Solidarity in crisis : a cross-country experimental study on the effect of the Great Recession on solidarity

(1)

Master Thesis

Faculty of Economics and Business

Solidarity in Crisis

A cross-country experimental study on the

effect of the Great Recession on solidarity

Alexandra Christopoulou

Supervisor: Prof. dr. Aljaž Ule

(2)

Statement of Originality

This document is written by Alexandra Christopoulou who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

(4)

Abstract

In a cross-country experimental study, conducted in Greece and the Netherlands, 93 subjects played the solidarity game. In both countries, subjects were divided into two treatments; the control and the crisis treatment. The difference between the two treatments is in the instructions provided to the subjects, creating a crisis mindset on those participated in the crisis treatment. We found that Greek subjects behaved significantly less egoistically in the crisis than in the control treatment, while Dutch subjects did not present any difference between the treatments. Additionally, we found that Greek subjects behaved in a significantly more altruistic manner than the Dutch in both treatments. We conclude that the crisis mindset has a positive effect on solidarity.

Keywords: Altruism; Greece; Crisis; Cross-country study; Experimental economics; Great Recession; Solidarity; The Netherlands.

Acknowledgements

First and foremost, I offer my sincerest gratitude to my professor and supervisor Aljaž Ule for his help during my thesis and for his support during my entire studies in Amsterdam. I would like offer my gratitude to my professor Joep Sonnemans for his valuable help in organising the experiment conducted in the CREED laboratory. Above all, I would like to thank my parents, Galani and Giannis for their unconditional love and support, my brother Konstantinos and my sister Diotima for their crucial unflagging belief in me.

(5)

(6)

Abstract ... iv

Acknowledgements ... iv

Contents ... vi

Introduction

... 1

The Great Recession ... 1

Research methodology ... 1

Experimental design ... 1

Solidarity game and response functions ... 2

Main results ... 4

Chapter 1 Literature ... 5

1.1 Great Recession and social preferences ... 5

1.2 The solidarity game ... 6

1.3 Response functions and social preferences ... 6

Chapter 2 Experimental design and procedures ... 7

2.1 Solidarity game ... 7

2.2 Response functions ... 7

2.3 Treatments ... 8

2.4 Experimental procedures ... 8

2.4.1 Differences between the experiments ... 8

2.5 Hypotheses ... 9

2.5.1 Comparison between the two countries ... 9

2.5.2 Comparison between the two treatments ... 9

Chapter 3 Results and discussion ... 10

3.1 Main results ... 10

3.1.1 Comparison between the two countries ... 10

3.1.2 Comparison between the two treatments ... 12

3.2 Response functions ... 13

3.3 Auxiliary results ... 17

3.3.1 Expected gifts ... 17

3.3.2 Gender and education effect ... 18

3.3.3 Comparison between the different number of losers ... 19

Chapter 4 Conclusion ... 20

Bibliography ... 21

Appendix 1 Instructions ... 23

Appendix 2 Decision Forms ... 24

(7)

Appendix 4 Gender effect ... 27

Appendix 5 Education effect ... 28

Appendix 6 Gifts with one and two losers ... 29

Appendix 7 Response functions ... 30

Appendix 8 Econometric analysis ... 34

Appendix 9 Demographics ... 36

List of figures

Figure 1

Flowchart of the developed research methodology ... 3

Figure 2

Comparison between Greece and the Netherlands in the control

treatment ... 11

Figure 3

Comparison between Greece and the Netherlands in the crisis

treatment ... 11

Figure 4

Comparison between control and crisis treatments in Greece ... 13

Figure 5

Comparison between control and crisis treatments in the Netherlands

... 13

Figure 6

Types of response functions observed ... 15

Figure 7

Response functions observed... 16

Figure 8

Correlations of expected and actual gifts ... 26

Figure 9 Frequencies of male and female subjects per behavior ... 27

Figure 10 Frequencies of subjects’ field of study per behavior ... 28

Figure 11 Frequencies of subjects per behavior in case of one and losers

†

.... 29

Figure 12

Response functions: Greece - Control treatment ... 30

Figure 13 Response functions: Greece - Crisis treatment ... 31

Figure 14 Response functions: The Netherlands - Control treatment ... 32

(8)

List of tables

Table 1

Country effect

†

... 4

Table 2

Treatment effect

†

... 4

Table 3 Treatments and number of participants ... 8

Table 4 Comparison between the experiments ... 9

Table 5 Types of behavior ... 10

Table 6

Mean averages of gifts donated ... 10

Table 7

Contingency table of observed types of behavior and response

functions ... 14

Table 8 Response functions observed per treatment ... 17

Table 9

Mean averages of actual and expected gifts

†

... 18

Table 10 Variation of donations based on the number of losers ... 19

Table 11 P-values, Fisher’s exact test for country effect, (two-tailed) ... 34

Table 12

P-values and effect size, Mann-Whitney U-test for country effect,

(one-tailed) ... 34

Table 13 P-values, Fisher’s exact test for treatment effect, (two-tailed) ... 34

Table 14 P-values and effect size, Mann-Whitney U-test for treatment effect,

(one-tailed) ... 34

Table 15 P-values, Fisher’s exact test for gender effect, (two-tailed) ... 35

Table 16 P-values, Fisher’s exact test for education effect, (two-tailed) ... 35

Table 17 Chi-square test for independence between types of behavior and

types of response functions ... 35

Table 18 Gender ... 36

Table 19

Nationality ... 36

(9)

Introduction

The Great Recession

The Great Recession was triggered by the financial crisis which began in 2007. Its causes have been analysed in depth by several economists and its consequences have been in the centre of discussion for many years now. Some countries are already showing signs of recovery, while others are still experiencing its effects on their real economy. There is already evidence, reporting changes in individuals’ behavior, regarding economic decisions, such as consumption, investments, borrowing and saving patterns (Hurd and Rohwedder

(2010), Mian and Sufi (2010), Mian et al. (2013)).

However, the Great Recession, as any other previous economic crisis, resulted (arguably, still is resulting) in behavioral changes that go beyond economic indicators. Long term unemployment and limited equity within a household are not only influencing economic decisions driven by bounded materiality but have a deeper psychological impact on individuals’ perceptions and emotions (Clark and Oswald (1994), Di Tella et al. (2001), Farré et al. (2015)). The generated inequality, poverty, instability and uncertainty are causing observable changes in decision making but more importantly they are changing the motives behind each decision. A different status quo could be forming, leading to rationally individualistic or on the contrary, to more cooperative societies.

Besides the sociological impact of this change in the existing status quo, a more cooperative society could absorb better the negative shocks of the economy. Social cohesion can alleviate the consequences of an economic crisis, substituting in many cases the role of the official institutions in inertia. Cooperation is thus of crucial importance, especially between the lower and middle strata of the society, which are affected the most. This cooperation often results in the formation of solidarity networks. The purpose of this study is to understand this phenomenon and to research the effect of the Great Recession on solidarity.

Research methodology

When asking what the effect of the crisis on solidarity is, there are many research methods – still constrained inside the experimental economics scope– that one could follow. Figure 1 gives an overview of the logical path followed in order to reach the developed research methodology. The subsequent sections discuss in detail the process of thinking that resulted in an experimental design, capable to answer our research question. Information regarding the experimental design per se and the procedures followed can be found in Chapter 2 .

Experimental design

The first step was to identify a suitable population to be studied. An analysis of such a population that has experienced the Great Recession and still is experiencing its consequences was considered suitable. Initially, the answer on how to address the research question was a comparison study within the same county and subsequently Greece was chosen as a good fit.

(10)

The idea was to compare experimental data collected before the crisis with recent data, collected after the crisis. However, the methodology had to be adjusted, since such historical data, regarding social preferences, let alone solidarity, were not available in case of Greece. Subsequently, to overcome the pre-mentioned difficulties, a comparison study between Greek subjects and subjects from a country which is not affected by the crisis, at least not in a significant magnitude, was conceived. The Netherlands was identified as such a country and a cross-country experimental design was developed.

Designing an identical experiment with one treatment, which would be repeated in the two countries, was considered problematic. The comparison would reflect mostly the differences in cultural characteristics and would not capture the crisis effect on the subjects’ social preferences. As a result, the final experimental design was conceived with two treatments, both to be repeated in each country, the control and the crisis treatment. In that way, an observation of a between the countries effect and a between the treatments effect (in the same country) is allowed.

Solidarity game and response functions

After developing the experimental design and the treatments, the game theoretical approach was still to be decided. According to Selten and Ockenfels (1998), the purpose of their experimental setup was “to create a situation in which subjects can show solidarity in the sense that they are willing to help others who by chance came to a much worse position than they themselves”, (p. 518). Their design, the solidarity game as played by their subjects, as well as their perception on solidarity as a variable, are completely in line with our study. Consequently, the solidarity game was replicated, with minor differences, as it was played in their experiments.

Furthermore, a primary interpretation of the variation in the motives behind solidarity is attempted in this study with the use of response functions as in Bolle et al. 2012.

(11)

(12)

Main results

The results of this study are analysed extensively in Chapter 3 and they are divided further into three sections. Section 3.1, is presenting and analysing the main results regarding the solidarity exhibited by the subjects. In section 3.2, the motives behind the observed behaviors in the previous subsection are analysed with the use of response functions. Section 3.3 is presenting an analysis of auxiliary results and conclusions for which no hypotheses were formed.

In the experiments conducted, subjects played the solidarity game with one and two losers in their group. The extended analysis on Chapter 3 presents the results separately on those two cases, however, in this introductory analysis the average donations will be used for simplification purposes.

Table 1 gives an overview of the country effect showing that subjects in Greece exhibited different behavior than those in the Netherlands, donating higher gifts.

Table 2 gives an overview of the treatment effect. Subjects in Greece exhibited different behavior between the two treatments, donating higher gifts in the crisis treatment. On the other hand, Dutch subjects did not present significant difference between the two treatments.

Table 1 Country effect†

Control treatment Crisis treatment

Egoistic behavior Weak altruistic behavior Strong altruistic behavior Egoistic behavior Weak altruistic behavior Strong altruistic behavior Netherlands 72.9% 25.0% 2.1% 77.1% 18.8% 4.2% Greece 25.0% 54.2% 20.8% 16.7% 23.8% 59.5% Difference -47.9% +29.2% +18.7% -60.4% +5.0% +55.3% †

the percentages were calculated based on the average donations between the case of one and two losers

Table 2 Treatment effect†

Greece Netherlands Egoistic behavior Weak altruistic behavior Strong altruistic behavior Egoistic behavior Weak altruistic behavior Strong altruistic behavior Control treatment 25.0% 54.2% 20.8% 72.9% 25.0% 2.1% Crisis treatment 16.7% 23.8% 59.5% 77.1% 18.8% 4.2% Difference -8.3% -30.4% +38.7% +4.2% -6.2% +2.1%

† _{the percentages were calculated based on the average donations between the case} of one and two losers

(13)

Chapter 1 Literature

1.1 Great Recession and social preferences

Due to the limited body of knowledge regarding the impact of Great Recession specifically on solidarity, this section discusses the effect of Great Recession on social preferences in a broader way.

Margalit (2013), conducted a panel data study, in which he compared pre-crisis data, collected in 2007, with post-crisis data from 2013. He found that when personal economic circumstances are changing, there is a strong impact on individuals’ preferences regarding welfare spending. More specifically, his study finds that the effect of a negative economic shock, such as a job loss, is increasing support for welfare sharply, while when employment returns to its original levels, support for welfare is decreasing. Thus, Margalit’s study concludes that there is an inverted U-shaped relationship between the two variables.

Rakopoulos (2013), in a field research he conducted in Athens, Greece, he explored the anti-middleman organisations and their rise after the crisis. He found that local communities are cooperating to eliminate the middleman between the producers and the consumers in food distribution networks. In their point of view, economies should function under reciprocal relationships between cooperatives and markets and solidarity among people.

Sotiropoulos and Bourikos (2014), found evidence that after the crisis, social solidarity has increased in Greece. Their field research, shows that since 2010, the economic crisis has led to an increase of organizations, informal social networks and self-help groups. Moreover, NGO’s, active in the social solidarity sector, are relying more on volunteers, as the government no longer supports them financially to the same level as before the crisis and private sources of funding are limited. Furthermore, “both formal and informal organizations claim that there has been an increase in the numbers of volunteers offering their time and skills to a public cause, such as assistance to the victims of the crisis”, according to Sotiropoulos and Bourikos (2014, p. 51).

Ananyev and Guriev (2015), researched the effect of income on trust, in Russia, comparing data in the end of 2008 and the beginning of 2009. They found that the decrease on income results in a decrease in trust levels. Moreover, their study is taking into account that the effect of the crisis has been uneven among Russian regions, concluding that regions more affected by the crisis are experiencing a larger decline in trust than the others.

Fisman et al. (2015), studied the effect of the Great Recession on distributional preferences. In an experimental study with three treatments, their subjects played a modification of the dictator game. The participants had to allocate tokens between themselves and another subject unknown to them. The difference between the first two treatments is that the first one is using data collected in an experiment conducted before the Great Recession (2004), while the second one is using data collected in 2011. Their last treatment, namely loss treatment, is using data collected after the crisis (2010-2011) but is also simulating recessionary conditions in the laboratory. The authors found that subjects participating in the loss treatment exhibited higher selfishness than subjects participating in the other treatments.

(14)

1.2 The solidarity game

Selten and Ockenfels (1998), presented in an experimental study in which their subjects played a variation of the dictator game, the solidarity game as we know it today. As they explained, the solidarity game, as they design it, it adds to the dictator game the missing motivational element of trust in reciprocity. They found that only 25 out of their 120 subjects (21%) presented an “egotistical” behavior, not donating any amount of money to the losers of their group. The behaviors observed by their subjects were divided in categories based on the amount they donated in case there was one or two losers in their group. The majority of their subjects (52%), gave the same amount in both loser conditions (“fixed total sacrifice”), keeping the total amount of their ‘sacrifice’ constant, while 16% of their subjects presented a “fixed gift to loser” behavior, donating the same amount in both losers in their group. The rest of their subjects (11%) presented an “intermediate” behavior. In section 3.3.3, we compare our results to those of Selten and Ockenfels (1998) and present them based on their classification (see Table 10).

Ockenfels and Weimann (1999), conducted a cross-culture study of public good and solidarity experiments. Comparing the behavior of subjects from East and West Germany, they found that eastern subjects behave in a more selfish way than the western. Brosig-Koch et al. (2011), in an experimental study, compared again subjects from East and West Germany. They found that eastern subjects are still behaving more egoistically than the western and the differences persist even after 20 years of the fall of the Berlin wall.

Büchner et al. (2007), replicated the solidarity game as presented in Selten and Ockenfels (1998), with the purpose to test the robustness of the “fixed total sacrifice” effect they found. Büchner et al. (2007), only partially verified the results of Selten and Ockenfelts (1998). The authors found that the behavior of the donors is dependent on their endowment. When the endowment is kept constant, between the cases of one and two losers, they behave in a more egoistical way, while when it increases - in the case of two losers - then the subjects behave in more equitable way.

1.3 Response functions and social preferences

Bolle et al. (2012), analysed experimental solidarity games, researching the motives behind solidarity. After playing the solidarity game with one loser and two winners, their subjects were asked what they would give to the loser of their group, for every amount of the gift the other winner gave. Thus, different types of response functions were formatted and afterwards categorised based on their convexity and monotonicity. The authors concluded that different motives occur behind solidarity and diverse combinations of those may coexist. Further analysis of our results, regarding the derivation of the motives and an analysis of the shapes of the response functions can be found in section 3.2.

(15)

Chapter 2 Experimental design and procedures

Overall, 93 university students participated in two laboratory experiments conducted in Amsterdam and Athens.

The experimental design consists of three parts, which correspond to three decisions on behalf of the subjects and has two treatments, the control and the crisis treatment. In the first part subjects played the solidarity game as in as in Selten and Ockenfels (1998). The second part's purpose was to identify their expectations regarding the donations of other subjects. The third part aimed in understanding the motives behind the behaviors observed in the first part. In the end of the experiment questions of demographic nature were asked.

Section 2.1 describes the first two parts of the experimental design while section 2.2 the third one. Section 2.3 discusses the treatments while section 2.5 the hypotheses formed prior the experiments were conducted.

Section 2.4 describes the experimental procedures followed. Both experiments had an identical experimental design, although due to technical issues, some small differences occurred procedure-wise. However, treatments are comparable, as those differences do not result in changes in the subjects’ behavior. Those differences are analysed further in subsection 2.4.1.

The instructions provided to the subjects, as well as the decision forms can be found in Appendix 1 and Appendix 2 respectively.

2.1 Solidarity game

In the first part of the experiment, subjects played the solidarity game as in Selten and Ockenfels (1998). They were divided randomly in groups of three and with probability 2/3 they could win 10 €. Thus, it could happen that all three, two, one or none members of the group win. Subsequently, the subjects, without knowing if they will win or not, they were asked, how much they are willing to give to the loser(s) in their group as a gift, in both the cases of one and two losers, (Appendix 2 - Decision form 1). Then, in the second part of the experiment, subjects were asked about their expectations, regarding the gifts of the other participants, (Appendix 2 - Decision form 2), decision that was not incentivised.

As subjects were informed in the instructions, one group would randomly be selected for payment in the end of the experiment.

2.2 Response functions

The motives behind the observed behavior in the first part of the experiment were expected to vary. As a result, a further analysis was considered necessary to understand those motives, rendering the third part of the experiment. As in Bolle et al. (2012), subjects were asked if they won and there was only one loser in their group, what they would give to him, knowing the amount of the gift the other winner gave.

Thus, for 11 possible outcomes representing the gift of the other winner, they had to specify their own, such that . The extreme cases of 0 and 10 € and their integer

(16)

interval were included, such that . The third part of the experiment (as

the second) was not incentivized.

2.3 Treatments

Both experiments, in both countries, consist of two treatments; the control and the crisis treatment. The only difference between the two treatments is in the instructions provided to the subjects and not in the questions asked. More specifically, in the control treatment subjects received instructions only explaining the rules, while subjects in the crisis treatment were additionally informed that 'due to the recent economic crisis resources are limited'. It was considered prior the experiment that this phrase will be taken into account by the subjects in both countries, altering their behavior between the treatments significantly. A more extensive reference to the crisis would result into a framing effect, thus it was avoided. The availability heuristic is an idea originally explored by Tversky and Kahneman (1973). According to them, a person is making a judgement decision based on the availability of relevant instances which he can recall into his mind, in order to evaluate the frequency of classes or the probability of events.

2.4 Experimental procedures

The experiments were conducted with a month apart and on four separate days (one day per treatment in each country). The first experiment, was conducted at the CREED laboratory of the University of Amsterdam and had 24 participants assigned to the control treatment and another 24 to the crisis treatment. The second experiment was conducted at the Athens University of Economics and Business and had 24 participants in the control and 21 in the crisis treatment. Nobody participated twice in the experiments and anonymity was kept. In the end of each experiment, payments were made privately to the subjects in a separate room. All subjects were given a show up fee for their participation in the experiments, regardless of the outcome, to ensure sufficient participation. All participants were undergraduate students and the average age was 23.8, (22.5 in the experiment conducted in Athens and 25 in the experiment conducted in Amsterdam). Table 3 summarizes the sample. Additional information on the demographics of the sample can be found in Appendix 9.

Table 3 Treatments and number of participants

Greece Netherlands

Control 24 24

Crisis 21 24

Total N = 93

2.4.1 Differences between the experiments

Although the experiments had an identical experimental design, the lack of a laboratory in Greece resulted in small differences between them. More specifically, in Greece, in order to inform the participants for the experiment, flyers were distributed one hour prior each treatment at the university cafeteria. Since the experiment conducted in Greece was not

(17)

computerised, instructions and decision forms were provided in paper. For randomization purposes (6-sided, 14-sided and 16-sided) dices, were used. Moreover, while CREED laboratory ensures privacy and anonymity with its cubicles, to ensure the same result in the classroom experiment conducted in Greece, empty seats were kept between the subjects. Table 4 gives an overview of the differences between the two experiments.

Table 4 Comparison between the experiments

Announcement flyers emails Location classroom laboratory Instructions pen and paper computerized Decision forms pen and paper computerized Randomization dice computerized

In addition, it must be noted that subjects in the experiment conducted in the Netherlands are more experienced than those in Greece as they are frequently participating in experiments at the CREED laboratory. Moreover, the laboratory experiment conducted in the Netherlands was part of a larger experiment as subjects participated in subsequent small experiments in the same day.

2.5 Hypotheses

Based on the literature discussed in Chapter 1, the intuitional belief that creating a crisis mindset will result in an altruistic behavior by Greek subjects, as well as evidence from solidarity movements rising after the crisis in Greece and the absence of such in the Netherlands, the following hypotheses were formulated.

2.5.1 Comparison between the two countries

Hypothesis 1: Solidarity will be higher in Greece than in the Netherlands, comparing the control and crisis treatments respectively.

2.5.2 Comparison between the two treatments

Hypothesis 2: Solidarity will be higher in the crisis treatment than in the control treatment, in both countries.

(18)

Chapter 3 Results and discussion

3.1 Main results

The following subsections, 3.1.1 and 3.1.2, discuss the main results of the two experiments conducted, as well as the hypotheses formed in section 2.5.

3.1.1 Comparison between the two countries

Initially, subjects were categorised into three types of behavior: egoistic, weak altruistic and

strong altruistic behavior. It was considered necessary to differentiate in the categorisation

between the cases of one and two losers, since in case of two losers the gift donated is per loser and thus the subject’s loss is higher. For instance, if a subject donates 3 € when there is one loser in his team, his loss would be only 3 € but if he donates 3 €, when there are two losers in his team, then his loss would be 6 €.

Consequently, a subject is exhibiting egoistic behavior if he does not donate any amount to his team members, irrelevantly of the number of losers in his team. Weak altruistic behavior have the subjects who donate from 1 to 3 €, in case there is one loser in their team, or from 1 to 2 €, in case there are two losers in their team. Strong altruist behavior have the subjects whose gifts range from 4 to 10 €, in case there is one loser in their team, or from 3 to 10 € in case there are two losers in their team.

Table 5 shows the types of behavior corresponding to the range of gifts donated, in case of one and two losers. Table 6 gives an overview of the mean averages of the gifts donated per country, treatment and number of losers.

Table 5 Types of behavior

Behavior Gifts with one loser Gifts with two losers

Egoistic behavior 0 € 0 € Weak altruistic behavior [1-3] € [1-2] € Strong altruistic behavior [4-10] € [3-10] €

Table 6 Mean averages of gifts donated

Treatments Greece Netherlands

Control treatment with one loser 2.25 € 0.58 € Crisis treatment with one loser 3.29 € 0.25 € Control treatment with two losers 1.71 € 0.46 € Crisis treatment with two losers 2.48 € 0.54 €

(19)

Figure 2 and Figure 3 show the frequencies of the participants per behavior type, (egoistic, weak altruistic and strong altruistic) in case of one and two losers, in the control and crisis treatment respectively.

Figure 2 Comparison between Greece and the Netherlands in the control treatment

Figure 3 Comparison between Greece and the Netherlands in the crisis treatment

Dutch subjects show less solidarity as more than 70% exhibited egoistic behavior in all treatments (for both one and two losers), while the majority of Greek subjects exhibited weak altruistic behavior in the control treatment (54.2% for both one and two losers) and strong altruistic behavior in the crisis treatment (61.9% for one loser and 57.1% for two losers).

Result 1:

Subjects in Greece exhibited different behavior than those in the Netherlands, donating higher gifts in both control and crisis treatment. Thus, hypothesis 1 is verified.

The null hypothesis that Greek and Dutch subjects are equally likely to be egoistic, weakly altruistic and strongly altruistic has been rejected. Fisher’s exact test (for 2x3 contingency tables, two-tailed), yields the results significant, as for the control treatment and

6 13 5 17 7 0 0 5 10 15 20

egoistic behavior weak altruistic behavior strong altruistic behavior fr eq ue nc y

Frequencies with one loser

GR NL 6 13 5 18 5 1 0 5 10 15 20

Frequencies with two losers

GR NL 4 4 13 19 5 0 0 5 10 15 20

GR NL 3 6 12 18 4 2 0 5 10 15 20

GR NL

(20)

for the crisis treatment. A one-tailed Mann-Whitney U-test supports the view that Dutch subjects gave lower amounts yielding the results significant; for the control treatment and for the crisis treatment. (In case of one loser, in the crisis treatment results were highly significant as ). Table 11 and Table 12, give an overview of all the P-values rendered by Fisher’s exact test and Mann-Whitney U-test respectively. Both tables can be found in Appendix 8.

3.1.2 Comparison between the two treatments

Result 2:

Subjects in Greece exhibited different behavior between control and crisis treatment, donating higher gifts in the crisis treatment. On the other hand, Dutch subjects did not present any difference between the treatments. Thus, hypothesis 2 is verified partially. In case of Greece, the null hypothesis that subjects in the control and crisis treatment are equally likely to be egoistic, weakly altruistic and strongly altruistic has been rejected. Fisher’s exact test (for 2x3 contingency tables, two-tailed), yields the results significant as , in case of one loser and , in case of two losers. A one-tailed Mann-Whitney U-test supports the view that Greek subjects gave lower amounts in the control treatment, yielding the results significant; , in case of one loser and , in case of two losers. In case of the Netherlands, the null hypothesis that subjects in the control and crisis treatment are equally likely to be egoistic, weakly altruistic and strongly altruistic has not been rejected.

Table 13 and Table 14, give an overview of all the P-values rendered by Fisher’s exact test and Mann-Whitney U-test respectively. Both tables can be found in Appendix 8.

The majority of Greek subjects exhibited weak altruistic behavior in the control treatment (for both one and two losers), while in the crisis treatment the majority of them expressed strong altruist behavior (for both one and two losers). Moreover, the number of Greek subjects expressing egoistic behavior in the crisis treatment (19% for one loser and 14.3% for two losers), is less than the one in the control treatment (25% for both one and two losers). Dutch subjects had no alteration in their behavior as in every treatment, for any number of losers, the majority of them exhibited egoistic behavior.

Figure 4 and Figure 5 show the frequencies of the participants per behavior type, (egoistic, weak altruistic and strong altruistic) in case of one and two losers, in case of Greece and the Netherlands respectively.

(21)

Figure 4 Comparison between control and crisis treatments in Greece

Figure 5 Comparison between control and crisis treatments in the Netherlands

3.2 Response functions

The response functions, derived from the third part of the experiment, were initially divided in six categories as in Bolle et al. (2012): constant, weakly increasing, weakly decreasing, quasi-convex, quasi-concave and other. The constant category was divided further into constant zero and constant non-zero, rendering seven types of response

functions in total.

Constant zero response functions, ( ), are representing egoistic motives,

while constant non zero response functions, ( ), describe subjects experiencing warm glow. A weakly increasing response function is when the subject is increasing his donation as the other winner does, such as for example when exhibiting guild towards him. A subject with a weakly decreasing response function could be driven by altruism and/or guild towards the loser in his group.

When a subject is exhibiting a quasi-concave response function, he is increasing his donations, as the other winner does, up to a point and afterwards decreasing them. This behavior behind quasi-concave functions, according to Bolle et al. (2012), could be

6 13 5 4 4 13 0 5 10 15 20

control crisis 6 13 5 3 6 12 0 5 10 15 20

control crisis 17 7 0 19 5 0 0 5 10 15 20

control crisis 18 5 1 18 4 2 0 5 10 15 20

control crisis

(22)

interpreted as combination between envy towards the loser of the group and guilt towards the other winner and/or altruism. On the contrary, a subject with a quasi-convex response function, is decreasing his donations, as the other winner does, up to a point and afterwards he is increasing them. Altruism or/and guilt towards the loser in his group could explain the increasing response initially. Guilt towards the other winner in his group could explain the increasing part of the quasi-convex function.

To evaluate if the response functions could interpret the results obtained from the first part of the experiment, a chi-square test for independence was performed between the types of behavior, as they were described in Table 5, and the seven types of response functions as they were described in the first paragraph of this chapter. A further analysis is applicable as we reject the null hypothesis that the two variables are independent; .

Table 7 describes the frequency distribution of the observed response functions in relation to the observed behavior types in the solidarity game with one loser. The expected frequencies and chi-square statistic values can be found in Appendix 8.

Table 7 Contingency table of observed types of behavior and response functions

Constant non-zero Constant zero Weakly increasing Weakly decreasing Quasi-convex Quasi-concave Other Egoistic behavior 0 28 13 2 0 3 0 Weak altruistic behavior 4 1 8 8 1 2 5 Strong altruistic behavior 3 0 5 6 1 0 3

Figure 6 and Figure 7 give an overview of the response functions observed overall, in both countries and for both treatments. A further breakdown of the response functions per country and treatment is available in Appendix 7.

(23)

(24)

(25)

Table 8 gives an overview of the response functions observed in the experiments conducted in Greece and the Netherlands for both treatments. In the Netherlands, the majority of subjects exhibited a ‘constant zero’ response function in both control and crisis treatment. This absence of alteration in the response functions observed between the treatments, is in line with the absence of treatment effect observed in subsection 3.1.2. Subjects exhibiting egoistic behavior were consistent in their response functions, constantly donating 0 € to the loser of their group, regardless what the other winner donated.

Table 8 Response functions observed per treatment

Type of response function Greece Netherlands Control treatment Crisis treatment Control treatment Crisis treatment Constant non-zero 8.3% 19.0% 0.0% 4.2% Constant zero 8.3% 19.0% 41.7% 54.2% Weakly increasing 33.3% 14.3% 33.3% 29.2% Weakly decreasing 37.5% 19.0% 8.3% 4.2% Quasi-convex 4.2% 4.8% 0.0% 0.0% Quasi-concave 0.0% 0.0% 12.5% 8.3% Other 8.3% 23.8% 4.2% 0.0%

In case of Greece, the majority of the subjects in the control treatment is almost equally divided between weakly increasing and weakly decreasing response functions. In the crisis treatment, the results are more complicated as subjects’ response functions vary significantly among constant non-zero, constant zero, weakly increasing, weakly decreasing and other. As the results in the previous analysis regarding the treatment effect indicate, egoistic behavior is observed less in the crisis treatment than the control treatment, fact which does not explain the rise in the constant zero observed response functions.

The shift of the majority of the Greek subjects from weak altruistic behavior, in the control treatment, to strong altruistic behavior, in the crisis treatment, would make us expect an increase in the number of the weakly decreasing response functions, however this is not the case. On the other hand, the shift is in line with the increase of constant non-zero response functions in the crisis treatment.

3.3 Auxiliary results

The following subsections 3.3.1 - 3.3.3, analyse results derived from the experiments conducted for which no hypotheses were formed a priori.

3.3.1 Expected gifts

In the Netherlands, expected gifts in both control and crisis treatment (for both one and two losers) were higher than the actual gifts. On the other hand, in Greece, subjects responded differently between the two treatments. Expected gifts remained higher than the actual gifts in the control treatment, while there were lower in the crisis treatment, insignificantly of the number of losers. Thus, participants in Greece expected others to be less altruistic than

(26)

themselves in the crisis treatment. Table 9 gives an overview of the mean averages of actual and expected gifts per country, treatment and number of losers.

Table 9 Mean averages of actual and expected gifts†

Treatments

Actual gift Expected gift Actual gift Expected gift

Control treatment with one loser 2.25 2.33 0.58 0.83 Control treatment with two losers 1.71 2.17 0.46 0.50 Crisis treatment with one loser 3.29 2.24 0.25 0.83 Crisis treatment with two losers 2.48 1.62 0.54 1.17

†_{Mean averages of actual and expected gifts are in €}

Spearman rank correlation coefficient varied considerably between the countries and the treatments; , (rounded to two decimal points). In case of the Netherlands, results are considered statistically significant in both treatments (for both one and two losers). In case of Greece, results are only considered statistically significant in the control treatment (for both one and two losers); , (one-tailed test). The distributions of the deviations from the actual mean, – , and from the expected mean, – , ( , where , the number of losers), spearman rank correlation coefficients and P-values (one-tailed), can be found in Appendix 3.

Control treatments in both countries have stronger correlations between expected and actual gifts than the crisis treatments; and , for both countries. The reason could be, possibly, that the crisis factor was interpreted differently by the subjects.

For instance, a subject could make the conclusion that he is affected by the crisis more than the other participants and thus expect less donations from them. Another interpretation could be that he is not altruistic enough, because the crisis did not affect him as much as the other participants and predict that their gifts will be higher. Thus, when a subject in the control treatment is forming expectations about his team members, he is only comparing them with himself on the basis of solidarity solely. On the other hand, a subject in the crisis treatment, has to consider also the effect of the crisis on himself in comparison to his team members. Increasing the number of variables that have to be taken into consideration is weakening the correlations.

3.3.2 Gender and education effect

Gender effect has not been observed nor has education effect. The null hypothesis that men and women are equally likely to be egoistic, weakly altruistic and strongly altruistic has not been rejected. Similarly, the null hypothesis that participants who had economic studies and participants who had other studies, are equally likely to be egoistic, weakly altruistic and strongly altruistic has not been rejected as well. Table 15 and Table 16 give an overview of

(27)

the P-values rendered by Fisher’s exact test (for 2x3 contingency tables, two-tailed), for gender and education effect respectively. Both tables can be found in Appendix 8.

The frequencies of the male and female subjects per behavior, (egoistic, weak altruistic and strong altruistic), can be found in Appendix 4. The frequencies of the participants who had economic studies and those who had other studies, per behavior, (egoistic, weak altruistic and strong altruistic), can be found in Appendix 5.

3.3.3 Comparison between the different number of losers

Selten and Ockenfels (1998) classify their subjects’ donations based on how donations between one and two losers relate to each other. While the majority of their subjects gave the same total amount in both conditions (one or two losers) either on “exact” or “up to rounding”, (52%), this result is not replicated in none of the treatments in our experiment. On the contrary, “intermediate behavior” is observed on the majority of participants in Greece in both treatments. In the Netherlands the results do not replicate as well, as “egoistical behavior” is observed on the majority of the participants in both treatments.

Table 10 gives an overview of the results based on the classification Selten and Ockenfels (1998) presented.

Table 10 Variation of donations based on the number of losers

In our study, the number of losers did not influence the participants’ behavior. The null hypothesis that participants are equally likely to be egoistic, weakly altruistic and strongly altruistic, in case of one and two losers, has not been rejected in any of the treatments. In case of Greece, Fisher’s exact test (for 2x3 contingency tables, two-tailed), in the control treatment; and , in the crisis treatment. In case of the Netherlands, in the control treatment; and , in the crisis treatment.

The frequencies of the participants per behavior, (egoistic, weak altruistic and strong altruistic) in case of one and two losers, can be found in Appendix 6.

Egoistical Fixed total sacrifice Intermediate Fixed gift to loser

Other Exact Up to rounding

Treatments g1=g2=0 g1=2g2>0 2g2>g1>0 g1>2g2>0 2g2=g1>g2>0 g1=g2>0 -

Greece, control treatment 25.0% 8.3% 0.0% 4.2% 33.3% 20.8% 8.3% Greece, crisis treatment 14.3% 4.8% 0.0% 4.8% 42.9% 28.6% 4.8% Netherlands, control treatment 70.8% 12.5% 0.0% 0.0% 4.2% 4.2% 8.3% Netherlands, crisis treatment 75.0% 0.0% 0.0% 0.0% 0.0% 12.5% 12.5%

(28)

Chapter 4 Conclusion

In this study we researched the effect of the Great Recession on solidarity. After conducting experiments in Greece and the Netherlands, we found that Greek subjects exhibited a significantly more altruistic behavior than the Dutch in both treatments. We also found that while there is a difference between the crisis and the control treatment in Greece, Dutch subjects exhibited no variation in their behavior between the treatments. Based on our findings, we conclude that Great Recession has a positive effect on solidarity.

However, we understand the limitations of our study and the limitations of cross-country experimental studies in general. When comparing subjects from Greece and the Netherlands, besides the observed variable (behaviour exhibited) and the independent variable (control and crisis treatment), variables such as the cultural differences between the subjects remain uncontrolled.

Cultural differences as well as preceding historical circumstances, do naturally play a role in the observed variable. A natural tendency to moral stances (such as altruism and by extension, solidarity) is more prevalent in some countries than others. Thus, the spectrum of their acceptance and the extent of their diffusion when an economic crisis arises will vary and result in different behaviors.

The Great Recession had a global impact, however the crisis in each country had its differences. The depth of the crisis, the magnitude of the austerity measurements and the consequences in the true economy in Greece render the Greek crisis to be a particular situation. Hence, the effect of the Great Recession on solidarity found in our research cannot be generalised without further considerations to other countries.

(29)

Bibliography

Ananyev, M., & Guriev, S. (2015), Effect of income on trust: Evidence from the 2009 crisis in Russia, CEPR Discussion Paper (No. 10354), Centre for Economic Policy Research, London.

Bolle, F., Breitmoser, Y., Heimel, J. & Vogel, C. (2012), Multiple Motives of Pro-social Behavior: Evidence from the Solidarity Game, Theory and Decision, 72(3), 303-321.

Brosig-Koch, J., Helbach, C., Ockenfels, A., & Weimann, J. (2011), Still different after all these years: Solidarity behavior in East and West Germany, Journal of Public

Economics, 95(11), 1373-1376.

Büchner, S., Coricelli, G. & Greiner, B. (2007), Self-centered and Other-regarding Behavior in the Solidarity Game, Journal of Economic Behavior & Organization, 62(2), 293-303.

Clark, A. E., & Oswald, A. J. (1994), Unhappiness and unemployment, The Economic

Journal, 648-659.

Croson, R., & Gneezy, U. (2009), Gender differences in preferences, Journal of Economic

literature, 448-474.

De Oliveira, A., Eckel, C. & Croson, R. (2012), The Stability of Social Preferences in a Low-income Neighborhood, Southern Economic Journal, 79(1), 15-45.

Di Tella, R., MacCulloch, R. J., & Oswald, A. J. (2001), Preferences over inflation and unemployment: Evidence from surveys of happiness, American Economic Review, 335-341. Fafchamps, M. (1992), Solidarity Networks in Preindustrial Societies: Rational Peasants with a Moral Economy, Economic Development and Cultural Change, 147-174.

Farré, L., Fasani, F., & Mueller, H. (2015), Feeling Useless: The Effect of Unemployment on Mental Health in the Great Recession (No. 9235), IZA Discussion Papers.

Fisman, R., Jakiela, P., & Kariv, S. (2015), How Did Distributional Preferences Change During the Great Recession?, Journal of Public Economics, 128, 84-95.

Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., McElreath, R., Alvard, M., Barr, A., Ensminger, J., Smith, N., Hill, K., Gil-White, F., Gurven, M., Marlowe, F., Patton, J. & Tracer, D. (2005), “Economic man” in cross-cultural perspective: Behavioral experiments in 15 small-scale societies, Behavioral and brain sciences, 28(06), 795-815.

Hurd, M. D., & Rohwedder, S. (2010), Effects of the financial crisis and great recession on American households, (No. w16407), National Bureau of Economic Research.

(30)

Kuziemko, I., Norton, M. I., Saez, E., & Stantcheva, S. (2015), How elastic are preferences for redistribution? Evidence from Randomized Survey Experiments, American Economic

Review, 105(4), 1478-1508.

Levitt, S. & List, J. (2007), What do Laboratory Experiments Measuring Social Preferences Reveal About the Real World?, The Journal of Economic Perspectives, 21(2), 153-174. Margalit, Y. (2013), Explaining social policy preferences: Evidence from the Great Recession, American Political Science Review, 107(01), 80-103.

Mian, A. R., Rao, K., & Sufi, A. (2013), Household balance sheets, consumption, and the economic slump, Chicago Booth Research Paper, (No 13-42).

Mian, A., & Sufi, A. (2010), The great recession: Lessons from microeconomic data, The

American Economic Review, 51-56.

Ockenfels, A. & Weimann, J. (1999), Types and Patterns: An Experimental East-West-German Comparison of Cooperation and Solidarity, Journal of Public Economics, 71(2), 275-287.

Rakopoulos, T. (2013), Responding to the crisis: food co-operatives and the solidarity economy in Greece, Anthropology Southern Africa, 36(3-4), 102-107.

Selten, R. & Ockenfels, A. (1998), An Experimental Solidarity Game, Journal of Economic

Behavior & Organization, 34(4), 517-539.

Sotiropoulos, D. A., & Bourikos, D. (2014), Economic Crisis, Social Solidarity and the Voluntary Sector in Greece, Journal of Power, Politics & Governance, 2(2), 33-53.

Tversky, A., & Kahneman, D. (1973), Availability: A heuristic for judging frequency and probability, Cognitive Psychology, 5(2), 207-232.

(31)

Appendix 1 Instructions

Control Treatment

In this experiment you have the opportunity to win 10 Euros with probability 66%.

Groups of 3 participants will be formed. The composition of your group remains the same throughout the experiment. You will not know who is in your group.

All members of this group have individually 66% probability to win 10 Euros, so it can happen that all members, 2, 1 or none of the members win 10 Euros. In the end, randomly, one group will be selected for payment.

Before all that, you have to decide how much you are willing to hand over voluntarily from your winnings to the loser(s) in your group, in the case you are one of the winners. This will affect your payment. If you win, you will receive 10 Euros minus your voluntarily gifts to the loser(s) in your group. If you lose, you receive the gifts of the other members in your group who won.

Crisis Treatment

In this experiment you have the opportunity to win 10 Euros with probability 66%.

Groups of 3 participants will be formed. The composition of your group remains the same throughout the experiment. You will not know who is in your group.

All members of this group have individually 66% probability to win 10 Euros, so it can happen that all members, 2, 1 or none of the members win 10 Euros. Due to the recent economic crisis resources are limited. Thus, in the end, randomly, one group will be selected for payment.

Before all that, you have to decide how much you are willing to hand over voluntarily from your winnings to the loser(s) in your group, in the case you are one of the winners. This will affect your payment. If you win, you will receive 10 Euros minus your voluntarily gifts to the loser(s) in your group. If you lose, you receive the gifts of the other members in your group who won.

(32)

Appendix 2 Decision Forms

Decision form 1

If I win 10 Euros, I am willing to hand over to each loser in my group:

In the case of one loser: Euros.

In the case of two losers: Euros.

Decision form 2

What do you expect the other owners of 10 Euros, handed over to each loser on average?

In the case of one loser: Euros.

(33)

Decision form 3

In the case of one loser. If you would know what the other owner of 10 Euros gives, what would you give? (This will not affect your payment).

Other owner gives You 0 1 2 3 4 5 6 7 8 9 10

(34)

Appendix 3 Expected and actual gifts

(35)

Appendix 4 Gender effect

(36)

Appendix 5 Education effect

(37)

Appendix 6 Gifts with one and two losers

Figure 11 Frequencies of subjects per behavior in case of one and losers†

†

: solidarity game with one loser in the group, : solidarity game with two losers in the group

(38)

Appendix 7 Response functions

(39)

(40)

Figure 14 Response functions: The Netherlands - Control treatment

(41)

(42)

Appendix 8 Econometric analysis

Table 11 P-values, Fisher’s exact test for country effect, (two-tailed)

Treatment Losers P-value

Control one P = 0.00175873 Control two P = 0.00219438 Crisis one P = 0.00000114 Crisis two P = 0.00005758

Table 12 P-values and effect size, Mann-Whitney U-test for country effect, (one-tailed)

Treatment Losers P-value Effect size†

Control one P = 0.00045 r = 0.47917 Control two P = 0.00048 r = 0.47618 Crisis one P = 0.00000 r = - 0.64099 Crisis two P = 0.00005 r = - 0.57995

†

, where is the -score and the number of subjects per treatment

Table 13 P-values, Fisher’s exact test for treatment effect, (two-tailed)

Country Losers P-value

Greece one P = 0.01222504 Greece two P = 0.04754837 Netherlands one P = 0.73996750 Netherlands two P = 1.00000000

Table 14 P-values and effect size, Mann-Whitney U-test for treatment effect, (one-tailed)

Country Losers P-value Effect size†

Greece one P = 0.04006 r = - 0.26114 Greece two P = 0.03288 r = - 0.27471 Netherlands one P = 0.23270 r = - 0.10566 Netherlands two P = 0.49202 r = - 0.00297

†

(43)

Table 15 P-values, Fisher’s exact test for gender effect, (two-tailed)

Country Treatment Losers P-value

Greece control one P = 0.20215540 Greece control two P = 0.28254139 Greece crisis one P = 0.46873065 Greece crisis two P = 0.69686660 Netherlands control one P = 0.37423433 Netherlands control two P = 0.26853200 Netherlands crisis one P = 0.61462451 Netherlands crisis two P = 1.00000000

Table 16 P-values, Fisher’s exact test for education effect, (two-tailed)

Country Treatment Losers P-value

Greece control one P = 0.84506102 Greece control two P = 0.84506102 Greece crisis one P = 0.13659148 Greece crisis two P = 0.51653030 Netherlands control one P = 1.00000000 Netherlands control two P = 0.77688787 Netherlands crisis one P = 0.11808300 Netherlands crisis two P = 0.11808300

Table 17 Chi-square test for independence between types of behavior and types of response functions

Constant _non-zero Constant _zero _increasingWeakly _decreasingWeakly _convexQuasi- _concave

Quasi-Other Column totals Egoistic behavior 0(3.46) [3.46] 28 (14.34) [13.00] 13 (12.86) [0.00] 2 (7.91) [4.42] 0 (0.99) [0.99] 3 (2.47) [0.11] 0 (3.96) [3.96] 46 Weak altruistic behavior 4 (2.18) [1.51] 1 (9.04) [7.15] 8 (8.11) [0.00] 8 (4.99) [1.82] 1 (0.62) [0.23] 2 (1.56) [0.12] 5 (2.49) [2.52] 29 Strong altruistic behavior 3 (1.35) [2.00] 0 (5.61) [5.61] 5 (5.03) [0.00] 6 (3.10) [2.72] 1 (0.39) [0.97] 0 (0.97) [0.97] 3 (1.55) [1.36] 18

Row totals 7 29 26 16 2 5 8 (Grand total) 93

(44)

Appendix 9 Demographics

Table 18 Gender

Control treatment Crisis treatment Control treatment Crisis treatment Male 15 9 17 14 Female 9 12 7 10 Total 24 21 24 24 Table 19 Nationality

†_{All subjects participated in the experiment conducted in the Netherlands were} all able to speak the Dutch language, thus, they were treated homogeneously

Table 20 Field of education

Control treatment Crisis treatment Control treatment Crisis treatment Economics 17 15 14 9 Engineering 2 1 - - Humanities 1 2 1 5 Law 2 1 1 2 Medicine - - - 1 Science 1 - 3 2 Social sciences 1 2 2 4 Other - - 3 1

Total, other education

than economics 7 6 10 15

Total 24 21 24 24

Control treatment Crisis treatment Control treatment Crisis treatment Dutch - - 23 21 Greek 24 21 - - Other† - - 1 2 Total 24 21 24 24