%

-10%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(a) Firm-to-firm: ρff

-18%

-21%

-13%

-8%

-28%

-75%

-100%

-30%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(b) Worker-to-firm: ρwf

-21%

-17%

-9%

-3%

-19%

-52%

-52%

-100%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(c) Firm-to-worker: ρfw

-43%

-8%

-4%

-1%

-9%

-25%

-14%

-100%

Baseline

Education

Industry

Sector

Firm size

County

Co-worker class

Firm FE

Worker FE0 .05 .1 .15 .2 .25 .3

Coefficient

(d) Worker-to-worker: ρww

Figure A17. Coefficient change for path coefficients adding mediators. Point estimatesand 95% confidence intervals. Daughters only, sons shown in Figure A16.

61

+460%

+123%

+55%

+135%

+170%

+150%

+91%

+12%

+38%

+103%

+98%

+114%

+78%

-47%Most peers high-class?

Proportion high-class peers

Value added per worker

Operating profit per worker

Works in the capital?

Works in a large firm?

Is a top earner?

0 200 400 600Relative representation (percent)

Origin: Low income High income

(a) Firms, workplaces, peers

+90%

+3%

+161%

+207%

+188%

-89%

-77%

-84%

+71%

+18%

+246%

+247%

+121%

-63%

-76%

-88%

Agriculture/other

Health, public admin, services

Construction

Education and research

Trade and communication

Manufacturing

Energy

Finance and business

-100 0 100 200 300Relative representation (percent)

Origin: Low income High income

(b) Industry classification

Figure A18. Characteristics conditional on working in a high-paying firm, by fatherearnings (top vs bottom quintile). Daughters only, sons shown in the main text.

62


Top Related