Labour Market Preferences of Retrenched Australian Auto Industry Workers for Job Quality and Meaningful Work *

This study uses stated preference experiments to examine labour market preferences of 309 workers retrenched by the Australian automotive industry for non-pecuniary job attributes denoting job quality and meaningfulness. We ﬁnd that autonomy and employer reputation for good work policies and practices are the two most important non-pecuniary job attributes, with compensating wage differentials of roughly $5 per hour for greater autonomy and better employer reputation. Job security and skill utilisation are also important, but less so, with compensating wage differentials between $1 and $3 per hour for greater security and fewer training requirements. Workers’ strongest preference is not for a particular type of work, but rather for a particular type of employer, suggesting that labour market policy might pay more attention to regulating the quality of workplaces.


I Introduction
As labour-intensive manufacturing retreats from economically developed nations, there is an increasing need for displaced, mechanically specialised workforces to be accommodated into different industries.The increase in automation that technology is ushering in will further create displacement in manufacturing jobs into the future.New challenges associated with transitioning to a green economy and the computerisation of routine tasks through artificial intelligence will precipitate similar changes and cause widespread disruption to labour markets (Autor et al., 2003;Goos & Manning, 2007;Manyika, 2012;Autor et al., 2015;Ford, 2015;Susskind & Susskind, 2015;World Economic Forum, 2016;Bessen, 2019;Levy, 2018;Acemoglu et al., 2022).Given ongoing disruptions to the workforce due to these and other structural shifts caused by emerging technologies, it will be important to manage these transitions through labour policy.The goal is to match labour force skills to emerging industries and ensure that individual workers are not too adversely impacted during the impending industrial transformation.Korinek and Stiglitz (2018) reiterate a growing literature on the political repercussions of dislocation, especially in the rise of populist political movements (Norris & Inglehart, 2019;Rodríguez-Pose, 2022).They argue that without labour market intervention to insure individuals against the detrimental effects of innovation, there will be 'adverse political and economic consequences' and 'resistance from those in society who are losing'.They recognise that structural shifts in economic activity will create 'significant job dislocation', and that 'markets on their own are not good at structural transformation', especially at times and in places where jobs are destroyed faster than they are created.
Many studies on job displacement have focused on the outcomes of displacement in relation to wage loss and lack of job security following displacement.The literature has provided ample evidence that displaced workers can expect pecuniary loss as they look for work and re-engage with the workforce, and that this wage loss can continue for years post displacement (Ruhm, 1991;Carrington, 1993;Jacobson et al., 1993;Carrington & Zaman, 1994;Stevens, 1997;Eliason & Storrie, 2006).In an effort to understand the heterogeneity of worker outcomes, the existing research on displaced workers tends to focus on the skill dimension: the mismatch of displaced workers' skills with those in demand in the labour market, or on the narrowing of skills that results from long tenure in now-obsolete occupations or on workers' personal characteristicsin particular, age (Fallick, 1996).
There has been less research on the potential mechanisms at play in relation to the preferences of displaced workers as they re-enter the workforce.This paper focuses on these preferences.It is informed by a broader, cross-disciplinary body of work that considers meaningfulness of work, job interest and job quality, which assist with understanding the mechanisms shaping preferences and therefore predict likely worker behaviour.In the past, the economic literature on displaced workers assumed that workers primarily want the highest rate of pay relative to effort (Roy, 1951), on the assumption that rational workers seek to maximise expected discounted lifetime earnings (Chapman, 1981, p. 240). 1 But this is an oversimplified notion of worker preferences.In 1987, Kovach compiled results from 40 years of surveys about what workers valued and found that 'good wages' consistently ranked fifth on the scale, and followed factors such as 'interesting work', 'full appreciation of work done' and 'feelings of being in on things'.More recently, a robust literature on gender-based disadvantages in the labour market has highlighted the importance of the non-pecuniary aspects of work (such as flexibility) and their value to women with caring responsibilities (Wiswall & Zafar, 2018).While the literature establishes the importance of non-monetary motivations, the crisis of job displacement and its emotional effects might bring different priorities (such as job security) to the fore or might stifle non-pecuniary motivations when the imperative is to find a new source of income.Therefore, accepting that factors other than pay rates are important, there is a need to identify which factors are valued in the crisis of job loss, and their importance relative to monetary rewards, in the context of displacement.
Research on meaningful work shows that workers value autonomy, competence and relatedness, and that meaningful work and the related concept of job interest can be used to predict outcomes such as absenteeism, retirement intentions, task and training performance, and organisational citizenship behaviour (Cassar & Meier, 2018;Nikolova & Cnossen, 2020;Asuyama, 2021).Job quality, which encompasses the pecuniary, social, safety, creative and other attributes of a job, can also be used to predict overall productivity and quality of output (Bryson et al., 2017).Despite this literature establishing strong links between worker preferences and economic outcomes, we are not aware of any previous studies examining displaced workers' employment preferences in a systematic way.How important are displaced workers' non-pecuniary preferences, in particular in relation to job quality and meaningful work?
This paper answers this question by gauging displaced workers' preferences in terms of compensating wage differentials.Its content is based on an empirical study employing stated preference (SP) experiments administered to a randomly selected group of workers who were retrenched during the closure of Australia's automotive industry.The closure created an unusually informative context for the study of non-pecuniary preferences because workers received generous redundancy payments that softened pecuniary imperatives, while simultaneously finding themselves with diminished opportunities in the types of jobs similar to their pre-retrenchment position (Irving et al., 2022).These circumstances forced workers to think about their work preferences.The SP experiment comprised a survey of 309 retrenched auto workers in 2020-21, 3 years after their retrenchment.The sample is generally consistent (across areas such as gender, age, education, industry sector and country of birth) with the wider group of auto workers recorded in the 2016 Census, and as such our findings can be taken to be broadly representative (for further detail, refer to Section IV).The survey focused on non-pecuniary job attributes, namely job security, skill utilisation, autonomy, supervision responsibilities and employer reputation for good work practices.Participants were asked about their attitudes towards previous, current and future employment; perceptions of assistance and training programs; and their labour market preferences.
This paper contributes to the existing literature on non-pecuniary preferences of displaced workers in three key ways.First, we offer new insights into the employment preferences of displaced workers, and compensating wage differentials for different non-pecuniary job attributes related to concepts of job meaningfulness, job interest and job quality.We show that autonomy and employer reputation for good work practices are the two most important non-pecuniary job attributes to workers.Drawing on the established reservation wage literature, our interpretation of the SP experiment results focuses on the compensation workers would require to accept jobs that deviated from their non-pecuniary preferences.Our results suggest that the average worker in our sample wants to work for an employer that offers some autonomy, and would expect the employer to pay a $5 per hour penalty for a job offering little autonomy.Workers also want to work for an employer with a reputation for good work policies and practices, and they value this preference by roughly the same amount ($5 per hour), suggesting that a firm with a poor reputation would have to pay $5 per hour more to attract these workers.In 2020-21, when the survey was conducted, the minimum wage in Australia was $20.33 per hour, meaning that $5 per hour is a significant amount, suggesting that workers place a high value on autonomy and quality work practices.Job security and skill utilisation are also found to be important, but less so than these attributes.The average worker preferred a permanent contract over a 1-year fixed term contract, and would expect to be paid $3.40 per hour more to compensate for lack of job security.The average worker preferred work that does not require pre-employment training and would expect $2.30 per hour more to take a job that required reskilling.We interpret this as a preference for work within the range of workers' existing skills sets.Supervision duties did not have a statistically significant effect on employment preferences.In order of importance, then, displaced workers show on average that they value autonomy, respect, security and career maintenance.
Second, our analysis finds some heterogeneity in these estimates.For example, compensating wage differentials for greater job autonomy or better employer reputation varied between $4 and $6 per hour for the majority of the respondents.Exploring the patterns of correlation between compensating wage differentials for different non-pecuniary job attributes and demographic variables reveals that, in general, women, older workers and those born outside Australia put a higher mean value on autonomy, employer reputation and job security than do men, younger workers, and those born in Australia.Additionally, the first group of workers also puts a higher value on improving their existing skills, and/ or undertaking pre-or post-employment training, if necessary, for a job.This information takes on significance when considered from the point of view of potential workplace discrimination.Workers in these groupswomen, older workers and those born outside Australiaare already subjected to greater workplace discrimination than men, younger workers and those born in Australia (Syed & Pio, 2010;Australian Human Rights Commission, 2015;Kosny et al., 2017;O'Loughlin et al., 2017;Daly et al., 2018;Gauci et al., 2022;Satyen & Becerra, 2022).Further, the characteristics that lead to discrimination are often combinedfor example, the Australian it is important that these workers are protected through policy to ensure that they are not exploited.Third, our findings hold important implications for labour and industrial policy.One of the enduring gaps in understanding labour market processes is the issue of occupational choice and its relationship to the willingness to take up training.As Blaug (1976, p. 837) opined: 'The human capital explanation of labour market training founders on the failure to provide a testable theory of occupational choice.Nothing is said about the non-pecuniary attractions of alternative occupations.'The challenge for labour market policy and for interventions related to worker displacement is to find the optimal match of available jobs to available workers.It has always been difficult to isolate the role of preferences in this equation because labour market outcomes reflect the intersection of workers' preferences and employers' preferences in the context of regulatory frameworks (Peck, 1996).Moreover, much of the existing literature focuses on and assumes the dominance of employer preferences, given that employers hold the decision-making power (Rubery & Wilkinson, 1994).Examining workers' preferences provides a means of resolving the problems of endogeneity that bedevil understandings of actual labour market outcomes.Workers' preferences for specific job attributes are likely to influence both their willingness to retrain and their 'reservation wage' decisions.Our findings show that workers' strongest preference is not for a particular type of work, but rather for a particular type of employer, suggesting that labour market policy might pay more attention to regulating the quality of workplaces.
The rest of the paper is structured as follows.Section II reviews previous studies that are relevant to the labour market preferences of retrenched and displaced workers.Section III presents a theoretical model of displaced worker preferences for employment.Section IV describes the data used for our analysis.Section V presents findings from our empirical analysis of the dataset.Section VI concludes with a discussion of our key findings.

II Labour Market Preferences of Retrenched
Workers Key concepts such as meaningful work, job interest and job quality are often considered when attempting to understand and predict worker behaviour.These concepts have robustly debated definitions and distinctions and are understood to be highly contextual.We consider them in the context of job displacement and worker preferences.Meaningful work has become an increasing focus for economists and researchers in other disciplines, including psychology, sociology, philosophy and health (Bailey & Madden, 2017;Laaser & Karlsson, 2021) principles of autonomy, competence and relatedness (Nikolova & Cnossen, 2020;Laaser & Karlsson, 2021).Job interest is another measure used across various disciplines and increasingly in economics.Job interest is closely connected to meaningfulness, with Krapp (2005) stating that interest is facilitated when needs for competence, autonomy and relatedness are met.However, job interest is also affected by pecuniary rewards, the perceived utility value of the task and the match between the worker's interests and the tasks involved in their work (Asuyama, 2021).Job quality refers to the characteristics of a job which meet workers' needs and encompasses pecuniary benefits as well as needs for 'social relationships at work, for creative purpose, meaningfulness and identity, for flexible time scheduling, and for a safe environment' (Green, 2021).Holman (2013) defines job quality in relation to psychological and physical well-being that fosters beneficial outcomes for the employee.
In terms of labour markets, these frameworks provide us with an opportunity to begin to measure and predict worker behaviour and related economic outcomes.Nikolova and Cnossen (2020) have demonstrated that 'meaningful work determines relevant economic outcomes, such as retirement, training, and absenteeism'.Asuyama (2021) showed that job interestingness is connected to 'positive predictors of task and training performance, organizational citizenship behavior, and job satisfaction, and negative predictors of labor turnover'.Bryson et al. (2017) showed that subjective job satisfaction, which is often discussed within the broader context of job quality, is connected to increased organisational performance, including financial performance, quality of goods and services and labour productivity.
Qualitative research on worker displacement has repeatedly found that displaced workers have preferences that are closely connected to the concepts discussed above of meaningfulness of work, job interest and job quality.For example, based on semi-structured interviews with 37 retrenched auto industry workers in Australia examining job quality and precarity, Barnes and Weller (2020) found that a majority of workers 'stated that their new jobs had inferior conditions compared to their previous careers in the auto industry.Many complained about a lack of job security . . .Such a fear of job loss forced many workers to accept inferior working conditions.'However, to the best of our knowledge, there has been no quantitative analysis of displaced workers' preferences for these and other related non-pecuniary job attributes.Quantitative studies have tended to focus on the measurement of impacts of job displacement in relation to wage loss.More research is needed to understand if non-pecuniary job attributes, such as autonomy, employer practices and reputation, skill utilisation and security, matter to displaced workers, and how much.
Traditionally, employee preferences for different job attributes have been measured using real-world labour market data.Studies have employed hedonic regressions (Rosen, 1974) where, for example, wage rates across different employees may be regressed against a number of individualspecific and job-specific factors to estimate compensating wage differentials for each factor of interest.However, as has been pointed out by, among others, Brown (1978), Hwang et al. (1998), Lang and Majumdar (2004) and Mas and Pallais (2017), hedonic wage regressions using cross-sectional data can suffer from multiple sources of bias and lead to problematic estimates of compensating wage differentials for non-wage job amenities.For example, Mas and Pallais (2017) argue that hedonic regressions using cross-sectional market wage data can frequently result in marginal estimates that are 'wrong-signed'.In particular, their own analysis of weekly market wages finds that 'more pleasant work arrangements are correlated with higher wages', even after controlling for individual-specific and job-specific factors, when the opposite should hold true.Hwang et al. (1998) and Lang and Majumdar (2004) show that the empirical cross-sectional relationship between wages and amenities can structurally deviate from actual worker preferences due to search frictions in the labour market.For example, Hwang et al. (1998) argue that 'firms with greater cost efficiencies in producing a job amenity will not only offer greater values of that amenity, they will offer job bundles that have an overall higher value.They will do this because their lower costs mean greater opportunity costs in having job vacancies go unfilled.If the estimation of hedonic wage equations does not control for differences in firms' cost efficiencies, then the error term in the hedonic wage equation will be positively correlated with the right-hand-side amenity variable.As a result, the associated MWP [marginal willingness-to-pay] estimates will be biased.Wiswall and Zafar (2018) make the related argument that compensating wage differentials for non-wage job amenities estimated using market data might be deflated due to search frictions, as workers might be willing to settle for less in order to end the search process as quickly as possible.However, what workers settle for in the short term is not a reflection of their long-term marginal willingness to pay for particular job amenities, and experimental data can offer better estimates of these long-term preferences.
Our study builds on previous work, such as Mas and Pallais (2017) and Wiswall and Zafar (2018), using SP experiments to measure worker preferences for non-wage job amenities.SP experiments are not subject to the same sources of bias as real-world market data.In fact, SP experiments offer considerable controleach of the variables affecting the decision is defined explicitly by the analyst, and there are no omitted variables and no measurement error (Mitchell et al., 1989;Louviere et al., 2000).Based on a comparison across different survey-based methods to estimate worker valuation of non-pecuniary job attributes, Feld et al. (2022) concluded that SP experiments perform the best.Notwithstanding these benefits, the use of SP experiments has been criticised by traditional economists based on concerns around incentive compatibility and hypothetical bias (e.g., Diamond & Hausman, 1994).It has been argued that due to the hypothetical nature of SP experiments, participants do not have any incentives to reveal their true preferences.The idealised environment of the experiment might encourage workers to separate their preferences from the practicalities of real life.When people prefer something that might not be in their best economic interest (e.g., wanting work autonomy, which is usually associated with higher-skill occupations involving decision-making capacities, yet rejecting the idea of training) this is attributed to imperfect information, to preferences based on false beliefs, or to irresolvable conflicts among competing preferences (Hausman & McPherson, 1996, pp. 76-7).Consequently, preferences measured using SP experiments are subject to hypothetical bias (e.g., Collins & Vossler, 2009;Fifer et al., 2014).However, some of these concerns can be addressed by making the experiment design as credible and realistic as possible, so that study participants can provide meaningful responses.The growing popularity of these experimental methods across different domains, including labour markets, attests to their robustness (Non et al., 2022).

III Model of Displaced Worker Preferences
for Employment Assume that displaced worker n is currently unemployed and looking for a job.Let w j denote the wages offered by job j, and x j be a vector of other non-pecuniary attributes of the job, such as autonomy, employer practices and reputation, skill utilisation and security.Further, let z n denote the vector of worker characteristics, such as age and education.The utility u nj derived by worker n from job j may be expressed as a function of each of these variables, and some stochastic component ε nj that captures the residual effect of unobserved variables: where we have assumed implicitly that utility is additively separable in the systematic and stochastic components.Assume further that the systematic component of utility is linear in parameters as follows: where α denotes the marginal utility associated with the wage rate, and β is a vector of parameters denoting sensitivity to the non-pecuniary job attributes.We specify β as a function of worker characteristics.For example, older workers might care more about job security, or be less willing to undertake training, than younger workers, or vice versa.From Equation (2), the marginal willingness to pay or compensating wage differential for the kth non-pecuniary job attribute x jk can be calculated as follows: Note that while we hold the marginal utility of the wage rate to be constant across workers, we allow the marginal utility for other non-pecuniary job attributes to vary across workers, and consequently, so do the compensating wage differentials.Implicitly, this is equivalent to estimating the model in willingness-to-pay space (Train & Weeks, 2005).
Assuming that all workers are seeking to maximise their utility, the preferred job d n for worker n is that which yields the maximum utility: Assuming that ε nj is independently and identically Gumbel distributed with location 0 and scale 1 across workers and jobs, the probability that worker n prefers job j can be given by the familiar multinomial logit expression (McFadden, 2001): Note that the function β z n ð Þ can be specified in different ways.The simplest is to assume that the marginal utility of different non-pecuniary job attributes is the same across workers, β z n ð Þ ¼ β.This yields a model that assumes homogeneous preferences across workers, with the following probability function: where α and β are the model parameters that need to be estimated.
Alternatively, we can assume that there is some stochasticity associated with these marginal utility estimates that cannot be captured by differences in observed worker characteristics, such that , where β is a random variable with a probability density function that is characterised by the parameter vector μ.This yields the familiar mixed multinomial logit model of McFadden and Train (2000) that can asymptotically approximate any model derived from the assumption of random utility maximisation.Usually, μ is assumed to be the same across different worker subpopulations, and β μ ð Þ characterises this distribution across the full population.This allows us to test for the presence of preference heterogeneity in the sample, and the resulting probability function is given by where α and μ are the model parameters that need to be estimated.Finally, we can assume that the marginal utility of different non-pecuniary job attributes varies systematically as a function of worker characteristics, such that This allows us to offer insights into the underlying sources of preference heterogeneity, and how they are correlated with different worker characteristics.The resulting probability function is given by where α and γ are the model parameters that need to be estimated.All models for this study were estimated using the software package PandasBiogeme (Bierlaire, 2023).

IV Dataset
We approached a total of 1,277 workers retrenched by the Australian automotive industry to take part in an online web-based survey conducted between December 2021 and February 2022.Of these, 309 elected to participate in our study, giving us a response rate of 24 per cent.As part of the survey, respondents were asked about their attitudes towards previous, current and future employment; perceptions of assistance and training programs; and their labour market preferences.The latter was measured using SP experiments, such as the one shown in Figure 1, where respondents were presented a hypothetical choice between two potential jobs that varied in terms of attributes such as degree of autonomy and type of contract, and asked to indicate their preferred job.Employed respondents were also shown the same attributes for their current job, as a baseline for comparison.Each respondent was presented eight such distinct scenarios.Job attributes were varied systematically across scenarios and respondents across the values listed in Table 1 based on an efficient experimental design.
To represent meaningful work across jobs and scenarios, the non-pecuniary job attributes of skill utilisation, supervision responsibilities, reputation for good work policies and practices, autonomy, working hours and employment contract were chosen.Successive to the methodological opportunities presented in Beer et al. (2019), the selection of these attributes was informed by prior in-depth qualitative work on the same retrenched population by Dinmore and Beer (2022) totalling a sample of 28 retrenched worker interviews conducted in 2021.Their discussions with retrenched workers on navigating the labour market post displacement and its associated precarity highlighted key issues around feelings of control and autonomy, quality of workplaces, and pathways to secure employment marked by permanent tenure and full-time working hours (Dinmore & Beer, 2022).Additionally, the inability to find work that utilised existing skillsincluding supervisory experiencewhich could advance current career paths was identified as a frustration of the interviewed sample (Dinmore & Beer, 2022).Table 2 summarises the sample distribution across key demographic and socioeconomic variables and compares this with 2016 Census data.Our sample comprises a mix of original equipment manufacturer (OEM) and supply chain workers.While the majority of workers in our sample are male and 40 years or older, we do have a sufficient number of female and younger workers to be able to test for differences in preferences as a function of age and gender.In terms of educational and technical qualifications, we have a good spread across different categories, and roughly one-third of our sample comprises workers who were born outside Australia.When comparing our sample with Census data, one can see that our sample is representative of the wider group of automotive manufacturing workers in Australia at the time of the survey, and that there are relatively insignificant differences between our sample and the wider population of workers.For example, our sample consisted of 84 per cent male and 16 per cent female workers, and the Census shows that workers in this industry were 83 per cent male and 17 per cent female.Similarly, the Census showed that 17 per cent of all workers in the industry had completed a degree or equivalent, and in our sample 19 per cent of all workers had obtained a degree or equivalent.Further, our sample contains enough diversity across these areas, plus household structure and language spoken at home, to test for statistical significance.As previously discussed, the heterogeneity found is not statistically significant, but has implications for policy development, especially in relation to workplace discrimination.
V Estimation Results Section V.i reports results from a multinomial logit model that assumes homogeneous preferences across workers; Section V.ii reports results from a mixed logit model that assumes worker preferences are heterogeneous and stochastic; and Section V.iii reports results from a multinomial logit model that assumes worker preferences are systematically heterogeneous.

(i) Model of Homogeneous Worker Preferences
We present estimation results from a multinomial logit model of homogeneous worker preferences in Table 3.The model has a McFadden's pseudo R 2 of 0.231, indicating a good fit.The table includes compensating wage differentials for different non-pecuniary job attributes, based on the formulae presented in Section III.The table also reports odds ratios for these attributes, that is, how much more or less likely are respondents to prefer a job with a given attribute, relative to a job without that attribute.Over subsequent paragraphs, we describe our estimation results in greater detail.
In terms of skill utilisation, we find that the average displaced worker in our sample values the opportunity to improve on existing skills, but is reluctant to undertake pre-employment training to learn a new set of skills.On average, respondents would be willing to take a job that does not require pre-employment training and be paid $2.3 less per hour, than take a job that requires pre-employment training.In general, respondents appear unwilling to seek employment in occupations that are too different from their job in the auto industry.For example, Table 4 summarises responses to statements measuring attitudes towards future employment opportunities.In most cases, the majority of respondents would prefer to work in their usual occupation, or one closely related to it, that makes use of their existing skills and abilities.This sentiment also seems to be consistent with responses to questions asking about willingness to work in specific occupations.As shown in Figure 2, the most popular occupations for employment post redundancy are those that used to exist within the auto industry, and on average, respondents are less willing to pursue occupations in fields that are far removed from the auto industry.
Most respondents prefer to have some supervision duties, but not too much.However, in both cases, the effect is small and statistically non-significant.In summary, supervision duties do not appear to be an  important determinant of job choice, and most respondents are relatively indifferent.
Reputation for good work policies and practices is found to be highly important.On average, respondents are roughly 1.4 times more likely to work for an employer with a medium or high reputation than one with a low reputation.Similarly, respondents are willing to be paid $4.5-$5.0less per hour to be able to work for an employer with a medium or high reputation, than one with a low reputation.Alternatively, employers with low reputations would have to pay workers $4.5-5.0 more per hour to be competitive with employers with a medium or high reputation.
Autonomy is also found to be highly important.On average, respondents are roughly 1.5 times more likely to work for an employer that offers some autonomy than one that offers none.Similarly, respondents would need to be paid roughly $5.0 more per hour to work for an employer that offers no autonomy, than one that offers some.
As one would expect, displaced workers in our sample exhibit a strong preference for more working hours, even if the hourly wages are lower.For example, the average respondent would prefer a full-time job that pays $10 per hour, than a part-time job with approximately 10 h of work per week that pays $30 per hour.
Respondents show a slight preference for permanent contracts over casual and fixed-term contracts, but not as strong as one would expect.For example, the average respondent would need to be paid roughly $3.4 more per hour to have a 1-year fixed-term contract compared to a permanent contract.Similarly, they would need to be paid roughly $1.4 more per hour to have a casual contract over a 1-year fixed term contract, but the latter estimate is not statistically significant.
The impact of commute time is statistically significant.The average respondent is willing to be paid $0.22 less per hour to shorten their one-way commute time by 1 min.For a full-time employee who commutes 5 days a week, this implies a value of time of roughly $50 per hour.For a part-time employee working 10 h a week and commuting 2 days a week, this implies a value of time of roughly $33 per hour.The average hourly wage rate for respondents in our sample, based on either their current or previous employment, is $32 per hour.Therefore, the estimated value of time is close to the wage rate, and this lends credibility to our estimation results.Finally, as we would expect, we find wage rate to be highly important and statistically significant.
In summary, we find that autonomy, and employer reputation for good work policies and practices, are the two most important non-pecuniary job attributes, with compensating wage differentials of roughly $5 per hour for greater autonomy and better employer reputation.Job security (as denoted by contract tenure) and skill utilisation are also important, but less so than the above attributes, with compensating wage differentials between $1 and $3 per hour.Finally, supervision duties do not seem to be an important attribute, showing up as nonsignificant in the model.

(ii) Model of Stochastic Heterogeneous Worker Preferences
In this section, we present findings from a mixed logit model that allows worker preferences for different non-pecuniary job attributes to vary randomly.Figure 3 plots the estimated marginal distributions of compensating hourly wage differentials for different non-pecuniary job attributes, based on the model specification, where we assumed that preferences vary normally.The normal distribution is by far the most popularly used mixing distribution in choice modelling (McFadden & Train, 2000).
There is some heterogeneity, but not significant.For example, compensating wage differentials for greater autonomy and better employer reputation vary between $4 and $6 per hour.In terms of supervision responsibilities and job security, preferences are homogeneous.In most cases, the range of variation appears to be small, suggesting that preferences are, by and large, homogeneous within our sample.This is somewhat surprising, given that our sample comprises a good mix of demographic and employment characteristics, including workers of different ages, genders, educational and technical training, cultural backgrounds, etc.Our findings suggest that similar experiences working in the auto industry prior to retrenchment might be overriding these personal differences, to the point that most displaced workers have similar preferences for different non-pecuniary job attributes, regardless of their own personal backgrounds and characteristics.
In terms of skill utilisation, the results are more interesting.Most workers in our sample would prefer a job that offers opportunities to improve on existing skills, and would have to be paid roughly $2-3 more per hour to accept a job that does not offer the opportunity.Conversely, most workers would prefer a job that does not require pre-employment training, and would have to be paid a similar amount to accept a job that requires such training.However, when it comes to post-employment training, the sample is split in terms of its preferences.Roughly half of the sample would prefer a job that does not require post-employment  training, while the other half would prefer to undergo post-employment training.For both groups, the compensating wage differentials are small.In the next section, we explore these patterns of heterogeneity further, by examining their correlation with different worker attributes.

(iii) Model of Systematic Heterogeneous Worker Preferences
We explore potential patterns of heterogeneity our sample further, this time by examining patterns of correlation between compensating wage differentials for different non-pecuniary job attributes on the one hand, and demographic variables on the other.Our analysis tested for the influence of key demographic variables, namely gender, age, education, country of birth, household composition, current employment status, housing tenure, and redundancy payout at the time of retrenchment.Figures 4-6 summarise our findings.
In general, we find that differences between different demographic subgroups are small and, in most cases, not statistically significant.It could be that our small sample sizes limited our ability to estimate statistically significant differences between different subgroups.Or, as indicated by our analysis in Section V.ii, it could suggest that displaced workers have similar preferences for different non-pecuniary job attributes, regardless of their own personal backgrounds and characteristics.
We find that women, older workers, workers with a university education, and those born outside Australia value autonomy, employer reputation and job security more than men, younger workers, workers without a university education and those born in Australia.Additionally, the first group of workers is also more willing to improve their existing skills, and undertake pre-or post-employment training, if necessary, for a job.Workers in these groupswomen, older workers and those born outside Australiaare subjected to greater workplace discrimination than men, younger workers and those born in Australia (Syed & Pio, 2010;Australian Human Rights Commission, 2015;Kosny et al., 2017;O'Loughlin et al., 2017;Daly et al., 2018;Gauci et al., 2022;Satyen & Becerra, 2022).Previous research with this population group has shown that being an older worker can enhance fear of precarity (Barnes & Weller, 2020;Irving et al., 2022), and being a skilled worker can reduce the chance of obtaining a suitable job (Dinmore & Beer, 2022).Coupled with our findings, this suggests that workers in these groups have, or perceive themselves to have, less power when bargaining with potential employers.Consequently, they value non-pecuniary job attributes such as autonomy, employer reputation and job security more; are more willing to learn new skills and branch into new and different occupations; and have a lower reservation wage.However, as shown in Figure 4, these differences are small and not statistically significant.Consequently, these findings should be interpreted with caution, and more research is needed to examine these issues in further detail.
In terms of household composition (Fig. 5), workers in households with children tend to have lower valuations for non-pecuniary job amenities, while workers in households with other adults tend to have higher valuations.In combination, these findings suggest that workers who have dependants, and no other adults to share responsibilities with, might not be in a position to seek better job quality attributes, and are more vulnerable to low-quality jobs, than workers without dependants who share their household with other adults.
Finally, we examine how preferences vary as a function of income and wealth-related attributes in Figure 6.We find that workers who are currently unemployed have higher valuations for non-pecuniary job amenities, and their current employment status might indicate that they are unwilling to settle for low-quality jobs.Workers who own their house outright value non-pecuniary job amenities more than workers who have a mortgage or rent their house, suggesting that greater financial security might allow workers in the former group to place a greater value on these attributes.Finally, we find the impact of redundancy payments at the time of retrenchment on preferences for non-pecuniary job amenities to be small and non-significant.Ruhm, 1991;Carrington, 1993;Jacobson et al., 1993;Carrington & Zaman, 1994;Stevens, 1997;Eliason & Storrie, 2006).There has been less research on the potential mechanisms at play in relation to the preferences of displaced workers as they re-enter the workforce.
Just as research in this field has been focused on wage loss and job security, the attention of governments has been similarly circumscribed, concentrating on the dual goals of buttressing industry and creating circumstances conducive to labour branching for workers.Looking globally at how governments have intervened, a pattern emerges of financial aid packages designed to prop up struggling industry.Economically advanced countries such as Germany and France provided tax incentives, green incentives, and finance packages to ensure the industry could survive significant disruptions such as the Covid-19 pandemic.In the United States, the Obama administration invested over $80 billion in loans via its Automotive Industry Financing Program to allow the industry to recover from the brink of collapse following the Global Financial Crisis (US Department of the Treasury, n.d.).Within the context of industry shutdowns in Australia, successive governments have directed support towards tariff assistance, competition-enhancing programs, economic growth in hard-hit regions, and grants to support traditional industries moving into growth or niche areas of manufacturing (Australian Government, 2023aGovernment, , 2023b;;Parliament of Australia, n.d.)On the worker side, programs in the USA and Australia have been primarily focused on moving redundant workers into new forms of employment.This support has included retraining, employment subsidies, relocation payments, employment services, financial aid for uniforms or childcare, income support, and cash payments for equipment (Evans-Klock et al., 1998;Neering & Davis, 2023;Australian Government, 2023b).In an extensive study, the International Labour Organisation looked at measures taken in Germany, Japan, France, Italy and the United States and concluded that measures generally fell into  So, while much has been done to support manufacturing in developed nations globally, very little effort has gone into understanding the preferences of workers who have been made redundant through industry closure, or into developing policies that support those preferences.We argue that the importance of job quality and meaningful work has been consistently overlooked within this landscape.Job quality and meaningful work provide benefits that include loyalty, efficiency, productivity, satisfaction, mental and physical well-being, and monetary reward.Further, we note that within the automotive manufacturing industry, the principle of keiretsu, entailing close collaborative relationships that value trust, care and support, is understood to have contributed to the success of companies such as Toyota.Our research supports the idea that workers value the nature of keiretsu relationships, in which there is long-term benefit for both parties, and a degree of mutual respect that is not always found in other workplaces.
Despite the lack of policy attention paid to supporting redundant workers' preferences, there has been a sizeable volume of research conducted on job satisfaction, and as would be expected of a large body of research, various job attributes have been studied for their impact on job satisfaction.For example, Hasegawa and Ueda (2016) looked at the impacts of gender and marriage, Mishra (2013) considered the quality and scope of training, management capability and teamwork, Badillo-Amador and Vila (2013) and Vila and García-Mora (2005) examined education level and skill mismatch, and Bryson et al. (2004) looked at the impact of union membership.However, while there is no clear consensus on precisely which non-pecuniary job attributes have the most significant impact on job satisfaction, substantial evidence exists to support our findings that autonomy (Davis & Bordieri, 1988;Finn, 2001;Benz & Frey, 2004;Lopes et al., 2014), employer reputation (Alniacik et al., 2011;Tziner et al., 2011;Helm, 2013), working hours and job security (Bryson et al., 2005;Sypniewska, 2014;Raziq & Maulabakhsh, 2015;Wilczy ńska et al., 2016) play an important role in job satisfaction.To reach our findings, we examined labour market preferences of 309 workers retrenched by the Australian automotive industry and informed our approach with a broader, cross-disciplinary body of work that considers meaningfulness of work, job interest and job quality.We utilised SP experiments to measure worker preferences for different non-pecuniary job attributes, namely skill utilisation, supervision responsibilities, employer reputation for good work policies and practices, job autonomy, working hours and job security.The results of our research provide a stunningly simple but elegantly profound insight: above all, workers want to be respected.Therefore, it is not the work that matters, it is the way people are treated at work; and workers' strongest preference is not for a particular type of work, but rather for a particular type of employer.
Our study offers new insights into the employment preferences of displaced workers and compensating wage differentials for different non-pecuniary job attributes related to concepts of job meaningfulness, job interest and job quality.We found that autonomy and employer reputation for good work policies and practices are the two most important nonpecuniary job attributes, with compensating wage differentials of roughly $5 per hour for greater autonomy and better employer reputation.Job security (as denoted by contract tenure) and skill utilisation are also important, but less so than the above attributes, with compensating wage differentials between $1 and $3 per hour.Finally, supervision duties do not seem to be an important attribute, showing up as nonsignificant in the model.There is some heterogeneity across our sample, albeit not significant.We found that women, older workers, workers with a university education, and those born outside Australia value these non-pecuniary attributes more than men, younger workers, and those born in Australia, but these differences are small and, in most cases, not statistically significant.
A key limitation to our methodology is the use of SP experiments to measure worker preferences.While there is considerable empirical evidence to indicate that preferences estimated using SP experiments are credible and reliable (e.g., Mas & Pallais, 2017;Wiswall & Zafar, 2018: Feld et al., 2022), some studies have criticised their use based on concerns around incentive compatibility and hypothetical bias (e.g., Diamond & Hausman, 1994).It has been argued that due to the hypothetical nature of SP experiments, participants do not have any incentives to reveal their true preferences.Future research could attempt to estimate preferences for similar non-pecuniary job attributes using real-world labour market data.For the purposes of this study, we examine how our estimates compare with estimates from other related studies.In general, international research on compensating wage differentials for non-pecuniary factors supports our findings.For example, Poggi (2007, p. 730) found 'a positive and statistically significant wage differential associated with satisfactory non-pecuniary working conditions' across all Mediterranean countries; French and Dunlap (1998) found a wage differential of between 3 per cent and 10 per cent for mental stress using a large sample that included manufacturing worksites; and Makridis (2017) found a compensating wage differential of 2 per cent applied to workplaces with higher organisational practices (using a scale that included satisfaction with employer, appreciation and managerial relationship).In Australia, there is a notable gap in research pertaining to compensating wage differentials for job satisfaction; with the exception of Vij et al.'s (2023) findings in relation to work from home compensating wage differentials, we were not able to find research that considered compensating wage differentials for non-pecuniary job attributes in Australia.Therefore, our findings that workers value autonomy and employer reputation for good work policies and practices are consistent with the international literature review and represent a contribution to new knowledge in the Australian context.Our conclusion that the quality of the workplace matters more to workers than the type of work would suggest that labour market policy, whether it relates to retrenched workers or otherwise, might benefit from paying more attention to regulating the quality of the workplace.Ensuring that the future workplaces of retrenched workers emulate the respectful work practices that were embedded at workplaces such as Toyota Australia, and where individuals were supported to thrive and know that they are contributing to their own goals as well as those of the organisation, may result in retrenched workers being more satisfied in their new roles, regardless of the industry they move into.
These findings can be applied to policy development, with an awareness of individual contexts, in the wider landscape of manufacturing decline and increased automation of unskilled and low-skilled tasks.Australia and other developed economies have experienced significant declines in manufacturing employment, but future shocks to the labour force are expected, with some concluding that hundreds of thousands of jobs will be lost as Australia decarbonises its economy (Centre for Policy Development, 2022).There will be an ongoing need to support large groups of workers experiencing redundancy in developed economies, but also, as automation of manual tasks increases, the manufacturing jobs that have migrated to developing nations will also disappear, leaving these workforces facing similar shocks, albeit in different economic and social climates.
The heterogeneity among respondents demonstrates that policy responses must pay attention to individual needs rather than seek to apply a universal solution.These two contextual concerns combinedthe nature of the industry and the economic and social circumstances in which it exists, plus the heterogeneous nature of individual workers' needsimplies that blunt policy instruments are unlikely to resolve the problem of displaced workers' dissatisfaction.Therefore, while these findings about workers valuing autonomy and reputation, placed in the context of the large body of related work on job quality and meaningful work, create a robust picture of the value of respectful workplaces, policy must carefully address the diverse needs of workers.
Further, while most structural adjustment programs have focused on questions such as how workers should be retrained and how to maximise workers' existing skill set, our findings suggest that these considerations are of secondary importance to the workforce, whose primary concern is the quality of the workplace.Workers value how they are treated at work, and workers in the Australian automotive manufacturing industry have reported being well treated in their workplaces.Therefore, re-employing displaced workers, the focus of structural adjustment programs needs to broaden.Rather than implementing programs which match unemployed workers with jobs based on their existing skills, or based on skills shortages within labour markets, our research indicates that these programs should have a primary focus on the nonpecuniary considerations that matter most to the individual worker, such as degree of autonomy and workplace reputation.

Ó
2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

FIGURE 1
FIGURE 1Example SP Experiment to Elicit Labour Market Preferences.

FIGURE 2
FIGURE 2Occupations that Redundant Workers Would be Interested in Undertaking.

Ó
2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

FIGURE 3
FIGURE 3 Probability Density Functions Showing Distribution of Compensating Hourly Wage ($) Differentials for Different Non-pecuniary Job Attributes across the Sample.

Ó
2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

FIGURE 4
FIGURE 4 Estimated Mean and Standard Deviation of Compensating Hourly Wage ($) Differentials for Different Non-pecuniary Job Attributes as a Function of Gender, Age, Education and Birth Country.

FIGURE 5
FIGURE 5 Estimated Mean and Standard Deviation of Compensating Hourly Wage ($) Differentials for Different Non-pecuniary Job Attributes as a Function of Household Composition.

FIGURE 6
FIGURE 6 Estimated Mean and Standard Deviation of Compensating Hourly Wage ($) Differentials for Different Non-pecuniary Job Attributes as a Function of Current Employment Status, Home Ownership Tenure, and Redundancy Payouts at Time of Retrenchment.

Ó
2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

TABLE 1
Range of Attribute Values Used in our SP Experiments to Describe Different Working Arrangements across Different Scenarios Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1475-4932.12797by National Health And Medical Research Council, Wiley Online Library on [27/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Ó 2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

TABLE 2
Summary Statistics for Australian Automotive Manufacturing Workers (Recorded in the 2016 Census) Compared to Sample Note: N/A indicates that there was no direct comparison available between the sample and publicly available Census data.†Information was missing from some respondents in these categories.The missing values have been ignored for the purpose of calculating sample distributions.Ó 2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.

TABLE 3
Estimation Results from Multinomial Logit Model of Homogeneous Worker Preferences Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1475-4932.12797by National Health And Medical Research Council, Wiley Online Library on [27/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License †Odds ratio and compensating hourly wage differentials calculated with respect to the reference category for categorical variables.Ó 2024 The Authors.

TABLE 4
Summary Statistics for Responses to Different Attitudinal Statements Ó 2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1475-4932.12797by National Health And Medical Research Council, Wiley Online Library on [27/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Ó 2024 The Authors.Economic Record published by John Wiley & Sons Australia, Ltd on behalf of Economic Society of Australia.