A Precarious life: The effect of non-permanent contractual employment on financial
hardship
Neha Swami
Melbourne Institute of Applied Economics and Social Research and
Department of Economics
University of Melbourne
Melbourne, VIC 3010 Australia
[email protected]
Abstract
The increasing incidence of non-permanent contractual forms of employment raises concerns
about the consequences of these types of employment on workers’ well-being. This study
uses data from the Household, Income and Labour Dynamics in Australia (HILDA) survey to
examine how fixed-term, and casual employment and other employment characteristics
affect financial hardship among individuals. I estimate multivariate ordered logit fixed-effects
models of the relationship between employment and financial hardship that account for
unobserved heterogeneity, and a rich set of individual level characteristics. My results
indicate that, for men, casual employment is associated with more financial hardships after
controlling for a rich set of observed factors. However, mechnisms analysis suggests once I
account for level of income, employment shocks (job loss and job change), and hours worked,
estimates become considerably weaker in magnitute as well as significance. For women,
particularly those with caring reposibilities, accounting for potential mechanisms (income,
employment shocks, and hours worked) has relatively small effect on the estimates of casual
employment and it continues to be directly and significantly associated with more financial
hardships. Fixed-term employment is not associated with more financial hardship for either
men or women.
Keywords: Non-permanent employment; financial hardship; HILDA Survey; blow-up and
cluster fixed-effects ordered-logit model
*The paper uses data collected from the Household, Income and Labour Dynamics in Australia
(HILDA) Survey. The HILDA Project was initiated and is funded by the Commonwealth
Department of Family, Housing, Community Services and Indigenous Affairs and is managed
by the Melbourne Institute of Applied Economic and Social Research. The paper’s analysis
data were extracted using PanelWhiz v4.0, a Stata add-on package written by Professor John
P. Haisken-DeNew. The .do files to retrieve and analyse the data are available upon request.
The findings and views reported in this paper are solely of the author. The author sincerely
thank Professor David Ribar, Professor Jenny Williams and Professor Mark Wooden for their
guidance and helpful comments, and all the seminar participants for their valuale feedback
and suggestions on this paper.
1. Introduction and Motivation
The changing nature of the overall structure of employment in many western countries (ILO,
2008a, 2008b) has given rise to various alternative forms of employment arrangements that
differ from the standard permanent full-time employment in terms of continuity and hours
worked in varying degrees. This growth in the proportional share of ‘alternative or nonstandard’ employment is often used as evidence that employment is now less secure and
more precarious.
In this study, I examine the relationship between non-ongoing or non-permanent
employment arrangements defined by a lack of a legal guarantee of a long-lasting
employment relationship, and financial hardship. I focus on non-permanent employment
because, by definition, such an employment relationship is likely to be associated with more
variability in employment status (Giesecke, 2009; Giesecke & Groß, 2003; OECD, 2015a;
Watson, 2013) which in turn could lead to more uncertain income flows and lower income
levels. Non-permanent employment is also likely to be less secure than standard permanent
employment in terms of non-wage benifits such as paid sick and holiday leave, wages, and
hours worked (Campbell, Whitehouse, & Baxter, 2009; Kalleberg, Reskin, & Hudson, 2000;
McGovern, Smeaton, & Hill, 2004). I consider two types of non-permanent employment
arrangements: Fixed-term employment and casual employment, and define permanent
employment as standard employment.
A growing body of evidence shows that while permanent contractual employment
remains the norm in all OECD countries, the share of jobs that are non-permanent has
increased in most OECD countries since the 1980s (ILO, 2015; OECD, 1996, 2002, 2014,
2015a). That said, when averaged across all OECD countries with the available data, the
increase in the share of non-permanent of the total dependent employment has been
1
modest, rising from about 9% in the early 1980s to 11% in 2014 (OECD, 2015b). However,
these trends in the non-permanent employment share differ considerably across countries
(OECD, 2014, 2015a) as do the definitions and regulations governing these employment
arrangements. For example, since the 1980s non-permanent employment has grown
considerably in Netherlands, France, Spain, Poland, and Portugal (Fuller & Vosko, 2008; OECD,
2002, 2014, 2015a). At the same time, it has remained more or less stable in the U.K
(McOrmond, 2004) and even declined in countries such U.S and Greece (Houseman & Osawa,
2003; Lane, 2001; OECD, 2002, 2014).
Australia also has a high incidence of non-permanent employment (Will, 2013;
Wooden & Warren, 2004). Two main types of non-permanent employment in Australia are
casual employment and fixed-term employment. A casual employment contract is defined to
be one in which the employee is paid by the hour for every hour of work. Fixed-term
employment is defined as jobs of fixed duration without a guarantee of employment renewal.
The major contributing factor to the increased use of non-permanent employment in
Australia has been casual employment (ABS, 2013; OECD, 1996, 2015a). There was a
trajectory of strong growth in the proportion of casual employees rising from 13% in 1984 to
around 20% in 1996 (ILO, 2008a; OECD, 1996), but after the mid-1990s the share has
remained relatively stable (Wilkins & Wooden, 2014). For fixed-term employment,
information from HILDA, forms of employment surveys and the Survey of Employment
Arrangements and Superannuation (SEAS) suggests no clear trend but supports the
conclusion that fixed-term employment constitutes a relatively small proportion of employed
persons in Australia (Will, 2013). By 2013, casual employees represented 20% of the
Australian workforce, and fixed-term employees accounted for 6% (OECD, 2015a).
2
Australia presents an interesting context to study the relationship between financial
hardship and employment because of its unique labour regulations (the ‘Award’ system) and
industrial regulation systems. Casual employees do not have standard rights and benefits like
protection against dismissal, a right to redundancy pay, paid annual leave, paid sick leave, and
guaranteed hours of work (Campbell, 1996; Watson, Buchanan, Campbell, & Briggs, 2003).
However, the Australian labour regulation requires casual employees to receive a premium
(casual loading) to compensate them for the lack of these entitlements benefits (Will, 2013),
but evidence suggests that casual employees still suffer from wage penalties compared with
permanent employees (Watson, 2005; Wooden & Warren, 2003). While it might be
problematic to compare different forms of non-permanent employment in various countries,
casual employment in Australia does share some or many characteristics of non-permanent
or temporary employment in other OECD countries.
Workers on fixed-term contracts in Australia are not eligible for termination and
redundancy pay at the completion of the contract, and they do not receive any special
compensation as they otherwise tend to be entitled to similar benefits, do similar jobs and
receive similar wages as permanent employees (Burgess & Campbell, 1998; Will, 2013;
Wooden & Warren, 2003). Hence, in Australia even though both fixed-term and casual
employment are non-permanent, they are very different in terms of their characteristics and
the purpose for which they are used.
The growth in non-parmanent employment in the majority of OECD countries has
raised concerns that these jobs may lead to labour market segmentation and dualism with
non-permanent employees generally receiving low and uncertain income, limited access to
statutory benefits, limited job security and limited prospects of career progression compared
to their permanent counterparts (Fuller & Vosko, 2008; Houseman & Osawa, 2003; OECD,
3
2014, 2015a). It is often claimed or speculated that these factors could have adverse
consequences in terms of various financial hardships or difficulties (Eamon & Wu, 2011; Lovell
& Oh, 2006; OECD, 2015a). However, non-permanent employment may be beneficial for
some workers. It may serve as a crucial entry point into the labour market for the unemployed
and people with limited or no work experience. Workers who wish to combine work and other
commitments (like child care) may also benefit from the flexible schedules associated with
some forms of non-permanent employment (OECD, 2014).
Despite these claims, there is a surprising lack of convincing empirical evidence on the
financial hardship consequences of non-permanent employment. Indeed, the only
quantitative study to examine financial hardship issues is the cross-sectional analysis of
Buchler, Haynes, & Baxter (2009) that used HILDA data and examined the differences in
financial difficulty between casual and permanent employees.
This paper aims to fill this gap in the literature by providing new evidence on whether
and to what extent non-parmanent employment affects Australian men's and women's risks
of facing financial hardships. My analysis uses panel data from the Household, Income and
Labour Dynamics in Australia (HILDA) survey and utilises self-reported information on
financial hardships and employment situations. I consider two very distinct forms of nonpermanent employment in Australia: fixed-term contract jobs and casual jobs. In order to
account for individual time-invariant characteristics I estimate fixed-effects models.
Moreover, the HILDA Survey also collects data on an extensive set of covariates which allows
me to control for a rich set of observable worker and job characteristics that might confound
the relationship between employment and financial hardship, such mental and physical
heath, family composition and job history.
4
Finally, evidence from the literature suggests that experience of financial hardship,
the consequence of non-permanent employment, and other characteristics such as hours
worked may differ significantly for men and women (McGovern et al., 2004; OECD, 2015a; Yu,
2012). Moreover, changes in family composition are likely to influence people’s preference
for flexibility in the job (Watson et al., 2003). For example, workers with family
responsibilities, especially women, who face greater pressure to accommodate employment
to the demands of household labour, might trade off a higher wage or a permanent job for
more flexibility to balance work and non-work commitments. This could mean that they may
have to compromise their working conditions to take care of the children (Boden, 1999) with
economic losses mitigated by saving the cost of formal child care (Presser, 1989, 1995).
Hence, it is important that when examining the differences between permanent and nonpermanent employees in terms of some outcomes, we compare persons who are otherwise
as homogeneous in terms of their family situation to have relevant comparison group.
Therefore, I disaggregate the analysis by gender as well as family type and fortunately the
HILDA data provides me with enough sample size to disaggregate the sample1.
2. Previous literature on the consequences of non-standard employment
There has been a considerable amount of research on the consequences of non-permanent
employment and non-regular employment on various labour market outcomes for workers.
Studies have examined the effect of non-regular or insecure employment on health and safety
outcomes (Kim & von dem Knesebeck, 2015; Llena-Nozal, 2009; McNamara, 2006), on overall
1
Disaggregating the analysis by gender and family type also allows the effects of other factors, including not
only employment type but also mental and physical health, income, the experience of unemployment, and parttime work to vary by gender and family type. For example, the degree to which the risk of facing financial
hardships will be responsive to job loss or change in hours of work is likely to be different depending upon if the
person has dependents to provide for while being the sole earner of the family that can add additional financial
stress.
5
job satisfaction (Buddelmeyer, McVicar, & Wooden, 2015; Green & Heywood, 2011; OECD,
2014; VandenHeuvel & Wooden, 2000; Wilkin, 2013; Wooden & Warren, 2004) as well as on
perceptions of job insecurity (Booth, Francesconi, & Frank, 2002; Graaf-Zijl, 2005; Green &
Heywood, 2011; OECD, 2014). The literature that is most relevant for this study includes
studies that have examined the economic consequences for those in non-permanent
employment such as wage penalties, access to training, the probability of job loss and job
mobility, which can have direct implications for a person’s financial situation.
Many empirical studies have examined the wage differences between non-permanent
and permanent employment (Booth et al., 2002; Forde & Slater, 2005; Fuller & Vosko, 2008;
Kahn, 2013; Kalleberg, 2000; OECD, 2002, 2014, 2015a). Evidence from these studies often
suggests a significant difference in earnings between non-permanent and permanent
employees. For example, the OECD (2015a) found that on average men and women who are
non-permanent (or temporary) workers in all OECD countries face wage penalties (11% for
men and 13% for women) even after controlling for individual, family and work
characteristics. But there were clear differences across countries with wage penalty for nonpermanent workers ranging from 2-3% in Australia to 19% in Greece. Moreover, the average
wage penalty across countries reduced when individual fixed-effects were taken into account
(5% for men and 8% for women), and for Australia the penalty declined to almost zero. The
study also revealed that in most of the countries, the earning gap between non-permanent
and permanent employees is significantly larger at the bottom of the wage distribution.
Apart from wages, work-related training is also an important aspect of employment in
terms of improving workers’ skills and productivity which in turn enhances their earning
potential and improves their ability to transition to better and more secure jobs. Evidence
from the literature clearly indicates that non-permanent employees receive considerably less
6
work-related training from their employers than their permanent counterparts (Booth et al.,
2002; Buddelmeyer, Leung, McVicar, & Wooden, 2013; ILO, 2015; OECD, 2002, 2014). For
example, OECD (2014) found that on average being on a non-permanent contract reduces the
probability of receiving employer-sponsored training by 14% (around 27% in Estonia, France
and the Slovak Republic). Buddelmeyer et al. (2013) used HILDA data and found that casual
employees are less likely to participate in work-related training than those in permanent and
fixed-term jobs. The reason for this could be that the expected duration of non-permanent
jobs is shorter and hence, employers have less time to recoup the cost of training (Booth et
al., 2002; Buddelmeyer et al., 2013).
Another very important aspect of job quality is job security. Some studies have
examined the precariousness of non-permanent employment using the probability of job loss
in the next 6-12 months as a proxy for job insecurity (OECD, 2014, 2015a; Watson, 2013).
These studies find that non-permanent employees typically face worse job security than
permanent workers, although there are differences across gender and contract types. The
results from an empirical analysis by the OECD (2015a) showed that compared with a
permanent job, holding a non-permanent job increases the likelihood of unemployment in
the next six months in all OECD countries, especially for men. Watson (2013) estimated
random intercept MNL models using HILDA data and found that casual employees are more
likely to become jobless (unemployed or out of labour force) than fixed-term employees.
The evidence from the studies reviewed so far suggests that on average nonpermanent employees have worse labour market outcomes than permanent employees in
terms of earnings, on-the-job training and job security. However, as discussed earlier nonpermanent jobs may provide people with opportunities to enter or re-enter labour force and
gain work experience to move into better and more stable jobs in the future. The question of
7
whether non-permanent jobs are “stepping stones” to more stable/permanent employment
has been the focus of considerable amount of research in Australia (Buddelmeyer & Wooden,
2011; Chalmers & Kalb, 2001; Watson, 2013) as well as in other countries (Booth et al., 2002;
de Graaf-Zijl, Van den Berg, & Heyma, 2011; Gash, 2008; Giesecke & Groß, 2003; Mooi‐Reci
& Dekker, 2015; OECD, 2015a; Yu, 2012). Empirical evidence from these studies suggests that
some non-permanent employees manage to move into permanent jobs while others get
trapped in non-permanent jobs based on the type of contract and other worker
characteristics. For example, mobility into permanent jobs tends to be higher for men than
women (Buddelmeyer & Wooden, 2011; de Graaf-Zijl et al., 2011; Watson, 2013); higher for
fixed-term employment or high status non-permanent jobs than seasonal or casual
employment or low status non-permanent jobs (Booth et al., 2002; Buddelmeyer & Wooden,
2011; Watson, 2013; Yu, 2012); higher for prime-age workers (Chalmers & Kalb, 2001); higher
for full-time worker than part-time workers (Booth et al., 2002; OECD, 2015a); and higher for
single persons compared to persons with partner (de Graaf-Zijl et al., 2011; Yu, 2012).
Several studies have examined how some labour market outcomes and negative
income shocks like job loss/unemployment are associated with financial hardship (Bauman,
2002; Eamon & Wu, 2011; Gundersen & Gruber, 2001; Lovell & Oh, 2006; OECD, 2015a).
Bauman (2002) examined the relationship between work, welfare payments and financial
hardship using the Survey of Income and Programme Participation (SIPP) in the US and found
that hardship was high among households that worked part year and those who were in the
past or currently on welfare receipts. Eamon & Wu (2011) used data from SIPP and
multivariate methods to find that unemployment and involuntary job loss increased the risk
of facing several financial hardships among single mother families. A recent study by the OECD
(2015a), on the link between non-permanent and part-time employment and household
8
income and poverty found that more than one-third of the low earning non-standard workers
belonged to low-income households.
As discussed earlier, to my knowledge, only one study has directly examined the
relationship between financial hardship and employment type, a cross-sectional analysis by
Buchler et al. (2009). The authors used the HILDA Survey data and restricted the sample to
include only employees to examine the difference in financial wellbeing between casual and
permanent employees. The authors estimated multivariate models along with various
controls to find that casual employees face significant financial stress, such as not been able
to pay utility bills and rent on time, going without meals, etc.
Buchler et al. (2009) did not control for individual specific time invariant
heterogeneity. Controlling for individual heterogeneity is important as many negative
consequences could be driven by individual differences in factors such as motivation to work.
Moreover, by restricting their sample to employees, the authors may have introduced sample
selection issues. Finally, it is also important to consider fixed-term employment because it
differs quite significantly from casual employment in terms of benefits and other conditions
(Wooden & Warren, 2003) and examining the two types of non-permanent could advance
our understanding of the factors that are important to avoid financial hardship. This study
aims to fill these gaps.
3. Conceptual framework linking non-permanent employment and financial hardship
To consider how non-permanent employment might contribute to financial hardship, I adapt
the simple framework used by Jappelli & Pistaferri (2010) and Meghir & Pistaferri (2011) that
explains how changes in labour income might affect the consumption level of an individual.
In this framework, an individual maximises his/her expected lifetime utility of consumption
9
over a fixed time horizon. The individual faces an intertemporal lifetime budget constraint in
which his/her expenditure on consumption goods and services in any given period is
constrained by his/her per-period labour income, other unearned income, returns on
investments, and borrowings in that period. The individual can lend and borrow at the same
interest rate (perfect credit markets); the utility function is state and time-separable; the
individual faces a terminal condition on wealth; and all variables in the model are known
except the individual’s labour earnings. Given these assumptions, ex-ante an individual’s
current level of consumption is the best predictor of the next period level of consumption;
ex-post, consumption will change only if expectations regarding labour earnings are not
realised. The assumptions made in this model are restrictive, but this model provides useful
insights into the effect of changes in level and stability of labour earnings on consumption.
In this framework, one can define financial hardship in a very simple way. A person
experiences financial hardship if his/her level of consumption of goods and services falls
below a particular threshold in any given time period (say C*(t)) due to lack of sufficient
income/financial resources required to satisfy consumption needs like food, shelter, and
utilities.
Under the maintained assumptions the model generates the following predictions:
First, current consumption will respond to an anticipated change in future income level. For
example, an increase in consumption might occur immediately after the announcement of a
tax reform or a promotion at work. Similarly, a decrease in consumption might occur if labour
income is expected to decline in the future as consumers will reduce current consumption to
increase savings, especially if they are liquidity constrained.
Second, current consumption will respond to unanticipated earning and income
shocks. For example, a sudden job loss or an unexpected decline in pay levels and earnings
10
might severely affect an individual’s level of consumption. The consumption response to
unanticipated income shocks depends critically on the degree of persistence in the shock,
with more persistent shocks leading to more volatile consumption and increased possibility of
experiencing financial hardship. However, if variations in labour income are transitory, the
effect on consumption will be small as individuals can smooth consumption by borrowing
today (if they can) and repaying in the future when income returns to the mean value.
This framework is very helpful in illustrating why people working in non-permanent
jobs are at a greater risk of experiencing financial hardship than those working in permanent
jobs. First, non-permanent employees are more likely to receive lower wages (even casuals)
compared to permanent employees (Forde & Slater, 2005; Kalleberg et al., 2000; McGovern
et al., 2004; Wooden & Warren, 2003).
Second, they are also more susceptible to experiencing labour income/employment
shocks due to a higher probability of experiencing job separations (Watson, 2013; Wilkins &
Wooden, 2013; Wooden & Warren, 2003). These employment shocks can make income highly
volatile and can severely affect a person’s ability to pay for necessary expenses.
Third, non-permanent jobs are more likely to be associated with pay/hours volatility
given many non-permanent employees, particularly casuals, are hired to work irregularly. For
these workers, there is no guarantee of fixed work schedules or hours which in turn can lead
to unpredictable and fluctuating pay (Campbell et al., 2009; ILO, 2008a, 2008b). Also, nonpermanent employees are more likely to be employed part-time compared to permanent
employees (ABS, 2010; Wooden & Warren, 2003) which can further aggravate the situation
due to insufficient earnings.
Finally, non-permanent employment is often associated with a lack of benefits like
paid sick/carer leave, redundancy pay and other workplace protections (Will, 2013), that can
11
help smooth consumption in case of unforeseen or unexpected events like illness and job loss.
Hence, although non-permanent employment might make it possible for workers to fit
employment hours around their family and other commitments, it might come at the expense
of financial security.
In conclusion, economic theory predicts that, non-permanent employees are more
vulnerable to experiencing financial hardship than permanent employees because they are
not only more likely to receive lower wages but also likely to face more unemployment
spells/shocks. Moreover, non-permanent employees are more likely to have irregular hours
of work leading to unpredictable earnings and this is further exacerbated by
underemployment as non-permanent employees are more likely to be employed part-time.
Also, non-permanent employees lack standard statutory benefits and protections. All these
employment conditions may adversely affect the level and stability of earnings, and
contribute to financial hardships.
4. Research data
The data used in this study come from the Household Income and Labour Dynamics in
Australia (HILDA) Survey (see Watson & Wooden (2012) for more details). Commencing in
2001, it is a nationally representative annual longitudinal survey of youths and adults from
more than 7,600 households with an emphasis on employment, income and family. Each
wave of the survey includes an interview-based personal questionnaire (PQ) and a written
self-completed questionnaire, as well as a household questionnaire that is administered by
the interviewer to one person in the household. The re-interview rates for HILDA have been
high, increasing from 87% in wave 2 to over 94% by wave 5. The first 13 waves of HILDA data
(but excluding wave 10 as hardship question was not asked in wave 10) are used in the
12
analysis. The interviews start in July-August of the calendar (wave) year, and the majority of
interviews (more than 85%) are completed by November of the calendar (wave) year.
4.1 Measuring financial hardship
With the exception of wave 10, the SCQ of the HILDA survey asks respondents about
seven types of hardship that they might have experienced since the start of the calendar year
‘because of a shortage of money’. The hardships include: 1) an inability to pay electricity, gas
or telephone bills on time; 2) an inability to pay the mortgage or rent on time; 3) pawning or
selling something; 4) going without meals; 5) an inability to heat the home; 6) asking for help
from welfare or community organizations, and 7) asking for financial help from family or
friends2. Following Butterworth & Crosier (2005), I use this information to create binary
indicator variables for affirmative responses to each of the seven hardship questions and sum
those variables to form an index of financial hardship with a possible range of 0 to 7. The index
is coded as missing if the response on any one of the hardship questions is missing for the
individual.
Following Bray (2001), there are some studies (Breunig & Cobb-Clark, 2006; Bubonya,
Cobb-Clark, & Wooden, 2014) that have used two subscales; one defining hardships (missing
meals, pawning something, inability to heat the home and asking help from welfare) and
other cash flow problems (inability to pay rent/mortgage, inability to pay utilities, and
borrowing from friends or family). However, Butterworth & Crosier (2005) showed that these
hardship factors could be validly summed into a single index since they are highly correlated
in HILDA data. Studies that have used the single index measure of financial hardship
2
As three of the seven hardship questions are measured at household level it is important couples agree in
these response to these questions. Hence, I also examined the concordance in responses between couples in
the HILDA and found that overall it 90.8% (see appendix).
13
constructed from the HILDA data include (Buchler et al., 2009; Cobb-Clark & Ribar, 2012;
Ribar, 2015).
It is important to note some limitations of this financial hardship measure. As noted
by Butterworth & Crosier (2005), hardship indicators are not neutral as they involve subjective
decisions about selecting what constitutes hardship or basic needs. It is also a crude measure
since it only counts the different types of hardship events occurring and not how many times
each occurred during the year or the severity. For this reason I consider my index as an
ordinal, rather than cardinal measure of financial hardship.
4.2 Measuring employment arrangement
Each wave of the HILDA Survey collects extensive information about the respondent’s
main job held at the time of the interview, including the nature of the contract of
employment. All employees are asked to classify themselves into one of three mutually
exclusive categories of employment arrangements: (i) permanent or ongoing; (ii) casual; (iii)
fixed-term; and a small fourth category ‘other’, for persons who cannot be easily
characterised in other three categories. Additionally, the survey identifies persons who work
without pay in a family business (403 observations).
The survey further identifies three types of non-employees: (i) persons who work in
their own businesses, (ii) the unemployed, and (iii) those that are out of the labour force
(OOLF). I also include theses three categories of non-employees to reduce sample selection
issues and to compare relative financial disadvantage of non-permanent employees with the
non-employees. I create indicator variables (0,1) for each of the 8 discussed categories and
use permanent employment as the reference category. Even though I include separate
dummies for the ‘other’ employment category and unpaid family members category but
14
because of very small sample size I am not able to estimate precisely the estimates for people
in these categories.
As with the hardship measure, employment measures have some limitations as well.
The type of employment contract of the main job is measured at the time of interview and
job start and end dates are not recorded. Thus, it indicates the type of employment that a
respondent had at the time of the interview and not over the entire year, unlike the hardship
measure.
4.3 Explanatory variables
My empirical analyses also account for many observed characteristics. For the reasons
justified earlier in the paper all of the empirical analyses are conducted separately for men
and women within the following family types: 1) Couples without dependent children; 2)
Couples with dependent children; 3) Lone parents with dependent children and 4) Single
persons3. The proportion of men who are lone parents with dependent children is very small
in the analysis sample with only 711 observations4. As a consequence, it is difficult to estimate
precisely the effects of non-permanent employment for this sub-sample. Hence, I do not
report results for this group.
I also include several time-varying explanatory variables that are arguably exogenous.
These include four dummy variables for the respondent’s age (aged 18-24, aged 25-34, aged
45-54 or aged 55-65, with individuals aged between 35-44 as the base category); five
dummies for the highest level of education achieved (bachelor's or honours, diploma degree,
3
The first category includes couples (either married or defacto) that don’t have any children or those who have
non-dependent children; the second category includes couples (either married or defacto) that have dependent
children <15 or dependent students; the third category includes lone parents with dependent children or
dependent students; the fourth category includes single unmarried individuals or lone parents but with nondependent children.
4 The proportion is small even in the overall data. Only 1.84% (1,595 obs.) of men are lone parents with
dependent children.
15
certificate lll or lV, year 12 and year 11 or less, with postgrad as the base category); and two
indicators for living either in regional or remote area with living in city as the base category.
My empirical analyses also incorporate time-varying controls that might be potentially
endogenous. These include controls for job-related factors including an indicator for being a
member of union or employee association; a control for the number of jobs held by a
respondent in the preceding financial year (held no job or held multiple jobs, with held one
job as the base category); controls for industry (19 categories with working in ‘other services’
as the base category); controls for occupation (8 categories with being a manager as the base
category); and a control for private sector employee. I also include various health related
controls including three indicators for the presence of a long-term health condition that
affects the individual’s ability to work (mild effect, moderate effect and severe effect with no
condition as base category); and controls for mental and physical health measured using SF36 derived index variables ranging between 0-1005; and a control for health shock which
identifies whether the respondent suffered a serious personal injury or illness in the
preceding 12 months.
I also include a control for interview mode, given some interviews are conducted on
the telephone. Moreover, given that my outcome variable has a reference period that varies
in length across persons depending on the date the interview is undertaken, I include
dummies for month of the interview to control for differences that might arise due to this.
A potentially important issue for my analysis is the issue of non-random attrition
associated with non-return of the SCQ. The non-response/return of the SCQ is likely
associated with the location, availability of time and, more importantly, the type of
5
The SF-36 is a multi-purpose, short-form health survey with only 36 questions. For complete information about
the SF-36 see Ware et al (1993, 2000). The variables used are ghgh to control for physical health and ghmh to
control for mental health.
16
employment arrangement (see Appendix B). I include an indicator variable proposed by
(Verbeek & Nijman, 1992), a regressor identifying whether the sample member is observed
at the next survey wave, that is intended to control for this selectivity bias.
4.4 Mechanisms
The conceptual framework suggests that non-permanent employment might lead to
hardships not only through low level of earnings and hence, low level of houshold income but
also through unstable or volatile earnings that one might experience when working in these
jobs as workers in these jobs are more susceptible to shocks and likely to work for insufficient
hours. I attempt explore these three potentially important mechanisms- low income,
susceptibility to employment shocks, and hours worked, in my empirical analyses by
examining whether the significance and magnitude of the effects change when I control for
these factors. The measures for the potential mechanisms are constructed as follows: a
continues measure of the log of respondent’s level of gross tatal household income from all
sources adjusted by OECD equivalence scale6; two binary indicators for employment shocksif a respondent was fired or made redundant in the preceding 12 months and if respondent
changed jobs in the preceding 12 months; and a continuous measure for hours worked per
week in main job.
Some of my controls variables such as controls for union membership, private sector,
industry, occupation, and hours worked are applicable for only employed people. To
accommodate people who were not working the values for all these controls were coded as
zero. Hence, in the multivariate analyses the interpretation for unemployment and out of
labour force coefficients is different in the presence of union membership, industry,
6
OECD equivalence scale assigns a value of 1 to the household head, of 0.5 to each additional adult member
and of 0.3 to each child.
17
occupation, private sector and hours worked controls. The reference category for fixed-term,
casual and self-employed persons is permanent employees in all the models. However, the
reference category for unemployed and out of labour force people, in models where controls
for union membership, industry, occupation, private sector and hours worked are added, is
permanent employees who are managers and working for a government organisation in
‘other services’7. This implies that the coefficient estimates for unemployed and out of labour
force categories are likely to increase once these controls are added because of the change
in the reference category. Also, in the case of unemployment and out of labour force
categories there is a possibility of multicollinearity when controls for job characteristics are
added because of the way these variables are incorporated (constructed) for the two
categories8.
4.5 Analysis sample
The original sample consists of 182,799 observations from 27,894 unique individuals.
Several restrictions are placed on the sample for this study. The sample is restricted to the
working-age population of persons aged 18 to 55 excluding full-time students aged less than
25 years. This results in a sample size of 112,972 observations from 16,620 unique persons.
The age restriction of 18 to 55 is based on the fact that in Australia the average retirement
age for the recent Australian retirees, that is, people who retired less than five years ago from
2013, was 61.5 years (63.3 years for men and 59.9 years for women) (ABS, 2013). I excluded
full-time students given employment while pursuing full-time studies is not likely to have a
large effect on one’s future employment and income.
7
Managers is base category for occupation controls; public sector employee is reference for private sector
employee; and ‘other services’ is base category for industry controls.
8 There is never an overlap with any of the union membership, industry, occupation, private sector and hours
worked measures, implying that unemployment and OOLF indicators never take value 1 when any of the union
membership, industry, occupation and part-time employment takes value 1.
18
Second, the sample is further restricted to persons in single-family households and
single-person households. In other words, the families that only include parents and children
and persons living alone are included in the sample and families that include other related or
non-related persons have been excluded. The HILDA survey identifies around 93% of people
to be living in such families. This restriction is imposed to avoid issues that arise due to the
ambiguity regarding how resources and expenses are shared within joint or grouped families.
This further reduces the sample size to 105,540 observations from 14,433 unique persons.
Third, the sample used for the analyses is restricted to cases with non-missing data on
all variables9. 13.54% values in the analysis sample (105,540 observations) had missing
observation for financial index (excluding cases missing from wave 10)10. The employment
contract variable had only 0.29% of the observations missing for the analysis sample. As some
of the explanatory variables control for events or shocks in the previous year, I further lose
first wave of the data. Hence, the analyses are based on 11 waves of data from release 13.
The final sample size is 73,149 observations from 15,687 unique persons, after restricting the
sample to cases with non-missing data11.
The HILDA survey dataset provides sampling weights that can be used to make some
types of analyses more representative of the population or initial survey universe. However,
the use of weights is not appropriate for this study because of two reasons. First, I have
imposed several restrictions on the sample as discussed earlier. Second, I use fixed-effects
methods on an unbalanced sample where analyses require at least two observations on the
9
Except for mental and physical health SF-36 index variables. SF-36 mental and physical variables had the high
percentage of missing values for my outcome variable. For these variables I filled up any missing value for the
respondent with his/her average value for the observed years to minimise the missing data.
10 Financial hardship questions were asked in self-completion questionnaire (SCQ), hence the missing values are
more likely to be a result of non-return of SCQ rather than item non-response.
11
Non-missing data here refers to having complete data on all variables for each wave that the individual
participated in the survey. If individuals had at least one complete wave of data they are included in the study.
19
same individual and hence those people who are only interviewed once are dropped. Solon,
Haider, & Wooldridge (2015) showed that weighting actually reduces the efficiency of the
estimates when estimating fixed-effects model on unbalanced panel. Therefore, I do not use
weights in my empirical analyses.
5 Empirical analyses
5.1 Descriptive analysis
I start by describing how the incidence and characteristics of financial hardship vary by
employment status and gender. The first row of panel A and panel B in Table 1 lists the means
of the financial hardship index for men and women, respectively. It shows that men and
women in all alternative forms of employment arrangements report a higher incidence of
hardships, on average, than men and women in permanent employment. The second
component of panels A and B (part ii) presents the distribution of the financial hardship index
by employment type for each gender. Compared to permanent employees, the distribution
of financial hardship is much less skewed towards zero for casual employees. Among
permanent employees, just around 20% of men and women report facing one or more
hardships. By contrast, among casual employees, 40 % of men and 33% of women report
facing one or more types of hardships. The incidence of financial hardships does not,
however, appear to be very different between fixed-term employees and permanent
employees.
The third component of panels A and B (part iii) presents the rates of different types
of financial hardships experienced by men and women, respectively. Not being able to pay
utility bills on time, asking friends and family for financial help, and not being able to pay
mortgage or rent on time are the three most commonly experienced hardships by both men
20
and women. But the incidence of each hardship is much higher for casuals as well as for not
working people compared to permanents. The differences between fixed-term and
permanent employees are again much smaller. Gender differences also appear in types of
hardships experienced by respondents. Men casual employees do much worse than women
casual employees and so do men that are out of labour force. This to some extent reflects the
secondary earner role of the women in family.
Table 2 summarises how various demographic, economic, health, and work-related
characteristics vary for people in different employment arrangements.
Fixed-term employees. Fixed-term employees are, compared to permanent
employees, more likely to be women, young, and single. In terms of employment and job
characteristics, fixed-term employees are less likely to have union membership, and more
likely to have held multiple jobs, to have experienced job loss, and to have changed jobs in
the past 12 months. However, fixed-term employees are also more likely than permanent
employees to have university qualifications, to be working as professionals, and to be
employed in the public sector, particularly in the education and training industries.
Casual employees. Casual employees are, compared to permanent employees, more
likely to be women, young, less educated, living in regional or remote areas, and of Aboriginal
or Torres Strait Islander origin. They are also more likely to be lone parents with caring
responsibilities or singles. In terms of employment and job related characteristics, they are
less likely to be members of a trade union or employee association and twice as likely as
permanent employees to hold multiple jobs in the past year. Moreover, casual employees are
very dominant in retail trade and food, and accommodation industries, they are less likely to
be professionals and government employees, and more likely to be a community and personal
service workers, sales workers or labourers. As expected almost none of them are provided
21
with paid leaves. Finally, compared with permanent employees, casual employees have lower
levels of household income, they are significantly more likely to have experienced job loss, to
have changed job in the past 12 months, and to be working significantly fewer hours.
These results suggest clear differences between permanent and casual employees.
The differences between fixed-term and permanent employees, however, are not as striking.
In fact, those working on fixed-term contract are more educated and more likely to
professionals compared to permanent employees, whereas casual appear to be less educated
and working in low skilled jobs compared to permanent employees.
Self-employed and unemployed persons and those out of the labour force (OOLF).
Table 2 suggests that self-employed workers are, compared to those in permanent jobs, more
likely to be men, older, less educated, and belong to a couple family with caring
responsibilities. Unemployed people, compared to those in permanent jobs, are more likely
to be women, young, less educated and single. They are also more likely to have poor mental
and physical health. People who are out of the labour force are more likely to women, old,
less educated and with caring responsibilities. They are also more likely to have poor mental
health and poor physical health.
Table 3 shows how incidence of financial hardships varies by employment type for
men and women in different family types. The table provides clear evidence that a person’s
family type and gender affect the relationship between employment arrangements and the
number of hardships experienced. For example, lone parents with dependent children and
those living alone face relatively more hardships compared to those belonging to couple
families where incomes can be pooled. Similarly, the table shows that in couple families, men
in casual jobs and those not working do much worse in terms of the experience of financial
22
hardship compared to their permanent counterparts than women. This is because men are
more likely to be primary earners of the family.
5.2 Multivariate empirical analysis
5.2.1 Econometric specification
I conduct multivariate analyses to further investigate how employment type is associated
with financial hardship once various confounding factors are controlled for. To allow for the
rank order in the hardship measure but without making any assumptions about the relative
degree of difference between each of the individual hardships, I treat my financial hardship
index variable as an ordinal measure and estimate fixed effects ordered logit regression
models of the following form for men and women in different family types12:
𝑓𝑖𝑡∗ = 𝛽𝐸𝑖𝑡′ + 𝛿𝑋𝑖𝑡′ + ƞ𝑖 + 𝜇𝑡 + 𝜀𝑖𝑡,
𝑖 = 1, … , 𝑁
𝑡 = 1, … , 𝑇
𝑓𝑖𝑡∗ is a latent index describing financial hardships of person 𝑖 at time period 𝑡. 𝐸𝑖𝑡 is a vector
of dummy variables identifying the type of employment arrangement held by person 𝑖; 𝑋𝑖𝑡 is
a vector of person’s other time varying observable characteristics; 𝛽 and 𝛿 are the vectors of
coefficients; ƞ𝑖 is a time-invariant person specific unobserved term; 𝜇𝑡 is a time (or wave)
fixed-effect; 𝜀𝑖𝑡 is an idiosyncratic error term.
The latent variable is tied to the observed ordered variable 𝑓𝑖𝑡 by the observation
rule:
𝑓𝑖𝑡 = 𝑘 𝑖𝑓 𝜏𝑖𝑘 < 𝑓𝑖𝑡∗ ≤ 𝜏𝑖𝑘+1 ,
12
𝑘 = 0, … ,7
I also estimated pooled ordered logit and random effects ordered logit models. Likelihood ratio test for the
presence of panel-variance rejected pooled logit estimator in favour of random effects logit estimators.
However, I could not apply Hausman specification test to test for fixed-effects in ordered logit models because
standard errors are clustered by construction in my fixed-effects ordered logit models. Therefore, I used linear
models that treat 𝑓𝑖𝑡 as a cardinal measure. Lagrange- multiplier (LM) test rejected OLS in the favour of random
effects with a p-value of 0.000 for both men and women in each of the subsample. The Hausman-Wu test
rejected random effects in favour of fixed-effects models with a p-value of 0.000 for both men and women in
each of the subsample, indicating a significant correlation between time-invariant omitting variables and
observed variables.
23
The fixed effects ordered logit model assumes that the 𝜀𝑖𝑡 terms are independently
and identically distributed with a logistic distribution. The estimation of non-linear fixedeffects models such as the ordered logit model is complicated by the incidental parameters
problem (Lancaster, 2000). The usual approach adopted to obtain consistent estimates of β
is to dichotomize the ordered outcome variable at a cut-off point and then apply conditional
maximum likelihood (CML) estimation (Chamberlain, 1980). However, this approach ignores
individuals who do not cross the dichotomous cut-off over time. I use an alternative estimator
proposed by Baetschmann, Staub, & Winkelmann (2015) coined as the ‘Blow-Up and Cluster’
estimator (BUC) that is consistent and asymptotically efficient. This estimator uses all the
available variation in the ordered outcome variable. The authors show that, compared to the
alternative estimators such as Generalized methods of moments (GMM), Empirical Likelihood
(EL), Minimum Distance estimator (MD), the BUC estimator is more efficient, clearly
outperforms existing estimators if the ordered dependent variable displays extremely low
responses for some categories (as in my study) , and produces robust estimates even in small
samples. Hence, for this study I use the BUC estimator since the distribution of my hardship
index measure is highly skewed and due to segregation by gender and family type there is a
possibility of relatively small numbers of observations within each response category.
In practice, the BUC estimator is implemented by replacing every person-year
observation in the sample by K-1 copies of itself (hence, ‘blowing up’ the data) and then
dichotomizing each of the K-1 copies of the person at each threshold /cut-off point. The
resulting binary logit models are estimated using CML and standard errors are clustered at
the individual level to account for the interdependence of person-year copies of observation.
The resulting coefficients are combined by jointly estimating all dichotomizations subject to
the restriction 𝛽 𝑘 = 𝛽 𝑘 = 2, … , 𝐾. Given its advantages, this estimator has been used in
24
many recent studies involving ordinal subjective outcomes such as job satisfaction, and selfassessed health (for example, see Bell, Otterbach, & Sousa-Poza (2012) and Buddelmeyer et
al. (2015)). The coefficients 𝛽 in my ordered logit models are interpreted as the approximate
effect of change in employment status on the odds ratio Pr(𝑓𝑖𝑡 > 𝑘)/ 𝑃𝑟(𝑓𝑖𝑡 ≤ 𝑘) , for all 𝑘.
My model exploits the within-individual variation in both the experience of financial
hardship and employment status. This mitigates the time-invariant individual specific
endogeneity arising from unobserved individual time invariant heterogeneity that might be
correlated with financial hardships and employment status. I also incorporate a rich set of
time-varying control variables that account for potential confounders in the relationship
between financial hardship and type of employment. I also control for common business
cycle shocks that could affect both employment and hardships by including year/wave fixed
effects. Further, following the method proposed by Verbeek & Nijman (1992), I also include
a variable that is intended to control for endogenous sample attrition, a regressor
identifying whether the sample member is observed at the next survey wave. Finally, given
my outcome variable has a reference period that varies in length across individuals
depending on the date on which the interview is undertaken, I also include dummies for
each month to control for the bias that might arise due to this difference. To draw casual
inferences, my estimation approach relies on key identifying assumption that all the
selection on unobservable characterises is due to time-invariant factors.
5.2.2 Estimation results
I estimate fixed effects ordered logit models separately for men and women within
each family type using the BUC estimator. Results for men and women are given in Table 4
and Table 5, respectively. I consider three specifications in increasing controls. Model A lists
estimates from my baseline specification that only includes individual and time fixed-effects;
25
The next column, Model B adds controls for arguably exogenous time-varying observed
factors including demographic controls (age, education and area of residence); an indicators
for the mode of interview; and an indicator of whether the respondent is observed in the
sample at t+1. Model C adds controls that may be potentially endogenous. These include
various health controls discussed earlier, union status, the number of jobs held in the
preceding financial year, industry, occupation, and whether the person is a private sector
employee.
Results for men
Table 4 lists regression coefficient estimates and clustered robust standard errors
from fixed effects ordered logit models for men in couple families (partnered men) with
dependent children (panel (i)), in couple families without dependent children (panel (ii)) and
for men living alone (panel(iii)). For brevity, the table only lists the coefficients and standard
errors for the measures of employment arrangements (complete results are provided in
appendix D).
In ordered logit model we can interpret the coefficients in terms of log odds by
calculating [exp(β)-1]*100. For example, for partnered men with dependent children, I find
that compared to being a permanent employee, being a casual employee increases the odds
of experiencing more financial hardships by about 72% (or exactly by [exp(0.544)-1]*100))
after I control for exogenous and potentially endogenous variables in the final specification
(model C). In contrast, men working on fixed-term contract appear to be not significantly
different from their permanent counterparts in terms of their hardship experience.
For partnered men without dependent children, again casual employment but not
fixed-term employment is significantly associated with more financial hardships when
compared to permanent employment in all the specifications. For men living alone, I find that
26
compared to permanent employment, casual employment is significantly associated with
more financial hardship in model A and B, but in the final specification (model C), significance
declines and the estimate is significant only at 10%. Also for single men, I find no difference
in the experience of financial between fixed-term and permanent employees.
Additionally, I find that being self-employed is significantly associated with more
financial hardship for partnered men without caring responsibilities but not for partnered
men with caring responsibilities when compared to their respective permanent counterparts.
Also, being unemployed and out of labour force significantly increases the odds of
experiencing more hardships for men in all sub-samples except for those living alone where
estimate become insignificant once controls for health and job characteristics are added13.
Results for women
Table 5 reports regression coefficient estimates and clustered robust standard errors
from fixed effects ordered logit models for women in couple families (partnered women) with
dependent children (panel(i)), in couple families without dependent children (panel(ii)), in
lone parent families with dependent children (panel(iii)) and for women living alone
(panel(iv)). Table 5 only lists the coefficients and standard errors for the measures of
employment arrangements (complete results are provided in appendix E).
The results reveal that for partnered women with and without caring responsibilities
as well as for lone mothers with caring responsibilities, being a casual employee but not fixedterm employee, significantly increase the likelihood of facing more financial hardships
comapred to their permanent counterparts in all the models/specifications. By contrast,
single women who are casual employees are not worse of in terms of their experience of
13 However, standard errors increase significantly possibility due to the multicollinearity problem that I discussed
earlier these estimates
27
financial hardships when compared to single women who are permanent employee. Similarly,
the fixed-term employment is not associated with financial hardship for single women. In fact
estimates for fixed-term employment for single women are negative and large, although
insignificant.
Additionally, I find that self emplyment is significantly associated with more hardships
for partnered women without caring responsibilitites but not for partnered women with
caring responsibilities. This findings is similar to the finding for men. This suggests that
compared permanent employees, being self-employed might not be disadvantageous for
partnered men and women with caring responsibilities in terms of the experience of financial
hardship as it might offer them greater flexibility to manage and share child care and other
household responsibilities resulting in economic gains by saving on the cost of formal
childcare and paying for other household work. For lone mothers and single women the
estimates for self-employment are positive and large in magnitute but imprecise due to small
sample size for these two groups.
Also in all family types, unemployed women and those out of the labour force face
significantly more financial hardships compared to their permanent counterparts in models A
and B. In model C that includes control for various job chracteristics, except for partnered
women without caring resposibilities, the estimates for all other sub-samples become
insignificant due to increase in the standard errors. This could be because of the issue of
multicollinearity as discussed earlier leading to imprecise estimates.
Effects of covariates
I find that for both men and women in all subsamples, poor mental health is associated
with significantly more financial hardships. Poor physical functioning is also associated with
more hardships for men and women in all subsamples but estimates are not significant for
28
partnered men without caring responsibilities and single men and single women. Also,
findings suggest that living in remote area is associated with less financial hardships for
women in all family types but for men this is true only for partnered men without caring
responsibilities. Finally, being a private sector employee is associated with significantly more
hardships for men and women in couple families without caring responsibilities.
6. Mechanisms and sensitivity analyses
6.1 Mechanisms
Motivated by my conceptual formwork, I attempt to explore three potentially
important mechanisms through which insecure jobs can lead to financial hardships: low
income, susceptibility to employment shocks, and hours worked, by examining whether the
significance and magnitude of the effects change when I control for these factors in my model.
If most of the negative effect of non-permanent employment, particularly casual employment
arises through these mechanisms, we would expect the estimates to decline noticeably in
magnitude and significance once I control for these mechanisms.
The results for the mechanisms analysis are presented in Table 6 for men and Table 7
for women. The first column of each panel of both the tables simply shows my main results
from model C, in second column I add control for log of the person’s gross total household
income adjusted by OECD equivalence scale, in third column I add control for two measures
of employment shocks-job loss and job change in preceding 12 months, and finally in fourth
column I add control for hours worked.
For men, consistent with my expectation, I find that controlling for low income,
employment shocks experienced, and hours worked leads to considerable decline in the
magnitude as well as significance of the estimates for casual employment in all subsamples.
The results for fixed-term and self-employment remain qualitatively unchanged, but for fixed29
term employment the positive estimates do decline noticeably in magnitude for partnered
men with caring responsibilities and single men. Coming to the direct effect of the three
mechanisms on financial hardship, I find that higher household income reduces the odds of
facing more financial hardships for men in all family types but the estimate is not significant
for single men. The experience of job loss and/or job change is positively associated with more
financial hardships for men in all family types but estimates are not significant for partnered
men without caring responsibilities. Finally, I find that hours worked has a significant and
negative effect on financial hardships for men in all family types.
For women, I find that once I control for low income, employment shocks experienced,
and hours worked, the magnitude and significance of casual employment estimates declines
for women without caring responsibilities but not for women with caring responsibilities. For
women with caring responsibilities (partnered as well as lone mothers) there still remains
unexplained/residual significant positive effect of casual employment on financial hardship.
These findings suggest that other factors such as having access to paid carer/sick leave might
also be equally important for women with caring responsibilities to manage the household
and child care duties and avoid expenses associated with formal child care. Moreover,
evidence from previous studies suggests that scheduling flexibility associated with casual jobs
more frequently is created to increase employment flexibility for employers rather than
workers’ time-use preferences (OECD, 2002; Pocock, Prosser, & Bridge, 2004).
Coming to the direct effect of the three mechanisms on financial hardship for women,
I find that higher household income reduces the odds of facing more financial hardships for
women in all family types but the estimate is not significant for partnered women without
caring responsibilities. Hours worked has a significant and negative effect on financial
hardships for women in all family types. The direct effect of experience of job loss and/or job
30
change is inconsistent for women. Job loss is significantly associated with more financial
hardship only for single women. Job change is significantly associated with more financial
hardship for partnered women but not for single women and lone mothers.
It is important to highlight the caveats associated with these results. The employment
shock measures capture events that can occur any time in preceding 12 months. This means
that the shock events might not necessarily capture the experience associated with the
contract of employment that the person report at the time of the interview. For example, if a
person was a permanent employee then lost his/her job and became a casual, than the job
loss cannot be attributed to the consequences of being a casual employee. But rather, casual
employment is a consequence of job loss in this case. Nonetheless, given that findings suggest
that job loss and job change is associated with more hardships, particularly for men and there
is enough evidence in the literature that these experiences are much more common among
casual employees, it is important to provide some protection against these shocks to improve
financial well-being of workers working in insecure jobs.
6.2 Sensitivity Analysis
A number of analyses were conducted to investigate the sensitivity of the main results
and mechanisms analysis results to alternative functional forms, additional controls,
alternative measure of the hardship index, and different specifications. First, I considered
whether my results are robust if I treat the financial hardship index (𝑓𝑖𝑡 ) as a cardinal measure
by running two-way linear fixed-effects regression models of the count of hardships. The
results remain substantively same as the fixed-effects ordered logit models for men and
women in all family types.
Second, I incorporate lags for the physical and mental health controls. Mental and
physical health might be endogenous if financial hardships affect health outcomes. This could
31
be more problematic in my analysis given that mental and physical health are measured at
the time (or previous four weeks preceding the interview) of the interview and hardships are
measured from the start of the year. Hence, I additionally control for mental and physical
health at the previous wave in the models. A comparison of the estimates suggests my
findings are qualitatively similar to those reported in the results section.
Third, I excluded the inability to heat home variable from my financial hardship index
as the responces to it could be influeced by geograpical location/climate of the place where
respondents live. The findings are substantively similar to those reported in the results section
for both men and women. Similarly, I restricted the index to include only true/actual
hardships and excluded variables like unable to pay utility bills on time, unable to pay
rent/mortgage on time and asking financial help from friends and family, which might are
termed as cash flow problems by some studies. However, it should be noted that the
incidence of the actual hardships is very low in the sample and particularly among women
(see Table 1).
I find that for men the results remain qualitatively similar but estimates for all
employment types are noticibly larger in magnitute. For women, I find that the results are
qualitatively similar for lone mothers with caring responsibilities and single women, and again
magnitutes of estimates for fixed-term and casual estimates are noticibly large. For partnered
women with caring responsibilities, the results remain qulitatively similar but the estimates
are insignificant due to large standard errors. Finally, for partnered women without caring
responsibilities the results are qualitatively similar for models A and B. However, the estimate
for casual employment (while still positive) becomes insignificant in model C and becomes
negative ( though alomost zero in magnitute) when controls for all mechanisms are added.
The estimate for fixed-term employment becomes large and significant in model C as well as
32
remain so even when controls for potential mechanisms are added. However, the incidence
of actual hardship is extreamly low for women in couple families without caring
responsibilities (maximum of 400 observations on affirmative response to any given
hardship), hence, within estimates are very likey to be imprecise due to limited variation.
Fourth, given that my outcome, experience of financial hardships, is measured over
the entire year until the date of interview but the majority of my explanatory variables are
measured at the time of the interview, I also estimated models where I replaced these
explanatory variables with their lags at t-1 (values for the previous wave). The results for this
sensitivity analyses were quite interesting. The majority of estimates for the employment
categories had expected sign but were insignificant for both men and women. One potential
reason for this could be that employment categories such as fixed-term contract, casual and
unemployment are highly fluid. I find that just half of the men and women who were casuals
at t-1 were casuals at t and around 40% of men and women who were fixed-term employees
at t-1 were fixed-term employees at t (see appendix F). Moreover, fixed-effects methods
already in some sense take into account the transitions in employment from previous periods
to the current period since they are based on within-individual variation. Hence, estimating
fixed-effects models with lagged explanatory variables would consider transitions between t2 and t-1 periods to estimate its effects on financial hardships faced during year t. The
associations between current financial hardship and lagged employment categories, in this
case, are likely to be very weak. All of my sensitivity analyses are available upon request.
7. Discussion and conclusion
This paper uses rich longitudinal data from the HILDA survey to examine the relationship
between non-standard employment and financial hardship faced by men and women. The
33
descriptive analyses reveal that the incidence of financial hardship is relatively high among
those in non-permanent employment arrangements (and especially casual employees)
compared to permanent employees. Also, as expected, I find that unemployed persons and
those out of labour force report significantly higher financial hardships compared to
permanent employees. The descriptive analyses further reveal that there are significant
differences among the people working in different employment arrangements in terms of
their personal, demographic, economic and work related characteristics.
Results from the main multivariate models indicate that for both men and women,
compared to permanent employment, casual employment but not fixed-term employment
significantly increases the likelihood of facing more financial hardships net of the effects of
individual and year fixed effects, standard demographic characteristics, mode and month of
interview, attrition bias, mental and physical health status, union status, the number of jobs
held in the preceding financial year, industry, occupation, and private sector employee status,
in all subsamples even though results are relatively weak for single men and women. It is
important to note that my econometric model makes an assumption that the effect of job
change (β) is symmetric on different job types (that is, the magnitude of effect of moving from
permanent to casual is same as moving from casual to permanent), which might not hold.
Results from mechanisms analysis suggest that low level of income, high volatility of
income that arises from experiencing employment shocks and insufficient hours of work
experienced by casual employees could explain lot of the association between casual
employment and financial hardships for all men but only for women without caring
responsibilities. For Women with caring responsibilities, these factors do explains some
association but still residual relation between casual employment and financial hardship
remains.
34
In conclusion, results from the main econometric analysis and mechanisms analysis
indicate that the relationship between non-permanent employment and financial hardship is
heterogeneous. The finding that compared to permanent employment, casual employment
increases the odds of experiencing more financial hardships but not fixed-term contract
employment, suggest that my findings are not driven by lack of permanency of a job. Factors
like stable pay and certainty regarding the length of employment are important to overcome
financial stress. Unlike casual employees, level of earnings of fixed-term contract employees
is high and not variable during the contract period, plus, while these jobs are for a fixed
period, but there is no uncertainty regarding the contract termination date, and hence these
people can plan accordingly about getting their next job. Moreover, the majority of fixedterm employees have access to paid time off. Additionally, it is very clear that having a job is
better than not having one as non-employment led to significantly more financial hardships
for both men and women in all subsamples.
From a policy perspective, the findings of this study are important. With around 20%
of the workforce engaged in casual employment in Australia, the findings from this study
advance the current understanding of the finer mechanisms underlying these people’s
employment experience and reveal the possible implications that the non-regulation of such
forms of employment can have on the financial wellbeing of workers.
There are some important caveats to findings in this paper. First, while my estimation
methodology and rich nature of HILDA data allows me to control for all unobserved timeinvariant individual factors, common shocks and a rich set of time varying observed
characteristics, there might still be other time-varying factors that I cannot control for, leaving
the possibility of endogeneity arising from omitted variable bias. Further, one cannot
35
completely rule out the possibility that experiencing severe hardships (such as going without
food for long, housing insecurity due to not paying rent/mortgage for long period of time)
could affect employment prospects for at least some people, although my initial examination
did not support any such association in the data.
36
References
ABS. (2010). Measures of Australia's Progress 2006, cat. no. 1370.0. Australian Bureau of Statistics,
Canberra.
ABS. (2013). Retirement and Retirement intentions, cat. no. 6238.0. Australian Bureau of Statistics,
Canberra
Baetschmann, G., Staub, K. E., & Winkelmann, R. (2015). Consistent estimation of the fixed effects
ordered logit model. Journal of the Royal Statistical Society: Series A (Statistics in Society),
178(3), 685-703.
Bauman, K. J. (2002). Welfare, work and material hardship in single parent and other households.
Journal of Poverty, 6(1), 21-40.
Bell, D., Otterbach, S., & Sousa-Poza, A. (2012). Work hours constraints and health. Annals of
Economics and Statistics/ANNALES D'ÉCONOMIE ET DE STATISTIQUE, 35-54.
Boden, R. J. (1999). Flexible working hours, family responsibilities, and female self‐employment.
American Journal of Economics and Sociology, 58(1), 71-83.
Booth, A. L., Francesconi, M., & Frank, J. (2002). Temporary jobs: stepping stones or dead ends? The
economic journal, 112(480), F189-F213.
Bray, J. R. (2001). Hardship in Australia: an analysis of financial stress indicators in the 1998-99
Australian Bureau of Statistics Household Expenditure Survey. FaHCSIA Occasional Paper(4),
Canberra.
Breunig, R., & Cobb-Clark, D. (2006). Understanding the factors associated with financial stress in
Australian households. Australian Social Policy, 2005, 13-64.
Bubonya, M., Cobb-Clark, D. A., & Wooden, M. (2014). A Family Affair: Job Loss and the Mental
Health of Spouses and Adolescents.
Buchler, S., Haynes, M., & Baxter, J. (2009). Casual employment in Australia The influence of
employment contract on financial well-being. Journal of Sociology, 45(3), 271-289.
Buddelmeyer, H., Leung, F., McVicar, D., & Wooden, M. (2013). Training and its impact on the casual
employment experience: NCVER.
Buddelmeyer, H., McVicar, D., & Wooden, M. (2015). Non‐Standard “Contingent” Employment and
Job Satisfaction: A Panel Data Analysis. Industrial Relations: A Journal of Economy and
Society, 54(2), 256-275.
Buddelmeyer, H., & Wooden, M. (2011). Transitions out of casual employment: the Australian
experience. Industrial Relations: A Journal of Economy and Society, 50(1), 109-130.
Burgess, J., & Campbell, I. (1998). The nature and dimensions of precarious employment in Australia.
Labour & Industry: a journal of the social and economic relations of work, 8(3), 5-21.
Butterworth, P., & Crosier, T. (2005). Deriving a measure of financial hardship from the HILDA
survey.
Campbell, I. (1996). Casual employment, labour regulation and Australian trade unions. Journal of
Industrial Relations, 38(4), 571-599.
Campbell, I., Whitehouse, G., & Baxter, J. (2009). Casual employment, part-time employment and
the resilience of the male-breadwinner model. Gender and the contours of precarious
employment.
Chalmers, J., & Kalb, G. (2001). Moving from unemployment to permanent employment: could a
casual job accelerate the transition? Australian Economic Review, 34(4), 415-436.
Chamberlain, G. (1980). Analysis of Covariance with Qualitative Data. Review of Economic Studies,
47(1), 225.
Cobb-Clark, D. A., & Ribar, D. C. (2012). Financial stress, family relationships, and Australian youths’
transitions from home and school. Review of Economics of the Household, 10(4), 469-490.
de Graaf-Zijl, M., Van den Berg, G. J., & Heyma, A. (2011). Stepping stones for the unemployed: the
effect of temporary jobs on the duration until (regular) work. Journal of population
Economics, 24(1), 107-139.
37
Eamon, M. K., & Wu, C.-F. (2011). Effects of unemployment and underemployment on material
hardship in single-mother families. Children and Youth Services Review, 33(2), 233-241.
Forde, C., & Slater, G. (2005). Agency working in Britain: Character, consequences and regulation.
British Journal of Industrial Relations, 43(2), 249-271.
Fuller, S., & Vosko, L. F. (2008). Temporary employment and social inequality in Canada: Exploring
intersections of gender, race and immigration status. Social indicators research, 88(1), 31-50.
Gash, V. (2008). Bridge or trap? Temporary workers’ transitions to unemployment and to the
standard employment contract. European Sociological Review, 24(5), 651-668.
Giesecke, J. (2009). Socio-economic risks of atypical employment relationships: evidence from the
German labour market. European Sociological Review, 25(6), 629-646.
Giesecke, J., & Groß, M. (2003). Temporary employment: chance or risk? European Sociological
Review, 19(2), 161-177.
Graaf-Zijl, D. (2005). The anatomy of job satisfaction and the role of contingent employment
contracts.
Green, C. P., & Heywood, J. S. (2011). Flexible Contracts And Subjective Well‐Being. Economic
Inquiry, 49(3), 716-729.
Gundersen, C., & Gruber, J. (2001). The dynamic determinants of food insufficiency. Paper presented
at the Second food security measurement and research conference.
Houseman, S., & Osawa, M. (2003). The growth of nonstandard employment in Japan and the United
States. S. Houseman and M. Osawa, Nonstandard Work in Developed Economies: Causes and
Consequences. Kalamazoo, Mich.: WE Upjohn Institute for Employment Research, 175-214.
ILO. (2008a). Globalization, Flexibilization and Working Conditions in Asia and the Pacific. ILO
publications. International Labour Office, Geneva.
ILO. (2008b). World of work report 2008: income inequalities in the age of financial globalization. ILO
publications, International Labour Organization, International Institute for Labour StudiesGeneva.
ILO. (2015). Non-standard forms of employment. Report for discussion at the Meeting of Experts on
Non-Standard Forms of Employment (Geneva, 16–19 February 2015). ILO publications,
Conditions of Work and Equality Department, International Labour Office-Geneva.
Jappelli, T., & Pistaferri, L. (2010). The consumption response to income changes. Retrieved from
Kahn, L. M. (2013). The structure of the permanent job wage premium: evidence from Europe.
Kalleberg, A. L. (2000). Nonstandard employment relations: Part-time, temporary and contract work.
Annual review of sociology, 341-365.
Kalleberg, A. L., Reskin, B. F., & Hudson, K. (2000). Bad jobs in America: Standard and nonstandard
employment relations and job quality in the United States. American Sociological Review,
256-278.
Kim, T. J., & von dem Knesebeck, O. (2015). Is an insecure job better for health than having no job at
all? A systematic review of studies investigating the health-related risks of both job
insecurity and unemployment. BMC public health, 15(1), 985.
Lancaster, T. (2000). The incidental parameter problem since 1948. Journal of econometrics, 95(2),
391-413.
Lane, J., Mikelson, Kelly, Patrick, Wissoker, Douglas (2001). Low-income and low-skilled workers'
involvement in nonstandard employment: Final report. The Urban Institute, Washington,
DC(October).
Llena-Nozal, A. (2009). The effect of work status and working conditions on mental health in four
OECD countries. National Institute Economic Review, 209(1), 72-87.
Lovell, V., & Oh, G.-T. (2006). Women's job loss and material hardship. Journal of Women, Politics &
Policy, 27(3-4), 169-183.
McGovern, P., Smeaton, D., & Hill, S. (2004). Bad jobs in Britain nonstandard employment and job
quality. Work and occupations, 31(2), 225-249.
38
McNamara, M. (2006). The hidden health and safety costs of casual employment. Industrial
Relations Research Centre paper, University of New South Wales, Sydney.
McOrmond, T. (2004). Changes in working trends over the past decade. Labour Market Trends,
112(1), 25-36.
Meghir, C., & Pistaferri, L. (2011). Earnings, consumption and life cycle choices. Handbook of Labor
Economics, 4, 773-854.
Mooi‐Reci, I., & Dekker, R. (2015). Fixed‐Term Contracts: Short‐Term Blessings or Long‐Term Scars?
Empirical Findings from the Netherlands 1980–2000. British Journal of Industrial Relations,
53(1), 112-135.
OECD. (1996). OECD Employment Outlook 1996. Retrieved from OECD Publishing, Paris:
http://dx.doi.org/10.1787/empl_outlook-1996-en
OECD. (2002). Taking the Measure of Temporary Employment, OECD Employment Outlook 2002.
Retrieved from OECD Publishing, Paris: http://dx.doi.org/10.1787/empl_outlook-2002-5-en:
OECD. (2014). Non-regular employment, job security and the labour market divide, OECD
Employment Outlook 2014. Retrieved from OECD Publishing, Paris:
http://dx.doi.org/10.1787/empl_outlook-2014-7-en
OECD. (2015a). In It Together: Why Less Inequality Benefits All. Retrieved from OECD Publishing,
Paris: http://dx.doi.org/10.1787/9789264235120-en
OECD. (2015b). Labour Market Statistics: Employment by permanency of the job: incidence, OECD
Employment and Labour Market Statistics (database). Retrieved from OECD Publishing,
Paris: http://dx.doi.org/10.1787/data-00297-en
Pocock, B. A., Prosser, R. W., & Bridge, K. J. (2004). Only a Casual... How Casual Work affects
Employees Households and Communities in Australia.
Presser, H. B. (1989). Can we make time for children? The economy, work schedules, and child care.
Demography, 26(4), 523-543.
Presser, H. B. (1995). Job, family, and gender: Determinants of nonstandard work schedules among
employed Americans in 1991. Demography, 32(4), 577-598.
Ribar, D. C. (2015). Is Leaving Home a Hardship? Southern Economic Journal, 81(3), 598-618.
Solon, G., Haider, S. J., & Wooldridge, J. M. (2015). What are we weighting for? Journal of Human
Resources, 50(2), 301-316.
VandenHeuvel, A., & Wooden, M. (2000). Diversity in employment arrangements: Office of Economic
and Statistical Research, Queensland Treasury.
Verbeek, M., & Nijman, T. (1992). Testing for selectivity bias in panel data models. International
Economic Review, 681-703.
Watson, I. (2005). Contented Workers in Inferior Jobs? Re‐Assessing Casual Employment in Australia.
The Journal of Industrial Relations, 47(4), 371-392.
Watson, I. (2013). Bridges or traps? Casualisation and labour market transitions in Australia. Journal
of Industrial Relations, 55(1), 6-37.
Watson, I., Buchanan, J., Campbell, I., & Briggs, C. (2003). Fragmented futures: New challenges in
working life.
Watson, N., & Wooden, M. P. (2012). The HILDA survey: a case study in the design and development
of a successful household panel survey. Longitudinal and Life Course Studies, 3(3), 369-381.
Wilkin, C. L. (2013). I can't get no job satisfaction: Meta‐analysis comparing permanent and
contingent workers. Journal of Organizational Behavior, 34(1), 47-64.
Wilkins, R., & Wooden, M. (2013). Gender differences in involuntary job loss: Why are men more
likely to lose their jobs? Industrial Relations: A Journal of Economy and Society, 52(2), 582608.
Wilkins, R., & Wooden, M. (2014). Two decades of change: The Australian labour market, 1993–
2013. Australian Economic Review, 47(4), 417-431.
39
Will, L. (2013). Forms of Work in Australia-Productivity Commission Staff Working Paper.
Wooden, M., & Warren, D. (2003). The characteristics of casual and fixed-term employment:
evidence from the HILDA survey: Melbourne Institute of Applied Economic and Social
Research, The University of Melbourne.
Wooden, M., & Warren, D. (2004). Non-standard employment and job satisfaction: evidence from
the HILDA Survey. Journal of Industrial Relations, 46(3), 275-297.
Yu, W.-h. (2012). Better off jobless? Scarring effects of contingent employment in Japan. Social
Forces, sor031.
40
Table 1.Incidence of financial hardship conditional on the employment arrangement
Panel A. Men
Permanent
Fixedterm
Index of financial hardship (mean)
0.357
0.427***
Average number of financial hardships experienced (%)a
0
80.82
78.22
1
9.90
10.44
2
4.95
5.74
3
2.50
3.32
4
1.16
1.16
5
0.40
0.69
6
0.19
0.26
7
0.09
0.17
Experience of individual financial hardships (%)
Could not pay utility bills on time
11.61
11.73
Asked financial help from friends or family
10.64
13.84***
Could not pay mortgage or rent on time
4.99
6.21**
Pawned or sold something
3.58
4.79***
Was unable to heat home
1.49
1.81
Went without meals
2.36
2.59
Asked help from welfare organizations
0.99
1.77***
Person-wave observations (N)
19,726
2,319
Panel B. Women
Permanent
Fixedterm
Index of financial hardship (mean)
0.371
0.406*
Average number of financial hardships experienced (%)b
0
79.52
1
10.56
2
5.79
3
2.61
4
0.83
5
0.45
6
0.15
7
0.08
Experience of individual financial hardships (%)
Could not pay utility bills on time
12.92
Asked financial help from friends or family
11.70
Could not pay mortgage or rent on time
5.25
Pawned or sold something
2.76
Was unable to heat home
1.52
Went without meals
1.76
Asked help from welfare organizations
1.15
Person-wave observations (N)
17,656
Casual
SelfUnemp.
OOLF
emp.
0.968*** 0.480*** 1.563*** 1.333***
59.82
14.53
10.37
6.82
4.01
2.56
0.99
0.89
77.24
9.88
5.90
3.87
1.67
0.86
0.32
0.26
46.20
14.66
12.10
9.10
6.98
5.74
3.00
2.21
50.55
15.08
12.41
7.44
6.19
4.77
2.47
1.09
24.44*** 14.87*** 31.45*** 28.43***
25.72*** 12.43*** 37.10*** 31.74***
15.45*** 10.12*** 18.46*** 12.86***
10.23*** 4.47*** 22.61*** 19.65***
4.44***
1.58
8.57*** 9.58***
7.96***
2.38
17.40*** 15.08***
8.53*** 2.18*** 20.67*** 15.93***
2,815
5,681
1,132
2,473
Casual
SelfUnemp.
OOLF
emp.
0.721*** 0.450*** 1.533*** 0.972***
77.50
11.32
6.40
3.16
1.20
0.34
0.08
0.00
67.35
13.24
8.43
5.50
3.14
1.42
0.57
0.35
77.61
10.14
6.36
3.50
1.18
0.62
0.34
0.25
44.38
15.99
13.18
9.52
7.18
5.85
2.89
1.01
60.64
13.37
10.04
7.17
4.39
2.27
1.49
0.63
14.22*
12.87*
5.76
2.60
1.77
2.07
1.28
2,658
21.32***
20.71***
10.51***
6.01***
3.35***
4.74***
5.48***
5,127
14.30**
11.79
8.84***
3.94***
2.17***
1.77
2.17***
3,224
38.22***
38.30***
18.49***
17.39***
9.13***
14.43***
17.32***
1,282
26.61***
24.27***
11.65***
10.27***
6.49***
7.60***
10.33***
8,903
Notes: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey. Asterisks indicate statistically
significant differences using permanent employment as reference category (first column) based on t test. *** indicate that differences are
significant at 0.01, ** at 0.05 and * at 0.10 level. a,b Results from 2 test of differences in the relationship distribution were significant at 0.001 level.
41
Table 2. Economic, demographic and job characteristics conditional on the employment
arrangement
Variable
Permanent
Fixedterm
Casual
Gender (Female=1) (%)
0.472
0.534*** 0.646***
Household type (%)
Couples with dependents
0.457
0.413***
0.444**
Couples without dependents
0.340
0.360*** 0.297***
Lone parents with dependents
0.046
0.047
0.090***
Single-persons
0.158
0.179***
0.169**
Age (%)
18-24 years
0.102
0.132*** 0.195***
25-34 years
0.278
0.306***
0.269*
35-44 years
0.290
0.261*** 0.258***
45-54 years
0.329
0.302*** 0.278***
Highest education achieved (%)
Postgraduate
0.052
0.111*** 0.016***
Bachelor, honours or grad dip
0.273
0.330*** 0.151***
Diploma or Cert. III or IV
0.343
0.288*** 0.305***
Year 11 or less
0.332
0.271*** 0.528***
Area of residence (%)
Living in regional area
0.315
0.326*
0.411***
Living in remote area
0.016
0.018
0.021***
Living in city
0.669
0.655**
0.568***
Aboriginal or Torres Strait
Islander (%)
0.014
0.020*** 0.027***
Health variables
Long term health condition affecting work
No condition
0.863
0.866
0.810***
Mild effect (no work limitation)
0.072
0.068
0.071
Moderate effect (limits in some
way but not completely)
0.065
0.066
0.118***
Severe effect (cannot work at
all)
0.000
0.000
0.001**
Suffered from serious personal
injury
0.062
0.060
0.063
Physical functioning (based on
SF-36)
72.848
73.287
70.792***
Mental health (based on SF-36)
75.843
74.597*** 72.857***
Person-wave observations (N)
37,382
4,977
7,942
Selfemp.
Unemp.
OOLF
0.362***
0.531***
0.783***
0.585***
0.274***
0.028***
0.113***
0.360***
0.285***
0.132***
0.223***
0.529***
0.215***
0.120***
0.135***
0.024***
0.181***
0.342***
0.452***
0.257***
0.275
0.237***
0.231***
0.095**
0.280
0.285
0.340**
0.046**
0.218***
0.400***
0.335
0.032***
0.114***
0.292***
0.562***
0.023***
0.144***
0.250***
0.584***
0.380***
0.034***
0.587***
0.385***
0.020
0.595***
0.410***
0.016
0.574***
0.006***
0.073***
0.044***
0.834***
0.059***
0.711***
0.076
0.607***
0.059***
0.107***
0.212***
0.313***
0.000
0.001***
0.021***
0.065
0.097***
0.117***
73.438** 65.615*** 61.981***
76.122
66.617*** 67.393***
8,905
2,414
11,376
Notes: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey.
Asterisks indicate statistically significant differences in characteristics using permanent employment as the reference
category (first column) based on t tests. *** indicate that the differences are significant at 0.01 level, ** at 0.05 level and *
at 0.10 level. The estimates for other variables used in the analyses like industry dummies are given in the appendix C.
42
Table 2 continued. Economic, demographic and job characteristics conditional on the
employment arrangement
Variable
Job characteristics
Jobs held in last financial year
No job
One job
Multiple (two or more )
Member of union
Occupation (%)
Managers and professionals
Technicians and Trades Workers
Community and Personal
Service
Clerical and Administrative
Workers
Sales Workers
Machinery Operators and
Drivers
Labourers
Private sector employee (%)
Other job characteristics (%)
Have access to paid holiday
leave
Have access to paid sick leave
Potential mechanisms
Level of equi. HH income of
respondent
Employment shocks
Fired or made redundant in past
year
Changed job in past year
Hours worked per week in main
job
Person-wave observations (N)
Permanent
Fixedterm
Casual
Self-emp.
Unemp.
OOLF
0.028
0.797
0.175
0.329
0.032
0.683***
0.286***
0.276***
0.070***
0.579***
0.352***
0.126***
0.029
0.803
0.168*
0.107***
0.417***
0.412***
0.171
0.000
0.740***
0.220***
0.040***
0.000
0.411
0.134
0.511***
0.119***
0.158***
0.099***
0.487***
0.213***
_
_
_
_
0.089
0.099**
0.196***
0.052***
_
_
0.181
0.061
0.144***
0.057
0.136***
0.134***
0.094***
0.033***
_
_
_
_
0.062
0.062
0.687
0.030***
0.040***
0.551***
0.077***
0.200***
0.809***
0.045***
0.077***
0.988***
_
_
_
_
_
_
0.963
0.964
0.863***
0.871***
0.047***
0.051***
_
_
_
_
_
_
11.012
11.005
0.021
0.154
0.043***
0.302***
10.635*** 10.806*** 0.929*** 10.420***
0.073***
0.348***
0.023
0.124***
0.419***
0.433***
0.038***
0.053***
40.030
39.319*** 25.101*** 40.985***
37,382
4,977
7,942
8,905
_
2,414
_
11,376
Notes: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey.
Asterisks indicate statistically significant differences in characteristics using permanent employment as the reference
category (first column) based on t tests. *** indicate that the differences are significant at 0.01 level, ** at 0.05 level and *
at 0.10 level. The estimates for other variables used in the analyses like industry dummies are given in the appendix C.
43
Table 3. Average number of types of financial hardship experienced by different types of
employment arrangements conditional on gender and family type
Couples with
dependents
Couple without
dependents
Lone parents with
dependents
Single-persons
Males
Employment contract type
Females
Males
Females
Males
Females
Males
Females
Permanent
0.345
(0.009)
0.328
(0.009)
0.284
(0.010)
0.267
(0.009)
0.676
(0.070)
0.792
(0.033)
0.502
(0.019)
0.513
(0.021)
Fixed-term
0.398*
(0.029)
0.358
(0.027)
0.345**
(0.029)
0.311*
(0.024)
0.783
(0.259)
0.825
(0.085)
0.630**
(0.062)
0.521
(0.047)
Casual
0.911*** 0.523***+++ 0.778*** 0.603***+++
(0.044)
(0.022)
(0.045)
(0.032)
Self-employment
0.412***
(0.017)
0.408***
(0.022)
0.398***
(0.015)
0.352***
(0.030)
1.0333**
(0.161)
Unemployment
1.451***
(0.090)
1.182***++
(0.077)
1.473***
(0.108)
1.287***
(0.091)
1.283*** 1.989***++ 1.833*** 2.158***+
(0.206)
(0.115)
(0.113)
(0.146)
OOLF
1.448*** 0.771***+++ 0.951*** 0.579***+++ 1.830***
(0.062)
(0.019)
(0.054)
(0.030)
(0.169)
Person-wave obs.
(N)
16,600
18,222
10,650
11,818
0.970*
(0.163)
711
1.325***++ 1.308***
(0.064)
(0.064)
1.063***
(0.115)
0.890***
(0.057)
1.162***
(0.062)
0.719***
(0.083)
1.963***
(0.053)
1.497***
(0.070)
1.606***
(0.067)
3,906
6,256
4,986
Notes: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey.
Asterisks indicate statistically significant differences in means using permanent employment as the reference category
(first row) based on t tests. *** indicate that the differences are significant at 0.01 level, ** at 0.05 level and * at 0.10 level.
Plus(+) indicate statistically significant differences in means between men and women in same employment type and same
family type based on t tests. +++ indicate that the differences are significant at 0.01 level, ++ at 0.05 level and + at 0.10 level.
44
Table 4. Two-way fixed effects ordered logit model regression results for men of number
of types of financial hardship faced since the start of calendar year
(Model A)
(Model B)
(Model C)
Plus controls
for health &
job
characteristics
Plus controls for
background &
No controls
attrition
Men in couple families with dependent children (N=8,783)
Fixed-term
0.102
0.114
0.031
(0.208)
(0.205)
(0.202)
Casual
0.671***
0.681***
0.544***
(0.158)
(0.159)
(0.162)
Self-employment
0.174
0.183
0.158
(0.197)
(0.198)
(0.210)
Unemployment
1.088***
1.091***
1.370***
(0.229)
(0.227)
(0.403)
Out of the labour force
0.957***
0.962***
1.172***
(0.213)
(0.215)
(0.422)
b
Wald-test for joint significance
0.870
0.000
Men in couple families without dependent children (N=3,977)
Fixed-term
0.359
0.369
0.399
(0.262)
(0.266)
(0.287)
Casual
0.727***
0.726***
0.627***
(0.206)
(0.205)
(0.220)
Self-employment
0.993***
1.016***
1.004***
(0.304)
(0.307)
(0.324)
Unemployment
1.174***
1.201***
2.648***
(0.282)
(0.285)
(0.554)
Out of the labour force
1.523***
1.457***
2.854***
(0.252)
(0.270)
(0.546)
b
Wald-test for joint significance
0.389
0.000
Men living alone (N=5,175)
Fixed-term
0.118
0.189
0.115
(0.290)
(0.293)
(0.307)
Casual
0.481**
0.476**
0.389*
(0.202)
(0.204)
(0.220)
Self-employment
0.957***
0.886***
0.872***
(0.262)
(0.265)
(0.264)
Unemployment
1.289***
1.324***
0.729
(0.265)
(0.274)
(0.548)
Out of the labour force
0.976***
0.987***
0.236
(0.260)
(0.259)
(0.555)
Wald-test for joint significanceb
0.246
0.000
Notes: a) The reference category is persons in permanent or ongoing employment. b) Model A
includes only dummies. b) P-value of Wald test of joint significance of variables added to each
subsequent model based on 2. d) Standard errors are clustered at the individual level and are given
in parentheses. Significant at the .10 level; ** at the .05 level; *** at the .01 level
45
Table 5. Two-way fixed effects ordered logit model regression results for women of
number of types of financial hardship faced since the start of calendar year
(Model A)
(Model C)
Plus controls
Plus controls for for health &
background &
job
No controls
attrition
characteristics
Women in couple families with dependent children (N=9,915)
Fixed-term
0.136
0.138
0.172
(0.205)
(0.207)
(0.209)
Casual
0.586***
0.583***
0.545***
(0.154)
(0.153)
(0.156)
Self-employment
0.241
0.258
0.296
(0.202)
(0.201)
(0.211)
Unemployment
0.566***
0.559***
0.695
(0.197)
(0.197)
(0.427)
Out of the labour force
0.363**
0.346**
0.467
(0.144)
(0.143)
(0.402)
Wald-test for joint significanceb
0.053
0.000
Women in couple families without dependent children (N=4,255)
Fixed-term
0.257
0.254
0.318
(0.253)
(0.259)
(0.266)
Casual
1.018***
0.910***
0.775***
(0.214)
(0.211)
(0.227)
Self-employment
0.942***
0.879**
0.857***
(0.334)
(0.342)
(0.309)
Unemployment
1.337***
1.300***
1.959***
(0.329)
(0.324)
(0.700)
Out of the labour force
1.285***
1.219***
1.760***
(0.242)
(0.236)
(0.636)
b
Wald-test for joint significance
0.000
0.001
Women lone parents with dependent children (N=4,591)
Fixed-term
-0.019
-0.005
-0.071
(0.336)
(0.331)
(0.319)
Casual
0.686***
0.702***
0.680***
(0.214)
(0.206)
(0.211)
Self-employment
0.599
0.568
0.897**
(0.398)
(0.395)
(0.454)
Unemployment
1.123***
1.138***
0.308
(0.269)
(0.262)
(0.765)
Out of the labour force
0.945***
0.971***
0.061
(0.211)
(0.209)
(0.726)
Wald-test for joint significanceb
0.680
0.000
46
(Model B)
Table 5 continued. Two-way fixed effects ordered logit model regression results for
women of number of types of financial hardship faced since the start of calendar year
(Model A)
(Model B)
(Model C)
Plus controls
Plus controls for
for health &
background &
job
No controls
attrition
characteristics.
Women living alone (N=3,908)
Fixed-term
-0.360
-0.401*
-0.420
(0.227)
(0.240)
(0.260)
Casual
0.186
0.201
0.239
(0.236)
(0.241)
(0.260)
Self-employment
0.677
0.574
0.933*
(0.506)
(0.519)
(0.515)
Unemployment
0.814***
0.744**
0.135
(0.306)
(0.327)
(0.663)
Out of the labour force
0.687**
0.698**
0.043
(0.305)
(0.321)
(0.688)
Wald-test for joint significanceb
0.461
0.000
Notes: a) The reference category is persons in permanent or ongoing employment. b) Model A
includes only dummies. b) P-value of Wald test of joint significance of variables added to each
subsequent model based on 2. d) Standard errors are clustered at the individual level and are
given in parentheses. Significant at the .10 level; ** at the .05 level; *** at the .01 level.
Table 6. The effect of potential mechanisms on coefficient estimates for men
(1)
Main
results
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
(2)
Plus control
for HH
income
Men in couple families with dependents (N=8,783)
0.031
0.041
(0.202)
(0.202)
0.544***
0.537***
(0.162)
(0.162)
0.158
0.162
(0.210)
(0.210)
1.370***
1.378***
(0.403)
(0.401)
1.172***
1.191***
(0.422)
(0.418)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.114**
(0.058)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
47
(3)
Plus controls
for emp.
shocks
(4)
Plus control
for hours
worked
0.018
(0.199)
0.395**
(0.164)
0.118
(0.212)
1.172***
(0.402)
1.078**
(0.419)
0.001
(0.201)
0.267*
(0.161)
0.092
(0.215)
0.507
(0.441)
0.465
(0.458)
-0.117**
(0.058)
-0.108*
(0.060)
0.552***
(0.172)
0.382***
(0.113)
0.539***
(0.170)
0.366***
(0.111)
-0.019***
(0.005)
Table 6 continued. The effect of potential mechanisms on coefficient estimates for men
(1)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
(2)
(3)
Plus control
Plus controls
Main
for HH
for emp.
results
income
shocks
Men in couple families without dependents (N=3,977)
0.399
0.387
0.381
(0.287)
(0.283)
(0.281)
0.627***
0.621***
0.528**
(0.220)
(0.223)
(0.221)
1.004***
0.961***
0.907***
(0.324)
(0.323)
(0.325)
2.648***
2.651***
2.459***
(0.554)
(0.549)
(0.565)
2.854***
2.819***
2.636***
(0.546)
(0.546)
(0.545)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.279*
(0.159)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Casual
Self-employment
Unemployment
Out of the labour force
Men living alone (N=5,175)
0.115
0.106
(0.307)
(0.308)
0.389*
0.379*
(0.220)
(0.220)
0.872***
0.872***
(0.264)
(0.263)
0.729
0.750
(0.548)
(0.551)
0.236
0.262
(0.555)
(0.558)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.053
(0.041)
Employment shocks
Fired or made redundant in past year
0.441
(0.288)
0.408*
(0.226)
0.918***
(0.311)
1.618***
(0.597)
1.854***
(0.581)
-0.269*
(0.152)
-0.213*
(0.115)
0.425**
(0.214)
0.149
(0.172)
0.343
(0.213)
0.155
(0.171)
-0.028***
(0.007)
0.077
(0.300)
0.230
(0.215)
0.840***
(0.258)
0.463
(0.544)
0.033
(0.550)
0.075
(0.304)
0.088
(0.222)
0.810***
(0.263)
-0.101
(0.595)
-0.516
(0.603)
-0.057
(0.041)
-0.060
(0.041)
Hours worked per week in main job
Fixed-term
(4)
Plus control
for hours
worked
0.762***
(0.232)
0.353*
(0.185)
0.740***
(0.236)
Changed job in past year
0.359**
(0.182)
Hours worked per week in main job
-0.018***
(0.007)
Notes: a) The reference category is persons in permanent or ongoing employment. b) Baseline results
include controls for observed background, attrition, health, and job characteristics.
c) Standard errors are clustered at the individual level and are given in parentheses.
Significant at the .10 level; ** at the .05 level; *** at the .01 level
48
Table 7. The effect of potential mechanisms on coefficient estimates for women
(1)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
(2)
(3)
Plus
control for Plus controls
Main
HH
for emp.
results
income
shocks
Women in couple families with dependents (N=9,915)
0.172
0.181
0.148
(0.209)
(0.208)
(0.210)
0.545***
0.518***
0.507***
(0.156)
(0.158)
(0.159)
0.296
0.264
0.290
(0.211)
(0.214)
(0.214)
0.695
0.692
0.674
(0.427)
(0.428)
(0.427)
0.467
0.438
0.451
(0.402)
(0.402)
(0.400)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.197***
(0.062)
Employment shocks
Fired or made redundant in past year
Changed job in past year
0.128
(0.208)
0.404**
(0.166)
0.212
(0.215)
0.326
(0.440)
0.125
(0.409)
-0.200***
(0.062)
0.088
(0.216)
0.261**
(0.122)
0.066
(0.216)
0.288**
(0.122)
-0.012**
(0.005)
Women in couple families without dependents (N=4,255)
0.318
0.319
0.293
(0.266)
(0.266)
(0.264)
Casual
0.775***
0.772***
0.689***
(0.227)
(0.227)
(0.230)
Self-employment
0.857***
0.824***
0.791**
(0.309)
(0.308)
(0.308)
Unemployment
1.959***
1.934***
1.905***
(0.700)
(0.700)
(0.692)
Out of the labour force
1.760***
1.732***
1.735***
(0.636)
(0.632)
(0.623)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.098
-0.092
(0.091)
(0.090)
Employment shocks
Fired or made redundant in past year
-0.014
(0.250)
Changed job in past year
0.385**
(0.155)
Hours worked per week in main job
49
Plus control
for hours
worked
-0.195***
(0.062)
Hours worked per week in main job
Fixed-term
(4)
0.279
(0.261)
0.491*
(0.254)
0.781**
(0.314)
1.346*
(0.725)
1.177*
(0.666)
-0.099
(0.094)
-0.020
(0.248)
0.399**
(0.156)
-0.018**
(0.008)
Table 7 continued. The effect of potential mechanisms on coefficient estimates for women
(1)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
(2)
Plus
Main
control for
results
HH income
Women in lone parent families (N=4,591)
-0.071
-0.033
(0.319)
(0.327)
0.680***
0.596***
(0.211)
(0.209)
0.897**
0.806*
(0.454)
(0.456)
0.308
0.253
(0.765)
(0.775)
0.061
-0.003
(0.726)
(0.735)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.680***
(0.180)
(3)
Plus controls
for emp.
shocks
(4)
Plus control
for hours
worked
-0.029
(0.330)
0.591***
(0.209)
0.809*
(0.455)
0.208
(0.787)
-0.036
(0.749)
-0.019
(0.332)
0.464**
(0.220)
0.808*
(0.464)
0.012
(0.822)
-0.214
(0.783)
-0.686***
(0.181)
-0.670***
(0.181)
0.259
(0.308)
-0.043
(0.182)
0.223
(0.311)
-0.037
(0.184)
-0.016*
(0.008)
-0.493*
(0.271)
0.130
(0.263)
0.748
(0.485)
-0.106
(0.679)
-0.260
(0.692)
-0.467*
(0.274)
0.002
(0.276)
0.708
(0.473)
-0.529
(0.698)
-0.626
(0.704)
-0.218**
(0.086)
-0.210**
(0.083)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Women living alone (N=3,908)
-0.420
-0.481*
(0.260)
(0.265)
0.239
0.203
(0.260)
(0.258)
0.933*
0.890*
(0.515)
(0.491)
0.135
0.028
(0.663)
(0.668)
0.043
-0.088
(0.688)
(0.694)
Potential mechanisms
Log total gr. equi. HH income of the respondent
-0.213**
(0.084)
Employment shocks
Fired or made redundant in past year
1.059***
(0.221)
0.210
(0.196)
1.058***
(0.225)
Changed job in past year
0.198
(0.197)
Hours worked per week in main job
-0.018**
(0.008)
Notes: a) The reference category is persons in permanent or ongoing employment. b) Baseline results
include controls for observed background, attrition, health, and job characteristics.
c) Standard errors are clustered at the individual level and are given in parentheses.
Significant at the .10 level; ** at the .05 level; *** at the .01 level
50
Appendices
Appendix A. SCQ response conditional on employment arrangements
Permanent
Fixed-term
Casual
Selfemp.
Unemp.
OOLF
No response
9.97
10.12
12.42
12.30
16.19
11.69
Response
90.03
89.88
87.58***
87.70***
83.81***
88.31***
SCQ response status
Note: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey. ***
implies that means are statistically different from means in the first row, i.e., individuals employed on permanent contracts
at 0.01 level, ** at 0.05 level and * at 0.10 level
Appendix B. Summary statistics for industry variables conditional on employment
Permanent
Fixedterm
Casual
Self-emp.
Unemp.
OOLF
Agriculture, Forestry and Fishing
0.012
0.012
0.039***
0.097***
0.000
0.000
Mining
0.023
0.024
0.011***
0.003***
0.000
0.000
Manufacturing
0.112
0.067***
0.077***
0.069***
0.000
0.000
Electricity, gas, water & Waste serv.
0.014
0.011
0.007***
0.001***
0.000
0.000
Construction
0.058
0.051**
0.061
0.203***
0.000
0.000
Wholesale Trade
0.039
0.024***
0.027***
0.027***
0.000
0.000
Retail Trade
0.078
0.069**
0.138***
0.068***
0.000
0.000
Accommodation and Food Services
0.027
0.021***
0.150***
0.039***
0.000
0.000
Transport, Postal and Warehousing
0.045
0.026***
0.047
0.045
0.000
0.000
Information Media and Telecom.
0.027
0.033**
0.018***
0.020***
0.000
0.000
Financial and Insurance Services
0.053
0.037***
0.010***
0.027***
0.000
0.000
Rental, hiring & real estate services
0.014
0.011*
0.010**
0.016
0.000
0.000
Professional, Scientific and Tech.
0.077
0.075
0.044***
0.144***
0.000
0.000
Administrative and Support Service
0.022
0.022
0.041***
0.054***
0.000
0.000
Public Administration and Safety
0.100
0.097
0.027***
0.003***
0.000
0.000
Education and Training
0.111
0.216***
0.100***
0.023***
0.000
0.000
Health Care and Social Assistance
0.143
0.158***
0.137
0.059***
0.000
0.000
Arts and Recreation Services
0.012
0.021***
0.025***
0.023***
0.000
0.000
Other Services
0.033
0.026**
0.030
0.080***
0.000
0.000
Variable
Industry (%)
Person-wave observations (N)
37382
4977
7942
8905
2414
11376
Notes: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey.
Asterisks indicate statistically significant differences in characteristics using permanent employment as the reference
category (first column) based on t tests. *** indicate that the differences are significant at 0.01 level, ** at 0.05 level and *
at 0.10 level.
51
Appendix C. Percent of individuals for whom their answer to different hardship questions each
year matches/agree with their partner in HILDA data
2001
2002
2003
2004
2005
2006
2007
2008
2009
2011
2012
2013
Type of financial hardship
Could not pay
utility bills on time
83.85
84.01
86.73
85.71
86.25
87.79
87.19
86.60
86.21
86.67
86.09
86.68
Asked financial help
from friends/family
86.05
86.90
87.89
87.83
87.19
88.86
87.52
87.44
88.13
86.57
87.28
87.49
Could not pay
mortgage or rent
on time
88.75
89.03
89.83
90.11
90.99
91.62
90.21
90.94
90.49
90.39
90.58
91.22
Pawned or sold
something
89.71
91.14
91.52
91.53
92.40
92.43
92.26
92.83
92.47
91.37
91.15
91.38
Was unable to heat
home
92.35
92.84
93.78
93.08
93.75
94.62
93.70
93.92
93.93
93.22
93.12
93.62
Went without
meals when hungry
92.37
92.66
93.56
92.56
93.65
93.67
93.55
93.45
93.33
93.41
93.49
93.47
Asked help from
welfare
organization
91.78
92.77
93.20
92.91
93.13
93.45
93.60
93.61
93.59
93.37
93.24
93.36
Overall concordance
90.79
Appendix D1. Two-way fixed effects ordered logit model complete regression results for
men in couple families with dependents of number of types of financial hardship
experienced since the start of calendar year
VARIABLES
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
52
(1)
Model A
(2)
Model B
(3)
Model C
(4)
Model D
0.102
(0.208)
0.671***
(0.158)
0.174
(0.197)
1.088***
(0.229)
0.957***
(0.213)
0.114
(0.205)
0.681***
(0.159)
0.183
(0.198)
1.091***
(0.227)
0.962***
(0.215)
0.031
(0.202)
0.544***
(0.162)
0.158
(0.210)
1.370***
(0.403)
1.172***
(0.422)
0.001
(0.201)
0.267*
(0.161)
0.092
(0.215)
0.507
(0.441)
0.465
(0.458)
-0.078
(0.379)
0.106
(0.195)
-0.119
(0.255)
-0.177
(0.363)
0.086
(0.192)
-0.176
(0.253)
-0.189
(0.371)
0.083
(0.195)
-0.136
(0.254)
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
Remote area
Telephonic interview
Non-respondent in next wave
Health controls
Long term health condition affecting work (ref: Severe condition)
No condition
Mild effect (no work limitation)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
Managers or professional
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
Potential mediators
53
0.454
(1.043)
0.567
(0.838)
0.299
(0.310)
0.792
(1.184)
0.788
(0.964)
0.324
(0.320)
0.640
(1.181)
0.877
(0.919)
0.423
(0.316)
0.249
(0.247)
0.850
(0.865)
-0.016
(0.210)
-0.215
(0.193)
0.095
(0.238)
1.373**
(0.651)
0.021
(0.214)
-0.232
(0.184)
0.087
(0.244)
1.357**
(0.667)
0.066
(0.230)
-0.206
(0.185)
0.808
(0.496)
1.072**
(0.522)
0.870*
(0.481)
0.376***
(0.136)
-0.009**
(0.004)
-0.015***
(0.003)
0.789
(0.482)
1.090**
(0.513)
0.841*
(0.461)
0.360**
(0.140)
-0.009**
(0.004)
-0.016***
(0.003)
-0.106
(0.190)
0.294
(0.207)
-0.185
(0.150)
-0.108
(0.189)
0.081
(0.212)
-0.130
(0.152)
0.088
(0.238)
0.218
(0.207)
0.234
(0.350)
0.027
(0.288)
0.414
(0.303)
0.233
(0.251)
0.197
(0.230)
0.194
(0.230)
0.247
(0.204)
0.227
(0.326)
0.010
(0.279)
0.460
(0.302)
0.208
(0.251)
0.220
(0.230)
Log total gr. equi. HH income of the respondent
-0.108*
(0.060)
Employment shocks
Fired or made redundant in past year
0.539***
(0.170)
Changed job in past year
0.366***
(0.111)
Hours worked per week in main job
-0.019***
(0.005)
Wald-test for joint significancec
0.870
0.000
0.000
Observations
8,783
8,783
8,783
8,783
Notes: (a) In addition to those reported, all models include a dummy for ‘other’ contract, a dummy for unpaid
family worker, dummies for industry (19 categories) and dummies for each month to control for differences in
date of the interviews. (b) Model A, B and C shows complete results for model A, B and C of main regressions
and model D shows complete results for column (4) in mechanisms analysis. (c) The variable non-respondent at
next wave is measured as a dummy variable indicating whether the respondent was observed in the survey at
t+1. (d) c indicates P-value of Wald test of joint significance of variables added to each subsequent model based
on χ2. (e) Standard errors are clustered at the individual level. Significant at the .10 level; ** at the .05 level; ***
at the .01 level.
Appendix D2: Two-way fixed effects ordered logit model complete regression results for
men in couple families without dependents of number of types of financial hardship
experienced since the start of calendar year
VARIABLES
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
54
(1)
Model A
(2)
Model B
(3)
Model C
(4)
Model D
0.359
(0.262)
0.727***
(0.206)
0.993***
(0.304)
1.174***
(0.282)
1.523***
(0.252)
0.369
(0.266)
0.726***
(0.205)
1.016***
(0.307)
1.201***
(0.285)
1.457***
(0.270)
0.399
(0.287)
0.627***
(0.220)
1.004***
(0.324)
2.648***
(0.554)
2.854***
(0.546)
0.441
(0.288)
0.408*
(0.226)
0.918***
(0.311)
1.618***
(0.597)
1.854***
(0.581)
0.233
(0.498)
0.143
(0.356)
0.057
(0.525)
0.315
(0.520)
0.172
(0.369)
0.119
(0.498)
0.294
(0.507)
0.136
(0.363)
0.085
(0.517)
1.163
(1.452)
-0.940
(0.941)
0.112
(0.416)
2.890*
(1.524)
-0.538
(0.979)
0.160
(0.402)
3.474**
(1.466)
-0.345
(0.976)
0.142
(0.387)
-0.036
0.008
0.049
(0.288)
-1.532*
(0.784)
-0.185
(0.320)
0.084
(0.219)
Remote area
Telephonic interview
Non-respondent in next wave
Health controls
Long term health condition affecting work (ref: Severe
condition)
No condition
Mild effect (no work limitation)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
Managers or professional
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
Potential mediators
ln total gr. equi. HH income of the respondent
(0.297)
-1.767**
(0.799)
-0.237
(0.321)
0.024
(0.232)
(0.286)
-1.594**
(0.777)
-0.270
(0.317)
0.015
(0.243)
-0.825*
(0.474)
-0.530
(0.484)
-0.467
(0.408)
-0.139
(0.190)
-0.003
(0.006)
-0.019***
(0.005)
-0.651
(0.481)
-0.365
(0.488)
-0.299
(0.422)
-0.184
(0.195)
-0.004
(0.006)
-0.020***
(0.005)
0.053
(0.303)
0.280
(0.325)
0.071
(0.239)
0.018
(0.316)
0.123
(0.349)
0.109
(0.247)
0.444
(0.321)
0.665**
(0.314)
0.297
(0.451)
0.829**
(0.400)
0.504
(0.475)
-0.266
(0.354)
0.553*
(0.316)
0.690**
(0.341)
0.817**
(0.328)
0.332
(0.461)
0.971**
(0.398)
0.646
(0.474)
-0.159
(0.362)
0.587*
(0.330)
-0.213*
(0.115)
Employment shocks
Fired or made redundant in past year
0.343
(0.213)
0.155
(0.171)
Changed job in past year
55
Hours worked per week in main job
Wald-test for joint significancec
Observations
Notes: See footnote of Table D1
3,977
0.389
3,977
0.000
3,977
-0.028***
(0.007)
0.000
3,977
Appendix D3: Two-way fixed effects ordered logit model complete regression results for
men living alone of number of types of financial hardship experienced since the start of
calendar year
VARIABLES
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
Remote area
Telephonic interview
Non-respondent in next wave
Health controls
Long term health condition affecting work (ref: Severe
condition)
No condition
Mild effect (no work limitation)
56
(1)
Model A
(2)
Model B
(3)
Model C
(4)
Model D
0.118
(0.290)
0.481**
(0.202)
0.957***
(0.262)
1.289***
(0.265)
0.976***
(0.260)
0.189
(0.293)
0.476**
(0.204)
0.886***
(0.265)
1.324***
(0.274)
0.987***
(0.259)
0.115
(0.307)
0.389*
(0.220)
0.872***
(0.264)
0.729
(0.548)
0.236
(0.555)
0.075
(0.304)
0.088
(0.222)
0.810***
(0.263)
-0.101
(0.595)
-0.516
(0.603)
-0.275
(0.504)
0.081
(0.301)
-0.362
(0.288)
-0.458
(0.508)
0.054
(0.311)
-0.403
(0.273)
-0.576
(0.510)
0.023
(0.306)
-0.310
(0.260)
1.611
(1.657)
0.271
(0.545)
-0.515
(0.497)
2.103
(1.751)
0.248
(0.603)
-0.543
(0.542)
2.270
(1.856)
0.043
(0.527)
-0.604
(0.524)
-0.084
(0.256)
0.180
(0.455)
-0.708***
(0.268)
-0.052
(0.209)
-0.008
(0.265)
0.396
(0.545)
-0.517**
(0.240)
-0.034
(0.206)
-0.087
(0.259)
0.429
(0.539)
-0.471**
(0.240)
0.010
(0.214)
-0.862*
(0.458)
-0.877*
(0.487)
-0.860*
(0.467)
-0.900*
(0.488)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
Managers or professional
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
-0.588
(0.436)
0.131
(0.168)
-0.005
(0.006)
-0.021***
(0.004)
-0.604
(0.442)
0.071
(0.168)
-0.005
(0.006)
-0.022***
(0.004)
0.208
(0.212)
0.372
(0.260)
-0.348
(0.220)
0.190
(0.221)
0.104
(0.279)
-0.338
(0.227)
-0.593
(0.368)
-0.411
(0.288)
-0.386
(0.469)
-0.247
(0.387)
-0.102
(0.471)
-0.201
(0.360)
-0.174
(0.296)
-0.554
(0.362)
-0.369
(0.296)
-0.295
(0.467)
-0.236
(0.384)
-0.103
(0.475)
-0.060
(0.367)
-0.156
(0.300)
Potential mediators
ln total gr. equi. HH income of the respondent
-0.060
(0.041)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
Wald-test for joint significancec
Observations
Notes: See footnote of Table D1
5,175
57
0.246
5,175
0.000
5,175
0.740***
(0.236)
0.359**
(0.182)
-0.018***
(0.007)
0.000
5,175
Appendix E1. Two-way fixed effects ordered logit model complete regression results for
women in couple families with dependents of number of types of financial hardship
experienced since the start of calendar year
VARIABLES
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
Remote area
Telephonic interview
Non-respondent in next wave
Health controls
Long term health condition affecting work (ref: Severe
condition)
No condition
Mild effect (no work limitation)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
58
(1)
Model A
(2)
Model B
(3)
Model C
(4)
Model D
0.136
(0.205)
0.586***
(0.154)
0.241
(0.202)
0.566***
(0.197)
0.363**
(0.144)
0.138
(0.207)
0.583***
(0.153)
0.258
(0.201)
0.559***
(0.197)
0.346**
(0.143)
0.172
(0.209)
0.545***
(0.156)
0.296
(0.211)
0.695
(0.427)
0.467
(0.402)
0.128
(0.208)
0.404**
(0.166)
0.212
(0.215)
0.326
(0.440)
0.125
(0.409)
-0.010
(0.310)
-0.147
(0.170)
-0.003
(0.232)
0.036
(0.320)
-0.115
(0.170)
-0.011
(0.227)
0.009
(0.314)
-0.121
(0.168)
-0.048
(0.228)
-0.622
(0.774)
-0.655
(0.465)
-0.267
(0.252)
-0.248
(0.779)
-0.562
(0.451)
-0.231
(0.248)
-0.160
(0.769)
-0.459
(0.452)
-0.221
(0.247)
0.053
(0.203)
-3.737***
(1.143)
0.347*
(0.199)
-0.075
(0.184)
0.110
(0.196)
-3.767***
(1.134)
0.293
(0.207)
-0.036
(0.189)
0.100
(0.195)
-3.786***
(1.144)
0.326
(0.208)
-0.040
(0.189)
0.094
(0.504)
-0.018
(0.524)
0.194
(0.472)
0.031
(0.140)
-0.008**
0.118
(0.475)
0.014
(0.492)
0.208
(0.442)
0.040
(0.141)
-0.008**
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
Managers or professional
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
(0.003)
-0.020***
(0.003)
(0.003)
-0.020***
(0.003)
0.044
(0.127)
0.141
(0.165)
-0.110
(0.177)
0.065
(0.128)
0.032
(0.175)
-0.088
(0.181)
-0.303
(0.266)
-0.009
(0.368)
-0.264
(0.266)
-0.187
(0.287)
-0.126
(0.290)
-0.289
(0.476)
-0.168
(0.183)
-0.265
(0.271)
-0.007
(0.374)
-0.266
(0.264)
-0.183
(0.289)
-0.127
(0.295)
-0.311
(0.481)
-0.187
(0.184)
Potential mediators
ln total gross household income of the respondent
-0.200***
(0.062)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
Wald-test for joint significancec
Observations
Notes: See footnote of Table D1
0.000
9,915
0.053
9,915
0.000
9,915
0.066
(0.216)
0.288**
(0.122)
-0.012**
(0.005)
0.000
9,915
Appendix E2: Two-way fixed effects ordered logit model complete regression results for
women in couple families without dependents of number of types of financial hardship
experienced since the start of calendar year
VARIABLES
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
59
(1)
Model A
(2)
Model B
(3)
Model C
(4)
Model D
0.257
(0.253)
1.018***
(0.214)
0.942***
0.254
(0.259)
0.910***
(0.211)
0.879**
0.318
(0.266)
0.775***
(0.227)
0.857***
0.279
(0.261)
0.491*
(0.254)
0.781**
(0.334)
1.337***
(0.329)
1.285***
(0.242)
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
Remote area
Telephonic interview
Non-respondent in next wave
Health controls
Long term health condition affecting work (ref: Severe
condition)
No condition
Mild effect (no work limitation)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
60
(0.342)
1.300***
(0.324)
1.219***
(0.236)
(0.309)
1.959***
(0.700)
1.760***
(0.636)
(0.314)
1.346*
(0.725)
1.177*
(0.666)
1.118**
(0.520)
1.180***
(0.416)
-0.247
(0.404)
0.989*
(0.531)
1.180***
(0.431)
-0.392
(0.393)
1.073**
(0.509)
1.242***
(0.408)
-0.453
(0.398)
-18.067***
(1.156)
-2.710***
(1.025)
-0.465
(0.372)
-18.902***
(1.401)
-3.624***
(1.280)
-0.635*
(0.364)
-17.677***
(1.605)
-3.761**
(1.509)
-0.671*
(0.357)
0.231
(0.349)
-1.225**
(0.588)
-0.147
(0.309)
-0.182
(0.228)
0.287
(0.334)
-1.048*
(0.573)
-0.161
(0.289)
-0.166
(0.238)
0.262
(0.333)
-1.049*
(0.571)
-0.199
(0.294)
-0.136
(0.243)
0.113
(0.584)
0.148
(0.603)
0.117
(0.566)
0.139
0.219
(0.558)
0.256
(0.572)
0.173
(0.537)
0.128
(0.176)
-0.010**
(0.005)
-0.016***
(0.005)
(0.178)
-0.010**
(0.005)
-0.016***
(0.005)
0.006
(0.248)
-0.138
(0.268)
0.047
(0.285)
0.012
(0.248)
-0.310
(0.266)
0.076
(0.285)
Managers or professional
-0.095
(0.369)
-0.305
(0.445)
0.021
(0.454)
-0.048
(0.347)
0.384
(0.405)
1.054
(0.812)
0.543**
(0.233)
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
Potential mediators
ln total gross household income of the respondent
0.025
(0.364)
-0.214
(0.448)
0.020
(0.440)
0.024
(0.341)
0.395
(0.404)
1.077
(0.790)
0.557**
(0.231)
-0.099
(0.094)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
Wald-test for joint significancec
Observations
Notes: See footnote of Table D1
0.000
4,255
0.000
4,255
0.001
4,255
-0.020
(0.248)
0.399**
(0.156)
-0.018**
(0.008)
0.017
4,255
Appendix E3: Two-way fixed effects ordered logit model complete regression results for
women in lone parent families with dependents of number of types of financial hardship
experienced since the start of calendar year
VARIABLES
(1)
(2)
(3)
(4)
Model A
Model B
Model C
Model D
-0.019
-0.005
-0.071
-0.019
(0.336)
(0.331)
(0.319)
(0.332)
0.686***
0.702***
0.680***
0.464**
(0.214)
(0.206)
(0.211)
(0.220)
0.599
0.568
0.897**
0.808*
(0.398)
(0.395)
(0.454)
(0.464)
1.123***
1.138***
0.308
0.012
(0.269)
(0.262)
(0.765)
(0.822)
0.945***
(0.211)
0.971***
(0.209)
0.061
(0.726)
-0.214
(0.783)
-0.412
(0.431)
0.078
(0.265)
-0.193
-0.699
(0.447)
-0.045
(0.264)
-0.231
-0.788*
(0.453)
-0.026
(0.264)
-0.213
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
61
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
Remote area
Telephonic interview
Non-respondent in next wave
Health controls
Long term health condition affecting work (ref: Severe condition)
No condition
Mild effect (no work limitation)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
Managers or professional
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
62
(0.318)
(0.314)
(0.311)
-0.194
(1.811)
0.433
(0.546)
-0.086
(0.324)
-0.630
(1.735)
0.253
(0.509)
-0.159
(0.357)
-0.274
(1.780)
0.376
(0.494)
-0.077
(0.353)
-0.608
(0.423)
-0.275
(1.049)
0.145
(0.272)
0.111
(0.242)
-0.795*
(0.414)
-0.262
(1.024)
0.232
(0.288)
0.146
(0.256)
-0.919**
(0.436)
-0.297
(1.149)
0.212
(0.293)
0.162
(0.253)
-1.141*
(0.623)
-1.192*
(0.665)
-1.219**
(0.609)
-0.084
(0.191)
-0.011**
(0.005)
-0.018***
(0.004)
-1.019*
(0.584)
-1.099*
(0.626)
-1.142**
(0.569)
-0.054
(0.195)
-0.011**
(0.005)
-0.017***
(0.004)
-0.342*
(0.183)
-0.053
(0.258)
-0.468*
(0.281)
-0.307*
(0.180)
0.011
(0.274)
-0.423
(0.289)
0.220
(0.387)
-0.662
(0.688)
-0.235
(0.369)
-0.471
(0.421)
0.187
(0.385)
-1.227*
(0.678)
-0.060
0.273
(0.384)
-0.766
(0.743)
-0.175
(0.362)
-0.450
(0.417)
0.156
(0.397)
-1.144*
(0.688)
-0.072
(0.250)
Potential mediators
ln total gross household income of the respondent
(0.243)
-0.670***
(0.181)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
Wald-test for joint significancec
Observations
Notes: See footnote of Table D1
0.297
4,591
0.680
4,591
0.000
4,591
0.223
(0.311)
-0.037
(0.184)
-0.016*
(0.008)
0.001
4,591
Appendix E4: Two-way fixed effects ordered logit model complete regression results for
women living alone of number of types of financial hardship experienced since the start of
calendar year
VARIABLES
Type of employment arrangement (ref: on-going contract)
Fixed-term
Casual
Self-employment
Unemployment
Out of the labour force
Age (Ref: 35-44 years)
18-24 years
25-34 years
45-54 years
Highest level of education achieved (ref: Year 12 or less)
Postgraduate
Bachelor, honours or grad dip
Diploma or Cert. III or IV
Area of residence (ref: city)
Regional area
Remote area
Telephonic interview
63
(1)
Model A
(2)
Model B
(3)
Model C
(4)
Model D
-0.360
(0.227)
0.186
(0.236)
0.677
(0.506)
0.814***
(0.306)
0.687**
(0.305)
-0.401*
(0.240)
0.201
(0.241)
0.574
(0.519)
0.744**
(0.327)
0.698**
(0.321)
-0.420
(0.260)
0.239
(0.260)
0.933*
(0.515)
0.135
(0.663)
0.043
(0.688)
-0.467*
(0.274)
0.002
(0.276)
0.708
(0.473)
-0.529
(0.698)
-0.626
(0.704)
0.124
(0.594)
-0.323
(0.387)
-0.107
(0.409)
0.081
(0.614)
-0.240
(0.396)
-0.241
(0.390)
0.278
(0.621)
-0.086
(0.396)
-0.282
(0.401)
-0.146
(0.806)
0.821
(0.661)
-0.552
(0.498)
-0.353
(0.905)
0.807
(0.629)
-0.591
(0.467)
-0.278
(0.923)
0.944
(0.656)
-0.675
(0.477)
-0.494*
(0.299)
-0.058
(0.531)
0.141
(0.316)
-0.530*
(0.300)
-0.117
(0.551)
0.246
(0.332)
-0.440
(0.314)
-0.262
(0.590)
0.173
(0.323)
Non-respondent in next wave
0.045
(0.228)
Health controls
Long term health condition affecting work (ref: Severe condition)
No condition
Mild effect (no work limitation)
Moderate effect (limits in some way but not completely)
Suffered from serious personal injury
Physical functioning (based on SF-36)
Mental health (based on SF-36)
Job factors
Jobs held in last financial year (ref: held no job)
One job
Multiple (two or more )
Member of union/employee association
Occupation (ref: Labourers)
Managers or professional
Technicians and Trades Workers
Community and Personal Service Work
Clerical and Administrative Workers
Sales Workers
Machinery Operators and Drivers
Private sector employee
0.006
(0.236)
0.025
(0.236)
-0.636
(0.496)
-0.695
(0.505)
-0.268
(0.455)
0.237
(0.194)
-0.004
(0.005)
0.013***
(0.005)
-0.633
(0.606)
-0.654
(0.604)
-0.392
(0.577)
0.231
(0.194)
-0.004
(0.005)
0.013***
(0.005)
0.066
(0.242)
0.470*
(0.264)
0.141
(0.301)
-0.036
(0.259)
0.212
(0.299)
0.150
(0.269)
-0.831*
(0.468)
0.064
(0.570)
-0.188
(0.474)
0.190
(0.501)
-0.630
(0.522)
0.759
(0.654)
-0.374
(0.302)
-0.823*
(0.497)
-0.030
(0.575)
-0.124
(0.511)
0.149
(0.514)
-0.693
(0.533)
0.349
(0.627)
-0.295
(0.298)
Potential mediators
ln total gross household income of the respondent
-0.210**
(0.083)
Employment shocks
Fired or made redundant in past year
Changed job in past year
Hours worked per week in main job
Wald-test for joint significancec
Observations
Notes: See footnote of Table D1
0.009
3,908
64
0.461
3,908
0.000
3,908
1.058***
(0.225)
0.198
(0.197)
-0.018**
(0.008)
0.000
3,908
Appendix F. Contingency table for employment at t and t-1 for men and women (row
percentages and frequencies)
Panel A: Men
Permanent
Type of employment arrangement at t
Fixed-term
Casual
Self-emp.
Unemp.
OOLF
Type of employment arrangement at t-1
Permanent
86.62
(15,528)
2.97
(532)
2.66
(476)
1.27
(227)
1.34
(241)
Fixed-term
48.26
(1,013)
38.26
(803)
4.81
(101)
4.24
(89)
2.05
(43)
2.10
(44)
28.29 (687)
6.38
(155)
46.54
(1,130)
5.81
(141)
6.05
(147)
6.34
(154)
Self-employment
6.90
(361)
1.66
(87)
2.79
(146)
85.39
(4,465)
0.76
(40)
2.35
(123)
Unemployment
21.04
(211)
4.79
(48)
22.43
(225)
5.38
(54)
27.22
(273)
18.84
(189)
OOLF
6.80
(138)
2.17
(44)
7.68
(156)
5.07
(103)
8.18
(166)
70.05
(1,422)
Casual
4.99
(894)
Panel B: Women
Type of employment arrangement at t
Permanent
Fixed-term
Casual
Self-emp.
Unemp.
OOLF
Type of employment arrangement at t-1
Permanent
81.96
(13,031)
5.38
(855)
4.68
(744)
1.51
(240)
0.94
(149)
5.36
(853)
Fixed-term
43.19
(1,025)
40.29
(956)
6.32
(150)
2.02
(48)
1.73
(41)
5.94
(141)
Casual
22.01
(1,053)
5.92
(283)
53.11
(2,541)
3.76
(180)
3.62
(173)
11.35
(543)
Self-employment
6.86
(202)
2.45
(72)
5.30
(156)
76.89
(2,263)
1.02
(30)
7.31
(215)
Unemployment
18.10
(204)
5.68
(64)
18.10
(204)
2.40
(27)
24.58
(277)
30.88
(348)
OOLF
6.82
(550)
1.41
(114)
8.15
(657)
3.15
(254)
4.95
(399)
75.42
(6,082)
Note: I included, but do not report the persons in 'other' employment arrangements
65
Appendix G. Respondents' preference for work hours compared to the current working
hours
Panel A. Men
Permanent
Fixed-term
Casual
Self-emp.
Fewer hours
30.27
31.45
13.54***
39.36***
About the same
60.97
58.24 **
49.06***
51.04***
More hours
8.76
10.31**
37.40***
9.60*
19,717
2,318
2,813
5,676
Prefer to work
Person-wave obs.
Panel A. Women
Permanent
Fixed-term
Casual
Self-emp.
Fewer hours
31.76
32.97
9.34***
27.44***
About the same
59.21
55.78***
55.23***
58.21
More hours
9.03
11.25***
35.44***
14.34***
Prefer to work
Person-wave obs.
17,652
2,657
5,119
3,221
Notes: The statistics reported in the table were calculated using unweighted longitudinal data from the HILDA survey.
Asterisks indicate statistically significant differences in means using permanent employment as the reference category
(first column) based on t tests. *** indicate that the differences are significant at 0.01 level, ** at 0.05 level and * at 0.10
level
66