Are you the selfish type?
Population differences in trust and reciprocity:
It is time to prime to cooperative outcomes.
Master Thesis by: Coen Akkerman Student number: 6044298
Date: May 28th, 2015 Supervisor: Prof. dr. J.H. Sonnemans
Faculty of Economics and Business
Specialization track: Behavioral Economics and Game Theory University of Amsterdam
2 Abstract:
Trust and reciprocity are crucial to build and maintain connections. These connections cause higher welfare and happiness than when people only live out of self-interest. However, not everyone understands this. Especially when there is a single interaction with unknowns, trust and reciprocity are not always that high. Several experiments examined and confirm this (Berg et al, 1995; Bolle, 1998; Burks et al, 2003). There are groups in the population that lack this trust and reciprocal behavior. There are experiments that show that economists are the ones that trust less and behave selfish (Marwell and Ames, 1981; Carter and Irons, 1991; Frank et al, 1993). However, experiments of Yezer et al. (1996) and Laband and Beil (1999) contradict these findings. In this thesis I examine whether there is a possibility to prime population groups that lack trust and reciprocal behavior to more cooperative outcomes. This thesis shows the levels of trust and reciprocity today by the use of a simple trust-game. Using a questionnaire, I search for personal characteristics and general information of the participants to examine who are the ones that lack trust and reciprocal behavior. By the use of a scrambled sentence test before the trust-game, participants are primed in a direction to influence them unconsciously. I perform this test because recognizing the ones that lack trust and reciprocity is not enough to increase these levels. The thesis shows that the levels of trust and reciprocity today are still not optimal. It shows that economic educated participants do not show less reciprocity and trust than social educated participants. It shows that females trust less than males and that participants below the age of 23 behave less reciprocal than older participants. The thesis also shows that the scrambled sentence test was successful in priming participants to more reciprocal behavior. However, it failed in priming participants to more trusting behavior. I mention that further research is needed to examine the real power of scrambled sentence tests as a priming technique. Priming people to more reciprocal behavior is great to maintain connections but before connections can be maintained, they should have been built before. When the trust-level cannot be increased. the increase in reciprocity-level is less useful.
3
Table of Contents.
1. Introduction ... 4
2. Literature ... 5
2.1. What is trust and what does it contribute to the world? ... 5
2.2. Trust and reciprocity in past experiments. ... 7
2.3. The levels of trust and reciprocity in the population. ... 9
2.4. How to increase the levels of trust and reciprocity: The priming effect. ... 10
2.4.1. Priming with words: The scrambled sentence test. ... 12
3. Experimental design ... 13
3.1.1. Phase 2: The trust game. ... 13
3.1.2. Phase 1: The scrambled sentence test. ... 14
3.1.3. The questionnaire. ... 15
3.2. Hypotheses ... 17
3.2.1. Hypotheses on the priming conditions. ... 17
3.2.2. Hypotheses on the trust-game results. ... 17
3.2.3. Hypotheses on the differences between economists and social scientists. ... 18
3.2.4. Hypotheses on the differences of political affiliation and gender. ... 19
3.2.5. Hypothesis on the differences of age. ... 21
3.2.6. Hypotheses on the questionnaire equations... 21
3.3. Data collection ... 22
4. Results ... 22
4.1. Hypothesis test. ... 23
4.1.1. The priming results ... 23
4.1.2. The trust-game results. ... 26
4.1.3. The results of economic educated participants and social educated participants. ... 26
4.1.4. The results on political affiliation and gender... 29
4.1.5. The results on age and the questionnaire equations. ... 32
5. Discussion ... 36
5.1. Limitations and Improvements. ... 37
6. Application ... 38
7. Future study ... 39
8. Conclusion ... 40
Bibliography. ... 42
Appendix A - Result tables ... 45
4
1. Introduction
Trust and reciprocity. They are well-known phenomena. When I trust you, you feel happy because I trust you. When you behave reciprocal, I feel happy because you confirm why I trusted you. Having trust builds connections and behaving reciprocal maintains these connections. But there are some people who do not behave reciprocal or are not even trustworthy when an agreement is made. These people achieved that the people that trusted them became more wary and as a result trust less. But trust is important: trust can lead to a willingness to get involved in communities, higher rates of economic growth and to satisfaction with government performance (Putnam, 1993, 1995; Fukayama, 1995; Knack and Keefer, 1997). Countries with more trusting people have a better functioning government, more open markets and less corruption (Uslaner, 2008).
Previous experiments show that the level of trust and reciprocity is not always high. Several studies examined who are the ones that lack this trust and reciprocal behavior. For example, some studies conclude that economists are typically not cooperative (Marwell and Ames, 1981; Carter and Irons, 1991; Frank et al, 1993). There are also studies that show the opposite where economists are not guilty to non cooperative behavior, at least in 'real-world' situations (Yezer et al, 1996; Laband and Beil, 1999).
According to these findings, it is interesting to find out what the trust and reciprocity level is today and who are the ones in the population that lack trust and reciprocal behavior. Due to the importance of trust, it is not sufficient to only recognize the ones that lack trust and reciprocal behavior. To improve this behavior, it is important and interesting to check whether it is possible to prime these groups to more trust and reciprocity. Out of these points rises the following central question:
''Is there a possibility to prime population groups that lack trust and reciprocal behavior to more cooperative outcomes?''
This central question is very interesting because trust and reciprocity are crucial phenomena to build and maintain connections. Literature shows that the level of trust and reciprocity is not optimal yet (Berg et al, 1995; Bolle, 1998; Burks et al, 2003). However, these studies have been conducted long ago. What is the level of trust and reciprocity today? Who are the ones in the population that lack trust and reciprocal behavior? Can we prime these groups to more cooperative behavior?
To find out whether there is a possibility to prime population groups that lack trust and reciprocal behavior to more cooperative outcomes I run an online experiment. Before I can draw conclusions about priming, I need to know what the level of trust and reciprocity is today. Therefore, one part of the experiment is a trust-game. The trust-game is based on the experiment of Berg et al. (1995) and works as follows: you as subject are randomly selected to be one of the 2 possible roles: a sender or a
5
receiver. The sender is endowed with $10. He has a choice to send (a part of) the $10 to a receiver he is paired with. If the sender decides to send money to his receiver he is paired with, this amount gets tripled. The receiver has the choice to send a part of the tripled amount he just received back to the sender. The results of the trust-game show the levels of trust and reciprocity and table 1 and 2 in the literature section provide more information about this trust-game. It is also important to know which population groups lack trust and reciprocal behavior. Therefore, one part of the experiment contains a questionnaire to find general information and characteristics of the participants. To find out whether it is possible to prime participants in a certain direction, a scrambled sentence test is constructed before the trust-game. In the experimental design, I explain the test in detail. The experiment concludes whether it is possible to prime population groups that lack trust and reciprocal behavior to more cooperative outcomes.
In chapter 2 I give a review of previous literature about trust, trust-games, population parts that lack trust and reciprocal behavior, priming and the priming technique that makes use of the scrambled sentence test. In chapter 3 I give the experimental design of this thesis. The experimental design contains all information about the set up of the online experiment and how I use the experiment to answer the central question. It also gives the hypotheses, how I collect my data and how I analyze my data. In chapter 4 I show the results of the experiment and in chapter 5 I discuss these results. In chapter 6 I show some applications on the experiment and in chapter 7 I discuss whether further research can improve the results. Finally, I give my conclusions in chapter 8.
2. Literature
2.1. What is trust and what does it contribute to the world?
The concept of trust is well-known. Having trust in someone means that you believe in the honesty of this person. You believe that someone does what he says. Also without an agreement you can have trust in somebody: people that invest in stocks mostly have trust in the companies where they have these stocks. They trust that the companies do well so their stocks become worth more. Even when you give away some money to a person who wants to buy a winning lottery ticket with your money, you will hope that he is reciprocal and shares some of his millions he just won with you. When people trust each other and behave trustworthy or reciprocal, people feel great and happy: promises will be kept and connections will be made. As stated in the introduction: trust is mostly conceived as a "rational" response to trustworthy behavior by others (Uslaner, 2008).
Trust in people is not the only form of trust. Trust in the economy is also a well-known concept.
When there is trust in the economy, consumers tend to consume more. This will raise producers' profits which makes them able to increase their investments. These increases in consuming and investing will boost the economy.
6
Back to trust among people. What does trusting each other contribute to the world? We might expect that trust and economic performance are positively correlated. According to Knack and Keefer (1997), Trust and civic cooperation are associated with stronger economic performance. Individuals in higher-trust societies spend less to protect themselves from being exploited in economic transactions. Low trust can also discourage innovation. If entrepreneurs must devote more time to monitoring possible malfeasance by partners, employees, and suppliers, they have less time to devote to innovation in new products or processes. According to Putnam (1993), a group whose members manifest trustworthiness and place extensive trust in one another will be able to accomplish much more than a comparable group lacking that trustworthiness and trust.
According to Uslaner (2008), trusting people are more likely to volunteer their time, to give to charity, to be tolerant of others, and to support policies that both promote economic growth (open markets) and that provide support for the less fortunate. Countries with more trusters have better functioning government, more redistributive policies, more open markets, and less corruption. What distinguishes countries that are trusting from those that are not is the level of economic equality. Trust brings us all sorts of good things– from a willingness to get involved in our communities to higher rates of economic growth and, ultimately, to satisfaction with government performance and to making daily life more pleasant (Putnam, 1993, 1995; Fukayama, 1995; Knack and Keefer, 1997; Uslaner, 2008). People who believe that others can be trusted have an optimistic view of the world. They believe that things will get better and that they can make the world better by their own actions (Rosenberg, 1956; Lane, 1959, 163-166).
As stated before, trusting people are more likely to volunteer and give to charity. In the 1996 Giving and Volunteering survey trusters take up almost twice as many volunteering opportunities (1.606 versus .856) and charitable options (2.015 versus 1.221) as non-trusters. Volunteers who take part in eight or more organizations are twice as likely to trust others (by 64 to 34 percent), while people who give to seven or more different types of charities are more than twice as likely to put their faith in strangers (by 68 to 30 percent) (Uslaner, 2008).
Trust also has consequences beyond civic engagement. Trusters indeed do reach out to people who are different from themselves. They are more tolerant of gays and lesbians, have more positive views of blacks and immigrants, and are more willing to open markets. Trusters also support policies that remedy the wrongs in our social system–anti-discrimination laws and policies that make it easier for minorities to take their full place in society: military service and adoption for gays, affirmative action for African-Americans (Uslaner, 2008).
Cross-nationally, high degrees of trust lead countries to spend more on redistributing wealth from the rich to the poor. High trusting societies have greater transfer payments, spend more on education, and have larger public sectors more generally. They also have more open markets–and “better” government more generally: less corruption, less red tape in bureaucracy, and more efficient judicial systems (Uslaner, 2008). It is easy to have a well-functioning legal system, one where people have
7
faith in the law, when there is less for enforcement authorities to do. Trust seems to bring a faith that the legal system is fair–and a compliance with the law (Uslaner, 2002, chs. 7-8; cf. LaPorta et al., 1998; Rothstein, in press; Uslaner, 2008).
So according to the literature, trusting each other is crucial because of many things. We also read that trust needs to be answered with trustworthiness or reciprocal behavior to maintain the connections that are built.
2.2. Trust and reciprocity in past experiments.
How developed are the levels of trust and reciprocity in our world? Several experiments were done in the past to examine these levels with people that don't know each other and without reputation building. One of the experimental games used by experimenters to examine the levels of trust and reciprocity is the trust game as stated in the introduction. Results of these experiments show that the levels of trust and reciprocity are not always high. This is not strange, because in these experiments you interact with an unknown person and when following economic game theory, a strictly increasing indirect utility function is assumed, so a selfish utility maximizing receiver has the dominant strategy where he keeps all the money. If this is anticipated by the sender, his selfish utility maximizing choice is to send nothing. Table 1 shows the trust-game as a sequential game. The only Nash-equilibrium is the outcome where a sender keeps all the money.
So, when all subjects act selfish and don't trust each other, there will be no Pareto optimal results: trusting each other and behave reciprocal gives a higher outcome for both sender and receiver. As stated before, the levels of trust and reciprocity in these experiments are not always high, although these experiments also show that most subjects do not act as predicted by theory. In a lot of studies, there is a positive level of trust and reciprocity but optimal results where everybody trusted each other and behaves reciprocal were still not reached (Berg et al, 1995; Bolle, 1998; Burks et al, 2003).
In Burks et al. (2003) the average sent by the senders was $6.50 out of $10 and the average that the receivers sent back was 43.6% of their received tripled total. For both players this is a better result than the Nash-equilibrium. According to Bolle (1998), the level of trust and reciprocity is also greater than in the Nash-equilibrium. A slightly other approach of the trust game in the introduction of this thesis shows us that 76.2% of the participants decided to trust and that 49.3% of a doubled total is sent back. Results of Berg et al. (1995) show us that the average that senders sent was $5.16 out of $10 while the average amount that was sent back was $4.66 which is a percentage of 30.1.
So subjects do not all act selfish, but they also do not completely trust and behave reciprocal. As stated before, this last point is not that strange: subjects interact with unknown persons and there is no space for reputation building. So why is there a positive level of trust and reciprocity anyway? One explanation of the data is that senders were willing to place a trust, by risking some amount of money, in the belief that there would be reciprocity; the receivers who reciprocated kept this trust. The
8
receivers that did not reciprocated may have been acting out of self interest, or alternatively they may not have interpreted the senders' decisions as placing a trust and thus reciprocation was not an issue (Berg et al., 1995).
Table 1. The trust-game as a sequential game.
Table 2. Examples of possible scenarios in the trust-game of table 1.
Sent Amount -> Amount sent times 3 Amount received by receiver
Type receiver Final amounts [sender, receiver] Sender €0 times 3 -> €0 1,2,3,41-> [€10, €0] €3 times 3 -> €9 1 -> [€7, €9] €5 times 3 -> €15 2 -> [€20, €0] €8 times 3 -> €24 3 -> [€14, €12] €10 times 3 -> €30 4 -> [€15, €15] Type receivers:
1. Complete selfish receiver. 2. Complete altruistic receiver.
3. Receiver that shares amount received. 4. Receiver that shares final amounts.
1Because the sender keeps all the money in this case, it does not matter which kind of receiver is matched with this sender because the receiver will not get any amount.
9 2.3. The levels of trust and reciprocity in the population.
What are the levels of trust and reciprocity in the population? Who are the ones that act selfish? As we can read in the section 2.2 of this thesis, experimental studies show that the levels of trust and reciprocity are not always that high. Several studies were done in the past to examine who or what causes this lack of trust and reciprocity. I conclude that when following economic game theory, the level of trust and reciprocity is minimal. This is one of the reasons many studies (Marwell and Ames, 1981; Carter and Irons, 1991; Frank et al., 1993) examined and conclude that economists are the ones that are selfish. They argue that the economists should be familiar with the Nash-equilibrium and probably also reason like this in experiments.
According to Marwell and Ames (1981), who did an experiment on the provision of public goods, economics graduate students were much more likely to free ride than any of the other groups of subjects. In an ultimatum bargaining game of Carter and Irons (1991) where subjects needed to divide $10, economists on average accepted a minimum of $1.70 and proposed to keep $6.15. Corresponding figures for non-economists were $2.44 and $5.44. Economists' behavior is on average closer to that predicted by the economic model: economists accept less and propose to keep more (Carter and Irons, 1991).
Frank et al. (1993) faulted economists on several grounds. First, they reported that college economics professors were much less likely than college professors from a variety of other disciplines to contribute to private charities. Second, Frank et al. analyzed the behavior of undergraduate students placed in an experimental prisoner’s dilemma setting. They found that the defection rate was significantly higher among economics majors than among non-majors, even controlling for the possible influences of sex, age, and experimental conditions. Finally, Frank et al. analyzed student responses to two ethical dilemmas posed to students in two introductory microeconomics courses and students in an introductory astronomy course at Cornell University. The ethical dilemmas were posed to the students at the beginning of the fall semester and then again at the end of that semester. At issue was the extent to which the students became “less honest” over the course of the semester, with respect to their response to the two ethical dilemmas. Frank et al. report that the economics students generally showed more movement toward less honest responses than the astronomy students showed, and that the students who were taught microeconomic principles with a game theory emphasis showed greater movement toward less honest responses than did the students who were taught microeconomic principles without the game theory emphasis (Laband and Beil, 1999).
There are also studies that showed the opposite where economists are not the ones that act selfish, at least in 'real-world' situations (Yezer et al., 1996; Laband and Beil, 1999). Yezer et al. (1996) raise a host of cautions regarding the Frank et al. paper. Rather than relying on answers to hypothetical situations, they designed a “lost-letter” experiment, in which envelopes containing $10 in cash were left in classrooms, one envelope per classroom. The question of interest was whether
10
advanced undergraduate students majoring in economics would exhibit differential return behavior (cooperation) as compared to advanced undergraduate students from other disciplines. Contrary to the previous studies mentioned, Yezer et al. (1996: 181) observed that: Of the 32 letters left in economics classes, 18 (56 percent) were returned. Only 10 of the 32 (31 percent) letters left in non-economics classes were returned. This difference is large in magnitude and statistically significant at the 10 percent level. Contrary to the expectation one would have formed based on the sort of evidence presented by Frank et al. (1993), this experimental evidence indicates that economics students are far more cooperative than students studying other disciplines (Laband and Beil, 1999).
According to Laband and Beil (1999), who performed an experiment on paying professional Association dues, professional economists are no less cooperative than professional political scientists when it comes to paying their professional Association dues. However, sociologists exhibit a significant lower degree of cooperative behavior with respect to payment of their Association dues than either economists or political scientists. As stated above, the discussion whether economists are the ones that lack trust and reciprocal behavior is still unsolved.
Several studies examined the difference in trust and reciprocal behavior between people with a different political affiliation. For example, left-wing voters are more cooperative (Lunati, 1992), but also more risk-averse (Carlsson et al., 2005).
Several studies examined the difference between men and women in trust and reciprocal behavior. For example, females trust less; they are more risk-averse (Burks et al., 2003; Carlsson et al., 2005). Frank et al. (1993) found that there is a sharp tendency for males to behave less cooperatively in their experiments. In an experiment of List (2004) where subjects could do a donation, the proportion of females who donated was significantly larger than the proportion of male donators.
In the same experiment the difference in age was examined: mature individuals tend to provide greater voluntary contributions to public goods than younger cohorts (List, 2004). Frank et al. (1993) showed evidence that older students generally give greater weight to social concerns like the ones that arise in free-rider experiments.
2.4. How to increase the levels of trust and reciprocity: The priming effect.
As we can read in the abstract of this thesis, having trust and behave reciprocal and trustworthy is crucial to build and maintain connections. Is there a way to increase these levels of trust and reciprocity? According to the literature, one solution could be priming. According to the magazine Psychology Today, priming is a non-conscious form of human memory concerned with perceptual identification of words and objects. It refers to activating particular representations or associations in memory just before carrying out an action or task (https://www.psychologytoday.com/basics/priming).
11
The concept of priming looks like a good solution for the ones in the population that lack trust or reciprocal behavior, but does the concept really work? Several experimental studies show that the priming effect works (Fazio et al., 1986; Bargh et al., 1996; Kusev et al., 2002).
The findings of numerous experiments indicate that lexical decisions for concepts associated with the prime were facilitated by exposure to the prime (e.g., Meyer & Schvaneveldt, 1971; Neely, 1976, 1977). Thus, presentation of ‘‘doctor’’ as a prime facilitated identification of ‘‘nurse’’ as a word. The findings suggested that concepts associated with the prime are automatically activated from memory on its presentation (Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977) and, hence, facilitate responding to semantically related target words (Fazio, 2001).
Fazio et al. (1986) reasoned that a similar priming effect should be apparent for attitudes. Presentation of the attitude object as a prime should activate any associated evaluations and, hence, facilitate a related judgement. The paradigm that Fazio et al. (1986) developed, and that has been employed in much of the subsequent research, involved the participants’ performance on an adjective connotation task. The target word presented on each trial was an evaluative adjective and the participants were instructed to indicate whether the word meant ‘‘good’’ or ‘‘bad’’ as quickly as possible. The focus was on the latency with which this judgement was made and, in particular, the extent to which it was facilitated by the presentation of an attitude object as a prime. In three experiments, Fazio et al. (1986) found evidence of automatic attitude activation. It is important to keep in mind that the priming procedure employed by Fazio et al. (1986) did not require participants to consciously evaluate the primes (Fazio, 2001).
The priming effect has been observed in numerous subsequent experiments. It has proven to be a robust and replicable phenomenon apparent in experiments using a variety of priming stimuli, target stimuli, and specific task requirements. The effect has been found when the words serving as primes are presented subliminally (e.g., Greenwald, Klinger, & Liu, 1989; Greenwald, Draine, & Abrams, 1996; Wittenbrink, Judd, & Park, 1997), further attesting to its characterization as an automatic process. Moreover, evidence of automatic attitude activation has been demonstrated in later experiments, not only with the names of familiar attitude objects serving as primes (e.g., Bargh, Chaiken, Govender, & Pratto, 1992), but also with so-called ‘‘Turkish words’’ whose translation had been learned in an earlier phase of the experiment (De Houwer et al., 1998), with black-and-white line drawings of objects (Giner-Sorolla, Garcia, & Bargh, 1999), and with high-resolution colour images of the objects (e.g., Fazio, 1993; Hermans et al., 1994; Fazio, Jackson, Dunton, & Williams, 1995) (Fazio, 2001).
To apply these results on decisions of trusting and reciprocal behavior, we need to know whether priming works in decision-making experiments. Kusev et al. (2012) studied risky preferences using a semantic-priming paradigm where accessibility is manipulated independently of beliefs about the frequencies of risky events. They compared the risks taken for precautionary decisions primed by relevant information (enhancing accessibility to relevant events) with those taken for unprimed
12
decisions and decisions primed by irrelevant information. They found that both priming and the subjective frequency of beliefs independently influence decision making (Kusev et al., 2012).
2.4.1. Priming with words: The scrambled sentence test.
When reading the description of priming in the previous section, priming with words could be a solution to increase the levels of trust and reciprocity. Several experiments were done to find out whether priming works when subjects have to do a scrambled sentence test. A scrambled sentence test is a test where the subject needs to make sentences with words that are in a scrambled order. The prime here is that in each/or a part of the sentences are words that are primes. These primes will be of effect in a later task. For example: { be, do, greedy, not } results in { do not be greedy }. The word greedy here could be the intention of the experimenter to affect someone later on.
Bargh et al.(1996) tested this as follows: subjects were told that they had to do a task due to their language ability. This task was the scrambled sentence test. After finishing the test the subject was told that he should hand in his test to the experimenter who was in another room. The scrambled sentence test existed in three forms and each subject could get one of these three forms. Test 1 was a test with priming words that are related to rudeness, test 2 to politeness and test 3 was a control test with neutral words. The prime does only work when a subject does not recognize these primes, so the primes cannot be too obvious. That is why not all the sentences consist of primes. Bargh et al. chose 15 out of 30 sentences for priming. The experiment is designed in a way that when the subject wants to hand in his test, the experimenter is explaining the test to another person. This person is pretending to be a subject that does not understand the test. The results of Bargh et al. show that participants in the rude priming condition interrupted significantly faster (M = 326 s) than did participants in the neutral (M=519 s)or polite (M =558 s) priming conditions (Bargh et al, 1996).
This is just one example of a scrambled sentence test that successfully primed participants in an experimental setting. There are more experiments that show that priming with a scrambled sentence test was successful: Srull and Wyer (1979) show a scrambled sentence test making people more kind, Macrae and Johnston (1998) showed a test making people more helpful, Bargh et al. (2001) show another test which makes people more cooperative. Randolph-Seng and Nielsen (2007) show a test that religious words make more honest and Chartrand et al. (2008) show a test making people more thrifty.
Because the levels of trust and reciprocity are not always that high, solutions should be found. When recognizing the ones that do not trust and do not behave reciprocal, there is a possibility to prime these people to more cooperative outcomes, at least in an experimental setting.
13
3. Experimental design
In the literature section is stated that trust and reciprocity are crucial because of many things. I look at which part of the population does not trust much, does not behave reciprocal and I look whether priming the population to more cooperative outcomes is possible. To examine these subjects I run an online experiment. The experiment consists of three phases which will be explained separately below. Phase 1 is a scrambled sentence test, phase 2 contains a trust game as stated in the literature section and phase 3 contains a questionnaire.
There are instructions before every phase. The experiment is anonymous and double-blind: me as experimenter does not have any idea about who give which answers. Important parts of the general instructions are:
1. Participants have to do the experiment alone. When they have questions, they can ask me via phone or Facebook.
2. Because the experiment is running for a longer period, I ask the participants to not speak about decisions made until the experiment is no longer running.
3. The answers they give are permanent, they can't go back after submitting a phase. The complete general instructions are stated in appendix B.
Before I can draw conclusions about who are the ones that lack trust and reciprocal behavior and whether it is possible to prime the population, I first explain how the trust game works.
3.1.1. Phase 2: The trust game.
In phase 2 of the experiment I conduct a trust game as in the literature section. This means that as a subject you are randomly selected to be one of the 2 possible roles: a sender or a receiver. The sender is endowed with €10. He has a choice to send (a part of) the €10 to a receiver he is paired with. The pairs are randomly formed. If the sender decides to send money to his receiver he is paired with, this amount gets tripled. The receiver has the choice to send a part of the tripled amount he just received back to the sender.
The specific instructions of both roles mention that all players will not know who their match is. They play the game as if their match is an unknown person just as in previous experiments (Berg et al, 1995; Bolle, 1998; Burks et al, 2003). Because when a participant knows his match, his decisions are probably influenced by the relationship he has with his match. This could influence the results. Because the receiver does not know how much he gets from his randomly matched sender, he has to complete an entire strategy with a response for each possible amount sent. This is mentioned in both sender and receiver instructions. The sender definitely knows now that his sent amount will not be recognized by his matched receiver. In both sender and receiver instructions a general example of the
14
game is included. This example is included to make the game clear. The example only provides information of the game in general (amounts in letters). Because of this generality, participants will not be influenced by the example. Finally, in the specific instructions of both roles is mentioned that participants can earn money with the experiment: there will be two randomly matched pairs chosen for payment. The specific instructions and answer sheets for both sender and receiver are included in Dutch in appendix B.
According to previous experiments (Berg et al, 1995; Bolle, 1998; Burks et al, 2003), the levels of trust and reciprocity are not very high. However, much larger than in the Nash-equilibrium were the levels of trust and reciprocity are minimal. The trust game in my experiment examines whether the trust and reciprocity levels today have changed from the existing literature. The results of the senders give conclusions about the trust-level and the results of the receivers give conclusions about the reciprocity-level. In the questionnaire I ask, on a 5-point Lickert scale, how well the participant understood the trust game. Every participant who answers 1 (none) is removed from the results.
3.1.2. Phase 1: The scrambled sentence test.
According to the literature, the scrambled sentence test is a good test to prime the participants in a following task (Srull and Wyer, 1979; Bargh et al., 1996, 2001; Macrae and Johnston, 1998; Randolph-Seng and Nielsen, 2007; Chartrand et al., 2008). In Phase 1 I conduct a scrambled sentence test to see if I can prime the participants in a certain direction in the trust game of phase 2.
If the scrambled sentence test can prime the participants to more cooperative outcomes, the test is a solution for the population parts that lack trust and reciprocity. According to the literature, several scrambled sentence tests were done in the past. To find a good number of sentences and primes I studied these past experiments. Srull and Wyer (1979) constructed one test with 20% primes out of 30 or 60 sentences and another test with 80% primes out of 30 or 60 sentences. Bargh et al. (1996) constructed 30 sentences with 15 primes. Bargh et al. (2001) constructed 30 sentences with 10 primes. Macrae and Johnston (1998) and Randolph-Seng and Nielsen (2007) constructed 15 sentences with 10 primes. As read above, all these experiments had a different approach. However, these experiments show successful results. According to these studies, the most important point is that the primes are not consciously recognized by the participant. In phase 3 of the experiment, I ask questions whether they have consciously recognized the primes or not. There is more information about these questions in the design of phase 3.
I choose for 20 sentences with 15 primes and use the set up just as Srull and Wyer (1979) where the participants were thinking that the test was a language ability test. Each scrambled sentence consists of five words. The participant has to make a sentence with four words.
15
I have made three different scrambled sentence tests where every participant is randomly assigned to. One is intended to prime the construct more cooperative (the positive priming condition), one is intended to prime the construct less cooperative/selfish (the negative priming condition) and one is intended as a neutral/control test without these primes (the neutral priming condition). For the positive priming condition, the priming words are: respect, reasonable, share, ours, donates, civilized,
friendly, dependable, give, divide, tolerant, honest, cede, cooperatively and fair. For the negative
priming condition, the priming words are: cunning, retain, bank, money, uncivilized, greedy,
unreliable, rapacious, take, egoistic, moolah, deceptive, myself, keep and selfishly. For the neutral
priming condition, the words are: salad, umbrella, pencil, pizza, duck, green, paper, mountain, lead,
zebra, chocolate, cow, door, milk and teacher.
In the instructions of the test I mention that the participant has to make a sentence of four words out of five, that there are more possibilities for each sentence and that the participant has six
minutes time for the test. In each of the three tests I construct the same neutral example ({eats, monkey, the, plane, bananas} results in {the monkey eats bananas})
to make sure the participant understands the test. In appendix B, all three tests are included in Dutch.
3.1.3. The questionnaire.
To examine which part of the population lacks trust and reciprocity, whether participants have consciously recognized the primes in phase 1 and whether participants understood the trust game of phase 2, I conduct a questionnaire.
Many studies concluded that economists are a group what behaves more selfish and trust less than other groups (Marwell and Ames, 1981; Carter and Irons, 1991; Frank et al., 1993). Yezer et al. (1996) and Laband and Beil (1999) contradict this in their studies. In the questionnaire, I ask about the type of education. Combined with the results of the trust game in phase 2 I draw conclusions about this unsolved discussion.
According to Frank et al. (1993), college economics professors were much less likely than college professors from a variety of other disciplines to contribute to private charities. In the questionnaire, I ask participants on a 5-point Lickert scale about their willingness to give to charities. Combined with the type of education I'm able to discuss the results of Frank et al. (2003).
Several studies show that left-wing voters are more cooperative (Lunati, 1992), but also more risk-averse (Carlsson et al., 2005). In the questionnaire, I ask participants on a 5-point Lickert scale about their political affiliation. Combined with the results of the trust game I discuss the previous findings in the result section.
The difference between males and females in trust and reciprocal behavior is also examined in experiments. For example, females trust less; they are more risk-averse (Burks et al., 2003; Carlsson et
16
al., 2005). Frank et al. (1993) found that there is a sharp tendency for males to behave less cooperative in their experiments. In an experiment of List (2004) where subjects could do a donation, the proportion of females who donated was significantly larger than the proportion of male donators. The results of the trust game and the results of the question about giving to charities give information whether there are differences between males and females according to risk-averseness, cooperative behavior and donating.
List (2004) found that mature individuals tend to provide greater voluntary contributions to public goods than younger cohorts. Frank et al. (1993) showed evidence that older students generally give greater weight to social concerns like the ones that arise in free-rider experiments. In the questionnaire I ask participants their age. Combining these results with the receiver results in the trust game I draw conclusions whether age does matter looking at reciprocal behavior.
According to the studies on the scrambled sentence tests, it is very important that a participant does not consciously recognize the primes. That is why, just like all the experimenters of the literature on scrambled sentence tests in this thesis, I ask in my questionnaire whether the participant thinks the scrambled sentence test and the trust game were related and whether there was anything in the scrambled sentence test that influenced their decisions in the trust game. I also ask whether a participant has ever seen or done a scrambled sentence test for an experiment before, whether the participant remembers any of the words from the test and whether the participant thinks that any of the words in the test was unusual or distinctive and if so, why? Every participant who recognizes the primes will be removed from the results on the priming conditions.
The questionnaire also looks for new insights. On a 5-point Lickert scale I ask whether the participant thinks it is more important to make a lot of money above caring for their fellowman and helping to develop the society. It could be that participants who answer 1 or 2 (make a lot of money is more important) behave less reciprocal in the trust game of phase 2 than participants who answer 4 or 5 (care for your fellowman and helping to develop the society is more important).
Another new insight that the questionnaire examines is whether this difference also exists between participants with an economic education and participants with a social education. It could be that there are significant differences between the answers of participants with an economic education and participants with a social education.
Finally, the last new insight that the questionnaire examines, is whether participants who will choose spending all of their money in their own country above giving to charities and development in the third world will behave less reciprocal than participants who think vice versa.
17 3.2. Hypotheses
This chapter shows all hypotheses of the experiment. Section 4 shows the results of these hypotheses.
3.2.1. Hypotheses on the priming conditions. Hypothesis 1a. The effect of priming on trust.
As stated previously, the literature on priming with scrambled sentence tests shows that these tests can successfully prime participants in a following task. The null hypothesis is that there exists no difference in trust-level between a participant who did the negative priming condition and a participant who did the neutral priming condition. However, because of the literature on priming with scrambled sentence tests I expect that the negative priming condition shows less trusting in the trust game then the neutral priming condition.
The next null hypothesis is that there exists no difference in trust-level between a participant who did the positive priming condition and a participant who did the neutral priming condition. However, because of the same literature on priming with scrambled sentence tests I expect that the positive priming condition shows more trusting in the trust game then the neutral priming condition.
Hypothesis 1b. The effect of priming on reciprocity.
As stated previously, the literature on priming with scrambled sentence tests shows that these tests can successfully prime participants in a following task. The null hypothesis is that there exists no difference in reciprocity-level between a participant who did the negative priming condition and a participant who did the neutral priming condition. However, because of the literature on priming with scrambled sentence tests I expect that the negative priming condition shows less reciprocity in the trust game then the neutral priming condition.
The next null hypothesis is that there exists no difference in reciprocity-level between a participant who did the positive priming condition and a participant who did the neutral priming condition. However, because of the same literature on priming with scrambled sentence tests I expect that the positive priming condition shows more reciprocity in the trust game then the neutral priming condition.
3.2.2. Hypotheses on the trust-game results. Hypothesis 2a. The trust-level today.
The trust game in phase 2 has the same form as the trust games of Berg et al. (1995) and Burks et al. (2003). According to their combined results, the average amount sent was $5.83 out of $10 with a median of $6. The next null hypothesis is that the median has not changed overtime. I test the median because I expect no normality in the results of the trust game.
18 Hypothesis 2b. The reciprocity-level today.
According to the result of Berg et al. (1995), the average amount sent back by the receivers was 30.1%. Because every receiver has to complete an entire strategy for every received amount in my experiment, the maximum mean to give back is €15. According to the results of the literature above this was, recalculated, €4.52 with a median of €2. The null hypothesis is that the median has not changed overtime.
3.2.3. Hypotheses on the differences between economists and social scientists. Hypothesis 3a. The difference in trust between economists and social scientists.
The next null hypothesis is that there exists no difference in trust-level between economists and social scientists. According to the experiment of Carter and Irons (1991), economists' behavior on average is closer to that predicted by the economic model: economists accept less and propose to keep more. That 'proposing to keep more' leads to the expectation that the trust-level among participants with an economic education is lower than the trust-level among participants with a social education.
Hypothesis 3b. The difference in reciprocity between economists and social scientists.
The next null hypothesis is that there exists no difference in reciprocity-level between economists and social scientists. According to the literature in section 2.3, the discussion whether economists are the ones that lack cooperative behavior is still unsolved. However, I expect that economists are less reciprocal than social scientists because economists are known with game theory and the Nash-equilibrium. I expect that they will follow economic game theory earlier than participants with a social education will.
Hypothesis 3c. The difference between economists and social scientists in the attitude to charity giving.
Frank et al. (1993) mention that college economics professors were much less likely than college professors from a variety of other disciplines to contribute to private charities in their experiment. The null hypothesis is that there exists no difference between participants with an economic education and participants with a social education in the attitude to charity giving. However, according to this literature I expect that participants with an economic education would give less to charities than participants with a social education.
Hypothesis 3d. The difference between economists and social scientists in the willingness to make a lot of money above helping to develop the society.
The next null hypothesis is that there exists no difference between economists and social scientists in the willingness to make a lot of money above to care for their fellowman and help to develop the
19
society. However, I expect that economists are more tended to choose for making a lot of money above to care for their fellowman and help to develop the society than social scientists. This expectation should be valid because the expectation in hypothesis 3b shows that economists behave less reciprocal than social scientists. So it is also likely that an economist will be more likely to choose to make a lot of money than a social scientist.
3.2.4. Hypotheses on the differences of political affiliation and gender. Hypothesis 4a. The difference in trust between left and right-wing voters.
Carlsson et al. (2005) show that left-wing voters are more risk-averse. The null hypothesis is that the level of trust is not different between left and right-wing voters. However, I expect that left-wing voters send less than right-wing voters because sending money in the trust-game is a risk and above literature shows that left-wing voters are more risk-averse.
Hypothesis 4b. The difference in trust between males and females.
Several experiments were done to examine the difference between males and females in trust and reciprocal behavior For example, females trust less; They are more risk-averse (Burks et al., 2003; Carlsson et al., 2005). The null hypothesis is that the level of trust is not different between males and females. However, because of the above literature I expect that females send less than males in the trust-game.
Hypothesis 4c. The difference in reciprocity between left and right-wing voters.
The next null hypothesis is that there exists no difference in the level of reciprocity between left and right-wing voters. Studies show that left-wing voters are more cooperative (Lunati, 1992). This would contradict the expectation that left-wing voters do send less in the trust-game, but because this result came up in a case where there was no risk, I only expect this cooperative behavior in the form of reciprocity. So the expectation is that the level of reciprocity of left-wing voters is higher than the level of reciprocity of right-wing voters.
Hypothesis 4d. The difference in reciprocity between males and females.
The next null hypothesis is that there exists no difference in the level of reciprocity between males and females. Frank et al. (1993) found that there is a sharp tendency for males to behave less cooperatively in their experiments. This would contradict the expectation that females do send less in the trust-game, but because these results came up in a case where there was no risk, I only expect this less cooperative behavior by males in the form of reciprocity. So the expectation is that the level of reciprocity of females is higher than the level of reciprocity of males.
20
Hypothesis 4e. The difference in trust between a left-wing female voter and a right-wing male voter.
According to the literature in hypothesis 4a and 4b, left-wing voters are more risk averse than right-wing voters and females are more risk averse than males. The null hypothesis is there exists no difference in the level of trust between a female who votes left and a male who votes right. However, because of the results in this literature I expect that a left-wing female voter trusts less than a right-wing male voter.
Hypothesis 4f. The difference in reciprocity between a left-wing female voter and a right-wing male voter.
According to the literature in hypothesis 4c and 4d, left-wing voters are more cooperative and females are also more cooperative. The null hypothesis is that there exists no difference in the level of reciprocity between left-wing females and right-wing males. However, because of the results in this literature, I expect that a left-wing female voter behaves more reciprocal than a right-wing male voter. Using the same explanation as in hypothesis 4c and 4d, I only expect this cooperative behavior in the form of reciprocity.
Hypothesis 4g. The difference between males and females in the attitude to charity giving.
The next null hypothesis is that there exists no difference between males and females in the attitude to charity giving.. In an experiment of List (2004) where subjects could do a donation, the proportion of females who donated was significantly larger than the proportion of male donators. Because of this literature I expect that male participants would give less to charities than female participants.
Hypothesis 4h. The difference in reciprocity between a female who's attitude to charity giving is positive and a male who's attitude to charity giving is negative.
The next null hypothesis is that there exists no difference in the level of reciprocity between a female who would give to charities earlier than spending her money in her own country and a male who prefers the opposite. However, the literature above shows that male participants would give less to charities than female participants. The literature in hypothesis 4d shows that females behave more reciprocal than males. According to this literature I expect that a female who is more likely to give to charities than spending her money in her own country will behave more reciprocal in the trust game than a male who prefers the opposite. This expectation should be valid because it is likely that a female who prefers to give to charity above spending all her money in her own country behaves more reciprocal than a male who prefers the opposite. This expectation is strengthened by the fact that previous experiments show that females behave more reciprocal than males.
21 3.2.5. Hypothesis on the differences of age.
Hypothesis 5a. The difference in reciprocity between participants with different ages.
Another null hypothesis is that there exists no difference in the level of reciprocity between older participants and younger participants. List (2004) found that mature individuals tend to provide greater voluntary contributions to public goods than younger cohorts. Frank et al. (1993) showed evidence that older students generally give greater weight to social concerns like the ones that arise in free-rider experiments. Because of this literature I expect that older participants behave more reciprocal than younger participants.
I mainly ask students to participate in my experiment, so I make a distinction between students who are younger than 23 and students that are 23 or older. I expect to have a sufficient number in both groups then2. The null hypothesis here is that there exists no difference in the level of reciprocity between participants who are 23 years or older and participants below the age of 23. However, because of the literature above I expect the opposite.
3.2.6. Hypotheses on the questionnaire equations3.
Hypothesis 6a. The difference in reciprocity between participants who have a different attitude to charity giving.
The next null hypothesis is that there exists no difference in reciprocity level between a participant who would choose to give to charity above spending all of his money in his own country and a participant who prefers the opposite. However, I expect the opposite. This expectation should be valid because it is likely that a person who gives his money away will also behave less selfish in an experimental setting than a person who is not likely to give his money away.
Hypothesis 6b. The difference in reciprocity between a participant who prefers to make a lot of money above caring for his fellowman and help to develop the society and a participant who prefers the opposite.
Another null hypothesis is that there exists no difference in reciprocity level between a participant who prefers to make a lot of money (i) above caring for their fellowman and help to develop the society (ii) and a participant who prefers the opposite. However, I expect the opposite. This expectation should be valid because it is likely that when a participant prefers (i) over (ii) he behaves more selfish which then will result in less reciprocal behavior in the trust game.
2For each difference I want to examine, a minimum of 20 respondents for each type is needed to obtain valid conclusions. 3The questionnaire equations are the three questions on the 5-point Lickert scale about what participants' feelings think is more important.
Q1: (i) Make a lot of money - (ii) Care for your fellowman and help to develop the society.
Q2: (i) Spend all your money in your own country - (ii) Also give to charities and development in the third world. Q3: (i) Vote for a right-wing political party - (ii) Vote for a left-wing political party.
22 3.3. Data collection
I collected my data from an online experiment executed among friends and family.
The experiment started on the 5th of March and runs till May 20th. The experiment is generated at http://jeroen.browsertech.nl which is also the direct link to the experiment.
Because I examine population differences in trusting, reciprocal behavior and the role of priming, an online experiment was the best option possible. I also have some experience in website-building so this was an easy decision. The online experiment successfully distributes the participant randomly to one of the three priming conditions and to one of the two roles in the trust game of phase 2. In order to obtain valid conclusions, a minimum of 150 respondents are needed. This number of respondents is sufficient to examine the population differences and the role of priming. I mention that for each difference I want to examine, a minimum of 20 respondents for each type is needed to obtain valid conclusions. So for example, when I want to examine the difference between participants with an economic education and participants with a social education, I need at least 20 economists and 20 social scientists to obtain valid conclusions.
To collect at least 150 respondents, I asked all my friends and family by phone and through Facebook messages. After two months I collected 175 useful respondents. Useful in this case means the participants who did not answer 'none' in the question how well they understood the trust game in phase 2. These participants were skipped from the results.
The online experiment examines all questions that are interesting to know as stated in the first part of Chapter 3 of this thesis. For example it is interesting to know whether a positive scrambled sentence test successfully can prime the participants in a more reciprocal direction in the trust-game of phase 2. The results of the trust game make me able to compare participants who did a positive scrambled sentence test with participants who did a neutral test.
4. Results
As stated in the data collection section, there were 175 people participating in the online experiment. The website sent 87 participants to the role as sender (49.7 percent), 88 to the role as receiver (50.3 percent). The website sent 47 participants to the negative priming condition (26.9 percent), 80 participants to the neutral priming condition (45.7 percent) and 48 to the positive priming condition (27.4 percent). 96 participants are male (54.9 percent) and 79 female (45.1 percent). 130 participants are below the age of 30 and the average age was 27.46 years. 115 participants have HBO or WO as highest completed education (65.7 percent).
To check whether I can use t-tests for my results I conducted one-sample Kolmogorov-Smirnov tests for normality on the results of the senders, receivers and the results of the questionnaire
23
equations. The results of the average sent by senders (M=5.91 ; SD=2.868) show a KS-statistic of 0.176 with Sig. .000 which concludes that these results are not normally distributed. The results of the average sent back by receivers (M=6.88 ; SD=3.495) show a KS-statistic of 0.159 with Sig. .000 which concludes that these results are not normally distributed. Also all possible sent back amounts by the receivers were tested on normality. They gave the same conclusions as the average sent back amount of the receivers. The results of the first questionnaire equation (M=3.32 ; SD=1) show a KS-statistic of 0.249 with Sig. .000 which concludes that these results are not normally distributed. The results of the second questionnaire equation (M=2.66 ; SD=0,981) show a KS-statistic of 0.200 with Sig. .000 which concludes that these results are not normally distributed. The results of the third and last questionnaire equation (M=2.81 ; SD=1,142) show a KS-statistic of 0.171 with Sig. .000 which concludes that these results are also not normally distributed.
According to these outcomes, the hypotheses that need these results are not tested with independent sample t-tests but with Mann-Whitney U tests using SPSS (from now on MWU-test).
4.1. Hypothesis test.
In this section, the data obtained in the experiment is analyzed in five different parts to test the hypotheses outlined in the previous section.
The first part contains the results of the priming. The second part contains the results of the levels of trust and reciprocity in the trust-game. The third part contains all results that examine the differences between economic educated participants and social educated participants. The fourth part contains all results that examine the differences in political affiliation and gender. The fifth and last part shows the results that examine the differences in age and the results of the equations as stated in part 1 of the questionnaire. All results are summarized at the end of this thesis in appendix A.
Before I can test whether the priming conditions are able to successfully prime participants in a certain direction, it is crucial to examine whether there are participants who consciously recognized the primes. The results of 90 participants who did the positive or the negative priming condition show that only 5 participants consciously recognized the primes. I conclude that the scrambled sentence test is successful to hide these primes. The 5 participants who consciously recognized the primes are skipped from the tests that examine the priming conditions.
4.1.1. The priming results
Hypothesis 1a. The effect of priming on trust.
As stated in hypothesis 1a, I expected that participants who did the positive priming condition trust more than participants who did the neutral condition. The p-value of .374 in table 3 shows that there is not enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I cannot conclude that the positive priming condition successfully primes the participant to more
24
trusting behavior. This conclusion is the same when I compare the trust-level of the negative priming condition with the neutral priming condition. The p-value of .346 in table 3 shows that there is not enough statistical evidence to reject the null hypothesis at any significance-level.
Hypothesis 1b. The effect of priming on reciprocity.
As stated in hypothesis 1b, I expected that participants who did the positive priming condition show more reciprocal behavior than participants who did the neutral priming condition. The p-value of .001 in table 3 shows there is enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I can conclude that the positive priming condition successfully primes the participant to more reciprocal behavior.
Because participants who were receivers had to complete an entire strategy for every possible amount they could receive, I can check whether the positive priming condition was successful at any possible received amount. At any amount that the receivers could receive, the p-values in table 4 show that there is enough statistical evidence to reject the null hypothesis at any significance level. Due to these results I can conclude that the positive priming condition successfully primes the participant to more reciprocal behavior at any amount that the participant could receive.
As stated in hypothesis 1b, I expected that participants who did the negative priming show less reciprocal behavior than participants who did the neutral priming condition. The p-value of .051 in table 3 shows that there is enough statistical evidence to reject the null hypothesis at the 10% significance-level, however, not at the 5% significance-level. Due to this result I cannot conclude that the negative priming condition successfully primes the participant to less reciprocal behavior.
At an amount of €1 and €4-€10 sent, table 4 shows that the negative priming condition is successful in priming participants to less reciprocal behavior at a 10% significance-level. The p-values of €5-€6 and €8-€10 show this also at the 5% significance-level.
Figure 1 gives a clear view of the reciprocity levels of all priming conditions at each amount that is sent by the sender. It shows that participants who did the positive priming condition behave more reciprocal than the participants who did the neutral priming condition. As stated above, this result is significant. The figure also shows that participants who did the negative priming condition show less reciprocal behavior. However, this result is not significant at the 5%-level.
25
Table 3. Trust and reciprocity of the priming conditions.
Fig 1. Reciprocity differences between the priming conditions.
Table 4.Reciprocity of the priming conditions at all possible received amounts.
Sent amount Negative priming Neutral priming Positive priming P-value negative vs. neutral (1-tailed) P-value positive vs. neutral (1-tailed) €0 €0 €0 €0 1.000 1.000 €1 €0.95 €1.30 €1.90 .058 .001 €2 €1.81 €2.35 €3.45 .119 .001 €3 €2.90 €3.74 €5.25 .151 .001 €4 €3.67 €5.02 €7.15 .091 .001 €5 €4.67 €6.42 €8.90 .035 .002 €6 €5.33 €7.86 €10.80 .021 .001 €7 €7.14 €9.14 €12.85 .092 .001 €8 €7.76 €10.65 €14.75 .049 .001 €9 €8.57 €12.26 €16.90 .023 .002 €10 €9.43 €13.28 €18.90 .028 .001 0 2 4 6 8 10 12 14 16 18 20 0 1 2 3 4 5 6 7 8 9 10 Am ou nt sen t ba ck by r ec eiver
Amount sent by sender
Neutral Negative Positive Negative priming Neutral priming Positive priming P-value negative vs. neutral (1-tailed) P-value positive vs. neutral (1-tailed) Trust €5.70 €6.00 €5.81 .346 .374 Reciprocity €4.75 €6.55 €9.17 .051 .001
26 4.1.2. The trust-game results.
Hypothesis 2a. The trust-level today.
The average amount sent in the experiment was €5.91. Because of the not-normality of the results in the trust-game, I tested the hypothesis with a one-sample Wilcoxon signed rank test. The result in table 5 show a Sig. of .778 which means that there is not enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I may conclude that the median has not changed overtime.
Hypothesis 2b. The reciprocity-level today.
The average sent back amount in the experiment was €6.55. Using the same one-sample Wilcoxon signed rank test as above, the result in table 5 show a p-value of .000 which means that there is enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I can conclude that the median has changed overtime. I only used the average amount sent back of participants who did the neutral priming condition in my analysis. I did this because I compare my results with the results of Berg et al.. Because the positive priming condition in my experiment show that it was successful to prime participants to more reciprocal behavior, it would lead to an upward bias taking these results into account.
Table 5. Trust and reciprocity today vs. previous literature.
My experiment Combined results of the literature P-value (2-tailed)
Mean Trust €5.91 €5.83 -
Mean Reciprocity €6.55 €4.52 -
Median Trust €5.00 €6.00 .778
Median Reciprocity €6.32 €2.00 .000
4.1.3. The results of economic educated participants and social educated participants. Hypothesis 3a. The difference in trust between economists and social scientists.
As stated in hypothesis 3a, I expected that the trust-level among participants with an economic education is lower than the trust-level among participants with a social education. The p-value of .306 in table 6 shows that there is not enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I cannot conclude that trust among participants with an economic education is lower than participants with a social education. I mention that the number of senders with a social education was only 13 and that this number was only 19 among economic educated participants. This low N for both types causes that my conclusions are not very strong.
27
Hypothesis 3b. The difference in reciprocity between economists and social scientists.
As stated in hypothesis 3b, I expected that the amount sent back by participants with an economic education is lower than participants with a social education. The p-value of .065 in table 6 shows that there is enough statistical evidence to reject the null hypothesis at the 10% significance-level. However, it does not show enough statistical evidence at the 5% significance-level. Due to this result and based on the 5% significance-level, I cannot conclude that participants with an economic education behave less reciprocal than participants with a social education. What I do mention is that the number of these participants is again low (N=19 for Social scientists and N=18 for Economists). This low N for both types causes that conclusions are not very strong.
To check whether there are differences in reciprocity at any possible received amount, the p-values in table 7 show that when a sender sent an amount of €2-€6 or €10, participants with an economic education are less reciprocal at the 5% significance-level than participants with a social education. Again, because of the low N, this conclusion is not very strong.
Hypothesis 3c. The difference in attitude to charity giving between economists and social scientists.
Because the questionnaire equations are not normally distributed as well, I also conducted a MWU-test to analyze the result whether economists have a different attitude to charity giving than social scientists. The p-value of .003 in table 6 shows us that there is enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I can conclude that participants with an economic education would give less to charities than participants with a social education.
Hypothesis 3d. The difference between economists and social scientists in the willingness to make a lot of money above helping to develop the society.
As stated in hypothesis 3d, it is likely that a participant with an economic education is more tended to choose for making a lot of money above caring for their fellowman and help to develop the society than a participant with a social education. The p-value of .000 in table 6 shows that there is enough statistical evidence to reject the null hypothesis at any significance-level. Due to this result I can conclude that participants with an economic education prefer making a lot of money above caring for their fellowman and help to develop the society more than participants with a social education.
