The Dynamics of Satisfaction with Working Hours in Australia: The Usefulness of Panel Data in Evaluating the Case for Policy Intervention

Prevalence

Studies based upon the HILDA data consistently find that more Australians prefer fewer hours rather than more hours. 3 When examining other data sources, the prevalence of working hour mismatches seen in the HILDA data is largely supported. The exception to this is the Australian Survey of Social Attitudes (AuSSA), which reported underemployment as more prevalent than overemployment. 4 However, the AuSSA survey uses phrasing of the hour mismatch question that when used in other countries often reveals a larger underemployed population. 5

The demographic composition of the underemployed and overemployed is an area that has been well examined. Prime working-age Australians with working hour preference mismatches are more likely to be overemployed and less likely to be underemployed than people aged under 25. Likewise, university education is associated with higher rates of overemployment and lower rates of underemployment. These trends are found in both the HILDA dataset (Tseng & Wooden 2005; Wilkins 2006) and ABS data (ABS 2010, 2012). Tseng and Wooden (2005) also show that migrants from non-English-speaking backgrounds are more likely to desire more hours than their current arrangement and less likely to desire fewer hours. Partnered workers are more likely to be overemployed (ABS 2011), and the HILDA data show that people are likely to desire more (less) hours than their current hours if their partner also would prefer to work more (less) hours (Tseng & Wooden 2005). Reynolds and Aletraris (2007) construct a work–family conflict variable and associate it with an increased desire for fewer hours for both men and women, with the effects stronger for women if they have a young child. They also find that household income significantly impacts male hour mismatches in both directions, but not females in a couple.

Notable job characteristics that are correlated with hour mismatch include employment type and industry and occupation. Casual workers and the self-employed are more likely to be underemployed, with strong industry and occupation effects (Wilkins 2006; ABS 2010; Fear et al. 2010). It is uncertain what the causal factors for hour mismatch are as industry and occupation may influence the probability of being a salaried worker and other work conditions.

Due to the subjectivity of hour preference survey questions, causal relationships between satisfaction-based variables can be difficult to identify. For example, one may dislike his or her work environment and therefore desire less hours, making it difficult to ascertain which factor leads to lower job satisfaction. Tseng and Wooden (2005) show that people who enjoy having a job are unsurprisingly less likely to want fewer hours and more likely to want more hours. Similarly, Reynolds and Aletraris (2007) associate stress with a desire for fewer hours for both men and women in couples.

Labour demand factors are not present in the HILDA survey data and may be useful for hour mismatch analysis. Baum and Mitchell (2008) add some regional labour market information to HILDA Wave 1 in their analysis of 15- to 24-year-old Australians and find that higher proportions of low income jobs in a region increase the likelihood that the region's youth are underemployed. An area that cannot be explored using the HILDA dataset is the relationship between hour mismatch and employer/firm characteristics. Doiron (2003) used data from the 1995 Australian Workplace Industrial Relations Survey to match firm and worker data, finding that contracting firms do not hoard labour as they are less likely to have underemployed workers than expanding firms. Furthermore, non-contracting firms show a negative job tenure/probability of underemployment relationship for their workers. This suggests that changes in the characteristics of individual firms may influence hour mismatches of workers.

Severity

Similar to prevalence statistics, HILDA data consistently exhibit an average preference for less hours of work for both males and females. Appendix 1 Table A1 shows that where information on severity is produced from other Australian surveys, there is an average preference for less hours of work across the sample.

ABS categorises the most severely overemployed as workers who work 60 or more hours per week. Similarly, the most severely underemployed are workers of low hours, with the underemployed male workers wanting more additional hours than the underemployed females (ABS 2010, 2011). These trends are supported by Fear et al. (2010).

Tseng and Wooden (2005) measure the extent of both underemployment and overemployment in couple households by performing a conditional ordinary least squares regression on the hour mismatch for both groups. Of the overemployed, the average mismatch is greater for the self-employed and when the partner is overemployed, and the average mismatch is lower with a recent unemployment spell and if their partner is employed part-time. There are also significant age effects with prime age workers the most susceptible to overemployment. Of the underemployed, the mismatch is greater for casual, self-employed and non-English-speaking background workers, and the mismatch is lower for people with an overemployed partner.

Persistence

The longitudinal Negotiating the Life Course (NLC) survey (van Wanrooy 2004, 2005; McDonald et al. 2009) supports our results below by showing that overemployment is a more persistent state in the NLC data than underemployment.

Reynolds and Aletraris (2006) use the first two waves of HILDA data to show that a desire for more hours is more likely to be resolved than a desire for less hours between surveys. They also show that in creating and resolving hour mismatches, both preferred and actual hours adjust in most cases, with preferred hours often moving the most for resolving or desiring fewer hours. This suggests that people may acclimatise to their working situations. They find that men and women who change both their actual or preferred hours are more likely to resolve a mismatch than people who do not, and that men are more likely to resolve a desire for more hours than fewer hours.

Reynolds and Aletraris (2006) also test the effect of a range of job-specific, personal and family characteristics on creating and resolving hour mismatches. The strongest relationship found is an increased likelihood of resolving mismatches if people change their job. For females, lower household hours of work are associated with a lower chance of resolution, with younger and older workers more likely than the middle-aged to resolve a mismatch. For males, the presence of another earner in their household and feeling secure in their job improve the odds of resolving mismatches, and like females higher hours of work are associated with a decline in the odds. In creating hour mismatches, changes in preferred hours are positively associated with hour mismatches, whereas changes in actual hours are negatively related. People on higher incomes are less likely to develop a preference for more hours, and male professionals and managers are more likely to develop overemployment, likewise university-educated females. Women who have a teenage youngest child are more likely to create an unmet desire for more hours, while women who believe working harms their relationship with their child are more likely to develop overemployment.

In an article that is strongly related to ours, Wooden and Drago (2007) analyse the first five waves of the HILDA data asking the same set of questions that we do. They begin by looking at working hour preferences by sex and usual working hours. The patterns that they report in the first five waves are, unsurprisingly, very similar to what we report. They also look at the magnitude of hours mismatches by sex, and again our numbers for the early waves are similar to theirs. They examine persistence across one wave and multiple waves. Taking matched pairs of people employed at both time periods, they examine the reported hour mismatches at time t compared with those at time t + l for people with mismatches at time t, allowing l to take values from one through four. In this article, we primarily look at persistence over one time period, although we briefly discuss longer time horizons as well. Wooden and Drago (2007) do not examine the role of job changes nor do they look at what changes when hour mismatches are resolved, as we do below. However, in addition to examining longer time lags, they undertake some analyses that we do not. They look at annual hours versus weekly hours and the relationship between paid leave entitlements and whether or not they are taken up. They find no evidence that those working long hours are less likely to take up annual leave. They also provide some international comparisons.

When restricting the analysis sample to Australians with high working hours, Drago et al. (2009) reveals persistent high-hour overemployment for university-educated managers and professionals, fathers with a partner who works, and people with high debt to income ratios. Over time, fathers appear to accept their long hours by changing their preferences to match their long working hour situation.

Working Hour Mismatch and the Business Cycle

Analysis of the macro-dynamics of working hour mismatch tends to be limited to underemployment rather than all hour mismatches. The ABS (2010) and Campbell (2008) provide descriptions of the underemployment measures used by the ABS and show that underemployment varies less over the business cycle than unemployment. Rather, there appears to be an increase in the aggregate rate of underemployment at the beginning of an economic downturn with the new level maintained longer as the business cycle returns back to growth. Mitchell and Carlson (2000) use the ABS data to construct indexes of labour force underutilisation using the difference between reported and preferred working hours, and show that these indexes may evolve differently from the unemployment rate. The percentage of part-time work may increase at some points in the business cycle, and the Centre of Full Employment and Equity (2011) discussed the relationship between part-time work and mismatch between preferred and actual working hours.

We extend the above literature by looking at all three dimensions of mismatch between actual and preferred working hours across an 11-year span. In the next section, we describe our data before considering the three dimensions in detail.

Persistence: Complexity of the Dynamics

We next examine in more detail the dynamics of those who resolve a mismatch (who go from wanting to work more or less at t − 1 to being content with their hours at time t) and those who develop a mismatch (who report satisfaction with hours at t − 1 and then report dissatisfaction of either type at time t). We begin by looking at the question of what adjusts. Is it working hours that change or is it preferred hours that change. Recall that working hour satisfaction is only one aspect of overall job satisfaction and that it may depend upon other aspects of the job. Suppose that one goes from having an unpleasant boss to a pleasant boss without changing jobs or actual working hours. In this case, preferred hours may adjust, and the dissatisfaction with working hours may be resolved without changing actual hours worked.

Table 8 examines the average change in hours for those who develop a mismatch (rows one and two) and those who resolve a mismatch (rows three and four). Men are considered in columns 2 through 4 and women in the last three columns. For those who desire more hours at time t − 1 and resolve this mismatch in time t, actual hours worked increase by much more (twice as much for men; four times as much for women) than preferred hours change. So this group appears to resolve a mismatch by working more. For all other groups, we see that the adjustment of preferred hours is larger than the change in actual hours. Thus, a desire for fewer hours is generally resolved by a larger change in preferred hours than in actual hours worked. Likewise, the development of either type of mismatch is driven more by changes in preferred hours than changes in actual hours.

Averages can mislead. Tables 9 and 10 are an attempt to demonstrate the richness of patterns that are actually observed in the data for men and women, respectively. In columns 2 and 3, we examine the patterns of changes in actual hours and preferred hours for those who develop an hours mismatch. In columns 4 and 5, we examine the same patterns for those who resolve a mismatch. What we observe is that preferred and actual hours often move together, something not readily visible from Table 8. What we can also observe is that most mismatches are either resolved or created by preferred and actual hours both moving together rather than by just one or the other changing.

For example, in Table 10, we can see that for women a desire for more hours is resolved by a change only in preferred hours in 14.9 per cent of cases, by a change only in actual hours in 12.9 per cent of cases and by a change in both in the vast majority (72.2 per cent) of cases. Simple intuition might suggest that a desire to work more hours is resolved by working more hours, but at least in its simplest form this is the least likely way in which such a mismatch is resolved.

Figures 3 and 4 (for men) and Figures 5 and 6 (for women) further elaborate on this complexity. In these figures, we present non-parametric density estimates of the changes in actual and preferred hours for the four groups considered in Tables 8-10. We again can see the wide variety of experiences documented in Tables 9 and 10, where a change in both preferred and actual hours for all four groups is the norm.

Figure 7 (for men) and Figure 8 (for women) examine the difference in actual hours at time t and preferred hours at time t − 1 for the four groups that either resolve or develop a mismatch. This difference is largest (in positive terms) for those who resolve a desire for fewer hours. It takes on the largest negative values for those who resolve a desire for more hours. Both groups that develop mismatches are more centred around zero, indicating a smaller shift from past preferred hours to current actual hours.

Across all six figures, the differences between men and women are again very small.

Persistence: The Role of Changing Employer

As mentioned above, persistent mismatch may be viewed as indicating some type of rigidity in the labour market. If people can freely change jobs to improve their working conditions and move from a situation of dissatisfaction to one of satisfaction, then this would be a strong argument against public policy intervention.

In this subsection, we reexamine the question of transitions across wave between different states of satisfaction and dissatisfaction with working hours. We constrain ourselves to those who are employed in both periods, and we separately analyse those who change employers/jobs and those who stay with the same employer/job (as defined in Section 3.4).

Tables 11 and 12 are laid out in the same fashion as Tables 6 and 7. Looking at full-time employed males, we can see that changing employers between periods greatly facilitates resolving a desire to work less. Of full-time employed males who wished to work less, 65.5 per cent still felt that way a year later if they were with the same employer. Among those who changed employers, only 42.6 per cent now felt this way. Meanwhile, 32.2 per cent resolved a desire to work less hours without changing employer, whereas this number jumps to 46.9 per cent for those who changed employer.

The result is similarly dramatic if we look at part-time males who wish to work more. Of these, 52.2 per cent reported the same problem if they stayed with the same employer, whereas the desire to work more was only reported for 33.5 per cent who changed employer. Of those who changed employer, 55 per cent resolved the desire to work more hours, whereas only 40.8 resolved the problem without changing employer.

For women, the results are similar. Full-time employees resolved a desire to work less 46.5 per cent of the time when changing employer and only 31.6 per cent of the time if they stayed with the same employer. Part-time women resolved a desire to work more 52.7 per cent of the time when changing employers, and only 43.8 per cent of the time when not changing employers.