Restaurant Tipping: Free-Riding, Social

Restaurant Tipping: Free-Riding, Social
Acceptance, and Gender Differences’
WILLIAMJ. BoYEs2
WILLIAMSTEWART
MOUNTS,JR.
Department of Economics
Arizona State University
Stetson School of Business and Economics
Mercer University
CLIFFORDSOWELL
Department of Economics
Berea College
The practice of paying gratuities for services is a worldwide custom. Tipping is found only
in some professions, which suggests that it serves to increase the efficiency of specific
kinds of exchanges. The literature accepts the view that monitoring of employees by customers appears to be the logical rationale for the practice of tipping. This can be seen in
that gratuities are paid at the discretion of consumers after they receive the services for
which they are paying. However, it does not explain why, given the voluntary aspect of
tipping, rational people would not free-ride on the tipping of others. We found that both
men and women free-ride in their tipping behavior. Yet, we also found that men are more
influenced by social acceptance or approval in their tipping behavior than are women.
The practice of paying gratuities for services is a worldwide custom that
involves many professions. Some suggest that the practice originated in the
Middle Ages when journeying feudal lords purchased safe passage by tossing
handfuls of coins to beggars on the road (Schein, Jablonski, & Wohlfahrt, 1984).
Elsewhere, others claim that tipping grew out of the custom of vails, which
required visitors in Tudor, England, to pay their host’s servants for the extra
work their visit created (Shamir, 1984). Even the origin of the word tip is not
known. Tip may derive from (a) the Latin word stips, meaning gift; ( b ) the
Dutch word tippen, meaning to tap, as in tapping a coin on a table or glass to
attract a server’s attention; or (c) the words “To Insure Promptitude,” which were
written on collection boxes in 18th-century English coffee shops (Schein et al.,
1984).
‘The authors thank Linda Brennan, Corky Brown, Mona Land, and Jay Marchand for their comments. We remain responsible for any errors.
2Correspondence concerning this article should be addressed to William Boyes, Department of
Economics, College of Business, Arizona State University, Tempe, AZ 85257. E-mail: William
.Boyes(@asu.edu
2616
Journal of Applied Social Psychology, 2004, 34, 12, pp. 2616-2628.
Copyright 0 2004 by V. H.Winston & Son, Inc. All rights reserved.
RESTAURANT TIPPING
2617
Tipping is not observed in all professions and usually occurs only in
exchange for services. This suggests that it serves to increase the efficiency of
specific kinds of exchanges. Since many service employee inputs are difficult
and costly for employers (e.g., restaurants) to monitor, this function is
partially delegated to customers via the use of tips. The problem with this
convention is that gratuities are paid at the discretion of consumers after they
have received services. Also, why would anyone tip, since others in a party or in
the establishment will tip (or past patrons had tipped)? As such, a free-riding
effect is inherent in the tipping environment, which may work against the
monitoring role.
The interaction of monitoring with free riding is seen in a survey we conducted involving 360 people who were contacted at a shopping mall about their
motives for tipping at restaurants. In rank order of preferences, the reasons they
gave were as follows: (a) to ensure good service in the future, (b) to be fair to
servers, (c) to not be embarrassed, and (d) because everyone tips. Next, we asked
these same respondents why they might tip at a restaurant they would not frequent in the future. The four reasons noted did not change, even though customers could simply walk out without leaving a tip; that is, a free ride on the behavior
of others. In fact, no one said they would not tip, but several said they might tip
less than if they anticipated returning to the restaurant in the future. This was
followed with an inquiry into why the free-riding problem was apparently not an
important issue to the survey respondents. The principal reason given was that
they would be embarrassed not to tip.
In both the question of why a person tips and why a person might not freeride, we find that the patron desires not to be embarrassed or to be known as
“cheap.” This desire suggests that tipping and social approval (or wish to avoid
social disapproval) are related and that social approval or acceptance may mitigate the possibility of free-riding on the tipping of others.
The purpose of this paper is to examine the possible interaction between
monitoring, free riding, and social acceptance in restaurant tipping. Data derived
from a survey involving patrons and servers at 18 restaurants in Phoenix,
Arizona, are used to examine hypotheses about these aspects of tipping.
Tipping, Monitoring, and the Willingness to Free-Ride
Most restaurant owners partially delegate the monitoring of servers to patrons
and their tips. Servers want to be paid for the service they provide diners. Further,
they prefer to be paid more for these services as opposed to less. Free-riding
behavior on the part of patrons is partially the result of the absence of perfect
monitoring by servers, This is a result of the fact that tipping is ex post to service
and that servers are responsible for several tables. To overcome this, servers
can be expected to engage in certain types of behavior and to employ tactics
2618 BOYES ET AL.
that increase familiarity with the party, especially the payer. This may include
introducing oneself (Garrity & Degelman, 1990), touching patrons (Crusco &
Wetzell, 1984; Hornik, 1992), and bending down to be at the same height as the
patron (Lynn & Mynier, 1993).
Similarly, patrons want good service and may try to assure this outcome
through the prospect of a tip. Yet, a patron has an incentive to free-ride on the
behavior of others as a result of the absence of monitoring, which is partly a
result of the ex post (as opposed to ex ante) nature of the tipping relationship. For
example, they may free-ride on the tips of other patrons in the restaurant. This
would be likely if the patron (e.g., a tourist) had a good chance of never returning
to the restaurant. In another case, a patron may free-ride on others in his or her
party, regardless of whether the other members of the party tip. As stated previously, the ability to free-ride is more likely when individual behavior cannot be
monitored perfectly. The larger the party, the more difficult it is for the server to
monitor the behavior of any one individual and the harder it is for other members
of the party to monitor one of their own. Also, when the server is not able to identify the payer, the server is unable to give the payer extra service, as suggested in
the literature.
From a broader perspective, tipping can be viewed as a social norm
(Conlin et al., 2003). Holloway (1985) and Lynn and Grassman (1990) argued
that tipping is a means to purchase social approval. This argument follows from
the basic notion that compliance with social norms is motivated by a desire for
social approval and acceptance or a fear of social disapproval (Lynn and LatanC,
1984).
If tipping is an attempt to purchase social approval, from whom is a customer
purchasing social approval? If it is the server, then a lack of familiarity between
patron and server should enhance the possibility of free riding. Conversely,
familiarity and expected future interaction should increase the value of the
server’s social approval. Thus, customers who frequently dine at a restaurant or
who dine out in general should tip more. However, if social approval is being
purchased from other diners, then customers in groups should tip more than
patrons who are alone, until the size of the group is sufficient to encourage free
riding. This is supported by Snyder (1976) and Freeman, Walker, Borden, and
LatanC (1975) who found an inverse relationship between tip size and party size.3
Social approval also suggests that tips should be larger for couples or groups
whose members are not well acquainted versus couples who are married or
groups whose members know each other very well. In other words, if an impression is important-a date or a business meeting-then the tip should be larger
than if an impression is not important.
3The free-riding problem may be why, in some restaurants, a tip is automatically affixed to the
bills of large parties.
RESTAURANT TIPPING
2619
Data
An extensive survey of tipping behavior was conducted in November/
December 1995 in metropolitan Phoenix, A r i ~ o n a Spring
.~
1995 was used to
pre-test the survey instrument and the willingness of restaurant managers and
customers to participate. Following full payment, customers departing 18 different restaurants were asked a series of questions relating to their tipping b e h a ~ i o r . ~
Over 90% of those who were asked completed the surveys. Each survey was
completed during the hours of 6 p.m. and 10 p.m. on either Friday or Saturday.
The responses from every third completed survey were verified by an inquiry of
the server.
There were 20 surveys collected at each restaurant. Each questionnaire was
reexamined and deleted from the sample if it was incomplete or contained information that was obviously in error. Next, the remaining questionnaires were
classified into one of four restaurant categories where the categories are based on
the average bill. There were 40 surveys in each of the four categories for a total
of 160 observations.
The data are divided into three categories.6 The first category measures the
personal characteristics of the respondent. Income, gender, age, occupation, use
of a credit card, occupation, and so forth are included in this group. The second
category measures the tipping priors that consumers bring to a dining experience.
These include expected tipping percentage; frequency of dining in general and in
the current establishment; attitudes toward tipping and service; and other personal priors concerning tipping behavior, including the number of people in a
party. The final category characterizes the dining services just received. This
group includes the actual tip given, an appraisal of the actual food and service,
the number of people in the party, and whether the party’s mood was good and if
they consumed alcohol.
Method
The nature of the data and its internal structure affect the methodology
employed to measure empirically the relationships of interest described previously. The data are viewed as a set of four balanced (40 observations in each)
cross-sections. Within this setup, ordinary least squares (OLS) estimation of the
data pooled would not be appropriate. Such an approach would require the
restrictive assumption that the tipping behavior in each restaurant is identical.
4The efforts of 30 business students at Arizona State University who carried out the surveys are
appreciated. Any remaining errors are claimed by the authors.
SThe survey instrument appears in the Appendix A. The customer was contacted when he or she
left the table or exited the restaurant, depending on the desires of the restaurant manager.
6A dictionary of variables i s presented in Appendix B.
2620
BOYES ET AL.
Accordingly, an error-components methodology is deemed more appropriate
for the problem.7 Under an error-components model, the error term is assumed to
be comprised of three independent parts: (a) an error across a given cross-section,
(b) an error between cross-sections for a given observation, and (c) an error associated with a unique observation. Such a structure can be estimated using either a
fixed-effects (FE) or a random-effects (RE) format. Either could be used, and
each has advantages and disadvantages. In the present case, a fixed-effects model
would be useful if it is to be applied only to restaurants in the sample or in
Phoenix, Arizona. However, our view is that the four restaurant types are a random sample from a much larger population and, therefore, a random-effects
approach would be appropriate. We resolve this issue by using an empirical test
of fixed versus random effects.
Results
Initial results from a random effects setup are presented in Table 1. The
dependent variable is the tip left by the paying patron, in percentage terms. The
overall fit of the model, seen in the adjusted R2, is rather good, given the crosssectional nature of the data.8 Coefficients estimated with an error-components
model are interpreted as percentage-point changes in the tip percentage, holding
other factors represented by other explanatory variables constant. This is synonymous with a least squares multiple regression interpretation.
In tests not reported here, it was determined that there were significant differences in tipping behavior across the four classes of restaurants9 This argues for
the error-components format discussed earlier. Furthermore, the Hausman test
statistic shown in Tables 1 and 2 indicates that a random-effects model is more
appropriate than is a fixed-effects model.10
As presented, explanatory variables have been classified into three areas: personal attributes, tonight's attributes, and variables suggestive of free riding and
social acceptance. As shown, gender (men tip less than women) and a patron's
level of income are significant determinants of the percentage tip, holding other
factors controlled for in the model constant." That is to say, men tip less than do
women, independent of the effects of income, group size, service quality, and so
forth.
7A discussion of this methodology can be found in Davidson and MacKinnon (1993)
*Tests for heteroscedasticity were insignificant. In addition, judging by correlation coeffrcients,
multicollinearity appears not to be a problem. A correlation matrix is available from the authors.
9This is the result of an F test: estimated F(3, 144) = 9 . 5 3 , ~= .OOOO.
loThe Hausman
test statistic for a fixed-effects versus random-effects model is reported in
Table I .
"Other personal attributes that were insignificant include age (and age2), occupation, and
whether diners were related.
xz
RESTAURANT TIPPING
2621
Table 1
Results of the Random-Effects Model
R2 = .64
Sum of squared residuals = 1,152.80
Variance of residuals = 8.01
Adjusted R2 = .61
SE of regression = 2.83
Hausman test of HO: RE vs. FE: x2(3) = 12.07, p = .007
df = 146
Variable
Constant
Personal attributes
GENDER
1NC2
INC3
INC4
TIPAT
Tonight’s attributes
MOOD
DRINK
SERN
BILLN
Free riding and social acceptance
PHX
FREQO
PARTY
Note. HO = null hypothesis, RE
for variable definitions.
*lo% level of confidence.
Estimated
coefficient
SE
t
0.65
2.93
0.22
-0.67
2.32
2.82
3.24
0.06
0.45
1.47
1.45
1.56
0.07
- 1.47*
1.54
1.64
1.88
0.08
0.69
0.5 1
0.43
0.03
2.23*
3.21*
4.36*
2.88*
-0.65
0.1 1
0.04
0.48
0.1 1
0.23
- 1.36*
= random
1.58*
1.95*
2.08*
0.76
1.05
0.18
effects, FE = fixed effects. See Appendix B
The patron’s anticipated tip (TIPAT) was included to control for the tipping
priors brought to a restaurant. Its role is insignificant, however. It should be noted
that, in results not reported here, the relationship between the anticipated tip and
anticipated dining variables like quality and service were largely insignificant.
Anticipated tips may simply be a learned percentage.
As might be expected, the attributes of the night’s dining experience (MOOD,
DRINK, SERN, and BILLN), proxies for the monitoring role, were strong determinants of the tip percentage, holding other factors such as gender, income, party
2622 BOYES ET AL.
Table 2
Results of the Random-Effects Model With the Interaction of GENDER and
PARTY (GENPAR T)
R2 = .65
Sum of squared residuals = 1,121.3 1
Adjusted R2 = .62
Variance of residuals = 7.84
SE of regression = 2.80
Hausman test of HO: RE vs. FE: ~ ~ (=41 )2 . 5 0 , ~
= .014
df= 145
Variable
Constant
Personal attributes
GENDER
INc2
INC3
INC4
TIPAT
Tonight's attributes
MOOD
DRINK
SERN
BILLN
Free riding and social acceptance
PHX
FREQO
PARTY
GENPART
Estimated
coefficient
SE
t
1.99
2.96
0.67
-3.29
2.80
3.23
3.59
0.04
1.33
1.47
1.44
1.55
0.07
-2.47"
1.91*
2.24*
2.32*
0.59
1.30
1.70
1.93
0.08
0.69
0.51
0.43
0.03
1.89*
3.36*
4.51*
2.66*
-0.61
0.12
-0.48
0.90
0.47
0.11
0.33
0.43
-1.29*
1.12
-1.44*
2.09*
Nofe. HO = null hypothesis, RE = random effects, FE = fixed effects. See Appendix B
for variable definitions.
* 10% level of confidence.
size, and residency constant. The overall value of the night's service (SERN) had
the largest estimated coefficient.12 We suspect that these variables capture many
of the hedonic attributes of dining (e.g., the gender and attrativeness of the
12Service is being treated as a continuous variable.
RESTAURANT TIPPING
2623
server). Interestingly, as the night’s bill (BILLN) increased, percentage tipped
increased, holding other factors constant (e.g., group size). This may be suggestive of the overall quality of the restaurant and the meal. The role of income was
positive in tip determination.13 Holding other factors constant, increases in
income increased the percentage tipped.
The only significant variable measuring free riding and social acceptance is
whether or not the patron is a resident of metropolitan Phoenix. This variable is a
proxy for whether a patron will return to the particular restaurant. As shown, a
nonresident tip was 0.65 percentage points less than that of a resident, holding all
other things constant.’4
The monthly frequency of dining out (FREQO) and the number of people in
tonight’s party were marginally insignificant. Two opposing effects can be seen
playing a role in PARTY. The first is that the larger the party, the more likely the
patron will tip less. This effect implies that it is easier for a member of a party to
escape monitoring by other members of the party or the server; that is, to be able
to free-ride more easily. Yet, the larger the party, the greater the penalty in terms
of social acceptance on an individual who does free-ride. In this latter case, the
percentage tipped would rise as party size increases.l5
The results in Table 1 do not strongly support the presence of free riding,
social acceptance, or approval in the determination of percentage tipped. As a
result, the basic model was extended to see if there was a gender difference that
needed to be isolated so that the model could be estimated better. This was
prompted by the notion that men are punished more for failing (in general) than
are women (Aronson, Wilson, & Akert, 1994). This may suggest that there is a
gender difference in compliance with social norms. Given that free riding in tips
goes against the norm of tipping, there may be a gender difference in willingness
to engage in this type of tipping behavior.
The results of this extension are presented in Table 2. Again, a random effects
setup was employed.16 The results in Table 2 are an improvement over those
shown in Table 1.
The only difference between Table 2 and Table 1 is that a variable was added,
seen in an interaction between gender and the size of the party (GENPART).”
Given that PARTY is our proxy for the interaction of free riding and social
I3The omitted group is those individuals with incomes below $20,000.
I4This variable is highly correlated with the patron being a tourist or business traveler.
I5lt should be noted that the data were collected on Friday and Saturday nights. Also, a dummy
variable indicating whether or not a dinner was business-related was not significant. These considerations lead us to believe that groups were largely social in composition and in purpose.
I6The F statistic for differences in tipping behavior between restaurants was F(3, 144)= 9.46, p =
.OOOO.The Hausman x2 test statistic continues to reject the use of a fixed-effects model (Table 2).
‘7The interaction of gender with other variables, particularly frequency of dining and Phoenix
residence, were not significant.
2624 BOYES ET AL.
acceptance, ceteris paribus, GENPART represents the gender-specific aspect of
these two opposing factors. As shown, the estimated coefficient on gender
increased measurably, holding the influences of the other explanatory variables
constant.18 Furthermore, PARTY was negative and significant, indicating that
both men and women showed a net willingness to free-ride. However, the significant positive coefficient on GENPART suggests that the willingness of men to
free-ride in tipping was more than offset by their desire for social acceptance.
This reduces the overall tipping differential seen in the large coefficient on
GENDER.
Given that the remaining variable coefficients and their individual significance remained unchanged, this extension strongly supports the presence of a
gender differential in tipping behavior that goes beyond just the simple treatment
of gender differences. It must be noted that the estimated coefficient on
GENPART does not allow us to differentiate between the hypothesis that men are
trying to avoid social sanction by other party members or the hypothesis that men
are buying social acceptance.19 In either case, however, it seems to us that social
approval is being sought by men, given that other factors (e.g., party size, size of
the bill, level of income, etc.) are being held constant.
Tipping is a worldwide custom honored by most people. It stems, in part,
from the fact that some inputs, particularly those in the service industry, are
difficult and costly for managers to monitor. Consumers are placed in the
monitoring role through the act of tipping. In such an arrangement, however,
there is a possibility of a free-rider problem within multiperson dining parties
and across a restaurant that serves to reduce the efficiency of this form of
monitoring. The findings of this paper support the presence of free riding on
the part of men and women. Furthermore, we found this behavior on the part of
men to be offset by the presence of social acceptance sought in a tipping environment.
References
Aronson, E., Wilson, T. D., 19Akert, R. M. (1994). Socialpsychology: The heart
and mind. New York, NY: Harper Collins.
Conlin, M., Lynn, M., & O'Donoghue, T. (2003). The norm of restaurant tipping.
Journal of Economic Behavior and Organization, 52,291-321.
'8A reader suggested that the groups associated with male tippers may have greater range in size
and that this may explain male tipping. However, the impact of group size is being held constant
when GROUP is included in the estimation. Therefore, the role of gender shown in the coefficient is
independent of the range of group sizes. This is the same as assuming that the range of groups
between men and women is the same.
'9We would like to thank an anonymous reader for pointing this out.
RESTAURANT TIPPING
2625
Crusco, A. H., & Wetzel, C. G. (1984). The Midas touch: The effects of interpersonal touch on restaurant tipping. Personality and Social Psychology
Bulletin, 10, 512-517.
Davidson, R., & MacKinnon, J. G. (1993). Estimation and inference in econometrics. New York, N Y Oxford University Press.
Freeman, S., Walker, M. R., Borden, R., & Latane, B. (1975). Diffusion of
responsibility and restaurant tipping: cheaper by the bunch. Personalify and
Social Psychology Bulletin, 1, 584-587.
Garrity, K., & Degelman D. (1990). Effect of server introduction on restaurant
tipping. Journal of Applied Social Psychology, 20, 168- 172.
Holloway, J. C. (1985). Between gratitude and gratuity: Commentary on Shamir.
Annals of Tourism Research, 12,239-242.
Hornik, J. (1992). Tactile stimulation and consumer response. Journal of Consumer Research, 19,449-458.
Lynn, M., & Grassman, A. (1990). Restaurant tipping: An examination of three
“rational explanations.” Journal of Economic Psychology, 11, 169-181.
Lynn, M., & Mynier, K. (1993). Effect of server posture on restaurant tipping.
Journal of Applied Social Psychology, 23,678-685.
Schein, J . E., Jablonski, E. F., & Wohlfahrt, B. R. (1984). The art of tipping:
Customs and controversies. Wausau, WI: Tippers International.
Shamir, B. (1984). Between gratitude and gratuity: An analysis of tipping.
Annals of Tourism Research, 11, 59-78.
Snyder, M. L. (1976). The inverse relationship between restaurant party size and
tip percentage: Diffusion or equity? Personality and Social Psychology Bulletin, 2, 308.
2626 BOYES ET AL.
Appendix A
Tipping Survey
General Information
How many people in your party?
Your age
Your sex: Male
Female
Are the people in your party:
friends
business associates
family
Occupation
Average Annual Income:
Less than $20,000
$20,000 to $30,000
$30,000 to $50,000
Greater than $50,000
In what city do you reside?
Are you visiting the Phoenix metropolitan area? Yes
If yes, on vacation __ or on business -.
~
~
No-
Frequency
How many times a month do you eat out?
How many times have you frequented this restaurant in the past month?
~
Anticipated Tips
What percentage of the bill do you typically leave as a tip? __
On a scale of 1 to 5 where 5 is very important and 1 is not important, rate the
importance of the following factors in their influence on the percent of total bill
you leave as a tip?
Service Quality
Speed of Service
Quality of Food
Size of Bill
Number of People in Party
Time of Day
Friendliness of Server
RESTAURANT TIPPING
2627
Tonight 5 Considerations
Please rate on a scale of 1 to 5, where 5 is excellent and 1 is very poor:
the quality of your food tonight
the quality of service you received tonight
What was your bill (including tax) today prior to tip?
How much of a tip did you leave today?
How would you classify your mood today? Good bad __
Did you or anyone in your party consume any alcohol prior to or during your
meal?yes
no-
2628 BOYES ET AL.
Appendix B
Dictionary of Variables
Personal attribute variables
AGE = age of respondent (continuous)
GENDER = gender of respondent (1 = male)
OCC = occupation of respondent (0 = professional)
INC 1, INC2, INC3, INC4 = income class of respondent (categorical)
PHX = respondent is Phoenix resident (0 = Yes)
VAC = vacationing in area (0 = Yes)
General attitude variables
FREQO = number of times per month dine at this restaurant (continuous)
FREQH = frequency of dining out per month (continuous)
GENSERVICE = general importance of service in tipping (categorical)
GENSPEED = general importance of speed of service in tipping
(categorical)
GENFOOD = general importance of food in tipping (categorical)
GENBILL = general importance of bill in tipping (categorical)
GENPARTY = general importance of party size in tipping behavior
(categorical)
GENFRIEND = general importance of server friendliness in tipping
(categorical)
GENATT = general restaurant atmosphere in tipping behavior (categorical)
TIPAT = anticipated tip (continuous)
Tonight’s attributes, free riding, and social acceptance variables
PARTY = number in tonight’s party (continuous)
GENPART = PARTY and GENDER interaction
QFOOD = quality of food tonight (categorical)
SERN = quality of service tonight (categorical)
BILLN = size of tonight’s bill (continuous)
MOOD = mood tonight (1 = good)
ALCOHOL = consumed alcohol tonight (1 = Yes)
TIPN = percentage tip tonight (continuous)
CREDIT = paid by credit card (1 = cash)