Very preliminary and incomplete. Please do not cite or circulate without the authors’ permission.

What Makes You Click: An Empirical Analysis of Online Dating* Dan Ariely MIT Sloan School of Management

Günter J. Hitsch University of Chicago Graduate School of Business

Ali Hortaçsu University of Chicago Department of Economics

June 3, 2004

Abstract We utilize a unique data set obtained from a major online dating site to draw econometric inferences on revealed mate preferences of the site users. We first compare the reported demographic characteristics of site users to the characteristics of the population-at-large, and do not find large differences. We then investigate several behavioral models that allow us to interpret the observed browsing and e-mailing behavior of site users in a revealed preference framework. Our empirical results indicate, in particular, that while women’s choices depend strongly on the income and education of men, the choices of men appear to be driven much more by physical appearance at this initial “dating” stage of the marriage market. We also find evidence in observed “matches”(defined as the exchange of personal contact information between two site users) that corroborates this finding.

* We thank Babur de los Santos, Christopher Olivola, and Timothy Miller for outstanding research assistance. Hitsch gratefully acknowledges financial support from the James M. Kilts Center for Marketing at the University of Chicago Graduate School of Business. Hortaçsu acknowledges the generous financial support of a John M. Olin Junior Faculty Research Fellowship.

1. Introduction Understanding the workings of matching markets has been on the research agenda of the economics profession at least since the seminal works of Gale and Shapley (1962) and Becker (1973, 1974). Following these pioneering papers, economic models of marriage markets (or matching markets in general) have been built upon the specification of the following primitives: 1) the structure of preferences over match partners, 2) the matching protocol, i.e. the mechanism through which preferences are “declared” (either directly, or through the choice of an action from a set defined by the mechanism) and matches are made, 3) the information structure of the game, and 4) the strategic sophistication of the agents. Once these primitives are specified, the goal is to derive predictions regarding the structure of match outcomes, and, in particular, assessing the economic efficiency of realized matches. In this paper we analyze individual behavior and match outcomes among the users of a major online dating site. Members of this site post profiles that contain information regarding their socioeconomic and physical characteristics, along with some of their personal interests. The website allows users to narrow down the set of potential matches using database queries. Once a choice set is arrived at, the searcher can browse through individual profiles, and can respond by sending an e-mail to a person they are interested in. Thus, this is a matching market on which the matching protocol is well-defined: Searchers “declare” their preferences through their e-mails, and a “match” is arrived at if the recipient of the e-mail responds. The information structure of the game is also well-defined: all searchers have access to the same profile information about another user, at least before any e-mail contact is established. Hence, the online dating site can be thought of as a “field laboratory” in which the matching protocol and information structure are controlled for. In section 2 of the paper, we also argue that our analysis of this particular market may yield insights beyond what an isolated laboratory experiment can provide. Indeed, popular press coverage suggests that online dating is rapidly becoming an activity in which many Americans take part in.1 The dating site analyzed here is not According to a recent estimate based on ComScore Networks’ analysis of Internet users’ browsing behavior, 40 million Americans visited online dating sites in 2003, generating $214 million in revenues, making online dating the most important subscription-based business on the Internet. Match.com, which was founded in 1995 as one of the 1

2

an exception to this trend. Our data set contains information about the characteristics and actions of about 30,000 users located in two major metropolitan markets, Boston and San Diego. When we compare the (reported) socioeconomic characteristics of the users of this website to local population characteristics surveyed by the U.S. Census, we do not find stark differences, especially after controlling for Internet use.2 Our analysis then focuses on deriving empirical inferences regarding the structure of mate preferences in this market. To do this, we need to impose certain restrictions on the strategic behavior of agents and also on the allowable forms of preferences. Our main behavioral assumption is that, upon browsing a user’s profile, a searcher decides to send an e-mail if the profile is deemed worthy of a match. If we combine this behavioral assumption with the assumption that profiles can be ranked by a single dimensional “type,” we can use the number of e-mails received (per time unit) by a profile as a measure of this “type.” We can then use regression analysis to investigate how different socioeconomic or physical attributes enter into this single dimensional index of attractiveness. We should note that since people join online dating sites to find people they can exchange emails with, the regression analysis outlined above is of independent interest as a descriptor of “success” in this market. We also believe that the behavioral assumption imposed to interpret the result of these success regressions is not an unreasonable one. The interpretation of the empirical results becomes difficult if people are reluctant to send e-mails to profiles that they find attractive, but unattainable. Such behavior may lead to a very “high type” woman to receive very few e-mails, since men browsing her profile believe that they can never match with her. In conventional dating settings, such “strategic” concerns may be of considerable importance, since the social and psychological costs of rejection, or the effort involved in making the approach can be large enough to outweigh the expected benefit from a match. However, in the online dating

pioneering online dating sites, boasted 939,000 paying subscribers as of the 4th quarter of 2003. Although the sector is led by large and nationally advertised sites like Match.com, Matchmaker.com, and eHarmony.com, along with online dating services bundled by major online service providers (such as Yahoo!Singles), there are also numerous online dating sites that cater to more specialized audiences, such as JDate.com, which bills itself as the “The largest Jewish singles network,” Gay.com, BlackSinglesConnection.com, and ChristianSingles.com. 2 There do appear certain patterns in our sample that are distinct. Men are overrepresented on the dating site, minorities largely underrepresented, and the age profile appears more skewed to the 20-30 year old range.

3

context, the “marginal cost of e-mailing” is likely to be much smaller, especially given the ability to cut-and-paste from previously prepared introductory e-mails. With that caveat, the empirical analysis in Section 4 reveals the following findings: Many of the self-reported user attributes are strongly associated with online “success”, measured either by the number of times a user’s profile is browsed, the number of introductory e-mails received, or whether a user receives contact information (a phone number or an e-mail address) from another user. There are stark differences between the determinants of success of men and women. While self-reported good looks improve male and female outcomes similarly, we find that above a threshold of about 120 lbs female outcomes deteriorate strongly in their weight. Men, in contrast, do not suffer such a weight penalty. On the other hand, tall men, but not tall women, are more successful than shorter members. The most striking difference between the sexes is related to earnings. Self-reported incomes above $50,000 are associated with a strong increase in success for men only; the online success of women is mostly unaffected by the income variable. Men who are college educated or have a graduate degree are approached more often than less educated members, while, as in the case of income, no such effect is apparent for women. In section 4.2, we relax the assumption that users agree on the ranking of potential mates. Specifically, we segment the number of e-mails received by a user based on the demographic characteristics of the writers of that e-mail. This allows us to investigate, for example, whether a woman is more attractive among college educated vs. non college educated men. We find, interestingly, that high income and college educated women have a stronger preference for high income and highly educated men than low income and high school educated women. A similar pattern is apparent when we examine taste differences among men based on the income and education segmentation. In comparison to women, however, preference heterogeneity is less pronounced among men. Since Gary Becker’s seminal papers on the topic, models of marriage markets have also focused on analyzing whether matched couples sort along certain characteristics. Single-dimensional

4

“type” based models predict perfect assortative matching in the absence of search frictions.3 Such assortative sorting implications of marriage markets (along dimensions such as income, education, and ability) have been argued to be among determinants of long term trends in ability and income distributions. Unfortunately, our data set does not allow us to observe the formation of “productive” matches through marriage or long-term partnerships.4 However, utilizing information as to whether users e-mailed each other repeatedly, or whether they exchanged personal contact information (such as phone numbers and e-mail addresses), we can get a sense of whether a correspondence may have led to a physical “date.” In section 5, we indeed find that physical, demographic, and socioeconomic attributes of “dating members” exhibit statistically significant positive correlations that mimic similar findings among married couples surveyed in previous economic and sociological studies. Although it is difficult to conduct systematic comparison the strength of these correlations across our data and other data sets, this suggests that online dating sites do yield a matching environment conducive to sorting among their members. This allows us to conclude that observed sorting patterns among married couples are not solely due to institutional/search frictions encountered in the traditional channels they have met through.5 Our work relates to the economic literature on matching and marriage markets in several ways. A long line of empirical literature in economics, sociology and demography has focused on reporting correlations between married couples’ socioeconomic attributes. However, it is difficult to interpret these correlations in terms of underlying preferences without knowing the choice constraints faced by the matching parties. A long literature in psychology has thus taken the approach of measuring “stated” preferences through a wide variety of surveys in which participants are asked to rate hypothetical (or real) profiles. Another approach is to try to assess Shimer and Smith (2001) and Smith (2002) have established general conditions on the joint production function under transferable utility and non-transferable utility assumptions that lead to “positive assortative matching,” even when search frictions are present. 4 Information as to whether online dating achieves its goals in terms of marriages or long term relationships is sparse, with most of the information provided by the Web sites themselves for advertisement purposes. For example Match.com reports, based on surveys of users canceling their site subscriptions, that 200,000 of their users have found the person they were seeking on the site. eHarmony.com claims to have “more than 3,000 marriages to its credit” (http://www.eharmony.com/core/eharmony?cmd=community-facts) and claims to receive 10 notices of marriages, engagements or long-term relationships from its members every day. 3

5 I.e. an alternative explanation to sorting by educational attainment may be that many men and women meet each other at school – even if they might be indifferent towards each other’s educational attainment. On the dating site, however, one is not constrained by the availability of potential partners hailing from all walks of life.

5

“revealed” preferences by interpreting observed match outcomes through an explicit economic model which generates match outcomes as an equilibrium prediction (see for example Wong (2001) and Choo and Siow (2003)). However, the results of these studies may be criticized due to the strong assumptions they place on the form of the matching protocol, the information available to the agents, and the choice sets they face.6 Our work may thus be viewed as an attempt to measure “revealed” preferences using data from a setting where the matching protocol, information available to the agents, and the choices made by agents are observed. In this regard, the work that comes closest to ours is that of Fisman, Iyengarl and Simonson (2004), who investigate revealed preference determinants of mate selection using an experimental speed dating market. In contrast to their work, we emphasize how preferences and match outcomes are related to socioeconomic characteristics such as income and education. Our large and diverse sample is more ideally suited to analyze this question than theirs, which is mostly composed of graduate students at one U.S. university. The paper proceeds as follows. Section 2 provides a description of the workings of the dating site, and characteristics of site users. Section 3 outlines the modeling framework. Section 4 reports the main empirical results on the determinants of online dating outcomes, i.e. the number of profile browses, received “first contact” e-mails, and the number of e-mails containing contact information received from other users. Section 5 describes sorting characteristics of “matched” couples. Section 6 concludes.

2. Description of the Data and User Characteristics Our data set contains socioeconomic and demographic information and web site activities of close to 30,000 users of a major online dating site. Specifically, 15,034 users were located in the San Diego area (8,585 men) and 14,911 users in the Boston area (8,417 men). Our main source of socioeconomic and demographic characteristics is a mandatory survey filled out by new users 6 Even at the micro level, these empirical studies have very little information regarding the actual search behavior and constraints faced by the agents in their data sets. For example, Wong (2001) utilizes data from the NLSY on how long an individual stays single before marrying, but observes very little regarding the “alternatives” available to this individual aside from certain demographic characteristics of his/her spouse. Choo and Siow (2003) only utilize data from the U.S. Census regarding matched couples characteristics, following the sociology literature.

6

registering on the site.7 We have also obtained the profile photos of the users to construct “physical attractiveness” measures. In addition, we utilize information from detailed “weblogs” of registered users. Upon registering, users of the dating site can browse through user profiles. Typically, a user starts his search by entering a number of parameters (such as location, age range, income/education range) in a database query form. The query returns a set of “short profiles” displaying the alias, age, location, and (if available) a thumbnail version of the potential matches’ profile photos. The searcher then can click on one of these short profiles to “browse” a user’s detailed profile, which contains the socioeconomic/demographic information provided on the survey, a larger version of the profile photo (and additional photos), and answers to several essay questions. Upon reviewing this detailed profile, the searcher can decide whether to send an e-mail to this user. The receiver of the e-mail is notified of this action, and can choose to reply to this “first contact.” Our data set allows us to observe the detailed profiles that each user “browses,” the profiles to which s/he responds by sending an e-mail, and the e-mails to which s/he replies to. We also have additional information regarding whether the e-mail contained a phone number or an e-mail address, based on a text search looking for special characters in the e-mail.8 2.1 Who are the users of the Web site? A. Stated reasons for using the web site The registration survey asks users the reason they are joining the site. 35% of the users state they are looking for long-term relationships, 26% state that they just want to look around, and 10% declare they are looking for short term/casual relationships. Women are more likely to seek a long term relationship (36%) or friendship (12%) than casual relationships (5%). Perhaps not surprisingly, men seem to be more eager for short term/casual relationships (14%) than friendships (8%), though a large percentage are looking for long term relationships (35%). Sensibly, the percentage of users that are looking for a long-term relationship increases quite a bit when we condition on the sample having photos on file. 56% of men and 58% of women 7 8

The names, e-mail and physical addresses of the users were not provided to us to protect the privacy of the users. We do not have access to the actual text of the e-mail, or the e-mail address/phone number searched in the text.

7

indicate they are looking for a long-term relationship. Of these, only 6.1% of men indicate looking for a short term/casual relationship (compared to 14% of the whole sample). The registration also asks users about their sexual preferences. 93% of users on both markets declared that they were heterosexual. 9% of women declared they were bisexual or homosexual, whereas 5% of men in the respective markets did the same. Interestingly, close to 11% of users declare that they have had at least one homosexual experience, or could be persuaded to have a homosexual experience. In contrast, 10% of men declared that homosexuality offends them, whereas only 5% women said the same. B. Demographic/socioeconomic characteristics We now investigate the reported characteristics of the site users, and contrast some of these characteristics to representative samplings of these geographic areas as obtained on the CPS Community Survey Profile. In particular, we contrast the site users with two subsamples of the CPS. The first subsample is a representative sample of the Boston and San Diego MSAs (Metropolitan Statistical Area), and reflects information current to 2003. The second CPS subsample conditions on being an Internet user, as reported in the CPS Computer and Internet Use Supplement, which was administered in 2001. A visible difference between the dating site from the population-at-large is the overrepresentation of men on the site. In San Diego, 57% of users, and in Boston, 56.4% of users were men.9 Another visible difference is in the age profiles: site users are more concentrated in the 26-35 year range than both CPS samples (the median user on the site is in the 26-35 age range, whereas the median person in both CPS samples is in the 36-45 age range). People above 56 years are underrepresented on the site compared to the general CPS sample; however, when we condition on Internet use, this difference in old aged users attenuates somewhat.

9 When we restricted attention to members who have posted photos online (23.6% of registered users in Boson and 26.8% of users in San Diego), the percentage difference between male and female participation comes down slightly: in Boston 51.9% of users with posted photos are men, whereas in San Diego, 54.2%.

8

The ethnic profile of site users appears to roughly reflect the profile of the geographic regions covered by the site, especially when conditioning on Internet use, although Asians appear to be underrepresented on the San Diego site. 10 The reported marital status of site users clearly represents the fact that most of users are looking for a partner. More than 65% of site users have never been married, about 4% of the users are married but separated. The percentage of divorced and widowed women among site users is higher than male users in the same status. A rather surprising finding is the number of site users, especially men, who declare that they are “currently married, and not officially separated.” When we decompose this category further, we find in Boston 3.39% of men in Boston (out of 8% total) declare themselves to be “unhappily married,” 0.99% declare to be in an “open marriage,” and 0.59% of them claim to be a part of a “swinging couple.” However, this still leaves 3.12% of men who declare to be “happily married.” This suggests that the dating site may also be used as a search outlet by people in long term relationships. Of course, one may expect the true percentage of otherwise committed people on this site to be much higher than what is reported here. 11 The education profiles of the site users show that site users are in general more educated than the general CPS population. Above 50% of the site users have college degress or above. However, the educational profiles appear quite close to that of the Internet using population, with only a slightly higher percentage of professional degree holders.

We should note that the “Other” category in the site’s ethnic classification includes several ethnicities grouped under “White” by the CPS. Once we reconcile these classifications, the differences in the “White” category disappear. 11 Interestingly, only 9 out of the 258 “happily married” men actually posted a picture, and only 2 of these men (with photos) declared they were looking for a “long term relationship” or an “occasional lover/relationship.” However, 153 of the “happily married” men without pictures indicated they were looking for a long term or occasional romantic relationship. Somewhat similarly, out of the 281 “unhappily married” men, 174 were looking for romantic relationships, even though only 9 actually declared this purpose and posted a picture. Women appeared to be somewhat more cautious – of the 46 happily or unhappily married women declaring that they are looking for a long term relationship, none posted a picture. However, we do have to take into account that the pictures of men could be pranks played upon them by friends/colleagues. 10

9

The income profiles also reflect similar patterns to the education profile. Site users are in general a higher income sample than the overall CPS population, but not compared to the “Internet using” population. One visible difference between income profiles is that about 4% of site users declare their annual income to be above $200,000, whereas the CPS samples contain 0.0% of the population in this cell. LOOK INTO WHETHER THIS IS DUE TO TOPCODING. These comparisons suggest that although the online dating site appears to attract a single, somewhat younger, higher educated and higher income set of users than the representative population, once we condition on household Internet use, the discrepancies that remain are not very large. We believe this is a reflection of the fact that online dating technology has managed to diffuse quite thoroughly among the population of Internet users in recent years. Our data also contains information regarding the occupations of the site users. In Boston, the top occupation for men was engineering/science or other technical professions (16%). This was followed by men in executive/managerial professions (13%), students (11%), sales & marketing (10%), and men who are self-employed (10%). Women were about evenly divided between being students (10%) and in a health-care related profession, or administrative/secretarial jobs, executive/managerial posts, sales & marketing. For San Diego, the top 5 male professions were quite similar, but with more military personnel showing up in the sample than students. San Diego women also largely resembled their Bostonian counterparts. C. Reported physical characteristics of the users An interesting feature of our data set is that we have detailed (but self-reported) information regarding physical attributes of the users. As mentioned in the previous section, roughly a quarter of the users have posted photos online, however, for the rest of the users, the survey acts as the primary source of information for their appearance. The most direct question regarding “looks” asks the users to rate themselves on a rather subjective scale. 21% of men in Boston (20% in San Diego) claim to possess “very good looks,” with 22.8% of the women (25% in San Diego) doing the same. 44.6% of men (49.3% in San Diego) claim to have “above average” looks, with 46% of women (47.5% in San Diego) making 10

the same declaration. Finally, 29.0% of men (27.3%), and 27.5% (24.3%) of women classify themselves as having “average” looks. Interestingly, less than 1% of each gender describe themselves as having “less than average” looks (and about 2-3% do not report their looks, or choose a category that can be interpreted as a joking/non-serious answer to this question.) Somewhat interestingly, when we condition on having photos online, these percentages are not very different, but slightly revised towards the “above average” category. In Boston, 19.6% of men with photos describe themselves as having “very good looks” (19.0% in SD), 52.3% say they have “above average” (56.9% in SD), and 26.6% say they possess “average” looks (23% in SD). Among women, those who claim to have “very good looks” comprise 24.5% (24.7% in SD), “above average” 51.5% (56.6% in SD), “average” 23.1% (24.7% in SD). The registration survey also asks users to report their height and weight. We used this reported information to compare it with U.S. (white) population characteristics obtained from the National Health and Examination Survey Anthropometric tables (which report data from the 1988-1994 survey). Table 2 reports this comparison. As can be seen from the first few rows of the comparison, the reported weight of the women on the dating site is much less than the U.S. population averages. The discrepancy is largest for women in the 50-59 age group, where the reported weight is 20 to 25 pounds less than the national average. This may be driven by the relatively small sample size in this age group; however, women in the 30-39 age group also report weights that are about 20 pounds less than national averages. The discrepancy is still present, though somewhat less, for women in the 20-29 age group. Compared to women, reported weights of men on the dating system reflect national averages almost exactly. Where men differ somewhat is in reported heights – men on the dating system report heights that are about 1 inch taller than national averages. A similar discrepancy (of being 1 inch taller) is also present in the heights reported by women on the dating system. These differences translate into reported body-mass indices (BMI) that are 2 to 5 points less than national averages for women, and .6 to .9 points less than national averages for men. Aside from raw height and weight information, the survey also includes a question regarding “body type.” 10.5% of Boston men (8.9% in SD) report having a chiseled body that’s due to 11

working out everyday, with 42.5% (46.7% in SD) saying they keep themselves toned and fit. Women are less likely to claim they have a chiseled body (2.2% in Boston, 3.4% in San Diego), but many claim they have a toned, fit body (37.8% in Boston, 39.7% in San Diego). Almost as many (40.7% in Boston 39.3% in SD) women describe themselves as being “height/weight proportionate,” with 34.3% of men (32.1% in SD) doing the same. Finally, while only 8.3% of men in Boston (8.1% in SD) choose an option indicating that they may be heavier than “height/weight proportionate”, 15% (14.3% in SD) of women make the same admission. A very interesting comparison between the sexes results from the answer to the question “What is your hair color?” While only 9.2% of men in Boston (12.7% in SD) say they have blond hair, 26.2% of women (28.4% in SD) say they have blond hair.12 In comparison, whereas 74.4% of men report having black or brown hair in Boston (71.6% in SD), only 54.1% of women (50.6% in SD) do the same. Women’s second and third most-favored hair colors appear to be auburn (about 6% higher than male incidence of this color) and red (2-3% higher than male incidence). D. Measured Physical Characteristics of the Users [IN PROGRESS] We are currently in the process of conducting controlled laboratory experiments in which we ask subjects to provide 1-10 ratings of the member photos. We hope to obtain about 10 independent ratings per photo. We will then compute averages and standard deviations of these ratings (controlling for a subject-level mean and dispersion) across subjects to construct “attractiveness ratings” for the users. While the pool of experimental subjects, who are mostly University of Chicago students, is not likely to be representative of the web site users, a large literature in experimental psychology suggests that attractiveness rankings of human face photos are remarkably stable across very different subject pools. [CITATION]

12

This might, of course, be driven by natural blondes being more likely to participate in online dating.

12

3. A Modeling Framework for Analyzing User Behavior The seminal economic model of marriage markets is developed by Becker (1973 (a,b)) in the absence of search/matching frictions. Several authors, such as Morgan (1995), Lu and McAfee (1996), Burdett and Coles (1997), Shimer and Smith (2001), Smith (2002), Atakan (2004) have investigated Becker’s model by adding search frictions. We now describe the Smith (2002) model to motivate the way we will analyze our data on user behavior.13 Assume that there are a large number of men and women in the market at each instance in time. Both men and women are indexed by observable scalar types y and x ∈ [0,1] . Time is continuous, and men and women meet each other randomly in every period according to a Poisson process with parameter ρ , and agents discount the future at instantaneous rate r . Each man and woman can be in one of two states, either they are in a relationship, or they are single.14 Matches are exogenously broken at instantaneous rate δ , upon which both the man and the woman reenter the market. If woman x is in a relationship with man y, she gets an instantaneous flow of utility given by the function f(x,y), which is increasing in the partner’s type. To describe the strategies, we now define several primitives: 1) The acceptance set A(x) describes the set of men on the unit interval that a woman of type x would agree to match with. 2) The opportunity set Ω( x) = {y ∈ [0,1] | x ∈ A( y )} describes the set of men who would be willing to match with x.. 3) The matching set M ( x) = A( x) ∩ Ω( x) is the set of men with whom x has a mutual match with. 4) Let w(x) be the present discounted value to woman x of being single 5) Let w(x|y) be the present discounted value to woman x of being matched with man y. Smith (2002) is the non-transferable utility (NTU) version of Shimer and Smith (2002). Since the model setup and results are quite similar, we will forego a discussion of the Nash-bargaining assumption underlying the transferable utility (TU) model. Smith (2002) points out that both TU and NTU assumptions have been used in the marriage market literature. 14 Although it is possible for site users to “juggle” simultaneous relationships, users can easily monitor whether their “partner” is active on the site, at least under their online alias. 13

13

Let the (stationary) type density of unmatched women be given by u(y), and the density of men to be u(x). While unmatched, woman x enjoys zero flow utility. However, in each instant of time, she meets a man with whom she has a mutual match with probability ρ

∫ u ( y)dy

and enjoys a

M ( x)

“capital” gain of (w(x|y)-w(x))/r. Hence, the value of being single is given by the equation rw( x) = ρ

∫ [ w( x | y ) − w( x)]u ( y )dy

M ( x)

While in a relationship, the woman enjoys the flow utility of f ( x, y ) , but this match is destroyed with chance δ . Hence, the value of being in a relationship is given by the formula: w( x | y ) = f ( x, y ) − δ ( w( x | y ) − w( x )) / r

These two equations allow us to solve for the value of staying single, which is given by

ρ w( x) =

∫ f ( x, y)u ( y)dy

M ( x)

r +δ + ρ

∫ u ( y)dy

M ( x)

The acceptance set A(x ) is determined by this reservation value. I.e. y ∈ A(x) iff f ( x, y ) ≥ w( x) . The matching set M (x) imposes the additional restriction that f ( y , x ) ≥ w( y ) , i.e. the condition that x ∈ A( y ) . These definitions allow Smith (2002) to characterize several properties of equilibrium matching sets without further assumptions. In particular, we have: 1) (Choice Monotonicity) In equilibrium, if woman x is willing to match with y, she is willing to match with z>y. 2) (Increasing Opportunities) Because any woman willing to match with y is willing to match with z>y, Ω( x) ⊆ Ω( z ) 3) (Marginal partner) The acceptance set for woman x is characterized by A(x) = [a(x),1], where a(x) is the marginal man satisfying f(x,a(x))=w(x). Moreover, the matching set is characterized by M ( x) = {y : y ≥ a( x), x ≥ a( y )}

14

Smith (2002) then defines positive assortative matching in the following manner: Matching is positively assortative if when x1 < x2 and y1 < y 2 , y1 ∈ M ( x 2 ) and y 2 ∈ M ( x1 ) imply

y1 ∈ M ( x1 ) and y 2 ∈ M ( x2 ) -- that is, whenever a high type man is matched with a low type woman, and a high type woman is matched with a low type man, the two couples would be willing to switch partners (so that the high type man is matched with the high type woman, and the low type man is matched with the low type woman). Smith (2002) then shows that this definition is equivalent to the marginal man a (x) for woman x to be weakly increasing in the woman’s type, i.e. higher type women are choosier. The main theorem in Smith (2002) demonstrates that obtaining positive assortative matches in equilibrium defined as above depends on properties of the flow utility function. In particular, f being log supermodular (i.e.

∂ log f ( x, y ) ≥ 0 ) is a sufficient condition for positive assortative ∂x∂y

matching in equilibrium. B. Empirical Implications of the Smith (2002) Model The main assumption that allows us to use the Smith (2002) framework to interpret our data is to assume that searching members contact every member (or a constant fraction of them) they would be willing to match with. I.e. man y sends an e-mail to every woman x ∈ A( y ) , and a vice versa (or, as stated, a constant fraction of these women, where the fraction does not depend on the woman’s type). An additional assumption is that the rate at which different types of men and women “meet” each other in unit time is uniform across the population. Under these assumptions, the “choice monotonicity” and the “increasing opportunities” properties described above guarantee that higher “type” women ( men) receive e-mails at a higher rate. This allows us to use the observed rate of contacts received by an individual as a measure of her/his “type.” We can then project this type on various observed attributes to investigate the relative weights comprising the index. Notice that the strictly “vertical” nature of the demand system considered here (i.e. the rankability of men and women) can be relaxed somewhat. In particular, let’s assume that the flow

15

utility that woman x gets from being in a relationship with man y is given by f ( x, ~y ) , where ~ y = y + ε , and ε is a mean zero “match specific utility” specific to the woman, drawn iid from a distribution F (ε ) for each encounter. This allows women to disagree on the rankings of particular men; however, the basic structure of the equilibrium does not change. In particular, there will still be a “marginal man” a (x) describing the match set of woman x , where f ( x, a ( x )) = w( x ) . Since ε is iid mean zero, it will still be true, on average, that higher type

women and men will be found acceptable more often by people they meet, and will thus receive more e-mails. If there are nontrivial costs associated with writing e-mailing women (or men) that one finds acceptable, then the monotonic relation between a person’s “type” and the number of e-mails s/he receives may no longer hold true. Consider, for example, the Smith (2002) setup where there is no idiosyncratic match utility term. In this case, the equilibrium match set for man y , M ( y ) is deterministic. If it is also the case that higher type women have higher acceptance

thresholds, i.e. we are in PAM equilibrium, then, for certain men there will exist some women who are acceptable, but will never accept an offer. In this case, an equilibrium of the matching game is to send offers only to people within the matching set (if the cost of sending an e-mail is c , the equilibrium matching sets can be calculated by replacing f ( x, y ) by f ( x, y ) -c everywhere in the previous section’s discussion). In this equilibrium, however, the number of e-mails received by a woman is a monotonic function of the number (measure) of men located within her match set. As examples in Smith (2002) demonstrate, this implies that the number of e-mails received by a woman may not be a monotonic function of her “type.” Fortunately, unlike in conventional “marriage markets” where the costs of asking people out15 may indeed be nontrivial, online dating is designed to provide an environment that minimizes this cost. Aside from any psychological cost of rejection, the main cost associated with sending an email is the cost of composing it – however, the marginal cost of producing yet another witty email should not be exaggerated since one can always personalize a polished form letter. Moreover, with the addition of a “match-specific utility” component as above, match sets will no These costs may include the embarrassment of being rejected. In more traditional societies, the cost of “propositioning” the wrong woman (or man) may be the loss of life. 15

16

longer be deterministic, i.e. they will attach probabilities over the event that x and y will be matched. Furthermore, with a large enough ε , even the highest type woman may find the lowest type man acceptable. If the expected benefit from such a match exceeds the marginal cost of sending e-mail, we should not expect people to strategically refrain from e-mailing users they find acceptable.16

4. The Relationship between User Attributes and Online Outcomes In this section we explore how mate attributes such as the stated goals for being on the dating site, looks and physique, income, education, etc., are related to dating outcomes. We employ several outcome measures which correspond to the successive stages in an online relationship. First, we observe the number of times a user’s profile has been browsed.17 Second, we observe how many e-mails a user received as a first contact from another user (i.e., user A receives an email from B, but has not sent an e-mail to B previously). Finally, based on a keyword search of all e-mails that were exchanged between the users of the dating site, we know whether a user received an e-mail containing a phone number or an e-mail address from another user. This provides some information on the progress of the online relationship, possibly to an offline affair. We refer to the three outcome variables as “browses”, “e-mails”, and “keywords”. We have discussed before that under certain assumptions, the “type” or utility index of a user is positively related to the number of times the user is approached by other members of the dating site. Hence, if we can rule out strategic behavior as a major influence on the users’ decision making, our outcome measures reveal the ranking of a user’s type. A further qualification for this statement is that preferences over mates are homogeneous. We first present results that are based on this assumption that men and women have homogeneous preferences over their potential partners. We then relax this assumption, and assess the importance of preference heterogeneity by examining how outcomes vary across different segments of the dating population.

16 The exact characterization of match sets and whether match outcomes will still be assortative when there are explicit costs to making offers and where there is a “match-specific utility” term associated with each pair is an interesting problem on its own. We are currently making an effort to understand this problem. 17 This outcome variable records multiple browses, not only the first browse from a specific user.

17

We use a Poisson regression model to estimate the relationship between the outcome variables and the user attributes. A count data model, such as a Poisson regression, is particularly appropriate for the integer outcomes in our application.18 We specify the conditional expectation of the outcome variable as E( Y | x ) = exp( x ′ β ),

(1)

where x is a vector of user attributes. Under the Poisson assumption, this conditional expectation fully determines the distribution of the outcome variable. The Poisson assumption places strong restrictions on the data. In particular, the conditional variance of a Poisson distributed outcome variable equals the conditional expectation, Var( Y | x ) = E( Y | x ). However, as long as the conditional expectation (1) is correctly specified, the (quasi) maximum likelihood estimator associated with the Poisson regression model is consistent, even if the Poisson assumption is incorrect (Wooldridge 2001, pp. 648-649). We report robust (under distributional misspecification) standard error estimates for the regressions (Wooldridge 2001, p. 651). In our application, all regressors are categorical variables indicating the presence of a specific user attribute. If two users A and B differ only by one attribute that is unique to A, with the associated regression coefficient β j , the ratio of expected outcomes is E( Y | x A ) = exp( β j ). E( Y | x B )

(2)

exp( β j ) is called the incidence rate ratio. This factor measures the premium (or penalty) from a specific attribute in terms of an outcome multiple. For example, using the number of e-mails received as outcome variable, the coefficient associated with “some college” education is 0.18 for men. Hence, holding all other attributes constant, men with some college education receive, on average, exp(0.18) = 1.20 as many e-mails as the baseline group, men who have not finished high school yet. Alternatively, we can calculate the “college premium” for men as 100 × (exp(0.18) − 1) = 20% .

18

Alternatively, a linear regression model has the obvious disadvantage of predicting negative outcome values for some user attributes. A logarithmic transformation of the outcome variable avoids this problem, but would force us to drop many observations for which the outcome measure is zero. Furthermore, it is not clear how the estimated conditional expectation E(log( Y )| x ) is related to the object of our interest, E( Y | x ) . The same problem pertains to the transformation log(1 + Y ) , which is defined for outcome values of zero.

18

The results from the Poisson regressions are presented below for different types of user attributes, such as looks, income, and education. Separate regressions were estimated for men and women. As the outcome numbers are only meaningful if measured with respect to a unit period of time, we include the (log) number of days a user was active on the dating site as a covariate. Also, we include a dummy variable for users who were members of the website already before the start of the sampling period. Table 4-1 presents summary statistics of the outcome measures. Women are browsed more often, and receive more first contact e-mails and e-mails containing a phone number or e-mail address than men. Hence, a first contact is more likely to be initiated by a man. While men receive an average of 2.6 first contact e-mails, women receive 10.5 e-mails. 56.6% of all men in the sample did not receive a first contact e-mail at all. On the other hand, only 22.9% of all women were not approached by e-mail. 4.1 Outcome regression results: Homogeneous preferences over mates A. Effect of goals on outcomes The site members can state their goals for joining online dating system in their profile. The majority of all users are “Hoping to start a long term relationship” (35.8% of men and 37.5% of women), or are “Just looking/curious” (26.0% of men and 26.9% of women). An explicitly stated goal of finding a partner for casual sex (“Seeking an occasional lover/casual relationship”) is more common among men (14.5%) than among women (5.2%). The impact of these stated goals on online success differs strongly across men and women (table 4-2, figure 4-1). Men who indicate of a preference for a less than serious relationship or casual sex are browsed less often, and receive fewer first contact e-mails and “keywords” than men who appear more serious and state that they are “Hoping to start a long term relationship”. Women, on the other hand, are not negatively affected by these indications. To the contrary, women who are “Seeking an occasional lover/casual relationship” receives 38% more e-mails relative to the baseline, while men experience a 42% penalty in terms of first contact e-mails. The statement “I’d like to make new friends. Nothing serious.” is associated with a 30% decrease in e-mails for men, 19

but leaves the outcome of women virtually unaffected relative to the baseline. “Just looking/curious” is associated with a 14% penalty for men, and a 7% improvement in outcomes for women. Finally, men who are “Scouting for swinging couples” receive 42% fewer e-mails relative to the baseline, while women experience only a smaller and statistically insignificant drop of 22%. B. Looks and physical attributes We find consistent evidence that looks and physical attributes such as height and weight are strongly correlated with the members’ online dating outcomes (table 4-3, figure 4-2). Men who include a picture in their profile are browsed four times more often, receive three times as many first contact e-mails, and receive 72% more “keyword” e-mails than men without a photograph. Women with a photo in their profile get more than six times as many “browses”, almost four times as many e-mails, and receive 152% more “keyword” e-mails than women without a photograph. The members of the dating site can describe their looks in their profile. As reported before, only about a quarter of all members report that they “look like anyone else walking down the street”, while almost 70% claim that they have “above average” or “very good looks”. Whether our sample is in fact better looking than the rest of the population, overly optimistic about their own physical attractiveness, or inflates the statement of their looks for strategic reasons, is hard to assess. The data reveal, however, that members of “above average” or “very good looks” are browsed more often, and receive more e-mails and contact information keywords. Men with “above average looks” receive 27% more e-mails than average looking members, and “very good looks” are associated with an average increase of 57% in terms of received e-mails. The effect sizes are somewhat smaller for women; “above average looks” is associated with a 16% increase, and “very good looks” is associated with a 36% increase in e-mails received. The description “less than average looks” is associated with a decrease in outcomes, and carries a penalty of 21% fewer e-mails for women and 54% fewer e-mails for men. Further evidence on the importance of physical attributes is provided by the members’ description of their physique, height, and weight. Relative to “height-weight proportionate” 20

members, men who describe themselves as “chiseled, I work out every day!” have higher outcomes in terms of all three measures. “Toned, I keep fit” is associated with an improvement in outcomes for both sexes. On the other hand, the descriptions “Voluptuous/portly” and “Large but shapely” are associated with lower outcomes for both men and women. Weight has only a small impact on the outcomes of men, but has a large impact on the online success of women (figure 4-3). Men above 151 lbs are browsed more often and receive more first contact and keyword e-mails than men in the range of 111 to 150 lbs. However, this effect is small and not estimated precisely. On the other hand, the success of women declines dramatically above and below the 101-120 lbs weight range. For example, women with a self-reported weight of 151-160 lbs receive 53% fewer first contact e-mails than women in the 101-110 lbs range. For women in the 191-200 lbs category, the decline is 77%. As regards to height, we find some evidence that size matters mostly for men only (figure 4-4). The number of e-mails received is overall monotonically increasing in the height of a man. For example, men in the 6’3’’-6’4’’ range receive 40% more e-mails than men in the 5’7’’-5’8’’ range. In contrast, most women’s success is mostly unaffected by their height, although very short and tall women seem to suffer some penalty. The effect of height is not estimated precisely using the “browses” or “keywords” outcome variables. As regards hair color (using brown hair as the baseline), we find that men with red hair suffer a moderate outcome penalty. Blonde women have a slight improvement in their online “success” if measured by the number of times they were browsed, while women with gray or “salt and pepper” hair suffer a penalty. Men with curly hair receive only about half as many e-mails as men in the baseline category, “medium straight hair”. For women, “long straight hair” leads to a slight improvement, and any kind of short hair to a moderate decrease in outcomes. Interestingly, while the description “I’m bald on top with a fringe” leads to an almost 50% decrease in e-mails received for men, being “completely bald” is not associated with any decrease in online success. C. Income

21

About 65% of men and 53% of women report their income. Table 4-4 and figure 4-5 show how these self-reported income measures are related to the site members’ dating outcomes. We find that income strongly affects the success of men, as measured by the number of first contact emails received. While there is no apparent effect below an annual income of $50,000, outcomes improve monotonically for income levels above $50,000. Relative to incomes below $50,000, the increase in the expected number of contacts is at least 50%, and as large as 139% for incomes in excess of $250,000. In stark contrast to the strong income effect for men, the online fortunes of women are largely unrelated to their income. The same patterns, for men and women, can be found for the “browses” and “keyword” outcome variables. The effect size at different income levels (for men) is overall larger for “keywords” than for first contact e-mails, and larger for first contact e-mails than for “browses”. D. Educational attainment The relationship between online dating outcomes and education is less pronounced than the effect of income. However, we find some evidence that—similar to the income effect—higher levels of education increase the online success of men but not of women (table 4-5, figure 4-6). Measured by the number of first contact e-mails received, there is a college and graduate education premium for men. In particular, relative to men whose highest educational attainment is finishing high school, a college degree is associated with a 26% increase in the number of emails received. Higher levels of education lead to the approximately same increase in outcomes. In contrast to these findings for men, the outcomes of women do not improve with their educational attainment. To the contrary, college juniors and seniors, women in a post-graduate program, and women with a master’s degree incur a slight outcome penalty. The same overall patterns, for both men and women, can also be found for the “browses” and in particular for the “keyword” outcome variables. For men, we find somewhat larger education effects from the “keywords” than from the first contact e-mails measure, although the estimates from the “keyword” variable are not statistically significant. E. Occupation

22

We find that online success varies across different occupational groups (table 4-6). Here outcomes are measured relative to students, who are chosen as the baseline group. For men, the most successful occupations are “executive/managerial”, “financial/accounting”, “legal/attorney”, “law enforcement/firefighter”, various health related professions, and members of the military. Men in these professions receive an above average number of first contact emails, “keywords”, and are mostly browsed more often than men in other occupations. Some of the effects are large; lawyers, for example, receive 68% more first contact e-mails than students, and law enforcement officers and firefighters receive 73% more e-mails. The outcomes of women, on the other hand, are much less strongly influenced by their professions. In fact, students are the most successful group among women, while “law enforcement/firefighters” and members of the military are among the least successful.

4.2 Outcome regression results: Heterogeneous preferences (PRELIMINARY) This section provides evidence on the extent of preference heterogeneity over partner attributes. To capture differences in the way different users evaluate the characteristics of their potential mates, we create outcome measures with respect to more narrowly defined segments of the partner population. First, we segment users into a low income group, with annual incomes of up to $50,000, and a high income group, with annual incomes in excess of $50,000. Second, we create a segment of users whose highest level of education is finished high school, and another segment of users who completed or are working towards a college degree. The outcome measures, for men and women, are then created by counting the number of browses, e-mails, and keywords from users in one of the specific segments. Thus, we can assess how preferences differ by income and by education. We first focus on the importance of looks and physique. Table 4-7 confirms the importance of a photo in the user’s profile, self-reported looks, and weight and height that we found above in the case of homogeneous preferences. There is no stark evidence of preference heterogeneity, although women with weights above 120 lbs seem to be “punished” more strongly by high income men than by low income men.

23

With respect to income, there is stronger evidence for preference heterogeneity than in the case of looks. Table 4-8 shows that, as in the case of homogeneous preferences, high income men receive a larger number of e-mails than low income men. This effect, however, is more pronounced with respect to e-mails originating from high income women. Also, college educated women appear to have a slightly stronger taste for income than high-school educated women, although the difference is less clear than in the case of high versus low income women. For females, there is only little evidence that income is related to their outcomes. High school educated men seem to have a slight preference for women in the income range up to $75,000, while college educated men seem to prefer women above this income bracket. However, these effects are small in comparison to the income effect sizes found for men. We also find evidence for preference heterogeneity over education (table 4-9). High income women have a stronger preference over college educated men, and in particular over men with a graduate degree, than low income women. This pattern is even more pronounced if we compare the success of men with respect to high school and college educated women. Men with a doctoral degree receive 42% fewer first contact e-mails (relative to men who have finished high school only) from high school educated women. On the other hand, a doctoral degree is associated with an 82% increase in the number of first contact e-mails originating from college educated women. A similar, although less strongly pronounced pattern holds for women, i.e. women with a higher level of education are approached less often by low income and high school educated men relative to high income and college educated men.

5. Analysis of Sorting Patterns Who matches with whom? The previous section analyzed what attracts e-mails, but we have not analyzed whether these e-mails are responded to – a necessary precursor to forming a match. Moreover, the exchange of e-mails may not be sufficient to form a longer relationship; it may take several e-mails before the pair trusts each other enough to meet in person. Our data allows us to partially analyze the above question. As described in Section 2, we can observe whether an e-mail contains a phone number, an e-mail address, or certain keywords such 24

as “meet.” Using these data, we construct an indicator for a “match” whenever two e-mails containing such information are exchanged (i.e., for a “match” it is not enough for a man to offer his phone number, we also require that the woman responds by sending hers). Clearly, this definition of a “match” constitutes a very initial stage of a relationship. However, we should also consider that the main purpose of online dating sites is to bring two individuals close enough together to trust each other with their physical contact information. Tables 5-1 and 5-2 report Spearman rank correlations of various characteristics of matched couples in our data set, where a match is defined as above. First note that the age correlation is extremely high – the rank correlation across the nine age categories in our data set is 0.82. Table 5-1 reports the cross-tabulation over coarser age categories. Consistent with general demographic patterns, the diagonal entries in the table have the most mass, and younger women are more likely to match with somewhat older men. Table 5-2 uses this effective segmentation to report the correlations between partner attributes by age group (of the woman). First, notice that the number of e-mails received per week of activity is positively correlated across men and women for all age groups in the sample. The correlation coefficient is around 0.25, and highly statistically significant for all sub-samples. This is an encouraging finding, since it implies that couples sort along the “attractiveness index” we utilized in the previous section’s regressions. Recall, however, that the Smith (2002) model does not necessarily imply this positive correlation to be true in equilibrium – what is needed is the logsupermodularity of the match utility function f(x,y). Looking at further rows of the table, we see that there are distinct sorting patterns along income, education, and physical characteristic dimensions. Positive income sorting appears most prevalent for the 20-29 and 50-59 age groups. Sorting by education is also present for the 20-50 year old group, but not present for the 50-59 year old group. There also appears to be positive sorting along the body-mass index (BMI) dimension, though, interestingly, sorting patterns appear weaker when we use (preliminary) ratings of member photos. Note, however, that the “number of e-mails received” measure appears to display the most robust positive correlations compared to the other measures. This provides further 25

encouragement for the empirical methodology utilized in the previous section, since, according to the econometric model employed, factors such as income, education and physical characteristics make up only parts of a profile’s overall attractiveness. Therefore we might expect sorting to occur along this composite index of attractiveness, as opposed to along individual attributes. To the extent that they reflect actual matching preferences of men and women, our findings in section 4, especially those regarding the effect of looks and the differential effect of income across men and women, can be extrapolated to make some rather controversial, though intuitive predictions: if men care primarily about women’s physical attractiveness, but women do not care as much about men’s physical attractiveness, we might expect high income men to end up matching with physically attractive women.19 The last four rows of Table 5-2 indicate that this extrapolation may not be far from reality. Men’s income in all age categories appear negatively correlated with the BMI of the women. When we utilize (preliminary) ratings of member photos, we also find a positive correlation between men’s income and women’s beauty. Women’s income, however, does not exhibit as strong correlation with men’s physical attractiveness measures (except for younger women, who appear to be matched with lower BMI men).

19 We recognize that the proper manner to make this extrapolation is to use estimated values of preference parameters to calculate the matching equilibrium in model such as that of Smith (2002). This task is left for future research.

26

References Becker, G. S. (1973): “A Theory of Marriage: Part I,” Journal of Political Economy, vol. 81 (4), 813-846. Becker, G. S. (1974): “A Theory of Marriage: Part II,” Journal of Political Economy, vol. 82 (2), Part 2: Marriage, Family Human Capital, and Fertility, S11-S26. Choo, E. and A. Siow (2003): “Who Marries Whom and Why,” manuscript, University of Toronto. Fisman, R., S. Iyengar, and I. Simonson (2004): “Revealed Preference Determinants of Mate Selection: Evidence from an Experimental Dating Market,” manuscript, Columbia University. Gale, D. and L. S. Shapley (1962): “College Admissions and the Stability of Marriage,” American Mathematical Monthly, 69, 9-14. Shimer, R. and L. Smith: “Assortative Matching and Search,” Econometrica, 68 (2), 343-370. Smith, L. (2002): “The Marriage Model with Search Frictions,” manuscript, University of Michigan. Wong, L. Y. (2003): “An Empirical Study of Darwin's Theory of Mate Choice,” manuscript.

27

Tables Table 2-1. Dating System Members and County Profile of General Demographic Characteristics

Variable

Dating System

San Diego General Population

General Information Total Member and Population

15,034

2,026,020

1,180,020

14,911

2,555,874

1,581,711

Percentage of Males

57.0

49.9

49.4

56.4

49.0

50.6

Age Composition 18 to 20 years 21 to 25 years 26 to 35 years 36 to 45 years 46 to 55 years 56 to 60 years 61 to 65 years 66 to 75 years Over 76

8.2 15.0 30.5 25.8 15.3 3.0 1.0 0.7 0.5

6.0 9.5 21.3 23.0 18.5 6.3 2.9 6.9 5.7

6.4 11.5 18.8 28.6 19.0 6.5 3.6 4.8 0.8

6.8 16.0 32.6 25.3 15.1 2.3 0.9 0.5 0.4

5.8 9.3 17.2 23.1 17.6 7.3 4.3 8.8 6.8

7.2 12.0 19.7 26.8 20.1 6.9 3.7 2.9 0.7

Race Composition (1) Whites Blacks Hispanics Asian Other

63.4 4.7 11.8 5.5 14.8

61.9 4.8 19.5 13.0 0.9

71.3 4.2 9.8 13.6 1.1

71.8 4.9 4.4 4.3 14.6

84.2 7.4 4.4 3.8 0.3

89.1 4.2 2.3 4.2 0.2

67.1 7.0 4.1 1.8 20.1

31.8 57 1.2 2.3 8.1

28.5 62.0 0.7 1.5 7.4

69.4 8.0 4.3 1.3 17.0

35.3 54.1 1.1 3.6 6.0

36.8 56.7 0.3 1.0 5.2

63.1 3.8 3.5 3.2 26.5

20.2 57 3.9 6.3 12.3

23.9 62.5 1.9 2.0 9.7

66.7 2.7 4.2 2.8 23.6

28.0 49.0 2.4 13.4 7.2

32.7 55.9 0.9 3.5 7.0

1.9

12.1

3.0

1.9

9.2

3.2

Internet User

Dating System

Boston General Population

Internet User

Marital Status Males Never married Married & not separated Separated Widowed Divorced Females Never married Married & not separated Separated Widowed Divorced Educational Attainment (2) Have not finished high school

28

High school graduate Technical training (2-year degree) Some college Bachelor's degree Master's degree Doctoral degree Professional degree Income (3) Total Individuals with Income information

10.2

23.0

17.8

12.3

30.1

20.4

7.3 32.2 27.2 10.7 3.2 7.3

5.2 27.9 22.7 6.0 1.5 1.7

5.4 28.5 31.5 9.0 2.6 2.3

4.8 24.2 32.5 15.2 3.6 5.4

7.3 14.1 22.2 11.7 3.3 2.0

7.6 15.0 29.7 16.3 5.2 2.6

8,927

283,442

224,339

8,790

396,065

281,619

Less than $12,000 $12,000 to $15,000 $15,001 to $25,000 $25,001 to $35,000 $35,001 to $50,000 $50,001 to $75,000 $75,001 to $100,000 $100,001 to $150,000 $150,001 to $200,000 $200,001 or more

9.9 12.5 12.4 11.2 7.6 4.6 5.5 3.0 1.9 4.4 5.0 6.0 9.5 13.8 10.1 6.1 21.4 16.2 14.1 23.3 22.3 13.9 19.9 21.4 19.6 12.4 10.6 20.4 16.5 18.5 18.6 17.3 20.2 20.9 21.7 24.6 10.1 7.2 9.1 11.1 4.8 4.5 6.1 7.5 9.5 6.6 1.9 2.7 2.6 3.2 4.0 2.1 1.1 1.6 4.1 0.0 0.0 3.4 0.0 0.0 Source. Estimates from CPS Internet and Computer use Supplement, September 2001. All the CPS estimates are weighted. All the individuals for the CPS and members are constrained to be 18 years of age or older. The percentages for the column "Internet user" is calculated conditioning the CPS sample to those individuals who declared to use the Internet. Notes. Geographical information is based on Metropolitan Statistical Area. Boston PMSA includes a New Hampshire portion. San Diego geographic information corresponds to San Diego MSA. Member information as 2003. (1) The figures for Whites, Blacks and Asian and Other race for the CPS data correspond to those with non-Hispanic ethnicity. (2) Education excludes certain categories on member data that can not be translated into years of educational attainment. (3) The income figures from the CPS data were adjusted to 2003 dollars.

29

Table 2-2. Physical Characteristics of Dating System Members vs. General Population

Variable

Dating System Dating System (San Diego) (Boston)

General Population*

Women' s weight (lbs) 20-29 years 30-39 years 40-49 years 50-59 years 60-69 years 70-79 years

132.8 135.8 136.2 137.1 142.6 147.5

137.3 135.6 138.7 142.6 151.4 131.4

141.7 154.2 157.4 163.7 155.9 148.2

Men's weight (lbs) 20-29 years 30-39 years 40-49 years 50-59 years 60-69 years 70-79 years

172.8 183.6 187.3 185.8 185.3 182

170.4 183.1 187.1 188.7 189.3 177.6

172.1 182.5 187.3 189.2 182.8 173.6

Women' s height (inches) 20-29 years 30-39 years 40-49 years 50-59 years 60-69 years 70-79 years

65 65.1 64.9 64.5 65 65.1

65 64.9 64.8 64.7 64.4 64.9

64.1 64.3 64.1 63.7 63.1 62.2

Men' s height (inches) 20-29 years 30-39 years 40-49 years 50-59 years 60-69 years 70-79 years

70.2 70.5 70.6 70.5 70.2 69.3

70 70.3 70.2 70.3 69.5 70.2

69.3 69.5 69.4 69.2 68.5 67.7

Women' s BMI** 20-29 years 30-39 years 40-49 years 50-59 years 60-69 years 70-79 years

22.1 22.6 22.8 23.2 23.8 25.1

22.9 22.7 23.3 24 25.7 22.5

24.3 26.3 27 28.4 27.6 26.9

Men' s BMI** 20-29 years 30-39 years

24.6 26

24.5 26.1

25.2 26.5

30

40-49 years 50-59 years 60-69 years 70-79 years

26.4 26.4 26.5 26.8

26.7 26.9 27.6 25.7

27.3 27.8 27.3 26.7

*General population statistics obtained from the National Health and Nutrition Examination Survey, 1988-1994 Anthropometric Reference Data Tables. ** BMI (body mass index) is calculated as weight (in kilograms) divided by height (in meters) squared

31

Table 4-1. Description of Outcome Measures

Browsed

E-mails received containing phone E-mails recived no. or e-mail (first contact) address

Men Obs. Median Mean SD Min Max % Zero Obs. >0 Mean | >0 SD | >0

12635 12 78.4 191.4 0 3816 0.0 12629 78.5 191.4

12635 0 2.6 6.8 0 112 56.6 5480 6.1 9.2

12635 0 0.8 3.5 0 105 79.1 2640 4.0 6.9

Women Obs. Median Mean SD Min Max % Zero Obs. >0 Mean | >0 SD | >0

10539 45 218.0 420.0 0 7217 0.0 10535 218.1 420.1

10539 3 10.5 19.0 0 185 22.9 8127 13.6 20.7

10539 0 2.2 4.7 0 97 53.0 4958 4.7 5.9

32

Table 4-2. Stated reason for being on the dating site Men

Seeking an occasional lover/casual relationship I'd like to make new friends. Nothing serious. Looking for a pen-pal only Hunting for a roommate Looking for a travel partner Just looking/curious Scouting for others to do things with A friend put me up to this A higher power brought me here Looking for someone to share my lifestyle with Scouting for swinging couples

Browses -0.22 (0.05)*** -0.13 (0.05)*** -0.31 (0.16)** -0.61 (0.16)*** 0.00 (0.27) -0.05 (0.03) -0.11 (0.09) -0.20 (0.08)** -0.06 (0.08) -0.11 (0.05)** -0.34 (0.10)***

E-mails -0.55 (0.08)*** -0.35 (0.07)*** -0.15 (0.35) -1.30 (0.48)*** -0.30 (0.42) -0.15 (0.05)*** -0.22 (0.14) -0.18 (0.13) -0.03 (0.12) -0.21 (0.08)*** -0.54 (0.21)***

Note: The baseline category is "Hoping to start a long term relationship"

33

Women Keywords -0.82 (0.16)*** -0.49 (0.13)*** -0.52 (0.66) -1.71 (0.68)** 0.51 (0.40) -0.24 (0.10)** -0.50 (0.23)** -0.08 (0.30) -0.15 (0.14) -0.19 (0.14) -1.64 (0.55)***

Browses 0.13 (0.05)** -0.00 (0.03) -0.07 (0.12) 0.18 (0.19) -0.10 (0.17) 0.02 (0.02) 0.07 (0.08) 0.02 (0.04) -0.10 (0.06)* -0.06 (0.04)* -0.25 (0.12)**

E-mails 0.32 (0.06)*** 0.02 (0.04) -0.07 (0.17) 0.13 (0.30) -0.29 (0.28) 0.07 (0.03)** 0.14 (0.09) 0.07 (0.05) -0.10 (0.07) -0.08 (0.05) -0.25 (0.16)

Keywords 0.30 (0.08)*** -0.04 (0.06) -0.20 (0.21) 0.63 (0.31)** -0.16 (0.33) 0.01 (0.04) 0.03 (0.16) 0.01 (0.07) -0.04 (0.09) -0.09 (0.07) -0.36 (0.29)

Table 4-3. Looks and physique Men

Dummy variable if member has photo Very good looks Above average looks Less than average looks Bring your bag in case mine tears Chiseled, I work out every day! Toned, I keep fit Skinny, I could use some carbohydrates Voluptuous/Portly Large but shapely Rotund I look like a reflection in a fun house mirror! Less than 90 lbs (less than 41 Kg) 90 lbs - 100 lbs ( 41 Kg - 45 Kg) 111 lbs - 120 lbs ( 51 Kg - 55Kg) 121 lbs - 130 lbs ( 56 Kg - 59Kg) 131 lbs - 140 lbs ( 60 Kg - 64Kg) 141 lbs - 150 lbs ( 65 Kg - 68Kg) 151 lbs - 160 lbs ( 69 Kg - 73Kg) 161 lbs - 170 lbs ( 74 Kg - 77Kg) 171 lbs - 180 lbs ( 78 Kg - 82Kg) 181 lbs - 190 lbs ( 83 Kg - 86Kg)

Browses 1.41 (0.03)*** 0.20 (0.04)*** 0.10 (0.03)*** -0.24 (0.14)* -0.08 (0.12) 0.12 (0.05)** 0.08 (0.03)*** 0.01 (0.07) -0.18 (0.12) -0.05 (0.07) 0.12 (0.17) 0.02 (0.10) 0.15 (0.25) 0.15 (0.26) -0.11 (0.18) 0.17 (0.22) 0.10 (0.17) 0.06 (0.17) 0.22 (0.16) 0.22 (0.16) 0.25 (0.16) 0.27 (0.16)*

34

E-mails 1.10 (0.05)*** 0.45 (0.06)*** 0.24 (0.05)*** -0.78 (0.28)*** -0.00 (0.15) 0.26 (0.08)*** 0.11 (0.04)*** -0.16 (0.14) -0.63 (0.19)*** -0.18 (0.14) 0.05 (0.43) 0.03 (0.20) -0.63 (0.52) -0.35 (0.52) -1.30 (0.47)*** -0.76 (0.48) -0.87 (0.43)** -1.05 (0.42)** -0.67 (0.42) -0.66 (0.41) -0.58 (0.41) -0.52 (0.42)

Women Keywords 0.54 (0.10)*** 0.52 (0.13)*** 0.27 (0.11)** -0.78 (0.58) 0.09 (0.26) 0.31 (0.15)** 0.08 (0.08) -0.10 (0.25) -0.59 (0.33)* -0.20 (0.34) 1.00 (0.77) -0.48 (0.29)* -1.48 (1.07) -1.98 (1.21) -2.19 (0.89)** -0.80 (0.90) -0.90 (0.78) -1.09 (0.75) -0.83 (0.74) -0.66 (0.74) -0.65 (0.74) -0.46 (0.74)

Browses 1.85 (0.02)*** 0.13 (0.03)*** 0.10 (0.02)*** -0.19 (0.13) 0.14 (0.12) 0.05 (0.06) 0.09 (0.02)*** 0.01 (0.07) -0.11 (0.05)** -0.24 (0.06)*** 0.01 (0.12) 0.09 (0.08) -0.26 (0.17) -0.24 (0.07)*** 0.00 (0.04) -0.02 (0.04) -0.05 (0.04) -0.12 (0.05)** -0.14 (0.05)*** -0.20 (0.06)*** -0.17 (0.07)** -0.21 (0.11)*

E-mails 1.35 (0.03)*** 0.31 (0.04)*** 0.15 (0.03)*** -0.24 (0.16) 0.23 (0.19) 0.09 (0.07) 0.12 (0.03)*** -0.00 (0.08) -0.33 (0.07)*** -0.45 (0.09)*** -0.32 (0.24) -0.04 (0.12) -0.54 (0.25)** -0.28 (0.09)*** -0.03 (0.05) -0.19 (0.05)*** -0.32 (0.05)*** -0.55 (0.06)*** -0.76 (0.07)*** -1.03 (0.08)*** -1.14 (0.09)*** -1.22 (0.17)***

Keywords 0.93 (0.04)*** 0.30 (0.05)*** 0.11 (0.05)** -0.43 (0.21)** 0.21 (0.26) -0.05 (0.10) 0.11 (0.04)*** 0.04 (0.12) -0.44 (0.11)*** -0.44 (0.13)*** -0.49 (0.33) 0.01 (0.19) -0.55 (0.25)** -0.07 (0.12) 0.03 (0.06) -0.12 (0.07)* -0.28 (0.07)*** -0.45 (0.08)*** -0.76 (0.10)*** -0.97 (0.10)*** -1.15 (0.13)*** -0.84 (0.23)***

191 lbs - 200 lbs ( 87 Kg - 91Kg) 201 lbs - 210 lbs ( 92 Kg - 95Kg) 211 lbs - 220 lbs ( 96 Kg - 100 Kg) 221 lbs - 230 lbs (101 Kg - 105 Kg) 231 lbs - 240 lbs (106 Kg - 109 Kg) 241 lbs - 250 lbs (110 Kg - 114 Kg) Over 250 lbs (over 114 Kg) I broke the truck stop scale Under 5 (under 152 cm) 5 - 5 2 (152 cm - 158 cm) 5 3- 5 4 (159 cm - 163 cm) 5 7- 5 8 (169 cm - 173 cm) 5 9- 5 10

(174 cm - 178 cm)

5 11- 6 (179 cm - 183 cm) 6 1- 6 2 (184 cm - 188 cm) 6 3- 6 4 (189 cm - 193 cm) 6 5- 6 6 (194 cm - 198 cm) 6 7- 6 9 (199 cm - 206 cm) 6 10 - 7(207 cm - 213 cm) Over 7 (over 213 cm)

0.26 (0.16) 0.24 (0.16) 0.18 (0.17) 0.19 (0.17) 0.20 (0.18) 0.14 (0.18) 0.29 (0.20) -0.06 (0.24) 0.81 (0.25)*** -0.16 (0.17) 0.06 (0.14) -0.02 (0.06) -0.01 (0.06) 0.00 (0.06) -0.00 (0.07) -0.03 (0.08) -0.02 (0.11) -0.41 (0.15)*** 0.92 (0.31)*** -0.11 (0.17)

35

-0.55 (0.42) -0.61 (0.42) -0.72 (0.42)* -0.63 (0.42) -0.71 (0.43)* -0.80 (0.43)* -0.69 (0.49) -1.09 (0.52)** 1.32 (0.39)*** 0.10 (0.37) -0.29 (0.25) 0.03 (0.10) 0.19 (0.09)** 0.29 (0.10)*** 0.31 (0.10)*** 0.31 (0.12)*** 0.35 (0.17)** 0.08 (0.26) 1.40 (0.38)*** 0.47 (0.43)

-0.50 (0.73) -0.64 (0.74) -0.78 (0.75) -0.49 (0.77) -0.76 (0.76) -0.90 (0.78) -0.34 (0.99) -1.58 (1.01) 1.89 (0.61)*** 0.72 (0.53) -0.72 (0.41)* -0.06 (0.19) -0.00 (0.16) 0.06 (0.17) 0.04 (0.18) 0.04 (0.22) 0.09 (0.31) -0.42 (0.63) 0.37 (0.88) 0.74 (0.93)

-0.23 (0.13)* -0.16 (0.13) -0.13 (0.13) -0.46 (0.18)** -0.11 (0.20) -0.27 (0.14)* -0.29 (0.14)** -0.33 (0.10)*** 0.07 (0.09) -0.01 (0.03) -0.05 (0.02)** 0.00 (0.03) 0.01 (0.04) 0.00 (0.06) -0.03 (0.14) -0.47 (0.48) -0.06 (0.24) 0.52 (0.33) -0.64 (0.37)* -0.29 (0.24)

-1.47 (0.15)*** -1.27 (0.25)*** -0.79 (0.32)** -1.63 (0.30)*** -1.58 (0.22)*** -1.33 (0.27)*** -1.64 (0.25)*** -0.95 (0.16)*** -0.22 (0.13)* -0.14 (0.04)*** -0.13 (0.03)*** 0.03 (0.04) -0.00 (0.05) 0.04 (0.09) -0.09 (0.27) -0.50 (0.65) -0.10 (0.34) 0.97 (0.58)* -1.74 (0.42)*** -0.42 (0.37)

-1.63 (0.28)*** -1.08 (0.31)*** -0.62 (0.42) -1.31 (0.34)*** -1.46 (0.35)*** -0.99 (0.45)** -2.67 (0.53)*** -1.40 (0.21)*** -0.10 (0.19) -0.13 (0.05)** -0.08 (0.05)* 0.06 (0.05) 0.04 (0.07) 0.02 (0.13) 0.19 (0.36) -11.66 (0.59)*** -0.35 (0.56) 2.18 (1.11)** -10.85 (1.01)*** -0.00 (0.46)

Table 4-4. Income, and work and spending attitudes Men

Less than $12,000 US $12001 - $15,000 US $25001 - $35,000 US $35001 - $50,000 US $50001 - $75,000 US $75001 - $100,000 US $100,001 - $150,000 US $150,001 - $200,000 US $200,001 - $250,000 US Over $250,000 US

Women

Browses

E-mails

Keywords

Browses

E-mails

Keywords

-0.03

0.16

0.18

-0.29

-0.42

-0.42

(0.09)

(0.21)

(0.34)

(0.07)***

(0.09)***

(0.12)***

0.03

0.39

-0.16

-0.10

-0.13

-0.20

(0.12)

(0.25)

(0.39)

(0.10)

(0.13)

(0.17)

0.08

0.02

0.34

0.06

0.02

0.04

(0.08)

(0.15)

(0.25)

(0.06)

(0.08)

(0.12)

0.12

0.11

0.72

0.08

0.03

0.02

(0.07)*

(0.14)

(0.25)***

(0.06)

(0.07)

(0.10)

0.20

0.42

0.81

0.08

0.07

0.02

(0.07)***

(0.14)***

(0.24)***

(0.06)

(0.07)

(0.10)

0.26

0.55

0.75

0.10

0.18

0.04

(0.08)***

(0.15)***

(0.25)***

(0.06)

(0.08)**

(0.11)

0.27

0.62

0.89

0.02

0.05

0.09

(0.08)***

(0.15)***

(0.26)***

(0.09)

(0.11)

(0.15)

0.42

0.75

1.21

-0.04

-0.01

0.16

(0.13)***

(0.17)***

(0.38)***

(0.13)

(0.17)

(0.23)

0.07

0.32

1.32

-0.47

-0.53

-0.67

(0.12)

(0.20)

(0.46)***

(0.18)***

(0.21)**

(0.37)*

0.35

0.87

1.28

-0.01

0.11

0.14

(0.12)***

(0.18)***

(0.31)***

(0.15)

(0.21)

(0.20)

Only my accountant and the IRS know for sure

0.29

0.58

0.97

0.11

0.11

0.09

(0.07)***

(0.14)***

(0.23)***

(0.05)**

(0.07)

(0.09)

What, me work?

0.29

0.64

1.11

-0.01

-0.12

-0.14

(0.10)***

(0.16)***

(0.27)***

(0.07)

(0.08)

(0.11)

-0.00

0.11

0.55

-0.17

-0.10

-0.06

(0.12)

(0.22)

(0.36)

(0.06)***

(0.08)

(0.15)

0.06

0.09

-0.02

0.02

-0.01

0.02

(0.03)*

(0.05)

(0.09)

(0.03)

(0.03)

(0.04)

0.04

0.10

0.10

-0.01

0.00

0.09

(0.04)

(0.06)*

(0.12)

(0.03)

(0.04)

(0.06)

0.05

0.09

0.14

-0.01

0.02

0.10

(0.05)

(0.08)

(0.18)

(0.04)

(0.06)

(0.09)

-0.13

-0.16

-0.18

-0.00

-0.05

-0.12

(0.06)**

(0.09)*

(0.15)

(0.05)

(0.06)

(0.08)

-0.04

-0.30

-0.39

0.05

-0.00

0.10

(0.13)

(0.30)

(0.39)

(0.11)

(0.11)

(0.12)

I hate my job My work is very satisfying I can't believe I get paid to do this! If it were fun, it wouldn't be called work I don't work for wages 'Cheap' is my middle name

36

I adhere to a strict budget

0.01

-0.10

-0.02

-0.05

-0.07

0.04

(0.06) I spend all my cash like there's no tomorrow -0.09

(0.11)

(0.21)

(0.03)

(0.05)

(0.06)

-0.13

-0.31

0.02

0.05

0.02

(0.06)

(0.10)

(0.16)*

(0.04)

(0.05)

(0.07)

-0.01

0.02

0.03

-0.04

0.05

0.09

(0.03)

(0.04)

(0.08)

(0.06)

(0.06)

(0.09)

0.02

0.02

-0.03

-0.03

-0.07

-0.07

(0.07)

(0.09)

(0.18)

(0.09)

(0.12)

(0.12)

-0.19

-0.31

-0.53

-0.09

-0.03

-0.18

(0.08)**

(0.15)**

(0.21)**

(0.05)*

(0.07)

(0.09)**

I usually pick up the tab I could retire at any moment I'm living life to its credit limit

Note. The baseline income group is "$15001 - $25,000"

37

Table 4-5. Educational attainment Men

School of life Have not finished high school yet Some technical training (2 yr degree) College freshman/sophomore College junior/senior Some college 4 year degree In post-graduate program Master's degree Doctoral degree Professional degree/certification

Browses 0.03 (0.08) 0.07 (0.17) -0.04 (0.07) 0.07 (0.09) 0.10 (0.10) 0.07 (0.06) 0.05 (0.06) 0.03 (0.09) 0.02 (0.07) 0.01 (0.08) -0.01 (0.08)

E-mails 0.14 (0.13) 0.33 (0.21) 0.01 (0.12) 0.24 (0.16) 0.16 (0.15) 0.18 (0.10)* 0.23 (0.10)** 0.21 (0.13) 0.24 (0.11)** 0.24 (0.13)* 0.20 (0.12)

Note. The baseline education level is finished high school

38

Women Keywords 0.14 (0.22) -0.69 (0.67) -0.15 (0.21) 0.65 (0.28)** 0.39 (0.27) 0.23 (0.19) 0.24 (0.18) 0.32 (0.23) 0.27 (0.20) 0.34 (0.24) 0.30 (0.23)

Browses -0.12 (0.06)* -0.04 (0.10) -0.05 (0.06) -0.10 (0.07) -0.12 (0.06)* -0.07 (0.05) -0.02 (0.05) -0.14 (0.06)** -0.16 (0.05)*** -0.05 (0.08) -0.01 (0.06)

E-mails -0.11 (0.08) -0.00 (0.14) -0.02 (0.09) -0.13 (0.09) -0.16 (0.08)** -0.10 (0.07) -0.01 (0.07) -0.16 (0.08)** -0.21 (0.07)*** -0.00 (0.10) 0.01 (0.07)

Keywords 0.00 (0.11) -0.30 (0.17)* 0.09 (0.11) -0.19 (0.10)* -0.12 (0.11) -0.12 (0.08) 0.04 (0.08) -0.20 (0.10)* -0.15 (0.09)* 0.09 (0.13) 0.07 (0.09)

Table 4-6. Occupation Men

Women

Browses

E-mails

Keywords

Browses

E-mails

Keywords

0.10

0.24

0.56

-0.11

-0.15

-0.15

(0.07)

(0.13)*

(0.23)**

(0.05)**

(0.06)**

(0.08)*

0.08

0.01

0.26

-0.13

-0.23

-0.15

(0.16)

(0.25)

(0.31)

(0.05)**

(0.06)***

(0.09)*

0.18

0.27

0.79

-0.10

-0.19

-0.22

(0.09)**

(0.15)*

(0.26)***

(0.06)*

(0.07)***

(0.10)**

-0.11

-0.18

0.07

-0.06

-0.21

-0.34

(0.10)

(0.18)

(0.30)

(0.13)

(0.16)

(0.24)

0.13

0.17

0.50

-0.11

-0.14

-0.20

(0.08)*

(0.14)

(0.23)**

(0.05)**

(0.06)**

(0.08)**

Technical/ Science/ Engineering/ Research/ 0.03 Computers (0.07)

0.02

0.42

-0.14

-0.25

-0.12

(0.13)

(0.23)*

(0.06)**

(0.07)***

(0.09)

Clergy

-0.16

0.18

-9.76

-0.25

-0.12

-0.48

(0.27)

(0.45)

(0.88)***

(0.26)

(0.38)

(0.43)

0.15

0.19

0.69

-0.15

-0.18

-0.19

(0.09)*

(0.16)

(0.27)**

(0.05)***

(0.07)***

(0.09)**

-0.03

-0.26

0.04

-0.10

-0.08

-0.01

(0.10)

(0.18)

(0.33)

(0.08)

(0.11)

(0.15)

0.15

0.52

0.85

-0.10

-0.15

-0.16

(0.11)

(0.16)***

(0.27)***

(0.06)

(0.09)*

(0.12)

0.26

0.55

1.14

-0.35

-0.55

-0.56

(0.12)**

(0.20)***

(0.33)***

(0.16)**

(0.21)***

(0.24)**

0.12

0.23

0.65

-0.06

-0.05

-0.07

(0.10)

(0.16)

(0.28)**

(0.07)

(0.09)

(0.12)

Health/Medical/ Psychology/ Dental/ Nursing

0.06

0.26

0.66

-0.11

-0.15

-0.06

(0.09)

(0.14)*

(0.26)**

(0.05)**

(0.06)**

(0.09)

Veterinary/Farming/Ranching

-0.39

-0.72

-0.25

-0.02

-0.09

-0.23

(0.32)

(0.79)

(1.07)

(0.11)

(0.15)

(0.26)

Political/ Government/ Civil servant/ Social 0.02 services (0.10)

0.07

0.16

-0.18

-0.21

-0.16

(0.17)

(0.30)

(0.07)**

(0.10)**

(0.14)

Transportation

-0.11

-0.15

-0.02

-0.18

-0.30

-0.11

(0.11)

(0.20)

(0.31)

(0.10)*

(0.13)**

(0.19)

-0.34

-0.82

-0.16

0.03

-0.12

-0.39

(0.20)*

(0.31)***

(0.88)

(0.16)

(0.18)

(0.29)

Executive/Managerial Administrative/Clerical/Secretarial Financial/Accounting Manufacturing Sales/Marketing

Teacher/Educator/Professor Service/Hospitality/Food services Legal/Attorney Law enforcement/Fire fighter Artistic/Musical/Writer

Celebrity/ Personality/ Performer/ Actor

39

Entertainment/ Broadcasting/ Film Laborer/Construction Military Homemaker Self employed Retired

0.04

0.17

0.62

-0.12

-0.14

-0.33

(0.11)

(0.19)

(0.34)*

(0.10)

(0.12)

(0.18)*

-0.05

-0.15

-0.04

-0.13

0.04

-0.29

(0.10)

(0.18)

(0.30)

(0.18)

(0.18)

(0.50)

0.17

0.37

0.67

-0.11

-0.36

-0.21

(0.10)

(0.17)**

(0.30)**

(0.13)

(0.14)**

(0.25)

0.30

0.26

1.03

-0.22

-0.18

-0.23

(0.16)*

(0.49)

(0.45)**

(0.09)**

(0.15)

(0.20)

0.13

0.19

0.50

-0.09

-0.14

-0.17

(0.08)*

(0.13)

(0.23)**

(0.05)*

(0.06)**

(0.09)*

0.14

0.28

0.46

-0.20

-0.33

-0.32

(0.11)

(0.19)

(0.30)

(0.09)**

(0.14)**

(0.19)*

Note. The baseline occupation is "students"

40

Table 4-7. Heterogeneous preferences: Looks and physique Men

E-mails from segment: Dummy variable if member has photo

Women

E-mails Income up to Income above $50k $50k High school

1.59 (0.05)*** Very good looks 0.47 (0.07)*** Above average looks 0.25 (0.06)*** Less than average looks -0.39 (0.46) Bring your bag in case mine tears 0.12 (0.22) Chiseled, I work out every day! 0.15 (0.10) Toned, I keep fit 0.06 (0.05) Skinny, I could use some carbohydrates -0.02 (0.14) Voluptuous/Portly -0.51 (0.23)** Large but shapely -0.20 (0.18) Rotund -0.13 (0.42) -0.17 I look like a reflection in a fun house mirror! (0.22) Less than 90 lbs (less than 41 Kg) -0.40

1.44 (0.06)*** 0.50 (0.08)*** 0.25 (0.07)*** -0.36 (0.55) -0.05 (0.23) 0.28 (0.10)*** 0.17 (0.05)*** -0.32 (0.21) -0.78 (0.27)*** -0.49 (0.23)** -1.19 (0.82) 0.38 (0.24) -1.33

1.62 (0.08)*** 0.54 (0.11)*** 0.24 (0.09)*** -1.24 (1.10) -0.03 (0.32) 0.30 (0.14)** 0.07 (0.08) -0.35 (0.29) -0.68 (0.34)** -0.08 (0.25) -0.52 (0.60) -0.06 (0.36) -1.30

College 1.48 (0.05)*** 0.36 (0.07)*** 0.18 (0.06)*** -0.55 (0.79) 0.05 (0.20) 0.09 (0.10) 0.14 (0.05)*** -0.46 (0.19)** -0.40 (0.22)* -0.14 (0.15) -0.36 (0.61) 0.16 (0.22) 0.16

E-mails Income up to Income above $50k $50k High school 1.69 (0.03)*** 0.20 (0.05)*** 0.12 (0.04)*** -0.07 (0.20) 0.07 (0.20) -0.02 (0.10) 0.07 (0.04)* -0.20 (0.11)* -0.32 (0.07)*** -0.39 (0.11)*** -1.09 (0.44)** 0.11 (0.15) -0.32

1.57 (0.03)*** 0.35 (0.04)*** 0.17 (0.04)*** -0.17 (0.19) 0.19 (0.18) 0.16 (0.08)** 0.16 (0.03)*** 0.00 (0.10) -0.41 (0.08)*** -0.57 (0.10)*** -0.14 (0.29) -0.03 (0.12) -0.70

1.77 (0.05)*** 0.32 (0.07)*** 0.28 (0.06)*** -0.45 (0.44) 0.10 (0.26) 0.15 (0.14) 0.11 (0.05)** -0.04 (0.17) -0.36 (0.16)** -0.36 (0.20)* -1.39 (0.70)** 0.20 (0.20) -0.50

College 1.48 (0.03)*** 0.43 (0.04)*** 0.23 (0.04)*** -0.17 (0.23) 0.31 (0.17)* 0.12 (0.08) 0.14 (0.03)*** 0.05 (0.09) -0.45 (0.07)*** -0.56 (0.10)*** -0.10 (0.32) -0.06 (0.13) -0.89

90 lbs - 100 lbs ( 41 Kg - 45 Kg) 111 lbs - 120 lbs ( 51 Kg - 55Kg) 121 lbs - 130 lbs ( 56 Kg - 59Kg) 131 lbs - 140 lbs ( 60 Kg - 64Kg) 141 lbs - 150 lbs ( 65 Kg - 68Kg) 151 lbs - 160 lbs ( 69 Kg - 73Kg) 161 lbs - 170 lbs ( 74 Kg - 77Kg) 171 lbs - 180 lbs ( 78 Kg - 82Kg) 181 lbs - 190 lbs ( 83 Kg - 86Kg) 191 lbs - 200 lbs ( 87 Kg - 91Kg) 201 lbs - 210 lbs ( 92 Kg - 95Kg) 211 lbs - 220 lbs ( 96 Kg - 100 Kg) 221 lbs - 230 lbs (101 Kg - 105 Kg) 231 lbs - 240 lbs (106 Kg - 109 Kg) 241 lbs - 250 lbs (110 Kg - 114 Kg) Over 250 lbs (over 114 Kg)

(0.79) -0.17 (0.75) -0.71 (0.68) -0.37 (0.67) -0.24 (0.63) -0.42 (0.62) 0.04 (0.62) 0.03 (0.62) 0.04 (0.62) 0.20 (0.62) 0.12 (0.62) 0.07 (0.62) -0.03 (0.62) 0.07 (0.63) 0.14 (0.64) 0.07 (0.65) 0.15

(0.85) -0.84 (0.88) -1.36 (0.83) -1.24 (0.74)* -1.42 (0.66)** -1.66 (0.65)** -1.22 (0.63)* -1.20 (0.63)* -1.19 (0.63)* -1.05 (0.63)* -1.09 (0.63)* -1.19 (0.64)* -1.18 (0.64)* -1.08 (0.64)* -1.19 (0.65)* -1.30 (0.66)** -1.19

(0.87) -0.24 (0.75) -2.64 (1.10)** -1.32 (0.63)** -1.18 (0.55)** -1.55 (0.54)*** -1.25 (0.54)** -1.18 (0.54)** -1.13 (0.53)** -1.05 (0.54)** -1.15 (0.54)** -1.07 (0.54)** -1.33 (0.56)** -1.58 (0.57)*** -1.07 (0.59)* -0.91 (0.61) -0.98

42

(0.97) -0.06 (0.99) -1.44 (0.94) -0.52 (0.90) -0.93 (0.87) -1.10 (0.86) -0.80 (0.85) -0.80 (0.85) -0.78 (0.85) -0.73 (0.85) -0.81 (0.85) -0.82 (0.85) -0.83 (0.85) -0.81 (0.86) -0.78 (0.86) -0.94 (0.87) -0.92

(0.27) 0.02 (0.12) -0.05 (0.06) -0.00 (0.06) -0.10 (0.06) -0.34 (0.08)*** -0.46 (0.09)*** -0.49 (0.10)*** -0.71 (0.12)*** -0.87 (0.16)*** -1.16 (0.21)*** -1.18 (0.26)*** -0.41 (0.37) -1.41 (0.41)*** -2.22 (0.54)*** -1.27 (0.35)*** -1.29

(0.30)** -0.25 (0.10)** -0.05 (0.05) -0.18 (0.05)*** -0.31 (0.05)*** -0.61 (0.07)*** -0.78 (0.08)*** -1.02 (0.09)*** -1.16 (0.10)*** -1.32 (0.18)*** -1.55 (0.14)*** -1.25 (0.25)*** -0.75 (0.35)** -1.62 (0.28)*** -1.66 (0.27)*** -1.39 (0.30)*** -1.73

(0.37) -0.22 (0.16) -0.04 (0.09) -0.21 (0.09)** -0.34 (0.10)*** -0.61 (0.11)*** -0.80 (0.14)*** -0.98 (0.16)*** -0.99 (0.19)*** -1.17 (0.25)*** -1.94 (0.41)*** -2.11 (0.73)*** -0.58 (0.37) -2.33 (1.00)** -0.29 (0.47) -13.38 (0.37)*** -13.40

(0.26)*** -0.21 (0.11)* -0.04 (0.05) -0.20 (0.05)*** -0.33 (0.05)*** -0.61 (0.07)*** -0.82 (0.08)*** -1.06 (0.09)*** -1.23 (0.11)*** -1.26 (0.16)*** -1.33 (0.21)*** -1.20 (0.30)*** -0.90 (0.30)*** -1.38 (0.28)*** -1.74 (0.40)*** -1.32 (0.35)*** -1.58

I broke the truck stop scale Under 5 (under 152 cm) 5 - 5 2 (152 cm - 158 cm) 5 3- 5 4 (159 cm - 163 cm) 5 7- 5 8 (169 cm - 173 cm) 5 9- 5 10

(174 cm - 178 cm)

5 11- 6 (179 cm - 183 cm) 6 1- 6 2 (184 cm - 188 cm) 6 3- 6 4 (189 cm - 193 cm) 6 5- 6 6 (194 cm - 198 cm) 6 7- 6 9 (199 cm - 206 cm) 6 10 - 7(207 cm - 213 cm) Over 7 (over 213 cm)

(0.68) -1.17 (0.95) 1.07 (0.44)** 0.38 (0.46) -0.40 (0.31) 0.00 (0.13) 0.10 (0.12) 0.15 (0.12) 0.08 (0.13) 0.15 (0.15) 0.18 (0.20) -0.14 (0.35) 2.66 (0.44)*** -0.65 (1.09)

(0.70)* -3.06 (1.23)** 1.88 (0.47)*** -0.25 (0.76) -0.80 (0.41)** 0.16 (0.14) 0.31 (0.13)** 0.42 (0.13)*** 0.44 (0.14)*** 0.48 (0.16)*** 0.33 (0.24) -0.28 (0.51) 2.70 (0.47)*** 1.34 (0.65)**

(0.64) -1.59 (0.97) 0.32 (0.51) -0.15 (0.76) -0.06 (0.41) -0.13 (0.18) 0.00 (0.17) -0.00 (0.18) -0.04 (0.19) 0.05 (0.22) -0.40 (0.38) -0.47 (0.72) -12.96 (0.88)*** -14.23 (1.35)***

43

(0.89) -1.18 (1.01) 1.16 (0.42)*** -0.94 (0.46)** -0.63 (0.35)* 0.09 (0.13) 0.28 (0.12)** 0.42 (0.13)*** 0.39 (0.13)*** 0.43 (0.15)*** 0.47 (0.20)** 0.11 (0.40) 2.20 (0.44)*** 0.51 (0.74)

(0.37)*** -0.67 (0.18)*** -0.04 (0.15) 0.01 (0.05) -0.02 (0.04) 0.02 (0.05) -0.00 (0.06) -0.03 (0.13) 0.18 (0.37) -0.61 (0.59) 0.11 (0.63) 1.79 (0.67)*** -11.06 (0.86)*** -0.68 (0.69)

(0.38)*** -1.08 (0.18)*** -0.30 (0.15)** -0.14 (0.05)*** -0.12 (0.04)*** 0.05 (0.04) -0.01 (0.05) 0.05 (0.10) -0.35 (0.27) -0.55 (0.66) -0.76 (0.54) 0.61 (1.02) -10.63 (0.91)*** -0.57 (0.44)

(0.50)*** -0.83 (0.33)** -0.25 (0.27) -0.14 (0.08)* -0.05 (0.06) 0.09 (0.07) -0.04 (0.10) -0.33 (0.21) -0.85 (1.02) -0.84 (0.95) 1.06 (0.90) -10.86 (1.17)*** 0.79 (0.69) -0.80 (0.68)

(0.48)*** -1.08 (0.21)*** -0.32 (0.13)** -0.16 (0.05)*** -0.12 (0.04)*** 0.05 (0.04) 0.04 (0.05) -0.00 (0.10) -0.21 (0.30) -0.83 (0.57) -0.22 (0.42) -10.51 (0.67)*** -10.82 (0.89)*** -0.55 (0.47)

Table 4-8. Heterogeneous preferences: Income, and work and spending attitudes Men

E-mails from segment: Less than $12,000 US

Women

E-mails Income up to Income above $50k $50k High school

0.16 (0.26) $12001 - $15,000 US 0.21 (0.29) $25001 - $35,000 US 0.02 (0.18) $35001 - $50,000 US 0.23 (0.18) $50001 - $75,000 US 0.41 (0.17)** $75001 - $100,000 US 0.44 (0.18)** $100,001 - $150,000 US 0.44 (0.19)** $150,001 - $200,000 US 0.66 (0.22)*** $200,001 - $250,000 US 0.01 (0.24) Over $250,000 US 0.74 (0.23)*** Only my accountant and the IRS know for sure 0.42 (0.17)** What, me work? 0.50

-0.19 (0.41) 0.37 (0.48) -0.07 (0.31) -0.14 (0.29) 0.53 (0.28)* 0.66 (0.28)** 0.80 (0.29)*** 0.90 (0.30)*** 0.41 (0.32) 1.30 (0.32)*** 0.74 (0.28)*** 0.66

0.06 (0.35) 0.46 (0.35) -0.07 (0.26) 0.04 (0.26) 0.29 (0.25) 0.15 (0.26) 0.35 (0.27) 0.57 (0.33)* 0.10 (0.37) 0.75 (0.33)** 0.39 (0.24) 0.59

44

College -0.47 (0.36) 0.27 (0.37) -0.34 (0.25) -0.20 (0.23) 0.22 (0.22) 0.43 (0.22)* 0.49 (0.22)** 0.65 (0.24)*** 0.24 (0.27) 0.89 (0.24)*** 0.46 (0.21)** 0.41

E-mails Income up to Income above $50k $50k High school -0.38 (0.11)*** -0.14 (0.16) -0.02 (0.09) -0.04 (0.09) -0.05 (0.09) 0.03 (0.10) -0.28 (0.15)* 0.02 (0.22) -0.57 (0.38) -0.03 (0.20) 0.02 (0.08) -0.14

-0.44 (0.11)*** -0.16 (0.16) 0.08 (0.09) 0.12 (0.08) 0.15 (0.09)* 0.24 (0.10)** 0.13 (0.12) 0.15 (0.17) -0.56 (0.29)* -0.07 (0.24) 0.20 (0.08)** -0.07

-0.17 (0.17) 0.32 (0.22) 0.30 (0.15)** 0.31 (0.14)** 0.27 (0.15)* 0.12 (0.18) 0.18 (0.22) -0.25 (0.48) -0.18 (0.56) -0.09 (0.42) 0.35 (0.13)*** 0.12

College -0.42 (0.11)*** -0.18 (0.15) 0.04 (0.09) 0.09 (0.08) 0.12 (0.08) 0.24 (0.09)*** 0.17 (0.12) 0.28 (0.18) -0.61 (0.27)** -0.01 (0.21) 0.19 (0.08)** -0.03

I hate my job My work is very satisfying I can't believe I get paid to do this! If it were fun, it wouldn't be called work I don't work for wages 'Cheap' is my middle name I adhere to a strict budget I spend all my cash like there's no tomorrow I usually pick up the tab I could retire at any moment I'm living life to its credit limit

(0.21)** 0.25 (0.25) 0.14 (0.06)** 0.02 (0.07) -0.00 (0.10) -0.23 (0.11)** -0.29 (0.30) -0.11 (0.11) -0.26 (0.12)** 0.01 (0.05) -0.13 (0.14) -0.63 (0.18)***

(0.30)** 0.00 (0.33) 0.19 (0.07)*** 0.20 (0.08)** 0.26 (0.11)** -0.01 (0.12) -0.39 (0.46) -0.08 (0.14) -0.15 (0.14) 0.00 (0.06) 0.02 (0.12) -0.30 (0.20)

(0.28)** 0.16 (0.32) 0.10 (0.09) 0.08 (0.10) 0.11 (0.14) -0.27 (0.17) -0.13 (0.42) -0.11 (0.20) -0.09 (0.16) 0.06 (0.08) 0.00 (0.19) -0.24 (0.26)

Note. The baseline income group is "$15001 - $25,000"

45

(0.25) 0.07 (0.30) 0.08 (0.07) 0.06 (0.07) 0.10 (0.11) -0.16 (0.10) -0.41 (0.34) -0.24 (0.13)* -0.20 (0.12) 0.01 (0.05) 0.05 (0.10) -0.44 (0.20)**

(0.09) -0.07 (0.12) 0.03 (0.04) 0.01 (0.05) 0.07 (0.08) -0.20 (0.08)*** -0.01 (0.14) -0.17 (0.06)*** 0.06 (0.06) 0.02 (0.08) -0.23 (0.17) -0.12 (0.08)

(0.10) -0.15 (0.10) 0.03 (0.04) 0.01 (0.05) 0.11 (0.07) -0.06 (0.07) -0.13 (0.14) -0.11 (0.05)** 0.04 (0.06) 0.06 (0.07) -0.26 (0.12)** 0.03 (0.07)

(0.17) -0.34 (0.18)* -0.01 (0.06) 0.00 (0.08) 0.11 (0.11) -0.17 (0.12) 0.11 (0.24) -0.15 (0.09)* -0.12 (0.10) 0.12 (0.12) -0.46 (0.26)* -0.08 (0.11)

(0.09) -0.13 (0.10) 0.04 (0.03) 0.04 (0.05) 0.10 (0.07) -0.04 (0.07) -0.02 (0.14) -0.09 (0.05)* -0.01 (0.06) 0.05 (0.07) -0.26 (0.13)** 0.02 (0.07)

Table 4-9. Heterogeneous preferences: Educational attainment Men

E-mails from segment: School of life Have not finished high school yet Some technical training (2 yr degree) College freshman/sophomore College junior/senior Some college 4 year degree In post-graduate program Master's degree Doctoral degree Professional degree/certification

Women

E-mails Income up to Income above $50k $50k High school 0.05 (0.16) -0.18 (0.38) 0.11 (0.13) 0.32 (0.19)* 0.25 (0.16) 0.21 (0.12)* 0.24 (0.12)** 0.15 (0.17) 0.17 (0.12) 0.07 (0.16) 0.08 (0.15)

0.28 (0.19) 0.43 (0.34) 0.01 (0.19) 0.29 (0.24) 0.14 (0.25) 0.31 (0.16)* 0.36 (0.16)** 0.27 (0.20) 0.39 (0.16)** 0.36 (0.18)** 0.22 (0.18)

-0.11 (0.20) 0.28 (0.49) -0.31 (0.19) -0.07 (0.22) -0.40 (0.24) -0.19 (0.15) -0.36 (0.15)** -0.67 (0.23)*** -0.54 (0.18)*** -0.47 (0.24)* -0.64 (0.22)***

Note. The baseline education level is finished high school

46

College 0.36 (0.17)** 0.67 (0.26)** 0.12 (0.16) 0.22 (0.24) 0.28 (0.22) 0.27 (0.13)** 0.46 (0.13)*** 0.51 (0.17)*** 0.61 (0.13)*** 0.63 (0.15)*** 0.47 (0.15)***

E-mails Income up to Income above $50k $50k High school -0.16 (0.10) -0.11 (0.14) -0.03 (0.10) -0.07 (0.10) -0.15 (0.10) -0.06 (0.08) -0.04 (0.08) -0.25 (0.11)** -0.20 (0.09)** -0.08 (0.13) -0.04 (0.09)

-0.06 (0.10) 0.17 (0.19) 0.06 (0.10) -0.08 (0.11) -0.08 (0.10) -0.03 (0.08) 0.03 (0.08) -0.10 (0.09) -0.16 (0.08)** 0.02 (0.11) 0.07 (0.09)

-0.58 (0.16)*** -0.16 (0.20) -0.27 (0.15)* -0.30 (0.15)* -0.46 (0.14)*** -0.20 (0.11)* -0.28 (0.11)** -0.51 (0.16)*** -0.45 (0.12)*** -0.29 (0.18) -0.11 (0.13)

College -0.02 (0.10) 0.09 (0.17) 0.05 (0.09) -0.08 (0.11) -0.05 (0.09) -0.05 (0.07) 0.07 (0.07) -0.01 (0.09) -0.03 (0.08) 0.16 (0.10) 0.06 (0.08)

Table 5-1. Age breakdown of matched men and women in the sample Age of men Age of women 20-29 30-39 40-49 50-59 60-69 70-79 Total Correlation of ages of match partners:

20-29 133 29 7 0 0 0 169

30-39 198 447 56 3 0 0 704

0.8165*** N=1837, pval=0.00

40-49 22 223 371 56 0 1 673

50-59 0 4 130 131 2 1 268

60-69 0 0 0 14 6 0 20

70-79 0 0 0 0 2 1 3

Total 353 703 564 204 10 3 1837

Table 5-2. Rank Correlations of Characteristics of Matched Members Age Group Characteristics # e-mails received by M/week # e-mails received by W/week Income of M Income of W Yrs of education of M Yrs of education of W BMI of M BMI of W Beauty rating of M Beauty rating of W Income of M BMI of W Income of M Beauty rating of W Income of W BMI of M Income of W Beauty rating of M

20-29 30-39 40-49 50-59 0.2772*** 0.2368*** 0.2211*** 0.2436*** N=354, pval=0.00 N=704, pval=0.00 N=564, pval=0.00 N=204, pval=0.00 0.3017*** 0.0953 0.1364* 0.3474* N=117, pval=0.00 N=220, pval=0.16 N=144, pval=0.10 N=50, pval=0.06 0.2117*** 0.1804*** 0.165*** 0.0651 N=354, pval=0.00 N=704, pval=0.00 N=564, pval=0.00 N=204, pval=0.36 0.2301*** 0.1945*** 0.1756*** 0.1093 N=353, pval=0.00 N=700, pval=0.00 N=563, pval=0.00 N=203, pval=0.12 0.0019 0.2216*** 0.0396 -0.0398 N=73, pval=0.99 N=187, pval=0.00 N=196, pval=0.58 N=64, pval=0.75 -0.1184 -0.2407*** -0.3043*** -0.1259 N=230, pval=0.07* N=471, pval=0.00 N=386, pval=0.00 N=107, pval=0.20 0.0072 0.1946*** 0.2420*** 0.0765 N=80, pval=0.95 N=192, pval=0.01 N=194, pval=0.00 N=57, pval=0.57 -0.148** 0.0055 -0.0231 0.2067 N=183, pval=0.05 N=333, pval=0.92 N=213, pval=0.73 N=59, pval=0.12 0.1103 0.0257 0.1383 0.0829 N=95, pval=0.29 N=208, pval=0.71 N=132, pval=0.11 N=38, pval=0.62

Spearman rank correlations are reported between a woman's characteristics, and the average characteristic of her matches. "N" denotes sample size, "pval" is the p-value of the test for the independence of the two variables. *** 1% significance, ** 5% significance, * 10% significance

48

Figures Figure 4-1

Effect on Emails Received by Reason for Joining Site

Men

Hoping to start long term relationship Seeking an occasional lover/casual rel. I’d like to make new friends. Nothing serious. Looking for a penpal only Hunting for a roommate Looking for a travel partner Just looking/curious Scouting for others to do things with A friend put me up to this A higher power brought me here Looking for someone to share my lifestyle with Scouting for swinging couples

Women

Hoping to start long term relationship Seeking an occasional lover/casual rel. I’d like to make new friends. Nothing serious. Looking for a penpal only Hunting for a roommate Looking for a travel partner Just looking/curious Scouting for others to do things with A friend put me up to this A higher power brought me here Looking for someone to share my lifestyle with Scouting for swinging couples

100

50

0 50 % Difference

Note: The boxes represent 50% confidence intervals, the whiskers extend these to 95% confidence intervals.

100

Figure 4-2

Emails Received and Looks

Men

Dummy Variable if member has photo Very good looks Above average looks Less than average looks Bring your bag in case mine tears Chiseled, I work out every day! Toned, I keep fit Skinny, I could use some carbohydrates Voluptuous/Portly Large but shapely Rotund

Women

Dummy Variable if member has photo Very good looks Above average looks Less than average looks Bring your bag in case mine tears Chiseled, I work out every day! Toned, I keep fit Skinny, I could use some carbohydrates Voluptuous/Portly Large but shapely Rotund

100

50

0

100 200 % Difference

300

50

0

50

Men

Note: Weight in lbs

51 250

250

100

% Difference

Figure 4-3

Effect on Mails Received by Weight Women

52 6’5- 6’6

6’3- 6’4

6’1- 6’2

5’11- 6’

5’9- 5’10

50

0

50

100

Men

5’7 - 5’8

5’5- 5’6

5’3- 5’4

5’ - 5’2

6’7- 6’9

6’5 - 6’6

6’3 - 6’4

6’1- 6’2

5’11- 6’

5’9- 5’10

5’7- 5’8

5’5 - 5’6

5’3 - 5’4

100

% Difference

Figure 4-4

Effect on Mails Received by Height Women

Note: Income brackets in $1,000

53 >250

200−250

150−200

100−150

75−100

50−75

0

100

200

300

Men

35−50

25−35

15−25

12−15

250

200−250

150−200

100−150

75−100

50−75

35−50

25−35

15−25

12−15