The Evolution of Religion:
Mixed Signals Regarding the Costly
Signaling Theory of Religion
Olmo van den Akker (5843049)
March 14, 2014
Bachelor’s Thesis Economics University of Amsterdam Supervisor: Boris van Leeuwen Number of words:
Abstract
In this study, the costly signaling theory of religion was tested using 135 religious
participants. In accordance with the theory, it was found that participants who scored higher on a costly signaling questionnaire, also scored higher on self-report items that measure willingness to cooperate, altruism, received cooperation, and received trust. However, no positive relationship was found when costly signaling was related to behavioral measures of willingness to cooperate (cooperation in a prisoner’s dilemma) and altruism (donations in a dictator game). When taking into account validity and robustness issues of the self-report measures, it was concluded that there was not enough evidence for an association between costly signaling and both willingness to cooperate and altruism. Further, these issues make the positive associations between costly signaling and received trust, and between costly signaling and received cooperation, only tentative. In all, the evidence for the costly signaling theory of religion is mixed.
Introduction
Studies show that presently more than 70% of the world population has some sort of religious belief (World Values Association, 2009; Maoz & Henderson, 2013). The fact that religiosity is so abundant is quite puzzling from an evolutionary perspective. This is because practicing a religion often involves behaviors that can decrease an individual’s reproductive success or chance of survival. Examples of this are the water and food deprivation during fasting periods, sexual abstinence, either before marriage or completely, and health risks due to not taking vaccines. Besides such specific examples there is also the non-trivial investment of time and energy people make while practicing their religion (e.g. visiting church and saying daily prayers). This time and energy could also be used in other, more fitness enhancing ways, making religious activity seem quite wasteful from an evolutionary viewpoint. The fact that practicing a religion can have adverse effects on an individual’s fitness raises the question why religiosity is so prevalent after thousands of years of natural selection.
A solution to this evolutionary paradox can be found in the idea that some facets of religion may entail fitness benefits that can overcome the previously mentioned fitness costs. William Irons (1996; 2001) is one of the people who endorse this perspective. He proposes the costly signaling theory of religion, which builds on the idea that religion is a means to signal cooperative intent to others. In general, the costly signaling mechanism has been proposed as a way to enhance cooperation and altruism, and can work with all kinds of signaling devices (Gintis, Smith, & Bowles, 2002). The proposed mechanism functions as follows. Let us suppose that an individual often makes personal investments in a specific group setting. Examples of this are the participation in group meetings or the following of certain group guidelines. Other members of the group can perceive these investments as signals of loyalty and commitment to the group. These other members will subsequently be more inclined to trust the signaling individual. On the basis of this trust, they are more likely to cooperate with, and be helpful to this individual. This is because they trust this individual to be cooperative and helpful himself. If most group-members signal their commitments in this way, the group as a whole becomes more cooperative. This can confer an evolutionary advantage over other groups since all kinds of activities that require effective cooperation can increase the fitness of individuals. One can think for instance of improved hunting, increased trade, and most importantly, more effective warfare (Irons, 1996). In short, signaling
cooperative intent can have beneficial effects on an individual’s fitness. Irons proposes that this signaling can be effectively achieved through religious participation. Moreover, he
argues that the fitness benefits that come with this religious participation are large enough to outweigh the previously mentioned fitness costs. The fact that religiosity has these net fitness benefits can explain why religiosity is so widespread throughout the world.
In developing his argument, Irons (1996) remains vague about why it is that religion is such an effective means of signaling cooperative intent, and why no other means are equally or more effective. Other authors did put forward specific arguments to explain this. Most arguments focus on the importance of religious rituals, because people can signal their loyalty and commitment to a group through participation in those rituals. Moreover, rituals are an essential part of religious tradition all over the world. Alcorta and Sosis (2005) argue that participation in rituals is an especially effective signaling device because most religious rituals are formal, public, and very salient happenings. The formality, communality and salience of rituals increase their visibility, and this makes them more effective as a signal. Besides this, religious rituals are associated with highly emotional experiences. Emotions are known for their memory enhancing capabilities (Willingham, 2007), which increase the retention of religious signals even more. Dawkins (1976) emphasizes that religious rituals often include counterintuitive concepts that violate the basic premises of laws of nature, examples of which are resurrections, virgin births, and animism. These counterintuitive concepts can easily grab people’s attention and will therefore be more easily memorized. This better memorization of course makes for a more effective signal. Henrich (2009) adds to this by emphasizing that supernatural entities, which play a major role in many religious rituals, are often difficult to falsify. This means that rituals that include supernatural entities will probably not be devalued, and therefore will consistently have a large signaling potential. A final argument comes from Sosis (2003), who states that the multimodal nature of religious ritual facilitates the transmission of signals. This argument is based on findings that signals including multiple senses have a higher potential for memorization than signals including only one sense (Rowe, 1999).
In all, it is plausible that partaking in religious rituals is very effective to signal cooperative intent. Even though this is the case, the signaling mechanism has its weakness; it is vulnerable to deception. The reason for this is that in many situations, it is difficult to decide to what extent other parties actually put effort into a cooperative venture. In a hunting group, you can never be sure whether someone is actually trying his best to catch a prey or whether he is slacking to save energy (Hawkes, 1993). Likewise, for firms in a cartel situation it is often impossible to predict whether the other firm will set their price upon the collusive level or will set a lower price and take a larger share of the profits (Stigler, 1964). In
cases like this, it can be tempting to signal to others that you intend to cooperate, but when push comes to shove actually do not do so. In this way it is possible to reap the benefits from the cooperative action (e.g. more food or higher profits) because the others believed your good intentions and cooperated, but you will not incur the cost of putting effort in yourself. Therefore, in situations where effort levels are not transparent, individuals in potentially cooperative groups have an incentive to not cooperate and free ride on other people’s cooperative actions. Whether or not they actually do so depends on many different factors, among which are personality, the social environment, and the costs of free riding (Albanese, & Van Fleet, 1985). When free riders exist, the cooperative actions of the other group members are less fruitful and this can lead to the breakup of cooperation in the group. This situation is dubbed the problem of collective action (Olson, 1971).
Luckily, the costly signaling mechanism includes a solution to this problem. The solution is based on the assumption that some people are more committed to a group than others, and will therefore be more likely to believe in and adhere to the group’s norms. With regard to the costly signaling theory of religion, Sosis (2003) labels these types of people ‘believers’ and ‘skeptics’ respectively. It can be argued that believers are, in general, less willing to free ride on the effort of group members. This is because cooperation and altruism are prevalent norms in many religious groups. So, if the majority of signalers are believers and not skeptics, there would be less free riding and sustainable cooperation can occur. However, this situation could only arise if the cost of signaling (i.e. religious participation) is lower for believers than it is for skeptics. It is reasonable to assume this is the case because religious participation is intrinsically valuable to those who firmly believe in the norms of a religious group. For example, people would have little problem with attending church every day if they firmly believed it would ensure them a place in heaven. People who do not believe that church attendance will ensure a place in heaven would be less willing to attend church regularly; for these people the cost of signaling is higher. What this means is that many skeptics will not use religious participation to signal cooperative intent. The majority of signalers are therefore believers that have a low willingness to free ride. Sosis (2003) has proven that when signals are sufficiently costly, the number of free riders is so low that a stable cooperative state can be achieved. In conclusion, only if signals are sufficiently costly can cooperative intentions be credibly signaled to others and can the signaling mechanism lead to sustained cooperation in a group.
Because of the issue raised above, it is critical to make a clear distinction between costly signals and non-costly signals. Sosis and Bressler (2003) propose that there are two
types of costly signals. The first type are behaviors that are required by a group that entail time, energetic, and/or financial costs that are not directed toward efficiently accomplishing evolutionary goals. The participation in religious rituals or meetings is a good example of a costly signal of this type, since it entails time and energy that could as well have been used for fitness enhancing activities. The second type are behaviors that might have entailed evolutionary benefits, but are restricted by group norms (e.g. sexual abstinence, and not eating healthy foods due to religious guidelines). This study will focus on the first type of costly signals because behavior that is exhibited can more readily be measured than behavior that is not exhibited.
Besides an accurate definition of costly signals, it is also important to define the other concepts that are of importance in the costly signaling mechanism: cooperation, and trust. In the literature, there have been some different, sometimes inconsistent, definitions of
cooperation. Some studies, implicitly or explicitly, define cooperation as “personally costly
behavior that benefits others” (Lehmann & Keller, 2006; Ernest-Jones, Nettle, & Bateson,
2011; West, El Mouden, & Gardner, 2011). This is a broad definition of cooperation in the sense that it incorporates phenomena like generosity and helping, that could also be labeled as altruism. The costly signaling mechanism is also relevant for such altruistic behaviors, but altruism will be defined separately in this study to avoid confusion. Cooperation will be
defined as “acts or instances of working or acting together for a common purpose or benefit.” (Dictionary.com, 2013). An added advantage of this narrow definition is that it is more
similar to the de facto meaning of the term cooperation, which originates from the Latin
cooperationem, meaning ‘a working together’. In accordance with Gray (2007), altruism will
be defined as “behavior that helps another while it decreases the evolutionary success of
oneself”.
The remaining concept that needs defining is trust. This is a difficult task, since trust has different connotations in different scientific disciplines. For this reason a broad definition is used that is based on several disciplines: “Trust is a psychological state comprising the
intention to accept vulnerability based upon positive expectations of the intentions or behavior of another.” (Rousseau, Sitkin, Burt, & Camerer, 1998). Only studies that used
definitions of cooperation, altruism, and trust similar to the ones formulated above were used as a basis for the current study.
The costly signaling theory of religion makes four predictions that will be tested in this study. A first prediction is that people who send out more costly signals have a higher
willingness to cooperate with others. This follows from two assumptions that are crucial for
the sustainability of the costly signaling mechanism. First, the majority of the costly signalers are believers because the signaling costs for believers are lower than the signaling costs for skeptics. Second, believers are more willing to cooperate than skeptics because they place a higher value on adherence to cooperative group norms.
In line with this first prediction, Ruffle and Sosis (2007) found that members of religious kibbutzim were more cooperative in a common-pool resource dilemma game than their secular counterparts. Moreover, the individuals in these kibbutzim who had the highest frequency of ritual participation displayed the highest levels of cooperation. Further, Soler (2012) used a tailor-made questionnaire to measure the level of costly signaling. He found that people who scored higher on this questionnaire were more cooperative in a public good game. Also, they claimed to be involved in more instances of intragroup cooperation.
The studies mentioned above intentionally set out to test the costly signaling theory of religion. However, studies that investigated related subjects can be instructive as well. For example, Ahmed (2009) found that students in India who were studying to become an imam were more likely to cooperate in a public good game than were nonreligious students. Similarly, Anderson, Mellor and Milyo (2010) found that higher attendance in organized religious services was related to higher contributions in a public good game. On the other hand, in studies by Anderson and Mellor (2009), and Paciotti et al. (2011), no relationship was found between religious participation and contributions in public good experiments.
The costly signaling theory not only states that costly signalers are more cooperative, but also that they are more altruistic. This second prediction follows from the theory because a group where reciprocal altruism is the norm would have an evolutionary advantage over a group with mainly selfish individuals. The literature on this prediction is limited, since many studies do not use religious attendance or costly signaling as independent variables. Instead, they focus on experimental manipulations where religious thoughts are manipulated (e.g. Shariff & Norenzayan, 2007; Pichon, Boccato, & Saroglou, 2007). Since the costly signaling theory is about actions and not about thoughts, these studies are not relevant. There are a few studies that do focus on religious actions. These studies did not find the predicted relationship between religious attendance and altruism (Eckel & Grossmann, 2004; Tan, 2006; Bulbulia & Mahoney, 2008).
The third prediction is that people who send out more costly signals are perceived as more trustworthy. This perceived trustworthiness will be labeled received trust in this study, because it involves trust that is received from others. Regarding this third prediction, Edgell, Gerteis, and Hartmann (2006) found that religiosity was positively associated with received
trust. Adding to that, Tan and Vogel (2008) found that religious people were trusted more than nonreligious people in a trust game. Equivalently, Orbell, Goldman, Mulford, and Dawes (1992) found that both religious and nonreligious people believe that religious people will cooperate more than nonreligious people in a prisoner’s dilemma game.
The fourth prediction of the costly signaling theory of religion is based on the increased trust that stems from the third prediction. This fourth prediction is that third parties are more likely to cooperate with people who send out more costly signals. This form of cooperation will be labeled received cooperation because it involves cooperation that is received from others. Not much research has been done regarding this prediction, but Soler (2012) found that actively signaling participants elicited more instances of received
cooperation within their religious community.
The existing literature on the costly signaling mechanism of religion, although not entirely consistent, does seem to underpin the theory to some extent. However, one can raise doubts about how the studies have measured the level of costly signaling. This is because almost all studies used attendance at religious meetings as their main independent variable. By doing this, these studies ignored other relevant aspects of religious costly signaling like the personal importance people attach to the religious meetings, and the intensity with which they participate in them. This study will therefore use a questionnaire that incorporates not only items on religious attendance, but also items on these other aspects.
Another point of critique on the existing literature is that many studies did not take into account the possibility of omitted variable bias (Stock, & Watson, 2007). Omitted variable bias occurs when variables that are correlated both with the dependent and
independent variable are excluded from a model. This is problematic because it could be the case that these excluded variables are responsible for the variance in the dependent variable, instead of the independent variable. For example, people who live in the countryside more often attend religious meetings than people in more urban environments (Chalfant, & Heller, 1991). In addition, it could be the case that these people learn to be more cooperative because small communities are largely dependent on close relations. This means that whether
someone is living in the countryside or not can account for a positive relationship between religious attendance and willingness to cooperate. This is troublesome, because we are interested in the causal relationship between religious attendance and willingness to cooperate. Another variable for which this is the case is the number of siblings. Many religious families are large, and it can be argued that cooperation and trust arise more naturally in large families. Therefore, also the number of siblings can be a confounding
variable. To avoid the risk of omitted variable bias, this study will take into account these two potentially confounding variables.
As stated before, this study will focus on four predictions of the costly signaling theory of religion that are related to willingness to cooperate, altruism, received trust, and received cooperation. A specifically designed questionnaire is used to measure the level of costly signaling of the participants. This questionnaire is loosely based on the questionnaire employed by Soler (2012). Further, willingness to cooperate is measured using a prisoner’s dilemma game. This game consists of two players that both have the option to cooperate or not cooperate with the other player. The game has been extensively used by economists and psychologists to measure people’s willingness to cooperate. Besides that, a dictator game is used to measure the degree of altruism of the participants. In this dictator game participants have the possibility to donate money to one of two charities. Finally, items on willingness to cooperate, altruism, received trust, and received cooperation are included in an additional questionnaire. Both the questionnaire, the prisoner’s dilemma, and the dictator game are part of an online survey that was made available to a selective group of religious students.
Following the predictions of the costly signaling theory of religion, it is expected that participants who score higher on the costly signaling questionnaire are also more likely to cooperate in the prisoner’s dilemma, donate higher amounts in the dictator game, and score higher on the items regarding willingness to cooperate, altruism received trust, and received cooperation.
Method
Participants
A large part of the participants in this study were approached by contacting the boards of various religious student fractions with the request of letting their members fill out the online survey. The student fractions of which members participated were Navigators Student
Association Amsterdam, CSFR Amsterdam, and Islamic Student Association Ibn Firnas from Delft. It is estimated that around 450 people were approached in this way, and that 105 chose to participate. Other participants in this study were approached through a system called ‘Digitaal Proefpersoonpunten Management Systeem’, designed by the University of
Amsterdam specifically to find research participants. Those participants were approached that in earlier, unrelated tests indicated some form of religious belief. It is estimated that this
amounted to 300 approached students, of which approximately 60 chose to participate. All of those that participated were paid a small sum of money. This amount varied from 1 to 5 euro and depended on the choices the participants made in the experiment.
Materials
One of the ways willingness to cooperate was measured was through a standard prisoner’s dilemma game. Participants had to play a 2-person game in which they had the choice to either cooperate with the other player or not. They were told that the other player was another participant that had to make exactly the same choice. The decision in the prisoner’s dilemma was purposefully labeled as ‘cooperate’ or ‘not cooperate’ instead of the more neutral ‘choice A’ and ‘choice B’ to make sure that the participants knew what the cooperative choice was. Furthermore, it was emphasized that all players in the game were totally anonymous, so participants did not know with whom they would be matched. The full instructions of the prisoner’s dilemma game can be seen in Appendix A.
In the prisoner’s dilemma, fictional monetary units called neuros were used, where each neuro corresponded to 2.5 eurocents. Participants could earn a minimum of 40 neuros (1 euro) and a maximum of 200 neuros (5 euro) with their decision. The full payoff table is shown in Figure 1. Whether or not participants cooperated in the prisoner’s dilemma game was taken as a measure of their willingness to cooperate.
Figure 1 *
The Prisoner’s Dilemma Game as Presented to the Participants
* the text is translated from Dutch
Besides playing the prisoner’s dilemma game, participants also had to play a so-called dictator game. In this game, participants were allocated 160 neuros and were given the choice to keep this amount or to donate a percentage of it to a given charity. The participants could decide for themselves what percentage they would like to donate. To see whether the level of the donations varied with the type of the charity, participants faced either a religious charity, The Salvation Army, or a secular charity, The Food Bank. Instructions of the dictator game can be found in Appendix B.
Before playing both games, participants were told that they would only receive payment for one of the games, and that it would be decided randomly which one this would be. Payments were made through bank transfer, and therefore participants had to fill in their bank account number at the end of the online survey. To assure their anonymity, they did not need to fill in the name that corresponded to the bank account number.
There was also a game included in the online survey without a monetary payoff. In this so-called snowy pictures task, 12 pictures were presented in which participants had to discern a figure. In some pictures there was indeed a figure, in others there was not.
Participants had to state what figure they saw, if any, and how sure they were that they saw this figure. This task was included as part of another study.
Aside from playing the games, participants had to fill in several questionnaires. One questionnaire was designed to measure the main independent variable, the level of costly signaling. The items in this questionnaire were loosely based on the items used by Soler (2012). Because Soler’s questionnaire was developed for Brazilian participants of a wide age range, the items were adjusted somewhat to make them more suitable for Dutch students. An overview of the costly signaling items is shown in Table 1. Some items on the level of
religiosity of the participants were also included in the questionnaire. These can be seen in Appendix C.
Another questionnaire included self-report items on prosocial behavior. Items that are of relevance to the costly signaling theory of religion are those regarding willingness to cooperate, altruism, received trust, and received cooperation. These items are shown in Table 2. The other items in this prosocial behavior questionnaire were relevant for another study and were not included in Table 2.
The items of the costly signaling questionnaire, as well as the items on prosocial behavior were Likert items with the following five answering possibilities: “I strongly disagree”, “I disagree”, “I neither agree nor disagree”, “I agree”, and “I strongly agree”. These answering possibilities corresponded to scores of 1, 2, 3, 4, and 5 respectively.
A final questionnaire was used to elicit demographic information from the participants. This questionnaire is shown in Appendix D.
Table 1 *
Items on Costly Signaling
Item Statement
1 I regularly attend religious activities
2 I often contribute to the organization of a religious activity 3 When there is a religious meeting, I wouldn’t want to miss it 4 Participating in religious rituals is not that important to me **
5 I sometimes go to a religious meetings, even though I prefer to do something else 6 I have given money to a religious organization or a religious charity
7 I have done volunteer work for a religious organization or a religious charity 8 When I go to a religious meeting, I want to participate in an active way * the items are translated from Dutch
** item that is phrased contra-indicative and was recoded for the statistical analyses
Table 2 *
Items on Willingness to Cooperate, Altruism, Received Trust, and Received Cooperation
Variable Item Measures
Co1 I’m a person that likes to cooperate with others Willingness to cooperate Alt I’m a generous person that likes to invest time and energy in others Altruism
Tru I have the feeling that other people trust me Received trust
Co2 Other people like to cooperate with me Received cooperation
* the items are translated from Dutch
Analysis
Because the choice made in the prisoner’s dilemma is dichotomous, a binary logistic regression was carried out. Besides that, linear regression analyses were done to investigate whether participants that showed more costly signaling donated more in the dictator game, and scored higher on willingness to cooperate, altruism, received cooperation, and received
trust. In all regressions, the same independent variables were used. The costly signaling variable was the main predictor. The environment a participant grew up in, and the number of siblings a participant grew up with, were control variables. Religious affiliation, level of religiosity, financial situation, and gender were added to explore whether some groups differed with regard to cooperation and trust.
Procedure
Participants got access to the online survey by clicking on a link that was made available to them. After receiving some information about the study, participants indicated whether or not they liked to participate. When they chose to participate, they were able to start with the first part of the survey.
While developing the online survey, close attention was paid to the order in which the tasks were presented. This was done to decrease the risk of carryover effects (Mook, 2001). Carryover effects occur if exposure to one task changes a participant’s behavior in a later task. This could be the case because people generally have a preference to behave consistently. This preference is common because most people will have feelings of cognitive dissonance when their beliefs, behaviors, or attitudes are inconsistent (Festinger, 1957). For example, if the self-report questionnaire would be presented before the prisoner’s dilemma, it is possible that participants who self-reported a high willingness to cooperate, would be more inclined to cooperate in the prisoner’s dilemma because that is the consistent thing to do. If the
prisoner’s dilemma was presented first, participants who cooperated in the prisoner’s dilemma may try to be consistent by self-reporting a higher willingness to cooperate. In this way, the behavior of the participants would change due to the order of the tasks. It was expected that these carryover effects would be smaller when the self-report questionnaire was presented first. The reason for this is that there was money at stake in the prisoner’s dilemma game. Because people tend to value money, this could deter them from changing their behavior because of consistency concerns.
The final order of tasks is listed in Table 3. On average, the online survey took approximately 15 minutes to finish.
Table 3
Order in which the tasks were presented
Task Type of task Shown in
1 Demographic information questionnaire Appendix D 2 Prosocial behavior questionnaire Table 2 *
3 Prisoner’s dilemma Figure 1 / Appendix A
4 Dictator game Appendix B
5 Religiosity questionnaire Appendix C
6 Snowy pictures task -
* only the items relevant for the costly signaling theory of religion are shown
Results
Participants
A total number of 165 people participated in this study. Of these 165, all participants that did not state themselves as religious were not taken into account in the analyses. This included 17 atheists, 7 agnostics, 5 deists (people that stated a belief in a supernatural entity but no
adherence to a specific religion) and 1 humanist. These 30 people were eliminated from the analysis because the costly signaling mechanism is not relevant to them; they do not have a religious group to signal their cooperative intent to. In all, 135 participants remained in the analysis. Their average age was 21.98 (SD = 4.21). Of all the participants, 71.1 percent chose to cooperate in the prisoner’s dilemma. The donations in the dictator game were on average 2.55 euro for The Salvation Army and 2.77 euro for The Food Bank. Finally, the participants scored 3.73, 3.90, and 4.21 respectively on the items measuring willingness to cooperate, received cooperation, and received trust. In Appendix E, descriptive statistics can be found for different categories of participants.
Principal Component Analysis
First, a principal component analysis was carried out over the costly signaling items that can be seen in Table 1. This was done to create an internally consistent measure of costly
signaling. In a principal component analysis, the correlations between various items are inspected. If one of the items does not correlate well with the other items, this is an indication that the item is measuring something different from what the other items are measuring. This
item may then be deleted from the analysis to create a more consistent measure.
Only with a sufficiently large sample size, reliable results will be obtained from a principal component analysis. Therefore, a Kaiser-Meyer-Olkin test was done to see if the sample size is large enough. The KMO test statistics ranges from 0 to 1, where values over 0.5 correspond to a sufficiently large sample size (Field, 2009). The overall KMO proved to be .86, which means the sample size in this study is more than sufficient to carry out a principal component analysis.
The principal component analysis was carried out using Varimax rotation. The results are shown in Table 4. It was found that two components had an Eigenvalue larger than 1. The first component was able to explain a large part of the variation in the questionnaire
responses (49.46%). The second component was able to explain markedly less variation (14.08%). Since the questionnaire was specifically developed to measure costly signaling, it is reasonable to assume that the underlying trait of the first component is the level of costly signaling. Items that are not significantly related to this component are probably measuring some other trait. These items will therefore be excluded from the statistical analyses. An often used cut-off point for determining what factor loading corresponds to a significantly low relation is 0.5. Therefore, item 4 and 5 were eliminated from the analysis. However, since the factor loading of item 4 did come close to the cut-off point level of 0.5, all statistical analyses were also done including item 4. This was done to control for the possibility that an arbitrary cut-off point influenced the results. The analyses including item 4 can be found in Appendix F. The scores for each participant on the 6 leftover items were summed, and this sum score was taken as a measure of the level of costly signaling of that participant.
The reliability analysis that was done over the 6 leftover items indicated that these items have a good internal consistency, Cronbach’s α = .88. The internal consistency of all 8 items together was α = 0.83.
Table 4
Rotated Component Matrix with Factor Loadings and Communalities of All Items, and the Eigenvalues of Both Components (N=135)
Binary Logistic Regression
Before carrying out the binary logistic regression, the assumption of multicollinearity was tested. There appeared to be no multicollinearity problems since the VIF for all variables was < 2.50, well below the cautionary value of 10 as suggested by Field (2009).
In the binary logistic regression, cooperation in the prisoner’s dilemma was regressed onto the costly signaling measure, level of religiosity, religious affiliation, gender, and
financial situation. When comparing different religious affiliations, the benchmark group was the one with the highest N, the protestant participants. Number of siblings, and youth
environment were included in the model as control variables. The result of this regression can be seen in Table 5. It was found that the level of costly signaling did not significantly predict whether participants cooperated or not, Wald χ²(1) = 1.91, p = .17. The significance level used in all regressions in this study is α = .05.
Component 1 Component 2 Communality
Item 1 .91 -.08 .83 Item 2 .88 .01 .77 Item 3 .68 .38 .60 Item 4 .48 .47 .45 Item 5 .21 -.82 .72 Item 6 .76 -.08 .58 Item 7 .78 -.11 .62 Item 8 .67 .25 .51 Eigenvalue 3.96 1.13 16
Table 5
Beta Coefficients (B), Standard Errors (SE), Odds Ratios and Nagelkerke’s R² for the Binary Logistic Regression on Cooperation in the Prisoner’s Dilemma (N=135)
B (SE) Odds Ratio
CS 0.09 (0.07) 1.10 Religiosity 0.03 (0.133) 1.03 Male -0.80 (0.44) 0.45 Financial Situation 0.13 (0.21) 1.14 Catholic 1.06 (0.84) 2.87 Undefined Christian -0.56 (0.82) 0.57 Muslim 1.03 (0.88) 2.79 Buddhist 0.49 (1.11) 1.63 R²N .13
Control variables: Number of Siblings, Youth Environment * p < .05 ** p < .01
Linear Regression on the Dictator Game Responses
For the linear regression on the dictator game responses, no problems of multicollinearity were found in the pretests (all VIFs < 6.5). Also, there appeared to be no dependence of error terms in the models. This was concluded because only random patterns were found when the standardized residuals were plotted against the model variables.
The results show a positive association between the level of religiosity and altruism, t = 2.91, p = .004, but no association between the level of costly signaling and altruism, t = -0.55, p = 0.78. Furthermore, it was found that there was no difference between the donations to the religious charity and the secular charity, t = -0.84, p = 0.40. Results of the linear
regression can be found in Table 6.
Table 6
Beta Coefficients (B), Standard Errors (SE), and R² of the Linear Regression on the Dictator Game Responses (N=135) B (SE) CS -0.43 (1.50) Religiosity 9.48** (3.26) Male -9.19 (10.78) Financial Situation -1.95 (5.01) Catholic -16.51 (19.43) Undefined Christian -.60 (21.15) Muslim -11.69 (18.11) Buddhist -34.12 (25.59) R²N .25
Control variables: Number of Siblings, Youth Environment * p < .05 ** p < .01
Linear Regressions on the Self-Report Items
For the linear regressions on the self-report items, no problems of multicollinearity were found in the pretests (all VIFs < 6.5). Also, there appeared to be no dependence of error terms in the models. This was concluded because only random patterns were found when the
standardized residuals were plotted against the model variables.
The linear regressions revealed a significant positive relationship between the level of costly signaling and willingness to cooperate, t = 2.30, p = .02, altruism, t = 2.10, p = .04, received cooperation, t = 3.11, p = .002, and received trust, t = 2.80, p = .01.
When comparing different religious affiliations, it was found that catholic participants scored higher on willingness to cooperate, t = 2.72, p = .01 and received trust, t = 3.82, p < .001, than protestant participants. Buddhist participants also scored higher on received trust than protestant participants, t = 2.18, p = .03. Furthermore, Muslim participants scored higher than protestant participants on altruism, t = 3.02, p = .003. The results of the regressions can be seen in Table 7.
Table 7
Beta Coefficients (B), Standard Errors (SE) and R² of the Linear Regressions on the Self-Report Items (N=135)
Co1 Alt Tru Co2
B (SE) B (SE) B (SE) B (SE)
CS 0.05* (0.02) 0.04* (0.02) 0.05** (0.02) 0.05** (0.02) Religiosity -0.08 (0.05) -0.09 (0.05) 0.01 (0.04) -0.03 (0.04) Male 0.12 (0.17) -0.05 (0.15) -0.08 (0.13) -0.14 (0.12) Financial Situation 0.02 (0.08) -0.09 (0.07) 0.02 (0.06) -0.02 (0.06) Catholic 0.52 (0.30) 0.49 (0.27) 0.87** (0.23) 0.60** (0.22) Undefined Christian -0.10 (0.33) -0.08 (0.30) 0.09 (0.25) -0.03 (0.24) Muslim -0.16 (0.28) 0.76** (0.25) 0.41 (0.21) 0.08 (0.21) Buddhist 0.33 (0.40) 0.59 (0.36) 0.66* (0.30) 0.44 (0.29) R² .09 .17 .19 .12
Control variables: Nunber of Siblings, Youth Environment * p < .05 ** p < .01
Conclusions and Discussion
In this study, four predictions of the costly signaling theory of religion (Irons, 1996; 2001) were tested. The first two predictions were that people who more often send out costly signals are both more willing to cooperate with others and more altruistic. Further, it was predicted that third parties perceive costly signalers to be more trustworthy and are therefore more willing to cooperate with them. It was found, through self-report questionnaires, that higher levels of costly signaling were indeed associated with higher levels of willingness to
cooperate, altruism, received cooperation, and received trust. However, willingness to cooperate and altruism did not seem to be related to costly signaling when behavioral measures were used.
An interesting finding was that the level of costly signaling was associated with the self-report measures of willingness to cooperate and altruism, but not with the behavioral measures of those variables. This dichotomy makes it difficult to interpret the results and raises questions about the validity of both the behavioral measures and the self-report measures. The validity of a measure can be defined as the extent to which the measure
actually measures what it is supposed to measure (Mook, 2001). With regard to the validity of the measures in this study, there are three points to be made.
First, it was possible to earn money in the prisoner’s dilemma game and the dictator game, while this was not the case for the self-report items. It is reasonable to assume that people think more deeply about what their true preferences are when money is at stake. And because they do, tasks involving monetary payoffs (e.g. the prisoner’s dilemma game) can induce responses that correspond more accurately with people’s actual preferences. In other words, tasks involving monetary payoffs are more valid.
Second, the self-report measures used the terms cooperation and trust without informing the participant about what is actually meant by these terms. We have seen previously that there are multiple definitions of the terms cooperation and trust, and it is therefore possible that participants interpreted these terms in different ways. If this was the case, the responses of the participants would differ due to differences in their interpretations and not due to differences in their willingness to cooperate, received cooperation, or received trust. In other words, the self-report measures have a validity problem. A solution to this problem would be to give a clear definition and some examples of cooperation and trust before presenting the items.
Third, it must be emphasized that the self-report measures consist of only one item. Due to this, it is possible that irrelevant factors had a large impact on the participant’s
responses. For example, a participant could be temporarily distracted or tired while answering this one item. If that is the case, the response of the participant probably does not correspond to his real preference. In other words, there is a validity problem. A solution to this problem would be to use multiple items because it is unlikely that a participant is distracted or tired while answering all of the items. The effects of distraction and tiredness will then have a smaller impact and researchers can be more certain that the responses of the participants are truly indicative of their actual preferences.
In short, there are several validity concerns regarding the self-report measures. These validity concerns warrant caution when drawing conclusions from the results in this study. While the positive relationships between costly signaling and both received trust and received cooperation do seem to support the theory, the validity issues limit these relationships to being tentative. Future research is required to substantiate the findings. This future research is preferably done using either behavioral measures that include a monetary incentive, or self-report measures that include multiple items and accurate definitions of trust and cooperation.
Regarding the self-report items on willingness to cooperate and altruism there is an 20
additional issue. As can be seen in Appendix F, the relationship between costly signaling and willingness to cooperate, and the relationship between costly signaling and altruism are no longer significant when item 4 is included in the analysis. This means that the arbitrary choice to exclude item 4 from the main analysis had a large effect on the results. In other words, the relationships are not very robust. This robustness issue and the previously
mentioned validity issues make it unwarranted to draw conclusions based on these self-report measures. The conclusion on the relationship between costly signaling and willingness to cooperate, and on the relationship between costly signaling and altruism will therefore be drawn solely on the basis of the behavioral measures. It can be concluded that there is not enough evidence for a relationship between costly signaling and both willingness to cooperate and altruism.
Based on the discussion above, it can be concluded that two predictions of the costly signaling theory of religion are tentatively supported, while two are not supported. The evidence for the theory is therefore mixed and further research is necessary.
Appendix A - Instructions of the Prisoner’s Dilemma game (translated from Dutch) The first task consists of making one single decision. You will be matched with a random other participant, who has to make exactly the same choice as you. Your choice and the choice of the other will determine both your earnings and the other’s earnings. You will make your choices independently of each other, and probably at different times. Since your
earnings depend on the choice of the other person, it will only later become clear how much money you will have earned in this task.
In a moment, you and the other participant will choose to cooperate or not cooperate with each other. When both of you decide to cooperate, you will both receive 160 neuro. When both of you decide not to cooperate, you will both receive 60 neuro. When you decide to cooperate, and the other decides not to cooperate, you will receive 40 neuro and the other receives 200 neuro. Finally, when you decide not to cooperate, and the other decides to cooperate, you will receive 200 neuro and the other will receive 40 neuro.
Appendix B – Instructions of the Dictator Game (translated from Dutch)
The religious charity
In this task you’ll have the opportunity to make a monetary donation to a charity. The charity in question is The Salvation Army. The Salvation Army is an evangelical movement that provides practical help and professional care to those in need. The money that is donated to Salvation’s Army will go to one of their social projects in The Netherlands. It is possible to donate an amount between 0 and 160 neuro to this charity. You can keep the money you don’t give away for yourself.
The secular charity
In this task you’ll have the opportunity to make a monetary donation to a charity. The charity in question is The Food Bank. The Food Bank is a charitable movement that provides food and water to those that are financially incapable of providing this for themselves. The money that is donated to The Food Bank will go to one of the regional food banks in The
Netherlands. It is possible to donate an amount between 0 and 160 neuro to this charity. You can keep the money you don’t give away for yourself.
Appendix C Table 8 *
Items on the Level of Religiosity of the Participants
Item Statement
1 I see myself as a believer
2 Religion is only a small part of my life **
3 I believe in the existence of God or a higher power * the items are translated from Dutch
** item that is phrased contra-indicatively and was recoded for the statistical analyses
Appendix D Table 9 *
Questionnaire on Demographic Information of the Participant
Item Answering possibilities
Gender Male / Female
Age (Open question)
Highest level of education VMBO / HAVO / VWO / MBO / HBO / WO / other I was born in … The Netherlands / Morocco / Turkey / Surinam / other My mother was born in … The Netherlands / Morocco / Turkey / Surinam / other My father was born in … The Netherlands / Morocco / Turkey / Surinam / other I was raised … In the countryside / In a village / In a city
I grew up with … brothers and/or sisters 0 / 1 / 2 / 3 / 4 or more I am currently in a relationship Yes / No
I see myself as … Protestant / Catholic / Muslim / Jewish / Atheist / other In general, I have enough money to do
the things I want to do
I strongly disagree / I disagree / I neither agree nor disagree / I agree / I strongly agree
* the items are translated from Dutch
Appendix E Table 10
Proportion of Cooperators in the Prisoner’s Dilemma Game, Proportion of Money Donated in the Dictator Game, and Mean Answers on the Self-report Items, by Gender
Gender PD DG Co1 Alt Tru Co2
Female (N=90) 0.76 0.69 3.68 3.81 4.23 3.93 Male (N=45) 0.62 0.61 3.82 3.84 4.18 3.82
Table 11
Proportion of Cooperators in the Prisoner’s Dilemma Game, Proportion of Money Donated in the Dictator Game, and Mean Answers on the Self-report Items, by Environment the
Participants Grew up in
Grew up PD DG Co1 Alt Tru Co2
In the countryside (N=8) 0.75 0.59 3.75 3.38 4.13 3.75 In a village (N=55) 0.69 0.67 3.65 3.85 4.31 3.93 In a city (N=72) 0.72 0.67 3.78 3.85 4.15 3.89
Table 12
Proportion of Cooperators in the Prisoner’s Dilemma Game, Proportion of Money Donated in the Dictator Game, and Mean Answers on the Self-report Items, by Number of Siblings
Siblings PD DG Co1 Alt Tru Co2
0 (N=6) 0.83 0.33 3.67 3.67 4.33 3.83 1 (N=32) 0.72 0.65 3.59 3.91 4.16 3.91 2 (N=34) 0.62 0.59 3.71 3.79 4.24 3.85 3 (N=35) 0.71 0.73 3.77 3.86 4.17 3.91 4+ (N=28) 0.79 0.76 3.86 3.75 4.29 3.93 24
Table 13
Proportion of Cooperators in the Prisoner’s Dilemma Game, Proportion of Money Donated in the Dictator Game, and Mean Answers on the Self-report Items, by Religious Affiliation
Religious Affiliation PD DG Co1 Alt Tru Co2 Protestant (N=83) 0.70 0.72 3.73 3.70 4.12 3.87 Catholic (N=17) 0.76 0.42 3.82 3.94 4.41 4.00 Undefined Christian (N=8) 0.63 0.78 3.50 3.63 4.25 3.88 Muslim (N=12) 0.83 0.69 3.58 4.33 4.33 3.83 Buddhist (N=6) 0.67 0.28 4.00 4.17 4.33 4.00 Other Religion (N=9) 0.67 0.73 3.67 4.00 4.44 4.00 Table 14
Proportion of Cooperators in the Prisoner’s Dilemma Game, Proportion of Money Donated in the Dictator Game, and Mean Answers on the Self-report Items, by Financial Situation
Financial situation PD DG Co1 Alt Tru Co2 1 (N=3) 0.67 0.56 4.00 4.67 4.33 4.33 2 (N=21) 0.71 0.60 3.71 3.90 4.14 4.00 3 (N=20) 0.65 0.77 3.85 4.00 4.25 3.80 4 (N=71) 0.68 0.65 3.61 3.73 4.18 3.80 5 (N=20) 0.90 0.69 4.00 3.75 4.35 4.15 25
Appendix F – Statistical analyses including item 4 Table 15
Beta Coefficients (B), Standard Errors (SE), Odds Ratio’s and Nagelkerke’s R² for the Binary Logistic Regression on Cooperation in the Prisoner’s Dilemma (N=135)
B (SE) Odds Ratio
CS 0.08 (0.06) 1.08 Religiosity 0.03 (0.14) 1.03 Male -0.77 (0.44) 0.46 Financial Situation 0.13 (0.21) 1.14 Catholic 0.94 (0.81) 2.57 Undefined Christian -0.58 (0.82) 0.56 Muslim 0.92 (0.86) 2.51 Buddhist 0.49 (1.11) 1.63 R²N .12
Control variables: Number of Siblings, Youth Environment * p < .05 ** p < .01
Table 16
Beta Coefficients (B), Standard Errors (SE), and R² of the Linear Regression on the Dictator Game Responses (N=135) B (SE) CS -.55 (1.34) Religiosity 9.77** (3.13) Male -9.23 (10.75) Financial Situation -2.02 (5.01) Catholic -16.82 (18.91) Undefined Christian -0.55 (21.137) Muslim -11.47 (17.88) Buddhist -34.26 (25.57) R²N .25
Control variables: Number of Siblings, Youth Environment * p < .05 ** p < .01
Table 17
Beta Coefficients (B), Standard Errors (SE), and R² of the Linear Regressions on the Self-Report Items (N=135)
Co1 Alt Tru Co2
B (SE) B (SE) B (SE) B (SE)
CS 0.03 (0.02) 0.03 (0.02) 0.04* (0.02) 0.05** (0.02) Religiosity -0.06 (0.05) -0.08 (0.05) 0.01 (0.04) -0.03 (0.04) Male 0.14 (0.17) -0.03 (0.15) -0.06 (0.13) -0.13 (0.12) Financial Situation 0.02 (0.08) -0.09 (0.07) 0.02 (0.06) -0.02 (0.06) Catholic 0.40 (0.30) 0.42 (0.27) 0.81** (0.22) 0.54* (0.22) Undefined Christian -0.11 (0.33) -0.09 (0.30) 0.08 (0.25) -0.05 (0.24) Muslim -0.24 (0.28) 0.71** (0.25) 0.35 (0.21) 0.03 (0.20) Buddhist 0.31 (0.40) 0.58 (0.36) 0.65* (0.30) 0.44 (0.29) R² .07 .16 .18 .11
Control variables: Number of Siblings, Youth Environment * p < .05 ** p < .01
References
Ahmed, A. M. (2009). Are religious people more prosocial? A quasi‐experimental study with madrasah pupils in a rural community in India. Journal for the Scientific Study
of Religion, 48(2), 368-374.
Albanese, R., & Van Fleet, D. D. (1985). Rational behavior in groups: The free-riding tendency. The Academy of Management Review, 10(2), 244-255.
Alcorta, C. S., & Sosis, R. (2005). Ritual, emotion, and sacred symbols. Human
Nature, 16(4), 323-359.
Anderson, L. R., & Mellor, J. M. (2009). Religion and cooperation in a public goods experiment. Economics Letters, 105(1), 58-60.
Anderson, L. R., Mellor, J. M., & Milyo, J. (2010). Did the devil make them do it? The effects of religion in public good and trust games. Kyklos, 63(2), 163-175. Bulbulia, J., & Mahoney, A. (2008). Religious solidarity: The hand grenade experiment.
Journal of Cognition and Culture, 8, 295-320.
Chalfant, H. P., & Heller, P. L. (1991). Rural/urban versus regional differences in religiosity.
Review of Religious Research, 33(1), 76-86.
Dawkins, R. (1976). The Selfish Gene. New York City: Oxford University Press. Dictionary.com (2013). Cooperation. Retrieved October 2, 2013, from Dictionary.com
webite: http://dictionary.reference.com/browse/cooperation.
Edgell, P., Gerteis, J., & Hartmann, D. (2006). Atheists as other: Moral boundaries and cultural membership in American society. American Sociological Review, 71(2), 211-234.
Ernest-Jones, M., Nettle, D., & Bateson, M. (2011). Effects of eye images on everyday cooperative behavior: A field experiment. Evolution and Human Behavior, 32(3), 172-178.
Festinger, L. (1957). A Theory of Cognitive Dissonance. Stanford University Press. Field, A. (2009). Discovering Statistics using SPSS, (3rd ed.). New York: Sage.
Gintis, H., Smith, E. A., & Bowles, S. (2001). Costly signaling and cooperation. Journal of
Theoretical Biology, 213(1), 103-119.
Gray, P. (2007). Psychology (5th ed.). New York: Worth.
Hawkes, K. (1993). Why hunter-gatherers work: An ancient version of the problem of public goods. Current Anthropology, 34(4), 341-361.
Henrich, J. (2009). The evolution of costly displays, cooperation, and religion: Credibility enhancing displays and their implications for cultural evolution. Evolution and
Human Behavior, 30, 244-260.
Irons, W. (1996). Morality, religion, and human nature. In W. M. Richardson and W. Wildman (Eds.), Religion and science: History, methods, dialogue (pp. 375-399).
New York: Routledge.
Irons, W. (2001). Religion as a hard-to-fake sign of commitment. In R. M. Nesse (Ed.), Evolution and the capacity for commitment (pp. 292-309). New York: Russel Sage.
Lehmann, L., & Keller, L. (2006). The evolution of cooperation and altruism: A general framework and a classification of models. Journal of evolutionary biology, 19(5), 1365-1376.
Maoz, Z., & Henderson, E. A. (2013). The World Religion Dataset, 1945-2010: Logic, estimates and trends. International Interactions, 39(3), 265-291.
Mook, D. G. (2001). Psychological Research: The Ideas Behind the Methods, (1st ed.). New York: W. W. Norton & Company.
Olson, M. (1971) [1965]. The logic of collective action: Public goods and the theory of
groups (Revised edition ed.). Harvard University Press.
Orbell, J., Goldman, M., Mulford, M., & Dawes, R. (1992). Religion, context, and constraint toward strangers. Rationality and Society, 4(3), 291-307.
Pichon, I., Boccato, G., & Saroglou, V. (2007). Nonconscious influences of religion on prosociality: A priming study. European Journal of Social Psychology, 37, 1032-1045. Paciotti, B., Richerson, P., Baum, B., Lubell, M., Waring, T., McElreath, R., Efferson, C., &
Edsten, E. (2011). Are religious individuals more generous, trusting, and cooperative? An experimental test of the effect of religion on prosociality. Research in Economic
Anthropology, 31, 267-305.
Rousseau, D. M., Sitkin, S. B., Burt, R. S., & Camerer, C. (1998). Not so different after all: A cross-discipline view of trust. Academy of Management Review, 23(3), 393-404. Rowe, C. (1999). Receiver psychology and the evolution of multicomponent signals. Animal
Behaviour, 58(5), 921-931.
Ruffle, B. J., & Sosis, R. (2007). Does it pay to pray? Costly ritual and cooperation. The BE
Journal of Economic Analysis & Policy, 7(1).
Shariff, A. F., Norenzayan, A. (2007). God is watching you: Priming God concepts increases prosocial behavior in an anonymous economic game. Psychological Science, 18(9), 803-809.
Smith, E., & Mackie, D. M. (2007). Social Psychology, (3rd ed.). New York: Psychology Press.
Soler, M. (2012). Costly signaling, ritual and cooperation: evidence from Candomblé, an Afro-Brazilian religion. Evolution and Human Behavior, 33(4), 346-356.
Sosis, R. (2003). Why aren’t we all hutterites? Human nature, 14(2), 91-127.
Sosis, R., & Bressler, E. R. (2003). Cooperation and commune longevity: A test of the costly signaling theory of religion. Cross-Cultural Research, 37(2), 211-239.
Stigler, G. J. (1964). A theory of oligopoly. Journal of Political Economy, 72(1), 44-61. Stock, J. H., & Watson, M. W. (2007). Introduction to Econometrics, (2nd ed.). New Jersey:
Pearson International.
Tan, J. H. W., & Vogel, C. (2008). Religion and trust: An experimental study. Journal of
Economic Psychology, 29(6), 832-848.
West, S. A., El Mouden, C., & Gardner, A. (2011). Sixteen common misconceptions about the evolution of cooperation in humans. Evolution and Human Behavior, 32(4), 231-262.
Willingham, D. T. (2007). Cognition: The thinking animal, (3rd ed.). New Jersey: Pearson International.
World Values Survey Association (2009). World Values Survey 2005. Retrieved from http://www.worldvaluessurvey.org.
