undergraduate dissertation - does intelligence affect susceptibility to anchoring?

39
School of Economics University of Nottingham L13500 Dissertation 2014-15 Does Intelligence Affect Susceptibility to Anchoring? Louis Adams Student ID: 4173657 Supervisor: Fabio Tufano Word Count: 7,492 This Dissertation is presented in part fulfilment of the requirement for the completion of an undergraduate degree in the School of Economics, University of Nottingham. The work is the sole responsibility of the candidate. I give permission for my dissertation proposal to be made available to students in future years if selected as an example of good practice

Upload: louis-adams

Post on 16-Jan-2017

89 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

School of Economics

University of Nottingham

L13500 Dissertation 2014-15

Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams

Student ID: 4173657

Supervisor: Fabio Tufano

Word Count: 7,492

This Dissertation is presented in part fulfilment of the requirement for the completion of an undergraduate degree in the School of

Economics, University of Nottingham. The work is the sole responsibility of the candidate.

I give permission for my dissertation proposal to be made available to students in future years if selected as an example of good

practice

Page 2: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 1

Abstract

The anchoring effect is now an accepted phenomenon in behavioural economics that leads to bias in decision-

making. The literature has progressed to examine the factors that affect susceptibility to anchoring, with a

small number of papers coming to contrasting conclusions on whether intelligence is one such factor. This

paper describes an online experiment conducted to explore whether people of higher cognitive ability are less

influenced by an arbitrary anchor number when making price judgement decisions of standard consumer

products.  Firstly,  it  finds  that  subjects’  stated  willingness to pay and willingness to accept decisions are affected

by the uninformative anchor. Secondly, it does not find evidence to suggest that cognitive ability is related to

susceptibility to anchoring.

Contents

1. Introduction .......................................................................................................................................................... 2

1.1 Research motivation .................................................................................................................................... 2

1.2 Research question ......................................................................................................................................... 2

1.3 Hypotheses ....................................................................................................................................................... 3

2. Literature review................................................................................................................................................. 3

2.1 Anchoring ........................................................................................................................................................ 3

2.2 Price judgement and anchoring ............................................................................................................... 5

2.3 Cognitive ability ............................................................................................................................................. 6

2.4 Testing of cognitive ability ......................................................................................................................... 6

3. Experimental design .......................................................................................................................................... 7

3.1 Valuation tasks .............................................................................................................................................. 7

3.2 Cognitive ability test .................................................................................................................................... 9

3.3 Demographic questionnaire ....................................................................................................................10

3.4 Subject pool ...................................................................................................................................................10

4. Results and discussion ....................................................................................................................................10

4.1 Cognitive ability ...........................................................................................................................................10

4.2 Hypothesis 1 ..................................................................................................................................................11

4.3 Hypothesis 2 ..................................................................................................................................................14

4.4 Discussion ......................................................................................................................................................17

5. Evaluation ............................................................................................................................................................18

5.1 Experimental procedure ...........................................................................................................................18

5.2 Further research ..........................................................................................................................................20

5.3 Conclusion ......................................................................................................................................................21

6. References ............................................................................................................................................................22

Page 3: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 2

1. Introduction

1.1 Research motivation

Since  Tversky  and  Kahneman’s (1974) pioneering research on judgement under uncertainty,

the anchoring effect, the phenomenon whereby exposure to an irrelevant number influences an

individual’s  subsequent  judgement  of  a  quantity  or  valuation,  has  become  an  accepted  psychological  

tendency. Ariely et al. (2003) (henceforth ALP) were the first to investigate the anchoring effect in

the economic context of valuing standard consumer products, such as a bottle of wine, by exposing

subjects to an arbitrary price prior to asking them to state the most they would pay for the good.

They found strong evidence of its existence, which has important economic implications as it

suggests that people do not always reveal, or even know, their true preferences, and that their price

judgements are not based on fundamental underlying value. This in turn draws into question the

neoclassical theory that individuals are rational economic decision makers.

More   recently,   research   has   progressed   to   study   the   factors   that   affect   an   individual’s  

susceptibility to anchoring; in other words, whether certain types of people are more influenced than

others. An area of the literature that is attracting new attention is the investigation of whether

intelligence plays a part. In 2010, Bergman et al. (henceforth BEJS) replicated  ALP’s  first experiment,

and  extended  it  to  find  that  subjects  with  higher  cognitive  ability  (CA),  a  measurement  of  “analytical

intelligence”   (Furnham, 2011, p.6), were less susceptible to anchoring. Oechssler et al. (2009) and

Stanovich and West (2008), however, do not find a significant link between the two. The very few

papers investigating this relationship and the discord among them suggest that further research into

the area would be worthwhile.

1.2 Research question

This paper has two aims. Firstly, it looks to replicate the findings of BEJS and ALP that

anchoring impacts upon an   individual’s   price   judgements; secondly, it aims to reaffirm the

robustness of   BEJS’s finding that there exists a relationship between CA and anchoring. They,

however, only consider this in the context of buying goods (willingness to pay) so I will extend the

investigation to check whether their results also hold in a selling (willingness to accept) scenario.

The potential finding that people with higher CA are less susceptible to anchoring is

important in an economic context because it implies that such people are more resistant to the

irrelevant anchor and therefore behave more rationally, displaying a lower disparity between their

revealed and true preferences.

Page 4: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 3

1.3 Hypotheses

As will be explained in Section 3, I conducted an online experiment with a between-subject

design in which each participant was put in either a buying or selling situation. They were then

exposed to a high (£17) or low (£3) anchor before giving price judgements of four products.

Following this, all subjects took a CA test. The following hypotheses are proposed in order to address

the aforementioned research questions;

Hypothesis 1: Within both the willingness to pay (WTP) and willingness to accept (WTA)

treatments, the group of subjects exposed to the high anchor will, on average, value the products

higher than the low anchor group. I therefore test the null hypothesis that the high and low anchor

groups’  mean  valuations  are  not  significantly  different.

Hypothesis 2: Drawing from BEJS’s findings, subjects who score higher in the CA test will be

less influenced by the anchor. I test the null hypothesis that the difference in anchoring effect

between subjects of high and low CA is not statistically significant.

2. Literature review

2.1 Anchoring

Tversky and Kahneman (1974) found that an irrelevant anchor between 1 and 99, randomly

generated  from  a  wheel  of  fortune,  had  a  significant  impact  on  people’s  subsequent  estimates  of  the  

percentage of African countries in the United Nations. For example, average estimates of 25% and

45% were given by those who landed on 10 and 65 on the wheel, respectively. To check the

robustness of their findings, however, the behavioural bias has since been investigated in a variety of

alternative contexts. Plous (1989) found evidence of the anchoring effect when asking students to

indicate their probability estimates of the outbreak of a nuclear war, and Carlson (1990) found it to

be robust in the context of selling three-outcome gambles.

Epley and Gilovich (2001) investigated whether people were influenced by an irrelevant

anchor when giving responses to general knowledge questions such as “When   was   George

Washington   elected   president?” but they extended their study to consider two different types of

anchor simultaneously: those generated by the subject (self-generated) and those given by the

experimenter (externally-provided). They found that anchoring plays a part in both cases.

Over time the anchoring effect has thus shown itself to exist in a range of contexts and is

described by Furnham and Boo (2011) as  “one  of  the  most  robust  cognitive  heuristics”  (2011, p. 35),

a heuristic being a cognitive rule of thumb. However, it is perhaps unsurprising that individuals are

influenced by an anchor when asked to make a judgement decision in an area in which they have no

expertise or experience. This is what Northcraft and Neale (1987) investigated in an innovative field

Page 5: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 4

experiment in which they asked both students (amateurs) and real estate agents (professionals) to

value properties, manipulating the listing price as the anchor. The finding that both groups were

significantly affected shows, firstly, that even people with expertise are susceptible to anchoring, and,

secondly, that the anchoring bias is  not  simply  a  “function  of  the  contrived  nature  of  the  laboratory  

setting” (1987, p.84), and therefore has external validity in the real world.

As discussed in Section 1, one of the first and most instrumental studies to consider

anchoring in the context of behavioural economics was that of ALP. They found strong evidence of

the effect across a range of WTP and WTA valuation tasks of both familiar consumer goods and

hedonic experiences. The results of their first experiment, which involved valuing consumer goods,

showed that people are susceptible to the effect even when they have experience of a product. Their

findings underline the economic relevance of anchoring, showing that arbitrary values can affect

individuals’  decision-making and preferences.

There are two primary psychological explanations as to why the anchoring heuristic exists.

The first, proposed by Tversky and Kahneman (1974), is the anchoring and adjustment hypothesis,

which explains that individuals initially pin their estimate to the irrelevant anchor and then fail to

sufficiently adjust away from it towards their true judgement. However, it is doubtful that this

hypothesis can be applied to every case of anchoring as it assumes that people actively use the

anchor as a reference point, which would be illogical in cases when the anchor is clearly

uninformative.

In response to this shortcoming, a second mechanism   of   “selective   accessibility”   was   put  

forward by Strack and Mussweiler (1997; Mussweiler and Strack, 1999), suggesting that people first

consider whether the anchor is plausible and then look for ways in which their own judgement is

similar, thus naturally resulting in bias towards it. The implication of this is that if the anchor is

implausible or irrelevant, it will not significantly affect the subsequent judgement, which is

supported by Sugden et al. (2013) in their study into different types of anchor.

The literature is divided in its support for each model, but Epley and Gilovich (2001, 2005)

conducted a series of experiments to show that the psychological mechanism employed is not the

same in every case, and can in fact depend on the type of anchor involved. Their papers argue, and

their results imply, that when an anchor is self-generated, there is no reason for an individual to

believe it could be perfectly correct, as they have derived it themselves from a typically irrelevant

source, so they employ the anchoring and adjustment process. On the other hand, when the anchor is

provided by the experimenter, it could feasibly be informative because the subject does not know of

its true origins, and so the process of selective accessibility takes places. Ultimately there is no single

mechanism that fully accounts for the anchoring phenomenon and so further research is required to

develop an all-encompassing model.

Page 6: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 5

More recently, the literature has moved on to consider the factors that affect susceptibility to

anchoring, which usually fall into one of two brackets: the type of anchor, and individual human

traits. The type of anchor refers to whether it is self-generated or externally-provided and whether it

is feasible or not; the relevance of each has already been discussed.

Regarding individual differences, it has been found that different types of people are

influenced by anchoring to varying degrees. In finding that introverts with high agreeableness are

more susceptible to anchoring, Eroglu and Croxton (2010) showed that personality plays a part,

while McElroy and Dowd (2007) concluded that the same could be said for people who are open to

new experiences. This area of the literature is developing and this paper aims to add to it by

considering the relationship between CA and anchoring. The small number of studies that have

already explored the link will be discussed in Section 2.3, but first the literature on anchoring in a

price judgement scenario will be covered.

2.2 Price judgement and anchoring

Asking individuals to state their maximum WTP or minimum WTA are two different ways of

establishing their valuation of a good. As Simonson and Drolet (2004) posit, the study of WTA is

becoming more important with the growth of online peer-to-peer market places such as Ebay.com,

because more consumers are becoming sellers as well.

Contrary to standard economic theory, behavioural economics  has  found  that  “willingness  to  

accept   is   usually   substantially   higher   than  willingness   to   pay”   (Horowitz and McConnell, 2002, p.

426). The primary explanation for this, supported by Thaler (1980) and Kahneman et al. (1990), is

the “endowment  effect”,  the  principle  that  a  good  is  more  valuable  to  somebody  when  they already

own it, due primarily to loss aversion.

In the literature studying the effects of anchoring on WTP and WTA judgements there is

inconsistency in results. Finding that anchoring has a greater effect on WTA than WTP, Sugden et al.

(2013) offer a convincing argument as to why this is the case. They explain that experiment

participants are more accustomed to buying than selling; people have less experience in accepting

payments for goods and are thus more influenced by the irrelevant anchor when making WTA selling

decisions.

Interestingly, Simonson and Drolet (2004) found the opposite to be true, concluding that

anchoring had a greater effect on WTP. In their WTA treatment, however, they asked participants to

imagine they had received the item as a gift and wanted to sell it. This introduces a potential framing

effect, whereby subjects are less likely to experience the endowment effect with a good that they

have neither bought themselves nor used, thus possibly explaining why the results are unusual.

Page 7: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 6

2.3 Cognitive ability

The effect of CA on economic preferences and behavioural biases has been explored in a

number of recent studies, and it   is   often   found   that  more   intelligent  people  behave   “more   like   the  

textbook  model  of  ‘economic  man’”  (Bergman  et  al.,  2010). Benjamin et al. (2006) and Dohmen et al.

(2007) looked into the effect of CA on small-scale risk aversion and impatience, respectively, finding

that  higher  CA  correlated  with  lower  susceptibility  to  these  “anomalous  preferences”  (Benjamin  et  al.,  

2006, p. 35). While the criticism could be made that the experiment in Benjamin et al. (2006) only

tested students, a very particular demographic, similar results were found by Dohmen et al. (2007),

who tested around 1000 German adults, implying that the homogeneous subject pool did not

influence results. Furthermore, Frederick (2005) had similar findings when running a lottery

experiment in hypothetical fashion.

Literature investigating whether CA is related to the anchoring bias, however, continues to be

underdeveloped and inconclusive. There is stark contrast between the results of BEJS on the one

hand, and of Oechssler et al. (2009), Stanovich and West (2008), and Furnham et al. (2012) on the

other. Initially, BEJS followed a very similar experimental design to the first ALP experiment, asking

participants for their maximum WTP for some familiar consumer goods. Subjects then completed a

CA test, with the results showing that those who scored higher on it were less susceptible to

anchoring. BEJS were simply testing for the existence of the relationship and no attempt is made to

explain why it is observed.

The other three papers, however, found no significant correlation between CA and anchoring

but this could be due to the way they measured CA. Whilst BEJS incorporated a professionally

developed 44-question intelligence test, Oechssler et al. (2009) used only a short three-question

cognitive reflection test, and Stanovich and West (2008) used self-reported SAT results, which

provide a much less reliable measure. Aside from this, it is worth noting that BEJS is the only one of

these papers to have considered anchoring in a price judgement scenario; the other three focused on

arbitrary topics such as the height of a redwood tree (Stanovich and West, 2008) and the population

of Ukraine (Furnham et al., 2012). This could explain the disparity in their results but further

research would be required to test this.

2.4 Testing of cognitive ability

In arguing that further research is required into the effects of CA on decision-making,

Frederick (2005) developed a simple three-question cognitive reflection (CR) test. These questions

are designed so that the immediately obvious answer is wrong, thus ensuring that a correct answer

requires   the  overcoming  of   the  spontaneous  “System  1  process”   (Frederick,  2005,  p. 27). Thus, the

test distinguishes between people who tend to make spontaneous decisions, and those who are more

reflective.

Page 8: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 7

There are two considerable drawbacks   to   Frederick’s   test,   however.   The   first is that CR

represents just one aspect of CA, and BEJS found it to have little effect on anchoring. The second is

that with only three questions, it is difficult to differentiate effectively between participants of

varying ability. To overcome these issues, Sousa (2010) developed a 12-question test, split into four

sections on sequential, verbal, and quantitative reasoning, as well as Frederick’s  CR. While perhaps

not as effective as a full intelligence quotient (IQ) test, it does provide an improved general measure

of  CA  compared  with  Frederick’s  test.

One further way of measuring CA is with self-reported SAT scores, as used by Stanovich and

West (2008). This is convenient for both the experimenter and the subject, however, as noted in

Kuncel et al. (2005), self-reported  grades  are  “less  construct  valid  than  many  scholars  believe”  (2005,

p. 63)  and  “unlikely  to  represent  accurately  the  actual  scores  of  students  with…low  ability”  (2005, p.

74).

3. Experimental design

I conducted an incentivised experiment to test the aforementioned hypotheses using the

online survey platform Survey Gizmo, although the format was similar to that of a laboratory

experiment. It was made up of three sections: valuation tasks, a CA test, and a demographic

questionnaire.

3.1 Valuation tasks

As with BEJS and ALP, the valuation tasks were used to test for anchoring and were of

between-subject design. Each participant firstly was placed at random into either the WTP or WTA

treatment, and then into a high anchor treatment or low anchor treatment. Subjects therefore took

part in any one of four possible treatments that shall be called WTP High, WTP Low, WTA High, and

WTA Low. Those in the High treatments were exposed to a £17 anchor figure, and those in the Low

treatments were exposed to a £3 anchor. Participants then valued four standard consumer products:

a bottle of wine, a box of Belgian chocolates, a recipe book, and an 8GB USB memory stick.

Before the valuation tasks commenced, instructions were given, and questions were asked to

check participants’  understanding of the task and payoff procedure. The instructions for the WTP and

WTA treatments were framed in a similar style, but different in terms of key points. Those in the WTP

procedure were first asked  to  state  ‘yes’  or  ‘no’  on whether they would pay a certain amount for the

product. The fact that this amount, either £3 or £17, was being used as an anchor was not explicitly

stated, so as not to inform the subject of the nature of the test. Subjects were then asked to give their

maximum WTP, which, according to neoclassical theory, should have represented the point at which

Page 9: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 8

they were indifferent between buying and not buying. On the other hand, subjects in the WTA

procedure were asked to imagine that they owned the product, and then state the minimum amount

for which they would be willing to sell (the WTA decision). Figure 1 shows the decisions faced by a

WTA High subject regarding the bottle of wine.

Figure 1 Bottle of wine, WTA High treatment

In line with BEJS, to incentivise the experiment, one participant was selected at random for

each product. A coin was flipped to determine whether their yes/no decision or their valuation

decision would be taken into account. In the WTA treatment, if the yes/no decision was selected then

it  was  implemented  and  the  subject  either  kept  the  product  (if  their  answer  was  ‘no’)  or  sold  it  to  me  

for the anchor amount (if their answer was   ‘yes’).  If  the  minimum  WTA  decision  was  implemented,  

the Becker-DeGroot-Marschak procedure (Becker et al., 1964) was used and an integer price between

£0 and £20 was drawn at random; if  it  was  higher  than  the  subject’s  minimum  WTA  then  they  sold  

the product to me for that price, and if it was lower, no transaction took place. This process was

similar for the WTP decision, except that when a transaction should have taken place, instead of

forcing the participant to buy the product, it was simply offered to them at the price, in order to avoid

unwanted real losses. The purpose of this incentive structure was to encourage participants to

behave truthfully.

Page 10: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 9

The use of an externally generated anchor is worthy of discussion. This study did not

replicate  BEJS’s use of a self-generated  anchor   (derived   from   the   subject’s   social   security  number)  

partly due to technical restraints but equally to ascertain whether their results held when using a

different type of anchor. As both types have been found to induce the anchoring effect, this was not

expected to significantly impact the results. Reasonable anchor prices of £3 and £17 were used in

light  of  Sugden  et  al.’s  (2013)  findings  that  anchors  are  more  effective  when  considered  to  be  feasible  

values.

The products displayed in this experiment were similar to those in BEJS and ALP: standard

consumer goods with which almost all participants would have had some experience. BEJS and ALP

asked subjects to consider six products but reducing this to four was sufficient for my analysis; ALP

were also investigating relative valuations, hence their use of two bottles of wine of differing quality

and two different types of chocolates. To control for any framing effect and thus allow for comparison

across treatments, every participant saw exactly the same products, with the same neutral

description.

3.2 Cognitive ability test

The valuation tasks were followed by the CA test developed by Sousa (2010), which was

made up of 12 questions testing four different categories   of   CA:   Frederick’s (2005) cognitive

reflection, quantitative reasoning, verbal reasoning, and sequential reasoning. This test gave a better

general  measure  of  CA  than  Frederick’s  three-question test, and the 12 questions allowed for a wide

enough range of scores to differentiate participants effectively, which was crucial for the statistical

analysis.

Figure 2 Example cognitive ability question

Page 11: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 10

Subjects were restricted to 12 minutes to complete the test, increasing the likelihood that

scores were based on ability and not on the willingness to spend additional time on it. Participants

were reminded of their anonymity and specifically asked not to cheat; however there was no way of

controlling for this and so a certain degree of goodwill is assumed. The questions were randomised to

avoid systematic bias, and, as specified by Sousa (2010), subjects were not told which category of CA

was being tested in each question.

3.3 Demographic questionnaire

The demographic questionnaire was administered at the end of the experiment, and asked

participants to disclose their gender, age, level of education, and income. Age and income were given

in ranges to increase anonymity. These details were acquired as such exogenous factors could

potentially impact an   individual’s  price   judgement   in   the   valuation tasks and therefore integrating

them into my regression analysis as control variables would isolate the effect of the anchor.

3.4 Subject pool

My subject pool was initially made up of willing family and friends, but a number of them

subsequently passed the link onto others to broaden the group. Almost all previous laboratory

experiments investigating the anchoring heuristic have been undertaken solely by university

students. I therefore looked to increase the heterogeneity of my subjects by asking people of a wide

range of ages and incomes to take part, thus adding to the external validity of the results.

4. Results and discussion

In total 203 people participated in my experiment; 58 in the WTP High treatment, 53 in WTP

Low, 47 in WTA High, and 45 in WTA Low. Participants were evenly split in terms of gender. 50% of

subjects were aged 18-24 and 41% were aged 45-64, reflecting the fact that the majority of

participants were either my peers or from my   parents’   generation. 81.59% of participants were

undertaking or had undertaken higher education studies.

The statistical analysis consists of a combination of nonparametric and regression analyses to

test the two hypotheses. Firstly, however, the results of the CA test shall be considered, as they play a

key part in the second hypothesis.

4.1 Cognitive ability

Scores in the CA test ranged from 2 to 12 (out of 12), with a mean of 7.28, a median of 8 and a

standard deviation of 2.6. Scores across all four treatments were not significantly different. Figure 3

Page 12: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 11

shows the full distribution of scores, with some skew towards the upper end, but a fairly even

distribution overall.

The cognitive reflection, quantitative reasoning and verbal reasoning questions had mean

scores (out of 3) of 1.43, 1.94, and 1.42, respectively, as well as relatively even distributions. The

scores of the sequential reasoning, however, had a mean of 2.48, and their distribution was heavily

skewed towards the higher end. This implies that the sequential reasoning questions were not as

successful as the other three types in distinguishing between people of different abilities, especially

considering 57% of participants scored 3 out of 3.

Nevertheless, the test as a whole was clearly effective in displaying a wide range of overall

ability.

4.2 Hypothesis 1

To test the first hypothesis, that the anchoring effect is present in my data, nonparametric statistics

are used initially. Table 1 shows the difference in average WTP between the WTP High and WTP Low

groups for each product individually and as an average across all four products. Mann-Whitney U

tests are run to test the null hypothesis that both groups are drawn from the same distribution, as

was done by Fudenberg et al. (2012) when comparing a high anchor group against a low anchor

group in their paper investigating the anchoring effect on WTP and WTA decisions. Due to the nature

of my high/low anchor binary variable, I felt this to be a more appropriate nonparametric test than

the Pearson Correlation used by BEJS. Table 2 displays the same information but for the WTA

treatments. Both tables of results give clear evidence of anchoring: in both the WTP and WTA

treatments, the high anchor groups gave higher valuations for every product.

Page 13: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 12

The ratios between low anchor maximum WTP and high anchor maximum WTP vary from 1.36 to

1.90 for each product, and the overall ratio is 1.56, showing that those in the WTP High treatment

valued the products at 56% higher than in the WTP Low treatment. These ratios are shown to be

significant by the Mann-Whitney U test, which allows me to reject the null hypothesis of homogeneity

across the two groups at the 1% significance level in all cases except that of the USB stick, where it

can be rejected at the 5% level.

The  story  is  similar  regarding  subjects’  minimum  WTA  with  ratios  ranging  from  1.83  for  the  

chocolates to 2.37 for the recipe book. The overall ratio is 2.08, showing that subjects exposed to the

£17 anchor stated minimum WTAs of more than double those of their low anchor counterparts. As

seen with the WTP treatment, the Mann-Whitney U test statistic is significant at the 1% level across

the board. These results therefore support the first hypothesis that the difference in price

judgements between the High and Low treatments is statistically significant and suggest that this

experiment has indeed induced the anchoring effect.

Table 1 Average WTP in GBP sorted by anchor and product

Anchor Wine Chocolates Recipe Book USB Stick Average for all products

Low (£3) (n=53) 6.32 6.28 5.49 5.74 5.96

High (£17) (n=58) 8.62 9.65 10.41 8.39 9.27

Ratio 1.36 1.54 1.90 1.46 1.56

Mann-Whitney test stat

z -2.802 -4.169 -4.575 -2.149 -4.540 (p-value) (0.005) (<0.001) (<0.001) (0.032) (<0.001)

Table 2 Average WTA in GBP sorted by anchor and product

Anchor Wine Chocolates Recipe Book USB Stick Average for all products

Low (£3) (n=47) 6.79 6.14 5.43 6.22 6.15

High (£17) (n=45) 14.1 11.26 12.86 12.91 12.78

Ratio 2.06 1.83 2.37 2.08 2.08

Mann-Whitney test stat

z -5.752 -5.253 -5.775 -4.810 -6.660

(p-value) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001)

Page 14: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 13

To check the robustness of these findings, regression analysis was implemented, allowing for

the control of certain   exogenous   factors   affecting   subjects’   price   judgements.   A   number   of   OLS  

regressions were run, with the dependent variables being the maximum WTP or minimum WTA

valuations (depending on the treatment). Logarithmic transformations were applied to the

dependent variables for two reasons: firstly, in logs, the distributions were closer to normality, and

secondly, this allowed for a percentage change interpretation of the independent variable

coefficients, which was more effective for comparison across treatments and between products. The

independent variable of interest was a high/low anchor dummy variable, the coefficient of which

represents the anchoring effect. The control variables were a gender dummy, a higher education

dummy, income category dummies and age category dummies. Table 3 displays the anchor dummy

coefficients that were extracted from the OLS regressions run.

These results represent the percentage difference in valuation between high and low anchor

subjects, when gender, education level, income and age are held constant. Ceteris paribus, the average

WTP for all products is 42.3% higher for those people who were exposed to the £17 anchor than

those exposed to the £3 anchor. The equivalent figure in the WTA treatment is 78.3%. The above

results are all statistically significant when tested at the 1% level, except for the wine WTP, which is

significant at the 5% level and the USB WTP, which is not significant. No other independent variables

in the regressions were found to be statistically significant.

The reason Table 3 includes fewer observations than Tables 1 and 2 is because not all

experiment participants were willing to disclose their demographic characteristics and thus their

data were dropped from the regressions. These results support the previous nonparametric findings

that the anchor does have a significant effect on a person’s  subsequent  price  judgement  and therefore

I am able to reject the null hypothesis of Hypothesis 1, providing further evidence to support the

findings of ALP and BEJS.

Comparing WTP and WTA, it is clear that the anchoring effect is greater when making selling

decisions; the WTP anchoring effects range from 26.8% to 75.2% whereas the WTA range is 72.3% to

Table 3 Anchoring effects (regression coefficients) in the WTP and WTA treatments

Wine Chocolates Recipe Book USB Stick Average for all products

WTP Coefficient 0.268 0.468 0.752 0.257 0.423 (n=101) (p-value) (0.012) (<0.001) (<0.001) (0.124) (<0.001)

WTA Coefficient 0.756 0.723 1.119 0.748 0.783 (n=87) (p-value) (<0.001) (<0.001) (<0.001) (<0.001) (<0.001)

Page 15: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 14

111.9%. The overall effect is 78.3% for WTA compared with only 42.3% for WTP. These results

support the findings of Sugden et al. (2013), who offer the convincing explanation that this difference

is due to the fact that subjects have more experience in purchasing goods than selling them, and are

therefore less influenced by the arbitrary anchor when in a buying scenario.

A further observation of interest from my results is that the anchoring effect is clearly

stronger  on  certain  products   than  others.  Following  on   from  Sugden  et  al.’s   (2013)  argument,   it   is  

reasonable to assume that most participants have had more transaction experience with wine and

chocolate than with a recipe book, so it is not surprising that the effect is greater on the recipe book.

This, however, would not explain why the anchoring effect is insignificant on WTP for the USB stick. A

second possible explanation could therefore be that in reality the price range of a typical recipe book

is wider than that of a USB stick, and so the range of perceivable values is narrower for the USB stick,

leading to a less pronounced anchoring effect. Further research would be required to explore these

points.

4.3 Hypothesis 2

Having established that the results support the existence of the anchoring effect, the study moves on

to consider the second hypothesis that more intelligent people are less susceptible to it. To test this,

the subjects were split into high and low CA groups and regressions were run in the same format as

those described above, so as to allow for comparison of the anchoring effect regression coefficient

between the two ability groups. Those who scored below the median, from 2 to 7 (out of 12), were

placed in the low CA group and those who scored the median or above, from 8 to 12, were placed in

the high CA group.

Table 4 WTP anchoring effects (regression coefficients) for low CA and high CA groups

Wine Chocolates Recipe Book USB Stick Average for all products

Low CA Coefficient 0.381 0.490 0.816 0.184 0.443 (n=51) (p-value) (0.024) (0.001) (0.001) (0.416) (0.003)

High CA Coefficient 0.201 0.493 0.674 0.014 0.377 (n=50) (p-value) (0.237) (0.008) (0.061) (0.961) (0.014)

Difference 0.180 0.003 0.142 0.170 0.066 (p-value) (0.442) (0.991) (0.734) (0.632) (0.749)

Page 16: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 15

Table 4 displays the results for the WTP treatments, which would not appear to support my

hypothesis. The differences between the coefficients are of small magnitude and, more importantly, a

number of the coefficients themselves are not statistically significant. For both the wine and the USB

stick, whilst the difference in coefficients is noticeable (0.18 and 0.17 respectively), only the low CA

wine result is significant. On the other hand, both Belgian chocolates anchoring coefficients are

significant but they are almost exactly the same, implying no difference between high and low CA.

However, to test the differences formally, Chow tests were conducted. This involved

introducing a high/low CA dummy variable and interacting it with all other independent variables,

which allowed me to test the null hypothesis that the two coefficients are not statistically different.

The coefficient of the CA/anchor interaction term represents the difference, which can be analysed in

line with its p-value. Not a single difference is found to be statistically significant, and most

importantly the overall difference for all products is only 0.066 and not significant (p=0.749). These

results lead to the conclusion that when making a purchasing price judgement, anchoring has a

similar effect on all participants, regardless of their CA.

The WTA results give a little more support to the second hypothesis, but not overly so. Shown

in Table 5, the anchoring effect regression coefficient for the average WTA of all products is 0.57 for

the high intelligence group, and 0.91 for the low intelligence group, with a difference of 0.34 that is

very close to significance at the 10% level, with a p-value of 0.102. Although all anchoring coefficients

are significant for the four individual products, only one difference is significant at the 10% level, and

that is of the chocolates (p=0.098). The USB treatment displays no difference at all across the two

intelligence levels.

Considering these results contradict those of BEJS, I decided to investigate further by

excluding from my regressions the participants who found themselves in the central 25% of CA

Table 5 WTA anchoring effects (regression coefficients) for low CA and high CA groups

Wine Chocolates Recipe Book USB Stick Average for all products

Low CA Coefficient 0.876 0.903 1.153 0.744 0.907 (n=42) (p-value) (<0.001) (<0.001) (<0.001) (0.001) (<0.001)

High CA Coefficient 0.538 0.447 0.932 0.754 0.566 (n=45) (p-value) (0.004) (0.074) (0.004) (0.020) (0.001)

Difference 0.338 0.456 0.221 -0.01 0.341 (p-value) (0.186) (0.098) (0.541) (0.977) (0.102)

Page 17: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 16

scores. In removing scores of 7 or 8 on the test, I could compare participants at more extreme ends of

the CA spectrum. Tables 6 and 7 summarise the findings for WTP and WTA respectively.

Regarding the WTP treatment, the differences do increase noticeably but it remains the case

that they are not significant. It should also be noted that only one of the high CA coefficients is

significant, which could imply that in the high intelligence group, the anchoring effect is not present;

however this is more likely to be the result of a smaller sample size. The average difference across all

products is still only 0.21, and found not to be significant by the Chow test.

The second set of WTA findings are not dissimilar from those found in Table 5. The overall

difference for all products has hardly changed, but it is further from significance with a p-value of

0.190. All ten anchor coefficients are significant, showing that subjects of all intelligence levels are

Table 6 WTP anchoring effects (regression coefficients), with CA test scores of 7 and 8 excluded Wine Chocolates Recipe Book USB Stick Average for all

products

Low CA Coefficient 0.398 0.420 0.995 0.070 0.423 (n=37) (p-value) (0.065) (0.026) (0.006) (0.823) (0.045)

High CA Coefficient 0.070 0.350 0.521 -0.132 0.210 (n=39) (p-value) (0.664) (0.061) (0.199) (0.646) (0.123)

Difference 0.328 0.069 0.474 0.202 0.213 (p-value) (0.206) (0.784) (0.368) (0.632) (0.365)

Table 7 WTA anchoring effects (regression coefficients), with CA test scores of 7 and 8 excluded

Wine Chocolates Recipe Book USB Stick Average for all products

Low CA Coefficient 0.941 0.883 1.169 0.739 0.911 (n=33) (p-value) (<0.001) (<0.001) (<0.001) (0.016) (<0.001)

High CA Coefficient 0.442 0.554 1.051 0.815 0.553 (n=31) (p-value) (0.049) (0.091) (0.028) (0.068) (0.024)

Difference 0.499 0.329 0.118 -0.076 0.358 (p-value) (0.122) (0.336) (0.804) (0.878) (0.190)

Page 18: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 17

influenced by the irrelevant £3 or £17 price but, again, no differences are significant. Having said this,

there is a clear contrast between certain products: the difference for wine is large and approaching

significance, whilst the differences for the recipe book and the USB stick are arbitrary and therefore

far from significance. This will be discussed in section 4.4.

Having conducted further analysis on both treatments, it is clear that the results of this

experiment fail to support the findings of BEJS in both the buying and selling contexts. Therefore, it is

not possible to reject the second null hypothesis that there is no significant difference in

susceptibility to anchoring across the different levels of CA.

4.4 Discussion

This paper set out to test for the existence of a relationship between CA and anchoring, and it

has ultimately found no significant link. While I may speculate below that a weak relationship does

exist, the overall findings of my experiment suggest that it does not, and this has certain economic

and psychological implications. In contrast to BEJS and the initial predictions, I find that, in the

context of price judgement anchoring, more intelligent people do not necessarily “behave more like

the   textbook   model   of   ‘economic   man’”   (BEJS,   2010,   p.67).   Participants   of   all   levels of CA are

influenced by the anchor in a significant way, regardless of whether they are making a buying or

selling decision, showing that more intelligent people are not more likely to reveal their true

preferences. In psychological terms, my results suggest that a person of higher CA is no more capable

of detaching themselves from the arbitrary anchor or referring to past experience when valuing a

good than someone of lower CA.

Furthermore, when conducting my statistical analysis, I checked whether any of the

individual types of CA, such as cognitive reflection or quantitative reasoning, were significantly

related to susceptibility to anchoring and found them not to be in all cases; however to test this

properly more than three questions per type would be required to elicit a good measure of each

individual ability. What my results are therefore showing is that the psychological processes that

affect how a person behaves in the presence of an anchor are not related to intelligence, and certainly

are not captured by the CA test implemented here.

On the face of it, my results would appear to completely contradict those of BEJS. The

variation in our results could potentially be explained by the variation in our experimental

procedure, especially regarding the CA test. BEJS argue that Oechssler et al. (2009) and Stanovich and

West (2008) find no relationship between CA and anchoring because they use poor proxies for

intelligence, and they would probably extend the same argument to my 12-question test, which was

not as comprehensive as their 44-question professionally developed psychometric test of general

intelligence. Perhaps if the same test had been used, higher significance would have been found.

Page 19: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 18

Having said this, our findings are not necessarily as different as they may first appear. For the

average of all products, BEJS did have a statistically significant difference of 0.550, but the differences

for three of their six individual products were not significant, showing that their results were neither

definitive   nor   entirely   convincing.   It   could   be   that  BEJS’s paper   is   an   example   of   a   “false   positive”  

(Maniadis et al., 2013, p.1), whereby their data reveal a relationship that does not actually exist.

In most cases, however, I did find a notable difference in anchoring effect between those of

high and low intelligence, though the Chow test found it not to be significant. My results would imply

that CA is not entirely irrelevant, but that further research is required to explore the relationship in

more depth. This is especially true considering that in both treatments the difference in anchoring

coefficient was typically larger and closer to significance when valuing the wine and chocolates, yet

non-existent when valuing the USB stick. The fact that the effect of intelligence varies between

products implies that it does play some part in certain cases. Similarly, while not statistically

significant, the differences are clearly larger in the WTA treatment than in the WTP treatment. Again,

this would superficially imply that the difference in intelligence has more of an effect in the selling

scenario, but this would need to be investigated further.

In summary, despite the speculation that a weak relationship may exist, the lack of

significance in the results fails to support the findings of BEJS. In running an experiment of similar

structure, I found strong evidence for the existence of anchoring and thus support for the first

hypothesis. However, the results  do  not  support  the  robustness  of  BEJS’s conclusion that intelligence

and anchoring are negatively linked in either the buying context, as BEJS investigated, or the selling

context; even though the differences in the WTA treatment were bigger than in the WTP treatment,

they still were not significant. This leads me to conclude that my findings instead support those of

Oechssler et al. (2009), Stanovich and West (2008), and Furnham et al. (2012).

5. Evaluation

5.1 Experimental procedure

The results are quite unambiguous in rejecting the second hypothesis. However, considering

the experimental procedure departed from the standard practice in a number of ways, the external

validity of these findings must be evaluated.

The primary issue in this regard is the incentive structure of the valuation tasks. While I

believe the payoff mechanism was sound for the WTA treatments, the WTP mechanism could have

been conducted differently. Rather than ensuring the buying transaction took place at the

appropriate price, subjects were only offered the opportunity to buy at that price. The purpose of this

was to avoid subjects making real losses, however this could have been accounted for if the relevant

participants had been given endowments from which they could have purchased the product. Had

Page 20: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 19

the incentive structure been stronger, the experiment may have induced more truthful responses in

the WTP treatments. Perhaps the findings would have then been more significant regarding the

second hypothesis, as subjects with higher CA scores may have applied more effort and concentration

in the tasks, and thus may have been less influenced by the anchor.

The first hypothesis would not appear to have been similarly affected, considering the WTP

treatments induced less anchoring, which does fall in line with the findings of past literature.

Additionally, the payment structure was correct for the WTA treatment and yet the Hypothesis 2

findings were still not significant when subjects were selling.

A further factor is that the CA test was not incentivised and I was unable to control the

amount of effort or time people were willing to put into it, which raises the pertinent question of

whether the test was effective in measuring intelligence. It is possible that someone of high

intelligence scored a low mark simply because they completed the experiment at a busy time or

without due attention. I believe this is the point that most draws into question the validity of the

results. The benefit of not incentivising such a test in an online environment is that there is no

resulting incentive to cheat for the purposes of monetary gain. The downside is that there is no

reason, aside from goodwill, for a participant to try and do the best they can, especially knowing that

the test was taken anonymously. While I do not think participants would have cheated, it is unlikely

that  every  person’s  score  perfectly  reflects  the  best  score they could have got, which has implications

for my results.

Whether the test was a good proxy for intelligence is another matter. Ideally, subjects would

have completed a full IQ test,  as  BEJS’s subjects did, but considering the extra time this would have

taken, the test used was a best alternative. In just 12 questions it measured four types of skill and

gave a good general measure of CA. That said, the difficulty of the sequential reasoning questions

could be increased due to reasons that have already been discussed: the mean score for these three

questions was 2.48 and 57% of subjects answered all three correctly, showing that they did not

successfully distinguish between people of different ability.

A more general problem with the experimental procedure is that it was conducted online,

which inevitably brings a lack of control when compared to a laboratory experiment. While the

approach was designed to control for as many relevant factors as possible, it could not fully account

for exogenous factors such as distractions, cheating, or even disinterest in the task. This is an

unfortunate side-effect of conducting research in this way.

Despite the above points, the experiment was considered successful in several areas.

Standard procedure was followed regarding the test for anchoring and as a result I found the

phenomenon to be present in the data. Additionally, products and anchor values remained constant

across all treatments, allowing proper comparisons to be made. Although the results do not support

Page 21: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 20

BEJS’s findings, they do have interesting implications and show that the area will benefit from more

attention.

Furthermore, the use of an online platform allowed for a subject pool that was more

heterogeneous than in the majority of the literature. Most previous anchoring experiments have

subject pools comprised entirely of students, representing a very specific demographic group,

typically with low incomes. Half of participants in my experiment were aged over 24, and incomes

ranged from £0 to £100,000+, making my subjects more closely representative of the general

population than is usually the case. This would therefore add to the external validity of my findings.

I should note that a total sample size larger than 203 would have been desirable to eliminate

the possibility that the lack of significance of my results is simply caused by too few participants. This

is especially the case regarding the second set of regressions that excluded all subjects who scored 7

or 8 in the test, seeing as a number of the differences were large but still not considered significant by

the Chow test.

5.2 Further research

This paper is unique in exploring the relationship between intelligence and anchoring in a

selling context and as a result experimental replication is required to check the robustness of the

findings.  As  Maniadis  et  al.  (2013)  argue,  “a  few  independent replications dramatically increase the

chances  that  a  given  original  finding  is  true”  (2013, p.1). The next step in this area of research would

therefore be to conduct an experiment similar to this one, in a more controlled laboratory setting,

taking into account the points I outlined above, including a revised incentive structure designed to

elicit responses that reflect real world behaviour.

Replication is especially important considering the contrast between these results and those

of BEJS. If it is found that there is indeed no relationship between intelligence and anchoring in a

price judgement scenario, then there would be two directions in which the literature could move. The

first would be to check whether these results hold in other anchoring contexts such as when giving

probability estimates. The second would be to continue investigating the individual differences, other

than CA, that do influence susceptibility to anchoring, such as personality and expertise.

The results do suggest, however, that intelligence has a small effect on anchoring for certain

products but not for others, and that this effect is different for WTP and WTA. I think these two points

are worthy of further investigation

Finally, this paper makes no attempt to distinguish between the different types of CA because

it was taking into account a general measure. Further research might replicate this experiment whilst

testing each type of CA individually and more comprehensively, to establish whether a relationship

does in fact exist, if only in certain cases.

Page 22: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 21

5.3 Conclusion

This   paper   set   out   to   test   the   robustness   of   BEJS’s findings, and to extend their research

question through an investigation into whether more intelligent people are also less susceptible to

the anchoring effect when stating their minimum WTA. An online experiment was run comprising

four valuation tasks of standard consumer products, followed by a 12-question test of CA. Having

found strong evidence of the anchoring heuristic within the data, I proceeded to check for the

existence of a relationship between intelligence and anchoring in order to test the second hypothesis.

In both the WTP and WTA treatments, the results do not support those of BEJS. Evidence was

not found to suggest that intelligence and anchoring are significantly related, implying that when

making a price judgement, people of higher CA are just as influenced by an arbitrary number as those

of lower CA. This in turn suggests that more intelligent people are no more likely to act rationally or

reveal their true preferences in this context. Whilst it could be that the lack of significant results is

due to issues with experimental procedure, it is also possible that these findings are robust and

externally valid; further research is required to establish whether this be true.

This paper contributes to the literature by drawing into question the findings of BEJS, in both

a buying and a selling context. In the few papers that explore the relationship between intelligence

and anchoring, there is already discord, giving support to the argument that further investigation is

required to develop a deeper and more nuanced understanding of the matter.

Page 23: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 22

6. References Ariely, D., Loewenstein, G., Prelec, D. (2003). Coherent arbitrariness: stable demand curves without

stable preferences. Quarterly Journal of Economics, 118, 73-105.

Benjamin, D., Brown, S., Shapiro, J. (2006), Cognitive ability and anomalous preferences. Mimeo, Harvard University.

Bergman, O., Ellingsen, T., Johannesson, M., Svensson, C. (2010). Anchoring and cognitive ability.

Economics Letters, 107 66-68.

Carlson, B. W. (1990). Anchoring and adjustment in judgments under risk. Journal of Experimental Psychology: Learning, Memory and Cognition, 16, 665-676.

Dohmen, T., Falk, A., Huffman, D., Sunde, U. (2007). Are risk aversion and impatience related to

cognitive ability? IZA Discussion Paper, 2735.

Epley, N., Gilovich, T. (2001). Putting adjustment back into the anchoring and adjustment heuristic:

differential processing of self-generated and experimenter-provided anchors. Psychological Science,

12, 391–396.

Epley, N., Gilovich, T. (2005). When effortful thinking influences judgemental anchoring: differential

effects of forewarning and incentives on self-generated and externally provided anchors. Journal of Behavioural Decision Making, 18, 199-212.

Eroglu, C., Croxton, K.L. (2010). Biases in judgmental adjustments of statistical forecasts: the role of

individual differences. International Journal of Forecasting, 26, 116–133.

Frederick, S. (2005). Cognitive reflection and decision making. Journal of Economic Perspectives, 19

(4), 25-42.

Furnham, A., Boo, H.C. (2011). A literature review of the anchoring effect. The Journal of Socio-Economics, 40, 35-42.

Furnham, A., Boo, H.C., McClelland, A. (2012). Individual Differences and the Susceptibility to the

Influence of Anchoring Cues. Journal of Individual Differences, 33(2), 89–93.

Horowitz, J., McConnell, K. (2002). A Review of WTA / WTP Studies. Journal of Environmental Economics and Management, 44, 426-447.

Kahneman, D., Knetsch, J., L., Thaler, R. (1990). Experimental tests of the endowment effect and the

Coase theorem. Journal of Political Economy, 98, 1325-1348.

Kuncel, N., Credé, M., Thomas, L. (2005). The validity of self-reported grade point averages, class

ranks, and test scores: a meta-analysis and review of the literature. Review of Educational Research,

75, 63-82.

Maniadis,  Z.,  Tufano,  F.,  List,   J.,  A.   (2013).  One  Swallow  Doesn’t  Make  a  Summer:  New  Evidence  on  Anchoring Effects. CeDEx Discussion Paper Series, 2013-07.

McElroy, T., Dowd, K. (2007). Susceptibility to anchoring effects: how openness-to-experience

influences responses to anchoring cues. Judgment and Decision Making, 2, 48–53.

Page 24: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 23

Mussweiler, T., Strack, F. (1999) Hypothesis-consistent testing and semantic priming in the anchoring

paradigm: a selective accessibility model. Journal of Experimental Social Psychology, 35, 136-164.

Northcraft, G.B., Neale, M.A. (1987). Experts, amateurs, and real estate: an anchoring-and-adjustment

perspective on property pricing decisions. Organizational Behavior and Human Decision Processes, 39,

84–97.

Oechssler, J., Roider, A., Schmitz, P.W. (2009). Cognitive abilities and behavioural biases. Journal of Economic Behaviour and Organization, 72, 147-152.

Plous, S. (1989). Thinking the unthinkable: The effects of anchoring on likelihood estimates of nuclear

war. Journal of Applied Social Psychology, 19, 67-91.

Simonson,  I.,  Drolet,  A.  (2004).  Anchoring  effects  on  consumers’  willingness-to-pay and willingness-

to-accept. Journal of Consumer Research, 31, 681-690.

Sousa, S. (2010). Are smarter people really less risk averse? CeDEx Discussion Paper Series, 2010-17.

Stanovich, K.E., West, R.F. (2008). On the relative independence of thinking biases and cognitive

ability. Journal of Personality and Social Psychology, 94, 672-695.

Strack, F., Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: mechanisms of selective

accessibility. Journal of Personality and Social Psychology, 73, 437–446.

Sugden, R., Zheng, J., Zizzo, D.J. (2013). Not all anchors are created equal. Journal of Economic Psychology, 39, 21–31.

Thaler, R. (1980). Toward a positive theory of consumer choice. Journal of Economic Behaviour and Organization, 1, 39-60.

Tversky, A., Kahneman, D. (1974). Judgement under uncertainty: heuristics and biases. Science, 185,

1124-1131.

Page 25: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 24

Appendix A: Questionnaire

A.1 Introduction

Page 26: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 25

A.2 WTP Explanation page

Page 27: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 26

A.3 WTP Questions to check understanding

The text in blue appeared after the corresponding question had been answered.

Page 28: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 27

A.4 WTA Explanation page

Page 29: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 28

A.5 WTA Questions to check understanding

The text in blue appeared after the corresponding question had been answered.

Page 30: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 29

A.6 Wine valuation task

This is an example from the WTP High treatment.

Values are entered for demonstration purposes; subjects had blank spaces in which to answer.

Page 31: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 30

A.7 Belgian chocolates valuation task

This is an example from the WTP Low treatment.

Values are entered for demonstration purposes; subjects had blank spaces in which to answer.

Page 32: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 31

A.8 Recipe book valuation task

This is an example from the WTA High treatment.

Values are entered for demonstration purposes; subjects had blank spaces in which to answer.

Page 33: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 32

A.9 8GB USB stick valuation task

This is an example from the WTA Low treatment.

Values are entered for demonstration purposes; subjects had blank spaces in which to answer.

Page 34: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 33

A.10 Cognitive ability test

Questions were randomised.

Correct answers are entered here for demonstration purposes.

Page 35: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 34

Page 36: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 35

Page 37: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 36

A.11 Demographic questionnaire

Page 38: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 37

Appendix B: Regression output tables

All regressions took the same format, so one WTP and one WTA example are given.

B.1 Regression output table for ‘WTP: Average for all products’ (found in Table 3)

Number of obs = 101 F (14, 86) = 2.66 Prob > F = 0.003 R² = 0.302 Adj R² = 0.189

Average WTP (logs) Coefficient Standard Error t P  >  │t│ [95% confidence interval]

High anchor .4228681 .0946679 4.47 0.000 .2346746 .6110616

Male .0286447 .0941634 0.30 0.762 -.1585461 .2158354

Higher education

.0554399 .1323124 0.42 0.676 -.2075885 .3184682

Income

Group 2 -.0962399 .1822303 -0.53 0.599 -.4585017 .2660219

Group 3 -.048309 .2014169 -0.24 0.811 -.4487126 .3520945

Group 4 .0531143 .3686012 0.14 0.886 -.6796406 .7858693

Group 5 .0337771 .2391853 0.14 0.888 -.4417076 .5092618

Group 6 -.520335 .2936453 -1.77 0.080 -1.104082 .0634124

Group 7 -.2499797 .233069 -1.07 0.286 -.7133055 .213346

Age

Group 2 .2493917 .2550043 0.98 0.331 -.25754 .7563234

Group 3 -.1318741 .2797127 -0.47 0.639 -.6879245 .4241763

Group 4 .0068182 .2055535 0.03 0.974 -.4018086 .4154451

Group 5 -.0570527 .1924786 -0.30 0.768 -.4396874 .325582

Group 6 .0894002 .3037076 0.29 0.769 -.5143505 .6931508

Constant 1.713019 .1465369 11.69 0.000 1.421713 2.004324

Page 39: Undergraduate Dissertation - Does Intelligence Affect Susceptibility to Anchoring?

Louis Adams 38

B.2 Regression output table for ‘WTA: Average for all products’ (found in Table 3)

Number of obs = 87 F (14, 86) = 5.96 Prob > F = 0.0000 R² = 0.5366 Adj R² = 0.4465

Average WTA (logs) Coefficient Standard Error t P  >  │t│ [95% confidence interval]

High anchor .7829722 .0930785 8.41 0.000 .5974235 .9685209

Male .0650118 .0916341 0.71 0.480 -.1176573 .247681

Higher education

-.0895331 .1345627 -0.67 0.508 -.357779 .1787128

Income

Group 2 -.2058362 .219582 -0.94 0.352 -.643565 .2318925

Group 3 -.2613322 .1763257 -1.48 0.143 -.612831 .0901666

Group 4 -.3138622 .2294593 -1.37 0.176 -.7712811 .1435566

Group 5 .083074 .2100037 0.40 0.694 -.3355607 .5017088

Group 6 -.0271214 .2450597 -0.11 0.912 -.515639 .4613962

Group 7 -.0985883 .2125234 -0.46 0.644 -.522246 .3250694

Age

Group 2 -.0272748 .2770223 -0.10 0.922 -.5795085 .524959

Group 3 .7028294 .4623622 1.52 0.133 -.2188728 1.624532

Group 4 .0818781 .1815544 0.45 0.653 -.280044 .4438001

Group 5 .0242351 .1890821 0.13 0.898 -.3526931 .4011633

Group 6 .1121862 .4582636 0.24 0.807 -.8013456 1.025718

Constant 1.7946 .1492053 12.03 0.000 1.497165 2.092036