Class and Earnings Inequality in the Australian Labor Market

aboriginalconspiracyUrban and Civil

Nov 16, 2013 (3 years and 9 months ago)

67 views

1


Class

and

Earnings Inequality in
the Australian Labor Market

Mark Western

Institute for Social Science Research

The University of Queensland

m.western@uq.edu.au


This paper uses unit record data from the
Household, Income and Labour Dynamics in

Australia
(HILDA) Survey. The HILDA Project was initiated and is funded by the Australian

Government
Department of Families, Housing, Community Services and Indigenous Affairs

(FaHCSIA) and is
managed by the Melbour
ne Institute of Applied Economic and Social

Research (Melbourne
Institute). The findings and views reported in this paper are

those of the author and should not be
attributed to either FaHCSIA or the Melbourne

Institute.

I would like to thank Laura Cox fo
r helpful research assistance.

Draft only. Not to be cited without permission
.



Paper prepared for
2011 RC28 Conference, Essex, April 13
-
1
6, 2011
.
2



"Class

and Earnings

Inequality in the Australian Labor Market".


Mark Western,

Institute for Social Science Research,

The University of Queensland, Australia.



This paper examines how class relations structure earnings inequality in the Australian labor
market using data from Australia’s national household panel study, Housing, Income and
Labour Dynamics in Australia (HILDA). Contemporary class theorists such as
Wright and
Goldthorpe have argued that class relations, however conceptualized, matter for earnings
inequality because of distinct “rent
-
type” processes associated with particular class
locations. Key examples would include the monopoly rents associated wi
th professional
occupations and the strategic or efficiency wage type rents associated with managerial and
(in the Goldthorpe conceptualization) service class jobs. Other approaches to class analysis,
in contrast, assert that class relations structure dist
inct labor markets within which earnings
determination processes vary. This
paper

examines some of these issues using 8 waves of
longitudinal data from Australia’s HILDA study, a national househo
ld panel of (in wave 1),
over 17
,000 individuals from 7,000 h
ouseholds. The first part of the
paper

explores the
extent to which the
work life

mobility of women and men

takes place within distinct
segmented working class or middle class labor markets. The second half of the
paper

estimates longitudinal regression mo
dels for earnings that attem
pt to specify how class
location
shape
s

earnings inequality over the
life course
.




3


Class and
Earnings Inequality in
the
Australia
n Labor Market

Sociologists have long argued that class relations c
entrally explain economic ineq
uality in
capitalist societies (Wright 1
985, 1997; Sorensen 1991
). Recent
ly

class theorists have

attempt
ed

to systematically link individuals’ class locations to their economic circumstances
through the operation of different kinds of economic rents. While these
theorists

propose
specific mechanisms to account for class variations in income or
earnings

the
ir

arguments
still allow
some
indeterminacy

about how these processes should be operationalised and
empirically examined. This paper
draws on recent theories to develop

a number of
hypotheses about how class relations are related to earnings inequality
and em
pirically
examines the
structuration of Australian labour markets by gender, class and ethnic
relations.

The paper attempts to make two contributions. First it attempts to develop a
range of arguments about how class and earnings are potentially related on

the basis of
recent theoretical work. Second it empirically examines these arguments through an
analysis of women’s and men’s earnings determination in Australia.


Class and Earnings Inequality

The assertion that class relations are
a

fundamental cause of

economic inequality
(understood in this paper as the unequal distribution of earnings) has been a hallmark of
sociological analysis.
Since the mid
-
1980s,

one common feature of both Marxist and non
-
Marxist class analysis is the view
that
class variations in income
are

caused by the operation
of

different kinds of economic rents. Individuals derive different incomes because they
occupy positions in the class

structure

that either enable or fail to enable

them to extract
economic returns tha
t are in some sense “disproportionate”.

Formally, an economic rent is
a return to an asset that exceeds the cost of producing or bringing about the employment of
that asset (Sorensen 2000).

The different rent components of income

that arise from the
resour
ces and powers attaching to classes

explain

why
income differences
exist
among
classes.

The first recent attempt to propose economic rents as the source of income differences
among classes was by Erik Olin Wright (1985, 1997). Although Wright was initially

primarily
4


concerned with conceptualising the class structure of contemporary societies in the face of
increasingly complex occupational structures and employment relations
,

his concept of class
structure is embedded in a theory ab
out the distribution of i
ncome.

For Wright (1985, 1997) the capitalist class structure is defined in terms of ownership
relations with respect to three types of assets used in production: property (or the means of
production), organisational resources and skills. Ownership relatio
ns with respect to
property, skills and organisational assets or resources define locations in the class structure
and allow individuals in these locations to extract certain kinds of economic rents.

Owners of productive property derive income t
hrough the
ir property ownership, and since
this income rests on the value of goods and services produced by employees, it represents
an exploitative appropriation of the surplus value
employees produce. Among owners of
productive property Wrig
ht distinguishes large
and small employers (capitalists), and the
self
-
employed who

employ no
-
one and work their own means of production (the petty
bourgeoisie).


Non
-
owners of the m
eans of production, employees, may
have

skills

and/

or

control

organisational resources, which are also a basis for income. Professionals or experts, for
instance, control scarce occupational skills that are typically regulated through credentialing
or other restrictions on labour market access.

Credentialing

artifi
cially regulates entry into
professional occupations and labour markets thereby driving up

the price of expert labour
relative to a situation

where

labour supply was not restricted. Employed professionals and
others with scarce occupational skills

therefor
e have “monopoly re
nt” components in their
incomes (Wright 1985, 1997).

Managers, on the other hand, occupy positions
with

some of the rights and powers of
owners (to make decisions about the distribution and use of resources, to hire and fire, to
supervis
e others, for instance) and some of the characteristics of employees (
managers

can
be dismissed, and cannot sell or appropriate organisational assets for their own
consumption). These features of managerial jobs give managers bargaining power.
Managers’

de
cisions and actions can be costly for organisations,
but

owners have delegated
much of the responsibility for ensuring appropriate performance
, including of managers,

to
managers.

Paying high salaries and providing other forms of remun
eration such as bonu
ses,
5


shares and

stock options, helps secure managerial loyalty and ties managers’ interests
directly to their employing organisations. Managerial incomes therefore contain “strategic
rents”, rents that reflect the strategic position and responsibilities of

managers
in
hierarchical organisations (Wright 1985, 1997, 2005).

This argument is analogous to
efficiency wage arguments in economics,
where

higher than market clearing wages are paid
to certain categories of employees to prevent them from shirking, and/
or to minimise
costly
employee
turnover (Shapiro and Stiglitz 1984).

Different employee class locations may have different levels of skill or organisational
control, and some locations may have both scarce skills and managerial rights and
responsibilities
(so
-
called expert managers). Work
ing

class locations, on the other hand, are
defined by the non
-
ownership of productive property and the absence of substantial
occupational skills and organisational assets. Individuals in working class jobs are therefore
u
nable to extract rents from their employers. Wright’s conceptualisation of the class
structure in terms of relationships of ownership and control of productive property, skills
and organisational resources is therefore also a theory of income determination

in which
the incomes of property owners, managers and experts/professionals all vary with the level
of their assets and all
should
exceed those of workers (Western 1991; Talmud, Kraus and
Yonay 2003).

Like Wright, Goldthorpe
conceptualizes
class structure in terms of employment relations
linked to

a partial theory of income determination.
The initial

class

distinction is

the
ownership and non
-
ownership of productive property (Breen 2005; Erikson and Goldthorpe
1992). Among non
-
owners, the ke
y feature of the employment relation is the nature of the
employment contract


in particular, whether or not there is a strict labour contract in
which a worker exchanges specified duties in return for wage, or whether there is a more
diffuse “service rel
ationship” built
around long
-
term employment,
less
-
specific and more
variable duties
, and expectations about lifetime and future rewards
. The former case can be
relatively easily supervised and monitored and contract breaches can be enforced.
Employees in
a “service relationship” with their employer, on the other hand, cannot be so
easily supervised or monitored, particularly since employers have
frequently
delegated
these duties

to
these
agents. Employers therefore have

to
create incentives that align the
interests of employers and managers, that is, buy

the loyalty of service class employees
6


through remuneration

packages
, career
opportunities and other rewards (Breen 2005;
Goldthorpe 2000).

Goldthorpe’s argument for the high income of the service class

as

a rent
arising from the strategic
nature of service class jobs is
thus
very similar to Wright’s
argument about the strat
egic rent in managerial incomes (Western 1991; Talmud et al.
2003)

Wright and Goldthorpe thus focus on similar resources and mechanisms
to link class to
income inequality, and both produce concrete “maps” of the class structure that identify
similar groupings (the self
-
employed, professionals managers, various skilled workers)
(
Western 1991;
Talmud e
t al. 2003). They differ where

Wright vi
ews the monopoly and
strategic rents of managers and professionals as distinct mechanisms, while Goldthorpe
emphasises the strategic rent of a
combined
professional/managerial service class

(cf
Savage, Barlow, Dickens and Fielding 1992)
. They also differ
in that Wright links the theory
of income differen
ces to a theory of exploitation

in which one class derives income at the
expense of another. This means that classes also have
intrinsically opposed

class interests

thereby

linking

class structure to
social

change

(Wright 2005
).

Sorensen (2000
) also argued that rents should be used as the basis of a class structural
t
heory of income inequality

and
linked
class structure
to class formation and s
ocial change
as groups attempt

to destroy or protect rents. L
ike
Wright, Sorensen (2000
) recognised the
importance of monopoly and strategic rents and also acknowledged the role of composite
rents, in which the returns to the use of two or more assets exceeds the sum of the separate
returns to each asset.

Unlike Wright,

h
owever, Sorensen (2000
) also generalised arguments
beyond the occupational groups emphasised by Wright and Goldthorpe to

define
employment rents
generated by

circumstances such as
union closed shops, crafts and trades
regulated through apprenticeships, a
nd
the possession of valued
natural and cultural
endowments.

Empirically, employment rents should be reflected in higher earnings among rent generating
asset holders than among those who lack such assets. Income differences among classes are

therefore
cons
istent with the operation of employment rents, but are not sufficient to
demonstrate that re
nts, rather than other mechanis
ms, such as
greater

productivity
,

are
responsible
for

observed differences. However, to the extent that other sources of variation
7


in

earnings are effectively controlled, income differences among classes provide some
evidence
for class
-
based employment rents
.

This assumption has motivated most empirical
analyses

of

class differences in earnings

which attribute

net class differences in earnings
to
such rents

(e.g. Western 1991; Halaby and Weakliem 1993; Talmud et

al 2003
; Morgan and
McKerrow 2004
).

While intercept shifts in a regression equation
might reflect

the operation of class
-
based
rents
in the labour
market, class analysis assumes

that class matters for life
-
chances, the
formation of identities and interests, and attitudes and behaviour because it is a durable
aspect of individual and household circumstances. Class “membership” is an enduring
component

of people’
s lives
that

over time
,

through circumstance and experience, produces
the outcomes
for which class
is belie
ved to be causally important
.

This

temporal component of class is missing or underemphasised in theoretical and
empirical work on class an
d income. But if class has to be a temporally enduring feature of
people’s lives to matter for class outcomes
,
class relations
are

more appropriately thought
of in terms of class structured careers

or
class structured
labour markets in which careers
are em
bedded

(cf Esping Anderson 1993)
.
The view that some classes, at least, are
temporally structured, is explicit

in

Goldthorpe’s arguments about the “prospective” nature
of service class employment organised in terms of a career in a firm internal labour mar
ket.
This
argument also
conversely implies a separate working class labour market organised in
terms of discrete finite employment contracts. Likewise, Sorensen (2000) and Wright (1985,
2005) acknowledge the temporality of certain class locations, particul
arly managerial
locations,
that they link

to internal labour markets.
The importance of temporality is also
noted by stratification researchers such as Hause
r and Warren (1997), who defend

the use
of occupational status

as a casual

factor in other outcomes because occupation is
temporally stable.

Thinking about class in terms of careers
implies

that class relations potentially structure

class
-
specific labour markets in which advantages and disadvantages accrue because of
class
-
spec
ific lifetime differences in the returns which attach to various earnings generating
resources. Class
-
based employment rents, on this view
, are not

reflected in net intercept
shifts in earnings linked to class location, but to interactions between class lo
cation and
8


other determinants of earnings such as education, experience, seniority, job tenure and job
shifts. In an interactive model

informed by a temporal view of class structure
, returns to
education or experience, for instance, might vary across diffe
rent class structured labour
markets. Equivalently, class differences in earnings might vary at different levels of labour
force experience

or educational qualification
.
This latter point is noted by Sorensen (2000)
and Wright (
2005)
who
argue that manager
s typically start off in lower wage jobs, but finish
their careers with higher than market wages. Employees in managerial careers may
initially
earn

negative rents (less than market clearing wages), only earning positive rents after they
have accrued enoug
h job experience or seniority (Sorensen 2000).

Conversely, l
ow wage
jobs may
be

“sticky” over the
life course

(Bernhardt, Morris, Handcock and Scott 2001, ch.
7), if individuals begin employment in industries and occupations from which it is difficult to
m
ove, and which do not result in wage growth

as experience accrues.
Being stuck in a low
wage working class labor market

will lead to a significant accumulation of disadvantage over
the course of a

lifetime

employment trajectory
.

These arguments have particular relevance for managerial and professional
employee
labor
markets where
class structured
employment rents imply larger returns to the

human capital
determinants of earnings than exist in working class labor markets. In parti
cular returns to
characteristics such as seniority
, experience,
education

and job shifting,

should be higher

for
professionals and managers

than in working class labour markets.

M
anagers in Australia are
more likely
than workers
to have their pay and remun
eration set by individual agreements
negotiated with employers (Australia 2008). Such agreements allow flexible forms of
remuneration. Individuals in working class jobs, on the other hand, are more likely to have
their pay set according to industrial award
s, which specify fixed wage rates,

or collective
agreements, which can increase wage rates over award minimum levels

and allow some
variation within specified limits
. Awards and collective agreement
s thus

impose wage
uniformity on workers and limit
wage di
spersion in the manner noted by Goldthorpe in
relation to working class jobs.

The flexibility

of individual agreements as a method

of pay
setting, t
he prospective nature of managerial jobs and the existence of organisational
hierarchies and firm internal l
abour markets all imply greater returns to seniority and
experience in

managerial than

working class labour markets where specific employment
contracts and
fixed

returns for effort predominate. We should also expect

greater

9


managerial returns to education
at higher levels of formal qualification

because

higher
levels of education may signal the strategic importance of the managerial job and give
managers more bargaining power with respect to the negotiation of individual employment
contracts with employers.


We should also expect greater returns to
experience in professional labo
r

markets

because
monopoly rents

premised on skills shortages should increase for more experienced
professional employees. Formal educational qualifications are typically a condition of entry
into Australian professional labour markets, as they are elsewhere. This will truncate
education
al variation in professional labour markets and
undermine an
education premium

for levels of education that are consistent with standard labour market entry requirements.
However, individuals with higher than en
try
-
level education may receive

a
premium bas
ed
on

a

further

monopolistic skill claim within an already monopolistic labour
market.
Conversely,
individuals with lower than average levels of education employed in
professional labour markets may suffer a negative employment rent.

Among the self
-
employe
d, employment rents arise through levels of capital ownership

rather than education, experience or seniority
. The latter are
all factors behind earnings
determination for employees
. There are not
strong

grounds for believing that
experience,
seniority or e
ducation
will be associated with greater returns for the self
-
employed
.
However, to the extent that the self
-
employed determine their own income level
s, among
larger owners at least

we
would expect

returns to experience
to

exceed those of workers, as

wealt
hier

property owners progressively pay themselves more.


Globalisation, Uncertainty and Gender

While a temporal view of class
-
structured careers implies class differences in returns to
some determinants of earnings, t
he arguments
largely assume

a “s
tandard


male

life
-
cours
e
in which individuals

enter full
-
time employment in a predictable labor market and remain
until retirement.

It is less clear how they apply to women, who
typical
ly

interrupt or limit
their labor force participation to manage households an
d care for children.

In addition
,
however, social and economic changes associated with
post industrialism
, neoliberal public
10


policy and globalisation have potentially undermined
standard male
life courses

for

men as
well
(
Block 1990;
Baxter and

Western 20
01a;

Wes
tern et al, 2007
).

These transformations
may erode class structured career trajectories and class based employment rents.


R
elevant
changes
include the growing internationalisation of economic activity and the
declining importance of national borde
rs, global interconnectedness through information
and communications technologies, increased national economic deregulation, liberalisation
and privatisation
,

and the growth of world markets for capital, labour, services and goods
(Mills, Blossfeld and Ber
nardi 2006).
Globalisation has intensified competition in world and
national economies

(Mills et al. 2006)

which has been localised in competition between

firm
s and
organisation
s

(Bernhardt et al 2001
)
.
Neoliberalism has
led to

economic
deregulation, the growing importance of markets as mechanisms of social allocation and the
withdrawal of
the state from the provision of soc
ial services (Western et al 2007
).

Post
industrialism

has seen rising female labour force participation ra
tes, increased occupational
and industrial differentiation, the emergence of service sector industries and jobs, and the
relative decline of manufacturing

(Block

1990; Esping Anderson 1993
)
.

Some

of
these
changes
potentially result

in choice and

opportunit
y

for individuals

but

they

also increase

uncertainty

and individual responsibility for managing social risk.


Since the 1980s Australia
has progressively deregulated

and opened

the national economy

to international competition

(floating the dollar, opening the financial industry to foreign
banks, removing protective tariffs

in industry
, privatising and corporatizing government

services and

business enterprises). L
abor market institutions which ensured

coordinated

economy
-
wide
c
entral
ised wage bargaining
,

have given way to decentralized, deregulated
agreements
(Western et al 2007
)
.
Labor market uncertainty has been reflected in the
growth of precarious and non
-
standard employment (casual and fixed term employment
contracts, irreg
ular working hours,
the polarization of working hours,
the growth of part
-
time work)
,

the
confining

of
central
award regulation to low
-
wage sectors of the economy
,

the disappearance of collective bargaining in the private sector and
the
emergence

of
indivi
dual bargaining and methods of pay setting

(Van Gellecum et al.

2008
)
.

The potential consequences for the class structuring of labor markets are twofold

(cf Mills et
al. 2006)
.

According to some,
economic changes such as internationalisation and
increased

11


labor market
flexibility
, and non
-
economic changes such as

environmen
tal
challenges
, the
threat of global health pandemics

and new technologies
,

universalise

risk

across the
population
(Beck, Giddens
)
. At the same time,

educational expansion
,
increased
oc
cupational

and industrial differentiation

and
increased geographic

and social

mobility
create
new
possibilities

for socioeconomic attainment and weaken the

effects of social
origins. Coupled with
the decli
ning
importance

of
class, family and community
on
the
formation of social identities
,

a culture of individualisation and “reflexivity” emerges in
which individuals
constantly
face new choices and have to actively make their own
biographies

through strategic life planning

(Phillips and Western 2005; Mills
et al. 2006).
These processes

see class structured inequalities replaced by complex individualised
inequalities

(Pakulski 2005)

as class
-
structured
life courses

and careers
give way to

individually differentiated life
-
pathways
. C
lass structured employment
rents are
also
eroded
by

deunionisation,

the rise of precarious employment,

part
-
time employment and
unemployment,

and

the disappearance

of internal labor markets
and occupational and
organisational careers.

An alternative perspective questions the univers
alisation of risks and opportunities. In
particular, labor market uncerta
inty

is more prevalent in

certain occupations and industries

(Bernhardt et al. 2001
)
, access to higher education

remains unequally distributed by class
and educational origins

(
Blossfeld and Shavit

1993
), and

opportunities to activ
ely construct
one’s life
-
pathway

in the pursuit of desired goals

depend on

having

appropriate
economic,
cultural and social resources (Phillips and Western 2
005;
Adams 2007
). R
ather than
undermining cla
ss inequalities
,

economic and cultural changes in contemporary societies
adversely affect the

opportunities of
some



especially
low skill, low wage workers in certain
occupations and industries
-

while

advantaging
others, particularly professional and
man
agerial employees.


Adjudicating these arguments

ideally require
s

comparisons

of longitudinal cohorts

who
entered the labor market at

different times over the last
forty to fifty years (e.g. Bernhardt
et al., 20
01
; Cobb
-
Clark and Khoo 2006
). Although I cannot directly examine these
arguments
in this

paper, evidence of class structured differences in earnings and stable
class
-
based careers will undermine claims that globalisation and individualisation have led
to the disappearance of class s
tructured inequality.

12


L
abor market uncertainty also
potentially affects
class structured careers for women and
men
differently
because gender relations also shape

earnings inequality
and

labor market
structuration. Gender relations
affect

women’s and men’s

labor force participation

and
earnings trajectories

through

differences in labor force participation

rates
, occupational sex
segregation,
interconnections between paid work and domestic responsibilities and gender
discrimination
.

In Australi
a in

the last
thirty

years,
women’s and men’s labor force
participation rates have converged as

male participation has
fallen

while
female

participation has grow
n

(Van Gellecum et al, 2008).
G
rowth has been particularly
pronounced among middle aged women and mothers

and concentrated

in part
-
time and
casual jobs.

In addition,
the Australian labor market

is

highly sex
-
segregated (Kidd 1993)

with women concentrated in clerical, sales, service and some professional occupations.
Occupational sex segregation shapes earning
s because wage profiles vary across
occu
pations (Kidd 1993). W
omen’s continued responsibilities for domestic labor and
childrearing

also
mean that women, rather than men, adjust their involvement in paid
employment to accommodate work in the home (
Western
and Baxter 2001
b
). Such choices
include interrupting paid employment to care for children, choosing part
-
time rather than
full
-
time work
, limiting working hours in full and part
-
time work,

and
choosing jobs with
hours and employment conditions that enable
employee flexibility.

These factors influence
earnings directly and indirectly through their effects on women’s abilities to establis
h
continuous employment careers, and through the impact they have on employer
perceptions and thus on direct and statistica
l gender discrimination.
In recent years, the
combined effect of these processes in an increasingly decentralised and deregulated labor
market has been to expose women to employment in sectors
in which

employment
regulation
s

prescrib
e only

minimum conditions and standards (Van Gellecum et al, 2008).
W
omen, rather than men

are
disproportionately
subject to labor market uncertainties that
may

erode
class
-
based employment rents.

Ethnicity

also structures labor market outcomes

in Australia

but

in ways that differ from
those of gender. Immigrant stat
us
is particularly consequential and
itself
is shaped by
immigration policy which regulates migrant selection

(Cobb
-
Clark and Khoo 2006
)

through
Humanitarian (refugee) and Migration programs
. The
Mig
ration program
is
further divided
into Family, Skilled a
nd Special Eligibility streams
(Chiswick and Miller 2006).

I
n Australia
,
13


unlike the United States and Canada,

immigrant status primarily affects

unemployment
duration, rather than earnings
, especially

for recent immigrants
,

because the jobs to which
immigrant

are drawn are largely

regulated

by industrial awards
that strictly
specify

pay and
working conditions. Recent immigrants who lack “Australia
-
specific skills”

(Chiswick and
Miller 2006), such as En
glish language abili
ty, education, and labor
market experience, are
not competitive for such jobs, but
become so over time,
especially
as English langua
ge
ability
increases
. At this point

immigrants
typically
enter
jobs

regulated by industrial awards
.
The
limited wage dispersion in this sector of the Australian labor market means that
immigrant status, language ability and country of origin more typically affect the
employment status of immigrants

than their relative earnings, unlike the United States and
C
anada, where wage dispersio
n
,

a
larger low
-
wage sector
, and lower relative minimum
wages

see more variability in the relative wages of immigrants (Miller and Neo 2003;
Chiswick and Miller 2006).


Methods

To investigate how class
relations influence

earnings inequality in Australia, I analyse
longitudinal data from
eight
annual
waves of
the Household, Income and Labour Dynamics
in Australia

Survey (HILDA)
. HILDA is a federally funded national household panel study that
began in 2001 with special emph
asis on family and household formation, income and work
(Watson et al., 2010). The original
multi
-
stage probability
sample consisted of over 19,000
individuals

aged 15 years

and over

from

over 7,000 households
. The wave 1 sample

was
broadly
representative
of the Australian population on key social, demographic and
economic characteristics.
1

Since wave 1, wave
-
on
-
wave attrition rates have varied from 13
percent to 5 percent (Watson et al. 2010), declining with each wave.

Most data are
collected by face
-
to
-
fa
ce interviews with a supplementary self
-
completed questionnaire.

I restrict the analysis to respondents who were aged 25
-
54 years in wave 1 and employed
between 5 and 80 hours per week in any of the eight waves. The age restriction is designed
to capture
workers in

prime working age
” or “mid
-
career”

in wave 1
,

while the
hour’s




1

The wave 1 sample slightly under

represents immigrants from
non
-
English speaking backgrounds, but the
discrepancies are small (Watson and Wooden 2002).

14


restriction selects those

who are not working excessively long or short regular hours. I also
omi
tted

observations whose regular hourly or weekly wage fell below the federally
manda
ted minimum wage in
that survey

year. The result is an unbalanced panel
consisting
of 27617 person year observations, with complete data on all variables, and longitudinal
information for between one and eight waves.



Variables

The dependent variable is
log hourly earnings in the main job deflated to constant 2001
dollars using the consumer price index for each survey year. I focus on earnings in the main
job, since the occupational information used to construct class location is based on the main
job. I
model hourly earnings (usual total weekly earnings divided by usual hours worked per
week) because many workers


wages are based on an hourly wage rate multiplied by hours
worked. I model log hourly earnings rather than raw earnings to normalise the depend
ent
variable and pull in pos
itive outliers, and because it
enables class differences in returns to be
thought of as multiplicative increments or decrements

on a class base rate

rather than
absolute shifts.

The key independent variable is class location, ba
sed on information about a respondent’s
main job in
each survey wave.

The class concept I use is a modified version of Wright’s
(1985) productive assets typology. It first distinguishes property owners from employees,
and among property owners distinguishe
s employers,
who employ others
, from the petty
bourgeoisie who do not. Among employees, I distinguish professionals (experts), managers,
trades and technical workers, white collar clerical sales and service workers, and blue
-
collar
skilled, semi and unskil
led workers. Each employee category is defined using the Australian
and New Zealand Standard Occ
upational Classification (Australia 2006a
) a skill
-
based
occupational classification that uses standardised information about job title and job tasks.
Ideally I

would measure occupational skills/expertise using information about the average
education of occupations, and organisational resources using information about decision
-
making and supervisory responsibilities at work. However the latter information is not
available in HILDA and thus the occupational classification using title and task information is
the only way to identify managerial employees.

15


E
mployers and the petty bourgeoisie correspond directly to Wright’s (1985) classes, and
professionals and manager
s are roughly equivalent to experts and managers

in that
classification
. However there is no way

with these data

to identify individuals who have both
scarce occupational skills and managerial responsibilities

(expert managers)
.
Technical/trade, white coll
ar and blue collar employees make up the working class. I
introduce these distinctions to capture skill differences among workers and also because
occupational sex segregation means that women are typically located in white collar
working class jobs, while

men are increasingly found in both blue and white collar working
class

jobs (cf Esping Andersen 1993
; Erikson and Goldthorpe 1992
).

I use a version of Wright’s concept
because there is a strong

conceptual linkage between the
three productive assets, prop
erty, organisation and skills,
and three kinds of rents that maps
more directly onto arguments about class variations in earnings.
However the distinctions I
introduce into the working class

also reflect Goldthorpe’s (Erikson and Goldthorpe 1992
)

arguments

about the need to distinguish class fr
actions by skill and occupation, particularly
in light of occupational sex segregation.

Previous
research

that has compared Wright’s and
Goldthorpe’s schemes finds that they are similarly associated with earnings, alt
hough
Wright’s scheme tends to perform somewhat better (Western 1991; Talmud et al., 2003).

The regression models that follow include other variables capturing human capital (age,
education and work experience

with squared terms for age and experience
), la
bor market
characteristics

(employment sector, industry,
union membership, job move in the last 12
months, full
-
time/part
-
time status, dummy for multiple jobs), family responsibilities (marital
status, resident dependent children, presence of a preschool c
hild) and ethnicity (country of
birth, English as first language). With the exception of the ethnicity variables, all are time
-
varying.

Apart from industry, the above variables are largely self
-
explanatory. Industrial classification
is based on information

coded to the Australian and New Zealand Standard In
dustrial
Classification (Australia 2006b
), collapsed to capture the theoretical concept of the post
-
industrial service economy (Esping
-
Anderson
1993
1999: ch. 6). Primary and secondary
industry sectors ar
e defined as Agriculture, Forestry, Fishing
,
Mining, and Manufacturing,
Construction
,
Power

generation
, respectively. The service sector is divided into Business or
16


Producer Services (finance, insurance, real estate, business professional services),
Distri
butive Services (wholesale and retail trade, transportation and communications),
Consumer and Personal Services, Social Services, (Health, Education, Social Care), and Public
Administration and Defence (Government, Emergency Response, Defence Forces).

I also include
two

attitudinal

measures, one of which is time
-
invariant. Instrumental work
orientation is
a reverse coded seven point Likert response item “I would enjoy having a job
even if I didn’t need the money”, measured in wave 1.
Wave 1 values are c
arried forward to
wave 8.
Financial situation is a time
-
varying covariate measured on an 11 point scale asking
respondents how important their financial situation is to their life at the time of interview.
These two measures are included to attempt to take account of any endogeneity between
class and earnings that might arise
if

individuals choose jobs based on their expected
earnings. If individuals choose jobs according to their anticipated earnings, and a
nticipated
earnings and actual earnings are correlated, class location will be endogenous in earnings
equations if the propensity to choose jobs based on earnings varies over time for
individuals, regardless of whether one uses fixed or random effects to e
stimate longitudinal
earnings equations. If propensity to choose jobs on the basis of anticipated earnings is time
invariant, but correlated with a job’s actual earnings, class location will be endogenous in
random effects regression models, but not
in
fix
ed effects models. I use random effects
models to estimate longitudinal regressions and therefore include both variables in an
attempt to measure time invariant and time
-
variant propensities to choose jobs according
to expected earnings.

Finally, to captur
e any survey
-
specific or year
-
specific aspects of earnings, such as
unaccounted for wage growth, or effects of attrition and non
-
response, I also include
controls for survey wave.

Summary statistics for all variables for women and men are shown in Table
s

1

and 2
.



17


Analytic Strategy

My arguments about class structured labor markets assume that individuals remain in a
particular class location over the employment career, or at least that mobility patterns are
structured meaningfully rather than being random.

To investigate this issue I first present
descriptive information on the class mobility sequences of individuals over the eight waves
of data.

To investigate the class structuring of earnings I

next

regress log hourly earnings on the
explanatory variable
s separately for women and men using random effects (i.e. random
intercept) models estimated with maximum likelihood. The models allow time
-
varying
individual
-
specific errors (i.e. e
ij
where i indexes time and j indexes individuals) to vary
heteroscedastic
ally between the self
-
employed and employees. This setup is a two
-
level
linear mixed model of time points (survey waves) nested within individuals (level 2). I all
ow
level 1 residuals to have different variances for the self
-
employed and employees because
there is greater earnings dispersion among the self
-
employed at each survey wave.
I use a
random effects model rather than fixed effects
.
Although the random effects model
assumes that random intercepts are uncorrelated with explanatory variables, and the
fixed
effect
s

model

does not, it is not possible to straightforwardly account for time
-
varying
subject
-
specific heteroscedasticity with fixed effects. It is also not possible to estimate
effects of time
-
invariant covariates such as country of birth and fir
st language.

I begin with a model that interacts class location with relevant labor market variables,
education, age, age squared, experience, experience squared, industry, recent job shift, and
survey wave.
Class interactions with e
ducation, age, age squared and experience capture
the idea that human capital and educational credentials have different returns in class
structured labor markets. Class by industry interactions reflect the recognition that insider
and outsider labor marke
ts
may

have an industry basis,
with low
-
wage personal and
consumer service industries, for example
,

likely to have smaller class
-
based employment
rents than producer services or, in Australia, mining

and parts of agriculture

(c.f. Bernhardt
et al. 2001, ch
.7; Morgan and Tang 2007).
Job shifts also have effects on earnings that
potentially vary by class. Where job shifts reflect voluntary moves and upward mobility we
would expect them to be associated with wage growth. This is the image of the
18


individualised

middle class career in which professionals and managers view mobility as part
of the individual “project of the self” (
Savage 2000) built around the tactical mobilization of
individual resources. Alternatively,
when they occur in occupations and industrie
s with high
turnover, no career ladders or internal labour markets and unstable and insecure
employment (Bernhardt et al., ch 6)
job shifts may be associated with wage penalties
.

Having fit
the

interactive model, I then use multivariate Wald tests and like
lihood ratio tests
to identify and eliminate non
-
significant

interactions and develop preferred earnings models
for women and men.

Results

Table
3

contains preliminary descriptive information on the employment trajectories of
women and men over the eight H
ILDA waves. There are potentially as many different
trajectories as individuals. The columns labelled “Within” and “Between” fractions identify
the proportion of individuals whose trajectories over the eight waves remain within a single
class fraction, or
move between two class fractions in a
single

class (i.e. between employers
and the petty bourgeoisie, professionals and managers, or any pair of trade/technical, blue
collar or white collar). Over half of men and sixty percent of women remain within a sing
le
class fraction

over the eight waves, and over two thirds of men and seventy percent of
women have trajectories that are either confined
to

one class fraction, or move between
fractions within self
-
employment, the middle class, or the working class.
2

Wom
en
demonstrate slightly more class immobility than men, which could reflect occupational sex
segregation, labor market interruptions or more limited mobility up career ladders. The
concentration/dispersion index
expresses the number of distinct sequences a
s a proportion
of the number of possible sequences (i.e. the number of units to which the sequences
apply).
3

There are 2724 men
with

86 different trajectories and 2534 women with 59
different trajectories, thus the concentration measures for women and men are 3.16
(86/2724) and 2.33 (59/2534) respectively. Class trajectories over the eight waves are thus
highly concentrated.




2

Recall that the data are an unbalanced panel so an individual’s employment trajectory may vary between one
and eight elements over the eight waves. On average eac
h trajectory is just over five years for women and
men.

3

“Different” sequences here means sequences with different elements (class locations, regardless of order). A
more restrictive definition of similarity would require the same elements in the same ord
er in which case
concentration would be lower.

19


Tables 4 a
nd 5 show the elements
of

the ten most common employment trajectories
. The
se

trajectories account for 66 percent of the possible tr
ajectories for men (1786/2724)
and 80
percent of possible trajectories for women (2019/2534) reinforcing the inference that
t
rajectories

are highly concentrated. Six of the ten most common male trajectories stay
within a single class fraction (only the petty

bourgeoisie is not represented)

and, three of
the remaining four are confined within the middle class or the working class
. Although

women’s trajectories are more concentrated than men’s, only four of the top ten female
trajectories stay within a single class fraction. Two of these are working class (white
-
collar
and blue
-
collar). Employment trajectories between professional

and managerial locations,
white collar and managerial locations
,

white collar and blue collar,
and

professional and
white collar locations are also relatively common.
Women’s class trajectories over this
period thus remain largely within the working or mi
ddle classes, or cross boundaries
between white collar jobs on one hand, and professional or managerial jobs on the other.
Overall, these simple descriptive results show that men’s and women’s career trajectories
over the eight waves of the survey remain l
argely within class fractions or classes, or show
systematic mobility between pairs of linked locations. Although the observation window is
comparatively short, there is prima facie evidence that employment careers are
systematically structured by class re
lations, rather than being individualistic and variable.
4

Regression results for the preferred models for men and women

are contained in Tables 6
and 9
. The preferred model for men incorporates interactions between class and industry,
education, the linear

and quadratic terms for age, the quadratic term for experience, job
shift and survey wave. These results imply that there are class differences in returns to age,
education
, and job shifting, that class differences in earnings vary by industry, and that n
et
of these factors, classes also experienced different levels of real wage growth or stagnation
from one year to the next.


Table 6 presents derived coefficients
from the men’s random effects equation
for each class

for education, age, experience, job

shift and survey wave.
For brevity I do not present the
industry results.




4

If highly mobile respondents are more likely to leave the survey, selection bias will inflate results for
immobility. This is clearly plausible but as indicated earlier, attrition is greatest in the earliest

waves and since
wave 5 has stayed around five percent (Watson et al., 2010).

20


T
he
squared

terms
for age
are significantly

negative for all three working class locations
(technical/trade, white collar and blue collar workers) but not for any of the other class
es.
The differences between
Employers, Professionals and Managers
, on one side and
Technical
and Trade, White Collar and
Blue Collar Workers
, are
also
statistically significant
.
Lo
g hourly
wages
for all three working class groups

thus

have a c
urvilinear ag
e
-
earnings profile rising
and then declining
,

while those
for the self
-
employed

and the middle class do not.

The model
also
specifies a
single

linear experience
effect

w
ith class
-
specific non
-
linear
experience trends
. The Blue Collar experience

squared

coe
fficient is significantly positive,
implying that the

net effect of experience in blue collar jobs is for earnings to increase at an
increasing rate. By contrast, for Professionals and Managers,
log
-
hourly wages increase with
experience up to a point, afte
r which they begin to decline.

The education coefficients indicate that in all
classes’

individuals with Bachelor degrees or
higher qualifications receive a similar

earnings premium in comparison to individuals with
high school completion only (the reference category). Class differences in the effects of
education are most evident for respondents who did not complete high school
. In
employee
classes

not
finishing

hig
h school

results in lower hourly wages than completing high school.

Among employers and the petty bourgeoisie, in contrast, there is no wage penalty

for not
completing high school; equivalently, employers and the petty bourgeoisie do not receive an
income
benefit for finishing high school rather than leaving at the end of compulsory
schooling.

I
n classes based on property ownership,

then,

there is no return to completing
high school rather while among employees leaving before the end of high school carries

a
wage penalty of between 12 and 17 percent.
Finally the coefficients for trade/vocational
qualification show
that this increases log hourly income by about 14 percent for petty
bourgeois men compared to finishing high school, while in some other classes

(white and
blue collar workers, managers)

there is actually a wage penalty for having a trade
qualification rather than a high school certificate.

Looking at the point estimates, the job shift coefficients show that in some classes changing
jobs is associ
ated with increased hourly earning
s (employers, petty bourgeoisie)

while for
white collar workers, jobs shifts result in decreased hourly earnings. These differences are
statistically significant. The employer coefficient is also significantly larger than
the Blue
21


collar coefficient, meaning that job mobility largely has opposite effects on wages for the
self employed and the working class. For workers, job shifts are associated with static or
declining hourly wages, whereas for the self
-
em
ployed changing j
obs results in increased
earnings.

Finally the coefficients on survey wage capture any remaining year by year shifts in hourly
earnings over the survey period. With the exception of employers most classes show yearly
net wage growth of about one to two per
cent. However, middle class professional and
managerial employees, and skilled trade and technical workers experienced higher net wage
growth than blue collar employees did.

The
model to this point allows

all parameters

that interact with class to vary fre
ely.
However, table 6

suggests

that a simpler model that constrains some class interactions to be
equal might also describe the data well. I therefore next fit a final model for men that
imposes selected equality constraints on the class by education, age, age squared,
experience squar
ed, survey year and job shift interaction parameters, in light of the results
in table 6. The pattern of equality constraints is shown in Table
7
.

The model allows four
class varying interactions for the linear effect of age, with Employers and the Petty
Bourgeoisie sharing one level, Managers and Technical/Trade employees the second,
Professionals and White Collar Workers the third, and Blue Collar workers the fourth. The
quadratic age effect has three levels comprising employers, middle class locations a
nd
working class locations: Employers; Petty Bourgeoisie, Managers and Professionals, and
Technical Trade, White Collar and Blue Collar. Other interactions are as shown.

In a c
omplex model such as this,
assessing the overall earnings trajectories for each

class

from

coefficients

is difficult. Figure 1 therefore
plots

predictive margins

derived from model
estimates
. These are

average predicted log hourly earnings, for each class,
calculated

separately
at fixed values of
age,

work
experience

or survey year

(
random intercept
estimates are not included
in the calculation
but have a zero expected value)
.

These three
measures of tim
e
capture career trajectories based on
life course

measures

(age and years
of experience)
and
wage trends over the survey period.

The

predictive margins
assign

t
he
values of age,
experience

or survey year

to specified values to generate predicted values

while other variables take the

values

found in the data
.

22


All

employee classes have age
-
earnings profiles that increase with age before decreasing.
Professionals and managers have the highest mean earnings of employees across the
life
course

and also the flattest age
-
earnings trajectories. Working class employees ha
ve l
ower
mean earnings at all ages and
their earnings peak earlier
than professionals and managers

and decline more sharply.

The age earnings profiles of white collar and technical trade
employees are virtually identical.
Middle class advantages in returns

to age thus increase
over the
life course

(i.e. middle class employment rents increase with age).
Petty Bourgeois
incomes
, on the other hand,

decline at an increasing rate

over the life course, while

e
mployers’
incomes

initially decline with age,
bottomin
g out

in the mid forties and then
beginning to increase again. By the time they are in their late fifties, employers


e
arnings
equal those of managers

but are still less than those of professionals.

Professionals, managers and the self
-
employed also have u
pwardly sloping experience
-
earnings profiles
that peak at about thirty years

in the case of employers, professionals and
managers, and then decline. The earnings profile of the petty bourgeoisie is still increasing
after thirty years. Working class employe
es on the other hand have continuously rising
returns to experience, but at all but the highest levels of experience, still earn less on
average than middle class employees. Again, the technical/trade and white collar
experience earnings profiles for men a
re virtually indistinguishable, especially in the middle
of the trajectory, while blue collar employees start off at lower hourly wages and experience
slower wage growth through the first half of the career.

Compared to male workers, the
male middle class
premium for experience is thus largest over the first half of the career
before declining middle class returns and increasing working class returns lead to a
convergence in wage rates.

The final panel in Figure 1 shows net w
ag
e growth between 2001 and
2002. Growth rates
are similar across all classes

(1.4 to 2 percent)

except for employers and blue collar workers,
both of whom experienced lower growth rates

(less than 1 percent)

over the seven years.

Together the results indicate that there are distinct

wage trajectories by age, experience
and over time for each of these classes, consistent with the existence of class structured
labor markets yielding different returns to human capital, and different
levels

of wage
growth from year to year.

23


Table 8 show
s remaining coefficients for ethnicity

and constrained class variations in returns
to education and job moves. Immigrants from English speaking countries earn on average
about 8 percent more than the Australian
born

and about ten percent more than migrants

born in non
-
English speaking countries. These factors could reflect a combination of
selection effects associated with country of origin, with migrants from English speaking
countries much less likely to have come under the humanitarian program and more l
ikely to
have come under skilled and other economic migration programs. The point estimate for
speaking English as a first language is positive, but not statistically significant.

The model posits two Bachelor degree effects. In blue collar labor markets,
a university
degree is worth approximately 14 percent over finishing high school, but in other classes
completing a Bachelor degree is associated with 20 percent earnings premium. Among the
petty bourgeoisie a trade qualification is worth almost as much as

university degree (17
percent on earnings compared to completing high school only), while in other classes a
trade qualification incurs a 6 percent wage penalty. Conversely, in employee classes
completing high school is associated with a 14 percent wage p
remium over leaving school
early, whereas in the Petty bourgeoisie again, leaving school early is associated with a 12
percent wage premium (although the coefficient is not statistically significant).

Finally, for
self
-
employed men, changing jobs leads to
an eight percent increase in income, on average,
a significantly greater premium than for managers, white and blue collar workers.

Turning now to results for women, Table 9

contains prelim
inary results for the model with
full class interactions be
tween the

linear effects of age

and experience, and education.
There is a si
gnificant negative linear trend

in age for Blue Collar women
that is less than the
point estimates for Employers, the Petty Bourgeoisie and Professionals. There is also a
significant class
-
specific ef
fect of having a Bachelor

degree or higher qualification in all
classes, with a particularly large effect for the Petty Bourgeoisie, and a smaller effect for
Professional women. The small earnings advantage associated with getting a degree for
P
rofessional women is consistent with expectations, given that a Bachelors degree is
typically an entry requirement for professional jobs. Relative to completing high school,
professional women also receive a wage penalty for having less than a high school
education or for having a trade qualification. In this regard they differ significantly from
petty bourgeois women for whom these
qualifications

are associated with a small
24


nonsignificant wage premium.

There is also a class variation in the linear experien
ce effect
with employers, petty bourgeois women and managers all showing a negative linear
relati
onship between earnings and exp
er
ie
nce, while for white
-
collar and especially blue
collar women, the effect is significantly positive.

Based on a comparison of

these coefficients, I then fitted a model with constrained class
interactions

(Table 10)
.

There are three class by age linear interactions: the self
-
employed,
blue collar workers and others. There are four class by experience interaction levels: the self
employed; professionals, trade/technical and white collar employees; managers and blue
collar workers. There are also differential class effects by education as shown, with common
interaction structures for interactions involving having a Bachelor degree
a
nd a trade
qualification.

Figure 2 plots the age and experience income trajectories by class, and net patterns of wage
growth. There is no interaction between class and survey year, so all classes show a
common level of wage growth of approximately 1.6
percent per year, which is similar to that
for some men. Age earnings profiles for employee women are generally flatter than for
employee men, and begin to decline at quite young ages (early 30s) a factor which is
possibly associated

with differences in ea
rnings between mothers and non
-
mothers,
particularly if mothers incur a wage penalty when they re
-
enter the labor market. The
decline in earnings is slightly sharper for

blue collar

working class women than it is for
professional
,
managerial
and other working class
women. Among self
-
employed women on
the other hand, earnings increase with age, but wage growth begins to slow in the early
fifties. There are

relatively

few person
-
year observations for self
-
employed women so the
upward trend obser
ved in the graph may not be robust.

In contrast to
age, the experience earnings profile for employee women increases
monotonically. In all employee classes, earnings peak at around 35 years of labor force
experience. Again, working class women start at lo
wer wage levels
but wage growth rates
are relatively similar in all classes
. Self
-
employed women, in contrast, see wage rates decline
with experience, but again
these
results may not robust. Overall, figure 2 shows distinct
wage trajectories for working cl
ass women and middle class women
, which are largely a
consequence of the higher starting wages of professional and managerial women, rather
25


than substantial class differences in rates of growth or decline over the
life course
.

Working
class women’s earnin
gs

do, however,

begin to decline
at a slightly younger age

than middle
class women’s earnings do, which
partly
may reflect class differences in the timing of
fertility and labor market re
-
entry
.

Selected regression coefficients for the constrained class in
teraction model are shown in
Table 11. As for men, immigrant women from English speaking countries enjoy a small wage
premium over the Australian born and immigrants from non
-
English speaking countries,
however there is significant disadvantage associated
with not being a native English
speaker. Petty Bourgeois Women receive substantially greater returns for completing a
Bachelor degree than professional women do, while the university/college premium for
women in other classes is about twice that of the pre
mium for professional women. As
noted previously this result is consistent with a university degree being an entry
requirement for a professional labor market but not for other class segmented labor
markets. The disadvantage of not having a university degr
ee in professional labor markets is
shown by the eight percent wage penalty professional women incur for having a trade
qualification rather than completed high school, and the eighteen percent wage penalty for
a trade qualification relative to university
degree.

Failing to complete high school is also
associated with a substantial wage penalty for female employers and professionals and a
much small penalty for women from other employee classes.

The model also shows the negative impact of childcare responsi
bilities on women’s
earnings. Net of other factors there is a three percent wage penalty for having a resident
child under 15, and an additional three percent wage penalty for a preschool child. Bearing
in mind that the male and female earnings equations a
re specified differently, there is no
wage penalty for parenthood for men.

Conclusions

I have argued that a temporal view of class relations suggests that class structured or class
segmented labor markets may be associated differential returns to age, expe
rience and
education and different patterns of net wage growth.
Class structured labor markets

may
also affect the impact of job mobility/job shifts on earnings

and levels of net wage growth.
In particular, if individuals in middle class locations earn more than individuals in working
26


class jobs because of the existence of various kinds of employment rents, the value of these
rents may themselves vary over time.

For both men and women there was reasonable evidence of career immobility, at least over
a narrow window of eight years. This pattern is broadly consistent with the existence of
class segmented labor markets and undermines a central claim of theorists of
complex
inequality (cf Pakulski 2005) that individual differentiation rather than class structured
inequality characterises contemporary Australia. For men and women, there is also
evidence that middle class wage rates consistently exceed working class wag
e rates at
almost all levels of age and experience. This is the
life course

equivalent finding of earlier
cross
-
sectional analyses that showed systematic intercept differences in wage rates
between working class and middle class women and men (e.g. Western

1991; Halaby and
Weakliem 1993; Talmud et al, 2003). However, the analysis also showed that education
effects on earnings varied by class, for women and men, with trade and vocational
qualifications being particularly valuable when combined with self
-
empl
oyment, and with
the university degree premium being worth less in professional labor markets than others.
Returns to age and experience also showed class
-
specific variations for women and men
.

W
orking class men experienc
ed

earlier and steeper declines in
wage rates than professional
and managerial men
, blue collar men experienced lower levels of net wage growth over
time than all other classes except employers, and middle class men received greater returns
to early job experience than working class men, bu
t lesser returns to later job experience.

For women,
employee
class trajectories by age and experience were more similar than for
men, and class differentials tended to be preserved across the
life course
, more closely
reflecting different starting wage ra
tes.
The relative “flattening out” of class effects on
earnings for women compared to men

is consistent with certain unifying gendered aspects
of female labor force participation patterns


a relative concentration in a small number of
occupations and indu
stries with standardised conditions of employment and narrow wage
dispersion and a common experience of labor force interruption and accommodation to the
demands of domestic labor and childrearing. These gendered processes undermine class
differences in ea
rnings. However, to the extent that they also vary by class they have the
potential to magnify as well as suppress class variations.
In particular, class variations

in
27


things such as the timing of fertility and labor market re
-
entry, and the prevalence of
part
-
time rather than full
-
time work may act to magnify class variations over the
life course
.

Nonlinearities in the earnings effects of age, education and experience mean that class
structured labor markets do not easily lend themselves to conclusions t
hat returns are
systematically greater for one class than another

across the
life course
. Rather, class
trajectories vary

because class careers are produced from the intersection of class and other
proce
sses that involve organisations

and the institutions
of work, family, education and the
welfare state

(Mills, Blossfeld and Bernardi 2006, 2007).

Incorporating these elements into
theory and analysis is necessary for a comprehensive understanding of how class relations
shape earnings inequality through the s
tructuring of labor markets and careers.



28


References

Adams, M., (2007). Self and Social Change. Los Angeles, CA, Sage.

Australia (2006a). Australian
and New Zealand Standard Classification of Occupations
. Australian
Bureau of Statistics. Catalogue number

1220.0.

Canberra, Australian Bureau of Statistics.

Australia (2006b).
Australian and New Zealand Standard Industrial Classification.
Catalogue number
1292.0. Canberra, Australian Bureau of Statistics.

Australia 2008. Employee Earnings and Hours Australia.

Australian Bureau of Statistics, Catalogue
number 6306.0.
http://www.abs.gov.a
u/ausstats/abs@.nsf/Latestproducts/6306.0Main%20Features9999Au
g%202008?opendocument&tabname=Summary&prodno=6306.0&issue=Aug%202008&num
=&view
=. Accessed 28 September 2010. Canberra, Australian Bureau of Statistics.

Bernhardt, A., Morris, M., Handcock, M.S.
and Scott, M.A. (2001).
Divergent paths: Economic
mobility in the new American labour market
. New York, Russell Sage Foundation.

Block, F. 1990.
Postindustrial Possibilities
:
A Critique of Economic Discourse
. Berkeley, University of
California Press.

Bloss
feld, H.
-
P., Mills, M. and Bernardi, F., Ed. (2006).
Globalization, uncertainty and men's careers:
An international comparison
. Cheltenham; Northampton, Edward Elgar.

Blossfeld, H.
-
P., and Hofmeister, H. (2006).
Globalization, uncertainty and women's caree
rs: An
international comparison
. Northampton, Edward Elgar.

Breen, R. (2005). Foundations of a neo
-
Weberian class analysis. In Wright, E. O. (2005).
Approaches

to class analysis
. Cambridge; New York, Cambridge University Press. 31
-
50.

Butler, T. a. S., M.
, Ed. (1995).
Social change and the middle classes
. London; Bristol, UCL Press.

Chiswick, B. R., and Miller, P.W. (2006). Immigration to Australia during the 1990s: Institutional and
labour market influences.
Public policy and immigrant settlement
. D. A. a
. K. Cobb
-
Clark, S
-
E.
Cheltenham; Northampton, Edward Elgar
:
3
-
24.

Cobb
-
Clark, D. A. a. K., S
-
E., Ed. (2006).
Public policy and immigrant settlement
. Cheltenham;
Northampton, Edward Elgar.

Erikson, R., and Goldthorpe, J.H. (1992).
The constant flux: A
study of class mobility in industrial
societies
. New York, Oxford University Press.

Esping Andersen, G. (1993). Post
-
industrial class structures: an analytical framework. In Esping
-
Anderson, G. (ed).
Changing Classes: Stratification and Mobility in Post
-
In
dustrial Societies
.
London, Sage.

Esping
-
Andersen, G. (1999).
Social Foundations of Postindustrial Economies
. Oxford, Oxford
University Press.

Goldthorpe, J. H. (1995). The service class revisited.
Social change and the middle classes
. T. a. S.
Butler, M.

London; Bristol, UCL Press
:
313
-
329.

Goldthorpe, J. H. (2000).
On Sociology: Numbers, Narratives and the Integration of Research and
Theory.
Oxford: Oxford University Press.

Halaby, C. and Weakliem, D. (1993). “Ownership and authority in the earnings func
tion: Nonnested
tests of alternative specifications.”
American Sociological Review
.
58
: 16
-
30.

Hauser, R. M., and Warren, J.R. (1997). Socioeconomic indexes for occupations: A review, update,
and critique.
Sociological Methodology 1997
. A. E. Raferty.
Boston; Oxford, Blackwell
Publishers.
27:
177
-
298.

Kidd, M. P. (1993). "Sex discrimination and occupational segregation in the Australian labour
market."
Economic Record

69
(204): 44
-
55.

Kingston, P. (2000).
The classless society
. Stanford, Stanford Univers
ity Press.

Miller, P.W. and Neo, L. (1997). “Immigrant unemployment: the Australian experience”.
International Migration
.
35:

155
-
185.

Mills, M., Blossfeld, H
-
P., and Bernardi, F. (2006). Globalization, uncertainty and men's employment
careers: A theoretic
al framework.
Globalization, uncertainty and men's careers
. H.
-
P.
29


Blossfeld. Cheltenham; Northampton, Edward Elgar
:
3
-
37.

Morgan, S. L., and McKerrow, M.W. (2004). "Social class, rent destruction, and the earnings of black
and white men, 1982
-
2000."
Resear
ch in Social Stratification and Mobility

21
: 215
-
251.

Morgan, S. L., and Tang, Z. (2007). "Social class and workers' rent, 1983
-
2001."
Research in Social
Stratification and Mobility

25
: 273
-
293.

Pakulski, J. (2005). Foundations of a post
-
class analysis .
Wright, E. O. (2005). Approaches to class
analysis. Cambridge; New York, Cambridge University Press. 152
-
180.

Phillips, T., and Western, M. (2005). Social change and social identity: Postmodernity, reflexive
modernisation and the transformation of social i
dentities in Australia.
Class, consumption
and identity
. F. Devine, Savage, M., Crompton, R. and Scott, J. London, Palgrave
:
163
-
185.

Savage, M., Barlow, J., Dickens, P. and Fielding, T. (1992).
Property, bureaucracy and culture: Middle
-
class formation in
contemporary Britain
. London; New York, Routledge.

Savage, M., Barlow, J., Dickens, P. and Fielding, T. (2000).
Class analysis and social transformation
.
Buckingham; Philadelphia, Open University.

Savage, M. (2000).
Class analysis and social transformation
. Buckingham; Philadelphia, Open
University Press.

Shapiro, C., and Stiglitz, J. (1984). “Equilibrium Unemployment as a worker discipline device.”
American Economic Review
.
74:
433
-
444.

Shavit, Y., and Blossfeld, H
-
P. (1993). Persistent Inequality: Changin
g Educational Attainment in
Thirteen Countries. Boulder CO, Westview Press.

Sorensen, A. B. (2000). “Toward a sounder basis for class analysis”.
American Journal of Sociology
.
105: 1503
-
1528.

Talmud, I., Kraus, V. and Yonay, Y (2003). "Class analysis and e
arnings inequality: Nested and non
-
nested comparisons of two class models in Israel."
Quality and Quantity

37
: 21
-
41.

van Gellecum, Y., and Baxter, J.H. (2008). "Neoliberalism, gender inequality and the Australian
labour market."
Journal of Sociology

44
(1)
: 45
-
63.

van Gellecum, Y., Baxter, J., and Western, M. (2008). "Neoliberalism, gender inequality and the
Australian labour Market."
Journal of Sociology

44
(1): 45
-
64.

Watson, N. (2010).
HILDA User Manual


Release 8
. Melbourne, Melbourne Institute of Appli
ed
Economic and Social Research, University of Melbourne.

Watson, N. and Wooden, M. (2002). “Assessing the Quality of the HILDA Survey Wave 1 Data”.
HILDA Project Technical Paper. Melbourne, Melbourne Institute of Applied Economic and
Social Research, Univ
ersity of Melbourne.Western, M. (1991). The process of income
determination.
Class analysis and contemporary Australia
. J. Baxter, Emmison, M., Western,
J. and Western, M. Melbourne, The Macmillan Company of Australia.

Western, M., Baxter, J., Pakulski,
J., Tranter, B., Western, J., van Egmond, M., Chesters, J., Hosking,
A., O'Flaherty, M., and van Gellecum, Y. (2007). "Neoliberalism, inequality and politics: The
changing face of Australia."
Australian Journal of Social Issues

42
(3): 401
-
418.

Western, M.
C., and Baxter, J.H. (2001). The links between paid and unpaid work: Australia and
Sweden in the 1980s and 1990s.
Reconfigurations of class and gender
. J. H. a. W. Baxter,
M.C. Stanford, Stanford University Press
:
81
-
104.

Wright, E. O. (1985).
Classes
. Lon
don, Verso.

Wright, E. O. (1997).
Class counts: Comparative studies in class analysis
. Cambridge; New York,
Cambridge University Press.

Wright, E. O. (2005).
Approaches

to class analysis
. Cambridge; New York, Cambridge University
Press.





30


Table 1. Summa
ry Statistics: Men

Variable

Mean

Standard
Deviation

Min

Max

Log
Hourly
Earnings

3.117

0.408

2.388

6.173

Survey
Wave

4.187

2.312

1

8

Employer

0.073

0.261

0

1

Petty Bourgeois

0.029

0.168

0

1

Manager

0.146

0.353

0

1

Professional

0.218

0.413

0

1

Technical/Trade

0.176

0.381

0

1

White Collar

0.126

0.332

0

1

Blue Collar

0.231

0.422

0

1

Age
-
19 (years)

3.124

8.317

-
14

23

(Age
-

19) Squared

78.93

92.754

0

529

Bachelor Degree

0.3

0.458

0

1

Trade Qualification

0.416

0.493

0

1

High School

0.104

0.305

0

1

Less than High School

0.18

0.384

0

1

Experience
-

19 (years)

4.413

8.987

-
18.75

27.833

(Experience
-

19) squared

100.238

124.975

0

774.694

Agriculture, Forestry,
Fishing, Mining

0.056

0.229

0

1

Manufacturing,
Construction, Power

0.283

0.45

0

1

Business Services

0.148

0.355

0

1

Distributive services

0.226

0.418

0

1

Consumer, Personal
Services

0.047

0.212

0

1

Social Services

0.124

0.33

0

1

Public
Administration/Defence

0.117

0.321

0

1

Public Sector Employee

0.295

0.456

0

1

Job shift
last 12 months

0.139

0.346

0

1

More than one job

0.078

0.268

0

1

Full
-
time worker

0.903

0.296

0

1

31


Australian born

0.776

0.417

0

1

English speaking country

0.12

0.325

0

1

Non
-
English speaking
country

0.104

0.305

0

1

English first language

0.918

0.274

0

1

Married/de facto

0.794

0.405

0

1

Separated, divorced,
widowed

0.082

0.274

0

1

Never married

0.125

0.331

0

1

Resident dependent child

0.583

0.493

0

1

Resident preschool child

0.231

0.421

0

1

Instrument
al

Work
Orienta
t
ion

3.444

1.782

1

7

Importance of Financial
situation

7.984

1.589

0

10




32


Table 2. Summary Statistics: Women

Variable

Mean

Standard
Deviation

Min

Max

Log

Hourly

Earnings

2.994

0.367

2.388

5.6

Survey
Wave

4.19

2.333

1

8

Employer

0.04

0.195

0

1

Petty Bourgeois

0.016

0.126

0

1

Manager

0.083

0.276

0

1

Professional

0.329

0.47

0

1

Technical/Trade

0.031

0.172

0

1

White Collar

0.327

0.469

0

1

Blue Collar

0.175

0.38

0

1

Age
-
19 (years)

3.761

8.318

-
14

22

(Age
-

19) Squared

83.333

93.098

0

484

Bachelor Degree

0.356

0.479

0

1

Trade Qualification

0.258

0.438

0

1

High School

0.136

0.342

0

1

Less than High School

0.25

0.433

0

1

Experience
-

19 (years)

2.015

8.375

-
19

26.889

(Experience
-

19) squared

74.201

94.411

0

723.012

Agriculture, Forestry,
Fishing, Mining

0.012

0.111

0

1

Manufacturing,
Construction, Power

0.069

0.253

0

1

Business Services

0.166

0.372

0

1

Distributive services

0.187

0.39

0

1

Consumer, Personal
service

0.04

0.195

0

1

Social Services

0.448

0.497

0

1

Public
Administration/Defence

0.079

0.269

0

1

Public Sector Employee

0.485

0.5

0

1

Job shift in last 12 months

0.137

0.344

0

1

More than one job

0.098

0.297

0

1

Full
-
time worker

0.506

0.5

0

1

Australian born

0.781

0.414

0

1

English speaking country

0.106

0.308

0

1

33


Non
-
English speaking
country

0.113

0.317

0

1

English first language

0.904

0.294

0

1

Married/de facto

0.729

0.445

0

1

Separated, divorced,
widowed

0.152

0.359

0

1

Never married

0.119

0.324

0

1

Resident dependent child

0.605

0.489

0

1

Resident preschool child

0.144

0.351

0

1

Instrument
al

Work
Orienta
t
ion

3.166

1.665

1

7

Importance of Financial
situation

8.042

1.57

0

10




34



Table 3 Class Immobility Trajectories over Eight Survey Waves by Gender


Within Class
Fractions

Between
Fractions

Total

Concentration/Dispersion

N

Men

0.52

0.15

0.67

3.16

2724

Women

0.61

0.11

0.71

2.33

2534




35


Table 4. Top Ten Employment
Trajectories

Over Eight Waves: Men

Trajectory Elements

Frequency

Percent

Blue collar only

394

22.06

Professional only

263

14.73

Trade/Technical Only

262

14.67

Employer Only

193

10.81

Manager

-

Professional

145

8.12

Trade

-

Blue Collar

125

7

White Collar only

121

6.77

Manager Only

108

6.05

White Collar


䉬Be⁃ 汬慲



5⸱

坨P瑥⁃o汬慲

-


M慮慧敲



4⸷

呯瑡W

1ⰷ86

100




36


Table 5. Top
Ten Employment Trajectories Over Eight Waves:

Women

Trajectory Elements

Frequency

Percent

White collar only

514

25.46

Professional only

479

23.72

Blue collar only

327

16.2

Professional


坨P瑥⁃ 汬慲

147

7⸲.

坨P瑥⁃o汬慲a
-

䉬Be⁃ 汬慲

112

5⸵.

M慮慧敲e


坨PW攠Co汬慲

108

5⸳.

M慮慧敲e
-

Pro晥獳fon慬



4⸵.

䕭p汯y敲nly



4⸴.

P牯f敳獩on慬a


䉬B攠Co汬慲



4⸳.

M慮慧敲e


Prof敳獩en慬


坨P瑥⁃o汬慲



3⸲.

呯瑡W

2ⰰ19

100

37


Table 6.

Men’s Class Specific Returns to Education, Age, Experience, Job Shifting and Survey Wave: Unconstrained
Class Interaction Model.


Age
(coefficie
nt by 100)

Age
squared
(coefficie
nt by
1000)

Experienc
e squared
(coefficie
nt by
1000)

Bachelo
r
Degree

Trade
qua
lification/po
st secondary

Part
High
Schoo
l

Job Shift
(coefficie
nt by 10)

Survey
Wave
(coefficie
nt by 10)

Employer


-
0.75
*


0.54

-
0.37

0.19
***

-
0.07

-
0.05


0.86
*

0.08

Petty
Bourgeois


-
0.56

-
0.19

-
0.09

0.17
*


0.14
*


0.09


0.59

0.20
*

Manager


-
0.07

-
0.12

-
0.31
*

0.23
***

-
0.08
**

-
0.13
**
*

-
0.04

0.25
***

Professional


0.17

-
0.17

-
0.49
***

0.19
***

-
0.04

-
0.12
**


0.07

0.22
***

Technical/Tra
de


-
0.17

-
0.57
**


0.20

0.20
***

-
0.03

-
0.17
**
*


0.08

0.21
***

White Collar


0.12

-
0.65
**


0.05

0.19
***

-
0.05
*

-
0.14
**
*

-
0.53
***

0.17
***

Blue Collar


-
0.26
***

-
0.77
***


0.33
*

0.12
***

-
0.08
***

-
0.16
**
*

-
0.03

0.13
***


*
p < .05
**
p < .01
***
p < .001

Notes:

Age: Manager, Professional and White Collar coefficients significantly larger than Blue Collar coefficient.

Age Squared: Employer, Manager and Professional coefficients significantly larger than Blue Collar coefficient.

Experience Squared: Employer, Manager and Professional coefficients significantly smaller than Blue Collar
coefficient.

Job Shift: Employer c
oefficient significantly larger than Blue Collar Coefficient.

Education: Petty Bourgeois Trade and Part High School coefficients significantly larger than Blue Collar coefficients.

Survey Wave: Professional, Manager and Technical/Trade Coefficients signifi
cantly larger than Blue Collar
coefficients.



38



Table 7. Equality Constraints for Men’s Constrained Class Interaction Model of Log Hourly Earnings

Parameter

Constraints

Class by Age

Emp PB; Ma TT; Pr WC; BC

Class by Age Squared

Emp; PB Ma
Pr; TT WC BC

Class by Experience Squared

Emp Ma Pr; PB; TT; WC; BC

Class by Degree

Emp PB Ma P TT WC; BC

Class by Trade

PB; Emp Ma Pr TT WC BC

Class by Part High School

Emp; PB; Ma Pr TT WC BC

Class by Job Shift

Emp P
B; Pr TT; Ma BC; WC

Class by Survey Year

PB Ma Pr TT WC; Emp; BC



Notes: Constraint levels separated by semicolons (;). Emp

= Employer, PB = Petty Bourgeois; Ma = Manager, Pr =
Professional, TT = Technical/Trade, WC = White Collar, BC = Blue Collar.

39




Table 8.

Selected Regression Coefficients for Constrained Class Interaction Model of Log Hourly Earnings: Men

Independent Variable

Regression
Coefficient

Born in English speaking country

0.08
***

Born in non
-
English speaking country

-
0.02

English as first
language

0.03

Blue collar Bachelor Degree

0.14
***

Other classes Bachelor Degree

0.20
***

Petty Bourgeois Trade Qualification

0.17
***

Other classes Trade Qualification

-
0.06
***

Employer Part High School

-
0.04

Petty Bourgeois Part High School

0.12

Other Classes Part High School

-
0.14
***

Employer and Petty Bourgeois job move (coefficient by 10)

0.76
*

Professional and Technical/Trade job move (coefficient by 100)

0.78

White collar job move (coefficient by 10)

-
0.53

Managerial and blue collar job m
ove (coefficient by 100)

-
0.52




40




Table 9. Women’s Class Specific Returns to Age, Education and Experience: Unconstrained Class Interaction Model


Age (coefficient
by 10)

Bachelor
Degree

Trade
qualification/post
secondary

Part High
School

Experience
(coefficient by
100)

Employer


0.19

0.21
***

-
0.02

-
0.15

-
0.14
***

Petty Bourgeois


0.11

0.37
***


0.09


0.18

-
0.13
*

Manager

-
0.002

0.20
***

-
0.01

-
0.06

-
0.5
*

Professional


0.01

0.09
***

-
0.08
***

-
0.15
***


0.03

Technical/Trade

-
0.05

0.21
***


0.003

-
0.09


0.07

White Collar

-
0.04

0.19
***

-
0.01

-
0.03


0.07
***

Blue Collar

-
0.31
*

0.20
***


0.01

-
0.05


0.57
***


*
p < .05
**
p < .01
***
p < .001


Notes:

Age: Employer, Petty Bourgeois, and Professional coefficients larger than Blue Collar coefficient.

Education: Professional degree, trade and part high school coefficients smaller than Blue Collar coefficients.

Experience: Employer and Petty Bourgeois coefficient smaller than Blue Collar coefficient.

Experience: Employer and Petty Bourgeois coefficient
smaller than Blue Collar coefficient.



41


Table 10. Equality Constraints for Constrained Class Interaction Model of Log Hourly Earnings: Women

Parameter

Constraints

Class by Age

Emp PB; Ma Pr TT WC; BC

Class by Experience

Emp

PB; Pr TT WC; Ma; BC

Class by Degree

Emp Ma TT WC BC; PB; Pr

Class by Trade

Emp Ma TT WC BC; PB; Pr

Class by Part High School

Emp Pr; PB; Ma TT WC BC



Notes: Constraint levels separated by semicolons (;). Emp

= Employer, PB = Petty Bourgeois; Ma = Manager, Pr =
Professional, TT = Technical/Trade, WC = White Collar, BC = Blue Collar.





42


Table 11. Selected Regression Coefficients for Constrained Class Interaction Model of Log Hourly Earnings:

Women.


Independen
t Variable

Regression
Coefficient

Born in English speaking country

0.05
**

Born in non
-
English speaking country

-
0.04

English as first language

-
0.03

Employer, Manager, Tech/Trade White Collar and Blue collar Bachelor
Degree

0.20
***

Petty Bourgeois
Bachelor Degree

0.38
**

Professional Bachelor Degree

0.10
***

Employer, Manager, Tech/Trade White Collar and Blue collar Trade
Qualification

-
0.01

Petty Bourgeois Trade Qualification

0.08

Professional Trade Qualification

-
0.08
**

Employer, Professional
Part High School

-
0.15
***

Petty Bourgeois Part High School

0.16

Other Classes Part High School

-
0.04
*

Resident Child

-
0.03
**

Resident Preschool Child

-
0.03
***

43



Figure 1.

Predicted Log Hourly Earnings by Age, Experience and Survey Year: Men


2.8
3
3.2
3.4
Log Hourly Earnings
20
30
40
50
60
Age: Years
Age
2.8
3
3.2
3.4
0
10
20
30
40
Experience: Years
Experience
2.8
3
3.2
3.4
Log Hourly Earnings
2000
2002
2004
2006
2008
Year
Year
Emp
PB
Man
Prof
Trade
WC
BC
44




Figure 2. Predicted Log Hourly Earnings by Age, Experience and Survey Year: Women

2.8
3
3.2
3.4
3.6
Log Hourly Earnings
20
30
40
50
60
Age: Years
Age
2.8
3
3.2
3.4
3.6
0
10
20
30
40
Experience: Years
Experience
2.8
3
3.2
3.4
3.6
Log Hourly Earnings
2000
2002
2004
2006
2008
Year
Year
Emp
PB
Man
Prof
Trade
WC
BC