Experimental Economics

, Volume 15, Issue 4, pp 589–603

Are social preferences related to market performance?

Authors

    • Department of EconomicsMonash University
Article

DOI: 10.1007/s10683-012-9315-y

Cite this article as:
Leibbrandt, A. Exp Econ (2012) 15: 589. doi:10.1007/s10683-012-9315-y

Abstract

This paper combines laboratory with field data from professional sellers to study whether social preferences are related to performance in open-air markets. The data show that sellers who are more pro-social in a laboratory experiment are also more successful in natural markets: They achieve higher prices for similar quality, have superior trade relations and better abilities to signal trustworthiness to buyers. These findings suggest that social preferences play a significant role for outcomes in natural markets.

Keywords

Social preferencesMarket performanceExternal validityQuality misrepresentationOpen-air markets

JEL Classification

C91C93D49

1 Introduction

A fundamental question in behavioral economics is which role social preferences play in natural markets. Pro-social behavior is omnipresent in the laboratory environment (Güth et al. 1982; Roth 1995; Fehr and Gächter 2000; Camerer 2003) and these observations have led to the formulation of other-regarding preference theories (Andreoni 1990; Rabin 1993; Fehr and Schmidt 1999; Bolton and Ockenfels 2000; Charness and Rabin 2002; Dufwenberg and Kirchsteiger 2004; Sobel 2005; Falk and Fischbacher 2006; Cox et al. 2007; López-Pérez 2008). Laboratory studies also suggest that pro-social behavior can affect outcomes in market settings, pay off for employers and provide explanations for phenomena such as price rigidities and relational contracts (Fehr et al. 1993, 2009; Brown et al. 2005).

There is also suggestive evidence that pro-sociality is a positive predictor for earnings and productivities outside the laboratory environment (Bowles et al. 2001; Barr and Serneels 2009; Dohmen et al. 2009; Carpenter and Seki 2010). There are at least three explanations for the positive impact of pro-sociality on job performance. The first potential explanation is that pro-social individuals are more likely to select into environments where earnings and productivities are higher than selfish individuals. The second potential explanation is that pro-social individuals are more productive because they interact better with their colleagues and are better integrated into the social network at the workplace (Barr et al. 2009). The third potential explanation is that pro-social individuals interact more smoothly with employers/buyers because they are less likely to refrain from opportunistic behavior that hurts the employers/buyers and thus can reap reputational benefits (Bowles et al. 2001).1

This paper investigates professional individual sellers in open-air markets and whether pro-social sellers achieve different prices for their products and have different trade relations than selfish sellers. The hypothesis based on previous evidence is that pro-social sellers outperform selfish sellers. To test this hypothesis and whether differential seller-buyer relations are responsible for such a relationship, I observe the behavior of the same professional shrimp sellers in open-air markets and the laboratory, and in addition collect additional information about them in surveys. More precisely, I use an anonymous laboratory experiment to isolate the sellers’ levels of pro-sociality. On the markets, I record trade outcomes to measure prices, qualities and quality misrepresentations. Finally, I conduct surveys to collect data on the sellers’ trades, trade relations and their characteristics.

The data confirm the hypothesis and show that social preferences are positively related to performance in natural markets. Sellers, who are more pro-social in a laboratory public goods experiment, achieve significantly higher prices for goods of similar quality than less pro-social and selfish sellers. The features of the field setting render two of the three mentioned explanations unlikely for the observed relationship between pro-sociality and market performance. First, it is unlikely that this relationship is driven by selection into different job environments because the study investigates one unique job environment. Second, it is unlikely that the relationship is explainable by the idea that pro-social individuals interact better with colleagues because the individuals in this study operate on their own, i.e. they catch and sell shrimp alone. However, as the subjects in this field setting are in steady and direct contact with buyers it seems likely that the third potential explanation is valid, i.e., that pro-social sellers outperform selfish sellers because they interact more smoothly with buyers.

I find mixed evidence for the explanation based on differential seller-buyer relations depending on the seller’s level of pro-sociality. On the one hand, the data show that more pro-social sellers have significantly more stable and longer lasting trade relations than less pro-social sellers. Moreover, I observe that more pro-social sellers self-report to be better able to signal trustworthiness to buyers than less pro-social sellers. However, on the other hand, I find no evidence that pro-social sellers misrepresent quality less than selfish sellers. Thus, I find that pro-social sellers interact more smoothly with buyers than selfish sellers but it remains unclear why this is the case. One possibility is that pro-social sellers cheat less on other unobservable dimensions. Another possibility is that buyers know the seller’s type but hold false beliefs about their level of quality misrepresentation.

This paper is also related to other studies which combine laboratory data on pro-sociality with field data (Karlan 2005; List 2006; Fehr and Leibbrandt 2011) and studies which observe the link between prices, reputation and trade relations (Weisbuch et al. 2000; Jin and Kato 2006). For example, Fehr and Leibbrandt (2011) study individuals drawn from the same subject pool and find that more pro-social fishermen exploit fishing grounds less. List (2006) studies sellers’ pro-sociality in the laboratory and also relates it to their quality misrepresentation in natural markets. His study suggests that the main determinant for quality misrepresentation in natural markets is reputation and that pro-sociality plays a negligible role. In contrast to List, this studies combines both laboratory and market data from the same individuals to directly study whether the extent of individual pro-sociality is related to individual quality misrepresentation in natural markets.

The reminder of this paper proceeds as follows. Section 2 presents the field setting and the collected data. Section 3 links the data on market performance with the laboratory data. Section 4 provides concluding remarks.

2 Field setting and the data

2.1 Field setting

The study took place in Brazil, using fishermen who live by selling their caught shrimp in open-air markets.2 The participants in this study usually catch shrimp five to seven days per week and sell their shrimp once a week in markets. In the field setting, there is one bigger and several smaller markets. Reputation plays a crucial role in these markets, as long-term trade relations between sellers and buyers are very common. The terms of the trades are not common knowledge. Typically, fishermen are able to sell their whole catch within few hours, and often to one buyer. Figure 3 in the Appendix illustrates one shrimp market.

The traded shrimp differ on one main quality dimension that significantly affects their price: shrimp size. Bigger shrimp are tastier, and are typically sold for significantly higher prices per liter than smaller shrimp. However, bigger shrimp are more difficult to catch. First, to catch a large fraction of bigger shrimp fishermen need to use larger hole sizes in their shrimp traps which however leads to a lower quantity of smaller shrimp caught (as they can escape from these traps).3 Second, because many fishermen exploit the shrimp population by catching huge quantities of smaller shrimp which have not reached sexual maturity, there is pressure on the shrimp population and it is particularly difficult to encounter large quantities of bigger (i.e. mature) shrimp. Besides differences in the shrimp size, there are differences in the shrimp color as some sellers color their shrimp to be suggestive to be tastier. There are no visible differences in the shrimp type and freshness is less important because the shrimp are sold dried.

The shrimp are represented in large piles (frequently containing more than 100 liters of shrimp) and there is incomplete and likely asymmetric information about the average size of the shrimp.4 I also realized that some sellers place the bigger shrimp on top of the pile and in this way misrepresent quality.5 In principal, buyers could identify such quality misrepresentation by scanning the piles. However, I have not observed such behavior on these markets. Note in this regard that the average shrimp size differs from week to week, i.e., even if buyers measured the exact size in a certain week, they would not know the exact size in a different week. Moreover, although sellers have considerable influence over the average shrimp size they sell (as it is largely determined by the holes in their shrimp traps), other factors not under control of the fishermen, like season or luck, also influence average shrimp size.6

2.2 Field data

The field data comes from three sources: (i) records of actual trades on one big and three smaller shrimp markets, (ii) surveys with sellers, and (iii) two laboratory experiments with sellers. Table 1 provides a detailed overview of the field data. Appendix Table 3 provides some additional information on the data collection.
Table 1

Summary statistics

 

Mean

Standard error

Lower quartile

Upper quartile

N

Price (per liter in Reais)

1.31

0.04

1

2

143

Shrimp size (in cm)

2.99

0.04

2.68

3.26

147

Quantity shrimp (in liter)

71.93

6.24

20

100

137

Color of shrimp

0.41

   

145

Size difference (in %)

7.32

2.35

0.1

15.94

33

Trade stability

0.39

   

114

Trade duration (in years)

5.26

0.48

2

6.75

112

Signaling ability

2.29

0.06

2

3

150

Market recording

0.34

   

138

Pro-sociality

3.67

0.19

1

5

216

Risk-aversion

6.98

0.15

5

8

216

Experience

17.72

0.81

9

22

215

Outside income (monthly in Reais)

95.71

21.39

0

50

212

Income/wealth

394.81

406.86

158.13

493.33

216

Shrimp seller

0.75

0.03

  

216

Male

0.78

   

216

Notes: Color of shrimp=1 if shrimp is colored, 0 if shrimp is not colored. Size difference indicates differences between shrimp on top and bottom of shrimp pile. Trade stability=0 if seller reports to have already lost an important buyer, i.e. a buyer who frequently buys at least 20 liters, 1 otherwise. Trade duration defines the duration of a trade relation to a buyer who buys frequently at least 20 liters. Signaling ability defines the sellers self-estimation of signaling trustworthiness compared to other sellers (1=worse, 2=equal, 3=better). Market recording=1 if data on price and shrimp size was recorded on the market during the market exchange, 0 otherwise. Pro-sociality defines the amount of monetary units contributed in the public goods experiment. Experience defines the years being a fishermen selling catch. Outside income=Income from agriculture. Shrimp seller=1 if individual only sells shrimp in markets, 0 if seller also sells fish. Male=1 if gender is male, 0 if female

2.2.1 Records of actual trades

I collected information on the trade outcomes (prices and quantities of sold shrimp) and the characteristics of the traded shrimp (average size, color of shrimp). To measure precisely shrimp size, I collected samples from the sold shrimp and then measured 30 randomly picked shrimp from this sample and averaged their size. Experimenters were able to encounter approximately one third of the sellers at the markets and collected data on the trade outcomes, average size and color of shrimp immediately after the purchase (N=47 sellers). They asked sellers about the details of the trade (price and liter sold) and collected samples from the sold shrimp to measure the average size and color. The other two thirds of the data were collected some hours, or some very few days after the purchases have occurred.7 Experimenters either visited these sellers in their houses to ask them about the details of the trade and to collect a sample of the sold shrimp or they asked them about trade outcomes during meetings to which they also brought a sample of the sold shrimp. In addition, I collected data on the shrimp size on top and bottom of the shrimp pile. This data was exclusively collected at the markets (N=33 sellers).

Sellers achieve on average 1.31 Reais (1 Real, pl. Reais; 1 Real equaled US$ 0.60) per liter shrimp and sell on average 71.9 liter shrimp. The average size of the shrimp is 2.99 centimeters and 41% color their shrimp.8 The shrimp size is larger on top of the shrimp pile in 25 of the 33 samples. On average, the sample from the top contains shrimp that are 7.3% larger (one sample T-test that mean equals zero, t=3.12, p<0.004; variable: size difference).

2.2.2 Surveys with sellers

I have data from two surveys that were conducted individually and such that other sellers could not listen to or observe the responses to the survey questions. In the first survey, I collected information about the sellers’ attributes such as their gender, their experience in selling shrimp and whether they also generate income from other activities than catching shrimp. 78% of the sellers are male, and they sell their catch on average for already 17.7 years (variable: experience). 75% in our sample are fishermen who specialize in selling shrimp (variable: shrimp seller) whereas the remaining 25% sell shrimp and fish. Approximately 36% generate an additional small income by selling agricultural products.

In the second survey which was conducted several months after the first survey, sellers from the same subject pool were asked about their trade relations and signaling abilities. To identify sellers who were/are involved in long-term trade relations, I asked participants whether they had or currently have a buyer who frequently bought/buys at least 20 liters shrimp from them (76% said yes). I asked the sellers who responded with yes whether they have already lost such an important buyer to another seller (variable: trade stability). In addition, I asked these sellers for how long this trade relation existed/exists (variable: trade duration). For signaling ability, I asked sellers (independent whether they have long-term trade relations) about their self-estimation of how well they can appear trustworthy to the buyers relative to the other sellers (the categories were: (it is) more difficult, similar, or easier (for me)).

60.5% of the sellers who were/are involved in long-term trade relations report to have already lost an important buyer. The average trade relation lasted for 5.26 years. Trade stability and trade duration measure two distinct aspects of trade relations and are not significantly related (z=0.361, p=0.718). With regard to signaling ability I find that 13.3 percent report to have more difficulties signaling trustworthiness compared to the other sellers whereas 42% believe it is easier for them to signal trustworthiness (the remaining 44.7% say it is equally difficult).

2.2.3 Laboratory experiments with sellers

Sellers took part in experimental sessions (N≥15) where they played a public goods experiment (PGE) and a risk-aversion experiment (RAE), both with high stakes.9 The experiments took place during village meetings, typically in a local school building and before the survey and market data was collected. The experiments were conducted individually and anonymously, i.e. participants were seated in a way such that they could not look at or listen to the decisions of other participants. Most participants knew each other as they were fishermen from the same village but they did not know who was in their group in the PGE.

In the PGE, the participants were divided in groups of three and played this experiment anonymously for one period.10 Each participant had to decide how many out of ten monetary units (MUs) he transfers from a private to a group account. The experimenter gave the participants two envelopes, one containing 10 MUs (the ‘private account envelope’) and one containing 0 MUs (the ‘group account envelope’). The participants could transfer MUs from the private account envelope to the group account envelope and thereafter put the envelopes in a box. At the end of the experiment, each MU in the group account envelope was multiplied by 1.5 and then divided equally between the three group members. Thus, it was not in the monetary self-interest to contribute because the net return from contributing 1 MU was only 0.5 MU. However, for the group it was optimal if all group members contributed maximally. If all three individuals in the group decided not to contribute, each of them only earned 10 MUs (10−0+0), compared to 15 MUs (0+0.5×10×3) if all of them contributed all ten MUs. To minimize scrutiny, the letters were only identifiable by codes (no names were written on the envelopes) and the experimenters turned their backs to the participants during their contribution decisions. The experimenters explained all rules individually to the sellers and no seller was informed about the identity of his group members.

I denote the contribution decision pro-sociality. The more sellers contribute, the more pro-social they are. I find that most sellers contribute to the public good; only 16.2% did not contribute and 11.1% contributed only one MU. Approximately half of the participants contribute between zero and three MUs (51.4%), 10.2% four MUs, 19.4% five MUs and 14.8% more than five MUs.

In the RAE, participants had to decide how many out of ten MUs they invest in a lottery with a payoff of 2.5 times the invested amount and a winning probability of 50%; i.e. the expected payoff of the lottery is 1.25 times the invested amount. The experiment lottery was implemented in a simple manner with the help of a coin flip. Participants had to announce which side of the coins shows up after tossing the coin. I observe high levels of risk-aversion: 21.8% do not invest at all, 35.6% invest only two or three MUs and only 7.4% invest more than five MUs.

3 Pro-sociality, market performance and trade relations

In this section I link the different data sets. Figure 1 provides a first raw impression of the relationship between laboratory pro-sociality and field market performance. It illustrates the cumulative percentage of achieved shrimp prices depending on the level of contributions in the public goods experiment. Because of the relatively small number of observations for some contribution levels, I split the sellers into two equally large samples according to their contributions in the PGE: the less pro-social sellers who contributed less than four MUs (N=74) and the more pro-social sellers who contributed at least four MUs (N=69). The figure for example shows that a larger percentage of the less pro-social sellers achieve prices below 1.5 Reais (blue solid line, 71.6%) as compared to the more pro-social sellers (red dashed line, 52.2%). The more pro-social sellers achieve on average 1.41 Reais per liter shrimp which is approximately 15 percent more than the less cooperative sellers (average=1.22 Reais; t=2.07, p=0.04). Sellers who free-ride in the public goods experiment and contribute nothing achieve on average only 1.1 Reais per liter shrimp. The pure correlation between pro-sociality and shrimp price is significant at p=0.029 (r=0.18, Pearson).11
https://static-content.springer.com/image/art%3A10.1007%2Fs10683-012-9315-y/MediaObjects/10683_2012_9315_Fig1_HTML.gif
Fig. 1

Pro-sociality and price

To measure the relationship between pro-sociality and market performance in a more precise manner, I use the achieved selling price per liter as the dependent variable in an OLS regression and control for shrimp size and other potential covariates. More precisely, in Table 2, model 1, I investigate whether public goods contributions are related to shrimp prices after controlling for features of the shrimp sold (size and color) and trade specifics (quantity, location and date) as well as other seller attributes (risk-aversion, specialization, outside income, knowledge and gender) and how the data was collected (immediately or shortly after transaction; variable: market recording). The model shows that pro-sociality is significantly linked to prices at p=0.062. The positive coefficient of 0.020 says that sellers who contribute ten instead of zero MUs in the public goods game achieve 0.2 Reais more per liter shrimp (approximately 15 percent of the average shrimp price) after controlling for quality and the previously mentioned variables.
Table 2

Determinants of market performance

Model

dependent variable

(1)

Price per liter shrimp

OLS

(2)

Trade stability

Probit

(3)

Trade duration

OLS

(4)

Signaling ability

Probit

(5)

Size difference (in %)

OLS

Pro-sociality

0.020*

(0.011)

0.030**

(0.015)

0.322*

(0.171)

0.019**

(0.008)

1.333

(0.919)

Shrimp size

0.154**

(0.070)

    

Quantity shrimp

−0.672E−03

(0.482E−03)

    

Risk-aversion

0.020

(0.016)

−0.002

(0.022)

−0.136

(0.199)

−0.013

(0.011)

1.935

(1.330)

Shrimp seller

0.106

(0.073)

−0.270***

(0.098)

1.760*

(0.933)

−0.039

(0.043)

 

Outside income

−0.030E−03

(0.051E−03)

−0.290E−03

(0.234E−03)

−0.063E−03

(0.227E−03)

−0.193E−03

(0.118E−03)

 

Experience

0.020E−01

(0.017E−01)

−0.012E−01

(0.039E−01)

0.156***

(0.047)

−0.468E−03

(1.766E−03)

 

Male

0.050

(0.056)

−0.165

(0.113)

−0.212

(1.263)

0.011

(0.050)

 

Color of shrimp

0.010

(0.073)

    

Market recording

0.147*

(0.078)

    

Market fixed effects?

yes

yes

yes

no

yes

Date fixed effects?

yes

no

no

no

no

Constant

0.255

(0.227)

 

1.890

(2.034)

 

−6.728

(12.531)

R-sqr

0.726

 

0.247

 

0.365

N

133

113

111

148

33

Notes: ***p<0.01, **p<0.05, *p<0.1. Robust standard errors in parentheses. Coefficients in probit models present average marginal effects. Observations are on individual level

Besides pro-sociality, only size and market recording are significant variables for shrimp price. As should be expected, sellers are able to achieve higher prices when they offer larger shrimp (p=0.030). We also observe that market recording is positive and significant (p=0.062) showing that shrimp prices are higher when measured during the trade. All other covariates in model 1 are not significant at the 10%-level.

Result 1

Sellers who contribute more in a laboratory public goods experiment achieve higher prices (per liter for shrimp of similar quality) in natural markets.

Next I investigate the relationship between pro-sociality and trade relations. I find that pro-sociality is significantly linked to both trade stability and trade duration. 71.2% of the less pro-social sellers (contributions in PGE<4 out of 10 MUs) report to have already lost an important buyer whereas the corresponding number is only 49.1% for the more pro-social sellers. The sellers at highest risk of trade termination are selfish sellers who did not contribute in the PGE (81.8%). In Table 2, model 2 I use a Probit model to estimate the impact of pro-sociality on the trade stability controlling for the seller attributes in model 1 (risk-aversion, specialization, outside income, knowledge and gender). Model 2 shows that pro-sociality is significant and positively related to trade stability (p=0.039). The coefficient represents the marginal effect of one additional MU contributed in the public goods experiment; i.e., a seller who contributed ten instead of zero MUs is approximately 30 percent more likely to report that he has not lost an important buyer. The model also shows that specialized sellers face a significantly higher risk of trade termination which is natural since they are more likely to have more long-term trade relations.

The level of public goods contributions is also positively related to the duration of trade relations to important buyers. Figure 2 illustrates the duration of the trade relation for the less and more pro-social sellers. We can see for example that a larger percentage of the less pro-social sellers has trade relations which lasted for maximally three years (blue solid line, 59.3%) as compared to the more pro-social sellers (red dashed line, 37.7%). On average, the more pro-social sellers have trade relations which exist for more than six years whereas the less pro-social sellers have trade relations which exist for 4.5 years (t=1.58, p=0.117). The average duration for trade relations from selfish sellers who did not contribute in the PGE is only 3.8 years. In Table 2, model 3, I use an OLS regression to control for the seller attributes used in the previous models. The regression shows that pro-sociality is positive and significant at p=0.063. Each additional MU contributed in the PGE is associated with a 0.322 years longer trade duration. Furthermore, we can see in this model that more experienced sellers (p<0.01) and specialized sellers (p=0.062) have longer lasting trade relations.12
https://static-content.springer.com/image/art%3A10.1007%2Fs10683-012-9315-y/MediaObjects/10683_2012_9315_Fig2_HTML.gif
Fig. 2

Pro-sociality and trade duration

Result 2

Sellers who contribute more in a laboratory public goods experiment are at a lower risk of losing important buyers and have longer lasting trade relations in natural markets.

There are also interesting links between pro-sociality and signaling abilities. I find a positive and significant relationship between the sellers’ public goods contributions in the laboratory and their self-estimate about the extent to which they can signal trustworthiness. Only 10.8 percent of the more pro-social sellers have problems signaling trustworthiness; i.e., they report that it is more difficult for them to signal trustworthiness compared to the other sellers. In contrast, the percentages are substantially higher for the less pro-social (15.8 percent) and selfish sellers who do not contribute in the PGE (20 percent).

In Table 2, model 4, I use an ordered Probit model which includes the previously mentioned other seller attributes as controls. I find that pro-sociality is significantly linked to signaling abilities even after controlling for risk-aversion, gender and other variables. The coefficients are in average marginal effects and show that each additional MU contributed in the PGE is associated with a 1.9 percent increase in the probability to report that one can signal trustworthiness better than other buyers (p=0.021).

Result 3

Sellers who contribute more in a laboratory public goods experiment report to be better able to signal trustworthiness to buyers.

While there are significant relationships between pro-sociality and prices, trade relations and signaling abilities, pro-sociality is not significantly correlated to quality misrepresentation (N=33, r=−0.01, p=0.93, Spearman). More pro-social sellers have on average a difference of 6.1 percent in shrimp size between top and bottom shrimp whereas less pro-social sellers have a slightly but not significantly higher difference of 8.3 percent (t=0.46, p=0.648). The selfish sellers who do not contribute in the PGE are also not different from the other sellers; their difference is 7.4 percent. Likewise, in Table 2 we observe that pro-sociality is not linked to the extent to which sellers place bigger shrimp on top of their shrimp pile after controlling for covariates. In model 5, we can see that, if at all, pro-sociality is rather positively than negatively related to quality misrepresentation (t=1.36, p=0.186).

Result 4

Sellers who contribute more in a laboratory public goods experiment are not less likely to place bigger shrimp on top of the shrimp pile.

4 Concluding remarks

This paper observes professional sellers in a laboratory environment and their performance in natural markets. I find that sellers who are more pro-social in the laboratory outperform less pro-social and selfish sellers and achieve higher prices for the same goods. I also provide empirical evidence that more pro-social sellers are involved in more stable and longer lasting trade relations and can better signal trustworthiness to buyers. These data provide new evidence that laboratory pro-sociality is related to outcomes in natural markets and therefore also corroborate the relevance of pro-sociality findings in laboratory experiments and other-regarding preference theories. In addition, this paper observes the level of quality misrepresentation from a small sample of sellers in their natural environment. I find no evidence that more pro-social sellers in the laboratory misrepresent quality less in markets as compared to less pro-social sellers.

An interesting feature of this study is that the observed link between pro-sociality and market performance can hardly be explained by selection into jobs depending on pro-sociality or differences in relations to co-workers depending on pro-sociality. In addition, an explanation based on reverse causality is unlikely to account for the main finding. I do not find that sellers who achieve higher prices are richer and as a result become more pro-social as richer individuals are not significantly more pro-social in the public goods experiment (r=0.05, p=0.43). Moreover, including income/wealth in the regression analysis does not affect the relationship between prices and pro-sociality (Appendix Table 4, models 3–4). Finally, it is unlikely that certain seller characteristics that affect the likeability of a seller play an important role. I asked six buyers who frequently buy shrimp from the studied sellers to rank the importance of the seller’s (i) trustworthiness, (ii) price, and (iii) likeability, for their choice of a trading partner. All buyers report that the seller’s trustworthiness is the most important factor followed by the price providing suggestive evidence that likeability plays if at all a relatively minor role.

Taken together, the data suggests that individual differences in social preferences are important for understanding market performance of sellers and that social preferences facilitate the smoothness of seller-buyer relations. A crucial next step is to pin down why and how pro-social sellers are able to sustain superior trade relations as compared to selfish sellers.

Footnotes
1

Bowles et al. (2001) argue that there are different “predisposition(s) to truth-telling” and that truth-telling is an “incentive-enhancing” preference that attenuates tensions in principal-agent relationships.

 
2

In this setting, there is free access to the fishing grounds and capital requirements for becoming a fisherman selling shrimp in markets are low.

 
3

The vast majority uses modified plastic bottles to catch shrimp. Thus, there are no significant differences in equipment used.

 
4

I tested the existence of incomplete/asymmetric information in a “guessing game” where buyers and sellers took part. In this game the most accurate guess about the average shrimp size in a pile was rewarded with a high monetary reward (worth several days’ income). The sellers guessed the average shrimp size in their pile and the buyers guessed the average shrimp size in a pile they were about to buy. It turned out that both sellers and buyers significantly overestimated the average shrimp size by on average 0.297 centimeters (t=2.29, p=0.027, that overestimates are equal to zero) and that buyers overestimated the average size considerably more than sellers (0.437 centimeters buyers vs. 0.243 centimeters sellers).

 
5

There may be other ways to misrepresent quality or to cheat which I could not observe. For example, sellers often have the chance to cheat on the quantity of shrimp sold.

 
6

I collected data on the average shrimp size over consecutive weeks from 24 sellers. As expected there is a significant correlation of average shrimp size across weeks (Spearman Rank Correlation, r=0.493, p=0.014), but considerable variance as well.

 
7

Sellers do not always go to the market to sell shrimp. They also sometimes commission other sellers to sell their shrimp.

 
8

Sellers color their shrimp red with natural or chemical substances to be suggestive to be tastier. I use a binary measure to assess the probability that a sample of shrimp was colored. This measure was derived from the experimenters’ estimation of the redness of shrimps.

 
9

Participants earned significantly more than a typical daily income. They took also part in other experiments than the PGE and RAE (a stag-hunt experiment, competition experiment, time preference experiment, charity experiment). To minimize the risk that there are behavioral spillovers between experiments, participants did not know the behavior of the other participants before the end of all experiments. In addition, participants were told that they did not get to know whether their behavior in any experiment became payoff relevant before the end of all experiments because only two experiments were chosen for payment. The experimental instructions are available in an on-line appendix.

 
10

Note that the group size in the PGE was four in one experimental session (N=16). The behavior in this session is very similar compared to all other sessions (average contribution in this session=3.75, in all other sessions=3.66; t=−0.12, p=0.90). Excluding this session from further analysis would not lead to systematic different results.

 
11

Appendix Table 4 shows that there is a significant relationship between pro-sociality and shrimp price for both samples. The sample which uses prices that were directly collected at the market is highly significantly related to pro-sociality (p=0.007, N=47) and the other sample that uses prices which were collected outside the market is marginally significantly related to pro-sociality (p=0.096, N=91).

 
12

Not specialized shrimp sellers sometimes temporarily only sell fish. This may explain why they have more problems keeping trade relations alive.

 

Copyright information

© Economic Science Association 2012