Effect on Restaurant Tipping of a Helpful Message Written on the Back of Customers' Checks BRUCERIND' DAVIDSTROHMETZ Uonmouth University Temple Universiw Research has shown that servers can increase their tip percentages by writing \"Thank you\" or by drawing a happy face on the backs of customers' checks. In the current study, a third approach of this type was tested. An experiment was conducted in which a female server either did or did not write a helpful message about an upcoming dinner special on the backs of checks before delivering them to customers. It was predicted that adding the helpful message would increase tip percentages because of reciprocity, in which customers would tend to respond to the server's \"tip\" with an increased tip of their own. Results were consistent with this prediction: Mean tip percentages increased from about 17% to 20%. More than 1 million people in the United States work as waiters or waitresses (Statistical Abstracts, 1990). Although these servers are generally paid by their employers, their major source of income usually comes in the form of tips from customers (Lynn & Mynier, 1993; Schmidt, 1985). Because tips are so important to the livelihood of most servers, knowledge about factors that affect customers' tipping behavior is valuable. Over the last decade, a growing number of studies have identified a variety of such factors. Factors affecting tipping can generally be grouped into three categories (Rind & Bordia, 1996). The first concerns characteristics of the dining party, including party size, method of payment, alcohol consumption, and mood (Cunningham, 1979; Freeman, Walker, Borden, & Latant, 1975; Lynn, 1988; Lynn & Latank, 1984). The second concerns characteristics of the server, including attractiveness, dress, and gender (Lynn & Latant, 1984; May, 1978; Stillman & Hensley, 1980). The third concerns server-diner interactions, such as having servers briefly touch their customers (Crusco & Wetzel, 1984; Hornik, 1993; Stephen & Zweigenhafi, 1986), make additional nontask visits (May, 1978), squat during their initial interaction with customers (Lynn & Mynier, 1993), personalize their interaction by giving customers their first names during the initial contact (Garrity & Degelman, 1990), and display a 'Correspondence concerning this article should be addressed to Bruce Rind, Department of Psychology, Temple University, Philadelphia, PA 19122. e-mail: rind@vm.temple.edu. 139 Journal of Applied Social Psychology, 1999, 29, 1 pp. 139-144. Copyright 0 1999 by V. H.Winston 8 Son, Inc. All rights reserved. I 140 RIND AND STROHMETZ maximal smile during initial interaction with customers (Tidd & Lockard, 1978). More recently, two additional studies involving server-diner interactions were conducted, which focused on manipulating what appeared on the checks delivered to customers at the end of their meals. In one study, a female server received higher tip percentages when she wrote \"Thank you7'on the back of the check than when she did not (Rind & Bordia, 1995). In the other study, a female and male server either did or did not draw a happy face on the back of the customers' checks, with the result that tip percentages increased for the female but not for the male server (Rind & Bordia, 1996). The current investigation similarly involves manipulating what appears on the back of customers' checks. A server either did or did not write a message on the check informing the customer of a dinner special sometime in the future. The message consisted of four sentences, representing a nontrivial effort by the server. The content of the message was designed to be helpful to the customers-that is, to alert them to a good deal. As such, it was expected that customers would perceive the message as a helpful \"tip\" for them that cost the server some time in writing up, which would have the effect of increasing their willingness to return the favor by leaving a larger tip. This expectation follows fiom research on reciprocity, in which individuals feel obligated to return a favor to a person doing them a favor (Regan, 1971). Method Subjects Eighty-one dining parties eating dinner at an upscale restaurant in northem New Jersey served as subjects. The restaurant was located in a private country club, and its menu was buffet style; servers catered to all needs of customers, except for serving them the food. The dining parties consisted of a total of 3 15 customers, with a mean of 3.89 customers per party (SD= 2.72); the range of customers in the 81 dining parties was from 2 to 10. Procedure The experiment was conducted over a 3-week period during March and April 1997. A female server in her 20s. who worked the dinner shift, acted as the experimental accomplice. She had had 2 years' experience waiting on tables, all at the current restaurant, prior to this study. She was given a stack of 3 x 5 index cards. On half of these cards was written \"message,77while on the other half was written \"no message.\" The cards were thoroughly shuffled, such that the order of the two types of cards was random. At the end of a TIPPING BEHAVIOR 141 dining party's meal, when it came time for her to present the dining party with a check, the server reached into her pocket and randomly selected a card. This procedure ensured random assignment of dining parties to the message and no-message conditions. If the server selected a card with \"message\" written on it, then she wrote on the back of the check the message \"We have a special dinner on (date specified). The menu will feature delicious seafood. Why not give it a try? It's great!\" The server wrote down the appropriate date for the next time a special seafood buffet was being served, usually in about 1 week's time. If the server selected a card with \"no message\" written on it, then she wrote nothing on the back of the check. To avoid potential confounding, the server was instructed to behave in the same way, regardless of condition, when delivering the check. The server delivered the check with a neutral facial expression, while saying \"Here's your check,\" and then immediately left to avoid further contact with the dining party. After the dining party left, the server recorded on the same 3 x 5 index card just used to determine the dining party's condition the amount of tip left by the party, the amount of the bill before taxes, and the number of customers in the dining party.2 The dependent measure was the tip percentage. Results Of the 8 1 dining parties served, 40 were in the message condition, while 41 were in the no-message condition. The mean bill amount per dining party was $5 1.52 (SD= $47.64), which worked out to be a mean of $13.25 per per- son. Tip percentages were calculated for each dining party by dividing the tip size by the bill amount and then multiplying this quotient by 100. It was hypothesized that writing a helpful message would increase tip percentages, compared to writing nothing. To assess this hypothesis, the mean tip percentages in these two conditions were contrasted. Consistent with expectations, the mean tip percentage was statistically significantly greater in the message condition (M= 19.91%, SD = 7.77%) than in the no-message condition (M= 16.95%, SD = 5.91%), r(79) = 1.93, p = .028, one-tailed, effect size r = .21. We also examined the relationship between party size and tip percentage. Consistent with previous research findings (Freeman et al., 1975; Lynn & LatanC, 1984; May, 1978; Pearl & Vidman, 1988; Rind & Bordia, 1996), 2The server also attempted to record information on gender of the bill payer for each dining party by observing who paid or by checking names on credit-card slips. However, because she was unable to determine gender for a sizable minority of bill payers through these methods, these data were excluded from analysis. In retrospect,recording gender composition of the dining parties would have been more feasible. 142 RIND AND STROHMETZ dining party size was negatively correlated with tip percentage, 479) = -.23, p < .05, two-tailed. Lynn and Bond (1 992) have shown that correlations between dining party size and tip percentages can be biased because of the use of tip percentages as the dependent measure. They presented a method for correcting the tip percentage: corrected percentage tip = (T - A)/B, where T = the tip, B = the bill, and A = the y intercept when predicting T from B. The correlation between corrected tip percentages and dining party size was nonsignificant, r(79) = .17, p > . l o , two-tailed. To examine the relationship between tipping and dining party size further, we calculated the semipartial correlation between these two variables, adjusting tip size by holding bill amount constant. This correlation was also nonsignificant, r(78) = .16, p > .lo, two-tailed. Discussion Results of the current experiment support the prediction that adding a helpful message to a check would increase tip percentages. This finding adds to those of previous studies that have shown that adding information to the back of customers' checks can increase tip percentages. Rind and Bordia (1995, 1996) found that writing \"Thank you\" and drawing a happy face on the backs of checks increased tips; they attributed these increases to the beneficial effects of creating a perception among the customers of the servers' friendliness. In the current experiment, adding a message may also have engendered perceptions of friendliness. More likely, however, adding the message created the perception on the part of customers that the server was doing something extra for them-that is, giving them a useful tip about a future dinner, and incurring the cost of spending the time to write out a relatively lengthy message to do so. This perception, it follows, then heightened the likelihood that customers would reciprocate the tip and extra effort by the server with an increased tip of their own (cf. Regan, 1971). As plausible as this explanation may be, however, the resolution concerning the precise mechanism for the increased tip percentage in the message condition requires further research. In the current experiment, the female server's customers spent a total of $4,173 over a 3-week period, which is equivalent to $1,39 1 per week. By the addition of a helpful message concerning a future dinner special on all checks, the server would have increased her tips from $236 to $277 per work week, which represents an increase of $4 1, or 17.4%. For the more than 1 million servers in the United States, systematic use of this technique could mean millions of dollars of extra income annually. The current study also examined the relation between dining party size and tip percentage, which has been found to be an inverse one in many TIPPING BEHAVIOR 143 previous studies (cf, Lynn & Bond, 1992). Initially, a statistically significant inverse relation was uncovered. After correcting tip percentages based on Lynn and Bond's (1992) procedure, however, this inverse relationship disappeared, as it has in other reanalyses (e.g., Lynn & LatanC, 1984; Rind & Bordia, 1996). The current result, along with those of previous analyses, adds W h e r to the weakening of this relationship. The current experiment was conducted in an upscale restaurant in a country club in the northeastern United States using only one female server. The generalizability of the current findings thus needs to be examined in future research by varying location and restaurant-type factors, and by using different servers, male as well as female. References Crusco, A. H., & Wetzel, C. G. (1984). The midas touch: The effects of interpersonal touch on restaurant tipping. Personality and Social Psychology Bulletin, 10, 5 12-517. Cunningham, M. R. (1979). Weather, mood, and helping behavior: Quasiexperiments with the sunshine Samaritan. Journal of Personality and Social Psychology, 37, 1947-1956. Freeman, S., Walker, M., Borden, R., & LatanC, B. (1975). Diffusion of responsibility and restaurant tipping: Cheaper by the bunch. Personality 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. Homik, J. (1993). Tactile stimulation and consumer response. Journal of Consumer Research, 19,449-458. Lynn, M. (1 988). The effects of alcohol consumption on restaurant tipping. Personality and Social Psychology Bulletin, 14, 87-9 1. Lynn, M., & Bond, C. F. (1992). Conceptual meaning and spuriousness in ratio correlations: The case of restaurant tipping. Journal of Applied Social Psychology, 22, 327-341. Lynn, M., & Latane, B. (1984). The psychology of restaurant tipping. Journal of Applied Social Psychology, 14,549-56 1. Lynn, M., & Mynier, K. (1993). Effect of server posture on restaurant tipping. Journal of Applied Social Psychology, 23,678-685. May, J . M. (1 978). Tip or treat: A study of factors affecting tipping behavior. Unpublished master's thesis, Loyola University of Chicago. Pearl, R. B., & Vidman, J. (1988). Tippingpractices ofAmerican households in restaurants and other eating places: 1985-1986. Champaign, IL: Summary report to the IRS under contract TIR 86-279 with the Survey Research Laboratory, University of Illinois. 144 RIND AND STROHMETZ Regan, D. T. (1971). Effects of a favor and liking on compliance. Journal of Experimental Social Psychology, 7,627-639. Rind, B., & Bordia, P. (1995). Effect of server's \"Thank you\" and personalization on restaurant tipping. Journal of Applied Social Psychology, 25, 745-75 1. Rind, B., & Bordia, P. (1996). Effect on restaurant tipping of male and female servers drawing a happy, smiling face on the backs of customers' checks. Journal of Applied Social Psychology, 26,2 18-225. Schmidt, D. G. (1985). Tips: The mainstay of many hotel workers' pay. Monthly Labor Review, 108,50-6 1. Statistical Abstracts of the United States. (1 990). Washingtin, DC: Department of Commerce. Stephen, R., & Zweigenhaft, R. L. (1986). The effect on tipping of a waitress touching male and female customers. Journal of Social Psychology, 126, 141-142. Stillman, J. W., & Hensley, W. E. (1980). She wore a flower in her hair: The effect of ornamentation on nonverbal communication. Journal of Applied Communication Research, 1,3 1-39. Tidd, K., & Lockard, J. (1 978). Monetary significance of the affiliative smile: A case for reciprocal altruism. Bulletin of the Psychonomic Sociery, 11, 344-346. Research Methods Research Validity Learning Objectives Explain the meaning of research validity. Explain the meaning of statistical conclusion validity and its importance in research. Explain the meaning of internal validity and its importance in making causal inferences. Explain how to eliminate the threats to internal validity. Explain the meaning of external validity and describe the conditions that threaten external validity Describe the relationship between internal and external validity. Overview Research Validity Statistical Conclusion Validity Threats to Statistical Validity - Type I error - Type II error Internal Validity Threats to Internal Validity - Confounding extraneous variable External Validity - Population Validity - Ecological Validity - Temporal Validity Relationship between Internal and External Validity Research Validity Research Validity: Truthfulness of inferences made from a research study. To conduct a valid research study, you must develop a plan to use and follow it closely. This plan must involve the use of strategies for obtaining valid results. We'll focus on three major types of quantitative research validity: statistical conclusion validity internal validity external validity Required Conditions for a \"Cause and Effect\" Relationship Understanding these conditions is important in assessing research validity. Criteria for identifying a causal relation - cause (IV) must be related to the effect (DV) (relationship condition) - changes in IV must precede changes in DV (temporal order condition) - no other plausible explanation must exist for the effect Example A study shows that there is a correlation between coffee drinking and likelihood of having a heart attack. Does drinking coffee cause heart attack? Does it meet all the three requirements? One problem with concluding that coffee drinking causes heart attacks is that cigarette smoking is related to both of these variables. The researcher would have to \"control for\" smoking in order to determine if this alternative explanation accounts for the original relationship. Statistical Conclusion Validity The validity with which we can infer that the independent and dependent variables covary. Covary means with every variation in the IV there is a corresponding variation in the DV. i.e. IV and DV are statistically related. We make this inference relationship between the IV and DV, from the results of the statistical analysis computed on the data collected in the research study. Statistical Significance Not only do we want to see if the data show a relationship, but we must also determine if the observed relationship is statistically significant. A relationship is statistically significant when the analysis indicates that the observed relationship is probably not due to chance. Threats to Statistical Validity Sometimes the inferences researchers make from their statistical analysis to the populations of interest are wrong. Type I error: When researchers conclude that there is a relationship between IV and DV, when there actually isn't. Type II error: When researchers conclude that no relationship between the variables exists when it really does. The lack of a sufficient number of data points in one's sample is one of the important factors that can threaten statistical conclusion validity Internal Validity Internal validity is concerned specifically and only with the issue of causation. Internal validity is the degree to which you can correctly conclude that the relationship between an independent variable and a dependent variable is causal. Internal validity boils down to ensuring that the observed effect, is caused only by the variation in your independent variable. Threats to Internal Validity Primary threat - confounding extraneous variables Extraneous variable - a variable that competes with the IV in explaining the DV Confounding extraneous variables - an extraneous variable that systematically varies with the independent variable and this variable also affects the dependent variable. Example - IV = tutoring, DV = grades - You don't use random assignment, but use two intact classrooms to serve as experimental and control groups - The experimental group (who receives extra tutoring) shows significant improvement in grades. - Can you conclude that the difference is due to the tutoring? - What if the class serving as your experimental group was an honors class? - An alternative explanation could be due to the presence of a confounding extraneous variable such as prior achievements or intelligence. Example - cont. If the students who received and did not receive tutoring were of the same intelligence level, any difference in grades could not be attributed to intelligence. Intelligence level, in this case, would represent an extraneous variable, but it would not represent a confounding extraneous variable It did not vary systematically with the independent variable. Controlling for Extraneous Variables Controlling for the effect of extraneous variables does not mean totally eliminating their influence because eliminating the influence of many extraneous variable is not possible. Eliminate the confounding influence of extraneous variables by: - holding their influence constant; equating the groups - arranging factors so that extraneous variables are equated across the groups and do not differentially influence the results through control - using random assignment to balance their influence External Validity Generalization is a major goal of scientific research. Degree to which the study results can be generalized to and across other people, settings, treatments, outcomes, and times. External validity is an inferential process because it involves making broad statements based only on limited information. Stating that a particular study conducted on 100 college students is fully externally valid would imply that the results are true for all college students responding in a variety of settings to variations in the treatment and and at different times. Factors Impacting Generalization Lack of random selection - In order to generalize, your sample must be representative of a target population of people, settings, treatment variations, outcome measures, and times. - For a variety of reasons (e.g., cost, time, accessibility), most experimental research studies are not based on a random sample from the defined population. Chance variation - Findings usually vary slightly from study to study because of the operation of chance. - Occasionally this variation will be large simply because of the operation of chance. - Replication can help reduce this factor Types of External Validity: Population Validity Population validity: degree to which the study results can be generalized to and across the people in the target population. Target population: The large population to which the researcher would like to generalize the study results. Accessible population: The population of research participants that is practically available to the investigator Inferential Steps First, the researcher has to generalize from the sample to the accessible population from which the sample was drawn. This step is easily accomplished if the researcher randomly selects the sample from the accessible population. The second step requires moving from the accessible population to the target population. This ultimate generalization seldom can be made with confidence because often the accessible population will not be representative of the target population