Employment Contracts and E ort: Why Do Temporary Workers Take Less Absence?

Transcription

1 Employment Contracts and E ort: Why Do Temporary Workers Take Less Absence? Steve Bradley, Colin Green y and Gareth Leeves z March 7, 2007 Abstract Worker e ort is di cult to measure. One related measure occasionally available in personnel and survey data is individual absenteeism. Recently, it has been demonstrated that workers on temporary contracts take substantially less absence than permanent workers, which suggests that employment contracts can in uence worker e ort. Using personnel data drawn from a large public sector labour force, we investigate to what extent, and through which channels, temporary employment contracts in uence worker absence. Speci cally we focus on the role of job security, promotion opportunities (to permanent contracts), tenure and workplace norming on the absence behaviour of a cohort of temporary workers. We demonstrate that a large proportion (61% to 66%) of the temporary-to-permanent absence di erential can be explained by these factors. Nonetheless, a contract-based absence di erential remains once the contribution of these factors are accounted for. KEYWORDS: Absenteeism, E ort, Temporary Contracts. JEL CODE: J29, J45, M54. Economics Department, University of Lancaster, UK y Centre for e-science, University of Lancaster, UK z School of Economics, St Lucia, University of Queensland, Australia

2 1 Introduction Worker e ort is di cult to measure. Often researchers have turned to the analysis of worker absenteeism as a negative measure of e ort. It has been demonstrated that absence behaviour provides a predictor of promotion and wage growth within the rm (Audas, Barmby and Treble 2004). This suggests that absence may be used as a selection device within an organisation. Hence workers have an incentive to modify their absenteeism so as to signal e ort. A situation where the incentive to lower absence may be particularly acute is during the period a worker is seeking to gain permanency in an organisation. Researchers have looked for an increase in absenteeism rates after probation periods as evidence of a response from worker e ort to changes in incentives (Riphahn and Thalmaier 2001, Ichino and Riphahn 2005). The increasing use of temporary contracts as a pathway to permanent employment has lead other researchers to investigate if there are systematic di erences in absenteeism between temporary and permanent contract workers (Jimeno and Cortes 1996, Engellandt and Riphahn 2005). The expectation is that temporary workers would display lower rates of absenteeism than permanent employees, but results, to date, have been mixed. This paper analyses the e ects of temporary and permanent contracts on worker absenteeism using longitudinal personnel data for an entire public sector workforce. This data is advantageous in a number of ways and enables us to substantially extend previous research. Our data e ectively covers a multiorganisation, multi- rm workforce. This gives rise to variation in not only work- 2

3 ing conditions, but other pertinent factors such as the relative use of temporary contracts, likelihood of transition to permanent employment and risk of nonrenewal for temporary workers. Previous research has been based on either single rm data where this variation is not present, or survey data which lacks the same workplace level information. With this data we aim to establish the existence of a di erence in absenteeism behaviour by contract type, and then to examine the extent to which this di erence in behaviour can be explained by a number of factors. Speci cally we examine the role of (1) tenure, (2) workplace characteristics, (3) individual speci c and workplace unobserved heterogeneity, (4) promotion opportunities and (5) separation risk (i.e the risk of contract non-renewal). We focus on a cohort of temporary workers, in particular the transition from temporary to permanent employment The remainder of the paper is set out as follows. The next section provides an overview of the existing literature. Section 3 provides background on the public sector workforce, an outline of the data and preliminary evidence on the size of the temporary to permanent absenteeism di erential. Section 4 provides a theoretical framework of contract e ects on absence behaviour. Section 5 presents the results of the cohort analysis, while section 6 concludes. 2 Previous Evidence The in uence of a worker s contract status on absenteeism has been investigated by a number of authors. For instance, Barmby et al (1995) and Barmby and 3

4 Stephan (2000) show that full time workers, or those with longer contracted hours, are more likely to be absent when compared to part time workers. The rst study speci cally examining temporary and permanent worker absenteeism behaviour was Jimeno and Cortes (1996). The study used Spanish survey data and concluded that after controlling for occupational and industrial di erences, there were signi cantly lower rates of absenteeism amongst temporary workers. In studies using data from the German Socioeconomic Panel, Riphahn and Thalmaier (2001) and Ichino and Riphahn (2004) found that workers in the German public sector, who were subject to very high degrees of employment protection 1, had absence rates of 6.7% which compared to only 4.3% for those in the public sector who were not covered by the extended protection. This di erence remained robust (1.57% higher absence rate for protected compared to non-protected) even after conditioning on a range of observable characteristics. Public sector worker s absence also increased signi cantly around the six month period when the e ect of stringent lay-o laws come into e ect. Ichino and Riphahn (2004) also reported a statistically signi cant di erence in absenteeism between workers in small and large Italian rms. Workers in large rms are subject to very strict employment protection legislation whereas workers in small rms are exempt. Recently, Ichino and Riphahn (2005), using personnel data from an Italian bank, exploited legislative rules on probation periods to demonstrate that once 1 In the German public sector workers who have accumulated fteen years of service in the public sector and are over forty years old cannot be dismissed except for gross misconduct. These provisions do not apply to other workers in the sector. 4

5 employment protection legislation becomes active, workers take markedly more absence, approximately 0.04 of a day per week. This suggests that workers view the cost of absenteeism, in terms of lost earnings due to the detection of shirking and subsequent dismissal, as being high during probationary periods when dismissal is easier. In contrast to the above ndings, a study of temporary worker absenteeism by Engellandt and Riphahn (2005), using longitudinal data from the Swiss Labour Force Survey, concluded that there was no signi cant di erence between the absence rates of temporary and permanent workers after controlling for a range of observable characteristics and controls for individual xed e ects. However, there was only a very small sample of temporary workers who recorded any absenteeism and this may account for the lack of any signi cant e ect. In addition, Swiss absence rates are very low by international standards and it may therefore be the case that temporary workers are less able to signal e ort through absenteeism. They demonstrate that an alternative indicator of e ort, unpaid overtime, is 60% higher amongst temporary workers than permanent workers. 3 Background and Preliminary Evidence 3.1 The Queensland Public Sector Workforce The data used in this study are based on the administrative personnel records of the Queensland State Government. This data was collected in order to facilitate human resource management and is known as the Minimum Obligatory Human 5

6 Resources Information (MOHRI) database. It represents the minimum level of human resource information that the Queensland Government s agencies are required to collect and report to the central government agency for industrial relations purposes. The database holds quarterly information on approximately 180,000 public sector workers and in this study refers to the period from 2001(1) through to 2004(2). In Australia, state governments account for 65.8% of all public sector employees and have responsibility for core services, such as education, health, emergency services and law enforcement. The remaining public sector employees work in the federal government (23%) and local government (11.2%). The Queensland State Government is the third largest in Australia, servicing a population of approximately 3.6 million people and making up approximately 12.5% of the total employed labour force in the State. Workers in the Queensland public service can be employed on one of four types of contract, casual, temporary, permanent or contract. The latter is a special case involving instances such as one of sub-contracting, and these individuals do not represent ongoing members of the workforce. As a result they are excluded from further analysis. Casual employees are, in essence, hired on a common law basis each time they are commissioned for work. They have no entitlements to annual holiday pay or paid sick leave (ABS 1996). Hence we also exclude this group from the subsequent analysis. Temporary contracts are xed term contracts, that are renewable. Permanent contracts are tenured, and hence permanent employees are very di cult to re. There are, however, no 6

7 di erences in holiday or sick pay entitlements between the two types of workers. 3.2 Do Temporary Workers Take Less Absence? INSERT TABLE 1 Table 1 provides an overview of the Queensland public sector labour force, split according to contract status. The permanent worker takes on average 5 and a half hours more absence a quarter than temporary workers. Temporary workers are, on average, younger, with less tenure and they work less hours (have lower average Full Time Equivalency, FTE). There are also noticeable di erences in occupational structure between the two groups, with temporary contracted workers less likely to be employed in management positions or as teachers and nurses, but more likely to be in other professional, intermediate or elementary clerical, sales and service work. Temporary workers are also more likely to be Aboriginal or Torres Strait Islander (ATSI) or of an Asian background. These di erences in observable characteristics between temporary and permanent contracted workers may in turn explain some of the di erence in observed absence behaviour. Therefore, as a rst step, we examine the contribution of these charaterisitics by estimating the following model: a it = X it + W it + P erm it + " it (1) Where a it is the hours taken absence in quarter t, X i is a vector of personal characteristics, W i is a vector of workplace and work related characteristics (including tenure) and P erm i is the workers contract status, here denoted by 7

8 a dummy variable for permanent contract (the omitted variable is temporary employment). Estimation of (1) provides a baseline estimate of the di erence in permanent and temporary absence behaviour conditional on workplace and personal characteristics. This model is intially estimated on the complete 14 quarter panel of MOHRI using OLS, and the results are reported in columns 2 and 3 of Table 2. INSERT TABLE 2. Looking at the coe cient on permanent contract status, the OLS estimates suggest that just under half of the temporary to permanent di erence in absence reported in Table 1 were due to observable characteristics that make temporary workers less likely to take absence. Nonetheless, a 3 hour per quarter di erence in absenteeism remains. Our longitudinal data allows for the estimation of (1) with the inclusion of individual speci c xed e ects to control for unobservable di erences in absence propensity between permanent and temporary contracted workers. Time invariant personal or workplace characteristics drop out of this model: a it = i + X it + W it + P erm it + it (2) Estimates of are identi ed in this model by workers moving between contract states. Columns 4 and 5 provide estimates from maximum likelihood estimation of (2). The contract e ect on absence reduces by an hour when compared to the OLS estimates, but remain at 2 hours per quarter and this is statistically signi cant at the 1% percent level. 8

9 These results demonstrate that there are di erences in temporary and permanent workers absence behaviour, and that these are not solely due to observable or unobservable characteristics. However, this data is pre-sorted insofar as individuals have already been assigned into temporary and permanent contracts within the public sector labour force. Previous research on the role of temporary employment as a port of entry into permanent employment suggests this will not be a random process (Booth et al 2002, Green and Leeves 2004). As a result estimates of from (1) or (2) are biased. For example, if there is a screening process of temporary workers then one may expect the poorer job matches to be terminated rst. Temporary workers exerting less e ort are less likely to have contracts renewed. Thus, one might expect the di erence between temporary and permanent worker absence to vary by duration of temporary employment (Engellandt and Riphahn 2005). We examine the di erence between temporary and permanent workers by categories of employment duration. The results for this exercise are presented in Table 3, where we report only estimates of the P erm coe cient. As expected the di erence between temporary and permanent increases as tenure increases and this remains true if we include individual xed e ects. In the latter case, the absence di erence is only signi cant for temporary workers with over three years tenure. Hence, in contrast to Engellandt and Riphahn (2005) we do nd a general di erence in absenteeism between the two contract types, which may re ect the coarser de nition of absenteeism that was used in their study. 2 2 Absence was de ned as whether the individual had missed work for the entire work week 9

10 INSERT TABLE 3 Given there are selection e ects; an alternative approach to control for presorting is to follow a cohort of new temporary workers through their employment in the public service. This is the empirical strategy adopted to examine the e ects of temporary to permanent transition on absenteeism. In the next section we present a model of a worker s absence decision to examine why contract status and prospects for change in that status may a ect workers absence behaviour. 4 A Model of Absence and Contract Status Assume that a workers utility is derived from work and absence Work e ort causes disutility and absence lowers overall e ort, hence absence (a) is positively related to utility. Wages are also postively related to utility. U i = U(w; a i ) = w + a i (3) The discounted utility of temporary employment is given by: rv ET i = w T + a i + (1 g)c(a i )(V U V ET i ) (4) where the discount rate is r and the value of being employed on a temporary contract is V ET. The variable g is the degree to which temporary contracts within his/her department are subject to evaluation of worker e ort as they previous to the survey due to illness, accident, personal or family matters, or other reasons. 10

11 lapse, where g = 0 implies all contracts are evaluated as they lapse and g = 1 means all contracts are terminated as they lapse and 0 < g < 1: The probability of non-renewal is a function of worker absenteeism C(a) (C = > 0 and C == > 0). The variable g is not directly observable for the worker but would be estimated from current information supplied by the employer and from recent history of termination and renewal in the department. Equation (4) states that the discounted utility of employment for a worker is equal to the instantaneous utility of employment, the rst two terms of the right hand side, and the loss in utility from the probability of non-renewal, the second term. The loss in utility is equal to the di erence between the value of being unemployed (V U ) and the value of employment. It is assumed that all workers are hired initially on temporary contracts before conversion to permanent contracts. The discounted utility of permanent employment is given by: rvi EP = w P + a i + (1 )C(a i )(Vi ET V EP i ): (5) The value of being employed on a permanent contract is V EP :The probability that a worker is o ered a conversion to a permanent position is represented by (1 )C(a i ). The probability of of conversion is a function of absenteeism C(a). The variable represents the proportion of temporary contracts within the department where the worker is located that are being evaluated for conversion to permanent positions. The evaluation of by the worker would be based on information within the department and the recent levels of conversion to permanent employment. If = 0 then all contracts are subject to evaluation 11

12 and if = 1 no contracts are subject to evaluation and this lies in the range 0 < < 1: The worker will choose absenteeism to maximise rv EP i which, for all workers, is given by the condition: = 1 (1 )C = (a)(v ET V EP ) = 0: (6) Equation (6) illustrates how a worker will take absence until the marginal bene t is equated to the marginal cost We deal separately with the evaluation decisions relating to temporary and permanent work and temporary work and unemployment. The two expressions for the value of temporary and permanent work can be used to produce the following expression for V EP V ET V EP V ET = (w P w T ) r + (1 g)c(a) + (1 )C(a) (7) This expression can be combined with equation 6 to obtain the condition: S = r + (1 g)c(a) + (1 )C(a) ((1 )C 0 (a))(w P w T ) = 0 (8) To determine the e ect of higher rates of conversion to permanent positions on absenteeism we need to evaluate this equation which through implicit di erentiation gives the expression 12

13 da d = X X a = r + (1 g)c(a) (1 g)c 0 (a) + (1 )C 0 (a) (w P w T )((1 )C 00 (a)) > 0 (9) From this we see that workers will increase absence as level of evaluation for conversions increases. If a temporary worker is not renewed then the value of being in unemployment is de ned by income when unemployed b (same for all workers) and the probability of nding another job multiplied by the di erence in value between employment and unemployment. rv U = b + (V ET i V U ): (10) The level of absenteeism which will maximise rv ET i is given by = 1 (1 g)c = (a)(v ET V U ) = 0: (11) The two expressions for the value of work and unemployment can be used to produce the following expression for V ET V U. V ET V U w T + a = r + + (1 b g)c(a) (12) This expression can be combined with equation 11 to give the following condition. 13

14 X = r + + (1 g)c i (a) (w T + a b)(1 g)c = i (a) = 0 (13) To determine the e ect of higher levels of non-renewal on absenteeism we need to evaluate equation (13) using implicit di erentiation, which gives the expression da dg = X g r + = X a (1 g)c 00 (a)(w T + a b) > 0 (14) If an individual is in work then (w T + a b) > 0 and workers will increase absence as level of evaluation for renewal increases. 5 Absenteeism and the Transition to Permanent Employment In an attempt to address the selection e ects into contract type identi ed in section 3, the following analysis is based on a cohort of recent entrants to the public sector labour force. The cohort used is all individuals who entered the public sector workforce during the rst year of our MOHRI data. We then follow these individuals for up to 12 additional quarters, subject to their remaining in the public sector workforce. Workers can initially enter the public sector labour force on any contract type. Summary statistics for the cohort sample, disaggregated by contract status, are reported in the Appendix as Table A1. Brie y, individuals who start in temporary contracts take markedly less absence 14

15 in that quarter, than those who start on permanent contracts. It also displays contract non-renewal rates and transition rates to permanent employment for workers who start on a temporary contract. Similar to the data presented for the entire public sector labour force (Table 1), new entrants to public sector who are on temporary contracts are younger and on average work less hours and are less likely to be in managerial positions and more likely to be low skill occupations. INSERT FIGURES 1 and 2 Using this data we can plot the absence behaviour over time of both temporary and permanent workers, these are provided as Figures 1 and 2, respectively. Each gure plots mean individual absence by quarter in the public sector (i.e. the x-axis runs from the quarter of entry into the public sector through to the last quarter) according to whether they started on a temporary or a permanent contract. What is immediately noticeable, is that both temporary and permanent workers take very little absence in the rst two quarters. For the rst quarter that absence is lower is not surprising as this may not represent a full quarter in the workforce. With this said, if we assume that the start dates are evenly distributed across the quarter this alone does not explain the magnitude of this di erence. It is interesting to note that the period in which the largest increase in absenteeism is observed, the 1st to 3rd quarter of tenure, covers a similar interval to that over which an increase in absenteeism is observed by Ichino and Riphahn (2005), which in their data is associated with beginning of employment protection provisions. Here, we observe similar increases in absence 15

16 but without any change in employment protection. Though as these authors noted in an earlier study (Ichino and Riphahn 2004) this could represent the e ect of early career concerns (Holmstrom 1992). Workers have an incentive to take less absence early in their career as supervisors have little knowledge of their abilities, but this incentive will decrease with time as monitoring will be likely to decrease. Hence, absenteeism will increase with tenure. For permanent workers, absence levels continue to climb until around 2 years of tenure, for temporary workers absence levels appear to plateau by the 1st year of tenure. INSERT FIGURE 3 Our particular interest is in the change in absence behaviour that occurs once a worker moves from temporary to permanent employment. Figure 3 presents temporary workers absence normalised to the time at which they transit to permanent employment. Speci cally, time 0 on the x-axis refers to the quarter in which the worker made the transition from temporary to permanent contract. As a result period 0 covers a quarter within which we do not observe perfectly the timing of the transition between contract types. Hence we do not know exactly how much of the absence in this period occurred in temporary and also in permanent contract. As a result, the correct comparison is between the periods denoted -1 and 1. This gives an unconditional di erence of 2.1 hours absence. Alternatively, one could look at mean di erences in absenteeism over the whole period before and after transition, this provides a mean di erence of 4.1 hours. We now seek to examine more formally if temporary cohort absence increases 16

17 following a move to permanent employment. First consider the simple empirical model, where an individual s absence (a it ) is a function of personal characteristics (X it ), workplace characteristics (W it ) and contract status (P erm it ): a it = X it + W it + P erm it + " it (15) As noted above, we imperfectly observe the timing of the transition to permanent employment. As a result, we exclude the quarter of transition from the analysis 3. Estimation of (15) model by OLS provides a temporary to permanent di erential of 2.28 hours per quarter (reported in column 2, Table 4). Hence, just under half of the observed 4.1 hour di erence in behaviour between when an individual is on a permanent contract versus when they were on a temporary contract is due to observable personal and workplace characteristics. As in Table 2 we include a measure of an individual s contracted hours, their the full-time equivalent status (FTE). In a study of UK absenteeism Bridges and Mumford (2001) noted that absenteeism should be an increasing function of contracted hours of work as it lowers schedule exibility. We found this to be the case in the results reported in Table 2. As before the FTE coe cient is positive but it is about half the size, this would be indicative of both temporary and permanent workers having an incentive to better manage the reduction in exibility early in their tenure (Holmstrom 1992). There remains a di erence 3 In all estimations this has a small positive e ect on the estimate of the contract status e ect. For instance, for the OLS model reported in Table 4 the permanent contract e ect on absenteeism is including the quarter of transition and excluding this quarter. 17

18 and this may re ect the fact that individuals working less hours are monitored more intensively than others or their duties are such that absence is more di - cult to take. Alternatively, it maybe that these individuals are more motivated to reduce absenteeism, perhaps to secure their position in the organisation. For instance, they may be individuals who are presenters rather than absenters (Chatterji and Tilley 2002). We control for the in uence of unobservable factors by extending our model to include xed e ects controls. We can utilise the panel structure of MOHRI and include individual time invariant xed e ects such that: a it = i + X it + W it + P erm it + " it (16) Where i is an individual speci c time invariant xed e ect. Estimation of this model by MLE leads to a reduced estimate of temporary to permanent di erential of 1.42 hours. When compared to the OLS estimates, this lower estimate of the impact of permanency on absenteeism suggests that individuals with lower absenteeism propensities are more likely to make the transition to permanent employment. This is consistent with previous research demonstrating a connection between lower absenteeism and promotion (Audas, Barmby and Treble 2004). The FTE coe cient decreases when allowing for individual xed e ects but remains signi cant and positive. Workplaces are also likely to vary in terms of absenteeism levels (Ichino and Maggi 2000, Bradley et al 2007), and this is likely to be due to factors such as management policy that will not be observed in data. As a result, unobservable 18

19 workplace characteristics may bias estimates of the contract e ect on absenteeism. We introduce controls for unobservable workplace characteristics into equation 16 above. Ideally, this would include a xed e ect for every workplace in the public sector. However, this would require the inclusion of over 1500 xed e ects. Instead we include xed e ects for each of the 51 agencies in the public sector (where Z ij is a dummy that is equal to one if worker i is in agency j at time t). This controls for di erences in departmental policies and management practices: a it = i + Z ijt + X it + W it + P erm it + " it (17) Estimation of (17) by MLE (columns 2 and 3, table 5) leads to a small but statistically insigni cant reduction in the estimated e ect of permanency on absenteeism to 1.41 hours. Peer absenteeism behaviour may e ect individual absenteeism (Ichino and Maggi 2000, Bradley et al 2007). We investigate this by introducing a control for average workplace absenteeism (excluding the ith individual). Estimating this model does not have a marked e ect on the estimate of permanent contract e ects on absenteeism (columns 4 and 5 Table 5). Absenteeism is likely to have a seasonal component, the underlying rates of actual absence may not be stationary. If there is also seasonality in the timing of temporary to permanent transitions these factors could lead to spurious estimates of contract status on absenteeism. We therefore introduce a control for the quarter of the year, and omit the rst quarter (January to March). Estimates of the transition e ect from this model are 1.55 hours, suggesting 19

20 that seasonality in the data leads to some minor understatement of the contract status e ect. In this section we have sought to quantify the impact of contract status on absence behaviour by examining workers who enter the public sector labour force on temporary contracts and contrasting their absence behaviour to that when they gain a permanent contract. The evidence presented suggests that between 61 to 66 percent, depending on the exact model speci cation, of the overall observed di erence of 4.1 hours in absenteeism can be explained by controlling for observed and unobserved personal and workplace characteristics. 5.1 The role of renewal and transition rates We now formally consider the role of separation risk and the probability of transition to permanency on the e ort of temporary workers. The contemporaneous rate of (rate of conversions) and g (separation risk) are unknown to the worker when considering how much e ort to exert. Hence, we assume workers base their expectations on recent history and so take the lagged values of these variables as an indicator of their lieklihood in the current period. Our expectation is that e ort will respond positively to increased probability of renewal and separation. We take equation (17) and add a measure for the probability of contract non-renewal (separation risk) and the temporary to permanent transition rate. Each of these variables are measured in half yearly lags and exclude the ith individual. This leads to the following empirical speci cation, where R is the rate of renewal of temporary contracts in worker i 0 s workplace j and S is 20

21 the rate of non-renewal of temporary contracts in worker i 0 s workplace j: a it = i + R ijt 2 + 'S ijt 2 + X it + W it + P erm it + " it (18) INSERT TABLE 6 These results are reported in Table 6. The coe cient estimates indicate that both the risk of contract non-renewal and the probability of transition decrease absenteeism (increase e ort), which is consistent with the theoretical model of absence outlined earlier. These estimates are robust to including both variables within the same model (columns 6 and 7). This suggests that the threat of non-renewal and the likelihood of transition have separate e ects on individual s absence behaviour. In terms of magnitude, changes in the transition rate to permanent employment has roughly twice the impact that changes in non-renewal rates has on individual absenteeism. Hence temporary workers absence is more sensitive to their probability of gaining permanent employment, rather than risk of contract non-renewal. Despite the statistical signi cance of these e ects, they do not markedly change the estimated overall permanent contract e ect on absenteeism, which remains at 1.8 hours per quarter. 4 In unreported tests, we interacted these two variables with a dummy indicating permanent employment. We would expect that neither of these variables should in uence absenteeism once an individual gains a permanent contract. Indeed these interaction terms were insigni cant and approximately zero, whilst 4 These results are robust to the inclusion of time-invariant agency xed e ects. 21

22 the sign and size of the coe cients on separation risk and temporary to permanent transition were una ected 6 Conclusion It has previously been demonstrated that a worker s contract status in uences their level of e ort or absenteeism (Ichino and Riphahn 2005, Arai and Thoursie 2005). Using longitudinal personnel data drawn from a large public sector labour force, we investigate to what extent and through which channels temporary employment contracts in uence worker absence. Our data is advantageous insofar as it covers a multi-organisation, multi- rm workforce. As a result, we observe variation in not only working conditions, but also temporary contract non-renewal and transition rates to permanent employment. We nd a di erence between temporary and permanent worker absence that cannot be accounted for by obervable and unobservable characteristics. This di erence widens as tenure in the public sector increases, which is indicative of sorting over time. This leads us to use a cohort approach where we follow workers who enter the public sector labour force on temporary contracts and later move to permanent employment. Conditional on observed and unobserved characteristics we again nd a di erence between temporary and permanent absence. A lack of job security and a desire to signal suitability for a permanent appointment represent two factors that may lead temporary workers to take less absence. We examine the relative role of job security or separation risk and 22

23 promotion opportunities (to permanent contracts), where our theoretical model suggests that both these workplace charateristics should condition temporary workers to exert more e ort. We nd this to be the case for both factors and these e ects are statistically signi cant. Quantitatively, the promotion opportunity e ect is stronger. This could be regarded as a more direct signal to temporary workers of the reward to e ort than the threat of contract nonrenewal. We demonstrate that a large proportion (61% to 66%) of the temporary to permanent absence di erential in rates of absenteeism can be explained by observed and unobserved personal and workplace characteristics. Nonetheless, a temporary to permanent contract absence di erence still remains. Some contract based di erence in e ort exists that is not readily explained by factors such as di erences in unobserved individual and workplace absence propensity; di erences in observable absence associated characteristics (such as tenure) between temporary and permanent workers; temporary workers inherently lower job security; or temporary workers attempts to signal suitability for permanent employment through increased e ort (lower absence). REFERENCES Audas, R. Barmby, T. and Treble, J. G., Luck, E ort, and Reward in an Organizational Hierachy,.Journal of Labor Economics 22, Australian Bureau of Statistics Standards for Labour Market Statistics, 23

24 Cat.No , Canberra, AGPS. Arai, M. and Thoursie, P. S Incentives and selection in cyclical absenteeism, Labour Economics, 12, Barmby, T. Orme, C. D. and Treble, J. G., Worker Absence Histories: a Panel Data Study, Labour Economics, 2, Barmby, T. and Stephan, G., Worker Absenteeism: Why Firm Size May Matter. The Manchester School 68, Barmby, T. Ercolani, M. G. and Treble, J. G., Sickness Absence: An International Comparison. The Economic Journal 112, F315-F331. Booth, A., Francesconi, M. and Frank, J Temporary Jobs: Stepping Stones Or Dead Ends?, The Economic Journal, Vol 112, pp Bradley, S., Green, C. and Leeves, G., 2006 Workers Absence and Shirking: Evidence from Matched Teacher-School Data, Labour Economics (forthcoming). Chatterji, M. and Tilley, C. J., Sickness, Absenteeism, Preseteeism and Sick Pay. Oxford Economic Papers 54, Engellandt, A. and Riphahn, R. T., Temporary contracts and employee e ort, Labour Economics, Green, C. and Leeves, G Casual Employment and Internal Labour Markets, Manchester School, 72,

25 Holmstrom, B., Managerial Incentive Problems - A Dynamic Perspective in Essays in Economics and Management in Honor of Lars Walbeck. Helsinki: Swedish School of Economics, reprinted in Review of Economic Studies, 66, Ichino, A. and Maggi, G., Work environment and individual background: Explaining regional shirking di erentials in a large Italian rm, Quarterly Journal of Economics, August, Ichino, A. and Riphahn., Absenteeism and Employment Protection: Three Case Studies, Swedish Economic Policy Review, Ichino, A. and Riphahn, R. T., The e ect of employment protection on worker e ort: a comparison of worker absenteeism during and after probation, Journal of European Economic Association, 3, Jimeno, J. F. and Cortes, L. T., E ort, absenteeism and xed term employment contracts, Revista Espanola de Economia, 1 3, Bridges, S. and Mumford, K., Absenteeism in the UK: A Comparison Across Genders, The Manchester School, 69, Riphahn, R. T and Thalmaier, A Behavioral E ects of Probation Periods: An Analysis of Worker Absenteeism, Journal of Economics and Statisitics, 221,

26 Table 1: Summary Statistics, Queensland Public Service 2001(1)-2004(3), Age Full Sample Temporary Permanent Absence (hours in quarter) [20.834] [35.133] Male Ln Hourly Wage [0.353] [0.399] Age [11.367] [10.163] Tenure [4.763] [9.117] Non-English Speaking Background (NESB) European Asian Other Aboriginal or Torres Strait Islander (ATSI) Disability Full-time Equivalency (FTE) Occupation: 1. Manager Other Professional Teacher Nurse Associate Professionals Tradespersons Advanced Clerical and Service Workers Intermediate Clerical, Sales and Service Workers Intermediate Production and Transport Workers Elementary Clerical, Sales and Service Workers Labourers and Related Workers Observation 69, ,365 6 Source: MOHRI data. Standard errors reported in parentheses where appropriate. 26

27 Table 2: Determinants of Hours Absent per Quarter, Full MOHRI 2001(1) to 2004(3) 8 OLS Fixed E ects Coe Std. Err Coe Std. Err Permanent Contract Male Age Age Tenure Tenure FTE Disabled Non-English Speaking Background (NESB) European Asian Other Aboriginal or Torres Strait Islander (ATSI) Occupation 1. Manager 2. Other Professional Teacher Nursing Associate Professional Tradespersons Advanced Clerical and Service Workers Intermediate Clerical, Sales and Service Workers Intermediate Production and Transport Workers Elementary Clerical, Sales and Service Workers Labourers and Related Workers Observations Source: MOHRI data. 27

28 Table 3: Absenteeism and Contract Status by Tenure, Full MOHRI 2001(1) to 2004(3), period 10 OLS Fixed E ects Coe Std. Err Coe Std. Err Tenure 0-1 Year Tenure 1-3 Year Tenure 3 Years plus Source: MOHRI data. 28

29 Table 4: Absenteeism and Transition to Permanent Employment, Temporary Entrant Cohort 12 OLS Fixed E ects Coe Std. Err Coe Std. Err Permanent Contract Male Age Age Tenure Tenure Disabled FTE NESB European Asian Other ATSI Occupation: 1. Manager 2. Other Professional Teacher Nurse Associate Professionals Tradespersons Advanced Clerical and Service Workers Intermediate Clerical, Sales and Service Workers Intermediate Production and Transport Workers Elementary Clerical, Sales and Service Workersz Labourers and Related Intercept Observations 45, Source: MOHRI data. 29

30 Table 5: Absenteeism and Alternative Controls, Temporary at Start Cohort 13 Agency FE Peer E ects Seasonality Coe Std. Err Coe Std. Err Coe Std. Err Permanent Contract Source: MOHRI data. All models include controls for individual time varying characteristics, individual time invariant xed e ects and workplace characteristics. 30

31 Table 6: Absence, Contract Security and Promotion Probability, Temporary at Start Cohort 15 Coe Std. Err Coe Std. Err Coe Std. Err Permanent Contract Probability of Contract Non-Renewal Temporary to Permanent Transition Prob Source: MOHRI data. All models include controls for individual time varying characteristics, individual time invariant xed e ects, workplace characteristics and agency level time invariant xed e ects. 31

32 mean of sick_hr_q Figure 1: Sick Hours by Tenure (quarters), Temporary at Entry

33 mean of sick_hr_q Figure 2: Sick Hours by Tenure (quarters), Permanent at Entry

34 mean of sick_hr_q Figure 3: Sick Hours by Time Until Temporary to Permanent Transition (quarters), Temporary at Entry

35 APPENDIX Table A1: Summary Statistics, Entrant Cohort Temporary Permanent mean std. dev mean std dev Individual Absenteeism (hours in starting quarter) Age (years) Male Non English Speaking Background (NESB) European Asian Other Disabled ATSI Full Time Equivalency (FTE) O ccupation 1. Manager Other Professional Teacher Nurse Associate Professionals Tradepersons Advanced Clerical and Service workers Intermediate Clerical, Sales and Service Workers Intermediate Production and transport Workers Elementary Clerical, Sales and Service Workers Labourers and Related Workers Observations