Consumer credit adverts - risky business?

(1)

Consumer credit adverts - risky business?

Amy-Louise Dorrell (11781521)

August 10, 2018

MSc Economics (Specialisation Behavioural Economics and Game Theory), 15 EC

Abstract

This thesis describes an experimental attempt to influence risk aversion through priming, with the advertisement of consumer credit products. The purpose was to further contribute to wider literature in risk preference, and to test whether advertisements alone can indeed shift them. The experiment did not yield any significant results, except reproducing existing literature, in that females were significantly more risk averse. However, it does provide some evidence that the frequency of viewing consumer credit adverts may affect an individual’s level of risk aversion.

(2)

Statement of Originality

This document is written by Student Amy-Louise Dorrell, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are 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

"Credit card interest payments are the dumbest money of all." - Hill Harper

Risk preferences and risk aversion is a thoroughly studied topic in economics, starting with the humble beginnings of the utility function in 1738 (see Bernoulli, 1954), and developing to the more recent Prospect Theory (see Tversky and Kahne-mann, 1992). Defined as "future uncertainty about deviation from expected earnings or expected outcome" (The Economic Times, 2018), risk has many applications in economics and beyond - and is present in our everyday lives.

Many studies have attempted to decompose exactly what formulates an individ-ual’s risk aversion level - from demographic factors, to overconfidence, and other personality traits (Campbell et al., 2004; Chang et al., 2004; Grable, 2000). How-ever, work on the role that cognitive biases play on influencing risk aversion is less well-studied. Some literature has emerged finding evidence that framing and prim-ing can influence individual risk aversion, by changprim-ing the choice architecture that an individual faces (Erb et al., 2000; Kühberger et al., 1999).

To further explore this topic, this research aims to considers other forms of priming (other than through choice architecture), and its effect on risk aversion, and consequent risky behaviour. One form of priming is advertisements. It reaches into nearly every society across the globe. Advertising messages communicate with the viewer into feeling a certain way about a product, encourage them to buy, or even take an unlikely bet (Newall, 2015). The research question is thus: "Can risk aversion be influenced by consumer credit adverts?".

Advertisements are nearly impossible to get away from. If they do have an effect on people outside of just stimulating buying behaviour, it seems important to know how. Current literature has found that advertisements impact self-confidence (Trampe et al., 2010). But what other effects are they having on the citizens of the world? Do they need to be more stringently regulated in order to avoid welfare losses in society?

The particular type of advertisements focused on in this thesis surround con-sumer credit. Credit plays a role in many classical economical models - but occa-sionally in practice, individuals violate the rules of rationality, and behave in ways that are inconsistent with the theory. Falahati and Sabri (2015) argue that the ex-pansion of money and financial services, and the increasing ease of access to money, has increased people’s propensity to become indebted. Furthermore, several psy-chological studies have found that priming with money can influence behaviour in

(5)

a variety of ways. With these ideas in mind, consumer credit adverts seems well positioned to study behavioural anomalies in the risk domain.

The experimental design of this thesis utilises a between-subjects survey exper-iment with two different treatments - an Advert and an Education treatment, to which participants are randomly assigned. Each treatment presents the partici-pants with different information and will be further discussed in the Methodology section, and is the independent variable in this study. A well-established method to capture subjects’ level of risk aversion is then used (see Holt and Laury, 2002), formulating the dependent variable in this study. The final section of the survey focuses on establishing the control variables in order to avoid [omitted variable] bias in the results. However, there are no key findings in this experiment which are ro-bust. Some initial results did emerge, finding that the frequency of viewing adverts significantly affected risk aversion level - but due to small sample size in this study, the reliability of this is questionable, and no conclusions have been drawn.

2 Literature and Theoretical Review

This section examines existing literature on the topic of risk aversion, and discusses consumer credit’s role in economic theory. There will then be a short section dis-cussing advertising as priming and its role in influencing behaviour, motivating the research question of this thesis. Then, the literature covering risk aversion in the financial domain are analysed.

2.1 A primer: Risk within Economic Literature

Early work concerning decision-making under risk began with Daniel Bernoulli in the 18th century - originally published in 1738. He proposed a mathematical theorem concerning an individual’s utility of any possible profit expectation - by deriving a mean utility "moral expectation", which then corresponded to the value of the risk of the expectation (see Bernoulli, 1954). It was the first formulation of marginal utility through its demonstration that a poorer person gains higher utility from one additional euro, than that same euro gives to a richer person. Overall, Bernoulli’s ideas are fundamental concepts in economics to this day.

Bernoulli’s work and Expected Utility Theory was further developed by von Neumann and Morgenstern (1944), in their book Theory of Games and Economic Behaviour. Their decision theorem proposes that a rational agent behaves according to four axioms, completeness, transitivity, continuity, and independence. They also developed a utility function for this rational agent, who maximises their expected

(6)

utility when faced with risky outcomes. Their work has been cited almost 35,000 times, and has broad applications for economics. However, additional theories, such as the Allais Paradox (see Allais, 1990) find that these axioms are easily violated by individuals in reality, when considering risky decisions. Newer theories have given way to different approaches towards decision making under risk in economics, veer-ing away from the more mathematical to the physiological - and thus, behavioural economical.

Kahneman and Tversky (1979) are the first notable authors that advanced away from Expected Utility Theory when considering decision-making under risk. They developed Prospect Theory, which theorises that an individual’s willingness to take risks can be influenced by the way choices are framed - and that the choices indi-viduals make are not always optimal. Additionally, they prove that outcomes are different depending on whether the choice is framed as a gain, or a loss. The utility function is steeper for losses, since "losses loom larger than gains" (Kahneman and Tversky, 1979, p. 279). This directly contests Expected Utility Theory, which as-sumes that preferences should remain consistent, and should not change regardless of how the choices are described. Prospect Theory thus explicitly recognises that cognitively, individuals are biased in their decisions depending on a loss or a gain frame. This behavioural theory was arguably ground-breaking with its consideration of irrationality, and is still widely regarded as the "best available description of how people evaluate risk" (Barberis, 2013, p. 173) since its inception in 1979.

More recently, there has been further work on decision-making under risk, with the Risk-as-Feelings Theory being one example. It further expands on Prospect Theory by highlighting the effect of emotions when undertaking a risky decision (see Loewenstein et al., 2001). Additionally, cognitive neuroscience has begun to explore the neural basis or risky decision making, and find some early neural evidence for Prospect Theory. There is "suggestive [neurological] evidence that the processing of reward depends upon the context in which it occurs" (Trepel et al., 2005, p. 46), which further suggests that indeed, individuals do act according to Prospect Theory. Clearly, decision-making under risk is a well-contested area of economics. Given that Prospect Theory and the Risk-as-Feelings hypothesis argue that decision mak-ing under risk can be influenced both by frammak-ing and emotions, one can assume that different framing may well influence and individual’s level of risk aversion in other settings. One such setting is the financial domain.

(7)

2.2 Consumption theories and risk in the consumer credit

domain

Several economic theories are consistent with the idea that an individual consumes in an optimal way across their lifetime, in order to maximise their utility. While the Permanent Income Hypothesis considers the role of changing "permanent income" when making periodic consumption decisions (Friedman, 1957), while the Life Cy-cle Hypothesis argues that an individual borrows while they are young, anticipating future income to pay it off (Modigliani and Brumberg, 1954). Within these theories, and other consumption theories, there is almost always a role for credit and borrow-ing. In practice, as well as theoretically, consumer credit does play an important role in consumer behaviour and is well documented.

Consumer credit is often attributed to being an outlet for individuals to smooth consumption over their life, in order to maximise their lifetime utility. Brady (2008) argues that there is evidence that better access to credit can increase the likelihood of consumption smoothing, suggesting that consumers do use credit for its intended purposes. There is a further suggestion that risk-averse households may borrow to keep their consumption levels constant (Morduch, 1995), which is further evidence of the fact that consumer credit is used for consumers to smooth their consumption across time.

This seems to suggest that consumer credit is rational behaviour. However, some literature proposes the contrary. Individuals are subject to levels of interest on a credit product or loan; and should they misuse their credit, they are likely to incur fees. Indeed, Heidhues and Köszegi (2010, p. 2279) argue "nonsophisti-cated consumers overborrow, pay the penalties, and back load repayment, suffering large welfare losses" which suggests a clear violation of rationality, since a rational consumer would not have overborrowed in order to suffer a welfare loss.

There is plenty more literature to suggest that borrowers are not always ratio-nal. McCoy (2005) raises the question as to whether are rational economic actors maximising their expected utility, or if lenders induce sub-optimal decisions by ex-ploiting anomalies in consumer behaviour through marketing messages. She further argues that there is evidence for the latter, and that lenders take advantage of such cognitive biases. Chuah and Devlin (2011) agree that this violation of standard eco-nomic theory in could be due to mental effort, faulty reasoning, or the complexity of the decision-making process. As well as this, they argue that this behaviour is consistently observed and is fairly robust.

(8)

are responsible for misuse of consumer credit, and its subsequent violation of ratio-nality. One example is the notion of Present Bias, which theorises that an individual favours consuming in the present rather than the future. This is down to impatience, and weights decisions more in the short-run, relative to their long-run preferences. Several papers attribute consumer debt to present bias, due to individuals wanting instant gratification - and conclude that more impatient individuals are more likely to be debtors, or have higher debts (Hardisty et al., 2013; Ikeda and Kang, 2015; Meier and Sprenger, 2010). This is further evidence to the notion that people indeed do have preferences which are inconsistent with standard economic theory.

Self-control is another reason put forward as to why individuals may not behave rationally in the consumer debt sector. Strömbäck et al. (2017) found that higher levels of self-control are more likely to have a positive effect on savings behaviour and general financial behaviour, which supports the notion that a lack of self control may affect one’s behaviour in the financial domain. Another effect could be optimism - Yang et al. (2007) show that unrealistic optimism of consumers leads them to underestimate future borrowing, and choose credit cards which are not the optimal choice for them.

Arguably then, the area of consumer credit is one that will benefit greatly from the insights of behavioural economics, since there seems to be some departure from standard economic theory. Further evidence for the fact that behavioural economics should be considered is due to "predictable growth in consumer credit is significantly related to consumption growth, a finding that is inconsistent with existing models of consumer behaviour" (Ludvigson, 1999, p. 434).

A further motivation for this research question is due to the trend in consumer credit in Europe. The number of defaults on consumer credit has increased, with the European Central Bank reporting that there was 4.5 billion euros worth of consumer credit write-offs in 2008, growing from 2.7 billion euros in 2003 (Gerhardt, 2009). Further exploring the reasons behind this is interesting, and has potentially good policy implications.

Given the fact that there is an array of evidence that individuals are boundedly rational when it comes to consumer credit, there is also reason to believe that other

cognitive biases could lead to irrational behaviour. This thesis will attempt to

decompose whether advertisement, a form of priming, can cause irrationality in risk behaviour. This is aligned with Prospect Theory, due to the fact that it can make the underlying risk more complicated to understand.

(9)

2.3 Money Priming?

Several psychology studies have found that the mere idea of money can change behaviour. A study found that in four simulated buying experiments, individuals primed with credit card stimuli were more likely to give money to charity and tip in restaurants, with a greater amount, and in a shorter period of time (Feinberg, 1986). As well as this, Feinberg (1986, p.354) concluded that "the presence of credit card cues may elicit spending responses".

Money priming has been shown to effect individuals across demographics and domains - even children, not understanding the economic functions of money, became more selfish in economic games when primed with money (Gasiorowska et al., 2012). Another study found that money priming causes participants to distance themselves from others, and they seemed to care more about themselves than others (Hansen et al., 2012).

Finally, Chatterjee and Rose (2011) found that when subjects were primed with credit cards, they focused more on the benefits of a product than the costs. Clearly, money and credit priming can influence many different behaviours.

2.4 Previous Studies into Financial Risk Aversion

Studies into what formulates an individual’s level of risk aversion have been explored since well before the turn of the century. They began analysing demographic factors, before looking into other factors, such as returns on investments and opinions of the economy.

Grable (2000) was one of the prominent authors in financial risk taking, analysing which demographic factors are likely to affect risk-taking in "everyday money mat-ters". Using a survey questionnaire, respondents received an assessment to find their level of financial risk-tolerance, and their answers were given a weight accord-ing to the riskiness of the response. Results indicated that gender, age, occupation, income, education, financial knowledge and economic expectations were significant in differentiating between the levels of risk tolerance. The authors concluded: "(a) males were more risk tolerant than females, (b) older respondents were more risk tolerant than younger respondents, (c) married respondents were more risk tolerant than single respondents, (d) professionals (occupational status) were more risk tol-erant than those with lower incomes, (e) respondents with higher incomes were more risk tolerant than those with lower incomes, (f) respondents with higher attained education were more risk tolerant than others, (g) respondents with higher levels of financial knowledge were more risk tolerant than respondents with less knowledge."

(10)

Weber and Hsee (1998) found that nationality also had an impact on the level of risk aversion.

A number of other researchers come to the same conclusions in many of these areas (Clark and Strauss, 2008; Gilliam et al., 2010; Hallahan et al., 2004). There is a further suggestion that females are less comfortable with debt than males and manage their money better (Davies and Lea, 1995), which aligns with the literature stating that females are more risk-averse than males. As well as this, Gathergood and Weber (2017) found that poor financial literacy has an effect on a broad range of financial choices.

Whilst these papers provide insight on the demographic factors affecting risk preferences in individuals, they tend to evaluate them with self-reported survey tests. Several authors have devised other ways of determining risk preference using more robust choice architecture, a better reflection of actual behaviour.

Arrow-Pratt measures of risk aversion (see Arrow, 1965; Pratt, 1964) became "workhorses" (Ross, 1981, p. 621) for analysing decision-making under uncertainty, being used to examine risk-taking in many different economic contexts. However, this measure is determined by deriving the utility function - something not always available to experimenters.

Holt and Laury (2002) noted that while risk aversion was a fundamental element in theories of lottery choice, asset valuation, contracts and insurance, there were few established methods to actually model risk aversion in experimental settings. They constructed choice architecture to formulate an individual level of risk aversion, and their design "should become an important tool for... experiments in which risk attitudes could play a role"(Harrison et al., 2005, p. 897).

Although these studies illuminate on the factors on what affects the level of risk aversion, and suggest different ways of modelling them, the question on how this can be manipulated is less studied. Kühberger et al. (1999) find that framing outcomes as gains can induce risk aversion, while presenting outcomes as losses encourages risk seeking. Erb et al., (2002) also attempt to influence preference for risk by using different choice architecture, and conclude that these preferences can be affected quite strongly by priming.

However, these studies focus on changing the choice architecture itself, rather than other external priming cues. This research aims to further address how priming through advertising can affect an individual’s level of risk aversion.

(11)

3 Methodology

An online survey experiment will be conducted to investigate the research question: "Can risk aversion be influenced by consumer credit adverts?".

Given the literature, a survey was determined the best method to capture the level of risk aversion of the subjects. Many experiments have used surveys to not only measure risk, but also financial literacy and understanding propensity for in-debtedness (Grable, 2000; Limbu, 2017; Sjöberg and Engelberg, 2009).

3.1 Variables and Explanations

The dependent variable in this study is individual level of risk aversion. Subse-quently, "higher risk aversion" translates to the subject being more risk-averse than another subject (or, they are less likely to take risks), or their individual level of risk aversion having increased. From this point onwards, when considered as a variable, it will be abbreviated to "risk uptake" for simplicity. The method of measurement for this variable is drawn from Holt and Laury (2002). The subjects must choose between two lotteries, whereby the first becomes progressively riskier in terms of the estimated payout. When the subject changes from the first lottery to the second, the subject’s level of risk uptake can be determined - depending at which question they switched. The scale is from 0-10 - the number representing the number of questions remaining when they made the switch from Lottery A to Lottery B. Therefore, 0 is the lowest level of risk uptake (the most risk averse individuals), and 10 is the highest level of risk uptake (the least risk averse individuals). As a result, risk up-take can be interpreted as the level of risk that the subject took with regards to the lotteries.

The lotteries were scaled up from the original payouts used in Holt and Laury’s original experiment, up to 25 euros - since the maximum payout of 3.85 euros is relatively small to incentivise participation in a 20-minute survey. The payments were all scaled up by the same amount so as to not affect the lottery itself - in the same manner as Holt and Laury did in their own experiment. The exact figures used can be found in the Appendix.

The independent variable is treatment. This study concerns two different treat-ments to analyse the effect of adverts on risk uptake. The two treattreat-ments in question are named "Advert treatment" and "Education treatment" from this point forward. Each subject will view only one of these treatments in the survey experiment, in the between-subjects experimental design format.

(12)

-two credit card adverts, -two payday loans, one car finance. The Education treatment displays eight different statements describing the negative effects of being indebted,

since adverts warning about the dangers of debt were harder to source. These

treatments will be further discussed in the Survey Method subsection. The exact advertisements and information used for the two treatments can be found in the Appendix.

There will also be several control variables, in order to control for and clearly identify the relationship between the independent and dependent variable. Given the previous studies discussed in the Literature Review, some factors have had an effect on risk uptake. Therefore, the following variables will be included in this study: age; gender; nationality; country of residence; highest level of education; employment status; employment sector; annual income; financial dependents; number of credit products; frequency of adverts viewed, and debt literacy.

Annual income and highest level of education are measured on a numerical scale - 1 being the lowest level for both. Each increment represents an increase from one group to the next, for example, "Less than 20,000 euros" and "20,000 to 34,999 euros" represent 1 and 2, respectively.

The variables number of credit products, frequency of adverts viewed and debt literacy command further explanation. The number of credit products owned may have an effect on risk uptake. This is because if a subject is shown an advert for a credit product but they already own one, the desire to get another credit product may not be as strong as a subject who didn’t own one. As well as this, if a subject views consumer credit adverts more often, their response to the advert treatment may be less strong than another subject who rarely views this type of adverts. In this number of credit products variable, the subject will choose from a list how many credit products they own, and with the frequency of adverts viewed variable, the subject will pick the statement closest to their estimation of how often they view adverts.

Financial literacy has been shown to affect risk aversion and behaviour (Gath-ergood and Weber, 2017; Cox et al., 2015; Guiso and Jappelli, 2008), and so also should be included as a control variable in this survey experiment. Debt literacy was chosen as the derivative of financial literacy, since many classic financial lit-eracy questions also concern knowledge of investments or economic indicators, but in this setting, only questions concerning debt seemed relevant. Three questions, drawn from a paper by Lusardi and Tufano (2015) will be used, and a score from 0-3 will represent the number of questions the subject answered correctly. The exact questions can be found in the Appendix, and the correct answers are highlighted.

(13)

3.2 Hypotheses

The hypotheses are below. An explanation as to why these results are expected then follows.

• H(1): Subjects in the Advert treatment will be significantly less risk averse than the subjects in the Education treatment

• H(2A): Males will be significantly less risk averse than females

• H(2B): Older subjects will be significantly less risk averse than younger sub-jects

• H(2C): Subjects with higher debt literacy will be significantly more risk averse H(1) is the primary hypothesis, and directly answers the research question. This hypothesis is derived from the notion that priming may influence behaviour, and consequently, risk aversion, highlighted in the Literature Review section of this thesis. Previous studies on priming in order to influence a subject’s level of risk aversion focus on choice architecture, whereas this one focuses on priming itself. Instead of varying the choices presented to the subject, the ’choices’ (Holt and Laury risk elicitation questions) remain constant, while the nature of the priming will vary.

The secondary hypotheses [H(2A), H(2B) and H(2C)] have been formulated from the results of previous studies. Gender and age are widely contested as affecting risk aversion and so they are highlighted as variables of interest in the current study.

Although debt literacy affecting risk aversion is less studied, intuitively hypoth-esis H(2C) should follow. Research has found that financial literacy can explain behavioural factors such as materialism, compulsive buying, and propensity to in-debtedness (Potrich and Vieira, 2018). As well as this, Disney and Gathergood (2013) find that individuals with lower financial literacy levels may underestimate the cost of credit and be more likely to become indebted.

3.3 Survey Method

The survey experiment will consist of four sections and be financially incentivised. The financial incentive will be used only in the risk elicitation section, in order to achieve a representative sample of the subject’s actual level of risk aversion - and to encourage effortful participation.

(14)

1. The respondent is shown at random, one of two treatments, the "Advert" treatment, or "Education" treatment. In order to disguise the research ques-tion, subjects are asked to choose or arrange adverts and statements in order of their preference.

2. Then, respondents are presented with a Holt and Laury (2002) risk elicitation section. This section is incentivised, and respondents will be informed that the decisions they make in this section will determine their payout. The exact wording for the explanation will be explicit, and can be found in the Appendix. One of the subject’s choices will be selected at random, and then that choice will be randomised according to their preferences. The maximum payout will be 25 euros.

3. In the third section, respondents are asked three questions to determine their level of debt literacy.

4. Finally, standard demographic questions will be asked of the subject. These will formulate control variables in the Results section.

The survey questions will be provided in the appendix for replication purposes. The experimenter will attempt to reduce ambiguity in some of the questions by adding explanations of words where possible to increase internal validity. This is in case the respondent may not be familiar with the terms used in the questions, due to potential regional differences.

The between-subjects design was chosen, since the experimenter believed it un-likely that a subject would respond differently to the same risk elicitation questions in such a short space of time. As well as this, asking the same set of questions twice may have confused some respondents, and made the research question fairly easy to guess. Should respondents guess the research question, their responses may not reflect their real-world decisions; perhaps they would behave according in a way they think the experimenter is hoping to observe.

Although a within-subjects design would have more likely captured the effect of the treatment, this may be better suited for a laboratory experiment, with more time passing between each section of the experiment. Another experimental method was considered - domain specific risk tolerance (DOSPERT), which measures the perceived benefits and risks of a decision, as well as the likelihood of undertaking that behaviour (Blais and Weber, 2006). However, this measure is less focused on the actual decision, and its behavioural economic application is unclear - since perception of risk does not always reflect real-world decisions.

(15)

Other methods of placing the adverts were also considered. Arguably, a better method for the adverts would have been to place an advertising banner on the survey. This would reflect what a subject may see in their natural environment. However, with this method, there was no guarantee that every respondent reviewed the information to the same degree as others. So, in order to best ensure that the method answers the research question, it seems important that the subject fully reviews the information. Even though people may choose to ignore adverts in real life, asking respondents to review the information can control for the fact whether they choose to ignore them or not.

4 Results/Regressions

There were 57 participants in total completing the survey, with 27 participating in the "Advert" treatment, and 30 participating in the "Education" treatment. The discrepancy in the two values is due to individuals starting the survey, and not completing it. However, 9 respondents were excluded from the results, due to in-consistent preferences arising from multiple switches between lotteries in the risk elicitation question. As a result, their risk aversion level is unable to be determined. Excluding subjects from analysis is standard practice in experiments using the Holt and Laury (2002) risk elicitation questions (see Dave et al., 2010; Harrison et al., 2005). Two exclusions were from the "Advert" treatment, and seven from the "Ed-ucation" treatment, leaving 24 in the Advert Treatment and 24 in the Education Treatment. These respondents will be excluded from all further analysis until the Other Findings subsection, where their risk aversion level is no longer relevant to the line of argument.

Two subjects were paid out according to their decisions in the risk elicitation question. The two subjects were chosen at random, using a randomiser. The ran-domiser then selected one of the ten lottery choices, at random. Then, the subject’s choice of Lottery A or B was run through the randomiser as well, according to the probabilities of that particular question.

One subject was paid 25 euros, and the other was paid 10.40 euros. This gives an average payout of 17.70 euros of the subjects that were paid, but a total of 62 cents on average for all 57 participants of the survey.

This Results section begins discussing the primary findings and whether the hypotheses can be rejected or not. Then the regression analysis for hypothesis H(1)

concerning all the control variables will be discussed. Finally, other interesting

(16)

of this section.

4.1 Primary Findings

In this subsection, each hypothesis will be considered in turn. A Shapiro-Wilk test for normal distribution was conducted on the dataset to determine whether the data was normally distributed around the mean for the variable risk aversion level. The mean risk uptake was 5.02. The data was found to be normally distributed - both across the entire dataset without treatments, and both treatment subsets of data.

The results of the Shapiro-Wilk tests can be found in the Appendix. The inter-pretation for the test is if p-value is greater than the significance level alpha=0.05, one cannot reject the null hypothesis H0. The null hypothesis in this case is that the data is normally distributed. Since all the data is normally distributed, t-tests are used to test the hypotheses of this study.

The full results of all tests in this Results section can be found in the Appendix.

4.1.1 H(1): Treatment Effects

The null hypothesis of H(1) is that the mean risk uptake is not significantly higher in the Advert Treatment than the Education Treatment. The alternative hypothesis is that the mean of risk uptake in the Advert Treatment is significantly lower than the Education Treatment.

The data showed that the mean risk uptake level in the Advert Treatment was lower (M = 5.00, SD = 2.24) than the Education Treatment (M = 5.04, SD = 2.25); one-sided t-test t(39) = −0.069, p = 0.47.

The p-value is higher than any relevant alpha, so the null hypothesis cannot be rejected.

Therefore, this t-test does not suggest any evidence in favour of the notion that adverts lead to an increase in risk uptake.

4.1.2 H(2A): Gender Effects

19 of the 48 respondents were female, with 9 females in the Advert treatment. There were 15 males in the Advert Treatment. The recorded responses for risk replicated results from previous studies, finding that on average, females had lower risk uptake [were more risk averse] than males.

The null hypothesis of H(2A) is that the mean risk uptake is not significantly higher for males than it is for females.

(17)

Figure 1: Gender and Mean Risk Uptake

In a two-sample, one-sided t-test (assuming unequal variances), the means are significantly different at the 1% level, with a one-tail P-value of 0.0094 (rounded). The mean risk uptake of females (M = 4.16, SD = 1.98) was significantly lower than for males (M = 5.59, SD = 1.96); one-sided t-test t(38) = −2.46, p = 0.0094.

This means that the null hypothesis of H(2A) is rejected, suggesting the alter-native hypothesis - that males have a significantly higher risk uptake than females. However, when considering the effect of the treatment across genders, comparing means of risk uptake appears to differ to H(1), across genders. While males appear to have risk uptake in support of H(1), females have the opposite. This data is summarised in Figure 1. Two t-tests were conducted to test these differences.

The initial results suggest that females have a higher risk uptake in

Educa-tion treatment, which seems to contradict H(1). In a two-sample, one-sided

t-test (assuming unequal variances), the mean risk uptake of females in the Advert treatment (M = 3.89, SD = 2.26) was lower than in the Education treatment (M = 4.40, SD = 1.78); one-sided t-test t(15) = −0.53, p = 0.297. Since the p-value is larger than any relevant level of significance, the null hypothesis that there is a difference between risk uptake of females across treatments cannot be rejected.

For males, the initial results suggest that there is a higher risk uptake in the Advert treatment, which seems to support H(1). In a two-sample, one-sided t-test (assuming unequal variances), risk uptake was higher in the Advert treatment (M = 5.67, SD = 2.44) than in the Education treatment (M = 5.50, SD = 1.34); one-sided t-test t(22) = 0.23, p = 0.41. Since the p-value is larger than any relevant level of significance, the null hypothesis that there is a difference between risk uptake of males across treatments cannot be rejected.

(18)

Age Group Advert Education Total

20-25 years 4.87 5.00 4.92

26-30 years 5.75 5.13 5.33

39-49 years 3.67 4.75 4.29

50+ years 6.50 5.33 5.80

Table 1: Age and Mean Risk Preference

4.1.3 H(2B): Age

The age of the respondents varied from 20 to 70 years of age. Twenty-three respon-dents were aged from 20-25, twelve were 26-30, seven were aged 39-49, and five were 50+ years of age. There were no respondents aged 31-38. Age was analysed in these groups from the raw data to compare the means.

Comparing means, the findings seem to suggest that there is no treatment effect on risk uptake. The means can be found in Table 1.

H(2B) hypothesises that older subjects would have a higher level of risk uptake than younger subjects. In order to test whether the means of the age groups were significantly different from each other, a single factor analysis of variance (ANOVA) test was conducted. The null hypothesis for this ANOVA test is that the mean risk uptake for each age group are all equal.

For each group, the mean and standard deviation were as follows: • 20-25 years, M = 4.92, SD = 2.38;

• 26-30 years, M = 5.33, SD = 1.61; • 39-49 years, M = 4.29, SD = 2.29; • 50+ years, M = 5.80, SD = 0.84.

If the F is larger than the F crit, then the null hypothesis is rejected.

In this ANOVA test, F = 0.63, F crit = 2.82. Therefore, the null hypothesis cannot be rejected, and the means of the age groups cannot be said to be different to each other. Thus, H(2B) can be rejected.

4.1.4 H(2C): Debt Literacy

The initial results did not show any clear relationship between debt literacy and risk uptake. The mean risk uptake for respondents by debt literacy level can be found in Table 2. The debt literacy level corresponds to the number of correct answers, 0 representing zero correct answers, and 3 representing three correct answers.

(19)

Debt Literacy Level Advert Education Total

0 (lowest) 6.00 4.00 5.33

1 4.71 5.25 5.00

2 5.00 4.80 4.92

3 (highest) 5.40 5.40

Table 2: Debt Literacy and Mean Risk Preference

Only five respondents managed to get three correct answers in this section, and they were all randomly selected for the Advert treatment. This means there is no data for the Education treatment for respondents with three correct answers.

An ANOVA test was conducted to test hypothesis H(2C). Once again, the null hypothesis is that the mean risk uptake of each level of debt literacy are all the same.

For each level, the mean and standard deviation were as follows: • Level: 0, M = 5.33, SD = 1.53

• Level: 1, M = 5.00, SD = 2.39 • Level: 2, M = 4.92, SD = 2.10 • Level: 3, M = 5.40, SD = 1.52

The ANOVA test found F = 0.093 and F crit = 2.82. Therefore, the null

hypothesis cannot be rejected, and the means of the debt literacy levels cannot be said to be different from each other. As a result, H(2C) can be rejected.

4.1.5 Regression Analysis

Dummies were used for many of the variables for the regression analysis. The dummy for being in the Advert treatment was named "AdvertTreatment"; the dummy for females was named "Female"; three nationality dummies were created: "Dutch", "British", and "Other Nationalities", due to the number of respondents from those nationalities; a "Current Student" dummy was created for those still in education; dummies were created for each level of education; there were dummies created for the respective bands of income; another was created for whether the respondent had financial dependants; dummies were created based on the frequency of adver-tisements the respondents reported. In the regression analysis, one of the dummies from each variable group was excluded in order to avoid the dummy variable trap (see Dougherty, 2011).

(20)

The "CurrentStudent" dummy was removed due to endogeneity with income -all current students reported an income of under 20,000 euros.

Seven separate regressions were run, adding an additional variable at a time. The regression is linear. Regression (7), with all control variables, is below.

RiskU ptake = β0 + β1Education + β2M ale + β3Dutch + β4N oDependants + β5N ever

An interpretation of regression (7) is that the reference group is male, is in the Education treatment, is Dutch, has no financial dependants, and never views adverts for consumer credit. An example of this using the table: the risk uptake is -0.11 from being in the Advert treatment than in the Education treatment.

The findings show that females have a significantly lower risk preference at the 5% level, except for regressions (4) and (5). This is most likely due to noise in the data or important variables missing.

The only other significant findings are that the frequency of viewing adverts has an effect on risk uptake. However, the coefficients are very large, suggesting a dramatic increase in risk uptake. Such a large increase is unlikely and is most caused by to outliers in the data.

The regressions show that there is insufficient evidence for the hypothesis H(1) that the treatments had an effect on risk uptake. They do suggest that the frequency of viewing adverts may have an effect on risk preference, however due to the small sample size, further research will need to be done. This is to investigate why it has such a significantly large effect on risk uptake.

(21)

Table 3: Regressions with Robust Standard Errors

(1) (2) (3) (4) (5) (6) (7) RiskPreference RiskPreference RiskPreference RiskPreference RiskPreference RiskPreference RiskPreference AdvertTreatment -0.077 -0.11 -0.11 -0.053 -0.11 -0.12 -0.11 (0.59) (0.63) (0.63) (0.62) (0.65) (0.70) (0.70) Age 0.0056 0.011 0.012 -0.022 -0.036 -0.056 -0.058 (0.019) (0.026) (0.027) (0.041) (0.038) (0.039) (0.041) Female -1.44∗ -1.50∗ -1.46∗ -1.16 -1.25 -1.86∗ -1.81∗ (0.58) (0.61) (0.62) (0.67) (0.71) (0.84) (0.85) DebtLiteracyLevel 0.0077 0.024 -0.049 -0.040 -0.089 0.0030 (0.42) (0.42) (0.40) (0.41) (0.51) (0.60) British -0.68 -0.74 -0.97 -0.88 -1.46 -1.40 (0.85) (0.89) (0.82) (0.85) (1.02) (1.06) OtherNationality 0.023 -0.087 -0.77 -0.63 -1.19 -1.34 (0.72) (0.75) (0.79) (0.88) (0.91) (0.99) EducationalScale -0.21 -0.28 -0.35 -0.56 -0.58 (0.45) (0.41) (0.43) (0.40) (0.40) IncomeBand 0.42 0.38 0.43 0.46 (0.25) (0.27) (0.23) (0.26) wDependants 0.72 1.22 1.19 (0.87) (0.96) (0.98) Daily 2.11 2.21 (1.53) (1.69) Almostdaily 1.62 1.77 (1.49) (1.61) twiceweekly 3.24∗ 3.25∗ (1.36) (1.37) onceweekly 2.41∗ 2.44∗ (1.04) (1.09) twicemonthly 3.10∗ 3.07∗ (1.17) (1.22) oncemonthly 0.70 0.75 (1.04) (1.06) lessthanmonthly 1.70∗ 1.65 (0.76) (0.84) numberofproducts -0.12 (0.24) _cons 5.46∗∗∗ 5.62∗∗∗ 6.53∗∗ 7.01∗∗ 7.66∗∗ 7.57∗∗ 7.71∗∗ (0.85) (1.05) (2.33) (2.20) (2.44) (2.28) (2.25) N 48 48 48 48 48 48 48 R2 0.118 0.145 0.149 0.217 0.225 0.376 0.381

Standard errors in parentheses

∗ _{p < 0.05,}∗∗ _{p < 0.01,}∗∗∗ _{p < 0.001}

4.2 Other Findings

This study asked respondents to report how many different consumer credit products they own, or have owned in the past.

There appeared to be a trend on how frequently adverts were viewed and number of credit products owned, on average. This correlation is interesting, and could be further explored in future research.

This trend is shown in Figure 2.

As well as this, respondents with higher levels of debt literacy owned more credit products on average. Although this finding is interesting, no conclusions can be made in terms of its relation to risk uptake.

(22)

Figure 2: Average Credit Products Owned, by Frequency of Adverts Viewed

Debt Literacy Level Mean Number of Products Owned

0 (lowest) 1.00

1 2.25

2 2.86

3 (highest) 3.4

Total 2.62

Table 4: Debt Literacy and Mean Credit Products Owned The results are found in Table 4.

4.3 Summary of Results

In summary, the results show little evidence for the hypothesis H(1), that adverts effect risk uptake. Using robust standard errors, the only notable result gathered from this survey experiment was that females had a significantly lower risk preference than males.

The finding of "frequency of adverts viewed" significantly affecting risk uptake is interesting. However, since the constants are so large, it is likely that these results are due to a small sample size, or noisy data. In order to draw conclusions on this finding, further research will be required. The variable also suffers from the fact that respondents may not be able to recall exactly how often they view adverts for credit products. As well as this, adverts may not be actively noticed, but may still be able to have a subliminal effect. Unfortunately however, respondents would be unlikely to recall the number adverts they had not actively viewed.

(23)

5 Discussion

The results found in this survey experiment clearly do not show any evidence in favour of the hypothesis H(1), that viewing advertisements results in an increase in risk uptake. There are a few possible reasons as to why this may be, and will be discussed in this section, grouped by main themes.

5.1 Experimental Design

The most crucial point necessary to discuss is whether the treatments themselves were enough to induce an effect on risk preferences, and whether the treatments were suitable.

There were two treatments in this survey experiment, an Education treatment, and an Advert treatment. The experimenter had deemed the treatments different enough, with one warning about the dangers of debt, and the other advertising debt. However, the fact that the respondent was presented with the idea of debt at all, regardless of the messaging, may have led to a change in the risk uptake. Thus, a control treatment could have been used, where nothing was shown to the respondent. This may help to determine whether both the Advert and Education treatment resulted in a change in risk uptake in the same direction. Additionally, comparing the treatments to each other may not be sufficient to conclude whether there was actually an impact on the risk uptake without a baseline, even if the results had been significant.

As well as this, it is difficult to pinpoint which particular advertisement had induced a change in risk uptake, if at all. The experimenter selected a wide variety of consumer credit products in order to reflect what a consumer may view in their day-to-day life, however, each different credit product could have resulted in moving the risk uptake in opposite directions. For further research, an experimenter could focus on one specific area of consumer credit, for example credit cards.

Another factor is whether the Advert treatment was similar enough to what an individual would see in their natural environment, and whether individuals of different nationalities would perceive them differently. This could be due to the fact that in some countries, consumer credit adverts are regulated in a certain way in that the messaging could differ to others. The effect of nationality on risk attempt was explored in this survey, however, no significant results were found in the regression. In the debt literacy questions, a few participants contacted the experimenter to ask whether they should consider time discounting in the third question. This suggests that the question was not clear enough, and may have resulted in some

(24)

mis-reporting of the debt literacy results. This is an element to consider when conducting future research, as a different method could be used to elicit debt literacy levels.

The other problem with the design is the fact that almost 15% of the participants needed to be excluded from the results, due to inconsistent preferences in the risk elicitation section. This is a rather high percentage. Perhaps the participants didn’t understand the question well enough, or were not motivated to complete the question in such a way that would have reflected their real-world behaviour. Either way, it raises the question as to whether further explanation should be provided when presenting such decision problems to the layman.

Another point to consider is that the question to determine how many consumer credit products the respondent owns is not hugely insightful. It did not account for multiples of each type of product - for example, one respondent may have owned three or four credit cards, and was compared in the same way as a respondent that owned one. Owning more credit cards may indicate more risky behaviour, but it also may not - it could be that the individual is confident with managing their money.

The structure of the question meant that no further conclusions could be made, controlling for age, gender or other relevant variables. Although it was not the purpose of this research, given the interesting initial results, it may be of the interest of researchers to explore further whether the number of credit products is indeed affected by the frequency of viewing of advertisements. As well as this, there was no question to determine whether a respondent had ever defaulted on one of their debts. Although this may not be explicitly linked to risk itself, this potentially could have been an omitted variable in the regression. This is because an individual who had previously defaulted on a loan could be affected stronger by the Education treatment, by bringing up bad memories for example.

Finally, the choice of methodology must be evaluated, based on the rejection of hypothesis H(1). Different choice architectures that reflect risky spending pat-terns more realistically could be considered when reproducing this research question. However, consumer credit products often concern intertemporal decisions, in that an individual may buy something on a credit card now in order to pay it back later. Current literature examines the effect of consumer credit products by utilising field experiments - and in the setting of consumer credit adverts, it would also likely be better suited than a survey in order to observe actual behaviour. However, it is probably almost impossible to completely alter someone’s wider environment for a sustained period of time, serving them exclusively consumer credit advertisements. As well as the above points, a within subjects design could have been used in order to isolate the effect of the treatment further, but would require more

(25)

par-ticipants, and most likely a laboratory-based design. This is due to the reasons discussed in the Methodology section.

5.2 Contextual Issues

There are several other issues to consider when evaluating this research. The first issue is whether consumer credit and financial risk preferences are indeed related. The link between the two is not really clear given the results, and indeed, there may be other reasons aside from risk as to why an individual takes out a credit product. One of these reasons may be solely to improve a credit rating, which is particularly relevant in the United Kingdom. However, the research question is more whether the priming of consumer credit is enough to influence risk uptake, not to explore the link between the two.

Although Holt and Laury’s (2002) method is well-established for determining the level of risk aversion, the link between it and actual behaviour in the context of consumer credit is ambiguous. Being more likely to take risks in a financial gamble does not necessarily mean that the individual would be more likely to make an irrational purchase.

Another suggestion may be that as online advertisements become more person-alised, some respondents may have already viewed more adverts than others, had they searched for a consumer credit product online. Although unlikely, it could have affected the results in some way. Either way, this raises an interesting idea of how the experiment could have been conducted.

In summary, although raising interesting points with an experiment fairly con-sistent with experimental economics literature, it seems using advertisements is very difficult to influence risk uptake in the context of this particular design. However, using a different design, perhaps a better reflection of behaviour could have been made.

5.3 Further Research

To further research this question, a field experiment could be carried out with a num-ber of different cues. A field experiment is likely to be a better reflection of human behaviour, and adverts could be placed more naturally in the subject’s environment. Another method could be to randomise the order that the subject receives the treatment and questions. Perhaps there were order effects in this experiment, and changing the order of the questions may have resulted in different results. Although, it makes sense to put the research variables in question at the beginning while the

(26)

subject’s attention is greatest.

Another method of measuring the impact of adverts on risk could be to use domain specific risk-tolerance (DOSPERT) measures. Subjects self-report their tol-erance to risk by judging the benefits and risk of each situation. Even though the risk perception may not necessarily convey to actual behaviour, DOSPERT has been described as being suitable for finding a "general risk-taking disposition", and that this appetite for risk can indeed predict real-world outcomes (Highhouse et al., 2017). This method may have captured a more accurate risk profile of the subject. This is because subjects may be more easily influenced in the perception of risk, rather than the likelihood of doing that behaviour.

6 Conclusion

This thesis provides little evidence for the hypothesis H(1), that advertisements on consumer credit can influence an individual’s risk uptake. The lack of any signif-icant results for this hypothesis not only suggests that risk uptake is perhaps not very easily influenced, but that there is little evidence in favour of this hypothesis. However, this thesis does provide some evidence that the viewing consumer credit advertisements more frequently could have an effect on risk uptake. This particular result this will need to be further studied to check its robustness.

Although the experimental design has some flaws, this is original, early literature exploring how exactly risk uptake can be influenced by external priming, and still contributes to the literature despite the lack of any solid conclusions. It also adds to the literature that males are less risk averse than females, given that H(2A) cannot be rejected. As well as this, it demonstrates a broader context to explore the effect of cognitive biases on different elements of behaviour.

Further research in this area will give a greater understanding about risk aversion, exactly what influences it, and whether marketing messages need to be more tightly regulated for "at risk" consumers.

References

Allais, M. (1990). Allais paradox. In Utility and probability, pages 3–9. Springer. 25

Arrow, K. J. (1965). Aspects of the theory of risk-bearing. Yrjö Jahnssonin Säätiö.

(27)

Barberis, N. C. (2013). Thirty years of prospect theory in economics: A review and assessment. Journal of Economic Perspectives, 27(1):173– 96.

Bernoulli, D. (1954). Exposition of a new theory on the measurement of risk. Econometrica, 22(1):23–36.

Blais, A.-R. and Weber, E. U. (2006). A domain-specific risk-taking (dospert) scale for adult populations.

Brady, R. R. (2008). Structural breaks and consumer credit: Is consumption smoothing finally a reality? Journal of Macroeconomics, 30(3):1246 – 1268.

Campbell, W. K., Goodie, A. S., and Foster, J. D. (2004). Narcissism, confidence, and risk attitude. Journal of behavioral decision making, 17(4):297–311.

Chang, C.-C., Devaney, S., and Anong, S. (2004). Determinants of subjective and objective risk tolerance. 3.

Chatterjee, P. and Rose, R. L. (2011). Do payment mechanisms change the way consumers perceive products? Journal of Consumer Research, 38(6):1129–1139.

Chuah, S.-H. and Devlin, J. (2011). Behavioural economics and financial services marketing: a review. International Journal of Bank Marketing, 29(6):456–469.

Clark, G. L. and Strauss, K. (2008). Individual pension related risk propensities: the effects of socio-demographic characteristics and a spousal pension entitlement on risk attitudes. Ageing & Society, 28(6):847–874.

Cox, R., Brounen, D., and Neuteboom, P. (2015). Financial literacy, risk aversion and choice of mortgage type by households. The Journal of Real Estate Finance and Economics, 50(1):74–112.

Dave, C., Eckel, C. C., Johnson, C. A., and Rojas, C. (2010). Eliciting risk preferences: When is simple better? Journal of Risk and Uncertainty, 41(3):219–243. 26

Davies, E. and Lea, S. E. (1995). Student attitudes to student debt. Journal of economic psychology, 16(4):663–679.

Disney, R. and Gathergood, J. (2013). Financial literacy and consumer credit portfolios. Journal of Banking & Finance, 37(7):2246– 2254.

Dougherty, C. (2011). Introduction to econometrics. Oxford University Press. Erb, H.-P., Bioy, A., and Hilton, D. J. (2002). Choice preferences without in-ferences: Subconscious priming of risk attitudes. Journal of Behavioral Decision Making, 15(3):251–262.

Falahati, L. and Sabri, M. F. (2015). An exploratory study of personal financial wellbeing determinants: examining the moderating effect of gender. Asian Social Science, 11(4):33.

(28)

condi-tioning interpretation. Journal of consumer research, 13(3):348–356.

Friedman, M. (1957). The permanent income hypothesis. In A theory of the consumption function, pages 20–37. Princeton University Press.

Gasiorowska, A., Zaleskiewicz, T., and Wygrab, S. (2012). Would you do some-thing for me? the effects of money activation on social preferences and social be-havior in young children. Journal of Economic Psychology, 33(3):603–608.

Gathergood, J. and Weber, J. (2017). Financial literacy, present bias and alter-native mortgage products. Journal of Banking & Finance, 78:58–83.

Gerhardt, M. (2009). Consumer bankruptcy regimes and credit default in the us and europe: A comparative study. ceps working document no. 318, 27 july 2009.

Gilliam, J., Chatterjee, S., and Zhu, D. (2010). Determinants of risk tolerance in the baby boomer cohort. Journal of Business & Economics Research, 8(5). 27

Grable, J. E. (2000). Financial risk tolerance and additional factors that af-fect risk taking in everyday money matters. Journal of Business and Psychology, 14(4):625–630.

Guiso, L. and Jappelli, T. (2008). Financial literacy and portfolio diversification. Hallahan, T. A., Faff, R. W., and McKenzie, M. D. (2004). An empirical in-vestigation of personal financial risk tolerance. FINANCIAL SERVICES REVIEW-GREENWICH-, 13(1):57–78.

Hansen, J., Kutzner, F., and Wänke, M. (2012). Money and thinking: Reminders of money trigger abstract construal and shape consumer judgments. Journal of Consumer Research, 39(6):1154–1166.

Hardisty, D. J., Appelt, K. C., and Weber, E. U. (2013). Good or bad, we want it now: Fixed-cost present bias for gains and losses explains magnitude asymmetries in intertemporal choice. Journal of Behavioral Decision Making, 26(4):348–361.

Harrison, G. W., Johnson, E., McInnes, M. M., and Rutström, E. E. (2005). Risk aversion and incentive effects: Comment. American Economic Review, 95(3):897–901.

Heidhues, P. and Kőszegi, B. (2010). Exploiting naïvete about self-control in the credit market. American Economic Review, 100(5):2279–2303.

Highhouse, S., Nye, C. D., Zhang, D. C., and Rada, T. B. (2017). Structure of the dospert: Is there evidence for a general risk factor? Journal of Behavioral Decision Making, 30(2):400–406.

Holt, C. A. and Laury, S. K. (2002). Risk aversion and incentive effects. Amer-ican economic review, 92(5):1644–1655.

Ikeda, S. and Kang, M.-I. (2015). Hyperbolic discounting, borrowing aversion and debt holding. The Japanese Economic Review, 66(4):421– 446.

(29)

under risk. Econometrica, 47(2):263–292. 28

Kühberger, A., Schulte-Mecklenbeck, M., and Perner, J. (1999). The effects of framing, reflection, probability, and payoff on risk preference in choice tasks. Organizational behavior and human decision processes, 78(3):204– 231.

Limbu, Y. B. (2017). Credit card knowledge, social motivation, and credit card misuse among college students: Examining the informationmotivation-behavioral skills model. International Journal of Bank Marketing, 35(5):842–856.

Loewenstein, G. F., Weber, E. U., Hsee, C. K., and Welch, N. (2001). Risk as feelings. Psychological bulletin, 127(2):267.

Ludvigson, S. (1999). Consumption and credit: A model of time varying liquidity constraints. The Review of Economics and Statistics, 81(3):434– 447.

Lusardi, A. and Tufano, P. (2015). Debt literacy, financial experiences, and overindebtedness. Journal of Pension Economics & Finance, 14(4):332–368.

McCoy, P. A. (2005). A behavioral analysis of predatory lending. Akron Law Review, 38(4).

Meier, S. and Sprenger, C. (2010). Present-biased preferences and credit card borrowing. American Economic Journal: Applied Economics, 2(1):193–210.

Modigliani, F. and Brumberg, R. (1954). Utility analysis and the consumption function: An interpretation of cross-section data. Franco Modigliani, 1:388–436.

Morduch, J. (1995). Income smoothing and consumption smoothing. Journal of Economic Perspectives, 9(3):103–114.

Newall, P. W. (2015). How bookies play with your emotions to make you place unlikely bets. The Conversation.

Potrich, A. C. G. and Vieira, K. M. (2018). Demystifying financial literacy: a behavioral perspective analysis. Management Research Review. 29

PRATT, J. W. (1964). Risk aversion in the small and in the large’. Economet-rica, 32(1-2).

Ross, S. A. (1981). Some stronger measures of risk aversion in the small and the large with applications. Econometrica: Journal of the Econometric Society, pages 621–638.

Sjöberg, L. and Engelberg, E. (2009). Attitudes to economic risk taking, sen-sation seeking and values of business students specializing in finance. Journal of Behavioral Finance, 10(1):33–43.

Strömbäck, C., Lind, T., Skagerlund, K., Västfjäll, D., and Tinghög, G. (2017). Does self-control predict financial behavior and financial well-being? Journal of Behavioral and Experimental Finance, 14:30 – 38.

(30)

-The Economic Times. [online] Available at: https://economictimes.indiatimes.com/definition/risk [Accessed 10 Aug. 2018].

Trampe, D., A. Stapel, D., and W. Siero, F. (2010). The selfactivation effect of advertisements: Ads can affect whether and how consumers think about the self. Journal of Consumer Research, 37(6):1030–1045.

Trepel, C., Fox, C. R., and Poldrack, R. A. (2005). Prospect theory on the brain? toward a cognitive neuroscience of decision under risk. Cognitive brain research, 23(1):34–50.

Tversky, A. and Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and uncertainty, 5(4):297–323.

Von Neumann, J., Morgenstern, O., et al. (1944). Theory of games and economic behavior.

Weber, E. U. and Hsee, C. (1998). Cross-cultural differences in risk perception, but cross-cultural similarities in attitudes towards perceived risk. Management sci-ence, 44(9):1205–1217.

Yang, S., Markoczy, L., and Qi, M. (2007). Unrealistic optimism in consumer credit card adoption. Journal of Economic Psychology, 28(2):170– 185.

7 Appendix

This appendix includes the survey experiment used in this research, as well as the results of the Shapiro-Wilk tests for normal distribution, and t-tests at the end. The Appendix begins on the next page.

(31)

Introduction Thank you for opting to participate in this survey experiment. This survey should take you around 1015 minutes. Please ensure you can complete this survey in one s truthfully as possible. In order to encourage effortful participation in this survey, two participants will be chosen at random to be paid. This survey will consist of 4 sections. There are specific instructions at the beginning of the each section. Payment Two participants will be chosen at random for payout in a lottery. You can earn up to €25 (or equivalent in your currency), depending on the choices you make. The rel survey. In order to be eligible to win, you will be asked to leave your email address so you can be contacted. You can choose not to enter your email, but if you do not, there is be entered into the payout lottery. Advert Treatment Section 1 In this section, imagine you are advising a credit lender on their advertising. Review the adverts carefully below, and then choose the advert which you think is the best one at the end.

(32)

(33)

Advert 1 Advert 2 Advert 3 Advert 4 Advert 5 Which advert do you think is the best? Education Treatment

(34)

2 years or less Between 2 and 5 years Between 5 and 10 years 10 years or more Never, you will continue to be in debt Do not know In this section, imagine you are advising the government on what messages to include in their new education schemes on debt. Review the sentiments carefully below, and then rank the statements in the order which you think is the best. Drag and drop the statements to rank them. Holt & Laury Question Section 2 In this section, you will need to choose from a lottery of choices. In the case that you are drawn for payment, one of the decisions you make below will be chosen at ra Then, the Decision that is chosen will be played, and you will be paid out according to what the randomiser shows. Example: You are selected for payment. Decision 6 is chosen (at random) for payout. You chose Option B. There is a 60% chance you will be paid €25, and a 40% chance you will be paid €0.65. Read the options on the left and then choose your option in the right. Please note, on mobile devices, you may have to scroll left or right to select the option. Option A Option B OPTION A: 10% chance of €13, 90% chance of €10.40. OR OPTION B: 10% chance of €25, 90% chance of €0.65 OPTION A: 20% chance of €13, 80% chance of €10.40. OR OPTION B: 20% chance of €25, 80% chance of €0.65 OPTION A: 30% chance of €13, 70% chance of €10.40. OR OPTION B: 30% chance of €25, 70% chance of €0.65 OPTION A: 40% chance of €13, 60% chance of €10.40. OR OPTION B: 40% chance of €25, 60% chance of €0.65 OPTION A: 50% chance of €13, 50% chance of €10.40. OR OPTION B: 50% chance of €25, 50% chance of €0.65 OPTION A: 60% chance of €13, 40% chance of €10.40. OR OPTION B: 60% chance of €25, 40% chance of €0.65 OPTION A: 70% chance of €13, 30% chance of €10.40. OR OPTION B: 70% chance of €25, 30% chance of €0.65 OPTION A: 80% chance of €13, 20% chance of €10.40. OR OPTION B: 80% chance of €25, 20% chance of €0.65 OPTION A: 90% chance of €13, 10% chance of €10.40. OR OPTION B: 90% chance of €25, 10% chance of €0.65 OPTION A: 100% chance of €13, 0% chance of €10.40. OR OPTION B: 100% chance of €25, 0% chance of €0.65 Debt Literacy Section Section 3 You may use a calculator, and a pen and paper to answer these questions if you wish. Question 1 Suppose you owe €1,000 on your credit card and the interest rate you are charged is 20% per year compounded annually. If you didn’t pay anything off, at this interest rate, how many years would it take for the amount you owe to double (grow to €2,000)? Debt hurts your credit score. Debt can negatively affect your marriage and your family. Debt can lead to stress, anxiety, and even depression. Debt can prevent you from obtaining a mortgage, and owning your own home. Debt keeps you from meeting your financial goals. Debt costs money, and high interest rates mean you can pay back more for things than they actually cost. Debt is borrowing from your future income. Debt encourages you to spend more than you can afford.

(35)

Less than 5 years Between 5 and 10 years Between 10 and 15 years Never, you will continue to be in debt Do not know Option (a) Option (b) They are the same Do not know Male Female Other Yes No Did not complete High School/Secondary School High School/Secondary School qualifications College/Sixth Form qualifications Bachelor's degree or equivalent Master's degree or equivalent Doctoral degree or higher, or equivalent Employed fulltime Employed parttime Selfemployed Unemployed You pay a minimum payment of €30 each month. At an Annual Percentage Rate of 12% (or 1% per month), how many years would it take to eliminate your credit card charges? Question 3 Suppose you purchase an appliance which costs €1,000. To pay for this appliance, you are given the following two options: a) Pay 12 monthly instalments of €100 each b) Borrow at a 20% annual interest rate and pay back €1200 one year from now. Which is the more advantageous offer, in other words which one will cost you less? Demographic/Other Questions Section 4 This section will ask some demographic questions about you. Please answer as truthfully as possible. How old are you? What is your gender? What is your nationality? In which country do you live? Are you currently in fulltime education? What is your highest level of education? If you are still in education, please choose your current level. What is your employment status? In which sector do you work?

(36)

Less than €20,000 €20,000 to €34,999 €35,000 to €49,999 €50,000 to €74,999 €75,000 to €99,999 Over €100,000 Yes No Phone loan (including monthly phone contracts) Car finance (regular instalments) Other instalment credit (such as furniture/large appliances/etc) Personal credit card Store credit cards Overdraft on a current account Student loan Payday loan Other type of shortterm loan (e.g. not including mortgages or any other type of real estate) Never Less than once a month Around once a month Twice a month Once a week Two or three times a week Almost daily Daily What is your annual income? If you earn a salary in a different currency, please give your best estimate of it in €. Do you have any financial dependants? (Children, spouses, relatives, etc who depend on you for money) Which consumer credit products do you currently have, or have had in the past? Please check all the boxes which apply to you. How often do you see adverts for consumer credit products? (credit cards, car finance, mobile phone contracts, furniture finance, payday loans, etc) Please give your best estimate. What question do you think this survey is trying to answer? Please leave your email address below if you would like to be entered into the payout lottery to earn up to €25. The only purpose it will be used for is to contact the winners of the lottery.

(37)

Figure 3: Two-Sample t-test: Treatment

Figure 4: Two-Sample t-test: Gender

Figure 5: Two-Sample t-test: Female, Treatment

W 0.954

p-value (Two-tailed) 0.058

alpha 0.05

(38)

Figure 6: Two-Sample t-test: Male, Treatment

W 0.953

alpha 0.05

Table 6: Shapiro-Wilk Test for Advert Treatment Subset

W 0.947

alpha 0.05

Consumer credit adverts - risky business?