What makes you click?—Mate preferences in online dating

We estimate mate preferences using a novel data set from an online dating service. The data set contains detailed information on user attributes and the decision to contact a potential mate after viewing his or her profile. This decision provides the basis for our preference estimation approach. A potential problem arises if the site users strategically shade their true preferences. We provide a simple test and a bias correction method for strategic behavior. The main findings are (i) There is no evidence for strategic behavior. (ii) Men and women have a strong preference for similarity along many (but not all) attributes. (iii) In particular, the site users display strong same-race preferences. Race preferences do not differ across users with different age, income, or education levels in the case of women, and differ only slightly in the case of men. For men, but not for women, the revealed same-race preferences correspond to the same-race preference stated in the users’ profile. (iv) There are gender differences in mate preferences; in particular, women have a stronger preference than men for income over physical attributes.

1. To be precise, we do not observe the site users’ opportunities and mate choices outside the online dating environment, but control for the effect of these opportunities on mate choices using person-specific fixed effects.

2. Lee (2009) and Banerjee et al. (2009) follow Hitsch et al. (2010) by estimating a discrete choice preference model and simulating equilibrium match outcomes using the Gale–Shapley algorithm. They use, respectively, data from a South Korean matchmaking agency and a Bengali newspaper’s matrimonial ads section. The data used by Lee (2009) allow her to follow the users of a matchmaking service through several stages of the dating process until an eventual marriage, and she adds a learning component to the choice model.

3. Neither the names nor any contact information of the users were provided to us in order to protect the privacy of the users.

4. Our sample includes only heterosexual users.

5. In Hitsch et al. (2010) we employed a smaller sample of users because we randomly discarded some observations due to computer memory constraints.

6. Biddle and Hamermesh (1998) report a Cronbach alpha of 0.75.

7. Reservation utilities of agents are also interdependent in that exogenous shocks to a subset of agents’ reservation utilities will cause other agents to re-optimize their threshold-crossing rules. Adachi (2003) shows a strategic substitutability property in that more selective behavior (higher reservation utilities) by women (men) leads, in equilibrium, to less selective behavior (lower reservation utilities) by men (women).

8. Although especially Adachi (2003) pushes the realism of these models significantly forward by allowing agents to possess very general preferences.

9. We estimated the model in MATLAB using the KNITRO nonlinear optimization solver. Instead of concentrating out the fixed effects, we estimated all fixed effects directly along with the preference parameters. Using an analytic gradient and Hessian, convergence always occurred in less than 10 steps and in less than 120 s.

10. The main cost associated with sending an e-mail is the cost of composing it. However, the marginal cost of producing yet another witty e-mail is likely to be small since one can easily personalize a polished form letter, or simply use a “copy and paste” approach. The fear of rejection should be mitigated by the anonymity provided by the dating site. Furthermore, rejection is common in online dating: in our data, 71% of men’s and 56% of women’s first-contact e-mails do not receive a reply.

11. To be precise: The probability that m receives a reply from w is determined by the utility function U _W (x _w ,x _m ), i.e. the preference of a woman with attributes x _w for a man with attributes x _m .

12. We resample over individuals rather than individual choice instances to preserve within-person dependence structure.

13. However, it is not clear whether a longer “time on market” in the context of dating should be considered a good or a bad signal of quality. A costly signalling story may suggest that “good” types can separate themselves from “bad” types by holding out longer. The opposite interpretation of time on market is possible if bad types reveal their unobserved quality during a date, are then rejected and hence stay longer in the market.

14. The results reported in Table II are based on the predicted reply probabilities including the “days since registration” variable. The estimates based on the predicted reply probabilities without an excluded (from the first-contact decision) variable are similar.

15. The effect is slightly positive and statistically significant for men in the 30–39 and 40–49 age groups, and statistically insignificant otherwise.

16. Most recent speed dating papers do not report age preferences, due to the small amount of variation in age among the students that comprise many of the analyzed samples. The exception is Kurzban and Weeden (2005), who consider only preferences over the age level, but not the age relative to a potential partner. Our results, however, show that the preference for a partner’s age is strongly contingent on own age.

17. If weight is measured in pounds and height is measured in inches, the BMI is calculated as \(\mbox{BMI}=(\mbox{weight}\cdot703)/\mbox{height}^{2}.\)

18. More precisely, we estimate preferences over BMI differences that are at least 2 in absolute value.

19. Note that according to this scheme, users with a two-year degree or similar education are subsumed in the “some college” group.

20. Using data from speed dating events, Eastwick et al. (2009) find that among whites, the relative preference for a white partner over a black partner is stronger for conservative than for liberal speed dating participants, while relatively conservative blacks have a stronger relative preference for a white than for a black partner compared to liberal blacks. Their study, however, does not report gender differences.

We thank Babur De los Santos, Chris Olivola, Tim Miller, and David Wood for their excellent research assistance. We are grateful to Elizabeth Bruch, Jean-Pierre Dubé, Eli Finkel, Emir Kamenica, Derek Neal, Peter Rossi, Betsey Stevenson, and Utku Ünver for comments and suggestions. Seminar participants at the 2006 AEA meetings, Boston College, the Caltech 2008 Matching Conference, the Choice Symposium in Estes Park, the Conference on Marriage and Matching at New York University 2006, the ELSE Laboratory Experiments and the Field (LEaF) Conference, Northwestern University, the 2007 SESP Preconference in Chicago, SITE 2007, the University of Pennsylvania, the 2004 QME Conference, UC Berkeley, UCLA, the University of Chicago, UCL, the University of Naples Federico II, the University of Toronto, Stanford GSB, and Yale University provided valuable comments. This research was supported by the Kilts Center of Marketing (Hitsch), a John M. Olin Junior Faculty Fellowship, and the National Science Foundation, SES-0449625 (Hortaçsu).

Authors and Affiliations

1. Booth School of Business, University of Chicago, 5807 South Woodlawn Avenue, Chicago, IL, 60637, USA

   Günter J. Hitsch

2. Department of Economics, University of Chicago, 1126 East 59th Street, Chicago, IL, 60637, USA

   Ali Hortaçsu

3. Fuqua School of Business, Duke University, 1 Towerview Drive, Box 90120, Durham, NC, 27708-0120, USA

   Dan Ariely

Corresponding author

Correspondence to Günter J. Hitsch.

Note that previous versions of this paper (“What Makes You Click?—Mate Preferences and Matching Outcomes in Online Dating”) were circulated between 2004 and 2006. Any previously reported results not contained in this paper or in the companion piece Hitsch et al. (2010) did not prove to be robust and were dropped from the final paper version.

Reprints and permissions