6. Factors influencing the pay and training rates of temporary workers

6.1 Introduction

Section 4 compared the employment outcomes of temporary and permanent employees using a range of different measures, and found disparities on some measures. Among those differences, it is notable that temporary employees earned significantly less per hour on average, and were substantially less likely to have undertaken employer-funded training in the last year.

These outcome differences are not necessarily caused by the nature of the job contract or the non-permanence of the employment relationship, because temporary workers differ from permanent workers on a range of other characteristics, such as age, work experience, education, and hours of work. This section of the paper discusses the factors that may be contributing to the wage rate and training rate differentials that are found between temporary and permanent employees. Regression methods are then used to estimate the size and significance of the wage and training differentials after taking into account additional information on measured personal and job characteristics.

Section 6.2 draws on the economic literature to review the reasons why temporary workers might earn less than permanent workers and undertake less training. Section 6.3 presents an analysis of the gap in average hourly earnings, and section 6.4 examines the gap in training rates.

6.2 Theory and research evidence

In a competitive labour market, in which workers have a choice about what form of employment contract they enter into and most workers prefer permanent jobs, temporary workers could be expected to receive a wage premium to compensate for the fact that their job is not permanent (Booth et al, 2002). The theory of compensating differentials suggests that temporary employees should earn more than permanent employees.

There are a number of reasons why a wage premium might not be paid, or why the wage premium that is paid to compensate temporary workers for the non-permanency of their job is offset by other factors, leading to lower observed wages. One reason is that employers with high quality jobs could use temporary employment contracts as a screening devise, requiring employees to work on a temporary basis for an initial period, at a lower rate of pay, before being hired on a permanent basis. If the likelihood of promotion to more attractive permanent positions is relatively high, these temporary jobs may readily filled by workers despite lower pay rates (Booth et al, 2002).

Another reason why the wages of temporary worker might be lower is that it is not efficient for workers in temporary jobs to invest heavily in job-specific skills or human capital, or for firms to provide them with large amounts of training. Employees in temporary jobs are less likely to receive training that is funded by employers than permanent employees, because the post-training period in which benefits can be gained (through improvements in employee performance) is usually shorter, making the investment less attractive. Temporary employees may also be less likely to agree to undertake training because they are less likely to gain future benefits through higher wages or promotions. Lower job-specific skills could to lead to lower wages being paid to employees in temporary positions.

If temporary jobs tend to offer lower wages and less training, this could have implications for the typical characteristics of the workers who hold temporary jobs (Booth et al, 2002). The workers holding these jobs may be the ones for whom there is a greater probability of wanting to leave (in order to change jobs, leave the labour market, or migrate), or those who face a higher cost (or lower benefit) to acquiring specific human capital. For example, young workers and new migrants might be disinclined to make a large investment in a particular job if they are unsure of their future employment and locational preferences, leading to an over-representation of these groups in temporary jobs. Older workers who are near to retirement might also be less inclined in invest in specific skills, and therefore be over-represented in temporary jobs.

These theoretical ideas imply that whether temporary jobs offer higher or lower wage rates, and are filled by high ability or low ability workers, will depend on the circumstances. For example, if firms maintain a pool of temporary workers in order to adjust staffing levels to fluctuations in activity, and do not promote these workers to better jobs, they may be filled by workers who have low ability or low desire to acquire specific skills. These workers will go through a succession of low-paid temporary jobs. If firms use temporary jobs as a screening device and the likelihood of promotion to more attractive permanent positions is high, then those temporary jobs could potentially be filled with high ability workers. In other circumstances, if a worker's productivity depends on general and transferable skills rather than skills that are specific to a job or firm, then highly productive and high ability workers may be in a position to negotiate wage premiums in temporary jobs, and may actually prefer to be employed in a series of temporary jobs (Booth et al, 2002).

A number of overseas studies have estimated the wage penalty or premium that is associated with temporary work. Typically they use models to adjust the 'raw' or observed average wage gap for differences between temporary and permanent workers in skill-related characteristics, with the aim of better estimating the wage differential that is due to the temporary nature of the job contract. Dekker (2001) finds evidence of significant wage penalties for temporary workers in Germany, the United Kingdom, and the Netherlands, in wage regressions estimated using national longitudinal data that included controls for gender, age, marital status, qualifications, industry, size of firm, tenure in the current job, and unobserved individual heterogeneity. Houseman (1997, p29) estimated that US workers employed on a short-term basis earned 10-16 percent less per hour than permanent employees, after controlling for a range of personal, job and workplace characteristics. Workers employed through temporary employment agencies earned 16-19 percent less, and on-call workers earned 5-10 percent less. The OECD (2002, p143) estimated the independent effect of holding a temporary job on hourly earnings using data for 13 OECD countries. It reports that standardising for worker and job characteristics reduces the wage penalty associated with holding a temporary job, but does not eliminate it.

Although many of the published overseas studies have found that temporary workers earn less per hour than permanent workers even after adjustments have been made for differences in other characteristics, this is not always the case. Bono and Weber (2008) studied the earnings of seasonal employees in Austria. Although seasonal workers earn 3 percent less than non-seasonal workers on average, Bono and Weber estimate that an 11 percent wage premium is actually paid for seasonal work, after taking other factors into account.

Studies that have compared the workplace or job-related training rates of permanent and temporary employees have generally found an association between temporary work and lower training (Long et al 2000, p42; OECD 1999; Arulampalam and Booth, 1998). Arulampalam and Booth (1998) find that both measured and unmeasured individual characteristics contribute to the observed training gap between permanent and temporary employees in the United Kingdom, but do not full explain it, suggesting that temporary job contracts do in fact lead to a lower level of training. The OECD (2002, p.158) estimated the relationship between temporary employment and participation in job-related training using data for 12 OECD countries. It found evidence of a lower rate of training in eight countries, but no significant difference between the training probabilities of temporary and permanent workers in the other four countries.

The existing studies vary in the quality of the data used and the extent to which other relevant factors are controlled for. The ideal dataset for identifying whether temporary jobs offer wage premiums or wage penalties, or differing opportunities for training, is a longitudinal dataset in which individuals move between temporary and permanent jobs, because this allows the effects of movements between temporary and permanent positions on the outcome of interest (wages or training) to be isolated from the effects of individual heterogeneity in characteristics, preferences and behaviour. Estimates obtained from cross-sectional data sets, which are not able to control for the effects of unmeasured individual heterogeneity, may be biased.

6.3 Hourly earnings

In the March 2008 quarter the raw wage differential between temporary and permanent employees in New Zealand was relatively large. The average hourly earnings of temporary employees were $18.50, 79 percent of the average hourly earnings of permanent employees ($23.40 percent). As explained above, a simple comparison of the average temporary worker wage with the average permanent worker wage is unlikely to reveal the true wage effects that are associated with entering a temporary job, because temporary employees differ from the norm on a number of dimensions.

We explore the contribution of demographic, educational, and job characteristics to the temporary-permanent gap in average hourly earnings in New Zealand using the SoWL data. The sample used for this analysis was all employees who supplied valid data on their average hourly earnings, 91 percent of the total. The earnings of males and females were modelled separately. The log of hourly earnings was modelled as a function of personal and job characteristics X_i, whether employed in a temporary job T_i, and an individual-specific error term ε_i:

Ln W_i = X_iβ + T_iδ + ε_i

In a first set of regressions we included age, ethnicity, parental status, immigrant status, geographical location, and highest educational qualification in X_i, the vector of individual attributes. In a second set of specifications the following job characteristics were also included: part-time status, occupation defined at one-digit level, industry defined at one-digit level, the employer's business type (using indicator variables for public sector and non-profit organisations), and the size of the employer's enterprise. We do not control for variations in job tenure because this is likely to be determined jointly with temporary job status.

The way in which temporary employees are identified (T_i ) varies across equations. Initially we identify all temporary workers using a single dummy variable that is set to one for all temporary workers and to zero otherwise. In a second set of equations, we use four dummies which separately identify each type of temporary worker (casual, fixed-term, temporary employment agency and seasonal not otherwise classified). In a third set, we use a pair of dummy variables that identify both seasonal workers and temporary non-seasonal workers, distinguishing them from permanent workers.

Table 15 shows the coefficient estimates and standard errors that were obtained for the temporary employment dummy variables, in each of the different specifications.^[15] The coefficient estimates represent the approximate percentage difference between the earnings of temporary and permanent employees. If negative, they represent the wage penalty that is associated with working in a temporary job, holding the other factors constant. If positive, they represent a wage premium.

Table 15: Estimates of the gap in average hourly earnings between temporary and permanent employees

The first column of the table shows the unadjusted or 'raw' log wage gap between temporary workers (or a particular subgroup of temporaries) and permanent employees. The second column gives the estimated log wage gap once the effects of differences in personal characteristics are controlled for. Controlling for personal characteristics dramatically reduces the size of the temporary-permanent gap in log hourly earnings. For instance, there is a 64 percent reduction in the wage gap estimated for all male temporary workers (which declines from a 30 percent penalty to approximately 11 percent), and a 44 percent reduction in the wage gap estimated for all female temporaries (which declines from 13 percent to 7 percent). Temporary employees are younger and less qualified than permanent employees, on average, and controlling for these and other differences reduces the wage 'penalty' that is assigned to the temporary work indicators in the regression estimates.

Controlling for job characteristics as well as personal characteristics further reduces the estimated wage penalties associated with temporary work. For all temporary and for most subgroups of temporary employees, there is no longer a statistically significant difference between the wages of temporaries and the wages of permanent employees. The hourly earnings of fixed-term employees are now estimated to be higher than those of permanent employees, but not significantly so. Female casual workers are an exception: they are estimated to earn 6.5 percent less than female permanent employees, after taking part-time status, industry and occupation as well as demographic characteristics into account.

To summarise, the analysis indicates that the temporary-permanent gap in average hourly wages can be largely or entirely attributed to differences in measured demographic, educational, and job characteristics. Although we continue to find that a small wage penalty is associated with casual work for women, we have not been able to make any adjustment for a number of other factors that prior research findings suggest may influence the casual/non-casual wage differential, such as differences in previous work experience and unmeasured heterogeneity in attitudes and skills.

Overall, the results of the analysis undertaken here suggest that most temporary workers earn roughly the same amount per hour as similar permanent workers, after taking both their personal characteristics and some key job characteristics (industry, occupation and hours) into account. They imply that temporary workers in New Zealand are not systematically employed at lower rates of pay than permanent employees, as a result of their type of employment relationship. Some types of temporary work may in fact attract pay premia. Due to the limitations of the SoWL dataset, the wage premiums or wage penalties that are associated with different kinds of temporary work can't be estimated using all the information that ideally would be brought to bear on the issue, and for that reason the finding that most temporary workers earn roughly the same hourly wages as similar permanent workers in similar types of work should be regarded as a provisional one.

The finding that temporary employment is not generally associated with a wage rate penalty relates only to hourly compensation. Because temporary employees work fewer hours per week than permanent employees on average, their weekly earnings will be lower. If temporary workers also work for fewer weeks of the year than permanent workers, which seems very likely, their annual earnings will also be commensurately lower.

6.4 Job-related training

In the descriptive statistics reported above, the absolute difference between temporary and permanent employees in rates of training participation was quite large. Eighteen percent of temporary employees, and 32 percent of permanent employees, reported they had undertaken some employer-funded study or training in the past twelve months, a difference of 14 percentage points. Clearly, this 14 percentage point gap does not represent the marginal effect of working in a temporary job on the probability of receiving training.

To estimate the marginal effect of working in a temporary job on the probability of having received training in the last year, we modelled the probability of having received training using binominal logistic regressions. The sample used for this analysis was all employees in the SoWL. The training rates of males and females were modelled separately.

The reduced-form model takes the form:

Y = f(X, T) + ε

The dependent variable Y is the incidence of any employer-funded study of training in the last year (which takes the value one if the individual reported that they undertook training and zero otherwise). The explanatory variables comprise individual characteristics (X) and temporary job status (T). T varies across specifications. It is either a single dummy variable that is set to one for all temporary workers and zero otherwise; or a set of four dummies that separately identify each type of temporary worker (casual, fixed-term, temporary agency and seasonal not otherwise classified); or a pair of dummies that identify seasonal workers and temporary non-seasonal workers, distinguishing them from permanents.

Initially, we model the training participation probability of men and women as a function of their measured personal characteristics (age, ethnicity, parental status, immigrant status and years in New Zealand, geographical location, and highest educational qualification) and temporary job status (or type of temporary job). In a second set of regressions, the following job characteristics were added to the explanatory variables: part-time status, occupation defined at one-digit level, industry defined at one-digit level, size of employer, and the employer's business type (private sector, public sector or non-profit sector). We do not control for variations in job tenure because this is likely to be determined jointly with temporary job status. With only one observation for each respondent, we are unable to make any statistical adjustments for heterogeneity on unmeasured factors such as career aspirations or the motivation to learn job-related skills. We are also unable to control for differences across individuals in weeks or months of employment during the previous year, a factor that is likely to have a direct impact on our measure of training rates (because people who worked for only part of the year would have had less time in which to receive training, all else being equal).

Table 16 presents the key results, showing the marginal effects of being in a temporary job on the estimated probability of having received training.^[16] The first column of the table shows the unadjusted or raw percentage point gap in training rates between a particular group of temporary workers (as specified by the row heading) and permanent employees. The second column shows the marginal effect of temporary employment on training rates after controlling for personal characteristics. The third column shows the marginal effect of temporary employment on training rates after controlling for both personal and job characteristics. These marginal effects represent the difference between the predicted training participation rates of temporary and permanent employees, calculated while holding the effects of other variables constant.^[17]

Table 16: Estimates of the gap in training rates between temporary and permanent employees

For all male temporary employees and for each sub-group of male temporaries, the predicted training probabilities remain substantially lower than those of permanent employees after adjustments for personal and job characteristics have been made. The training probability for all male temporary workers is estimated to be approximately 15 percentage points lower than that for males in permanent employment, with similar personal and job characteristics. (The unadjusted difference was 21 percentage points.) The estimated training probabilities are also significantly lower for males in casual, fixed-term, and seasonal jobs.

Women in temporary jobs face a smaller gap in training rates than men, relative to women in permanent jobs, both before and after adjustments are made for differences in characteristics. In the specification with a full set of controls for personal and job characteristics (ie model 2), the estimated training probability for all females in temporary work is 5.6 percentage points lower than that for 'similar' female permanents in similar jobs. (The unadjusted difference was 9 percentage points.) The disaggregated results show that women in casual and temporary employment agency jobs have significantly lower training probabilities than those in permanent jobs, but women in fixed term and seasonal jobs do not.

Temporary and permanent employees are likely to differ in their continuity of employment, and this can be expected to directly influence training rates when the reference period for measuring training participation is the previous year. Although it isn't possible to control for differences between survey respondents in weeks worked, we explore the potential impacts of employment continuity on our results by restricting the estimation sample to individuals with tenure of at least one year with their current employer, and re-estimating the models. The results obtained are shown as 'model 3' in the fourth column of table 16.

Restricting the sample to employees with job tenure of one year or more reduces the size of most of the marginal effects estimated for temporary work, consistent with the idea that differences in employment continuity are likely to matter. Males in casual and seasonal employment, and females in casual employment, continue to have significantly lower estimated probabilities of participation in training in these regression estimates. However, it is highly likely that some differences in employment continuity remain even in the restricted sample. The job tenure restriction is unlikely to equalise weeks worked across the two groups, because the job tenure measure in the SoWL does not specify that employment with the current employer must have been continuous to be counted as part of the current job. So it is unclear whether temporary employment would continue to be associated with a lower training probability if it was possible to fully control for the effects of differences in weeks worked.

The implications of a lower rate of training in temporary work will depend on the role of temporary employment in individuals' working lives. Teenagers who work in temporary jobs may receive higher levels of training when they move on to permanent jobs at a later age, while older employees may have received substantial training in the permanent jobs they held earlier in their working lives. In the final column of table 16, we show the differentials for temporary work that were estimated when the sample was restricted to prime-aged employees, ie persons aged 25-54. In the case of males, these marginal effects are smaller than those estimated for all age groups. In the case of females, the age restriction makes little difference.

If a causal relationship between temporary employment and a reduced likelihood of receiving training does exist, the implications are also likely to depend on the duration of time that individuals spend working in temporary jobs. Short spells may have relatively little impact on skill accumulation and wage growth, but if some individuals work in a succession of temporary jobs over a number of years, adverse long-term impacts may be more likely. The OECD (2002) reviewed the evidence on rates of mobility from temporary to permanent jobs in Europe, and found the patterns were complex. In most countries a significant proportion of temporary workers move into permanent jobs within a two-year horizon, but this proportion varied across countries from a minority of temporary workers to a majority. Transition probabilities also varied across different types of temporary work and different types of worker. It is possible that a minority of temporary workers remain 'trapped' in temporary jobs for an extended period of time, but firm evidence of this phenomena is not currently available.

To summarise the findings of this section, the temporary-permanent gap in training rates appears to be partly due to differences in the personal characteristics of temporary and permanent employees. On average temporary employees are less highly educated than permanent employees and more likely to be aged under 25 years, and both these factors are correlated with a lower likelihood of receiving employer-funded training. Differences in job characteristics, including part-time status, industry, occupation and job tenure, also make a contribution to the temporary-permanent training gap. These factors do not fully account for the lower training rates of temporary employees. It is likely that some of the remaining gap is due to the fact that temporary employees typically worked for fewer weeks in the previous year than permanent employees.

The finding that there is a significant association between temporary employment and a lower probability of having undertaken training is consistent with the hypothesis that employers offer less training to temporary workers. It is also possible that there are unmeasured differences between temporary and permanent employees on other factors such as weeks worked during the year and the motivation to undertake training that are contributing to the gap in participation rates. Greenhalgh and Mavrotas (1994) found that employees' career aspirations and attitudes to training were significant determinants of their participation in employer-based training.

6.5 Summary

This section of the paper has considered the factors that may be contributing to the wage rate and training rate differentials that are found between temporary and permanent employees. Regression methods were used to estimate the size and significance of those wage and training differentials after taking into account differences in the measured personal and job characteristics of temporary and permanent employees.

Little evidence was found of a statistically significant gap between temporary and permanent employees in hourly average earnings after adjusting for personal and job characteristics. Females in casual jobs are the only sub-group of temporary workers whose hourly earnings were significantly (approximately 6 percent) lower than those of permanent employees of the same gender. This result could indicate the existence of pay penalty for casual employment, but the presence of uncontrolled differences in unmeasured characteristics such as prior work experience (which is likely to be associated with skill accumulation and therefore earnings), is also a plausible explanation.

Adjusting for differences in personal and job characteristics substantially reduced the size of the estimated training rate 'penalty' associated with temporary work, but did not eliminate it. The model results imply that males in temporary jobs were approximately 15 percentage points less likely to have undertaken employer-funded training than males in permanent jobs, while females in temporary work were approximately 6 percentage points less likely to have done so than females in permanent jobs. The analysis was not able to control for differences in weeks of employment during the reference period, or differences in prior employment histories, or individual heterogeneity, factors that are likely to influence both earnings and training rates and may account for the remaining differences in training probabilities. Consequently, although the results obtained suggest that temporary employment is associated with a lower likelihood of receiving employer-funded training, this should be treated as a provisional finding.

---