Differences in time and risk preferences between public and private sector workers

(1)

Differences in time and

risk preferences between

public and private sector

workers

Bachelor Thesis

Student: Sanne Eitjes Student Number: 10421785

Supervisor: Simin He University of Amsterdam Faculty of Economics and Business

Academic Year: 2015/2016 Submission date: 29-06-2016

Abstract

This paper sheds light on the literature of time and risk preferences applied to an individuals’ choice for either public or private sector employment. Both Bellante and Link (1981) and Pfeiffer (2011) have found that individuals, who are on average more risk averse, are more likely to select themselves into public sector employment. This relationship has been confirmed in this paper again and can be declared by the higher job security and less volatile wage compensation in the public sector. A novel contribution of this paper is that it investigates whether (higher qualified) public and private sector workers differ in terms of their level of patience. However, no significant evidence has been found for the hypothesis that (higher qualified) public sector workers are more patient than their private sector counterparts. Whereas recent research showed that time and risk preferences are correlated, another goal of the analysis is to find out to what extent the differences in risk and time preferences between public and private sector workers can be declared by these correlations. However, no significant relationship has been found between time and risk. Therefrom, time and risk preferences and selection of workers into the public sector do not form a triangular relationship in this paper.

(2)

Statement of Originality

This document is written by Student Sanne Eitjes who declares to take full responsibility for

the contents of this document. I declare that the text and the work presented in this

document is original and that no sources other than those mentioned in the text and its

references have been used in creating it. The Faculty of Economics and Business is

responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

In 2013, the public sector workforce accounted for over 14% of total employment in Germany (OECD, 2015). Because these people form a substantial fraction of the labour force, understanding the characteristics of these workers has always been of primary importance. In an environment where agency problems have been found to be more severe than in the private sector (Tonin & Vlassopoulos, 2014), it is from upmost importance to attract public sector workers who are capable of dealing with these problems. In order to attract these workers, it is important to know the differences in characteristics of private and public sector workers.

Since both time preferences and risk preferences play a major role in economics and they have been found to be the underlying drivers of the decision process (Wölbert & Riedl, 2013), this paper focuses on the question to what extent public and private sector workers differ from each other in terms of their level of patience and level of risk aversion. The economic importance of this question seems indisputable: if public sector workers are on average more risk averse and more patient than private sector workers, this will influence all aspects of their decision process, including all decisions they make in the workplace (Eckel & Grossman, 2008).

An example that emphasizes the important role of both time and risk preferences is the real-life example of the bankruptcy of the Lehman brothers. The focus of the brothers lied on making ‘quick money’ (making impatient decisions) and on risky investments (risk-taking behaviour). These decisions finally resulted in the downfall of the global financial services firm in 2008 and created a snowball effect that resulted in the infamous financial crisis. The impact of this crisis is still felt nowadays. Conclusive, time and risk preferences have considerable implications for the economy and the functioning of the government.

The question then arises: to what extent can this impatient and risk taking behaviour of the Lehman brothers be generalised to all private sector workers? What is true of the preconception that private sector workers, in general, act more like Nick Leeson, take more risks and act faster? On the other hand: what is true of the preconception that public sector workers reconsider their decisions at least three times and take less risk? The aim of this paper is answering this question, by looking at the extent in which public and private sector workers differ in terms of their level of patience and level of risk aversion. The extent in which public and private workers differ in terms of their level of patience has never been empirically tested before, and is therefore a novel contribution in this paper. In the analysis, the 2006 wave of the German Socioeconomic Panel (GSOEP) is exploited to estimate the probability of being employed in the public sector, using a simple probit model.

The first research question is formulated as follows: to what extent do public sector workers differ in their degree of risk aversion compared to private sector workers? Many studies have already supported the hypothesis that people who are on average more risk averse, are more likely to select themselves into public sector jobs (Bellante & Link, 1981; Pfeifer, 2011; Buurman et al., 2012). This can be reasoned by employment security and working conditions, which are perceived larger and more stable in the public sector than in the private sector (Pfeifer, 2011). In this paper, this hypothesis is confirmed again. Individuals, who are one point more willing to take risks on an eleven-point scale (on a scale from 0 to 10), are less likely to work in the public sector.

The second research question is: to what extent do (higher qualified) public sector workers differ in their degree of patience compared to private sector workers? Previous research into the differences in earnings distributions between the public and private sector found that public sector workers on the higher tail of the wage distribution earn less than their private sector counterparts (Jürges, 2002). As skill levels in these higher tails are normally acquired through a degree or post-compulsory

(5)

2

education (Office for National Statistics, 2014), these people are referred to as higher qualified public sector workers in the rest of this paper. Thus: wage penalties exist for higher qualified workers in the public sector. But why do higher qualified workers accept these lower wages? According to Pfeifer (2011), the demand of private sector firms for higher qualified workers is large and stable. For this reason, the acceptance of lower wages could not be contributed to the lack of demand for private sector workers. In the analysis of this paper, the aim is to identify to what extent existing wage penalties for higher qualified workers in the public sector can be declared by the height of the level of the discount rate (thus the degree of patience) of higher qualified public sector workers. Buelens & van den Broeck (2007) found that private sector workers are more extrinsically motivated by salary and attach more importance to pay and economic rewards than private sector workers. For this reason, my hypothesis is that higher qualified private sector workers are more impatient, simply because they might be more motivated to earn better earnings on a faster pace. Unfortunately, the hypothesis is confirmed for neither higher qualified private sector workers nor all private sector workers. In fact, results turn in the opposite direction instead: impatient people tend to select themselves into public sector employment, instead of private sector employment. However, whereas the results are not significant, it is difficult to derive any conclusions.

Additionally, it has also been tested to what extent risk measures are a reliable predictor for the height of the discount rates and vice versa. Whether and how risk and time preferences are stable over time and across different decision domains is an important, ultimately empirical question that economists have only recently begun to address (Andreoni & Sprenger, 2012). I conducted a Spearman’s rank correlation test but found no significant relation between risk aversion and impatience.

This paper proceeds as follows. In the theoretical framework a theoretical background to the determinants of heterogeneity in time and risk preferences of public and private sector will be discussed. Preceding time preferences will be introduced shortly. In section 3, empirical research concerning differences between public and private sector employees in terms of risk attitudes and level of impatience will be reviewed. The method and data details will be presented in section 4, whereas the results will be discussed in section 5. Section 6 discusses and concludes.

2. Theoretical Background

In the increasing body of development literature, both the issue of selection of workers into public sector employment as the possible existence of correlations between time and risk preferences have received a lot of attention recently. This section provides a theoretical background to the determinants of heterogeneity in time and risk preferences of public and private sector workers. First, a short introduction to time preferences will be given. Next, the most important factors that influence time and risk preferences will be examined, followed by a review of existing empirical literature on the relationship between time and risk preferences. This section concludes with a review of the factors that are of influence in choosing either public or private sector employment.

2.1 – Introduction to time preferences

In economics, time preference is the relative valuation placed on a good at an earlier date compared with the relative valuation at a later date. Time preferences can be captured mathematically in a discount function, which expresses time preference in a so-called discount rate. It is assumed that people with higher discount rates are less future oriented (are more impatient) than people with lower discount rates, i.e. people with higher discount rate invest less in their future than people with lower discount rates (Mincer, 1958).

(6)

3

In general, three different discount models are distinguished to measure time preferences in economics. The first model, the discounted expected utility model (henceforth DUE model) is introduced by Paul Samuelson (1937) and known for its simplicity. The most important feature of the model is that it assumes that the discount rate is constant over time. This results in a model that is dynamically consistent, which means that preferences do not change over time. The weighting placed on a certain utility is always the same at any point in time. For example: if one prefers a larger amount of money paid in one year from now versus a smaller amount of money paid today, dynamic consistency implies that the discount rate remains the same if one prefers the larger later amount paid in two years from now versus the same smaller amount paid in one year (Frederick et al., 2002). Loewenstein & Thaler (1989) found that discount rates are not constant and introduced the hyperbolic discount function. Hyperbolic discounting implies that the discount rate over short horizons is relatively lower than the discount rate over long horizons. This declining rate makes that the function becomes dynamically inconsistent, which means that preferences change over time. Dynamic inconsistency arises if preferences between two rewards will reverse in favour of the larger reward if the time horizon gets longer (Herrnstein, 1990).

The last model, introduced by Phelps & Pollak (1968), is the quasi-hyperbolic discount model. It combines both of the two former models and is therefore both hyperbolic as exponential, thus quasi-hyperbolic. The most important feature of the model is that the discount rate for short time horizons is hyperbolic (so relatively impatient). For the longer time horizons however, the discount rate becomes exponential (so relatively patient).

2.2 – Factors influencing risk and time preferences

This section provides a brief overview of the factors that are on influence on time and risk preferences. As there are many factors that are on influence on these preferences, I will only focus on the following factors that might be correlated to occupational choice as well: gender, education and intelligence, age and wealth and income. First, the findings on time preferences will be discussed, subsequently, the findings on risk preferences.

2.2.1 – Factors influencing time preferences 2.2.1.1 – Gender

In the field of psychology and biology, both Barkow et al. (1995) and Campbell (2013) found that men are on average more impatient than women. In economics, the relationship between gender and time preferences does not move in such a clear direction. Previous experimental research on this relationship showed the following varying results:

First, Coller & Williams (1999) show, in line with the biologists and psychologists, that women have lower discount rates and thus are more patient. Dohmen et al. (2010) however, found the completely opposite. These differences might be reasoned by the differences in the length of time horizons used in the two experiments. Whereas the former used a time horizon of two months, Dohmen et al. (2010) used a time horizon of one year.

(7)

4 2.2.1.2 – Education and intelligence

According to rational theory, one could assume that intelligent people are more patient, as their higher cognitive skills should enable them to make more rational choices. Dohmen et al. (2010) showed that subjects with higher cognitive ability are significantly more educated; for that reason, education and intelligence can be mentioned in one breath. Dohmen et al. (2010) confirm the positive relationship between higher cognitive ability and patience, in terms of the exponential discounting model (DUE model). Also, Benjamin et al. (2006), who conducted an experiment with Chilean students, found that students with lower scores on their intelligent tests made more impatient choices. Furthermore, both Burks et al. (2009) and Meier & Sprenger (2015) found that subjects with higher cognitive ability, are more patient in both their discount factor r as in their present bias β in terms of the quasi-hyperbolic discount model of Laibson (1997).

2.2.1.3 – Age

First, Tanaka et al. (2016), who focused on relatively short time horizons from three days to three months, found that average discount rates decline as people get older. They tested it both in terms of the exponential (DUE model), the hyperbolic and the quasi-hyperbolic discount model. They found that older people are less present-biased. Additionally, both Meijer & Sprenger (2015) as Harrison et al. (2002) confirm that younger people are on average more impatient, which is similar to the findings of Tanaka et al. (2016).

2.2.1.3 – Wealth and income

The first person who thought about the effect of wealth on time preferences was Irvin Fisher in 1930. Since it’s not sure if wealth is affected by discount rates or vice versa, clear causality in not ensured because of this possible endogenous relationship between the two variables. Nevertheless, much research has been done on this relationship. First, Kirby et al (2002) and Coller & Williams (1999) found a positive relationship between discount rates and wealth. On the other hand, Hausman (1979) found a negative relationship. These differences might be reasoned by the differences in sample composition: if for example a student sample is used, it could be that the incomes are more fluctuating than the incomes of a non-student sample. In general it can be concluded that wealthier people are more patient but more present-biased. Results based on the quasi-hyperbolic model have turned out to be the strongest.

2.2.2 – Factors influencing risk preferences

Risk preferences are the underlying factors that drive our decisions. Risk preferences are reflected in nearly all (economic) choices, such as choices regarding wealth, health, education, love and moreover: occupational choice. As risk preferences are important, it is interesting to know to what extent the four factors mentioned before are on influence on risk preferences.

2.2.2.1 – Gender

In general it is assumed that women take less risk in their financial-decision making. But is this truly the case? First Dohmen et al. (2006) confirm this assumption and found that women are in general less willing to take risk. In their research, subjects were asked to indicate their willingness to take risk on an eleven-point scale, similar to the way risk preferences are measured in the analysis of this paper. Harrison et al. (2007) used a more experimental way to measure risk: subjects were asked to choose between two lotteries that varied in probability. They found no significant effect of sex.

(8)

5

Whereas the results are mixed, it tends to be that women are slightly more risk averse than men in the end.

2.2.2.2 – Education and intelligence

First, Frederick (2005) found that more intelligent people are less-risk seeking in the loss domain and less risk-averse in the gain domain. This is due that it tends to be that more intelligent people are in general better in making rational-choices and that their reasoning is more punctual and faster, so that it leads to more rationality. Dohmen et al. (2010) found the same result in their analysis for both risk measures based on survey-data as risk measures within an experimental setting.

2.2.2.3 – Age

Dohmen et al. (2006) found that the willingness to take risks decreases with age. They used both the results of a survey-question that simply asked the subjects to indicate their willingness to take risk as results of a risk experiment, to find out whether the results were in line with each other. The results turned out to be in line, as both results confirmed a negative relationship between willingness to take risk and age. Additionally, Harbaugh et al. (2002) focused on children and found more or less the same: children are less risk averse and risk aversion increases with age.

2.2.2.4 – Wealth and income

Tanaka et al. (2016) conducted an experiment in villages in Vietnam and found that wealthier families are on average less loss-averse. Dohmen et al. (2006) found the same negative correlation between wealth and risk aversion, using both the GSOEP survey data for Germans as experimental data. Lastly, Gsuio & Paiella (2008), who used an Italian survey, found that people become more risk averse if aspects such as low income or income insecurity come into play. Concluding, it tends to be that on general, wealth is negatively related to risk aversion.

2.3 – Relationship time and risk preferences

Most economic decisions involve cost and benefits that are uncertain and affect the future as well as the present. Optimal decisions thus depend on both time and risk preferences. An example of a situation where both time aspects and aspects of risk matter is the decision to go to college versus finding a job. Whereas one who decides to start working will start to earn money right now, one who decides to go to college will have to wait for a couple years before he can start earning money. Thereby, deciding to go to college brings two aspects of risk: first, it is not sure yet if one will find a job after completing college that will have higher earnings and can cover both the study costs as the opportunities costs that have been lost during the years of education. Second, it is not sure whether one will have the capabilities to complete college anyways.

Conventional economic theories assume that risk aversion does not coincide with patience. However, Loewenstein & Prelec (1992) were the first who pointed out a relationship between the two preferences. How are these two preferences related? To what extent are risk attitudes a reliable indicator of time preferences and vice versa? This section summarizes the most important findings regarding the relationship between these two behavioral characteristics.

First, Psychologists Ida & Goto (2009) found that addictive behavior like drinking, smoking and gambling is related to impatience combined with low risk aversion. Economic research from Anderhub et al. (2001) found also a relationship: they found that individuals who are more risk averse have higher discount rates, thus are more patient (compared to individuals who are less risk

(9)

6

averse). This conclusion supports the findings of economists Keren & Roelofsma (1995). Lampi & Nordblom (2011) only found that men’s time and risk preferences are related. Lastly, Dohmen et al. (2010) found that time and risk preferences are related as well and that higher discount rates can be associated with a higher level of risk aversion.

Whether and how risk and time preferences are stable over time and across different decision domains is an important, ultimately empirical question that economists have only recently begun to address, according to Andreoni & Sprenger (2012). For that reason, a test will be conducted in the results part of this thesis to find out time and risk preferences are related to one other.

2.4 – Differences job-related characteristics between the public and private sector

In this section, a brief overview of the differences in job characteristics between the public and private sector will be provided. The employment choice for either public or private sector is based, in effect, on the specific job characteristics of the sectors. Therefore, it is important to know what the differences are.

Public sector labour markets have two important features: First, they are large. In Germany, more than 14% of the labour force worked for the government in 2013 (OECD, 2015). Second, public sector labour markets differ from private sector labour markets in many different ways. The first and most important differences are the earning differentials. There are a number of reasons why these differences could exist (Giordano et al., 2010): first, the objectives of private firm owners differ from those of politicians or bureaucrats. Whereas the public sector is subject to political constraints, the private sector is subject to profit constraints. For this reason issues of pay equity and pay security have a bigger chance to survive in the public or political market place than in the private market place.

On average, net earnings of public sector workers are thirteen percent higher than net earnings of private sector workers in Germany (Dustmann & van Soest, 1998). Although private and public sector workers are subjected to the same income tax rules, civil servants do not pay any social security contributions because of their own social security system. Since these costs are completely covered by their employer, the average wage of public sector workers is higher. On the other hand, comparable and recent research showed that average wages in the public and private sector were almost identical (Office for National Statistics, UK). After controlling for different jobs and personal characteristics, the estimated pay gap ranged between 0.9% in favour of the public sector workers and 0.1% in favour of the private sector workers.

Nevertheless, if we take a closer look at differences in earnings distributions between the two sectors, we find that the wage distribution in the private sector is more dispersed (Pfeifer, 2011). Whereas workers on the lower tail of the wage distribution benefit from public sector employment, workers on the higher tail earn less than their private sector counterparts (Jürges, 2002). Hence, wage penalties exist for higher qualified workers in the public sector. But why do these public sector workers accept these lower wages? According to Pfeifer (2011), the demand of private sector firms for highly qualified workers is large and stable. For this reason, the acceptance of lower wages cannot be declared by the lack of demand for private sector workers. A possible explanation is that individuals select themselves into public sector employment, as they prefer the characteristics of a public sector job to the characteristics of a private sector job.

Several characteristics have been found to be on influence on the decision for public sector employment. First, employment security is perceived larger in the public sector (Pfeifer, 2011). This

(10)

7

feature might be prevalent in the decision for public sector employment. For that reason, preferences towards risk play an important role in the decision process for public or private sector employment. Another characteristic of public sector employment is that many jobs involve helping people in need or contributing to society (education, health care, law enforcement). It can therefore be assumed that people with a strong willingness to serve others or the public interest select themselves into public sector employment sooner (Buurman et al., 2012).

Lastly, Buelens & van den Broeck (2007), who analysed the differences in working motivations between public and private sector workers, found the following interesting results that can declare why public sector workers do accept a wage penalty and therefore not switch to a private sector job with better earnings. First, they found that civil servants are less extrinsically motivated by salary. Second, they found that civil servants are less motivated by self-development. Third, they found that civil servants are strongly motivated to work in a supportive working environment. The last interesting finding was that civil servants significantly report fewer working hours than private sector workers. In particular the findings that civil servants are less extrinsically motivated by salary and report less working hours can be used in finding a declaration why public and private sector workers might differ in terms of their level of patience.

3. Related research

A rich literature in economics has examined the differences in motivations and preferences between public and private sector workers. In this section, the most important literature concerning differences between public and private sector workers in terms of risk attitude will be reviewed. The direct relationship between time preferences and employment choice has never been empirically tested before. For that reason, the most important findings on the indirect effects towards time preferences and employment choice will be discussed instead.

For the findings regarding the risk preferences a distinction will be made between literatures that used stated preferences measures (e.g. by asking subjects how risk taking or patient they are) and literatures that used revealed preferences measures; i.e. data on what people actually do and not on what they say. The discount rate measure as used in the analysis in this paper is an example of a revealed preferences measure: here, preferences are inferred from the stated behavior in the time preferences experiment (Buurman et al., 2012). Findings in this section will be compared to the results in the discussion and conclusion section.

3.1 – Related research using stated risk preferences

The first who examined whether public and private sector workers differed in risk preferences, using stated preference, were Bellante & Link (1981). In their study, they used answers to questions like the condition and insurance of owned cars, smoking and drinking habits, the extent of medical coverage and the use of seatbelts as their measure of risk aversion. They found that risk-averse individuals are more likely to select themselves into public sector employment. Roszkowski & Grable (2009) found the same result. As measure of risk aversion, data on clients of financial planners who had completed a survey concerning their financial risk tolerance was used. They found that public sector employees scored significantly lower than their private sector counterparts, even after adding a rich set of controls. Bonin et al. (2007), Luechinger et al. (2006), Dohmen et al. (2011) and Pfeifer (2011), who all used the 2004 wave of the German Socio-Economic Panel (GSOEP), found similar results. The panel contains questions on people’s attitude toward both general risk as career-risk. The analysis in this paper also exploits the GSOEP, but then using the 2006 wave. First, Bonin et al. (2007) confirmed the very important assumption that working in the private sector implies a

(11)

8

significantly higher earnings risk than working in the public sector. Moreover, they found that individuals who are more risk averse are more likely to end up working in a job with lower earnings risk, such as jobs in the public sector. Luechinger et al. (2007), Dohmen et al. (2011) and Pfeifer (2011), who directly estimated the effect of risk attitude on sector employment choice, reported similar findings and thus found that public sector workers are in general more risk averse. Additionally, Pfeifer (2011) also looked at the direct effect of career-specific risk attitudes on employment choice, and found that the impact of general risk attitudes is smaller compared to career-risk attitudes. Hartog et al. (2002) and Guiso & Paiella (2008) found the same result as well. By using household survey data on people’s willingness to pay, they constructed a direct measure of absolute risk aversion based on the maximum price a subject is willing to pay for a hypothetical lottery or a hypothetical risky security. They found that the amount that private sector workers were willing to pay was significantly higher than the amount of public sector workers. Lastly, Lewis & Frank (2002), who tried to investigate what kinds of people are attracted to governments jobs, analyzed the 1989 and 1998 General Social Survey (GSS) and found that job security is an important selling point of public sector employees. In fact, among the job aspects asked about, the GSS respondent gave it the highest priority. Where 57% of the 1998 respondents called job security ‘very important’, another 37% considered it as ‘important’. The respondents who considered job security as ‘very important’ were one-third more likely to prefer government jobs compared to the respondents who only labeled job security as ‘important’.

3.2 – Related research using revealed risk preferences

Buurman et al. (2012) were the first who used revealed preferences data rather than stated preferences. Their study took place in the Netherlands. They exploited a survey where subjects were offered a reward for completion: a gift certificate, a lottery ticket or a charitable donation. Subsequently, they looked at whether public and private sector employees differed in their choices in rewards and found that public sector employees are significantly less likely to choose the lottery ticket. Moreover, they also found that public sector employees were significantly more likely to choose the charitable donation at the start of their career. Buurman et al. (2012) were therefore the first who found strong support public sector workers are more risk averse than private sector employees, by using revealed preferences.

Conclusive, the hypothesis that compared to private sector employees, public sector employees have a higher degree of risk aversion is widely supported by many studies, both by using stated as revealed preferences.

3.3 – Related research on time preferences

An interesting study, that tries to investigate how time and risk preferences are related to cognitive ability, is the study of Dohmen et al. (2010). They found that lower cognitive aversion is related to more pronounced patience. Even after adding several control variables for educational attainment, income and measures of credit constraints, these results stayed significant. As the public sector tends to include a larger proportion of high skilled workers (Office for National Statistics, 2014) than the private sector sector, it could be assumed that public sector workers are more patient, simply because the proportion of high skilled workers in the public sector is larger. Additionally, Coller & Williams (1999) found that males have marginally significantly higher discount rates than females. They found that the discount rate of males was on average 13 percentage points higher than those of female. Whereas Pfeifer (2011) showed that females are more likely to select themselves into public sector employment, it could be assumed again that public sector workers are more patient, simply because of the higher proportion of females who work for the public sector.

(12)

9

As these finding towards time preferences and sector employment choice are indirect, I am aware no valid conclusions can be derived from this. For that reason, the findings of the existence of a direct relationship on time preferences and sector employment choice are presented in the following two sections.

4 Methodology and design

4.1 – Data and method

In this section, statistical tests are conducted to determine how public and private sector workers differ in their degree of risk aversion and in the height of discount rate. Also, the relationship between time and risk is analyzed to find out these characteristics are independent or not, by conducting a Spearman’s Correlation Test.

For the analysis, data from the large-scale German Socio-Economic Panel (GSOEP) are used. The GSOEP is an interdisciplinary longitudinal survey of approximately 11000 private households and 70000 persons of both the old and new states of Germany from 1984 to now, produced by DIW Berlin (European University Institute, 2016). Next to a stable set of core variables on general demographic and socio-economic characteristics (e.g. income, education, household composition, employment, health and satisfaction indicators) waves of data also include additional topics like measures of individuals’ discount rates and risk attitudes (Pfeifer, 2011). For a detailed summary, see Wagner et al. (2007).

4.1.1 Preferences measures

The data waves of 2004 and 2006 till 2013 include measures of risk aversion in the general and specific contexts. The datasets report an index of Risk Aversion that ranges from 0 to 10, where larger levels correspond to a lower level of risk aversion or greater willingness to take risks (Bellante & Link, 1981). The exact question is worded in the following way: ‘How do you see yourself: are you generally a person who is fully prepared to take risks or do you try to avoid taking risks? Please tick a box on the scale 0 to 10, where 0 corresponds to complete risk aversion and 10 to complete risk taking’ (cited from Skriabikova et al., 2014). Additionally in 2004 and 2009, individuals were asked to indicate their willingness to take risk on the same eleven-point scale in six different contexts. The exact wording of that question was: ‘People can behave differently in different situations. How would you rate your willingness to take risks in the following areas: financial matters, during leisure and sport, in your occupation, with your health and your trust in other people (scales from 0 to 10)?’ (cited from Skriabikova et al., 2014). Both the general and the specific question has been validated by Dohmen et al. (2011). They showed that the questions are related to actual behaviour in many important real life decision domains, by using incentivized experiments.

The measure for the discount rate is based on the findings of Wölbert & Riedl (2013) and the actual data on time preferences in the GSOEP. In their article the reliability, stability and domain specificity of different measures of time and risk preferences are analysed. In the GSOEP data wave of 2006 an experiment on time preferences was conducted. All subjects were asked to choose between a smaller-sooner amount of money of €200 (received by check immediately) or a larger-later amount of money (received later, with interest). As the subjects were divided into three subsamples it was possible to test the possible incentive effects by conducting three experiments that varied in time horizon and interest rate (Richter & Schupp, 2006).

(13)

10

Richter & Schupp (2006) explain that in all three experiments, subjects faced two lists of 20 choices between an earlier amount of money (which was €200 for all options) and a higher delayed amount of money. For example subjects were asked: Would you rather receive €200 today or receive €205 in 12 months? Furthermore, the larger-later payments increased from row to row at specific rates. How the three variants of the experiment varied in time horizon and interest is summarized in table 1. Appendix A shows the complete choice lists of all the three experiments.

Table 1

Time Horizon Discount Rate

Experiment 1

Z1 (list 1) Early payment: immediately

Delayed payment: 12 months

2.5%

Delayed payment : 6 months

1.25% Experiment 2

2.5%

Delayed payment: 1 month

0.2% Experiment 3

Delayed payment: 1 month

0.4%

Z6 (list 2) Early payment: 12 months

0.4%

Time preferences can be measured by calculating the average switching point from the early to the delayed payment of the two lists. Subjects who prefer the later options will switch earlier and discount the future by less (i.e. are more patient) than subjects who switch later. Hence, lower/higher values imply earlier/later switching rows and thus a higher level of patience/ impatience. To rule out a present bias exists between the immediate and delayed awards, I tested whether the average switching point from the Z6 list differs from the average switching point from the Z5 list in experiment 3. To do this, I used an Friedman test and found that they do not differ significantly (Friedman test, Χ2 = 0.891, N = 542).

As the time preferences experiment was conducted in 2006, the analysis in this paper will be based on four 2006 waves. The first wave includes the results of the time preferences experiment; the other three waves contain data on different general demographic and socio-economic characteristics including the measure of general risk aversion. The 2006 wave did not include measures of risk aversion in the specific contexts. To solve this, I will use the measure for Career Risk taking from the older 2004 wave. In total these waves comprise information on more than 20000 subjects. However, data will be reduced in sample size. The final dataset only includes individuals who are aged between 18-65 years, who have the German citizenship and subjects who are either employed in the public or private sector private (so self-employed and unemployed subjects are disregarded). This results in a final sample size of 8132individuals in 3256different households.

(14)

11 4.1.2 Method

To find out whether workers with different time and risk preferences have different probabilities of choosing public versus private sector employment a simple binary choice model - the probit model is used. The model is based on the model used in the comparable analysis of Bellante & Link (1981) and Pfeifer (2011). The dependent variable estimates the probability of choosing public sector employment and is a dummy: it takes value one if an individual is employed in the public sector and takes value zero if an individual is employed in the private sector (Pfeifer, 2011). Next to the two variables measuring the time and risk preferences, the probability of public sector employment will be estimated in terms of several independent control variables which are considered as relevant in explaining occupational choice (Blank, 1985). These variables include age (measured in years), gender (a female dummy), employment status (dummy for fulltime employment), education (a dummy for a degree in tertiary education), marital status (dummy for married, living together), fulltime working experience (measured in years), life satisfaction (measured on an eleven-point scale), monthly gross labour income (measured in Euros) and height (in cm). Additionally a dummy variable for workplace in the new federal states is added. Table 2 shows the descriptive statistics with the means and standard deviations of the main variables of the time and risk measures and most important variables. For the complete table of descriptive statistics, see Appendix B.

Table 2: Descriptive statistics

Variable

Observations Mean Std.

Dev.

Min Max

Public Sector (dummy)

8,132

0.29

0.46

0

1 Age (in years)

8,132

42.18 11.31

18

65 Female (dummy)

8,132

0.49

0.50

0

1 Higher Education (dummy)

8,132

0.27

0.44

0

1 General Risk Taking (0: low, 10: high)

8,132

4.95

2.11

0

10 Career Risk Taking (0: low, 10: high)

8,010

4.02

2.47

0

10 Discount Rate (1: complete patience, 21:

complete impatience)

574 10.92 7.01

1

21 Satisfaction with life at today (0: not

satisfied, 10: highly satisfied)

(15)

12

4.2 Hypotheses

Hypothesis 1

H

1

:

Compared to private sector employees, public sector employees have a higher degree of risk

aversion. Explanation:

According to economic reasoning, individuals with a higher degree of risk aversion are more likely to seek employment in the public sector, since working conditions in the public sector are more stable (Bellante & Link, 1981; Pfeifer, 2011). The higher the degree of risk aversion is, the higher the value individuals place on stable working conditions. Also, (Bloch & Smith, 1979) showed that the probability of becoming unemployed is lower for workers in the public sector than for workers in the private sector. Therefore, one should expect that more risk averse people prefer to work in the public sector.

Hypothesis 2

H

2

:

Compared to (higher qualified) private sector employees, (higher qualified) public sector employees have on average lower discount rates (are more patient).

Explanation:

This hypothesis is based on both a theory about the earnings differences between the public and private sector and on existing theories that support a negative correlation between a high degree of risk aversion and a high discount rate.

First, recent research from the Office for National statistics (2014) showed that average wages in the public and private sector are almost identical. After controlling for different jobs and personal characteristics, the estimated pay gap ranged between 0.9% in favour of the public sector workers and 0.1% in favour of the private sector workers.

However, if we take a closer look at differences in earnings distributions between the two sectors, we find that the wage distribution in the private sector is more dispersed. If one would order each person in terms of their hourly pay and then compare the 95th percentile of earnings with the 5th percentile in each sector, earnings are around 5.3 times higher in the 95th percentile compared to the 5th percentile in the private sector. In the public sector these earnings are only 4.3 times higher. From the 90th percentile onwards, the wages of the private sector start to exceed the wages in the public sector. Whereas the top 1% of workers in the public sector earn £50.11 or more per hour, the minimum top 1% wage in the private sector is higher: £59.09 per hour (see table 1). After conducting a simple t-test (a t-test comparing two means using matched pairs), it is found that private sector wages in the higher tails are significantly higher than public sector wages (t = 2.37, rejection region t ≥ 1.833 at the 5% level). Thus: Wage penalties exist for higher qualified workers in the public sector. According to Pfeifer (2011), the demand of private sector firms for highly qualified workers is large and stable. So why do these higher qualified workers accept these lower wages and do not choose to switch to a comparable job with better earnings in the private sector?

(16)

13

Table 3: distribution of hourly earnings in the public and private sector, April 2014

£/ hour £/ hour Public Private Percentile 90 26,01 25,07 Percentile 91 26,84 26,22 Percentile 92 27,71 27,55 Percentile 93 28,82 29,10 Percentile 94 29,83 31,02 Percentile 95 31,37 33,40 Percentile 96 33,39 36,54 Percentile 97 36,40 40,65 Percentile 98 41,11 46,74 Percentile 99 50,11 59,09

Source: ONS Annual Survey of Hours and Earnings

The question is, how does the wage penalty in the top 10 percentiles in the public sector relate to the height of the discount rate of a higher qualified private sector worker? Because there are no existing theories about this relationship yet, I will come up with my own one.

Let us take a closer look at the different factors that explain the employment choice for public or private sector employment. Buelens & Van den Broeck (2007) found that private sector employees and managers attach more importance to economics rewards than public sector employees and managers do. They also found that pay is from much more importance for private sector employees than it is for their public sector counterparts. My hypothesis is that people who value pay and economics rewards higher will attach more importance to the better earnings prospects in the private sector. The fact that private sector workers are more extrinsically motivated by salary can be a factor why impatient people tend to prefer to work in the private sector, simply because they are more willing to make promotion and attach more value to higher wages, compared to public sector workers (Buelens & Van den Broeck, 2007). Therefore, I expect that compared to higher qualified public sector employees, higher qualified private sector employees have on average higher discount rates, or in other words, are more impatient.

Another theory that supports the negative relationship between risk aversion and the discount rate, applied to all public sector workers instead of only the higher educated ones, is that it is assumed that intelligent people are more likely to be patient and thus have lower discount rates. Benjamin et al. (2005), Harrison et al. (2002) and Dohmen et al. (2010) all confirm the assumption that more intelligent people, both those with a higher IQ and more education, are more patient. The Office for National Statistics (2014) showed that the percentage of low skilled workers (6%) is lower in the public sector than in the public sector (15%). Moreover, the public sector tends to include the largest proportion of high skilled workers (43% vs 22%, see table 2). Since both the skill levels ‘High’ and ‘Upper middle’ are normally acquired through a degree or post-compulsory education it can be assumed that the relative number of educated people is higher in the public sector. Therefore, one should expect that employees in the public sector have on average lower discount rates and are more patient than private sector workers (Office for National Statistics, 2014).

(17)

14

Table 4: Percentage of employees by skill level of their job in the public and private sector, April 2014, UK. Source: Office of National Statistics, UK

Percentage

Skill-level Public sector Private sector

High skill 1.1 43 22

Upper middle 1.2 20 25

Lower middle 32 38

Low skill 6 15

1.1 High – This skill level is normally acquired through a degree or an equivalent period of work experience. Occupations at this level are generally termed ‘professional’

or managerial positions, and are found in corporate enterprises or governments. Examples include senior government officials, financial managers,

scientists, engineers, medical doctors, teachers and accountants.

1.2 Upper-middle – This skill level equates to competence acquired through post-compulsory education but not to degree level. Occupations found at this level include

a variety of technical and trades occupations, and proprietors of small business. For the latter, significant work experience may be typical.

Examples of occupations at this level include catering managers, building inspectors, nurses, police officers (sergeant and below), electricians and plumbers.

5 Empirical results

In the following section the determinants of public sector employed will be analysed by estimating probit models in which the dependent variable is a dummy variable that takes the value one if an individual is employed in the public sector. Note that the discount rate variable is not included in the risk preferences part for the reason the number of participants in the time experiment is limited. This would have reduced the sample size by too much.

5.1 Risk Preferences

Table 5 reports marginal effects of the probit estimates. To begin with, column one estimates a probit model where the measure for general risk attitude is the only explanatory variable. In specification two, the measure for career risk attitude is used instead of the general measure to find out which effect is stronger. The third specification includes both measures to find out which measure actually drives the results. The first specification (1) shows that individuals, who are one point more willing to take risks in general, are on average 1.07 percentage points less likely to work in the public sector. Specification two (2) shows a significant but weaker effect if career-specific risk-taking behaviour is used. Individuals who are one point more willing to take risks in their career, are on average 0.57 percentage points less likely to work in the public sector. In the third (3) specification both variables are included. It shows that general risk-taking behaviour mainly drives the results, as the significance level for Career Risk disappears. In order to assess whether these relationship are robust to controlling for personal characteristics, the next step is to estimate a probit model on a set of explanatory variables. Specification 4 (4) shows the marginal effects of the added variables. The variable age2 (age * age) is added to control for a potentially nonlinear relationship of age on self-selection into public sector employment.

(18)

15 Table 5 - Probability of public sector employment (risk

preferences)

Marginal effects

Mean (Standard Deviation) (1) (2) (3) (4)

Dependent Variable: public sector (dummy) 0.29 (0.46)

General Risk Taking (0: low, 10: high) 4.95 (2.11) -0.0107279 *** -0.0101436

***

-0.0088528 ***

Career Risk Taking (0: low, 10: high) 4.02 (2.47) -0.0056874 *** -0.0028199 -0.0008384

Age (years) 42 (11.3) 0.0032016

Age (squared, in years) 1907 (941) 0.0000237

Female (dummy) 0.49 (0.50) 0.1190848 ***

New Federal States (dummy) 0.22 (0.41) -0.0087271

Married (dummy for married, living together) 0.61 (0.49) -0.00799

Higher Educated (dummy for owning a tertiary degree) 0.27 (0.44) 0.1839387 ***

Income (monthly, in euros) 2437 (1783) -0.00000285

Height (in centimeters) 173.06 (9.26) 0.0006799

Discount Rate (1: complete patience, 21: complete impatience) 10.92 (7.01) Not included

Fulltime (dummy for fulltime) 0.68 (0.47) 0.030865 **

Satisfaction with life at today (0: not satisfied, 10: highly satisfied) 7.01 (1.61) 0.0129572 ***

Working Experience (in years) 15.83 (11.61) 0.0007854

Number of observations 8132 8010 8010 7985

Pseudo R2 0.0021 0.0008 0.0024 0.0677

(19)

16

Specification 4 (4) shows us that the effect of General Risk Taking remains significant at the 1% level after controlling for additional characteristics. The effect of Career Risk Taking is not significant, which corresponds with the findings in specification 3. Also, higher qualified workers are more likely to be employed in the public sector: individuals, who own a degree in tertiary education in German are on average 18.4 percentage points more likely to work in the public sector. This result is highly significant and corresponds with previous findings of Bellante and Link (1981), Blank (1985) and Luechinger et al. (2007). This might be declared by the higher demand for higher qualified workers in the public sector (Gregory and Borland, 1999). Furthermore, females are on average 11.9 percentage point more likely to work in the public sector. This effect is significant at the 1% level. This corresponds with the finding of Dohmen et al (2005), who confirm this assumption and found that women are in general less willing to take risk. Another interest finding is the significant effect of life satisfaction: individuals, who are one point more satisfied with life at today, are on average 1.29 percentage points more likely to work in the public sector. Lastly, all other control variables have not been found to be significant on any of the three levels.

Conclusive, the findings support a causal interpretation of the effect of risk preference on self-selection into public sector employment. These results therefore support the first hypothesis that public sector employees have a higher degree of risk aversion. The results correspond with previous findings of Bellante and Link (1981) and Pfeifer (2011) who have found the same result.

5.2 Time preferences

First, the differences in discount rate between public sector workers will be analysed, to find out whether public sector workers are more patient. Subsequently, it will be tested if higher qualified public sector workers are more patient in specific. For the first test, all subjects who participated in the time experiment are included. This results in a final sample size of 574 individuals. The second test only includes subjects who participated in the time preferences experiment and own a tertiary degree. This results in a smaller sample size, consisting of 138 subjects only. Table 6 reports marginal effects of the two probit analyses.

Specification 1 shows the average marginal effect of the estimated probit model with the measure for time preferences as the only explanatory variable for all subjects who participated in the time experiment. Specification 2 reports the marginal effects for all additional explanatory variables. Gender, age and height are included because they showed to be related to patience in previous studies (Dohmen et al. 2010). In specification 1, the estimate for the discount rate suggests that individuals’, whose average switching point from the early to the delayed payment is one point later, are 0.028 percentage points less likely to work in the public sector. Since higher values of switching points imply higher discount rates (thus higher levels of impatience), this result suggests that more impatient people tend to be less likely to work in the public sector. However this result is very small and not significant. This negative effect becomes positive after adding the additional controls: more impatient individuals are 0.004 percentage point more likely to work in the public sector now. However, the effect is not significant again. Nevertheless, the effect becomes stronger (and nearly reaches significance) in specification 3 and 4, where only higher educated individuals are considered. The fourth specification shows that individuals’, whose average switching point from the early to the delayed payment is one point later, are 0.8 percentage points more likely to work in the public sector. Despite the effect is not strong enough to be significant on either the 1%, 5% or 10% level, it is significant at the 20% level, as the p-value equals 0.155.

Besides, both the probit analysis on all subjects as on the higher educated subjects find again that females are more likely to select themselves into public sector employment (effect is significant at the 1% level). Also, the finding that higher educated individuals are significantly more likely to work in the public sector is confirmed. Additionally, height and age haven’t been found to be significant.

(20)

17

All participants in the time preferences experiment Participants holding a degree

Table 6 - Probability of public sector employment (Time preferences) Marginal Effects Marginal Effects Mean (Standard Deviation) (1) (2) Mean (Standard Deviation) (3) (4)

Dependent Variable: public sector (dummy) 0.30 (0.45) 0.46 (0.50)

Discount Rate (1: complete patience, 21: complete impatience) 10.92 (7.01) -0.0002847 0.0004 9.05 (6.74) 0.009 0.008

General Risk Taking (0: low, 10: high) 4.84 (2.17) -0.018* 4.76 (2.09) -0.009

Career Risk Taking (0: low, 10: high) 3.82 (2.56) 0.0167098** 4.24 (2.45) 0.027

Age (years) 43.33 (11.35) 0.016 45.42 (10.40) -0.0187

Age (squared, in years) 2006.95 (954.53) -0.00013 2170.3 (913) 0.0003

Female (dummy) 0.501 (0.50) 0.15*** 0.49 (0.50) 0.21*

New Federal States (dummy) 0.10 (0.31) -0.14** 0.15 (0.36) -0.228*

Married (dummy for married, living together) 0.59 (0.49) 0.012 0.61 (0.49) -0.027

Higher Educated (dummy for owning a tertiary degree) 0.24 (0.43) 0.19*** Omitted

Income (monthly, in euros) 2494 (2122) 0.000012 3641.38

(2823)

-0.00004

Height (in centimeters) 172 (9.30) -0.002 173 (9) 0.00386

Fulltime (dummy for fulltime) 0.67 (0.47) 0.11** 0.72 (0.44) 0.1198

Satisfaction with life at today (0: not satisfied, 10: highly satisfied) 7.44 (1.43) 0.006 7.64 (1.42) 0.0305

Working Experience (in years) 16.74 (12.21) 0.002 16.60 (12) 0.0079

Number of observations 574 574 556 138 138 136

Pseudo R2 0.000 0.0965 0.0126 0.0648

(21)

18

The size of this dataset is a lot smaller because it only includes the participants who participated in the time preferences experiment. This could be a reason why the results between the time preferences and risk preferences outputs differ. For example, the effect of life satisfaction was positive and very significant in table 5, but has not been found significant in table 6. These differences might be reasoned by the differences in characteristics of the samples and differences in sample size. However, after comparing the descriptive statistics and differences of the group composition of these groups, it tends to be that the group are composed more or less the same (see Appendix B). For that reason, it is hard to derive any conclusions about the influence of this on the differences in findings.

All in all, these results do not support my hypothesis that compared to private sector employees, (higher qualified) public sector employees have on average lower discount rates (are more patient). In fact, results turn in the opposite direction instead: impatient people tend to select themselves in public sector employment, instead of private sector employment.

5.3 – Correlations time and risk preferences

Another goal of these analyses is to find out about how possible differences between private and public sector workers in time and risk preferences workers could be declared by existing correlations between time and risk preferences. To find out to what extent risk attitudes are a reliable indicator of time preferences and vice versa, a significance test to decide whether there is any or no evidence to suggest that a relationship exist between time and risk is needed. I conducted a Spearman’s rank correlation test, which measures the strength of correlation of two variables. The results found show a spearman correlation coefficient ρ of -0.0545 (N=574, p = 0.1919). This indicates a negative correlation between time and risk, however, as the result is not significant, it can be concluded that the results of time and risk preferences found in this analysis cannot be declared by possible existing correlations between time and risk preferences. Whereas this finding is not in line with previous findings of for example Andreoni & Sprenger (2012), who found that time and risk correlations do relate to each other, it has a positive influence on the results found in this paper as internal validity is enhanced.

6. Discussion and conclusion

The main goal of this thesis was to investigate to what extent public and private sector workers differ in terms of their level of patience and in risk attitude. To determine this, a simple probit model was used to estimate the probability of being employed in the public sector.

In the first analysis, it is found that individuals, who are on average one point more willing to take risks on an eleven-point scale (a scale from 0 to 10), are less likely to be employed in the public sector. This is true controlling for several variables and exactly in line with previous findings of all studies mentioned in the related research chapter that conducted comparable research. The finding of Pfeifer (2011), that the impact of career-specific risk attitudes is larger compared with general risk attitudes, is not supported in this paper. In fact, the results indicate that only the impact of general risk attitudes drives the results and the effect of career-risk attitudes is not significant. Therefore Pfeifers’ conclusion on the importance to distinguish between different types of risk aversion when analyzing risk aversion is not confirmed. Other findings are that both higher qualified workers as well as females are more likely to be employed in the public sector. Lastly the analysis shows that individuals, who are one point more satisfied with life at today, are more likely to work in the public sector. All in all, the preconception that private sector workers act more like Nick Leeson and take more risk and that public sector workers take fewer risks seems to find support in this paper. Preferences towards risk play an important role in the decision process for public or private sector employment. This is worrisome and disadvantageous for the private sector. To make private sector employment more attractive, private sector recruiters will have to find creative ways to leverage this disadvantage in the future.

(22)

19

A limitation of this first analysis concerns the nature of the measures of time and risk preferences. Whereas for the time preferences measure revealed preferences are used, stated preferences are used to measure risk preferences. A possible limitation of the use of stated preferences is that they are vulnerable to self-serving biases, self-stereotyping, lack of attention by respondents and strategic motives (Buurman et al., 2012). On the contrary, Dohmen et al. (2011) show that these measures of risk aversion are related to actual behaviour in many important real life decision domains, by using incentivized experiments. A possible suggestion for future research is to use both revealed as well as stated measures (i.e. use both a survey question and a risk experiment), and find out whether the data of both measures are in line with each other. Applying this, internal validity will be enhanced. In the second analysis, no significant effect is found for the hypothesis that public sector workers are more patient than private sector workers. Neither the analysis where only higher qualified workers are considered nor the analysis where all workers are considered indicate significant results. Therefore, there is not enough evidence to state that the acceptance of the lower wages in the higher tails in the public sector can be contributed to the higher degree of patience of higher qualified public sector workers. Also, the combination of the finding that the relative number of educated people is higher in the public sector and the finding of Dohmen et al. (2010) that higher educated people are more patient does not lead to an actual higher proportion of patient employees in the public sector. All in all, it seems like that the impatient behaviour of the Lehman’s brothers cannot be generalised to all private sector workers and that the preconception about slower acting public sector workers who reconsider their decisions at least three times does not hold. Based on the analysis of this paper, one could therefore conclude that an individuals’ degree of patience has nothing to do with employment choice for either private of public sector employment.

However, this conclusion could be regarded as too premature due to the limitations the analysis is concerned with. An important limitation concerns the sample size. As most datasets do not contain or only use limited measures of time preference, this resulted in a small sample of respectively 538 and 138 subjects. The use of a bigger sample size is therefore recommended in future research. The second limitation concerns the use of time horizons. In the analysis of this paper varying time horizons of 1, 6 and 12 months are used. However, different uses of time horizons have shown varying results in related research to time preferences in the past, as shown in the theoretical background of this paper. This is therefore a limitation to measuring time preferences in general. A suggestion for future research is therefore first of all, to do research into the best way to measure time preferences. All in all, it is too premature to conclude that time preferences are not related to employment choice at all. It is therefore recommended to analyse this relation again in future research.

Lastly, it is found that risk and time are not related to one other. As time and risk are assumed to be independent in economic models, it would be a potentially important source of model miss-specification if it had turned out that time and risk are related. Fortunately, they have been found to be independent. This independence in turn has a positive influence on the results, because internal validity is enhanced.

A last general limitation to this entire research is the sample composition. To increase internal valid results, it is suggested to only consider individuals who have always worked for the public sector, and to disregard those who got involved into public sector employment via a detour (thus were employed in the private sector first). Applying this, individuals who prefer private sector jobs characteristics but are indicated as public sector worker in the sample (and vice versa) are

disregarded. This will result in a more representative sample and can therefore enhance the validity of future results even more.

Differences in time and risk preferences between public and private sector workers