Restaurant tipping in a field experiment: How do customers tip when they receive too much change?

Restaurant tipping in a field experiment: How do customers tip when they receive too much change?

Journal of Economic Psychology 50 (2015) 13–21 Contents lists available at ScienceDirect Journal of Economic Psychology journal homepage: www.elsevi...

341KB Sizes 0 Downloads 69 Views

Journal of Economic Psychology 50 (2015) 13–21

Contents lists available at ScienceDirect

Journal of Economic Psychology journal homepage: www.elsevier.com/locate/joep

Restaurant tipping in a field experiment: How do customers tip when they receive too much change? Ofer H. Azar ⇑, Shira Yosef, Michael Bar-Eli Ben-Gurion University of the Negev, Israel

a r t i c l e

i n f o

Article history: Received 3 April 2014 Received in revised form 11 June 2015 Accepted 22 June 2015 Available online 27 June 2015 JEL classification: C93 D03 D12 PsycINFO classification: 3900 Consumer Psychology 3920 Consumer Attitudes & Behavior 3000 Social Psychology Keywords: Tipping Field experiments Social norms Consumer behavior Moral licensing and moral cleansing

a b s t r a c t Tipping behavior is analyzed in a field experiment where restaurant customers received excessive change, either 10 or 40 Shekels (about $3 versus $12). One third of the tables reported the extra change to the server and returned it. Tips were higher with the higher level of extra change. Returning the extra change was negatively correlated with tips. The interaction between the level of extra change and whether it was returned had no effect. Possible explanations that tips are higher when the excessive change is not returned due to a positive income effect or perceiving paying a tip out of excessive change as less costly because it is a forgone gain and not a loss, are not supported by the data. Subjects may have exhibited moral licensing and moral cleansing effects. These effects, however, were possibly mitigated by a self-selection effect going in the opposite direction: those who are more generous or altruistic by their nature are more likely to return the extra change and also more likely to tip generously. Receiving the larger amount of extra change may result in feelings of good mood, perceived luckiness, or individuation and unconscious fear of being observed, which increase tips. Interestingly, these feelings seem to remain even if the extra change was returned. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction In recent years the economics literature has given increased attention to behavior that does not conform to the traditional assumption of a perfectly-rational and egoistic economic agent. Phenomena like heuristics and biases in economic decision making, other-regarding preferences, social norms, voluntary market payments and other behaviors that emphasize the psychological and social motivations in economic behavior constitute a much larger share of the economics literature today than a few decades ago. One of the topics in this area is tipping, which is related to the literatures both on social norms and on voluntary market payments (because giving a tip is voluntary and not a legal obligation). Tipping is an interesting social norm from an economic perspective for several reasons. First, as opposed to many other social norms, it is an economic activity. The customer receives service and pays for it. Consequently, tipping also has various ⇑ Corresponding author at: Department of Business Administration, Guilford Glazer Faculty of Business and Management, Ben-Gurion University of the Negev, P.O.B. 653, Beer-Sheva 84105, Israel. Tel.: +972 8 6472675; fax: +972 8 6477691. E-mail address: [email protected] (O.H. Azar). http://dx.doi.org/10.1016/j.joep.2015.06.007 0167-4870/Ó 2015 Elsevier B.V. All rights reserved.

14

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

business implications.1 Second, the monetary magnitude of tipping is large. Azar (2011) estimates, for example, that annual tipping in the US food industry alone is over $46 billion, and tips are given in dozens of industries and countries (Lynn, Zinkhan, & Harris, 1993; Star, 1988).2 Third, tips have significant impact on the lives of millions of workers who obtain much of their income, or even most of it, from tips.3 Fourth, as opposed to some other social norms, tips can be measured accurately. Consequently, it is easier to analyze tipping behavior than to analyze many other social norms both empirically and theoretically (see for example Azar, 2004). Finally, as opposed to social norms that following them can have also a rational explanation (e.g., following good table manners or etiquette in order to have better chances to make profitable business with the others who observe you in the future), tipping by a customer who dines alone in a restaurant that he does not plan to return to (e.g., in another city or a foreign country) is hard to reconcile with a rational and selfish customer (who does not have social or psychological motivations). In this article we analyze how tipping behavior was affected in a field experiment that we conducted, where customers who dined in an Israeli restaurant and paid with cash received too much change.4 The level of extra change given varied between treatments (10 versus 40 Shekels, equal to about $3 versus $12) and we know whether customers reported the extra change and returned it or not. This allows us to analyze how these variables affected tipping. We hypothesized that for customers who do not return the extra change the unexpected gain they get may result in higher tips (compared to those who returned the extra change). This can be due to a good mood5 or a self-perceived feeling of luckiness or good-fortune, which follow from the unexpected gain. Higher tips when customers do not return the extra change can also result from a positive income effect and having additional monetary resources from which to pay the tip.6 Alternatively, it can result from customers treating the increased tip as a small contribution out of the extra change they received, which may be perceived as less costly than increasing the tip in regular occasions. For example, applying prospect-theory (Kahneman & Tversky, 1979) in this context suggests that normally increasing the tip could be perceived as a larger loss, but when the increased tip is given from erroneous extra change, it is just a reduction of a gain and is therefore less costly (due to the asymmetry between losses and gains according to prospect theory). Additional reasons that may affect the interaction between the decision whether to return the extra change and the tipping behavior are the ideas of moral cleansing and moral licensing. Sachdeva, Iliev, and Medin (2009) suggest that affirming a moral identity leads people to feel licensed to act immorally (moral licensing), and on the other hand, when moral identity is threatened, a person may regain some lost self-worth by adopting moral behavior (moral cleansing). A prior expression that helps to establish a certain self-concept (e.g., being moral or non-prejudiced) then liberates an individual to make choices that are inconsistent with that expression (Kouchaki, 2011). These ideas apply to our context because subjects in our experiment faced two subsequent decisions: whether to return the extra change they received and how much to tip.7 A subject who returned the extra change should feel moral due to his honest behavior and then according to moral licensing may feel less need to tip generously, thus lowering his tip. One who did not return the extra change, on the other hand, may want to correct the dishonest self-image that results from it by tipping generously, according to the moral cleansing idea. The result in both cases is that returning the extra change will be associated with lower tips. However, while moral licensing and more cleansing suggest that one form of prosocial behavior may crowd out another prosocial behavior, there is some evidence that this does not always happen. Greenberg (2014), for example, examines seasonal differences in within-customer restaurant tipping behavior and finds that during the holiday season (i.e., around Christmas) when there is a prosocial norm to be more generous, tipping rates are also higher, suggesting that two prosocial norms may be complementary rather than competing. A reason that may create in our experiment an effect in opposite direction to that of moral licensing and moral cleansing is the subject’s personality and self-selection. If some people are more generous or altruistic than others, we may see them returning the extra change more often and also tipping more,

1 Azar (2011) discusses the implications of tipping for business strategy, asking questions such as whether firms are better off imposing a compulsory service charge in lieu of tipping, and can some industries use voluntary prices determined by the customers as a business model? Lynn and Wang (2013) ask why service firms allow their employees to be paid by tips despite the associated risks, such as collusion between employees and customers against the firm. 2 Some examples of tipping are well known and are even mentioned in travel guides (to educate the tourist about the customs in the country he visits), such as tipping in restaurants, taxis or hotels. Other examples are less familiar and may be country specific. For example, Saunders and Lynn (2010) study tipping car guards in South Africa. 3 Wessels (1997) suggests that over two million US workers are servers as their primary occupation, in addition to many others who are servers as their secondary occupation, and servers earn most of their income from tips. 4 In this article we analyze the tipping behavior in this experiment. In another article we analyze what affected the customer’s decision whether to return the extra change (Azar, Yosef, & Bar-Eli, 2013). 5 Gueguen and Legoherel (2000), for example, find that drawing the sun on the bill increased both the frequency and size of tips in bars, possibly due to creating a more positive frame of mind. 6 The positive income effect also implies that among customers who do not return the extra change, those who receive 40 extra Shekels will tip more than those who receive 10 extra Shekels. 7 Diners in Israeli restaurants almost always tip with cash and not a credit card. In any case, our experiment was limited to customers paying by cash (because we could not give too much change to someone paying with a credit card), and then tips are given when the customer leaves some cash on the table when exiting the restaurant. This means that the customer can decide on the tip after knowing that he received too much change and after deciding whether to return the extra change (even if he decided how much he wants to tip before that, he can easily update his decision about the tip afterwards). Therefore we assume that the decision whether to return the extra change is made before the decision about the tip. Moral licensing and moral cleansing then imply that returning the extra change should be associated with lower tips. This prediction, however, remains the same also if the decision about the tip was first and the decision whether to return the extra change was second.

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

15

compared to the others. That is, they self-select to belong to the group that returns the extra change due to their generosity or altruism, and their generosity or altruism also leads them to tip more. 2. Method In order to take advantage of the natural conditions of a field experiment (with subjects not being aware that this is part of an experiment), we attained the owners’ and management’s approval to conduct the experiment at a restaurant in central Israel. Over a period of about seven months (between March and September), waiters were asked to report to one of this article’s authors (who worked at that time in the restaurant as a manager) in a side room about tables with one or two adult customers who paid the bill with cash. In these tables, after paying the bill, the customers received too much change, either 10 extra Shekels (about $3) or 40 extra Shekels. Overall, 192 observations were collected. The experiment was conducted during all days of the week and all operating hours of the restaurant, from lunch until late night. The tables from which we collected observations were located in the same area, to ensure that light, temperature, music and atmosphere are similar across observations and do not create unnecessary noise in the data. As part of the regular restaurant’s procedures, waiters stay close to a table that received the bill, so that customers who want to leave are not delayed. This allowed the waiters to observe whether the diners share the bill, and also made it easy for customers who wanted to return the excessive change to do so. After the customers left the restaurant, the waiters reported to one of the authors (in a separate room that is not visible to the restaurant’s customers) several variables, including the customers’ satisfaction from the meal (asking them about it is part of the regular procedure in the restaurant), who paid the bill, the tip left by the customers, and whether they reported and returned the extra change. In addition, we collected some additional data. The author who served as a manager in the restaurant knew about incidents related to the meal, such as dishes that were delayed, dissatisfaction of the customers from the service, appraisal from the customers, etc. This together with the feedback collected from the waiters was combined to create a customer satisfaction variable (on a 1–3 scale with 3 being high satisfaction). The restaurant’s occupancy level was also recorded on a 1–3 scale (3 being high occupancy). The receipts were used to track additional data: the bill’s amount, the time when the order was made, whether the customers have the restaurant’s membership card, which meal was ordered, and whether it included alcoholic beverages. In addition, we recorded the date and day of the week, the time of the customers’ exit from the restaurant (and then we computed the duration of the customers’ visit), and the gender of the main server who provided service to the table. The customers were divided to repeat and one-time customers based on several parameters. First, the waiters reported whether they have seen the customer in the restaurant before. Second, one of the authors had worked in the restaurant for three years and knew many customers. Third, the regular procedure in the restaurant involved asking the customers whether they know the restaurant and/or the menu and if they dined in it previously. Finally, we knew whether the customer holds the restaurant’s membership card. Such a card can be purchased and then provides certain discounts and benefits, so the waiters are regularly obliged to ask the customers if they have this membership card in order to give them the relevant benefits. If one of the above considerations suggested that the customer had visited the restaurant before then he was classified as a repeat customer.8 The sample chosen created a balanced sample between the main variables of interest that we could control (the amount of extra change – 10 or 40 Shekels, repeat versus one-time customers, one versus two diners, and gender).9 The 97 observations of two-diners tables are divided roughly equally between two women, two men, and a woman with a man. 3. Analysis and results Table 1 presents the summary statistics and the explanations of the main variables. Tips ranged between 2 and 36 Shekels (all 192 tables tipped). Bill amount was between 7 and 254 Shekels. The duration of visit was between 20 and 111 minutes. Fig. 1 presents the tip averages in the two extra-change treatments. Fig. 2 shows the tip averages considering the interaction of the two extra-change treatments and whether the subject decided to return the extra change. Table 2 presents a correlation matrix of several main variables. To control for various other variables that may influence tipping (e.g., bill size), Table 3 presents three regressions that analyze the tipped amount. We use the tip amount (in Shekels) as the dependent variable. As Lynn, Jabbour, and Kim (2012) explain, because most customers have weak math skills and dislike effortful thinking, it seems unlikely that customers will take all considerations into account, compute a very precise percentage tip, and then multiply it by the bill and leave as a tip exactly this amount. Such strategy also requires more effort to find the correct change. Instead, a simpler and easier method to decide on the tip, which is more plausible as Lynn et al. explain in detail, is that the customer takes a certain percentage that he considers as a reasonable average tip or that he finds easy to compute (e.g., 15% of the bill), multiplies this by the bill, and then makes all adaptations (e.g., for service quality) on this amount. We can also look at the tip amounts versus tip percentages. Lynn et al. report that tip amounts are clustered around dollar amounts rather than percentages, which also suggests that 8

In tables of two diners, if one was a repeat customer the table was classified as having a repeat customer. The same applies to holding a membership card. All tables in the pre-determined area of the restaurant who paid with cash, with one or two diners, and in shifts when the co-author who served as manager was present, were added to the experiment with alternating amounts of extra change. Once a category (e.g., one male, repeat customer) had the pre-determined number of observations filled (due to the alternating amounts of extra change this also meant equal number of 10-Shekels and 40-Shekels observations), tables in that category were no longer added to the experiment. 9

16

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

Table 1 Summary statistics and explanations of the main variables. Variable

Explanation

Mean

STD

Tip-amount Tip-percent High-Change Report Bill Repeat Member Satisfaction Occupancy Booth Minutes Business Alcohol Male-waiter Lunch Evening Night Weekend Summer Sharing

Tip in Shekels Tip in percentage of the bill 1 if 40 Shekels extra change, 0 if 10 Shekels 1 if excessive change was returned, 0 otherwise Bill in Shekels 1 if a repeat customer, 0 otherwise 1 if the customer is a restaurant club member, 0 otherwise Satisfaction level on a 1–3 scale (3 = high) Restaurant occupancy on a 1–3 scale (3 = high) 1 if a booth table, 0 otherwise Duration of stay in the restaurant (in minutes) 1 if the customer had a business meal, 0 otherwise 1 if the customer consumed alcohol, 0 otherwise 1 if the server is male, 0 otherwise 1 if the order is made between 12:00 and 16:59, 0 otherwise 1 if the order is made between 17:00 and 21:59, 0 otherwise 1 if the order is made between 22:00 and 03:00, 0 otherwise 1 if weekend, 0 otherwise 1 if July–August, 0 otherwise 1 if the bill was paid by more than one person, 0 otherwise

12.05 13.34 0.50 0.33 97.94 0.51 0.06 2.84 1.87 0.38 48.69 0.61 0.11 0.31 0.40 0.26 0.34 0.19 0.53 0.34

5.80 9.87 0.50 0.47 47.24 0.50 0.23 0.48 0.75 0.49 13.78 0.49 0.32 0.46 0.49 0.44 0.48 0.39 0.50 0.47

14 12 10 8 6 4 2 0

10 Shekels extra change

40 Shekels extra change

Fig. 1. Average tip by extra-change treatment.

14 12 10 8 6 4 2 0

Returned 10 Shekels Returned 40 Shekels

Unreturned 10 Shekels

Unreturned 40 Shekels

Fig. 2. Average tip by extra-change treatment and decision to return.

people make the adaptations on the dollar tip rather than on the percentage. In our sample we observe similar behavior. When looking at tip percentages (with two numbers after the decimal point), only 7 out of 192 are integer numbers (e.g., 11%, 13%, etc.). When looking at the exact Shekels tip, however, 181 out of 192 tips are integer numbers, and out of the remaining 11, seven are round to the closest half Shekel (e.g., 14.5 or 17.5 Shekels). The Shekel is divided to 100 Agorot and a coin of 10 Agorot is very common, so it should not be a problem to give amounts such as 12.60 or 7.30 Shekels for someone who wants, implying that the amounts left being mostly round numbers are intentional and not due to technical difficulty to leave other amounts. Because people take into account the bill but then make their adaptations on the Shekel tip, it is more appropriate to use as the dependent variable the tip amount and not the tip percentage (but taking into account that the bill is one of the most important explanatory variables by adding it to the regressions). The variable High-Change equals 1 if the extra change was 40 Shekels and to 0 if the extra change was 10 Shekels. Report equals 1 if the subject reported the extra change and returned it and 0 otherwise. Bill is obviously expected to be positively

17

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21 Table 2 Correlation matrix of several main variables. Tip-amount Tip-amount High-Change Report Bill Repeat Member Satisfaction Alcohol Male-waiter % of male diners

High-Change

1.00 0.07 0.02 0.90 0.04 0.03 0.00 0.16 0.07 0.07

1.00 0.38 0.03 0.01 0.07 0.01 0.07 0.01 0.01

Report

Bill

Repeat

Member

Satisfaction

Alcohol

Male-waiter

% of male diners

1.00 0.03 0.41 0.16 0.02 0.01 0.06 0.26

1.00 0.04 0.10 0.06 0.18 0.04 0.06

1.00 0.24 0.10 0.04 0.04 0.02

1.00 0.08 0.02 0.02 0.10

1.00 0.02 0.03 0.00

1.00 0.03 0.03

1.00 0.03

1.00

Table 3 The effect of the extra change and whether it was returned. Variable

Intercept High-Change Report High-Change⁄Report Bill Bill-squared Repeat Member Satisfaction Occupancy Booth Minutes Business Alcohol Male-waiter Evening Night Weekend Summer Men-diners Mixed-diners Woman-diner Women-diners Sharing N R2

(1)

(2)

(3)

Coefficient

p-Value

Coefficient

p-Value

Coefficient

p-Value

2.756 0.669

0.112 0.054

2.775 0.986 0.934

0.103 0.003 0.085

0.063 0.00014 0.665 0.218 0.732 0.393 0.033 0.011 0.005 0.088 0.868 1.052 1.092 0.076 1.108 3.265 1.756 0.298 2.473 0.832 192 0.850

0.007 0.160 0.074 0.786 0.014 0.179 0.931 0.348 0.993 0.843 0.123 0.031 0.047 0.857 0.074 0.000 0.040 0.482 0.001 0.282

0.064 0.00014 1.026 0.087 0.655 0.383 0.008 0.013 0.012 0.068 0.814 0.998 1.101 0.192 1.080 3.382 1.861 0.026 2.892 0.979 192 0.854

0.007 0.170 0.031 0.915 0.023 0.178 0.984 0.280 0.982 0.879 0.139 0.039 0.048 0.656 0.073 0.000 0.024 0.951 0.001 0.178

2.769 0.957 0.998 0.099 0.064 0.00014 1.024 0.088 0.654 0.38 0.0002 0.013 0.017 0.061 0.812 1.004 1.101 0.187 1.085 3.375 1.850 0.027 2.888 0.981 192 0.854

0.104 0.026 0.266 0.912 0.008 0.179 0.030 0.915 0.024 0.178 1.000 0.282 0.977 0.896 0.139 0.038 0.049 0.672 0.080 0.000 0.033 0.949 0.001 0.176

Comments: The dependent variable is the tip in Shekels. The estimation uses robust standard errors.

and significantly associated with the tip. Bill-squared is added to observe if the relation between tip and bill is non-linear. The dummy variable Repeat captures whether the customer has visited the restaurant before and is therefore considered a repeat customer. We expected to find that repeat customers tip more. They may already know the waiter and therefore feel closer to him, and they are more likely to return in the future and may tip more due to this. The variable Member designates a customer who holds the restaurant’s membership card, which costs a small annual fee but gives certain discounts, and therefore is likely to be associated with a customer who visits the restaurant with relatively high frequency. Therefore we expected it to also be associated with higher tips. Satisfaction is also expected to be positively correlated with tips. Higher occupancy, a booth table, and a shorter visit may result in less contact with the waiter and possibly lead to lower tips. Business meals suggest an attempt to take advantage of deals with lower prices for the food and a tendency to save money, possibly also in the tip. Alcohol consumption is not cheap, suggesting smaller propensity to save, and may also be associated with better mood and therefore lead to higher tips. We also incorporate some other control variables without clear prior expectations about the direction of their influence, such as Weekend, Summer (July–August), Male-waiter, and Sharing (two people paying the check). We take Lunch (order between 12:00 and 16:59)10 as a benchmark and add Evening (order between 17:00 and 21:59) and Night (order between 22:00 and 03:00) as dummy variables. A single male diner is the benchmark and four dummy variables (Men-diners, Mixed-diners, Woman-diner, Women-diners) capture the other possibilities of diner composition (recall that all tables are of one or two diners). 10

Israeli lunch times are on average later than in the US, and an order around 16:00 is more of a late lunch than an early dinner.

18

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

In regression (1) we isolate the effect of the level of extra change given. The coefficient of High-Change is positive (0.669) and statistically significant at the 10% level (p-value = 0.054).11 Regression (2) adds Report as an explanatory variable and we see that Report has a negative coefficient ( 0.934) that is statistically significant at the 10% level (p-value = 0.085). Now the coefficient of High-Change becomes higher (0.986) and its statistical significance improves (p-value = 0.003). The reason is that in the experiment there was a positive correlation (0.376) between High-Change and Report and therefore in regression (1) when Report is omitted, High-Change captures also some of the negative effect of Report. Regression (3) adds also the interaction effect of High-Change and Report. Once again, High-Change is positive (0.957) and statistically significant (p-value = 0.026), and Report is negative ( 0.998). Here, however, despite the substantial coefficient, Report is not statistically significant (p-value = 0.266). The interaction term High-Change⁄Report is small (0.099) and far from being statistically significant (p-value = 0.912). Let us first address the negative coefficient of Report. As mentioned in the introduction, customers who do not return the extra change may tip more than those who return the extra change, due to good mood, self-perceived luckiness, positive income effect or considering the increased tip as a contribution out of the extra change they received, which may be perceived as less costly. Moral licensing and moral cleansing also lead to this expected result. The coefficient being negative and substantial is in line with these predictions. However, in regression (3) the coefficient of Report is not statistically significant. A possible reason is that the personality and self-selection ideas are also true: people who are more generous or altruistic in general, will be more likely to return the extra change and will tip more on average, creating an opposite (positive) effect on the coefficient of Report, which results in the total effect of Report being less significant. The interaction term High-Change⁄Report having a negligible (and not statistically significant) effect, whereas High-Change itself has a positive and statistically significant effect, are together interesting findings that are not what we expected, but do offer some insights about the reasons for the observed behavior. If tipping more when one receives the higher amount of extra change is driven by the positive income effect and having the additional resources of the extra change, or by treating a larger tip as a reduced gain rather than an increased loss, then the high level of extra change should have a positive effect on tipping if it is not returned, but not if it is returned. This implies a positive coefficient of High-Change and a negative coefficient of the interaction term High-Change⁄Report. We see the former but not the latter and therefore these potential explanations are not supported by the data. The potential reasons of mood or feeling lucky as a source for higher tipping are less straightforward in their possible implications on behavior. A reasonable assumption is that these feelings are stronger when one receives 40 extra Shekels than when he receives 10 extra Shekels. This is also consistent with the positive coefficient of High-Change in the regressions. The expected coefficient of the interaction High-Change⁄Report, however, is less clear. One possibility is that a better mood or perceived luckiness results when the subject receives an unexpected amount and keeps it, but not when he returns it. But another possibility is that receiving the extra change already creates good mood or self-perceived luckiness, and these do not change (or at least do not disappear completely) even if the subject returns the extra change. Assuming that High-Change is positive (as is indeed the case), the former possibility implies that the interaction High-Change⁄Report should be negative (so that the net effect of High-Change is close to zero if the change was returned), whereas the latter possibility implies that the interaction High-Change⁄Report should be small relative to the coefficient of High-Change, and possibly even close to zero. Because the interaction is indeed negligible in the data, it follows that to the extent that better mood and perceived luckiness are created when one receives too much change and encourage more tipping, the good mood and perceived luckiness remain even if the extra change is returned. An additional possible reason for the results is that a subject who experiences an uncommon event of getting too much change has a sense of individuation or an unconscious fear of being observed and therefore he tips more. In this case it is also plausible that this effect is stronger when receiving 40 extra Shekels than when receiving 10 extra Shekels, and that returning the extra change does not eliminate the reason for the higher tips, and so this potential reason is also consistent with the results. Another explanation that can be consistent with our findings is that cost-conscious consumers were more likely to return the excessive change and were also more likely to tip less.12 The data also allow us to explore some aspects of tipping behavior that are not directly related to our experimental intervention of giving too much change. Because these aspects are not the main focus of the paper, their analysis and discussion is relegated to the Appendix. 4. Discussion and conclusion In a field experiment, excessive change was given to restaurant customers paying in cash.13 We collected information on several variables pertaining to the dining experience, including tip and bill size. The extra change given was either 10 or 40

11 To better understand the magnitudes, recall that the average tip in the sample is about 12 Shekels. One US Dollar was about 3.5 Shekels at the period when the experiment was run. 12 In an attempt to gain more insights about the various possible reasons we also ran several regressions where additional interactions of High-Change and Report with other variables were added to the regression. However, these interactions did not have statistically significant coefficients or result in additional insights and therefore we do not report in detail these additional regressions here. 13 We could only use diners who paid with cash because those who pay with a credit card are not given change. Obviously, the customer decides how to pay and to the extent that characteristics of customers who choose to pay with cash are different from those of customers who pay with a credit card, our results are limited to cash-paying customers.

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

19

Shekels. One third of the tables reported the extra change to the server and returned it. We analyze how the level of extra change and the decision to return it are correlated with tipping behavior. We find that receiving the higher level of extra change was associated with larger tips. Returning the extra change was negatively correlated with tips,14 but despite the substantial magnitude of this result, it was not always statistically significant. The interaction between the level of extra change and the decision whether to return it had no effect. The results suggest that after receiving excessive change of 40 Shekels, tipping is higher than after receiving 10 extra Shekels, even when the excessive change is returned. This pattern of behavior implies that the income effect of having more monetary resources from which to pay the tip does not seem to significantly affect tipping behavior here. The results also do not support the prospect-theory explanation according to which tipping out of excessive change is less costly than usual tipping because it is a forgone gain instead of a loss. Several other potential explanations about how receiving excessive change and the decision whether to return it may be correlated with tipping behavior are consistent with the data. Subjects possibly exhibited some moral licensing and moral cleansing effects.15 Therefore those who decided to behave honestly and return the extra change felt liberated to tip less, whereas those who kept the extra change tipped more to regain a good self-image after behaving dishonestly. However, there may also be a self-selection effect driven by the subject’s personality: those who are more generous (or altruistic) by their nature are more likely to return the extra change and also more likely to tip generously. This goes in the opposite direction to the effects of moral cleansing and moral licensing, and thus mitigates their effect. Another possible explanation is that cost-conscious consumers were more likely to return the excessive change and were also more likely to tip less. The findings that subjects who received 40 extra Shekels tipped more on average than those who received 10 extra Shekels, but that this difference was similar when the extra change was returned and when it was not returned, yield two implications. One is that receiving the larger amount of extra change (40 Shekels) may result in feelings of good mood, perceived luckiness, or individuation and unconscious fear of being observed, which increase tips. The second is that these feelings remain even if the extra change was returned.16 Acknowledgements We are grateful to Shai Danziger, Simone Moran and two anonymous referees for helpful comments. We also thank for their valuable feedback the participants in the Deception, Incentives and Behavior conference in the Rady School of Management, UC San Diego and in the SABE-IAREP-ICABEEP 2013 conference, and in seminars given in the Max Planck Institute of Economics - Strategic Interaction Group, University Paris 13, University of Granada and Ben-Gurion University of the Negev. We also thank the staff of the restaurant for their cooperation and help in conducting the research and to the management for allowing us to run the experiment in the restaurant. Appendix. Additional results about tipping behavior The results of our experiment allow us to look at several aspects of tipping beyond the main topic of the extra change given. We will discuss the results from regression (3) in Table 3; the other two regressions yielded results that are qualitatively similar. The bill size is positively correlated with the tip as expected (p-value = 0.008). The coefficient seems a bit low, representing 6.4% of the bill; however, this is because we also include in the regression the squared-bill, which has a positive coefficient. When we run the same regression without the squared-bill, the coefficient of the bill becomes 0.1 and the p-value becomes 0.000 (t-statistic of 11.02). The effect of the squared-bill variable is positive but is not statistically significant. Repeat customers, as expected, tip more, by about one Shekel (p-value = 0.030). This can be the result of repeat customers feeling closer to the server (because they encountered him also in the past) and therefore being more generous.17 Alternatively, repeat customers may tip more because they are more likely to return to the restaurant in the future and tipping more generously can make them more comfortable in their future visits, or encourage the server to provide them good service in the future. The literature is mixed on the effect of repeat customers. For example, Kahneman, Knetsch, and Thaler (1986) interviewed people over the telephone and asked two alternative questions. Both questions began ‘‘If the service is satisfactory, how much of a tip do you think most people leave after ordering a meal costing $10. . .’’ One question ended ‘‘. . . in a restaurant that they visit frequently?’’ and the second ‘‘. . . in a restaurant on a trip to another city that they do not expect to visit again?’’ The average responses were $1.28 and $1.27 (the number of observations was 122 and 124). These answers indicate that people believed (probably based on their own experience as tippers) that repeat customers do not tip significantly more 14 We do not know whether the difference in tipping behavior between those who returned the excessive change and those who did not comes from the behavior of the former group, the latter group, or both. This is a possible direction for future research. 15 An interesting direction for future research is to test whether customers have some threshold for prosocial behavior that they want to attain. This can be thought of as prosocial or moral satisficing, i.e., customers care a lot about their prosocial behavior before reaching that threshold but are happy once they reach it and do not attempt to increase their prosocial behavior further. 16 Subjects who received 10 Shekels as extra change may also experience these feelings but the significant effect of the high-change treatment shows that if they experience such feelings, they are less pronounced than for subjects who received 40 extra Shekels. 17 Lynn and Mynier (1993), for example, found that when the server squatted during the initial visit to the table, tips were higher, a result that was replicated later by Davis, Schrader, Richardson, Kring, and Kieffer (1998). The proposed explanation for this result is that when servers squat, they are closer to customers, which may allow additional eye contact and foster higher rapport and enhanced communication.

20

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

than non-repeat customers. Lynn and McCall (2000) and Conlin, Lynn and O’Donoghue (2003), however, found a significant and positive correlation between patronage frequency and tip size. We thought that holders of the restaurant membership card, who tend to be customers with higher patronage frequency than other repeat customers, will tip more, but we found no difference between membership holders and other repeat customers. The variable of satisfaction has a positive coefficient as predicted, which is also statistically significant (p-value = 0.024). Satisfaction is ranked as either 1, 2 or 3, and an additional point raises the tip on average by about 0.65 Shekels. We predicted that occupancy level (ranked on a 1–3 scale) may have a negative effect, because a busier restaurant can result in less contact and closeness between the waiter and the customers, and also in service of lower quality (which is captured, but possibly not fully, in the satisfaction variable). The coefficient of Occupancy is negative as predicted but is not statistically significant. Booth, Minutes, Business and Alcohol have very small coefficients and are not statistically significant. Male waiters receive on average tips that are 0.81 Shekels higher than those of waitresses, but the difference is not statistically significant (p-value = 0.139). Dining in the evening or night results in a tip that is about one Shekel smaller than in lunch (p-value of 0.038 for Evening and 0.049 for Night). A major difference between the meal times is that lunch is often during a lunch break from work (and in the case of two diners, with a colleague from work) whereas dining in the evening or night is unlikely to be during work hours, and in the case of two diners they are more likely to be friends or a couple rather than work colleagues. To examine whether tables of two diners (and thus the different relationship between the diners in lunch versus evening or night) are related to the effect of Evening and Night we ran a regression where two interaction terms were added, of Evening and Night with a variable designating a two-diner table. The coefficient of the interaction term was close to zero for Night, close to 1 for Evening, and far from being statistically significant in both cases. Therefore it is hard to say what exactly is the reason for the different tipping behavior over the day. Tips over the weekend do not differ much from those during the mid-week, but in the summer months of July and August tips tend to be about one Shekel lower (p-value = 0.080). In the summer we are likely to see more cases of dining in one’s free time, with friends or family, than during the rest of the year. This is also true for evening and night versus lunch. But if this is related to the reason for the lower tips at evenings, nights and the summer, it is not clear why we do not observe also a negative coefficient for the weekend variable. Finally, let us look at the impact of diner composition (gender and number of diners). The benchmark here is a single male customer, and four dummy variables capture the other possibilities. We see that a single woman tips the same as a single man. Two diners, in all three gender possibilities, tip more than a single man. Two men tip the most (3.38 Shekels more than a single man, p-value = 0.000), two women are next (coefficient of 2.89, p-value = 0.001) and a mixed table of one man and one woman tips 1.85 Shekels more than a single man (p-value = 0.033). It is interesting that mixed tables tip below two men and below two women, but we are not sure what is the reason for this. Also interesting is the negative and relatively large (though not statistically significant) coefficient of Sharing, which suggests that among tables with two diners, when both paid the bill, the tip was lower by about one Shekel than the average tip in tables of two diners when only one diner paid the joint bill. The reason for higher tips with two diners may be that with a fellow diner there is some social pressure to be generous. In the literature the results about the correlation between group size and tips are mixed. For example, Freeman, Walker, Borden, and Latané (1975) examined 396 groups of restaurant diners and found a negative correlation between group size and tips, claiming that the findings are consistent with a theory of diffusion of responsibility. Lynn and Grassman (1990), however, found no correlation between group size and tips, and Conlin, Lynn and O’Donoghue (2003) found a positive correlation between tip percentage and group size. It should be pointed out that in our sample the variation of group size is only between one to two diners whereas in most studies the variation is larger and includes tables with more than two diners. References Azar, O. H. (2004). What sustains social norms and how they evolve? The case of tipping. Journal of Economic Behavior and Organization, 54, 49–64. Azar, O. H. (2011). Business strategy and the social norm of tipping. Journal of Economic Psychology, 32, 515–525. Azar, O. H., Yosef, S., & Bar-Eli, M. (2013). Do customers return excessive change in a restaurant? A field experiment on dishonesty. Journal of Economic Behavior and Organization, 93, 219–226. Conlin, M., Lynn, M., & O’Donoghue, T. (2003). The social norm of restaurant tipping. Journal of Economic Behavior and Organization, 52, 297–321. Davis, S. F., Schrader, B., Richardson, T. R., Kring, J. P., & Kieffer, J. C. (1998). Restaurant servers influence tipping behavior. Psychological Reports, 83, 223–226. Freeman, S., Walker, M. R., Borden, R., & Latané, B. (1975). Diffusion of responsibility and restaurant tipping: Cheaper by the bunch. Personality and Social Psychology Bulletin, 1, 584–587. Greenberg, A. E. (2014). On the complementarity of prosocial norms: The case of restaurant tipping during the holidays. Journal of Economic Behavior and Organization, 97, 103–112. Gueguen, N., & Legoherel, P. (2000). Effect on tipping of barman drawing a sun on the bottom of customers’ checks. Psychological Reports, 87, 223–226. Kahneman, D., Knetsch, J. L., & Thaler, R. (1986). Fairness as a constraint on profit seeking: Entitlements in the market. American Economic Review, 76, 728–741. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263–291. Kouchaki, M. (2011). Vicarious moral licensing: The influence of others’ past moral actions on moral behavior. Journal of Personality and Social Psychology, 101(4), 702–715. Lynn, M., & Grassman, A. (1990). Restaurant tipping: An examination of three ‘‘rational’’ explanations. Journal of Economics Psychology, 11, 169–181. Lynn, M., Jabbour, P., & Kim, W. G. (2012). Who uses tips as a reward for service and when? An examination of potential moderators of the service-tipping relationship. Journal of Economic Psychology, 33, 90–103. Lynn, M., & McCall, M. (2000). Gratitude and gratuity: A meta-analysis of research on the service–tipping relationship. Journal of Socio-Economics, 29, 203–214.

O.H. Azar et al. / Journal of Economic Psychology 50 (2015) 13–21

21

Lynn, M., & Mynier, K. (1993). Effect of server posture on restaurant tipping. Journal of Applied Social Psychology, 23, 678–685. Lynn, M., & Wang, S. (2013). The indirect effects of tipping policies on patronage intentions through perceived expensiveness, fairness, and quality. Journal of Economic Psychology, 39, 62–71. Lynn, M., Zinkhan, G. M., & Harris, J. (1993). Consumer tipping: A cross-country study. Journal of Consumer Research, 20, 478–488. Sachdeva, S., Iliev, R., & Medin, D. L. (2009). Sinning saints and saintly sinners: The paradox of moral self-regulation. Psychological Science, 20(4), 523–528. Saunders, S. G., & Lynn, M. (2010). Why tip? An empirical test of motivations for tipping car guards. Journal of Economic Psychology, 31, 106–113. Star, N. (1988). The international guide to tipping. New York, NY: The Berkeley Publishing Group. Wessels, W. J. (1997). Minimum wages and tipped servers. Economic Inquiry, 35, 334–349.