Fast or fair? A study of response times

Viewer
Transcript

Economics Letters 105 (2009) 193–196

Contents lists available at ScienceDirect

Economics Letters j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e c o l e t

Fast or fair? A study of response times☆ Marco Piovesan ⁎, Erik Wengström Department of Economics, University of Copenhagen, Øster Farimagsgade 5, Building 26, DK-1353 Copenhagen K, Denmark

a r t i c l e

i n f o

Article history: Received 8 April 2008 Received in revised form 22 July 2009 Accepted 23 July 2009 Available online 5 August 2009

a b s t r a c t This paper uses a modiﬁed dictator game to investigate the relationship between response times and social preferences. We ﬁnd that faster subjects more often chose the option with the highest payoff for themselves. Moreover, our within-analysis reveals that, for a given individual, payoff maximizing choices are reached quicker than choices expressing social preferences. © 2009 Elsevier B.V. All rights reserved.

Keywords: Response times Social preferences JEL classiﬁcation: C72 C91

1. Introduction The emerging ﬁeld of neuroeconomics seeks a biological foundation of social and economic decisions by studying activity in the neural circuitry.1 This approach is becoming widely used in economics but it is not without critiques. For instance, Rubinstein (2007) criticizes neuroeconomics as expensive and speculative. As an alternative, he proposes to follow psychology and explore the deliberation process of decision makers based on their response times (RT). Rubinstein (2007) studies strategic games and ﬁnds clear relationships between behaviour and RT.2 Our work follows a similar path by studying subjects' response times in the experiment of CabralesMiniaciPiovesanPonti08. Compared to the previous literature, the advantage of using this data is twofold. First, studying a modiﬁed dictator game, which represents a non-strategic situation, isolates the decision of caring for the other's payoffs or not. Second, we observe the same individuals making several repeated choices which enables us to study the within-subject relationship between RT and choices.

We ﬁnd that faster subjects more often chose the option with the highest payoff for themselves. Moreover, our within-analysis reveals that, for a given individual, payoff maximizing choices are reached quicker than choices expressing social preferences. Hence, the results from Rubinstein (2004, 2007) that fair decision making stems from fast responses, whereas more (rational) egoistic decision making requires more time, do not carry over to our setting. Our study differs from Rubinstein's in several respects which may explain the seemingly different conclusions. In our experimental design subjects were provided with monetary incentives, subjects made repeated choices and there was no equal split available to subjects in most of our decision tasks. We interpret our ﬁnding that egoistic decision making is faster as a result of the fact that maximizing own payoffs only requires the subject to pick the option with the highest payoff. Caring also about the other necessitates weighting your own payoff against the other's payoff; a process involving tradeoffs between emotions such as guilt, envy and efﬁciency concerns. 2. The experiment

☆ We are grateful to Pablo Brañas-Garza, Antonio Nicolò, Pedro Rey Biel, Rupert Sausgruber, Jean-Robert Tyran, one anonymous referee and seminar participants at Second Nordic Workshop in Behavioral and Experimental Economics in Göteborg and Alhambra Experimental Workshop in Granada for stimulating comments. We thank Antonio Cabrales, Raffaele Miniaci and Giovanni Ponti for letting us use the data. Financial support from MCyT (BEC2001-0980), Generalitat Valenciana (GV06/275) and the Instituto Valenciano de Investigaciones Económicas (IVIE) is gratefully acknowledged. ⁎ Corresponding author. Tel.: +45 35324409; fax: +45 35323000. E-mail addresses: [email protected] (M. Piovesan), [email protected] (E. Wengström). 1 See Camerer et al. (2005) for a survey of neuroeconomic research. 2 Another related study is Brañas-Garza et al. (2007) who investigate RT in an incentivized ultimatum game. 0165-1765/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2009.07.017

We use the data of the Cabrales et al. (2008) experiment. This experiment was conducted at the Laboratory of Theoretical and Experimental Economics (LaTEx) of the Universidad de Alicante with a total of 72 undergraduate students divided into three sessions. The experiment was programmed and conducted with the software zTree (Fischbacher, 2007).3 Participants played a modiﬁed dictator game lasting for 24 periods. At the beginning of each period, pairs were formed at random and

3

A translation of the instructions and example screens are available upon request.

80,51 70,30 75,40 70,32 71,42 70,30 47,37 39,39 46,43 39,39 41,36 42,38 51,48 44,44 52,51 46,45

Fig. 1. MRT over time.

Note: Numbers in the table gives the payoffs of the two players in Spanish Pesetas (1 Euro ≈ 166 ptas). LI, HI and MIX denote the degree of inequality of the options.

81,59 78,52 86,63 76,47 70,30 69,38 70,31 69,35 42,37 42,40 40,37 34,34 76,45 87,63 81,62 81,59 74,50 80,58 79,54 70,48 76,46 84,64 49,49 54,52 76,53 77,50 79,52 70,30 89,66 79,49 84,54 82,67 81,52 81,52 42,35 42,42 36,36 35,35 42,35 43,43 1 2 3 4

45,42 52,47 46,36 46,35

77,46 83,58 47,38 47,41

48,46 48,47 46,45 47,45

69,34 70,30 69,35 70,34

41,38 42,41 41,41 42,40

70,30 70,32 44,44 40,39

91,66 89,67 41,41 36,36

70,31 70,30 41,41 43,43

42,42 43,36 41,37 43,40

80,53 76,53 50,48 51,51

MIX

23 22

LI LI

21

HI HI

20 19 18 17

LI MIX LI

16 15

MIX HI MIX

14 13 12 11

LI HI HI MIX

10 9 8

LI

7 6

MIX LI HI HI MIX LI Options

5 4 3 2 1 Period

Table 1 Payoff options.

MIX

24

M. Piovesan, E. Wengström / Economics Letters 105 (2009) 193–196

HI

194

each participant had to decide how to allocate payoffs within the pair. Participants were restricted to a choice between four pre-speciﬁed options, describing the payoffs of both subjects. In each period, one participant (hereafter denoted as “Rich”) was randomly selected to have at least as high payoffs as the other player (hereafter denoted as “Poor”) in all 4 options. Moreover, the options differed in the degree of inequality and the choice set was composed of either: 1. four options with low inequality of payoffs (LI), 2. four options with high inequality of payoffs (HI), 3. two LI options and two HI options (MIX). See Table 1 for a list of the payoffs. The strategy method was used to collect observations for all players in each round; both subjects picked an option and one of them was thereafter randomly chosen to be the dictator and the payoffs were distributed according to her choice. The RT is measured for each of the 24 choices separately and refers to the time (seconds) between the moment the screen with the four options appears and the moment the subject clicks on her preferred option. Average earnings were about 11 euros and the sessions lasted approximately 30 min.

3. Results Fig. 1 displays the evolution of median RT (MRT) over time. We note a sharp decrease in MRT during the ﬁrst periods followed by a quite stable level throughout the remaining periods. Quite intuitively, as subjects get familiar with the format of the decision tasks they need less time to make their choice. Interestingly, the width of the interquartile ranges remains almost constant, indicating that experience does not have a clear effect on the heterogeneity in subjects' RT. There is also substantial variation in RT between different rounds of the experiment.

Table 2 Response times and decision types. Type of decision task

Descriptive statistics

Inequality

Player

25% percentile

Median RT

75% percentile

Average RT

LI HI MIX All All

All All All Poor Rich

4 4 4 5 4

6 7 6 7 6

10 11.5 10 11 10

8.4 8.7 7.6 8.5 7.9

M. Piovesan, E. Wengström / Economics Letters 105 (2009) 193–196

195

Table 3 Within-subject analysis of response times. (1)

Constant Period Period 1–6 Egoistic choice “Poor” LI “Poor” ⁎ LI HI “Poor” ⁎ HI Fairness efﬁciency Constant Number of obs. R2, overall Fig. 2. Fraction of egoistic choices.

One explanation to this variance may be the differences in inequality of the decision tasks. In Table 2 the summary statistics for the response times of the different decision tasks are displayed. In short, decisions were reached slowest on tasks characterized by high degree of inequality. This result is conﬁrmed by comparing MRT pairwise at the individual level with the Wilcoxon matched pairs (WMP) test. We reject the hypothesis that MRT for the HI decision tasks is drawn from the same distribution as LI and MIX (two-sided pvalues b0.050) but cannot reject the null comparing LI and MIX (twosided p-value = 0.874). Differences between periods could therefore to some extent be explained by the different degree of inequality of the options available to the subjects. From Table 2 it is also evident that the relative positions of the players were important for the speed of the decision making. Poor subjects made slower decisions than Rich subjects. Using the WMP test we can reject that individual MRT for Poor and Rich decisions is drawn from the same distribution (p-value = 0.001). A possible explanation for this ﬁndings is that it takes longer time to make a decision in the presence of envy. Guided by envy, players with low payoffs look at the payoff of the other in order to choose the pair with lower inequality. Overall the ﬁndings suggest that inequality and relative payoffs matter for RT, which motivates a more thorough investigation of the relationship between social preferences and RT.

3.1. Egoistic choices and RT To investigate the relationship between social preferences and RT we followed Rubinstein (2008) and divided subjects into four classes according to their MRT: very fast (the bottom 10%), fast (range of 10– 50%), slow (range of 50–90%) and very slow (the top 10%). For each subject we then calculated the fraction of egoistic choices across the 24 periods. That is, we calculated the fraction of decision tasks in which the subject chose the option that maximized his/her own payoff. In Fig. 2, the subjects' fractions of egoistic choices are displayed for each of the four classes. The ﬁgure shows that egoistic choices were much more prevalent among the faster subjects.4 Comparing the fraction of egoistic choices between the fast and the slow groups with the Wilcoxon/Mann–Whitney test we can reject the null hypothesis of similar distributions (p-value = 0.009). The result is even stronger if we compare the very fast with the very slow (p-

4

We obtain a similar picture if we look at the different player positions and degrees of payoff inequality separately. That is, if we split the sample into the four RT categories based on RT for the six different type decision tasks (mix-rich, mix-poor, high-rich, high-poor, low-rich, low-poor) we obtain analogous results.

(2)

Coef.

Std. Err.

Coef.

11.511⁎⁎⁎ − 0.164⁎⁎⁎ 1.258⁎⁎⁎ − 1.948⁎⁎⁎

0.342 0.026 0.413 0.497

10.183⁎⁎⁎ 0.571 − 0.16⁎⁎⁎ 0.026 1.26⁎⁎⁎ 0.409 − 1.892⁎⁎⁎ 0.342 0.93⁎⁎ 0.419 1.375⁎⁎⁎ 0.41 − 0.978⁎ 0.592 0.904⁎⁎ 0.415 0.268 0.587 0.194⁎⁎ 0.083 10.183⁎⁎⁎ 0.571 1728 (N = 72,T = 24) 0.130

1728 (N = 72,T = 24) 0.121

Std. Err.

Note: Results are obtained by ﬁxed effects within regressions using response time as dependent variable. Period takes values 1 to 24 according to the period of the experiment. Period 1–6 is a dummy variable for the ﬁrst six periods of the experiment. Egoistic choice is a dummy variable taking value one if the subject chooses the option that maximized his/ her payoffs. LI is a dummy variable for options with low inequality and HI is a dummy variable for options with high inequality. “Poor” is a dummy variable taking value one if the player has the low payoffs. Fairness efﬁciency is obtained, for each of the non-payoff maximizing options, by dividing the other player's gains with the costs for the subject. The highest of these gains-to-cost ratios is included in the regression. In case the egoistic payoff maximizing option is also payoff maximizing for the other subject, the variable is set to zero. The variables “Poor” ⁎LI (HI) are interaction terms. ⁎⁎⁎ denotes signiﬁcance at the 1% level and ⁎⁎ at the 5% level.

value = 0.001). Taken together, there is a clear relationship between RT and social preferences. More specially, faster subjects make more egoistic choices than slower subjects.5 3.2. Within-subject relationship between egoistic choices and RT The analysis above leaves open the possibility that the relationships between RT and egoistic choices are driven by some individual unobserved characteristic that is correlated with RT and preferences. To exclude this explanation, we now take a closer look at the withinsubject relationship between choices and RT. Table 3 summarizes results from ﬁxed effect regressions that exploit the panel structure of our data. In speciﬁcation (1), RT is regressed on the variable egoistic choice, a time trend and a dummy variable for the ﬁrst six periods.6 The variable egoistic choice is a dummy variable taking value one if the subject chose the egotistic payoff maximizing choice and value zero otherwise. The time variables are signiﬁcant, which reﬂects the decline in RT shown in Fig. 1. Importantly, we note that RT is negatively related to make an egoistic payoff maximizing choice. The relationship between RT and egoistic choices is robust to changes in the set of regressors. In speciﬁcation (2), RT is regressed on a more extensive set of variables including dummy variables for relative payoffs, degree of payoff inequality and interaction terms. To address the concern that the egoistic choice is salient in some decision tasks we include the variable fairness efﬁciency, which is obtained by dividing the amount the opponent would gain by the amount the subject would lose if the subject does not choose the egoistic option. Hence, the variable describes how much the other subject will gain if the player gives up one unit of payoff. In short, this regression exercise conﬁrms that the degree of inequality and relative payoffs affect RT. In addition, we observe a

5 This conclusion is conﬁrmed also if we classify the subjects using the Fehr and Schmidt (1999) preferences. Details are available upon request. 6 The results reported here are not sensitive to the choice of cutoff point. We chose period six since subjects experience two of each type of decision task (HI, LI, MIX) during the ﬁrst six periods.

196

M. Piovesan, E. Wengström / Economics Letters 105 (2009) 193–196

positive and signiﬁcant relationship between how cheap it is to increase the other's payoffs and RT. This indicates that decisions are reached more promptly if it is costly or impossible to increase the other player's payoff. Finally, the relationship between RT and egoistic choices holds also in this speciﬁcation. Hence, we can rather conﬁdently conclude that there exists a relationship between the degree of social concerns taken into consideration and RT, not only between subjects, but also within subjects. 4. Conclusions We ﬁnd that faster subjects make egoistic choices more often than slower subjects in a non-strategic setting. Using a within-subject analysis we rule out that this is a mere reﬂection of social subjects being slow decision makers. Our results point out that, for a given individual, egoistic payoff maximizing decisions are reached quicker than choices expressing social preferences. We believe that this ﬁnding manifests a relationship between RT and the level of complexity of the decision task and that the perceived complexity will depend on the preferences of the decision makers. Using the framework of Loewenstein and O'Donoghue (2004) our result may be interpreted as if egoistic decisions are faster since there is less conﬂict between affective and deliberative reactions. On the contrary, social decisions are slower since they do involve such conﬂicts that imply an

extra “cognitive load”. This interpretation is supported by a recent neuroeconomic study of the ultimatum game (Knoch et al., 2006). They ﬁnd that overriding or weakening self-interested impulses necessitate the activation of a particular brain area (dorsolateral prefrontal cortex), which slows down decision making. References Brañas-Garza, P., León-Mejía, A., Miller, L., 2007. Response time under monetary incentives: the ultimatum game. Jena Economic Research Papers # 2007-070. Cabrales, A., Miniaci, R., Piovesan, M., Ponti, G., 2008. Social preferences and strategic uncertainty: an experiment on markets and contracts. Discussion Papers 08-06, University of Copenhagen. Department of Economics. Camerer, C., Loewenstein, G., Prelec, D., 2005. Neuroeconomics: how neuroscience can inform economics. Journal of Economic Literature 43 (1), 9–64. Fehr, E., Schmidt, K.M., 1999. A theory of fairness, competition, and cooperation. Quarterly Journal of Economics 114 (3), 817–868. Fischbacher, U., 2007. z-Tree: Zurich toolbox for ready-made economic experiments. Experimental Economics 10 (2), 171–178. Knoch, D., Pascual-Leone, A., Meyer, K., Treyer, V., Fehr, E., 2006. Diminishing reciprocal fairness by disrupting the right prefrontal cortex. Science 314, 912–915. Loewenstein, G., and O'Donoghue, T. (2004): “Animal spirits: affective and deliberative processes in economic behavior,” mimeo. Rubinstein, A., 2004. Instinctive and cognitive reasoning: a study of response times. http://cess.nyu.edu/0012:2004-03.pdf. Rubinstein, A., 2007. Instinctive and cognitive reasoning: a study of response times. Economic Journal 117 (523), 1243–1259. Rubinstein, A., 2008. Comments on neuroeconomis. Economics and Philosophy 24, 485–494.