The Effect of Conscious Awareness on Framing by
Sabine Tijdeman – S11876956
University of Amsterdam
Master Thesis – Business Economics: Managerial Economics & Strategy 15 ECTS
1
Abstract
People are becoming increasingly aware of how framing effects can influence people’s decisions. When knowledge about framing becomes more widespread, it is interesting to see
how conscious awareness of framing affects decisions. This paper investigates the effect of conscious awareness on context framing. An experiment was conducted where subjects have
to play a game and have to decide whether to cooperate or not. The name of the game was either given a cooperative or a competitive frame and subjects were made aware of this frame
in the treatment as a measure of conscious awareness. It was found that conscious awareness decreases the difference in the average cooperation between the competitive and the cooperative frame and that the average cooperation decreased in both the cooperative and the
competitive frame when exposed to awareness. However, the differences were not found to be statistically significant.
Introduction
Kahneman & Tversky (1986) were the first to discover that the way a situation is framed has a significant impact on judgment and choice. Since then, framing has received much attention in the economics literature and awareness of the framing effect is increasing. However, how this awareness affects framing is unknown. With this paper, I would like to create a starting point in the literature about the effect of awareness on framing.
An experiment was conducted in which subjects had to play a game where they have to decide whether to cooperate or not. The two games are the same except for the name. The name of the first game consists of a cooperative frame, while the name of the other game consists of a competitive frame. Together these names establish the framing effect. In order to measure the effect of awareness of the frame on the framing effect itself, two other groups played the game again but were informed explicitly in the instructions that two games exist with different frames and that these were chosen to influence their behaviour.
It was found that conscious awareness decreases the difference in average cooperation between the competitive and the cooperative frame. However, the framing itself was not significant to begin with. Therefore, it can only be said that subjects behaved more similar. Furthermore, the average cooperation decreases in both the cooperative frame and
competitive frame. This supports the theory that people feel duped by framing and rejects the theory that people make more logical decisions when exposed to awareness of framing. However, this difference between the cooperative frame with and without awareness and the competitive frame with and without awareness is neither found to be significant.
2
Literature Review
Framing. Kahneman and Tversky (1986) were two of the first researchers to show that the theory of rational decision making is commonly violated and that how one frames a situation can have a significant impact on decisions. Researchers have found that by only changing the name of the game one can already achieve significantly different results. Changing the emphasis of the game is explained as context framing by Gerlach & Jaeger and ‘shapes players experience of the game by associating the game with different entities and/or stressing specific aspects of the game’ (p. 3, 2016).
Liberman et al. (2004) conducted an experiment where subjects had to play the same prisoner’s dilemma except the title of the game was different. They found that the difference in cooperation was around twice as high for the ‘Community game’ compared to the ‘Wall Street game’. Ellingsen et al. (2012) conducted an almost similar experiment (instead of Wall street game they named it the Stock Market game) and found also that in the ‘Community game’ cooperation was about twice as high. Furthermore, the significant difference between a cooperative and a competitive frame is also found to be significant in other types of games. Cookson (2000) and Rege & Telle (2004) both find that a more community oriented frame leads to higher cooperation rates in a public good game. Even subjects who had to play a game on the Nintendo cooperated more when given instructions with a cooperative perspective and acted more aggressive when given instructions with a competitive perspective (Anderson & Morrow, 1995).
How can it be explained that framing seems to have a considerable effect on choices of individuals? The accessibility-driven process explains this effect by that framing
‘influences accessibility of associations in memory, thereby increasing the likelihood that audience members will follow those associations when thinking about the issue’ (Brewer et al., p. 496, 2003; Cappella & Jamieson, 1997).
The changed behaviour due to framing can be either because people create
preferences or people create beliefs about expected behaviour (Bacharach, 2006; Camerer & Fehr, 2004). Ellingsen et al. (2012) compared the decisions of players paired to other players with players paired to a computer. It was found that the difference in the level of cooperation vanished when players played against a computer and thus indicates that the decisions of players depend on their beliefs of what other players will do. If these were preferences induced by framing, there should not have been a difference in cooperation when playing against a computer or another human being. This also indicates that it is second-order beliefs that drive decisions. Thus, a player bases their decision not on what they believe they should
3
do (first-order beliefs), but what they believe their paired player will do (second-order beliefs). This is a form of social learning (Call et al., 2005). Social learning means that humans learn by observing and imitating other humans. Therefore, as humans age, their decisions are increasingly influenced by others and framing effects become stronger (Reyna & Farley, 2006; Albert & Steinberg, 2011; Strough et al., 2011).
Thus, when it is second-order beliefs that determine the subjects’ decision, it can change over time, because beliefs can be updated whereas preferences are fixed. Fischbacher & Gächter (2010) state that beliefs are a weighted average of what others decide in the previous period and one’s own belief in the previous period. If a subject’s decision is driven by preferences and they prefer to cooperate, then their strategy likely will not change over time. However, when decisions are based on beliefs of what other players will do, then every round their beliefs will be altered. In public good games it is usually found that average contribution declines over rounds. One reason for this is because many subjects only cooperate when other subjects also cooperate. However, usually there are also free-riders present who do not contribute. When a free-rider does not contribute, but does get a share of the contribution the subjects is less likely to contribute to the public good in the next round (for example Fehr & Gachter, 2000; Camerer & Fehr, 2002). Because the framing effect sets a norm on contribution levels and one of the paired players does not contribute, this norm is violated. When the norm is violated, the framing effect is less effective. Furthermore,
Ellingsen et al. (2012) also tested if it made any difference if the prisoner’s dilemma with the ‘Community Game’ and the ‘Stock Market Game’ was played sequentially or
simultaneously. They found that the social framing effect vanishes when the game is played sequentially. This is consistent with the hypothesis that social frames serve as a coordination device (Ellingsen et al., 2012).
Awareness. Awareness can be gained by statistical learning which means gradually gaining knowledge (Goujon et al., 2014). Statistical learning is often implicit, because the learning is unintentional (Chun & Jiang, 2003). In such a setting, subjects are first
unconsciously aware and slowly, by statistical learning, become consciously aware. Much of the literature of awareness focuses on subjects gaining awareness through statistical learning. Statistical learning can also be called gaining experience. Gilad and Kliger (2008) found in their study that financial professionals were more influenced by framing than undergraduates enrolled in a finance degree. Professionals seemed to make decisions by intuition whereas undergraduates used a more analytical approach. Thus, the effects of framing in combination with unconscious awareness is unclear and can be context-dependent.
4
Another way humans can gain awareness is by stating explicit knowledge. When a person becomes consciously aware of their views, attitudes, or biases, that person can significantly change their perception and consequently change their behaviour. One major example of how conscious awareness can change perception and attitude is the decreasing amount of tobacco smokers in the United States. In the first half of the 20th century smoking was branded as a luxury good which famous people consumed and was popular among the general public. However, throughout the years increasing amount of studies were presented on the negative health effects of smoking. People became increasingly aware of the negative impact smoking has on their health and therefore the amount of smokers declined (Cummings & Proctor, 2014). Yet, how conscious awareness affected smoking behaviour is difficult to examine, because there are many other factors influencing smoking behaviour.
Most of the literature investigates whether people have irrational perceptions and biases. However, how people change their behaviour if they become aware of these
perceptions and biases has received little attention until recently. Pietri et al. (2018) conduct an experiment where male and female students are made aware of the gender bias in science, technological, engineering and mathematics (STEM) fields. It was found that as soon as women learn about how this bias negatively affects them, their sense of belonging and trust in the stem environment decreases.
Humans are sensitive to being taken advantage of in social contexts. When humans are being taken advantage of it leaves them feeling duped. Feeling duped is ‘a response to an inequitable and disadvantageous exchange outcome thein advance one hoped would turn out fairly, but on the other hand, one could have foreseen being taken advantage of’ (Vohs et al., p. 130, 2007). Being duped is threatening when someone else or something else has the control over the outcome and can lead to an aversive state in humans, which can influence motivation and cognition. Awareness can help prevent people from becoming duped again. Sagarin et al. (2002) let people rate how convincing advertisements were and then informed then that the advertisements were deceptive to show their vulnerability. After recognizing their vulnerability, subjects were better able to resist getting duped by illegitimate
advertisements. Furthermore, another research found that the more advertisements the subjects are shown, the more sceptic people become of the honesty of the advertisements (Koslow, 2000). Subjects become more sensitive to the possibility that they are being persuaded and therefore the intended effect might backfire.
5
Experimental Design, Procedures, Payoff Structure and Strategy
Experimental Design. In order to measure if the context framing effect decreases
when subjects are made aware of the framing, it was decided that a simple game would be played in which subjects have to decide to cooperate or not. Each round, subjects play anonymously against another player. Because subjects play anonymously, image or
reputation should not have an effect on their behaviour. A subject can earn a certain payoff when deciding to cooperate or not. It was chosen to give the name of the game a cooperative and a competitive frame, because the difference has been proven to be significant in many papers (for example Liberman et al., 2004; Ellingsen et al., 2012). Because the experiment was conducted with high school students in a rural area, it was not possible to choose the same names as Liberman et al. and Ellingsen et al. since students might not be familiar with Wall Street or the stock market. Therefore, two Dutch sayings ‘Samen staan we sterk’ and ‘Wie niet waagt wie niet wint’ were chosen. This translates into ‘together we are stronger’ and ‘who doesn’t take a chance doesn’t win’ and is equivalent to a cooperative and a
competitive frame. One group played the game with the cooperative frame and one with the competitive frame. The cooperative frame is expected to induce a more cooperative attitude since the name already says that if you work together you get better results. A cooperative frame generates a more cooperative attitude in subjects (Ellingsen, 2012). The competitive frame says that if you do not take a gamble you will never win and is expected to lead to a more competitive attitude. When a subject defines a situation as competitive, he or she is more likely to behave competitively (Anderson & Morrow, 1995). Both games therefore have a different reference point of what level of cooperation is expected of them. The difference in the cooperation level between the cooperative and the competitive frame establishes the baseline in order to analyse the effect of conscious awareness of framing.
In the treatment, subjects were informed in the instructions that there were two different games with different frames which were chosen to influence their decision and that they were in either one of the games. With this method, subjects are made explicitly aware of the framing. As a result, the effect of conscious awareness can be measured. Also, the method prevents confusion because subjects understand why framing is used. Another option was to only state the existence of the other game, but then one might risk that subjects do not completely understand that these different frames are used to influence their decision. Besides, the instructions were exactly the same in all 4 groups, so that when a difference in the cooperation level is found, it can be attributed to conscious awareness of framing.
6
Furthermore, it was decided to randomly award one student with a payoff during one of the rounds that had actual payoffs in order to incentivise students to participate. To make sure a student was randomly chosen, an application on a smartphone was downloaded that randomly allocated a round and a student. The student who was chosen was anonymously paid in cash after the experiment.
Procedures. The experiment was conducted at a local high school in the Netherlands. It was chosen to conduct the experiment with students from the classes with the two highest levels out of the six levels that exist in the Netherlands, so that there was a higher chance that subjects would understand the experiment. The high school was able to provide four groups with children aging from 13-15 years old. The groups consisted of 14 to 26 students.
The experiment was conducted in one day and every group was questioned right after another group. Thus, there was no time gap between the experiment and students were not able to communicate with each other about the purpose of the experiment. Furthermore, the groups who were not exposed to awareness played the game first and the groups who were exposed to awareness came second, so that even if they did talk about the experiment, they would not be aware of the purpose of the experiment.
Every group was divided in two, so that half of the group (players A) played against the other half of the group (players B). Every round, one player A would play against one player B anonymously. Subjects were matched randomly and each round, they played against a different subject. Subjects were aware that they played against another participant, because it was clearly stated in the instructions (see Appendix A). Each student had to answer their decision (cooperate or not) on one answer sheet with their tag (A or B matched with a
number) and one without. The answer sheet without a tag was given to the matched player, so that every subject knew their payoffs after each round. In order for the experiment to go smoothly, a relative helped the experimenter throughout the experiment.
Before the experiment started, all subjects were handed out instructions. The
instructions were first read to the subjects out loud and afterwards subjects had two minutes to ask questions and go through the instructions again, which were printed and available on their desk. In the instructions, it was explained how to play the game and that the game would be played for four rounds. The instructions also included a risk-aversion test (for why this test is included please refer to 5.3). Subjects were asked to fill in the risk-aversion test before the game began. Furthermore, in the end the students were told that this was the end of the experiment and that one student would be randomly chosen by the experimenter to receive their payoffs of that round.
7
Payoff Structure. Below, in table 1, the payoff structure can be found. Player 1 is the subject themselves and player 2 is their paired player in the corresponding round. The first payoffs in the table correspond to the payoffs of player 1 and the second payoffs correspond the payoffs of player 2. The subject can either choose to cooperate or not. Cooperating generates a payoff of €5 when the paired player also cooperates and €2.5 when the paired player does not cooperate. Not cooperating generates a payoff of €10 when the paired player cooperates and €0 when the paired player does not cooperate. The payoffs structure is the same for the paired player and therefore the payoff structure is symmetric.
TABLE I
Payoff Structure of the Game Player 2
Cooperate Not cooperate
Player 1 Cooperate €5, €5 €2,5, €10
Not cooperate €10, €2.5 €0, €0
Notes. First payoffs correspond to player 1’s payoffs, second payoffs to player 2’s payoffs.
Strategy. When subjects act as rational economic agents, their strategy will always be to optimize their own payoffs. When player 1 cooperates, player 2’s best response is to not cooperate because €10 is higher than €5. When player 1 does not cooperate, player 2’s best response is to cooperate because €2.5 is higher than €0. Player 2 will choose the same strategy, because their payoffs are symmetric. When a subject does not cooperate when the other player cooperates, they have no incentive to deviate. Therefore, the game has two Nash equilibriums.
Hypotheses
Framing effects have shown to have significantly higher cooperation rates in a cooperative frame versus a competitive frame in several kinds of studies and games
(Liberman, 2004; Ellingsen et al., 2012; Kay & Ross, 2003; Cookson, 2000; Rege and Telle, 2004; Anderson & Morrow, 1995). A competitive frame brings up a more aggressive attitude and subjects tend to see it as a competition where there can only be one winner (Anderson & Morrow, 1995). Therefore, it is expected that subjects will cooperate more on average in the cooperative frame in comparison to the competitive frame.
8
Hypothesis 1: Average cooperation is higher in the cooperative frame compared to the competitive frame
In Ellingsen (2012), the framing effect vanishes when the ‘Community Game’ versus ‘Wallstreet Game’ is played sequentially. In this experiment subjects will not play
sequentially, because they are matched with another player each round. However, Ellingsen (2012) also finds that it is second-order beliefs that drive decisions and not preferences. The framing effect serves as a coordination device and second-order beliefs are updated each round. Second-order beliefs are a weighted average of what others decide in the previous round and one’s own belief in the previous round (Fischbacher & Gächter, 2010). In each round, second order beliefs are updated since the subject knows the outcome of the previous round. Therefore, it is expected that the framing effect diminishes over rounds.
Hypothesis 2: The framing effect diminishes over rounds
When subjects become aware of the frame, they move from the implicit cognitive system to the explicit cognitive system. The implicit system is unconscious and automatic whereas the explicit system is conscious, controlled, and logical (Evans, 2002). In the implicit system humans respond fast, which is based on incomplete information. When stating explicit information, a person can re-evaluate and make adjustments to their decision-making process. When subjects make these adjustments it is expected that the adjusted reasoning is more logical and thus the framing effect should not influence their decision. Therefore, it is expected that conscious awareness decreases the difference in cooperation between the cooperative and the competitive frame.
Hypothesis 3: Conscious awareness decreases the difference in the average level of cooperation between the cooperative frame and the competitive frame
If the difference in average cooperation decreases between the cooperative and the competitive frame, then the question remains if the level of cooperation decreases in the cooperative frame, increases in the competitive frame or both. In the cooperative frame, subjects are aware that the name of the game is chosen to increase cooperation levels and therefore it is expected that when subjects are exposed to conscious awareness, they will
9
cooperate less. In the competitive frame, subjects are aware that the name of the game is chosen to decrease cooperation levels and therefore it is expected that when subjects are exposed to conscious awareness, they will cooperate more.
Hypothesis 3a: The cooperation level decreases in the cooperative frame Hypothesis 3b: The cooperation level increases in the competitive frame
However, Vohs (2007) found humans are sensitive to being taken advantage of in social context. When subjects have the feeling that they were duped, it could lead to aversive states in humans. When subjects are made explicitly aware of why that framing was used to influence their level of cooperation, this can trigger this aversive state in subjects. Therefore, it can also be that a subject’s average cooperation decreases in both the cooperative and the competitive frame when exposed to conscious awareness.
Hypothesis 4: The cooperation level decreases in both the cooperative and the competitive frame
Summary Statistics & Result
Parameters. The dependent variable is called ‘Player 1’ and is measured as a dummy
variable which equals 1 when a subject decides to cooperate and 0 when the subject decides not to cooperate. To measure the difference between the cooperative frame and the
competitive frame, a dummy variable is created which is called ‘Cooperative frame’.
‘Cooperative frame’ equals 1 when a subject is in the cooperative frame and 0 when a subject is in the competitive frame. To measure the effect of awareness, another dummy variable is created which is called ‘Awareness’. ‘Awareness’ equals 1 if the subject is exposed to
awareness and 0 when the subject is not exposed to awareness. ‘The gender variable is called ‘Female’ and equals 1 if the subject is female and 0 if the subject is male. Risk aversion ranges from 1 to 5 where 1 corresponds being risk-averse and 5 to being risk-taking (please read further for how and why risk-aversion is measured).
Summary statistics. A total of 88 subjects participated for four rounds in the experiment which sums up to a total of 352 observations. Unfortunately, it was not possible to evenly distribute the groups, thus some groups have more subjects than others (min 14 subjects and max 26). Below, in table 2, one can find the summary statistics. About half of the subjects in the sample is female and subjects are on average risk-neutral.
10
TABLE II Descriptive Statistics
Variable Obs Mean Std. Dev. Min Max
Player 1 352 .409 .492 0 1
Female 352 .511 .501 0 1
Risk 352 2.690 1.282 1 5
It was not possible to randomly assign people to groups beforehand, but it is still important to see if there are significant differences between the groups. Therefore an ANOVA test was conducted to see if there were significant differences between the risk-aversion in subjects and gender (see Appendix C). It was found that there is a significant difference between the average amount of females distributed between groups. By running a Bonferroni test it was found that the main difference is between the mean of females in competitive frame in the control group and the cooperative frame in the treatment group. This can cause a threat to the internal validity.
Usually, women are more risk averse than men (for example Booth, et al. 2014) and therefore the variables risk and female could be correlated with each other. Hence, it is important to check for multicollinearity and a correlation matrix was conducted (see Appendix D). There seems to be no problem with multicollinearity.
Results. Average cooperation rates across the groups can be found in table 3. Average cooperation rates are higher in the cooperative frame compared to the competitive frame (+3.13%) in the group that is not exposed to awareness. To test for significant differences a t-test was conducted because data was found to be normally distributed by a Shapiro-Wilk test. The t-test showed that the difference is insignificant. The average cooperation rate is also higher in the cooperative frame when subjects are exposed to awareness (+0.96%), but this difference is also insignificant. Therefore, it is inconclusive if the cooperative frame leads to higher cooperation levels compared to the competitive frame. By not having a strong framing effect in the control group to begin with, it is difficult to distinguish the effect of awareness on framing. The difference of cooperation levels between the frames with and without awareness decreased by 2.17% (3.13% - 0.96%). Also, awareness decreased the average cooperation level in the cooperative frame by 6.33% and 4.17% in the competitive frame. However, none of the differences are significant and thus results are inconclusive.
11
TABLE III
Average Cooperation of Player 1
No awareness Awareness Difference
Cooperative frame 44.79% 38.46% -6.33%
Competitive frame 41.67% 37.50% -4.17%
Difference +3.13% +0.96%
Notes. *p < .10, **p < 0.05, ***p < .01.
In figure 1, one can see the average level of cooperation across rounds. The average level of cooperation decreases relatively much after the first round for both the control and treatment group. In round 1, 2 and 3 cooperation rates are still higher in the cooperative frame when not exposed to awareness, but in round 4 the competitive frame has a higher average cooperation level. Results are inconclusive in regards to framing diminishing over rounds because the framing effect itself is not strongly present. When subjects are exposed to conscious awareness, the difference in average cooperation between frames in round 1 and 2 is relatively small. In round 3 the competitive frame’s average cooperation levels are higher than the cooperative frame and in round 4 the effect is reversed again. Therefore, no pattern can be detected across the difference in average cooperation across rounds.
12
FIGURE 1
Average Cooperation Levels of Player 1 Across Rounds
The vertical axis measures the average level of cooperation of player 1 in each round in percentages where 100% is everyone cooperates and 0% is no one cooperates. The horizontal axis shows rounds 1 to 4 when
subjects are not exposed to awareness and rounds 1 to 4 when subjects are exposed to awareness.
Cooperation rates of player 1 are most likely influenced by other factors than merely framing and awareness and therefore a regression is conducted to control these factors. Cooperation of player 1 is measured as the dependent variable. The variables of interest are the frame the subject is in, whether the subject is exposed to awareness or not, the joint effect of the framing and awareness (cooperative*awareness) and if the effect decreases over rounds (several methods are used). Other factors are expected to have an influence on the level of cooperation are risk, gender, the cooperation levels of the paired player and the round the subject is in.
Selten and Stoecker (1985) found that subjects tend to cooperate until the last round in a finite prisoner’s dilemma where subject are matched anonymously with another subject each round. Because in this experiment subjects are aware of the amount of rounds they will play, it is also expected that subjects in this game subjects cooperate less in the last round. Also, the framing effect is expected to have a more significant impact on average cooperation in the first rounds. Therefore, dummies are created for the first, second and last round with and without the framing effect (third round is omitted).
The level of risk-aversion of subjects can have a significant impact on results. Risk-aversion is measured by a simplified Holt & Laury test (2002) which can be found in
13
Appendix B. The more risk-averse the subject is, the lower their score is on the test will be. When a subject decides not to cooperate, they either receive €0 or €10 whereas cooperating gives them either €2.5 or €5. Therefore, not cooperating can be seen as more risky since they can lose more and it is expected that the more risk-averse a subject is, the more likely they are to cooperate. Prospect theory states that people prefer sure gains over taking a relatively small risk even when the expected profit is higher (Kahneman & Tversky, 1986). They also found that losses loom larger than gains and when they do not cooperate, a subject can lose more (€10 versus €2.5 when cooperating). Therefore, it is expected that the more risk-averse a subject is, the more they are expected to cooperate.
Men and women systematically behave differently in laboratory experiments (for example; Davis & Holt, 1993; Ball and Cech, 1996). Frank et al. (1993) found that women significantly cooperate more in a prisoner’ s dilemma and Cadsby (2007) also found that women cooperate more in public good games. Therefore, it is expected that being female will have a positive effect on cooperation levels.
Since it is second-order beliefs that drive subjects’ behaviour, it is expected that the decision of the paired player has a significant result on the behaviour of the subject. Beliefs are a weighted average of the decision of the paired player in the previous round and their own decision and are thus are adjusted each round. If the paired player decides to not cooperate, the subject is more likely to not cooperate in the next round and if the paired player decides to cooperate the subject is more likely to cooperate in the next round. A dummy variable ‘Player 2 cooperate’ is created which equals 1 if player 2 cooperated in the previous round and equals 0 if player 2 did not cooperate.
14
TABLE IV
Determinants of Cooperation Levels of Player 1
(1) (2) (3) (4) (5) Cooperative frame .056 (.138) .080 (.182) .075 (.191) .103 (.651) -.032 (.311) Awareness -.140 (.137) -.108 (.214) -.081 (.222) -.067 (.265) -.097 (.227) Cooperative frame*Awareness -.169 (.294) -.164 (.287) -.097 (.341) -.166 (.294) Risk -.230*** (.063) -.259*** (.077) -.245*** (.065) Female .232 (.161) .195 (.193) .250 (.165) Cooperative*round -.011 (.204) Round .069 (.156) First round*Cooperative frame .054 (.400) Second round*Cooperative frame .197 (.404) Last round*Cooperative frame .178 (.404) First round .776*** (.301) Second round -.155 (.303) Last round -.019 (.304) Player 2 cooperate .011 (.174) Pseudo R-square .000 .000 .053 .332 .111 N 352 352 352 264 352
Notes. Dependent variable: Cooperation level of player 1. The table represents coefficients from GLS regressions. It is chosen to use a Probit model with panel data, since this showed the lowest AIC and BIC scores. All regressions control for random effects. Standard errors are in parentheses; *p < .10, **p < 0.05, ***p < .01.
15
In table 4 one can find the different specifications of the model. In specification (1), one can see that the cooperative frame and awareness have the expected signs, but are small and insignificant. When adding the effect of the cooperative frame*awareness in specification (2), the coefficient of the cooperative frame increases and awareness decreases. The effect of the cooperative frame*awareness has the expected negative sign. This means that when a subject is in the cooperative frame and is exposed to awareness, he or she cooperates less. However, again all coefficient are insignificant. Female and risk are predicted to have a large effect on cooperation levels and are thus included in specification (3). The cooperative frame and cooperative frame*awareness coefficients change relatively little compared to
specification (1) and (2), but are still insignificant. The awareness variable’s coefficient decreases, but is also still insignificant. Risk has the expected sign and is significant. The variable female also has the expected sign, but is insignificant. In specification (4), the variables cooperative*round, round and player 2 cooperate are included. The coefficient of the round*cooperative frame is negative, which means that the higher the round a subject is in is, the less they will cooperate. However, the effect of which round the subject is in has a positive coefficient which means that subjects increase their cooperation throughout the rounds. The variable player 2 cooperate has a positive effect on cooperation levels, which means that if player 2 cooperated in the previous round, player 1 is more likely to also cooperate. Nevertheless, all three coefficients are insignificant. The other coefficients also remain insignificant except for risk. In specification (5) the effect of each separate round is measured because the effect is expected to decrease over rounds. When measuring the effect of the first round, it is not possible to include the effect of player 2’s cooperation levels on player 1. In the first round, the cooperation of player 2 cannot be measured because the first round cooperation levels has an effect on the second round cooperation rate of player 1 and therefore one of the variables has to be dropped. The effect of being in the first, second and last round and being in the cooperative frame has a positive effect on cooperation levels. However, all variables are all insignificant. Only being in the first round has a positive and significant effect on cooperation levels. Again, all other coefficients are insignificant except for risk. A regression was run with only the results of the first round, but this also did not lead to significant results (see Appendix E).
Limitations. The main limitation of the research is that the framing effect itself was
not found to be significant. Because the framing effect is not significant, it is not possible to make conclusions about how conscious awareness of framing affects the framing effect. Without having a baseline, it is not possible to correctly examine the treatment effect.
16
Why framing was not found significant could be largely due on the payoff structure. It was intended to be a prisoner’s dilemma, but a mistake was made in the payoff structure. When a subject cooperates and their paired player does not cooperate, they should have been given a negative payoff of €2.5 instead of positive. Firstly, this has a large consequence on the strategy of player 1. When it would have been a prisoner’s dilemma, not cooperating for both players strictly dominates cooperating. In this game, the Nash equilibrium is for player 1 and 2 to not cooperate when the paired player cooperates. Because in this game there are two Nash equilibriums, the strategy is unclear. The strategy over rounds of subjects is also
unclear. In a prisoner’s dilemma, when a subject cooperates and his paired player does not, the subject is more likely to not cooperate in the next round. The subject perceives the risk to be higher to cooperate since their second-order beliefs are updated about cooperation levels of their paired player. When not cooperating, the perceived risk of getting a negative payoff of €2.5 is higher while cooperating generates at least not a negative payoff. In this game, when a subject cooperated when his paired player did not cooperate, what strategy a subject will pursue in the next round is less clear. When a subject does not cooperate, their chances of getting nothing are perceived larger and cooperating always generates at least a payoff of €2.5 However, if the subject bases their decision on second-order beliefs, they might think that their paired player will cooperate, because of the previous reasoning and therefore will not cooperate. This is most likely the reason why the variable ‘Player 2 cooperate’ is small and insignificant. When the strategy is unclear, it subjects are not directed towards certain behaviour which the frame could have enhanced.
Secondly, when a subject decides not to cooperate in the intended prisoner’s dilemma, the consequences for their paired player are much higher since their payoffs decrease by €7.5 whereas in this game their paired player loses €2.5. Therefore, not cooperating in a prisoner’s dilemma is a worse violation of the norm induced by the cooperative frame than in this game. In addition, it is often proved that people do not only care about their own payoffs and have social preferences (for example Gneezy, 2005; Fehr & Schmidt, 1999). Therefore, when subjects are inequality averse, a loss of €7.5 for their paired player lowers their utility more than a loss of €2.5 when subjects can gain €5. Since the cooperative frame sets the norm to cooperate, the negative utility might be enhanced by more inequality.
Another potential reason why framing effects are not significant could be that the experiment was conducted with children. It was found that children are less influenced by framing than adult, because their qualitative reasoning is not fully developed (Reyna & Farley, 2006; Steinberg, 2011; Strough et al., 2011).
17
Another limitation is the data. First of all, there is a significant difference between the amount of females between the groups which leads to a sampling bias. Not all members of the population are equally represented. Also, since the framing effect was hypothesized to decrease over the rounds, the sample size might be too small. When, for example, the framing effect is only apparent in the first round and there are only two groups who are exposed to framing without awareness, there are only 48 observations.
Conclusion and Discussion
It was found that subjects cooperated on average 3% more in the cooperative frame compared to the competitive frame. However, this difference is small and insignificant. Therefore, the framing effect between the cooperative and competitive frame was not found. When awareness was introduced, subjects cooperated 1% more on average in the cooperative frame. Thus, there is a small decrease in the difference in the cooperation levels when
awareness is introduced, but all effects are small and insignificant. When controlling for other factors, the results were still not found significant. Therefore, it cannot be concluded that the cooperative frame leads to higher average cooperation levels and that awareness decreases the difference in cooperation levels between the frames.
One main reason why none of the framing effect is significant is because the payoff structure of the game which lead to having negative consequences on the strategy of the player. When the strategy is unclear because there are two Nash equilibriums and therefore it is difficult to examine the behaviour of subjects. The Nash equilibrium was for both players to cooperate when the other player does not cooperate. However, cooperating was seen as more risky was found to have significant impact on cooperation levels. Kahneman & Tversky (1986) also found that losses loom larger than gains and that people prefer sure gains over taking a relatively small risk even when the expected surplus is higher. This might explain why subjects see not cooperating as a riskier decision compared to cooperating even though their optimal strategy is to not cooperate when their paired player cooperate. All this leads to subjects behaving more similarly and therefore the difference might be less stark between the frames. Another potential reason why framing effects are not found to be significant is because the experiment is conducted with children. Children are less sensitive to framing effects compared to adults because their cognitive reasoning is not fully developed.
Because the framing effect is not significant, it is likewise not possible to state if the framing effect decreases over rounds. In round 1, subjects’ cooperation was 4% higher when not exposed to awareness, in round 2 cooperation was 8% higher, and in the last round
18
cooperation was 4% lower in the cooperative frame. No pattern can be detected and none of the different specifications in the regression lead to significant results.
When analysing the cooperative frame and the competitive frame separately, it was found that cooperation in both frames decreases. Subjects cooperated on average 6.33% less in the cooperative frame and 4.17% less in the competitive frame. This supports the
hypothesis that people feel duped when becoming aware of framing and this leads to an aversive state in subjects. Subjects generate feelings of distrust and maybe even anger, because they felt they were duped. Whether this behaviour is more logical is hard to say, because there was no framing effect. However, again the differences are small and all not significant. Hence, it is difficult to draw a conclusion. When controlling for risk, gender, the effect of the rounds and the level of cooperation by their paired player results were still not found significant. Therefore, there is not enough evidence to prove that the cooperative frame leads to more cooperation compared to the competitive frame, that awareness decreases cooperation in the cooperative frame and/or increases cooperation in the competitive frame, and that awareness decreases the difference in average cooperation between the cooperative and competitive frame. However, the results are too large to assign as random and therefore further research is needed.
References
Albert, D., & Steinberg, L. (2011). Judgment and decision making in adolescence. Journal of Research on Adolescence, 21(1), 211-224.
Anderson, C. A., & Morrow, M. (1995). Competitive aggression without interaction: Effects of competitive versus cooperative instructions on aggressive behavior in video games. Personality and Social Psychology Bulletin, 21(10), 1020-1030. Bacharach, M. (2006). Beyond individual choice: teams and frames in game theory. Princeton University Press.
Ball, S. B., & Cech, P. A. (1996). Subject pool choice and treatment effects in economic laboratory research. Research in Experimental Economics, 6(3), 239-292.
Booth, A., Cardona-Sosa, L., & Nolen, P. (2014). Gender differences in risk aversion: do single-sex environments affect their development?. Journal of economic behavior & organization, 99, 126-154.
Brewer, P. R., Graf, J., & Willnat, L. (2003). Priming or framing: Media influence on attitudes toward foreign countries. Gazette (Leiden, Netherlands), 65(6), 493-508. Cadsby, C. B., Hamaguchi, Y., Kawagoe, T., Maynes, E., & Song, F. (2007). Cross-national
19
gender differences in behavior in a threshold public goods game: Japan versus Canada. Journal of Economic Psychology, 28(2), 242-260.
Call, J., Carpenter, M., & Tomasello, M. (2005). Copying results and copying actions in the process of social learning: chimpanzees (Pan troglodytes) and human children (Homo sapiens). Animal cognition, 8(3), 151-163.
Camerer, C. F., & Fehr, E. (2004). Measuring social norms and preferences using
experimental games: A guide for social scientists. Foundations of human sociality: Economic experiments and ethnographic evidence from fifteen small-scale
societies, 97, 55-95.
Cappella, J.A. & Jamieson, K.H. (1997) Spiral of Cynicsm. New York: Oxford University Press.
Chun, M. M., & Jiang, Y. (2003). Implicit, long-term spatial contextual memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29(2), 224.
Cookson, R. (2000). Framing effects in public goods experiments. Experimental Economics, 3(1), 55-79.
Cummings, K. M., & Proctor, R. N. (2014). The changing public image of smoking in the United States: 1964–2014. Cancer Epidemiology and Prevention Biomarkers, 23(1), 32-36.
Davis, D. D., & Holt, C. A. (1993). Experimental economics. Princeton university press. Dufwenberg, M., Gächter, S., & Hennig-Schmidt, H. (2011). The framing of games and the psychology of play. Games and Economic Behavior, 73(2), 459-478.
Ellingsen, T., Johannesson, M., Mollerstrom, J., & Munkhammar, S. (2012). Social framing effects: Preferences or beliefs?. Games and Economic Behavior, 76(1), 117-130. Evans, J. S. B. (2002). Logic and human reasoning: An assessment of the deduction paradigm. Psychological bulletin, 128(6), 978.
Fehr, E., & Gächter, S. (2000). Cooperation and punishment in public goods experiments. American Economic Review, 90(4), 980-994.
Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and cooperation. The quarterly journal of economics, 114(3), 817-868.
Fischbacher, U., & Gachter, S. (2010). Social preferences, beliefs, and the dynamics of free riding in public goods experiments. American economic review, 100(1), 541-56. Frank, R. H., Gilovich, T., & Regan, D. T. (1993). Does studying economics inhibit cooperation?. Journal of economic perspectives, 7(2), 159-171.
20
economic games. Presented at Second International Conference for Norms, Actions, Games.
Gilad, D., & Kliger, D. (2008). Priming the risk attitudes of professionals in financial decision making. Review of Finance, 12(3), 567-586.
Gneezy, U. (2005). Deception: The role of consequences. American Economic Review, 95(1), 384-394.
Goujon, A., Didierjean, A., & Poulet, S. (2014). The emergence of explicit knowledge from implicit learning. Memory & cognition, 42(2), 225-236.
Holt, C.A. & Laury, S.K. (2002). Risk aversion and incentive effects. The American Economics Review, 92(5), 1644-1655.
Kahneman, D. & Tversky, A. (1986). Rational Choice and the Framing of Decisions. The Journal of Business, 59(4), S251-S278.
Kay, A. C., & Ross, L. (2003). The perceptual push: The interplay of implicit cues and explicit situational construals on behavioral intentions in the Prisoner’s
Dilemma. Journal of Experimental Social Psychology, 39(6), 634-643. Koslow, S. (2000). Can the truth hurt? How honest and persuasive advertising can unintentionally lead to increased consumer skepticism. Journal of consumer Affairs, 34(2), 245-267.
Liberman, V., Samuels, S. M., & Ross, L. (2004). The name of the game: Predictive power of reputations versus situational labels in determining prisoner’s dilemma game moves. Personality & Social Psychology Bulletin, 30 (9), 1175–1185.
Pietri, E. S., Johnson, I. R., Ozgumus, E., & Young, A. I. (2018). Maybe She Is Relatable: Increasing Women’s Awareness of Gender Bias Encourages Their Identification With Women Scientists. Psychology of Women Quarterly, 42(2), 192-219.
Rege, M., & Telle, K. (2004). The impact of social approval and framing on cooperation in public good situations. Journal of public Economics, 88(7), 1625-1644.
Reyna, V. F., & Farley, F. (2006). Risk and rationality in adolescent decision making: Implications for theory, practice, and public policy. Psychological science in the public interest, 7(1), 1-44.
Sagarin, B. J., Cialdini, R. B., Rice, W. E., & Serna, S. B. (2002). Dispelling the illusion of invulnerability: The motivations and mechanisms of resistance to persuasion. Journal of personality and social psychology, 83(3), 526.
Selten, R., & Stoecker, R. (1986). End behavior in sequences of finite Prisoner's Dilemma supergames A learning theory approach. Journal of Economic Behavior &
21
Organization, 7(1), 47-70.
Strough, J., Karns, T. E., & Schlosnagle, L. (2011). Decision‐making heuristics and biases across the life span. Annals of the New York Academy of Sciences, 1235(1), 57-74. Vohs, K. D., Baumeister, R. F., & Chin, J. (2007). Feeling duped: Emotional, motivational, and cognitive aspects of being exploited by others. Review of General
Psychology, 11(2), 127.
Appendix A
Instructions in Dutch & English
INSTRUCTIES – HET ‘WIE NIET WAAGT WIE NIET WINT’/’SAMEN STAAN WE STERK’ SPEL
Elke ronde speel je het ‘wie niet waagt wie niet wint’/’samen staan we sterk’ spel met een van je klasgenoten, alleen weet je niet wie diegene is. Alle resultaten zijn daarom anoniem en zullen dat ook blijven tijdens het spel, maar ook na het spel. Vergeet niet om ID nummer in te vullen waar het is aangegeven (rechtsboven in op de papiertjes). Je ID is het nummer dat je hebt gekregen van de begeleider en tevens de plek waar je zit.
ADDED IN THE TREATMENT
-Deze groep is gekozen om het ‘wie niet waagt wie niet wint’/’samen staan we sterk’ spel te spelen. De andere groep speelt het ‘samen staan we sterk’ spel. The spelletjes zijn precies hetzelfde, behalve de naam. Deze verschillende namen zijn bedoeld om jouw keuze te beïnvloeden in het samenwerken.-
Jouw taken:
Jij bent speler 1 en je speelt tegen een van je klasgenoten die speler 2 is. Jij en speler 2 moeten tegelijk kiezen of je wilt samenwerken of niet. Als jullie beiden samenwerken dan verdienen jullie allebei €5. Als speler 2 beslist om niet samen te werken dan verdien jij €2,50 en speler 2 verdient €10. Als jij niet samenwerkt en speler 2 wel samenwerkt dan verdien jij €10 en speler 2 €2,50. Werkt speler 2 ook niet samen, dan verdienen jullie allebei niets. In de tabel hieronder staan de verdienmogelijkheden. Het eerste getal in de tabel is voor speler 1 en het tweede getal voor speler 2.
22
Speler 2
Samenwerken Niet samenwerken
Speler 1 Samenwerken €5 , €5 €2,5 , €10
Niet samenwerken €10 , €2,5 €0 , €0
Volgorde van acties:
· Vul je ID in op één van de twee papiertjes van de ronde waarin je bent.
· Beslis of je wilt samenwerken of niet en vul dit in op beide papiertjes (dubbel). · Een begeleider komt langs je bureau om het papiertje zonder ID op te halen en die aan je medespeler te geven.
· Leg de antwoordenpapiertjes op de kop aan de rechterkant van de tafel.
· Wacht met het invullen van de nieuwe papiertjes van de volgende ronde tot wordt aangeven dat we naar de volgende ronde gaan.
Dit is het einde van één ronde. Aan het einde van de ronde weet je dus hoeveel je hebt verdiend in de ronde, omdat je de beslissing weet van je tegenspeler. Aan het einde van elke ronde laat je de papiertjes op de kop aan de rechterkant op je tafel liggen. Deze zullen na het experiment verzameld worden. Er zullen 4 rondes zijn en de begeleider geeft aan wanneer we naar de volgende ronde gaan.
Je kunt een prijs winnen die gebaseerd is op jouw keuze en die van je tegenspeler. Aan het einde van het experiment zal de begeleider willekeurig één ronde en één speler uitkiezen en hij/zij zal zijn verdiensten uitgekeerd krijgen van die ronde in cash.
Voordat we beginnen met de 4 rondes moet je een van de volgende opties kiezen die het beste bij je past. Als je bijvoorbeeld optie 3 kiest, dan heb je 50% kans op €8 en 50% kans op €3, vergeleken met optie 4, waarbij je 50% kans hebt op €10 en 50% kans op €1.
Opties Kans op 1 50% kans op €4 en 50% kans op €4 2 50% kans op €3 en 50% kans op €6 3 50% kans op €8 en 50% kans op €2 4 50% kans op €10 en 50% kans op €1 5 50% kans op €12 en 50% kans op €0
23
Als je nog vragen hebt dan kun je die nu stellen. We hebben 2 minuten en daarna begint het spel.
INSTRUCTIONS – HET ‘WIE NIET WAAGT WIE NIET WINT’/’SAMEN STAAN WE STERK’ SPEL
You will play the cooperation game every round against another classmate, however you do not know who this person is. Therefore, all results will be anonymous and are confidential. There will be four rounds played. Please write on every page on the top right corner your ID number you have been given at the beginning of the experiment.
ADDED IN THE TREATMENT
-This group has been chosen to play the .. game. The other group was chosen to play the .. game. The games are exactly the same except for the name. These different names are meant to influence your decision in the game.-
Your task:
You are player 1 and you are playing with one of your classmates who is player 2. You and your classmate will simultaneously decide to cooperate or not. If you both cooperate both of you receive win €5. If player 2 does not cooperate, you get €2.50 and player 2 wins €10. If you do not cooperate and player 2 cooperates, you win €10 and player 2 gets €2.5. If player 2 also does not cooperate, you both lose everything.
In the table below you can find the earning possibilities. The first number corresponds with the earnings of player 1 and the second with the earnings of player 2.
Player 2
Cooperate Not Cooperate
Player 1 Cooperate €5 , €5 €10 , €2.5
Not Cooperate €10 , €2.50 €0 , €0
Sequence of events:
· Fill in which player you are on the paper of the round we are in (double)
24
· An experimenter will come at your desk to collect one of the paper and will give this to your paired player. At the same time, an experimenter will give you the papers for the new round.
· Please wait with filling in the new papers until it is announced that we are starting the new round.
The above sequence of events completes one round. At the end of the round you will thus know how much you won because you know the decision of your paired player. At the end of the round, turn around the papers and leave them on the corner of the table. There will be 4 rounds. You will be paired with a different participant every round. At the end of the round, please wait until we announce that we go to the next round.
You can receive a price based on your decisions and the decisions of your paired player. At the end of the experiment, one round and will be randomly selected and everyone will be paid accordingly.
Before the we start with the 4 rounds, please choose one of the following options. For example, if you choose option 3 you have a 50% chance on earning €8 and a 50% chance on €2, compared to option 4, where you have 50% chance on €10 and 50% chance on €1.
Option Probability
1 50% chance on €4 and 50% chance on €4 2 50% chance on €3 and 50% chance on €6 3 50% chance on €8 and 50% chance on €2 4 50% chance on €10 and50% chance on €1 5 50% chance on €12 and 50% chance on €0
If you have questions, please ask them now. We have 2 minutes to answer your questions and then we will begin.
25
Appendix B TABLE V Risk-aversion Test
Option Payoffs
1 50% chance on €4 and 50% chance on €4 2 50% chance on €6 and 50% chance on €3 3 50% chance on €8 and 50% chance on €2 4 50% chance on €10 and 50% chance on €1 5 50% chance on €12 and 50% chance on €0
Appendix C TABLE VI Randomization Test
Group Frequency Female Risk
Cooperative frame = 0, Awareness = 0 27.27% 41.67% 2.656 Cooperative frame = 1, Awareness = 0 27.27% 66.67% 2.875 Cooperative frame = 0, Awareness = 1 15.91% 26.09% 2.643 Cooperative frame = 1, Awareness = 1 29.55% 61.54% 2.558
26 Appendix D TABLE VII Correlation Matrix Player1 Cooperative frame Awareness Cooperative frame*Awareness Risk- aversion Female Player1 1.000 Cooperative frame .0138 1.000 Awareness -.0517 .1508 1.000 Interaction -.0322 .5645 .7094 1.000 Risk-aversion -.2410 .0266 -.0644 -.0574 1.000 Female .1776 .2952 -.0664 .1348 -.4095 1.000 Appendix E TABLE VIII
Determinants of Cooperation levels of Player 1
(1) Cooperative frame .083 (.407) Awareness -.182 (.443) Cooperative frame*awareness -.365 (.584) Risk -.208 (.124) Female .427 (.327) Pseudo R quare .084 N 88
Notes. Dependent variable: Cooperation level of player 1. The table represents coefficients from GLS regressions. It is chosen to use a Probit model, since this showed the lowest AIC and BIC scores. Standard
27
Appendix F Statement of Originality
This document is written by Sabine Tijdeman who declares to take full responsibility for the contents of this document
