Seeing red: color effects on restaurant tipping may not be as significant as thought

ABSTRACT

Previous research found that female restaurant servers displaying the color red, either on their lips or shirts, tended to receive higher tips from their male patrons, in theory because the color red is thought to enhance perceptions of sexual attractiveness. This study reports findings from two survey experiments that were administered to a sample of MTurk workers to test the causal process linking red lipstick and, alternatively, a red face mask with customers’ tips. We are not able to replicate the previously observed red lipstick effect on customers’ tipping practices, although some evidence for the underlying attraction-based theory is offered. The findings related to red masks suggest that red masks are not equivalent to red lips and more closely align with alternate theories suggesting that clothing color may signal competence or professionalism rather than attractiveness.

KEYWORDS:

• Tipping
• mask
• COVID-19
• restaurant
• customer service
• tipping motives

Acknowledgements

The authors wish to thank anonymous reviewers for JFBR and David Merolla for helpful feedback on earlier drafts of this manuscript.

Declaration of conflicting interests

The authors have no relevant financial or non-financial interests to disclose.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. This study was performed with approval from the Wayne State University Institutional Review Board (IRB-20-12-3056). In lieu of written documentation of consent, to better preserve anonymity, participants were informed of the procedures, intent of the study, and potential risks at the very beginning of the survey with an information sheet and will be asked to indicate their agreement to participate in the study by clicking “I agree to participate”.

2. Workers were requested from the MTurk platform who reside in the U.S. Worker screening on the MTurk platform includes validating worker identity for tax purposes. Respondents were further asked at the end of the survey to report the state wherein they currently reside, and “I do not live in the United States” was an option in the dropdown menu. None reported that they do not live in the United States.

3. Of 1642 participants recruited, 1524 completed the survey. Of these, 96 were excluded because of missing data on the analyzed variables (including 7 who did not report their sex as either male or female). An additional 113 were excluded on the basis of outlying tip amounts. While the decision to exclude so many outlying cases was difficult, we chose the cutoff of 55% based on two factors. First, higher tips (both in dollar value and in percent) appeared increasingly random and less principled. For example, a large number of cases had tips reported around 100% of the bill amount, which suggests that respondents may not have understood the question and were reporting the bill amount rather than a tip. We also saw many very high tips of exactly $10 or $20, suggesting that those respondents may not have given much thought to the size or rationale for their tip. Second, had we chosen a cutoff value much higher than we did, values of skewness and kurtosis on the distribution of tip percent would have been too large to assume a normal distribution and would have made any statistical assessments increasingly difficult. We assessed the even distribution of omitted cases across our 15 experimental conditions (Pearson’s chi-squared = 20.825, p = .106), and as a robustness test we estimated our models in a sample that included the cases that were omitted from the main analysis as a result of questionable tip amounts. The results from these models did not alter the substantive conclusions that we draw from the main analyses.

4. https://stock.adobe.com/license-terms

5. Guéguen and Jacob (2012) included only female servers and male patrons in their sample. Model 4 is designed to identify an effect specific to this subset of our sample. As an additional test, we also estimated a direct model in our sample after isolating cases with male patrons and female servers; again, there is no discernible association.

6. While the finding that our male server was, on average, perceived to be less attractive with red lip coloration was not specifically predicted in earlier literature or in our hypothesis, it is not inconsistent. With only one male model, this study is unable to make generalizable claims about male lip coloration; however, future studies could explore this phenomenon to see whether this finding holds for men in general.

7. For the male server, the negative indirect effect of red lip coloration on tips through attraction is significant at a 94% confidence level. For female servers, the positive indirect effect is significant with 90% confidence. Further, because these effects operate in opposing directions, server sex clearly moderates this relationship (index of moderated mediation = .9240, BootLLCI .0606, BootULCI 2.0233).

8. Because this theory is based on sexual attraction, we also considered if the effect would be stronger when server and patron were of the same sex and the patron identified as homosexual or bisexual. No significant difference was visible, though the sample did not include a large group of homosexual-identifying participants so a small effect might not have been statistically apparent (90.6% identified their sexual orientation as heterosexual). Additionally, the theory is based on biological impressions of attractiveness that may have more to do with the sexual (gender) identity of the subject than the sexual orientation of the observer. Since female heterosexual patrons do not view red lips of a male in the same way that male heterosexual patrons view red lips of a female, there is likewise no expectation that homosexual patrons would find these characteristics to be sexually attractive in a way that mirrors that of heterosexual patrons.

9. When compared to either the black or red conditions, the difference in mean tip percent for the white mask conditions still did not rise to statistical significance (p = .098 compared to black and p = .073 compared to red). However, when mean tip percent for white mask conditions is compared to all other conditions, the difference is statistically significant (mean tip 1.49% less, p = .012).

10. Lynn et al. (2016) also found that female patrons were more discriminating of server shirt color, but in different ways than our findings regarding mask color. Female patrons reported the female server to be more attractive when wearing a red shirt compared to black, and the male server less attractive when wearing red compared to black or white.

11. While this experiment was independent of the experiment conducted in Brewster and Gourlay (2021), the two studies shared the same sample.