Risk attitude and working contracts

Master’s Thesis:

Risk Attitude and Working Contracts

Alexander Mundle

Student-ID: 11832525

MSc. Business Economics - Specialisation Managerial Economics and Strategy

June 21, 2018

Supervisor:

Prof. Dr. E.J.S. Plug

Credits: 15 ECTS

Abstract

In this thesis it is investigated whether the personal attitude towards risk has an influence on the probability of being employed in a temporary contract. A simple theoretical model predicts that the associated uncertainty about future employment attracts more risk tolerant individuals. Using a panel data set from the German Socio-Economic Panel shows the non-existences of this causality. Interestingly, heterogeneity analyses indicate the presence of this relationship for highly educated employees, individuals in a partnership, residents of West Germany and employees with a low earning prospect. The finding of no or a very small effects hints toward the fact, that the role of risk attitude regarding individual labour market decisions is probably mitigated by more important criteria and therefore rarely crucial.

1

Introduction

A working contract with a duration that is limited in time is called a temporary or fixed-term con-tract 1. It is prominently utilized to generate flexibility regarding a firm’s work force (Berton

and Garibaldi, 2012, Faccini, 2014). A contract limited in time does give the opportunity to intensively screen the employee and evaluate the suitability of the employee. This increased flexibility goes hand in hand with a disadvantage for the employee. The associated risk of end-ing up in unemployment due to the absence of a permanent contract imposes uncertainty about the future employment situation. Additionally temporary contracts lead to higher turnover rates compared to permanent contracts, which constitutes another disadvantageous aspect (Blanchard and Landier, 2002; Dolado et al., 2002). On the contrary the screening possibility exists for the employee as well. The employee can observe the firm’s culture, working conditions and col-leagues and can assess the quality of the match and future prospects like career opportunities and salary structures. Furthermore, fixed-term contracts increase the probability of having a permanent contract in the future and are thus considered stepping-stones (Picchio, 2008; Ha-gen, 2003). Assuming that fixed-term contracts in comparison to permanent contracts are less attractive, gives rise to the question why job seekers do sign into them? More than often, the answer would be the lack of choices, however sometimes the opportunity exists to decide one either of the two types of contracts. According to standard economic theory the employees’ choice should swing into the direction with highest expected utility, since a fixed-term contract is certainly accompanied by future uncertainty regarding the employment situation, the factor of risk is not negligible.

Previous literature very much concentrated only on the wage differential between the two con-tract types. Hagen (2002) found among others that employees in fixed-term concon-tracts suffer from wage penalties. He chose an approach to eliminate the possibility that unobservable character-istics are the reason for earning less and having a temporary contract. This resulted in an even larger wage differential compared to the previous literature. Hagen interpreted this outcome as evidence for a self-selection effect caused by unobservables. Still, the finding of the selection effect itself does not contribute to answer the question which unobservable characteristics are of importance in this matter. Besides other factors, risk preferences are often involved when it comes to individual decision making. Especially the labour market, which is characterized by individual decisions and is affected by monetary aspects. Therefore, the employee’s risk atti-tude influencing the expected utility of those monetary decisions, constitutes a major influence regarding the future prospect of a position of employment. Hence, the question arises whether personal risk preferences are of a fundamental influence with respect to working contracts. This study investigates if higher levels of risk tolerance lead to a higher probability of being in a temporary contract. To do so it made use of the German Socio-Economic Panel, a survey based panel data set conducted annually in Germany. Its diverse set of obtained variables allows for a detailed analysis of risk and contract type.

Two different strategies are applied in this analysis. The need for two different approaches origi-nates from an assumption regarding the variation of personal risk preferences. The applied fixed

effects analysis is theoretically harmed by the former standard assumption that risk preferences are stable across time (Stigler and Becker, 1977). Nevertheless, many recent articles suggest that risk preference is subject to change due to alterations in the personal environment (Chuang and Schechter, 2015; Dohmen et al., 2017). Hence, under the assumption that the detected vari-ation of risk preferences in the data set is of informative nature a worker fixed effects analysis is the appropriate design. In case the assumption of informative changes in risk attitudes does not hold and risk preferences are stable across time, no variation can be captured by the fixed effects approach and a model without worker fixed effects gives more appropriate estimates. The findings can be summarized as follows: A higher individual risk attitude does not result in an increased probability of being in a fixed-term contract under the former mentioned assump-tion. In contrast, by assuming stable risk attitudes a very small positive significant effect of risk attitude on the probability of being employed in a temporary contract can be detected. On the one hand, given the theoretical considerations, it is surprising that a greater risk attitude is not related to a type of contract that exposes more uncertainty. On the other hand, taking all other potential influences into account, the role of risk may not receive as much emphasis as other fac-tors in an employment decision. Another potential explanation for the obtained results could be that heterogeneity with respect to the role of risk tolerance is present among the population. This suspicion, that heterogeneity exist in the sample, can be verified. For highly educated individu-als, people living in a partnership or individuals being resident in the western part of Germany, it is shown that risk does indeed have a significant effect on the type of working contract. It must be said, however, that in terms of magnitude the effect is still very small.

The obtained results are robust to the concern of reverse causality. Neither the use of past risk attitudes, nor using contract type as explanatory variable do display a correlation in the opposite direction.

The remainder of this paper is organized as follows. First, some literature review on temporary contracts , selection effects and risk preferences are given and the theoretical foundations are explained. In Section 3. the data set used and the applied methodology is described. Section 4. contains the results of the conducted estimations as well as the heterogeneity analyses. In Section 5. risk attitude as key variable and some further explanations of the detected patterns are discussed. The paper is completed by the conclusion in Section 6.

2

Literature review and theoretical considerations

In the following the literature on temporary contracts utilized as screening devices, self-selection on basis of risk preferences and characteristics of personal risk attitude regarding its variation across time is summarized.

expiring date, which is followed by either cancellation of the employment relation or continu-ation with another working contract (of permanent or temporary nature). The appliccontinu-ation of temporary contracts provides an employer with more flexibility regarding the work force and provides the cheaper alternative. This is partly due to the fact that in case of a termination the employee does not need to be compensated by a severance pay (Berton and Garibaldi, 2012, Faccini, 2014) and employees in temporary contracts suffer from an wage penalty (Blanchard and Landier, 2002; Hagen, 2002). Those benefits offer the employer more flexibility, however on the contrary the involved employee is harmed by those contract components, as it exposes the employee entering a fixed-term contract to a greater amount of future job uncertainty compared to an individual employed on a permanent basis.

2.1

Wage Differential and Screening Hypothesis

According to Rosen’s Theory of Equalizing Differences (1986) an employee needs to be com-pensated for every negative aspect of a job. Hence, wage premiums are necessary to offset this inequality. From a long-term monetary perspective the uncertainty associated with a fixed-term contract does constitute a negative side effect. Based on this theoretical prediction, empirical patterns should be present according to which individuals in temporary contracts receive a higher wage compared to equivalent positions in permanent contracts. Nevertheless, most studies in-vestigating the wages of permanent and temporary positions do obtain contradicting results. Individuals with temporary contracts receive less pay than in permanent positions (Booth et al., 2002; Schömann and Kruppe, 1994; Hagen, 2002). This contradicting finding can be explained by taking a screening process as a fundamental feature of temporary contracts into account. The theory of self-selection in the labor market by Guasch and Weiss (1981) provides a the-oretical explanation for the detected wage differential based on screening and self-selection. The authors describe that a fixed-term contract incorporates an application fee, which occurs as the employee is paid less during the selection phase which is equivalent to the duration of the fixed-term contract and potentially receives a higher wage afterwards (transition to permanent contract). This fee leads to a two-sided selection effect. Firstly, only employees are selected to receive a permanent position, which pass the screening process successfully. Secondly, only employees who assess that they can master the screening process successfully are tempted to take the job. Several studies examine this screening hypothesis and successfully confirm it on empirical basis (Boockmann and Hagen, 2007; Booth et al., 2002; Faccini, 2014). Hagen (2002) conducts an analysis of the wage differential, by controlling for self-selection on unobservables through the application of a dummy endogenous variable model. The results obtained indicate an even greater difference with respect to wage between temporary and permanent contracts. The fact that the wages between the two types of contracts are further apart after taking self-selection based on unobservables into account, may lead to an interpretation of the fixed-term

contract as a sorting instrument, since job seekers with certain characteristics select themselves into these contracts.

2.2

Self-Selection and Risk Attitude

The presence of a temporary contract with its purpose as a screening device and the arising self-selection of individuals into these kinds of contracts give rise to the following question: Which characteristics have an influential role concerning self-selection? By taking the standard eco-nomic approach of expected utility maximizing individuals, it is evident that the utility obtained from such a working contract is partly determined by the individual’s risk attitude. The more risk tolerant an individual is, the less harmful is the uncertainty that emerges from a contract that is preset to be terminated with a certain probability after a particular period of time. So, does the personal willingness to take risks affect the choice of contract type?

Bellemare and Shearer (2010) investigated the issue of job selection based on risk preference. The authors were able to show that occupations with higher levels of income risk are practiced by individuals with an above average risk tolerance. Similarly a lab experiment by Prasad and Salmon (2013) reports the same sorting behaviour on the basis of risk. Less risk averse subjects do select themselves into riskier tasks. Hence, empirical patterns are present according to which personal risk attitude is influential in labour market decisions.

2.3

Risk Preferences

Risk attitude as the key variable in this analysis was previously assumed to be a stable individual characteristic (Stigler and Becker, 1977), meaning that environmental influence or social expe-rience do not trigger changes in the personal attitude towards risk. Much empirical research has been conducted to assess whether risk tolerance is subject to changes during a lifetime. In their study Cesarini et al. (2009) take advantage of a twin design to evaluate the variation of risk preferences. Their findings indicate that risk taking behaviour is to a great extent heritable, but is rather changing across time then stable. The variation can to a great extent be explained by genetics whereas environmental influences as a source of variation do only have a modest role. These results indicate that risk preferences are partly innate and heritable but are subject to changes over time. Dohmen et al. (2017) and Schurer (2015) show that risk tolerance declines with advancing age. The latter study additionally notes differences emerging between different levels of income and education. According to Guiso and Paiella (2008) an individual’s mone-tary endowment is influential regarding risk aversion. Moreover, the environment affects risk tolerance to a great extent. For instance, the authors found that income uncertainty is associated with an increasing degree of risk aversion. Given these literature findings it can be assumed that risk attitude constitutes a personal characteristic that is rather changing than stable across a

life-time. The actual level of risk attitude and a part of its development may probably be determined by birth and early life influences, however the exposure to environmental influences like social experiences, different degrees of education or income fluctuations can have a crucial effect on the personal attitude towards risk. Besides being exposed to long-term influences, short-term changes, for instance, uncertainty regarding income has a not negligible effect on risk prefer-ences.

Similar to this study, the assessment of personal risk attitudes often relies on survey questions. A potential concern about such a hypothetical, self-reported and non-incentivized question is whether it properly reflects the actual risk attitude. Several studies demonstrate the reliability of surveyed risk attitudes (Anderson and Mellor, 2008; Donkers, Melenberg, and Van Soest, 2001). A comparison between real-stake lottery decisions and the risk attitude obtained from a survey clearly shows that the self-indicated level of risk attitude is highly correlated with actual risky behaviour (Dohmen et al., 2011). Thus, the difference between indicated and ”real” level of risk aversion can be assumed as marginal and not decisive for the purpose of this study.

Therefore, based on the theoretical considerations , empirical findings and the assumption of temporally changing risk preferences the following hypothesis will be evaluated in the re-mainder of this thesis.

H1: A higher level of risk attitude has a positive influence on the likelihood of being in a temporary contract.

3

Data and Methodology

3.1

Institutions Germany

Since the 1980s the share of fixed-term contracts in many European countries increased rapidly (Booth et al., 2002; Schömann and Kruppe, 1994). Germany is no exception to this develop-ment. This is mostly due to the fact that the German legal system did not put many restrictions on the application of a fixed-term contract. This means, no specific reason for applying a temporary contract needed to be brought up by the employer. With the beginning of the year 2001 a new law referring to temporary contracts came into force. Its purpose was to restrict the excessive use of fixed-term contracts, since it constitutes a harm towards the employee’s long-term job security. In detail, the use of a temporary contract, without a very specific reason, is now only allowed if the employee’s working duration does not exceed two years or the employee was not employed at the same employer in the past (Bundesministerium für Arbeit und Soziales, 09.05.2018). Although, this leaves not much leeway to an employer, the presence of the two

mentioned situations does not prohibit the application of temporary contracts fully. The Ger-man legal systems still allows for fixed-term contracts if exceptional reasons exist, which may be interpreted broadly. Therefore, the legal system in Germany does not put many restrictions on the assignment of fixed-term contracts. It is assumed that a firm’s decision to offer a specific type of contract is independent of the employer’s legal framework in the considered case.

3.2

Data

The empirical analysis of the relationship between risk attitude and the duration of the working contract is conducted with data from the German Socio-Economic Panel (SOEP). This longitu-dinal data set originates from an annual conducted household survey representative for German private households. It consists of around 30,000 individuals in almost 11,000 households since 1984 and contains an extensive set of questions regarding topics concerning, inter alia, demo-graphics, employment and earnings. Most importantly, it allows observing the duration of the current employment contract and the ”personal willingness to take risks”, which will serve as a proxy for personal risk attitude in the following. In addition, the information gathering re-garding the respondents’ ”willingness to take risk” is only conducted irregularly before 2008. Since then the question was included consecutively. Thus, the considered time span consists of the nine consecutive survey years 2008 to 2016. This horizon was chosen, firstly to avoid an overlap with the law change in 2001 and the emerging effects on the application of fixed-term contracts and secondly because the information regarding the ”personal willingness to take risks” is obtained consecutively just since 2008. This provides a gapless panel data set for nine consecutive years. The finally used data set only comprises the individuals in the range of em-ployable age (25 - 66). Apprentices are dropped from the sample as well, since they constitute a specific environment for the application of temporary contracts. Apprentices always receive a temporary contract for the duration of their training period, which last between two and three years. Thus, apprentices are always accompanied by temporary contracts. Contract type does therefore no play a role in job selection and apprentices are not of interest for this analysis The main analysis is conducted with a full sample containing permanently, temporary, and self-employed as well as unself-employed individuals. Additionally, a sample restricted to the two types of contractually employed individuals is consulted since especially self-employed people are expected to show a different risk-taking behaviour compared to the rest of the sample (Cramer et al., 2002). Nonetheless, the share of observations of self-employed people in the sample is small with 8,457 (6.5%) and should not have a huge effect on the control group. In contrast the amount of observations of individuals in unemployment is with 33,337 not insignificant. Hence, the analysis is executed with the full panel data set consisting of 130,753 observations and the restricted case with 87,959 observations. Both samples contain 9,684 observation of individuals employed in temporary contracts. This corresponds to a share of 7% in the full and 11% in the

restricted sample.

The main variable of interest is the attitude towards risk which the respondents indicate on a scale from [0] - ”Not willing to take risks” to [10] - ”Very willing to take risks”. This indica-tion will serve as the only proxy for the actual risk attitude. The indicated risk preferences can hardly be clustered in terms of ”risk averse”, ”risk neutral” or ”risk seeking” individuals, since the offered scale is subject to individual interpretations. Due to different understandings of the measurement system, results regarding the actual levels of risk tolerance have to be interpreted with caution. All variables besides Age, #ofChildren, and Edu.Years are dichotomous variables. Temporary the main variable of interest, next to Risk, receives the value 1 if an individual is employed in a temporary contract in year t and zero otherwise. The variables Permanent, Self-employed, and Unemployed are defined accordingly.

The summary statistics in Table 1 show that the average level of risk attitude reported in the full sample is 4.58 compared to 4.68 in the restricted sample . Two-sample T-tests reveal that the disparity between the risk attitude for individuals in temporary contracts (4.93) and the rest of the sample is significantly different from zero, which holds for both of the considered samples. This gives a first indication of the expected pattern with respect to risk attitude and contract type. The descriptive statistics of the used control variables are in line with previous analyses of individuals’ characteristics employed in fixed-term contracts (Schömann and Kruppe, 1994). The share of 86% Germans, in temporary contracts is significantly lower than in the rest of the sample. Men represent 42% of the employees in fixed-term contracts compared an equal distribution of men and women in permanent contract for the restricted sample. The number of women exceeds the one of men in temporary contracts. The age averages for the different groups of employable individuals indicate a more frequent presence of young people in tempo-rary contracts. The average age of individuals in fixed-term contract lies with 39.01 years more than five years below the mean age in the control groups. Additionally, the descriptives show an unequal distribution between West and East Germany. 78% of the sample is domestic in the western part of Germany, whereas considering exclusively individuals in temporary contracts only 75% do live in western Germany. Furthermore, the years of education and the share of individuals with a university degree suggest that average education level is higher in fixed-term contracts.

3.3

Methodology

In order to investigate the relationship between an individual’s attitude towards risk and working contract, panel data is used to estimate a linear probability model. The model is computed with a set of exogenous controls consisting of ethnicity, gender, age, and region (West and East Germany) under the assumption that movement is fully exogenous. The set of extended controls

Table 1: Summary Statistics

All Contractually Employed

Overall Temp.=0 Overall Temp.=0 Temp.=1

Explanatory Var. Risk 4.58 4.56 4.68 4.65 4.93∗ (2.33) (2.33) (2.25) (2.23) (2.33) Groups Temporary 0.07 0.00 0.11 0.00 1.00∗ (0.26) (0.00) (0.31) (0.00) (0.00) Permanent 0.60 0.65 0.89 1.00 0.00∗ (0.49) (0.48) (0.31) (0.00) (0.00) Self-employed 0.06 0.07 0.00 0.00 0.00 (0.25) (0.25) (0.00) (0.00) (0.00) Unemployed 0.26 0.28 0.00 0.00 0.00 (0.44) (0.45) (0.00) (0.00) (0.00) Exogenous Controls German 0.89 0.90 0.92 0.93 0.86∗ (0.31) (0.30) (0.27) (0.26) (0.34) Age 46.14 46.71 44.91 45.64 39.01∗ (11.1) (11.9) (10.0) (9.71) (10.5) Male 0.46 0.46 0.49 0.50 0.42∗ (0.50) (0.50) (0.50) (0.50) (0.49) West 0.78 0.78 0.78 0.78 0.75∗ (0.42) (0.42) (0.41) (0.41) (0.43) Extended Controls Married 0.65 0.67 0.65 0.66 0.49∗ (0.48) (0.47) (0.48) (0.47) (0.50) Divorced 0.11 0.11 0.11 0.11 0.10 (0.31) (0.31) (0.31) (0.31) (0.30) #ofChildren 0.77 0.77 0.77 0.76 0.82∗ (1.10) (1.10) (1.03) (1.03) (1.07) Intermediate 0.56 0.57 0.56 0.57 0.45∗ (0.50) (0.49) (0.50) (0.49) (0.50) Highschool 0.41 0.40 0.43 0.41 0.53∗ (0.49) (0.49) (0.49) (0.49) (0.50) Vocational 0.69 0.69 0.72 0.73 0.58∗ (0.46) (0.46) (0.45) (0.44) (0.49) University 0.25 0.25 0.28 0.27 0.31∗ (0.43) (0.43) (0.46) (0.44) (0.46) Edu.Years 12.54 12.50 13.00∗ 12.81 13.00∗ (2.78) (2.76) (2.74) (2.71) (3.02) N 130753 121069 87959 78275 9684

Standard deviations in parentheses

Two-sample t-test on difference of means:∗p < 0.01, compared to both samples (full and restricted)

takes education and family characteristics into account. However, these estimates have to be treated with caution as full exogeneity cannot be assumed. Due to the assumption of changing risk preferences and the potential effect of situational circumstances, the use of a lagged value of risk attitude is considered. The individual risk attitude of the previous period ensures that risk attitude influences the type of contract and not the other way round. The first part of the analysis strategy to estimate the relation between risk attitude and contract type consists of the probability model with the sequential inclusion of the lagged value of risk attitude, exogenous controls and potential endogenous controls and does not include worker fixed effects. The final model can be written as:

Temporaryit =β0+β1Riskit+β2Riskit−1+ρXit+αi+ε

In this regression Temporaryit is the variable indicating whether the individual holds a tempo-rary contract, Riskit and Riskit−1denote the risk attitude in period t and t−1, respectively. Xit is the set of applied controls andαi contains the worker fixed effects. The analyses are executed for the full sample as well as for the restricted sample excluding the self-employed and unem-ployed population.

As the summary statistics demonstrate a heterogeneity in contract types for different popula-tion groups and risk aversion is not expected to be homogeneously dispersed among different groups, the main results are followed up by a heterogeneity analysis on the restricted sample to examine if the potential effect of the personal willingness to take risk on contract types varies for different parts of the population. Therefore, the sample is split analogously to the previously applied exogenous and extended control variables. Additionally, a heterogeneity analysis with respect to wage spreads is considered. Therefore, the sample is divided in industry sectors with a great and a small wage spread.

4

Results

4.1

Main Results

Table 2 contains the linear probability estimations with ”being in a temporary contract in pe-riod t” as dependent variable and the attitude towards risk as main explanatory variable. Since the dependent variable is dichotomous assigned with the value 1 if the individual i is holding a temporary contract in year t, a positive coefficient would indicate that with an increasing risk attitude (equal to ”decreasing risk aversion”) the probability of being in a temporary contract increases. The estimations are identically conducted for both, the full and the restricted sample. The first finding that needs to be mentioned is that apparently there are no tremendous differ-ences between the two samples, neither regarding the coefficient size nor the significance. The

T able 2: Linear Probability Model Dependent V ar: T emporary t All: Incl. Self-employed & Unemployed Restricted: Contractually Employed (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (1 1) (12) Risk t 0.0047 ∗∗∗ 0.0030 ∗∗∗ 0.0023 ∗∗∗ 0.0022 ∗∗∗ 0.0006 0.0053 ∗∗∗ 0.0032 ∗∗∗ 0.0030 ∗∗∗ 0.0028 ∗∗∗ 0.0009 (0.0003) (0.0004) (0.0004) (0.0004) (0.0005) (0.0005) (0.0007) (0.0007) (0.0007) (0. 0007) Risk t− 1 0.0038 ∗∗∗ 0.0022 ∗∗∗ 0.0014 ∗∗∗ 0.0012 ∗∗∗ 0.0000 0.0040 ∗∗∗ 0.0022 ∗∗∗ 0.0020 ∗∗∗ 0.0017 ∗∗∗ -0.0001 (0.0004) (0.0004) (0.0004) (0.0004) (0.0005) (0.0005) (0.0007) (0.0007) (0.0007) (0.0006) Age -0.0040 ∗∗∗ -0.0041 ∗∗∗ -0.0044 ∗∗∗ -0.0055 ∗∗∗ -0.0053 ∗∗∗ -0.0082 ∗∗∗ (0.0001) (0.0001) (0.0004) (0.0001) (0.0002) (0.0005) German -0.01 10 ∗∗∗ -0.0180 ∗∗∗ -0.0360 ∗∗∗ -0.0430 ∗∗∗ (0.0036) (0.0036) (0.0055) (0.0055) Male -0.0120 ∗∗∗ -0.0120 ∗∗∗ -0.0330 ∗∗∗ -0.0300 ∗∗∗ (0.0017) (0.0017) (0.0023) (0.0024) W est -0.0140 ∗∗∗ -0.01 10 ∗∗∗ -0.0230 ∗∗∗ -0.0200 ∗∗∗ (0.0020) (0.0020) (0.0029) (0.0029) #ofChildren -0.0094 ∗∗∗ -0.0120 ∗∗∗ -0.0047 ∗∗∗ -0.0007 (0.0010) (0.0024) (0.0014) (0.0030) Married -0.0210 ∗∗∗ -0.0150 ∗∗∗ -0.0340 ∗∗∗ -0.0240 ∗∗∗ (0.0019) (0.0047) (0.0027) (0.0057) University 0.0230 ∗∗∗ 0.0230 ∗∗∗ (0.0020) (0.0026) Const. 0.053 ∗∗∗ 0.050 ∗∗∗ 0.044 ∗∗∗ 0.27 ∗∗∗ 0.29 ∗∗∗ 0.29 ∗∗∗ 0.085 ∗∗∗ 0.079 ∗∗∗ 0.073 ∗∗∗ 0.39 ∗∗∗ 0.41 ∗∗∗ 0.49 ∗∗∗ (0.0015) (0.0018) (0.0019) (0.0061) (0.0069) (0.0180) (0.0024) (0.0027) (0.0029) (0.0093) (0.0099) (0.0220) Indv . FE ✓ ✓ N 130753 92587 92587 92587 92587 92587 87959 64394 64394 64394 64394 64394 Standard errors in parentheses, Specifications (1) -(5) and (7) -(1 1) are computed with robust standard errors ∗ p < 0 .10 , ∗∗ p < 0 .05 , ∗∗∗ p < 0 .01

estimations (1) and (7) do show a positive significant relation of 0.0047 (0.0053) 2, which is consistent with the expected direction of the effect.

Taking a look at the size of the standard deviation of the stated ”personal willingness to takes risk”, which is 2.33, does suggest that there is variation present in the subjects’ risk aversion, in line with the assumption of non-stable risk preferences. The circumstance mentioned by Guiso and Paiella (2008) that income uncertainty, such as in a temporary contract, can have an in-fluence on the level of risk aversion leads to the consideration of reverse causality. Hence, in the specifications (2) and (8) the explanatory variable is replaced with the lagged value of risk attitude. The size of the coefficient decreases slightly to 0.0038 (0.0040) but the significance remains stable at the one percent level. The inclusion of both variables, risk attitude in period t and the lagged value do still display a significant relationship. The coefficient of Risktdecreases to 0.0030 (0.0030) compared to the first specification with a coefficient size of 0.0047 (0.0053). A very evident explanation for these results is that Risk_t−1does capture part of the explanatory effect of Riskt, especially as the combined effect is close to the initial estimation. Moreover, these two variables may be primarily highly correlated. The first few results do hint towards a significant relationship, although the effect sizes are small. An increase of risk attitude about one standard deviation (2.33) would results in an enlargement of the probability of having a temporary contract by approximately one percent.

Note that by controlling for exogenous effects emerging from ethnicity, age, gender or region, the coefficient in the full and the restricted sample are altered just slightly downwards, whereas the significance of the estimates is not harmed. The signs of the control variables across all specifications follow the expected directions. In particular, becoming one year older decreases the probability of having a temporary contract by 0.4 % (0.5%), meaning that being 20 years older results in an decrease by almost 8-10%. Native Germans are 1.1% (3.6%) less likely to be employed in temporary basis. Similar men and residents of western Germany are more likely employed in a permanent contract. Even more surprising is the insight after the inclusion of control variables, for which full exogeneity cannot be assumed. However, the coefficients of Riskt decreases from 0.0023 (0.0030) to 0.0022 (0.0028) and stay significant at the one percent level due to the inclusion of the potential endogenous controls. The number of children and being married decreases the probability of having a fixed-term contract as well. In contrast, the possession of an university degree raises the chances to have a temporary contract by 2.3% (2.3%). The results obtained from the linear probability estimation without worker fixed effects indicates a relationship between personal risk attitude and working contract. However, the size is negligibly small. This holds for the two different samples.

2_{In the following the first value corresponds to the full sample and the value in parentheses to the restricted} sample, unless mentioned differently

The specifications (1) to (5) and (7) to (11) indicate a positive relation that is almost zero in size. But due to the application of individual fixed effects, the previous effect vanishes, as it can be obtained from the estimates in the columns (6) and (12). It seems that a lot of time-invariant factors are not captured by the initially used controls. The fixed effects regressions still do not differ much for the full (0.0006) and the restricted (0.0009) sample. Given the obtained results and the assumption of varying risk preferences the theoretical consideration, that risk attitude plays a role in the self-selection process into fixed-term contracts seems not to have an empirical basis and hypothesis H1 can be rejected. Neither does the omission of self-employed and unemployed subjects make a huge difference. In case the assumption about stable risk preference does not hold, it is obvious that the fixed effect coefficients are insignificant since no informative variation can be captured. Even so, the results without worker fixed effects point towards a relation close to zero. Therefore, in the following heterogeneity and robustness analyses will be conducted to investigate the potential issue of reverse causality and to analyze whether the role of risk attitude differs for certain population groups. The following analyses are exclusively conducted with the restricted sample.

4.2

Heterogeneity

In the following, heterogeneity test are computed in order to investigate whether the finding of no significant or a weak relation between risk attitude and contract type is due to the non-existence or heterogeneity within the sample. It turns out that for some population groups even after the inclusion of fixed effects, small positive and significant effects can be determined. The underlying idea of heterogeneity driving the results displayed in Table 2 is that the labour market does not always offer a free choice regarding contract type. Some characteristics may influence to what extent an individual can choose a job freely and thereby influence the potential role of risk attitude in this regard. One may consider that some individuals have less employ-ment opportunities because of characteristics, like age, gender or education. Thus, for groups with restricted job offers decisions in favor or against a certain occupation are determined by contextual differences, compared to individuals with various and more dispersed job options. These differences can determine the influence of risk attitude when it comes to the working contract.

Firstly, the effect sizes for young and old people are compared by splitting the sample at the age of 40. The difference between these groups mainly originates from the fact that a fraction of the young population is faced with the first job entry and moreover, does not have as much experience and references compared to the older share of the population. This leads to a more intensively needed and more often applied screening procedure. Hence, a young job applicant probably does not have the same freedom of choice regarding the contract as an experienced applicant of advanced age. Therefore, risk may play a subordinate role for the younger share of

T able 3: Heterogeneity I: (only contractually employed) Dep.V ar: T emporary t Age < = 40 Age > 40 German Non-German Male Female W est East (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (1 1) (12) (13) (14) (15) (16) Risk t 0.0034 ∗∗ 0.0022 0.0023 ∗∗∗ 0.0002 0.0027 ∗∗∗ 0.0009 0.0060 ∗∗ 0.0005 0.0021 ∗∗ 0.001 1 0.0039 ∗∗∗ 0.0009 0.0033 ∗∗∗ 0.0015 ∗∗ 0.0017 -0.0010 (0.0015) (0.0016) (0.0007) (0.0007) (0.0007) (0.0007) (0.0024) (0.0029) (0.0009) (0.0009) (0.0010) (0.0010) (0.0007) (0.0007) (0.0015) (0.0014) Risk t− 1 0.0015 0.0008 0.0016 ∗∗ -0.0006 0.0023 ∗∗∗ 0.0004 -0.0010 -0.0055 ∗ 0.0013 -0.0001 0.0028 ∗∗∗ 0.0000 0.0018 ∗∗ 0.0000 0.0027 ∗ -0.0003 (0.0015) (0.0016) (0.0007) (0.0007) (0.0007) (0.0007) (0.0025) (0.0029) (0.0009) (0.0009) (0.0010) (0.0010) (0.0007) (0.0007) (0.0015) (0.0014) Age -0.0170 ∗∗∗ -0.0210 ∗∗∗ -0.001 1 ∗∗∗ -0.0044 ∗∗∗ -0.0055 ∗∗∗ -0.0082 ∗∗∗ -0.0052 ∗∗∗ -0.0130 ∗∗∗ -0.0049 ∗∗∗ -0.0072 ∗∗∗ -0.0062 ∗∗∗ -0.0095 ∗∗∗ -0.0055 ∗∗∗ -0.0083 ∗∗∗ -0.0054 ∗∗∗ -0.0078 ∗∗∗ (0.0007) (0.0013) (0.0002) (0.0005) (0.0001) (0.0005) (0.0006) (0.0029) (0.0002) (0.0006) (0.0002) (0.0007) (0.0002) (0.0005) (0.0003) (0.0010) German -0.036 ∗∗∗ -0.048 ∗∗∗ -0.039 ∗∗∗ -0.034 ∗∗∗ -0.034 ∗∗∗ -0.088 ∗∗ (0.0100) (0.0063) (0.0074) (0.0083) (0.0056) (0.0360) Male -0.046 ∗∗∗ -0.028 ∗∗∗ -0.033 ∗∗∗ -0.032 ∗∗∗ -0.034 ∗∗∗ -0.027 ∗∗∗ (0.0055) (0.0023) (0.0024) (0.01 10) (0.0026) (0.0052) W est -0.030 ∗∗∗ -0.017 ∗∗∗ -0.022 ∗∗∗ -0.076 ∗∗ -0.026 ∗∗∗ -0.020 ∗∗∗ (0.0066) (0.0029) (0.0029) (0.0360) (0.0039) (0.0041) Const. 0.81 ∗∗∗ 0.88 ∗∗∗ 0.17 ∗∗∗ 0.29 ∗∗∗ 0.36 ∗∗∗ 0.46 ∗∗∗ 0.43 ∗∗∗ 0.74 ∗∗∗ 0.34 ∗∗∗ 0.41 ∗∗∗ 0.41 ∗∗∗ 0.54 ∗∗∗ 0.37 ∗∗∗ 0.46 ∗∗∗ 0.44 ∗∗∗ 0.48 ∗∗∗ (0.028) (0.045) (0.012) (0.024) (0.008) (0.022) (0.047) (0.130) (0.013) (0.028) (0.014) (0.032) (0.010) (0.024) (0.039) (0.044) Indv . FE ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ N 19665 19665 44729 44729 60301 60301 4093 4093 31432 31432 32962 32962 49548 49548 14846 14846 Standard errors in parentheses, Specifications (1), (3), (5), (7), (9), (1 1), (13) and (15) are computed with robust standard errors ∗p < 0 .10 , ∗∗ p < 0 .05 , ∗∗∗ p < 0 .01

the population and a lower effect can be expected.

The specifications (1) and (3) of Table 3 indicate a significant relation of 0.0034 and 0.0023. In contrast, the fixed-effects estimations (2) and (4) of personal risk attitude on the probability of having a temporary contract is insignificant for both parts of the split sample.

The necessity of including ethnicity and gender differences into the analysis follows a similar argumentation as the consideration of age. In general, Non-Germans and women are employed more frequently in fixed-term contracts as it can be obtained from the descriptives (Table 1). Potential language barriers or lack of education may lead to a higher share of non-Germans in temporary contracts than in permanent contracts (Schömann and Kruppe, 1994). For women the detected imbalance can emerge from the frequency of maternity breaks or short-time work due to childcare. Therefore, it is expected that the willingness of women and Non-Germans to take a temporary job is much higher generally. Consequently risk attitude presumably plays a more crucial role for men and Germans since their decisions are not influence by the previously mentioned characteristics.

The distinct regressions for Germans and Non-Germans display a unexpected relationship. Al-though, the coefficients of Riskt in both specifications are not significant, the coefficient of Riskt−1 for Non-Germans have a small but negative effect. Irregardless of the size, this would mean that a increased risk tolerance leads to a lower probability of having a temporary contract for foreigners. The direction is the exact opposite of the expected relationship. Still the coeffi-cient is close to zero and is only relevant if the assumption of informative variation in personal risk attitudes holds. The results for Gender are not different from the outcomes of the whole sample.

In the following a regionally separated approach is applied. The argumentation relies on the disparity between the labour market conditions in East and West Germany. The unemployment rate in West Germany in April 2018 was 5.5 percent, whereas it has been 7.1 percent in East Germany. This pattern reaches back to the German reunification (Bundesagentur für Arbeit, 01.06.2018). This imposes huge difference in labour force demand and supply. Hence, the application rate of fixed-term contracts is higher in East Germany and in addition, the greater amount of labour supply for the given demand exposes an individual to take a job independent of the offered contract. Thus, the impact of risk attitude is expected to be more essential in West Germany.

Specification (14) does display a positive significant effect at the five percent level, even for the fixed effect estimation. The effect of 0.0015 is not huge in size, but does show that a relation exists, which is, due to heterogeneity, not present in the whole sample. The poor labour market conditions in East Germany seem to mitigate the role of risk attitude in the self-selection mech-anism. In contrast to all specification without worker fixed effects before, no significant effect of risk attitude on having a temporary contract can be detected for East Germany

T able 4: Heterogeneity II: (only contractually employed) Dep.V ar: T emporary t Children No Children Partner No Partner Univ . degree No Univ . degree (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (1 1) (12) Risk t 0.0036 ∗∗∗ 0.0006 0.0026 ∗∗∗ 0.001 0.0026 ∗∗∗ 0.0015 ∗∗ 0.0032 ∗∗ -0.0009 0.0046 ∗∗∗ 0.0030 ∗∗ 0.0027 ∗∗∗ 0.0004 (0.001 1) (0.001 1) (0.0008) (0.0008) (0.0007) (0.0007) (0.0013) (0.0013) (0.0014) (0.0013) (0.0007) (0.0008) Risk t− 1 0.0035 ∗∗∗ -0.0002 0.0009 -0.00008 0.0020 ∗∗∗ 0.0004 0.0010 -0.0004 0.0003 0.0010 0.0025 ∗∗∗ -0.0003 (0.001 1) (0.001 1) (0.0008) (0.0008) (0.0007) (0.0007) (0.0013) (0.0013) (0.0014) (0.0013) (0.0008) (0.0007) Age -0.0055 ∗∗∗ -0.0096 ∗∗∗ -0.0059 ∗∗∗ -0.0073 ∗∗∗ -0.0036 ∗∗∗ -0.0066 ∗∗∗ -0.0071 ∗∗∗ -0.01 10 ∗∗∗ -0.0078 ∗∗∗ -0.0120 ∗∗∗ -0.0045 ∗∗∗ -0.0066 ∗∗∗ (0.0003) (0.0009) (0.0002) (0.0006) (0.0002) (0.0005) (0.0002) (0.0010) (0.0003) (0.0008) (0.0002) (0.0006) German -0.039 ∗∗∗ -0.044 ∗∗∗ -0.044 ∗∗∗ -0.048 ∗∗∗ -0.062 ∗∗∗ -0.031 ∗∗∗ (0.0074) (0.0083) (0.0060) (0.0130) (0.0130) (0.0060) Male -0.061 ∗∗∗ -0.013 ∗∗∗ -0.033 ∗∗∗ -0.030 ∗∗∗ -0.036 ∗∗∗ -0.031 ∗∗∗ (0.0038) (0.0030) (0.0026) (0.0046) (0.0047) (0.0027) W est -0.025 ∗∗∗ -0.019 ∗∗∗ -0.017 ∗∗∗ -0.023 ∗∗∗ -0.0087 ∗ -0.030 ∗∗∗ (0.0051) (0.0034) (0.0032) (0.0054) (0.0052) (0.0035) Const. 0.39 ∗∗∗ 0.51 ∗∗∗ 0.42 ∗∗∗ 0.44 ∗∗∗ 0.30 ∗∗∗ 0.38 ∗∗∗ 0.50 ∗∗∗ 0.62 ∗∗∗ 0.53 ∗∗∗ 0.66 ∗∗∗ 0.34 ∗∗∗ 0.39 ∗∗∗ (0.015) (0.037) (0.013) (0.028) (0.01 1) (0.025) (0.018) (0.041) (0.020) (0.038) (0.010) (0.025) Indv . FE ✓ ✓ ✓ ✓ ✓ ✓ N 26655 26655 37739 37739 42179 42179 22215 22215 18091 18091 46303 46303 Standard errors in parentheses, Specifications (1), (3), (5), (7), (9) and (1 1) are computed with robust standard errors ∗p < 0 .10 , ∗∗ p < 0 .05 , ∗∗∗ p < 0 .01

Further heterogeneity test were raised to investigate whether distinguished estimations on the basis of potentially endogenous variables yields more insights about the effect of risk pref-erences on the probability of having a fixed-term contract. Additionally, it should illustrate if heterogeneity in the sample is the reason for not detecting an overall effect. Nevertheless, the results presented in Table 4 should be treated with caution.

The first two distinctions are made with respect to domestic characteristics. The presence of a relationship or even children often imposes more responsibility on individuals. This risen responsibility could translate into a different behaviour when it comes to job selection. The additional burden of being responsible for someone and maybe an additionally decreased risk tolerance may switch the focus onto a long-term perspective. This expanded horizon changes the role of risk attitude, and it becomes a more influential aspect.

The results do show the expected positive and significant effect for people in partnerships after including fixed effects (Table 4 (6)). Apparently, the finding does only hold for partners and not with respect to being a parent. The non-existence could be due to a high correlation of both characteristics and the contamination of the ”no children” control group by individuals with a partner, since most likely the greatest share of the sample with children also is in a partnership. Finally, heterogeneity can be presumed between low and high educated individuals. Being equipped with a university degree opens up more opportunities than lower education does. To be provided with the possibility to choose between more alternatives, forms the need for more criteria. In this case the contract type becomes a more important factor and the influence of risk attitude increases simultaneously. In addition, a higher degree gives rise to more job movement, so that those individuals are confronted more often with employment decisions.

The estimation (10) does confirm the constructed hypothesis that risk preference have a greater impact for highly educated individuals. The coefficient of 0.0030, which is significant at the five percent level demonstrates that for highly educated individuals an increased level of risk attitude does lead to a higher probability of being in a temporary contract. This relationship is not present for the lower educated share of the population.

Finally, the consideration arises that the impact of risk attitude varies with the monetary prospect. The amount of potential future earnings may determine to which extent risk is tolerated. Picchio (2008) and Hagen (2003) demonstrated that temporary contracts can be considered as stepping stones towards permanent employment. Considering the fixed-term contract as a stepping stone incorporates much of the reasoning to bear the uncertainty of a fixed-term contract. However, due to the wage gap between the two contract types a higher compensation in a permanent con-tract can be expected. In case the expected future earnings are high, the employee is already partly compensated by the great monetary prospect and a high level of risk tolerance may not be required. Conversely, a low future compensation would bring more attention to the role of risk preferences. Therefore, the sample is clustered into 30 industry sectors which are listed in

the Appendix. These sectors are split according to the size of the wage spread in yearly labour income. To determine the wage spread the 95th percentile of earnings in each sector is used. The wage spread reaches from a minimum of 29,000€ up to 160,000€ with a mean of 65.393€. The 95th percentile is used to exclude huge outliers in yearly earnings.

Table 5: Heterogeneity: Wage Spread

Dep. Variable: Temporaryt

Wage Spread <=65,000€ >65,000€ (1) (2) (3) (4) Riskt 0.0032∗∗∗ 0.0021∗∗ 0.0028∗∗∗ 0.00040 (0.0008) (0.0008) (0.0010) (0.00010) Riskt−1 0.0016∗ 0.0003 0.0026∗∗ 0.0002 (0.0009) (0.0008) (0.0010) (0.0010) Age -0.0049∗∗∗ -0.0070∗∗∗ -0.0063∗∗∗ -0.0110∗∗∗ (0.0002) (0.0006) (0.0002) (0.0007) German -0.041∗∗∗ -0.032∗∗∗ (0.0070) (0.0090) Male -0.048∗∗∗ -0.0087∗∗ (0.0031) (0.0036) West -0.024∗∗∗ -0.022∗∗∗ (0.0038) (0.0043) Const. 0.38∗∗∗ 0.40∗∗∗ 0.42∗∗∗ 0.59∗∗∗ (0.012) (0.028) (0.014) (0.033) N 35699 35699 28695 28695

Standard errors in parentheses

∗_{p < 0.10,}∗∗_{p < 0.05,}∗∗∗_{p < 0.01}

The results display a similar positive significant effect for risk attitude in the regressions without worker fixed effects. Moreover, the worker fixed effect regression in specification (2) of Table 5 for industry sectors with a lower wage spread has a positive significant effect of 0.0021 at the five percent level. The group of industry sectors with a large wage spread does not show a significant effect of risk attitude on the probability of having a temporary contract for the fixed effect regression. Even though, a significant effect can be detected for industries with smaller income prospects, the size of the coefficient is still close to zero. The heterogeneity analysis is in line with the consideration that heterogeneity within the population leads to the finding of no relation between the personal willingness to take risk and the probability of being in a fixed-term contract. So finally, some sort of relationship is present, but it only affects certain groups of the population and the size of the effect is small.

4.3

Reverse Causality

Table 6: Reverse Causality

Only Contractually Employed

Dependent Var: Riskt

(1) (2) (3) (4) (5) (6) Temporaryt 0.27∗∗∗ 0.15∗∗∗ 0.14∗∗∗ 0.13∗∗∗ 0.056 (0.025) (0.044) (0.043) (0.043) (0.038) Temporaryt−1 0.25∗∗∗ 0.16∗∗∗ 0.16∗∗∗ 0.15∗∗∗ 0.0052 (0.032) (0.041) (0.040) (0.040) (0.036) Age -0.013∗∗∗ -0.010∗∗∗ 0.093∗∗∗ (0.001) (0.001) (0.004) German 0.045 0.027 (0.043) (0.043) Male 0.80∗∗∗ 0.81∗∗∗ (0.018) (0.018) West -0.037∗ -0.034 (0.021) (0.021) #ofChildren 0.049∗∗∗ -0.099∗∗∗ (0.010) (0.023) Married -0.20∗∗∗ -0.16∗∗∗ (0.021) (0.043) University 0.028 (0.019) Const. 4.65∗∗∗ 4.63∗∗∗ 4.63∗∗∗ 4.82∗∗∗ 4.76∗∗∗ 0.55∗∗∗ (0.008) (0.010) (0.010) (0.064) (0.067) (0.17) Indv. FE ✓ N 87959 59416 59416 59416 59416 59416 Standard errors in parentheses

Specifications (1) - (5) are computed with robust standard errors

∗_{p < 0.10,}∗∗_{p < 0.05,}∗∗∗_{p < 0.01}

One may consider that the obtained results are prone to potential reversed causality. Being exposed to a particular environment or situation may have an influence on the personal attitude. For instance, being in a contractual situation which is associated with a particular amount of uncertainty could lead to mechanisms of adaption. More formally, if the risk related to a situation is not likely to change in a short time span, a person may adjust her/his personal risk attitude simply because of the exposure.

In order to rule out this direction of action, the second estimation of the main analysis (Table 2) was conducted with the lagged value of the personal risk attitude, resulting in a coefficient close to initial specification (1). To sufficiently preclude reverse causality, the relation between risk attitude as dependent variable and contract type as explanatory variable is explored. Table 6 shows the estimates of those regressions. Estimation (6) displays a coefficient of 0.056, which is not significant. This corresponds to only contractually employed individuals and includes the initial set of controls and individual fixed effects. Therefore, no indication of reverse causality is present.

5

Discussion

In total, it has been demonstrated that considering either the full or the restricted sample the expected effect between risk attitude and working contract is absent. Also, the heterogeneity analyses does not show such a relationship, although for some groups there is a small positive effect of risk tolerance on having a temporary contract. In this section, the assumptions about risk preferences and the consequential interpretations of the obtained results are discussed.

The strategy of applying worker fixed effects to exclude individual and time-invariant factors fully relies on the assumption of informative variation in the obtained risk preferences. The standard deviation of risk attitude in the data set is with a size of 2.33 not negligible. Assuming that the detected variation is of informative nature makes the fixed effects approach an appro-priate strategy. The results derived with this method lead to the conclusion that the expected relationship between personal risk preferences and working contract is not existent. Although, for some population groups a effect ongoing from risk attitude is detectable, its size is around a quarter of a percent.

Now, consider the case that the variation of risk attitude is not of informative nature, but just noise or measurement error. By doing so the estimates of the worker fixed effects regression would be redundant, since there would be no explanatory power in the variation left to be capture by the fixed effects estimator. Specifications (4), (5), (10) and (11) of Table 2 would then re-flect the relation more properly. Although, the size of the effect would be larger, a coefficient of 0.0028 (specification (11)) is still small and close to zero. The existence of measurement error in the main variable of the analysis could also partly explain the small coefficients, since measure-ment error is mostly accompanied by a bias towards zero. Thus, irregardless of the assumption made with respect to the variation of risk preferences the effect on having a temporary working contract is not present. Another indication of the absence of the expected relationship is that the regressions do not display a tremendous difference between the samples with and without unemployed and self-employed individuals. The negligibility of the self-employed subjects can be explained by its small size compared to the whole sample. However, in contrast the share of unemployed people in the sample seems not neglectable, especially since they display divergent risk tolerances and characteristics. The circumstance that the analysis of two disparate samples leads to closely related results is probably another signal regarding the insignificance of this relation.

Even though no explicit relation between the personal willingness to take risks and being em-ployed on temporary basis was found in either direction, the heterogeneity analysis hints towards a small effect for specific population groups. It seems that for employees in West Germany and employees highly educate the effect is the most prominent. In both cases one explanation can be the different labour market circumstances. Such conditions influencing job search may be

more severe than age, gender or nationality.

This study has some limitations. First, the amount of variation emerging from measure of risk attitude cannot be assumed to be informative. The possibility of just detecting noise instead of real alterations could drive the results towards zero. Second, it is a major challenge to identify the correct reasons for the obtained outcomes since in this environment the complexity of the labour market and individual decision making interfere. The presence of many different influ-ences for various occupations or population subgroups makes it hard to isolate the direct impact emerging from personal risk attitude. Therefore, an in-depth research regarding specific sub-groups, which are expected to be the main carrier of the suggested effect, would be beneficial in the future. Overall, it seems to be the case that the personal risk attitude can play a role when it comes to choices which are decisively influenced by occurring uncertainty for particular parts of the population. It may thus be mostly the case that the decision in favour or against a job does not originate from one’s personal risk attitude, but relies on other criteria.

6

Conclusion

In this study the relationship between personal risk attitude and the duration of a working con-tract is analyzed based on nine survey waves from the German Socio-Economic Panel. Given the assumptions of standard economic theory, patterns should be present according to which more risk tolerant individuals sort themselves more often into fixed-term contracts. Apparently, the influential component of risk preferences for the self-selection into specific kinds of con-tracts has not been the subject of previous studies.

It was hypothesized that a higher level of risk tolerance leads to larger likelihood of being in a temporary contract. However, the obtained results suggest that such a relation is not empirically present. The non-finding of a significant correlation could be due to different influences. One may suspect, that the underlying concept of self-selection based on unobservable characteris-tics, in particular based on risk attitude, does not matter that frequently and can therefore not be detected. Theoretically and empirically the self-selection process has its justification. However, how often that process is fundamental in an individual’s labour market decision is unclear. The result hint towards an infrequent presence of self-selection based on the attitude towards risk. The suspicion that heterogeneity is driving the results can be confirmed on basis of the existence of a significant effect, however the effect sizes are too small to be of importance. Population subgroups, in particular individuals which are highly educated, live in a partnership, work in sectors that do not offer great earning prospects or are domestic in West Germany do show signs of the expected patterns.

proba-bility of having a temporary working contract exists, relies on the discussion if risk preferences are stable across time or not. If risk attitude is assumed to be stable there exists a significant effect, but it is so close to zero to not make a difference. In the opposite case of changing risk preferences, such an effect is only detectable for certain population groups and is still extremely small. Conclusively, it can be stated that according to the obtained results no relevant relation between risk tolerance and working contract is existent.

7

Appendix

Industry sectors by wage spread:

Wage spread lower or equal to 65,000 €:

[1] Agriculture and Forestry [2] Fisheries [3] Food Industry [4] Wholesale [5] Restaurants [6] Trash Removal [7] Not Applicable

Wage spread greater than 65,000 €:

[8] Energy, Water [9] Mining [10] Chemicals [11] Earth, Clay, Stone [12] Iron, Steel [13] Me-chanical Engineering [14] Electrical Engineering [15] Wood, Paper, Print [16] Clothing [17] Con-struction [18] ConCon-struction Relate [19] Other Trans. [20] Financial Institute [21] Insurance [22] Service Industry [23] Education, Sport [24] Health Service [25] Other Services [26] Voluntary, Church [27] Private Household [28] Public Administration [29] Synthetics [30]

8

References

1. Anderson, L.R., and Mellor., J.M., 2008. Predicting health behaviors with an experimental mea-sure of risk preference. Journal of Health Economics 27 (5): 1260–74.

2. Bellemare,C., Shearer,B., 2010. Sorting, incentives and risk preferences: Evidence from a field experiment. Economics Letters 108, 345–348

3. Berton, F., and Garibaldi, P., 2012. Workers and Firms Sorting into Temporary Jobs. The Eco-nomic Journal 122, F125-F154

4. Blanchard, O., and Landier, A., 2002. The Perverse Effects of Partial Labour Market Reform: Fixed-term Contracts in France. The Economic Journal 112, F214 - F244

5. Boockmann, B., Hagen, T., 2007. Fixed-term contracts as sorting mechanisms: Evidence from job durations in West Germany.Labour Economics 15, 984–1005

6. Booth, A.L., Francesconi, M., and Frank, J., 2002. Temporary Jobs: Stepping Stones or Dead Ends?. The Economic Journal 122, F189-F213

7. Cesarini, D., Dawes, C.T., Johannesson, M., Lichtenstein, P., and Wallace, B., 2009. Genetic Variation in Preferences for Giving and Risk Taking. The Quarterly Journal of Economics, 124 No.2, 809 - 842

8. Chuang, Y., and Schechter, L., 2015. Stability of experimental and survey measures of risk, time, and social preferences: A review and some new results. Journal of Development Economics 117: 151–70.

9. Cramer, J.S., Hartog, J., Jonker, N., and Van Praag, C.M., 2002. Low risk aversion encourages the choice for entrepreneurship: an empirical test of a truism. Journal of Economic Behavior & Organization Vol. 48, 29–36

10. Dohmen, T., Huffman, D., Schupp, J., Falk, A., Sunde, U., Wagner, G.G., 2011. Individual risk attitudes: Measurement, determinants, and behavioral consequences. Journal of the European Economic Association Vol 9, No.3, 522-550

11. Dohmen, T., Falk, A., Golsteyn, B.H.H., Huffman, D., and Sunde, U., 2017. Risk Attitudes Across the Life Course. The Economic Journal 127, F95 - F116

12. Dolado, J.J., Garcia-Serrano, L., and Jimeno, J.F., 2002. Drawing Lessons From the Boom at Temporary Jobs in Spain. The Economic Journal 112, F270 - F295

13. Donkers, B., Melenberg, B., and Van Soest, A., 2001. Estimating Risk Attitudes using Lotteries: A Large Sample Approach. Journal of Risk and Uncertainty 22 (2): 165–95.

14. Faccini, R., 2014. Reassessing Labour Market Reforms: Temporary Contracts as Screening De-vice. The Economic Journal 124, F167-F200

15. Guasch,T., Weiss, A., 1981.Self-selection in the labor market. American Economic Review Vol. 71, No. 3, pp. 275-284

16. Guiso, L., and Paiella, M., 2008. Risk Aversion, Wealth and Background Risk. Journal of the European Economic Association Vol.6 No.6, 1009 - 1150

17. Hagen, T., 2002. Do Temporary Workers Receive Risk Premiums? Assessing the Wage Effects of Fixed-term Contracts in West Germany by a Matching Estimator Compared with Parametric Approaches. LABOUR 16 (4), 667–705

18. Hagen, T., 2003. Do Fixed-term Contracts Increase the Long-term Employment Opportunities of Unemployed? ZEW Discussion Paper, No. 03-49

19. Picchio, M., 2008. Temporary Contracts and Transition to Stable Jobs in Italy. LABOUR 22 (Special Issue), 147-174

20. Prasad, K., and Salmon, T.C., 2013. Self Selection and Market Power in Risk Sharing Contracts. Journal of Economic Behavior & Organization 90, pp.71 - pp.86

21. Rosen, S., 1986.Theory of Equalizing Differences. Handbook of Labor Economics Volume I, Chapter 12

22. Schömann, K., Kruppe, T., 1994. Who enters fixed-term contracts: Evidence from East and West Germany. Vierteljahrshefte zur Wirtschaftsforschung Vol. 63, Iss. 1/2, pp. 69-74

23. Schurer, S., 2015. Lifecycle Patterns in the Socioeconomic Gradient of Risk Preference. Journal of Economic Behavior & Organization 119, 482 - 495

24. Stigler, G.J., and Becker, G.S., 1977. De Gustibus Non Est Disputandum. American Economic Review 67 (2): 76–90.

25. Bundesministerium für Arbeit und Soziales:

https://www.gesetze-im-internet.de/tzbfg/__14.html(09.05.2018) viahttp://www.bmas.de/DE/Service/Gesetze/teilzeit-und-befristungsgesetz.html (09.05.2018) 26. Bundesagentur für Arbeit https://statistik.arbeitsagentur.de/nn_4236/SiteGlobals/Forms/Themenauswahl/ themenauswahl-Form.html?view=processForm&resourceId=210342&input_=&pageLocale= de&regionInd=b&year_month=201805&topicId=17590&topicId.GROUP=1&search=Suchen (01.06.2018)