A closer look at risk-taking behaviour among bankers : why a sound risk culture can prevent stressed bank employees of taking too much risk

(1)

A closer look at risk-taking

behaviour among bankers

Why a sound risk culture can prevent stressed bank

employees of taking too much risk

Josefien Snoeij (1681957)

josefiensnoeij@gmail.com

Supervisor

Dr. J.W. Stoelhorst

Master Thesis

Amsterdam Business School

UNIVERSITY OF AMSTERDAM

August 10

th

, 2015

Utrecht, The Netherlands

(2)

2

Abstract

The increasing importance of the role of individual behaviour in risk-taking of financial organizations necessitates a strong emphasis on the drivers of risk-taking behaviour. The aim of this study was to examine if there is a positive relation between job stress and risk-taking behaviour among bank employees, whether this relation is mediated by risk perception and moderated by a sound risk culture. A sound risk culture is measured in general and on four underlying subscales as described by the Financial Stability Board (2014): tone at the top, accountability, effective communication and challenge and incentives.

Data was gathered through a survey among 328 bank employees of different departments. The present study demonstrates the importance of managing stress of employees through demonstrating a significant and positive relation between job stress and risk-taking behaviour. The mediator effect of risk perception was not demonstrated, which might be caused by limitations of the methods used. This study has answered the call of the FSB (2014) to pursue a sound risk culture in order to prevent financials from excessive risk-taking, with demonstrating a significant moderator effect of risk culture on the relation between job stress and risk-taking behaviour. Especially employees’ awareness of accountability is held

responsible for this moderator effect. A sound risk culture proves to be even more important than expected with the demonstration of the direct and negative relations between risk culture and job stress and risk culture and risk-taking behaviour.

The findings of this study imply that in financial organizations, mechanisms to cope with job stress, and a focus on creating and retaining a sound risk culture are crucial in trying to avoid excessive risk-taking behaviour.

(3)

3

3.2 Sample description ... 26 3.3 Measures ... 27 3.3.1 Risk-taking behaviour ... 27 3.3.2. Risk Perception ... 28 3.3.3. Job stress ... 28 3.3.4. Risk Culture ... 28 3.3.5 Individual differences ... 29 4. Results ... 30 4.1. Sample ... 30 4.2 Reliability analysis ... 32 4.3 Normality analysis ... 33 4.4. Correlations... 34

4.5 Differences between groups ... 40

4.6 Hypothesis Tests ... 40

4.6.1. Hypothesis 1. ... 40

4.6.2 Hypothesis 2 ... 42

(4)

4

4.7 Additional analyses ... 50

5. Discussion ... 52

5.1. Discussion of the results ... 52

5.2 Implications for theory and practice ... 56

5.3 Limitations and directions for future research ... 59

5.4. Conclusions ... 62

References ... 63

Appendix A. Survey ... 71

Appendix B. Normality Analysis ... 102

(5)

5

1. Introduction

The case: the banking sector

In 2008 John Robbins, long-time industry executive and former chairman of the Mortgage Bankers Association was asked: ‘What caused the financial crisis?’. He answered: ‘During the lending boom, the industry developed products that were extremely risky that were pushed by everybody up and down the food chain. We forgot about our customers, and making money and our commission checks were more important.’ (‘What caused the financial crisis?’, 2008)

As Robbins (2008) described, the financial crisis of 2007-2009 started after the sale of

extremely risky products. The people selling these products might have known that they were engaging in risk-taking behavior, but decided at that moment that the benefits outweighed the costs. Despite the fact that one would expect smart people to stop this risk-taking behavior in favor of money-making and the earning of commission checks, it didn’t happen. The financial crisis began and resulted in the largest realization of bank risk since the Great Depression (Cole & Ohanian, 2004). More than 3 trillion euros were erased from the market

capitalization of banks in Europe and the United States. This corresponds to a decrease of 82% in the stock market value of these banks between May 2007 and March 2009 (Altunbas, Manganelli & Marques-Ibanez, 2011). The impact of the problems in the banking sector on the global economy were severe, producing high levels of unemployment, loss in value and loss of trust in banks from customers, employees and governments (Campello, Graham & Harvey, 2009).

As the example of the Financial Crisis of 2007-2009 illustrates, the pre-eminent sector where risk management is of extreme importance is the banking sector. The Financial Crisis has forced a re-examination of macroeconomics, financial economics, regulation and risk management (Gray & Jobst, 2010). But still, as is described in the report of the European

(6)

6

Banking Authority (2013), there is limited improvement in market confidence of both debt and equity investors, and there are signs of a general weak macro-economic environment. Consequently, significant challenges within the EU banking sector continue to persist due to an increase in regulatory rules and continuing asset quality deterioration.

Furthermore, following the crisis, the current pressure on the banking sector is enormous: the clients, the government and the European Union (EU) are imposing pressure on banks in Europe. For example, clients who have lifelong relationships with their banks are losing trust in the financial organizations because of stories about affairs (for example Libor) in the news. On the national level, the Autoriteit Financiële Markten (AFM) and De

Nederlandsche Bank (DNB) tightened the requirements for banks and on the European level the European Central Bank uses stress-tests (monitoring and assessing of market

developments, identifying of trends, potential risks and vulnerabilities) to ensure the orderly functioning and integrity of financial markets and the stability of the financial system in the EU. However, in order to provide a clear and complete analysis of what causes bank risk, it is necessary to look at three different levels: the context-level, the firm-level and the individual-level. The context level is illustrated by the measures and analysis of the EU on the banking sector. The firm-level concerns the organization-specific methods to deal with the current pressures. For example the activities of the Risk Management Department. The individual level which is about individual risk-perception, decision-making and behaviour, is

underexposed.

Nevertheless, individual differences seem to play an important role in the process of risk-taking behaviour. As previous studies show, risk-taking behaviour is arising from a low risk perception (Baird & Thomas, 1985). Gender (Powell & Ansic, 1997), age, (Dohmen et al. 2009), genetics (Kuhnen & Chiao, 2009), attitude (Eckel & Grossman, 2002) and level of experience (Wiseman, 1998) all have major impact on individual risk perceptions, decisions

(7)

7

and behaviours which stresses the importance of individual characteristics as part of the subject of this study. Besides individual characteristics, the context the individual is surrounded by, or more specific the risk culture of an organization, also influences risk-taking. Risk culture is defined as the norms and traditions of behaviour of individuals and of groups within an organization that determine the way in which they identify, understand, discuss, and act on the risks the organization confronts and the risks it takes (FSB, 2014). As research shows, risk culture is related to firm-risk taking. For example, the type and degree of managers’ incentives are significantly related to firm risk taking (Wright et al. 2006). Also, the level of accountability, or the feeling of responsibility of individuals influences the degree of conservatism, or risk aversion, in organizational decision-making (Gomez-Mejia et al., 2000). In addition to individual characteristics and risk culture, different types of pressures are also related to the degree of risk taking of individuals. For example, peer pressure increases risky-decision making (Gardner & Steinberg, 2005) and imagine that one needs to fetch a target; it seems logical that a strong pressure to perform in any way might lead to individuals taking risks. Another form of pressure on employees is job stress, which is currently a serious problem in the financial sector. Due to the crisis, bank-employees are strictly controlled and being watched which results in intensive pressure on their behaviour and performance.

Job stress, individual differences and risk culture will play a central role in this study, and will be explained in more detail later.

Given the above described empirical findings and the urgency of guarding against excessive risk-taking behaviour in the financial sector, this study will focus on the dependent variable taking behaviour. The main question is: Is there a relation between job stress and risk-taking behaviour? More specific, do employees engage in risk-risk-taking behaviour earlier when under stress? And is it the risk perception of employees that changes under pressure or stress

(8)

8

and makes them behave in a more risky manner? In other words, is risk perception the mediating factor in the expected relation between job stress and risk-taking behaviour? The role of the risk culture as a moderator is also discussed: does the risk culture influence the expected relation between job stress and risk-taking behaviour mediated by risk perception? Additionally the expected effects are controlled for the variables gender, experience and risk propensity. The expected relations are shown in figure 1.

- +

- -

Figure 1. Research Model: The expected relation between job stress and risk-taking

behaviour, mediated by risk perception. A moderating effect is expected for risk culture on the relations between job stress and risk perception, and risk perception and risk-taking behaviour.

Job stress

Risk Culture Risk Culture

Risk Perception Risk-taking Behaviour

(9)

9

2. Theoretical background, conceptual model & hypotheses

The new rules and procedures of organizations, the national government and the EU described in the introduction, all try to control the same variable: risk-taking behaviour within financial organizations. Risk-taking behaviour of individuals was one of the important factors that caused the financial crisis. This stresses the importance of analyzing how individuals assess risk, why humans engage in risk-taking behaviour and which factors may increase or decrease the chance that humans make risky decisions.

In the following sections the variables and expected relations as shown in figure 1 will be discussed and the hypotheses will be specified. But before starting with this, the central theme of this study, risk, will be further defined.

2.1.1. What is risk?

Whether the activity is riding a bike or investing in the stock market, every day we are exposed to many forms of risk. Individuals thus encounter risk in a wide range of situations with different consequences associated with risk. Since scholars do not have the same opinion about how to define or measure risk, a wide range of descriptions and measurements have developed over time according to a sample of work by Slovic (1964a), Payne (1973a), and Weber (1988). According to Rohrmann and Renn (2000) there is no commonly accepted definition for the term risk, neither in the sciences nor in public understanding. In exact sciences such as engineering and physics, formal definitions based on the probability and physical measurement or corresponding utilities of negative outcomes are preferred; quantification of probabilities and outcomes lie at the core of this approach. In the social sciences , the ‘meaning’ of risk is a key issue, and qualitative aspects of risk are seen as crucial facets of the concept (Rohrmann & Renn, 2000). Elmiger and Kim (2003) define risk as follows:

(10)

10

Risk = Danger + Opportunity

This equation demonstrates risk as a combination of potential loss and gain. This point of view on risk reveals the relationship between danger and opportunity in which the trade-off that everyone has to make between the higher rewards that potentially come with the

opportunity and the higher risk that has to be borne as a consequence of the danger (Elmiger and Kim, 2003).

In traditional theories risk is classified as something quantifiable, so it is measured by the volatility of returns and individual trade-offs between risk and return (Diacon, 2004). This idea forms the essence of the expected utility model proposed by von Neumann &

Morgenstern (1944). They argue that individuals maximize their returns while minimizing their risks, with due consideration of the axioms that individuals are (1) completely rational, (2) able to deal with complex choices, (3) risk-averse, and (4) wealth maximizing.

Nonetheless, proponents of behavioral finance argue that individuals may not be rational at all times. Kahneman & Tversky (1979)propose an alternative theory, the prospect theory, which looks at the cognitive limitations of the individual decision-makers. Specifically, they argue that individuals will be risk-averse in a gain situation and risk-seeking in a loss situation. As described in classical decision theory, risk is most commonly conceived as reflecting variation in the distribution of possible outcomes, their likelihoods, and their subjective values. A risky alternative is one for which the variance is large; and risk is one of the attributes which, along with the expected value of the alternative, are used in evaluating alternative gambles (March & Shapira, 1987). To clarify, expected value is assumed to be positively associated, and risk is assumed to be negatively associated, with the attractiveness of an alternative (Arrow, 1965). Thus, individuals will make decisions based upon the attractiveness of alternatives.

(11)

11

2.1.2. What is risk-taking behaviour?

In existing literature attention is paid to various forms of risky decisions and, in the end risk-taking behaviour, including substance use, dangerous driving, promiscuous sex, and

delinquency (reviewed in Mishra & Lalumière, 2009). Gambling has been associated with various forms of risky behaviour (e.g., Martins, Tavares, da Silva Lobo, Galetti, & Gentil, 2004; Powell, Hardoon, Derevensky, & Gupta, 1999; reviewed in Van Brunschot, 2009), and shares instigative aspects associated with general risky behaviour (reviewed in Stinchfield, 2004). As becomes clear whilst reviewing the literature, various aspects of personality traits like sensation seeking, impulsivity and low self-control are related to risk-taking behaviour (Mishra et al., 2010). Decision-making tendencies and attitudes are again associated with risky behaviour and may form a causal mechanism underlying the risk-taking behaviour of individuals. Scholars have argued that risk-taking behaviour should be studied because of its relevance to three important issues in the field of psychology: the adaptiveness of human behaviour (Byrnes, 1998; Payne, Bettman & Johnson, 1993), the rationality of human thought (Baron, 1994), and the relative importance of genes versus the environment in determining the phenotypic expression of traits (Wilson & Daly, 1985, Zuckerman, 1991). In essence, risk-taking behaviour implies that someone faces some kind of danger, consciously or

unconsciously, but at the same time has some kind of opportunity in prospect. This research is focused on the context of the financial industry in which risk-taking behaviour is related to ethical and financial decision-making and behaviour. Balancing between a client’s interest and the self-interest of an employee illustrates the role of ethics in financial decision-making.

2.1.3. What influences risk-taking behaviour?

Risk-taking behaviour is a variable that is influenced by multiple factors. Among the different factors guiding one’s decision-making processes the willingness to take risk might be the most crucial one influencing one’s overt behaviour, as most judgments require us to deal with

(12)

12

risky situations or dilemmas. Risk-taking decisions could be considered as the product of complex interactions of several cognitive factors, including availability of cognitive resources, memory processing, and the choice of decision strategies. Recent research has suggested that mood exerts its influence on all these factors (Yuen & Lee, 2003, based on Teasdale, 1993). Likewise, psychologists have demonstrated that emotional reactions have a fundamental role as motivators of human behaviour and decision-making (Finucane et al., 2000) and can be used as information about the quality of decision alternatives or as a

feedback to evaluate the outcomes of one’s choices (Schwarz & Clore, 2003). Despite the fact that emotions can be useful to inform the decision maker, there are plenty of situations in which emotional reactions lead people to make mistakes, as shown by research on risk perception (which will be discussed later). Emotional feedbacks are often experienced in an automatic, unconscious way, which may cause a negative impact on decisions that are better taken in a conscious and informed manner (Finucane et al., 2000). For example people in a positive state turned out to be more optimistic and therefore more willing to take risks (Rolison & Scherman, 2002). More precisely, positive aroused feelings are associated with anticipation of gain (excitement) and may promote risk taking, whereas negative aroused feelings associated with anticipation of loss (anxiety) may promote risk aversion (Knutsen et al., 2005; Paulus et al., 2003).

2.2. Job stress

As mentioned above there are different factors that might influence someone’s risk perception and risk-taking behaviour. According to the literature, cognitive limitations, controllability and the degree of uncertainty influence someone’s risk perception and thus someone’s tendency to engage in risk-taking behaviour (Hamid et al. 2013). In the context of the financial sector, it seems logical that a combination of influencing factors at least partly explains why certain individuals are more likely to engage in risk-taking behaviour than

(13)

13

others. In the existing literature a number of influencing factors are emphasized.

Firstly, Diacon (2004) proves that an individual’s risk perception is influenced by the fact that humans try to reduce uncertainty when facing problems. When experiencing

uncertainty while making decisions about risk, individuals experience pressure to make a good decision. This experienced pressure influences the way the individual perceives the risky situation, or the risk perception of the individual and thus the tendency to engage in

risk-taking behaviour.

A second factor of importance which influences the way an individual perceives risk is about the effects of a group an individual is located in. According to Wallach et al. (1962) risk and conservatism are the extent to which the decision maker is willing to expose him or herself to possible failure in the pursuit of a desirable goal. They found that group interaction and the achieving of consensus eventuates in the willingness to make more risky decisions than those that would be made in the absence of group interaction. This implies that

circumstances of the group the individual is part of may influence the degree of risk-taking behaviour. Knudsen (2007) made an analysis of a few effects that are found to be important determinants of the risk strategy that is chosen based on the reference group one is part of. One of the effects is referred to as reference group sensitivity and is defined as the degree to which decision makers’ aspirations adapt to the average wealth of the population in which they are currently located in. An example is a manager who uses a benchmark against other firms, or an employee comparing his or her decisions with the ones of his or her colleagues. The main result of this analysis is that the reference group is seen as a dominating driver of performance. It provides a more important basis for risk strategies than the decision makers own situation (Knudsen, 2007). In addition, Osborn (1957) has reported that group interaction may lead to quite radical, bold, problem solutions. In line with the idea that the group may influence the risk-taking behaviour of the individual, Bougheas et al. (2013) investigated

(14)

14

whether individual choices under risk may be swayed by the opinions and decisions of others. They find that the level of making risky decisions is highest in groups, compared to the level of making risky decisions for individuals. Likewise, this implies that the risk-taking behaviour of individuals is influenced by the context the individual is surrounded by. Gardner and Steinberg (2005) also demonstrate that peer pressure increases risky-decision making and thus risk-taking behaviour.

A third factor that influences the way individuals make risky-decisions and thus influences the way individuals face risky situations and the way they behave in these situations, is the promised compensation. Ross (1981) investigated that individual human decision makers are risk averse. This means that when faced with one alternative having a given outcome with certainty, and a second alternative which is a gamble but has the same expected value as the first, an individual will choose the certain outcome rather than the gamble. The scenario becomes more complex when a more risky alternative might lead to nothing at all, or might lead to an even better scenario compared to the secure alternative. For example, a secure scenario explains that a person will earn €500 and the risky alternative explains that a person has a 50% chance of earning €1500 and the other 50% chance of earning nothing at all. This indicates that the compensation for a decision plays a major role when people decide to go for a certain or a risky choice. In a study regarding the effect of targets and incentives, portfolio holdings of mutual funds were examined in September and in December and it was demonstrated that mutual funds do alter the riskiness of their portfolios at the end of the year in a manner consistent with their incentives to receive (Chevalier and Ellison, 1995). The presence of a target or incentive thus seems to strengthen the pressure to perform on the individual and increases the degree of risk-taking.

A fourth factor of importance concerns the influence of past experiences and

(15)

15

influenced by the outcome history, the role of previously made decisions and their outcomes. The model is confirmed by their results. To illustrate, individuals who have experienced positive outcomes after risky-decisions in the past, are more likely to take risky-decisions when facing new situations. Also, Bowman (1984) demonstrated that individuals that suffer from low performance results, are more inclined to take risks, which increases bad gambles, or negative risky decisions, which again lowers future performance.

The four factors above all indicate an increase in risk-taking behaviour, and these are just a few of the possible factors that put pressure on employees in their organizational context. Summarized, all these factors cause employees to experience job stress, one way or the other. Therefore, in the case of the banking sector it is hypothesized that individuals who experience a high level of job stress from any source, differ in their risk perception and risk-taking behaviour of individuals that experience a lower degree of job stress. In more detail, because of the evidence that job stress, or pressure, pushes individuals to perform it is assumed that it makes individuals more eager to achieve results, with the effect that individuals pay less attention to risks, or attach less value to risks, which lowers their risk perception and thus makes them more willing to engage in risk-taking behaviour. Therefore, individuals who experience a high degree of job stress are inclined to less quickly perceive risks in a risky situation and thus engage in risk-taking behaviour, as individuals who

experience a low degree of job stress and vice versa. Hereby job stress can be defined as any stress of which the individual is suffering, caused by pressure to perform well, resulting from issues such as having a concrete target, facing a certain positive stimulus such as a financial compensation or promotion, dealing with previous disappointing results, feeling insecure about one’s job-position, experiencing pressure to perform from colleagues or from a manager etc.

(16)

16

Hypothesis 1: Individuals who experience a high degree of job stress are more inclined to engage in risk-taking behaviour.

2.3 Risk Perception

Before individuals proceed to act in a certain situation, and thus engage in risk-taking behaviour, they either consciously or unconsciously, form a perception about the situation. Risk perception is defined as an individual’s assessment of how risky a situation is in terms of probabilistic estimates of the degree of situational uncertainty, how controllable that

uncertainty is and the confidence in those estimates (Baird & Thomas, 1985). According to Sjöberg (2000) several factors have been proposed for the explanation of perceived risk. The primary factor is real risk, or general risk. To illustrate, research has shown that average estimated mortality rates of a number of common illnesses and accidents were strongly related to statistical data (Lichtenstein, 1978). A second factor in risk perception is the role of

heuristics or cognitive biases. As Diacon (2004) describes these ‘mental short-cuts’ are used to simplify information. A well-known example is the representativeness heuristic as

described by Kahneman and Tversky (1973). The representativeness heuristic is defined as the degree to which an event is similar in essential characteristics to its parent population, and reflects the salient features of the process by which it is generated. When people rely on representativeness to make judgments, they are likely to judge wrongly because the fact that something is more representative does not actually make it more likely (Kahneman & Tversky, 1973). The last and third factor in risk perception concerns the target of the risk. People do not make the same estimate when they rate the risk to themselves, their families or people or organizations in general (Sjöberg, 2000).

In a study of Sitkin and Pablo (1992) evidence is found for a model of the determinants of risk-taking behaviour in which risk perception plays an important role. According to their theory risk-taking behaviour is determined by two individual factors,

(17)

17

namely risk propensity and risk perception. Since risk propensity is seen as a stable individual characteristic, this will be discussed later in a paragraph about individual differences that influence risk-taking behaviour. The model explains that risk perception is influenced by problem framing, the either positive or negative way a problem is framed by the individual. In addition, the way individuals shape a risk perception varies across individuals and is

dependent on their situation (Sitkin & Pablo,1992).

In other studies evidence is found for the influence of the estimated degree of uncertainty, controllability, and confidence on risk perception in a problematic situation (Hamid et al, 2013). If changes in risk perception are the driving force in the estimation of risks and thus risky-decision making and in the end behaviour, then effective remediation should affect cognitive processes, with information aimed at more realistic risk perception (Weber & Milliman, 1997). According to Diacon (2004) risk perceptions are important factors in environments where individuals have limited information and are boundedly rational and when there is no universally agreed understanding of how risk should be conceptualized or measured. The case of the financial sector provides a prime example of such an environment, and therefore risk perception is expected to influence the relation between the independent variable job stress and the dependent variable risk-taking behaviour. The presumption is that a higher level of job stress, ensures a lower risk perception and thus an earlier tendency to engage in risk-taking behaviour. Assuming that risk perception declares the effect between job stress and risk-taking behaviour, risk perception is called a mediator variable.

Hypothesis 2: The hypothesized relation between job stress and risk-taking behaviour is mediated by risk perception: job stress is expected to lower the risk perception which in turn is expected to increase risk-taking behaviour.

(18)

18

2.4 Risk Culture

The research model used in this study not only concerns factors that might increase the expected relation between job stress and risk perception (such as gender, experience or propensity), but especially concerns a factor to diminish the expected relation. The research model emanated from the real-life case described in the introduction, the banking sector. Weaknesses in risk culture of banks are often considered a root cause of the global financial crisis. Therefore, after the recent issues in the financial sector, looking at factors to improve a solid risk culture at banks is important (De Nederlandsche Bank, 2009). The context the individual is surrounded by is part of the organizational culture. According to Schein (1990) the problem of defining organizational culture derives from the fact that the concept of

organization is itself ambiguous. He states that we cannot start with some cultural phenomena and then use their existence as evidence for the existence of a group. He proposes a three-level analysis when investigating culture. The three fundamental three-levels at which culture manifests itself are a) observable artifacts, b) values and c) basic underlying assumptions (Schein, 1990). The definition of risk culture that is most used in the financial sector is stated by the FSB (2014) and is as follows: the norms and traditions of behaviour of individuals and of groups within an organization that determine the way in which they identify, understand, discuss, and act on the risks the organization confronts and the risks it takes. In order to investigate whether employees are willing to take risks, one needs to focus on the risk culture of an organization.

Building on these insights the Financial Stability Board (FSB, 2014) aims to increase the intensity and effectiveness of supervision to reduce the risks posed by systematically important financial institutions and therefore proposes factors to support a sound risk culture. In their report of April 2014 the FSB describes that risk culture plays an important role when discussing risk perception or risk-taking behavior. A financial institution’s risk culture plays

(19)

19

an important role in influencing the actions and decisions taken by individuals within the institution and in shaping the institution’s attitude toward its stakeholders, including its supervisors. Therefore, according to the FSB, efforts should be made by financial institutions to understand an institution’s culture and how it affects safety and soundness. While various definitions of culture exist, supervisors are focusing on the institution’s norms, attitudes and behaviours related to risk awareness, risk taking and risk management, or the institution’s risk culture (FSB, April 2014). To support the financial institutions in Europe, the FSB has

identified four factors that contribute to the promotion of a sound risk culture.

The first factor of a sound risk culture stated by the FSB is tone at the top. Interest in the tone at the top of organizations has increased in recent years as stories have unfolded of major company crises in which the influence of CEOs and other senior managers has

contributed to a dysfunctional tone at the top. In particular, the financial crisis of late 2008 has focused attention on these issues again: dysfunctional practices took root, leading to a series of collapses (Amernic et al., 2010). Therefore, it is critical that the board and the senior management demonstrate adherence to sound risk management and the highest standards on integrity, as over time, their behavior will be emulated by the rest of the institution. As Schein (2004) describes, culture is created by shared experience, but it is the leader who initiates this process by imposing his or her beliefs, values and assumptions. In this study tone at the top is specified by the focus on exemplary behaviour of the manager of the individual. As specified by the FSB (2014), indicators of tone at the top are (1) leading by example; the manager has a clear view of the risk culture and of the behavioral and organizational consequences of this culture. He or she systematically monitors and assesses the prevailing risk culture and proactively addresses any identified areas of weakness or concern. Also, the manager (2) ensures common understanding and awareness of risk among his/her people and manages his/her people and him/herself to (3) learn of past experiences by communicating and

(20)

20

assessing lessons learnt from past events.

The second factor of a sound risk culture is accountability. In most settings,

accountability means one party (for example an individual, group, company or government) is said to be directly or indirectly accountable to another party for a certain action, process, output or outcome (Patton, 1992). Research indicates that the degree of accountability of individuals influences the degree of conservatism, or risk aversion in organizational

environments (Gomez-Mejia et al., 2000). In other words, individuals that feel accountable for their actions are less willing to take risks. According to the FSB (2014) accountability is shown by: (1) a policy of ownership of risk where employees are held accountable for their actions and (2) the awareness of individuals of the consequences for not adhering to the desired behaviours toward risk.

The third factor of a sound risk culture is effective communication and challenge. A financial institution’s sound risk culture encourages transparency and open dialogue with its employees on all levels. As is described in a recent research report, more interaction, open communication and ‘speaking up’ are seen as a good thing that should be encouraged (Power et al., 2012). Therefore, in order to promote a culture of effective communication and

challenge, mechanisms are established to facilitate communication, bring alternate views to the decision-making process, and support dialogue between manager and employee.

Indicators of effective communication and challenge are (1) openness to alternate views and (2) stature of control functions; control functions operate independently and have appropriate direct access to the management information and report periodically to the management (FSB,2014).

The fourth factor of a sound risk culture is incentives. The power of a positive culture in risk management lies in its ability to motivate employees to want to control risks because sound risk taking is valued and enforced. As research confirms, incentives can be used as a

(21)

21

means to stimulate employees to engage in certain behaviour (Jensen & Murphy, 1990). Indicators of incentives are (1) remuneration and performance management and (2) talent development (FSB, 2014).

The assumption is that in a sound risk culture, individuals always make accurate risk assessments. Therefore, in a sound risk culture individuals are not inclined to behave different or shape their risk perceptions in a different manner when they perceive pressure to perform. This so called moderator effect is expected on the two parts of the relation: between job stress and risk perception, since risk culture is expected to force employees to form accurate risk perceptions, and between risk perception and risk-taking behaviour, since risk culture is expected to force employees to not engage in risk-taking behaviour even after a risk

perception is formed. Thus, the hypothesis is that for individuals who experience the presence of (one of) the four factors proposed by the FSB, the expected relation between job stress and risk-taking behaviour, mediated by risk perception, will be less strong.

Hypothesis 3a: The expected negative relation between job stress and risk perception is reduced by the presence of a sound risk culture.

Hypothesis 3b: The expected negative relation between job stress and risk perception is reduced by the presence of the factor tone at the top.

Hypothesis 3c: The expected negative relation between job stress and risk perception is reduced by the presence of the factor accountability.

Hypothesis 3d: The expected negative relation between job stress and risk perception is reduced by the presence of the factor effective communication and challenge.

Hypothesis 3e: The expected negative relation between job stress and risk perception is reduced by the presence of the factor incentives.

(22)

22

Hypothesis 4a: The expected positive relation between risk perception and risk-taking behaviour is reduced by the presence of a sound risk culture.

Hypothesis 4b: The expected positive relation between risk perception and risk-taking behaviour is reduced by the presence of the factor tone at the top.

Hypothesis 4c: The expected positive relation between risk perception and risk-taking behaviour is reduced by the presence of the factor accountability.

Hypothesis 4d: The expected positive relation between risk perception and risk-taking behaviour is reduced by the presence of the factor effective communication and challenge. Hypothesis 4e: The expected positive relation between risk perception and risk-taking behaviour is reduced by the presence of the factor incentives.

Hypothesis 3 and 4 will be combined later into hypothesis 3, when the absence of the mediator effect is demonstrated.

2.5 Controls: Individual differences

The expected relation between job stress and risk-taking behaviour might not be the same across all individuals because the expected relation can be influenced by individual

differences. Therefore, this study will report a few control variables to control for the effect of gender, age, experience and risk propensity. The three main control variables will now be further explained.

2.5.1. Gender

In recent research, responses of men and women investors on risky decisions were analyzed and it turned out that men tend to take more risks than do women (Olsen & Cox, 2010). Also, research in the field of health and environmental hazards investigated that one of the most consistent findings to come from people’s perceptions on risk is that women express far

(23)

23

greater concern than men with regard to a large number of health and environmental hazards (Barke et al. 1997). Specific explanations for the gender differences in risk perception and risk-taking behaviour are first, that women are socialized to be nurturers, caregivers, and maintainers of life, whereas men are socialized to be economic providers and exploiters of resources (Steger and Witt, 1989). Second, women’s experiences as victims of crime, rape, and domestic violence create in them a greater sense of vulnerability that may sensitize them to other risks as well (Riger, Gordon, and LeBailly, 1978). Lastly, research into gender differences in financial risk decisions demonstrated the effect of testosterone in risk-taking behaviour. Individuals high in testosterone were more likely to choose risky careers in finance compared to individuals low in testosterone. Since men experience higher levels of

testosterone than women, it is reasonable to expect men to more quickly incline in risk-taking behaviour (Sapienza, Zingales & Maestripien, 2009).

Based on the above mentioned explanations and findings from other research it is expected that men have different risk perceptions than do women in the same situations and are more inclined to engage in risk-taking behaviour.

2.5.2 Experience

Byrnes (1999) describes that the experience and organizational level of an individual reduces but not eliminates the gender differences in risk-taking behaviour. Focusing on level and expertise of individual workers Diacon (2004) found that experts tend to perceive fewer risks, and are thus more willing to take risks, than do less experienced individuals. A decrease of uncertainty, an increase of controllability and positive former experiences might account for the fact that experts perceive less risks than do less experienced individuals and are thus more inclined to engage in risk-taking behaviour.

(24)

24

2.5.3. Risk propensity

Another characteristic that is supposed to play an important role in building a risk perception and assessing whether to take a risk, is risk propensity. Risk propensity is seen as a stable individual-difference construct, that is a variable that does not change over different stimulus domains or contexts (Weber & Milliman, 1997). Risk propensity refers to the notion that many decision makers have consistent tendencies to either take or avoid actions that they feel are risky (Kogan and Wallach, 1964; Harnett and Cummings, 1980; Sitkin and Pabo, 1992, reviewed in Keil et al. 2000). Risk propensity influences the relative salience of a situational threat or opportunity and thus leads to biased risk perceptions (Brockhaus, 1980). These biased risk perceptions might influence the degree of risk-taking behaviour as is expected in this study. Elaborating on this idea Sitkin & Pablo (1992) found support for the hypothesis that the higher the risk propensity of an individual, the lower the level of perceived situational risk, and thus the higher the chance the individual takes a risk. A possible explanation for this effect might be the influence of sensation-seeking characteristics. Higher sensation-seeking tendencies were affiliated with more risk-taking (Isen et al., 1982).

(25)

25

3. Methods

3.1 Procedure and research setting

The data of this research was gathered by use of an online survey. Participants were invited to participate in this research via an email with an invitation letter and direct link to the online survey (included in Appendix A). This email was sent via the research tool MWM²

(http://www.mwm2.nl/) to all members of the sample. The directors of the departments represented in the sample received an email with information one day before (included in the Appendix).

To stimulate the participants to fill in the survey, three dining vouchers were promised to be randomly distributed among the participants.

Survey administration started on June 2 2015 and was closed two weeks later on June 16 2015. The use of an online survey has a number of advantages. The respondents are

allowed to answer the survey at a convenient time, at their computer, smartphone or tablet and the administrative task of sending and receiving questionnaires and inputting the data into a data base is considerably reduced (Evans & Mathur, 2005). Moreover, with use of an online survey, the anonymity of the participants is easily ensured. Potential weaknesses of online surveys are a low response rate, privacy and security issues and the possibility of uneven sample distribution among test and control groups. Fortunately, in this research this was not the case.

To perform the statistical analysis, the statistical Software Package for the Social Sciences (SPSS) was used.

(26)

26

3.2 Sample description

In total 1613 employees of a financial organization were approached to participate in this research. 328 employees participated which means the response rate was 20,3%. The

employees were working in different departments throughout the organization, for the biggest part in the Netherlands but also abroad. The sample selection was made on a random basis by the Human Resources department. The departments representing the sample were business units ranging from frontline business to the back-office, and from local areas on the country side to metropolises in the world.

The first department contained employees of the local banks, working in the frontline business. 105 employees of this department contributed to the research. The second

department contained 50 internal auditors of the department Audit. The third department consisted of 30 employees of Credit Risk Management, and the fourth department was Risk Management with 119 employees participating. The last department consisted of traders working on the dealing room worldwide, of this group 24 employees participated. It is notable that the sizes of the departments in the sample vary, which is logical given the varying sizes of the departments in the organization. To check for possible significant differences between the departments, the departments will be incorporated into the control variables.

Of every department representing the sample, more or less half of the employees was placed in the control-group and the other half in the test-group. The survey of the test-groups started with a few questions about the participants’ jobs and their company, to prompt them to think of their identity as bank employees. (For example: ‘Do you see yourself as a real

banker?’, and ‘What is your function at the bank?’). Reminding the participants in the test-group of their identity as bank employees aims to launch a process that psychologists call ‘priming’. By filling in the work-related questions the participants’ awareness of being a bank employee is activated. The participants in the control-group are not primed, they started with

(27)

27

a few neutral questions about the employees’ hobbies and private life (For example: ‘How many times per week do you practice sports?’ and ‘How many times per week do you cook dinner?’).

3.3 Measures

3.3.1 Risk-taking behaviour

To measure risk-taking behaviour (dependent variable), a selection of items of the Risk-Behaviour Scale of Weber et al. (2002) was used. The original questionnaire consisted of a total of 101 items in five domains of risk: financial, health/safety, recreational, ethics and social. Considering the relevance to the case, only the items of the domains financial and ethics are used in this research. This resulted in a total of 20 items. Participants were asked to rate the likelihood that he/she will engage in the described behaviour. To illustrate, an

example of an item of the financial scale is: ‘Investing 10% of your annual income in a very speculative stock.’ An example of an item of the ethics scale is: ‘Using office supplies for your personal business.’ The degree of risk-taking behaviour was scored on a five point scale rating from 1= extremely unlikely to 5= extremely likely (that one engages in a certain behaviour).

To attempt to measure risk-taking behaviour as well as possible, a second measure for risk-taking behaviour was incorporated in the survey. This measure consisted of two cases of the Choice Dilemma Questionnaire (CDQ) of Kogan & Wallach (1964), and the idea was that these two cases forced the participants to make a choice about risks in concrete behaviour. The participants read the case and then the question is what the probability is that they would consider the risk to take. The two measures are included in the survey, included in Appendix A.

(28)

28

3.3.2. Risk Perception

As risk-taking behaviour, risk perception (mediator variable) is measured with a selection of items of the Risk-Behaviour scale of Weber et al. (2002). The exact same items are used, but with a different question and answer scale. This time, the participants were asked to rate the degree of riskiness of the described situation. Risk perception was scored on a five point scale rating from 1= not at all risky to 5= extremely risky.

3.3.3. Job stress

In this research job stress (independent variable) was set up as an overarching variable, measuring ‘group pressure’, ‘concrete targets’, ‘financial compensation’, and ‘past results’. Given the difficulty of creating a questionnaire that measures all of these subscales, and keeping in mind the length of the total questionnaire, it was decided to use one existing scale for this variable, ‘job stress’. To measure job stress the survey of Parker and Decotiis (1983) was used. The survey contained 13 items that were scored on a 4-point Likert scale (ranging from 1= I do not agree to 4 = definitely agree). Examples of the items are: ‘Working here makes it hard to spend enough time with my family’ and ‘I have too much work and too little time to do it in’.

1 item was reverse coded meaning that a relatively low score indicated a relatively high level of job stress. The item was recoded.

3.3.4. Risk Culture

The variable risk culture (moderator) is measured with a self-made instrument. The variable is built up from four subscales, based on the indicators of risk culture as described by the FSB (2014). All subscales contain four items. The first subscale is ‘tone at the top’. An example of an item is ‘The members of the Executive Board are ‘walking the talk’: they behave according to the goals an guidelines they maintain for all employees of the organization’. The second subscale is ‘accountability’ and is illustrated by the following example: ‘I am aware of the

(29)

29

risks I am responsible for in my function’. The third subscale is ‘effective communication & challenge’, an example of this subscale is: ‘I feel comfortable to share my ideas with my colleagues even if they have a different opinion’. The last subscale is ‘incentives’. An example of an item is: ‘The compensation I earn supports me to act in the interest of the greater good of the company, rather than for myself or my business line’. Risk culture was scored on a five point scale ranging from 1= strongly disagree to 5=strongly agree. 1 item was reverse coded meaning that a relatively low score indicated a relatively high level of a sound risk culture. This item was recoded.

3.3.5 Individual differences

To measure individual differences (control variable) a combination of questionnaires was used. First, the participants were asked to fill in their gender, age, and years of experience in this or a similar job. Second, to measure risk propensity six items of the General Risk

Aversion Scale of Mandrik and Bao (2005) were submitted. The answers were rated on a five point scale from 1= strongly agree to 5= strongly disagree. An example of an item is: ‘I do not feel comfortable about taking risks’.

The questionnaire also contained two items in which the participant had to choose between a risky alternative and a less risky alternative (Kahneman & Tversky, 1979). Initially the idea was to use these two items to measure risk propensity. However, it was confusing whether these items measured risk propensity, risk perception or risk behaviour and therefore it was decided to remove these items from the analysis.

(30)

30

4. Results

4.1. Sample

First, the collected data was screened to check for missing values and outliers, and

consequently analysed to provide an insight on the descriptive statistics of the sample. Of the 1613 invited participants, 434 started the survey (26,9%) and 328 (20,3% completed the survey. There were no missing values thus the total of valid cases (328) were kept for the analysis. Of the sample, 236 participants are male (72%) and 92 are female (28%). The ages of the participants were reported in categories. 6 participants (1,8%) were younger than 25 years old, 85 participants (25,9%) were between 25 and 35 years old, 111 participants

(33,8%) were between 35-45 years old, 98 participants (29,8%) were between 45 and 55 years old and 28 participants (8,5%) were 55 years or older. The number of years of experience in the current, or in a similar job was also measured in categories. 28 participants (8,5%) had experience of 1 year or less, 98 participants (29,9%) had between 1 and 5 years of experience, 80 participants (24,4%) had between 5 and 10 years of experience, 50 participants (15,2%) had between 10 and 15 years of experience and 72 participants (22%) had 15 or more years of experience. The 328 respondents were randomly assigned to either the test-group or the control-group. The test-group contained 154 participants and the control-group contained 174 participants.

(31)

31

Variable Level N % Cummulative %

Gender Male 236 72% 72%

Female 92 28% 100%

Age < 25 years old 6 1,8% 1,8%

25 – 35 years old 85 25,9% 27,7% 35 – 45 years old 111 33,8% 61,5% 45 – 55 years old 98 29,8% 91,3% > 55 years old 28 8,5% 100% Experience < 1 year 28 8,5% 8,5% 1 – 5 years 98 29,9% 38,4% 5 – 10 years 80 24,4% 62,8% 10 – 15 years 50 15,2% 78% > 15 years 72 22% 100%

Department Local bank - business 105 32,0% 32,0% Audit 50 15,2% 47,3% Credit Risk Management 30 9,1% 56,4% Risk Management 119 36,3% 92,7% Trading 24 7,3% 100% Group Control-group 174 53% 53% Test-group 154 47,0% 100%

Variable Min Max Mean SD

Risk-taking behaviour (measure 1) 15,00 44,00 25,23 6,17 Risk-taking behaviour (measure 2) 2,00 12,00 7,55 2,14 Risk Perception 38,00 79,00 57,83 9,16 Job stress 11,00 50,00 23,92 7,89 Risk Culture – general 18,00 73,00 35,73 7,10 Risk Culture – tone at the top

4,00 20,00 9,87 2,61 Risk Culture – accountability 4,00 20,00 7,34 2,33 Risk Culture – effective communication and challenge 4,00 20,00 8,79 2,38 Risk Culture – incentives 4,00 20,00 9,72 2,12

(32)

32

4.2 Reliability analysis

A reliability analysis was performed to assess to what extent the items that were supposed to measure one single construct, actually did measure one and the same construct. The reliability of the four variables was measured with Cronbach’s alphas.

Most of the scales had reliable alpha’s, exceeding the general acceptance level of α = .70 (DeVellis, 2003). A scale that did not exceed the general acceptance level was the second measure for risk-taking behaviour (the two cases of the Choice Dilemma Questionnaire (CDQ) of Kogan & Wallach (1964)) with a Cronbach’s alpha of α = .515. In combination with the absence of a significant correlation with the first measure of risk-taking behaviour (as will be described in paragraph 4.4), it was decided that this measure is not taken into further analysis.

The Cronbach’s alpha of the first measure of risk-taking behaviour was α= .780. The Cronbach’s alpha of job stress was α = .833 The Cronbach’s alpha of risk perception was α= .862 and the Cronbach’s alpha of the general scale for risk culture was α= .820.

There were also Cronbach’s alpha’s computed for the subscales of risk culture. Risk culture’s subscale tone at the top had a Cronbach’s alpha of α= .71, risk culture’s subscale accountability had a Cronbach’s alpha of α= .78, risk culture’s subscale effective

communication and challenge had a Cronbach’s alpha of α= .71 and lastly, risk culture’s subscale incentives had a Cronbach’s alpha of α= .45.

This last Cronbach’s alpha is below the general acceptance level of α= .70, which can be explained by the uncertainty that prevails about this topic among the employees of the financial organization. Another explanation for the low alpha can be the fact that the measure for risk culture was hand-made. Nevertheless, because of the low Cronbach’s alpha it is decided to not take this subscale into further analysis.

(33)

33

The control variable risk propensity was also tested for reliability and the results show that the Cronbach’s alpha of risk propensity was α = .791. All scales that are used in the analysis seem to be reliable and can be used to test the hypotheses.

4.3 Normality analysis

An assessment of the normality of data is a prerequisite for many statistical tests because normal data is an underlying assumption in parametric testing. To test the normality of data, the data is analysed graphically and numerically.

The dependent variable risk-taking behaviour had a mean of 25.23 and a 5% trimmed mean of 25.04, which indicate that the more extreme scores on risk-taking behaviour did not have a strong influence on the average mean (Appendix B). To assess whether the data was normally distributed, it was checked how symmetrical the distribution was by using the results for skewness, and to what extent the shape of the data distribution matched a Gaussian distribution by using the results of kurtosis. The positive skewness and negative kurtosis indicate that the scores were clustered to the left and more flat distributed than the Gaussian distribution (Appendix B).

The independent variable job stress had a mean of 23.92 and a 5% trimmed mean of 23.53, which indicate that the more extreme scores on risk-taking behaviour did not have a strong influence on the average mean (Appendix B). The positive skewness and positive kurtosis indicate that the scores were clustered to the left and peaked.

The mediator variable risk perception had a mean of 57,83 and a 5% trimmed mean of 57.74, which indicates that there were no extreme scores having a strong influence on the average mean. The positive skewness and the negative kurtosis indicate that the scores were clustered to the left, and that the distribution is flatter than the Gaussian distribution

(34)

34

The moderator variable risk culture had a mean of 35,73 and a 5% trimmed mean of 35,54. Again, this indicates that there were no extreme scores having a strong influence on the average mean. A positive skewness and a highly positive kurtosis were reported, which indicate that the scores were clustered to the left and highly peaked (Appendix B).

The significant results of the Kolmogorov-Smimov statistics for all four variables indicate a violation of the assumption of normality. However, by plotting the Q-Q plots for each of the variables, it is clearly showed that the dots reasonably follow a straight line (Appendices B). Therefore it is concluded that the assumption of normality is not seriously violated and the statistical test can be conducted.

4.4. Correlations

In this section the correlations of all variables are checked. Before the correlation analyses, the categorical variables: group, gender and department are transformed into dummy variables. The dummy variable became respectively for gender: 0 = female, 1 = male. For group the dummy variable became: 0 = control-group, 1 = test-group. The variable

department consisted of 5 departments. The baseline is chosen to be the department Audit, because it is expected that this department shows the least risk-taking behaviour.

Subsequently, there are four dummy variables made that are used in the analyses: the first is coded 1 = local business, 0 = risk, 0 = credit risk and 0 = trading. The second 1 = credit risk, and all other departments are coded 0, the third 1 = risk, and all other departments are coded 0 and the last: 1 = trading and all other departments are coded 0. The fifth dummy variable for the department Audit is included in the correlation matrix. It is coded: 1 = Audit, and all other departments are coded 0.

The correlation matrix, presented in table 2, was checked using the Pearson correlation coefficients and their significance. In the matrix a ** indicates a significant coefficient at

(35)

35

0.01-level (2-tailed) and * indicates a significant coefficient at the 0.05-level (2-tailed). The first column in the matrix shows the correlation between the dependent variable risk-taking behaviour and the other variables. Risk-taking behaviour (1) does not show a significant correlation with the second measure for risk-taking behaviour (2). Because risk-taking behaviour (2) also has a low Cronbach’s alpha, this might not be a good measure which may also account for the lack of correlation between the two measures for risk-taking behaviour. Therefore, the second measure of risk-taking behaviour will not be used further in the analysis. As is expected, risk-taking behaviour is positively correlated with the independent variable job stress (r= .159**, p= >.01), and negatively correlated with the mediator variable risk perception (r= -.612**, p= >.01). The moderator variable risk culture is significantly and negatively correlated to risk-taking behaviour (r= -.205**, p= >.01) as do the subscale’s tone at the top (r= -.141*, p= >.05), accountability (r= -.244**, p= > .01) and effective

communication and challenge (r= -.155**, p= >.01).

The control variable group and gender are not significantly correlated with risk-taking behaviour, in contrast to risk perception which is significant and positively correlated with the dependent variable risk-taking behaviour (r= .215**, p= >.01). In line with the expectations, this means that the employees that score higher on the personality trait risk propensity also score higher on risktaking behaviour. Age shows a significant negative correlation (r= -.164**, p= >.01) and the results also indicate a correlation for experience, which is significant and negative (r= -.133*, p= >.05). Judging from these correlations, as the years and

experience of employees increase, the tendency to engage in risk-taking behaviour declines. In the third column it is shown that age (r= .179**, p= >.01) and experience (r= .169**, p= >.01) correlate significantly and positively with the mediator variable risk

perception. As expected, risk perception might play a role in the estimation of risks and in the end, risk-taking behaviour. From the statistics we might infer that the personality trait risk

(36)

36

propensity also plays an important role within this effect, given the significant and negative correlation between risk propensity and risk perception (r= -.257**, p= >.01).Further, risk perception is significant and positively correlated with risk culture (r= .157**, p= >.01) what is explained by the significant positive correlation of the two subscales accountability (r= .205**, r= >.01) and effective communication and challenge (r= .119*, p= >.05). Risk perception is also significantly correlated with the control variables test-group (r= .149**, p= >.01) and gender (r= .125*, p= >.05). These results imply that there is a correlation between group (test or control) and the way the participants perceive risks. Possibly, the primer effect has worked on risk perception instead of risk-taking behaviour. Of the department variables only local banks (r=. 250**, p= >.01) and risk (r=-.133*, p= >.05) are significantly correlated with risk perception. The local banks correlate positively and the risk department correlates negatively with risk perception. This implies that compared to the other departments, the risk perception of employees of local banks are more clearly present, and the risk perceptions of employees of risk are less clearly present.

In the fourth column the results show that the moderator variable job stress is significant and negatively correlated for all types of risk culture (risk culture general; r= .312**, p= >.01, subscale tone at the top; r= .302**, p= >.01, subscale accountability; r= -.193**, p= >.01, subscale effective communication and challenge r= -.274**, p= >.01. These scales are all negatively related, which implies that a lower level of job stress is related to a more solid risk culture.

The moderator variable risk culture naturally correlates significantly with its subscales, but also shows a significant negative correlation with age (r= -.171**, p= >.01). A possible explanation is that older workers are more critical about their cultural environment and therefore report lower scores on risk culture.

(37)

37

(r= -.145**, p= >.01). This seems to imply that an increase in age is related to a decrease in the experience of the factor accountability in the organizational culture. Perhaps younger workers who recently started their job, feel more accountable for their actions than older workers, who have possibly already experienced that you are not always responsible for your own actions. Accountability as a subscale of risk culture is also significantly related with the department local business (r= -.114*, p= >.05) and the department risk (r= .160**, p= >.01). In comparison with the other departments, the local business department indicates a low presence of accountability and the department risk indicates a high presence of accountability. The subscale effective communication and challenge is found to be significantly related with gender (r= -.114*, p= >.05). This means that men have indicated a lower level of effective communication and challenge than women. Age is also significantly related to this subscale (r= -.126*, p= >.05) which means that older employees have reported a lower level of

effective communication and challenge. Looking at the department variables, accountability is significant and negatively correlated with the department trading (r= -.114*, p= >.05). Thus, the department trading has a low presence of accountability, compared to the other

departments. Some significant correlations can be found among the control variables. For example, gender is positively correlated for the departments local business (r= .110*, p= >.05) and trading (r= .149**, p= >.01) which means that within those two departments there are significantly more men working than women. Age and risk propensity are also significantly correlated (r= .162**, p= >.01). This positive correlation means that older people score higher on risk propensity. Because risk propensity is seen as a stable personality construct, this might illustrate a difference between generations. As is expected, employees working at the trading department are positively correlated with risk propensity (r= .176**, p= >.01), which

indicates that people working at the trading department are more intended to take risks by nature. Surprisingly, credit risk also shows a significant and positive correlation with risk

(38)

38

perception (r= .133*, p= >.05). Because this department is responsible for the approving of credit proposals, the management might be aware of the fact that their employees score high on risk propensity and are thus willing to take risks by nature. This offers them the chance to install extra controls to prevent excessive risk-taking. Age and experience are logically

correlated (r= .540**, p= >.01), and the departments risk (r= -.219**, p= >.01) and trading (r= -.283**, p= >.01) are negatively correlated with experience, which infers that more people working at these departments are less experienced in comparison with the other departments.

(39)

39

* correlation significant at the 0.05 level (2-tailed) and ** correlation significant at the 0.01 level (2-tailed)

Table 2. Correlationmatrix Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1. RTB (1) 1 2. RTB (2) .006 1 3. Risk Perception -.612** -.058 1 4. Job Stress .159** .022 -.032 1 5. Risk Culture -.205** .-77 .157** -.312** 1 6. RC – t.a.t -.141* .025 .101 -.302** .787** 1 7. RC – acc. -.244** .070 .205** -.193** .743** .393* * 1 8. RC – e.c.c -.155** .080 .119* -.274** .790** .566* * .504** 1 9. Test-group -.062 .020 .149** -.013 -.071 .009 -.057 -.043 1 10. Gender (male) .094 -.023 -.125* -.014 -.084 -.012 -.088 -.114* -.011 1 11. Risk Propensity .215** -.079 -.257** -.066 -.084 -.013 -.044 -.089 -.005 .088 1 12. Age -.164** -.061 .179** -.072 -.171** .011 -.145** -.126* -.061 .105 .162** 1 13. Experience -.133* -.069 .169** .027 -.061 .073 -.061 -.054 .068 .059 .081 .540** 1 14. Local business -.097 .012 .250** .061 -.023 -.021 -.114* .064 .075 .110* -.077 -.103 .026 1 15. Credit Risk -.024 -.067 -.021 -.095 -.015 -.070 .066 .018 -.023 -010 .133* -.128* -.022 -.218** 1 16. Risk .062 .037 -.133* -.092 .066 -.016 .160** -.053 -.062 .019 -.065 -.070 -.219** -.518** -.239** 1 17. Trading .014 -.029 -.048 .044 -.077 -.018 -.092 -.114* .064 .149** .176** .070 -.283** -.193** -.089 -.212** 1 18. ARG .053 .010. -.095 .087 -.026 -.118* .052 -.054 -.042 .000 -.048 .073 .072 -.291** -.135* -.320** -.119* 1

(40)

40

4.5 Differences between groups

To investigate whether the primers effected the test-group as expected, an independent samples t-test was conducted. The following table illustrates the samples.

Variable Groups N Mean SD

Dependent: Risk-taking behaviour Control-group 174 25,59 0,46 Test-group 154 24,83 0,51 Mediator: Risk Perception Control-group 174 56,55 9.16 Test-group 154 59,28 8.97

Table 3. Description of samples

This study found that there were no significant differences between the two groups on the dependent variable risk-taking behaviour, t(326) = 1.115, p = .423. This means that the primers did not work sufficiently on the dependent variable.

Surprisingly, a significant difference for the two groups is found on the mediator variable risk perception, t(326) = .007, p = <.05. This means that the primers effected the way in which employees perceive risks. The limitations and possible explanations of these findings will be discussed in the discussion.

4.6 Hypothesis Tests

4.6.1. Hypothesis 1.

The first hypothesis predicts that individuals who experience a high degree of job stress are more inclined to engage in risk-taking behaviour. In other words, a positive relation between job stress and risk-taking behaviour is expected.

A hierarchical regression technique is applied, as can be seen in table 4, model 1, first the explained variance of the control variables on the dependent variable risk-taking

A closer look at risk-taking behaviour among bankers : why a sound risk culture can prevent stressed bank employees of taking too much risk