August 24th, 2015

Internet Dating


Using internet dating sites to find partners is now completely mainstream. Panelists will consider the sociology of how this system works. Two panelists will use their access to data from internet dating sites revealing what groups are on the site and who contacts and responds to whom. One panelist will show how race affects who is contacted or responded to among those seeking same-sex or other-sex partners. Another panelist will use new methods to uncover what people are looking for in partners, and how market constraints shape people’s search and contact strategies. Another panelist will report on a survey ascertaining which groups are more likely to have used the internet to find partners, and will put internet dating into a broader historical perspective focusing on macrosocial trends affecting mate choice and family formation.


PAULA ENGLAND: Please welcome Michael Rosenfeld. [APPLAUSE]
MICHAEL ROSENFELD: All right. I want to thank president Paula England for organizing and I want to thank the National Science Foundation for generously supporting the How Couples Meet and Stay
Together project, whose data I'm going to be referring to a little later in this talk. You may have heard about this Vanity Fair story. I want to start with this. The title of the Vanity Fair story,
which is out now, is, "Tinder and the Dawn of the Dating Apocalypse," okay. And the title sounds ominous, right?
I want to suggest that neither Tinder nor hookup culture are going to lead to the end of civilization. That's the first thing. I feel like that's a safe statement. If I'm wrong, come see me later. I
also don't think even that Tinder and hookup culture are going to necessarily lead to the end of dating as we used to know it, right, which is really what this article is about. But I do believe that
the cellphone and the Internet, because of their ubiquity, because of their novelty, because young people love them, are kind of the natural repositories for all of our fears, founded and unfounded.
So the story is about 20-some things in New York who are using Tinder and hooking up. Now, if you know a little bit about family demography, you know that the median age at first marriage in the
United States has been climbing for about 50 years, and it's now in the late 20s. So when you're interviewing people who are in their 20s, a lot of those people are not yet at the life stage when
they're thinking of settling down, right? Hooking up to date does not preclude settling down at some time in the future, which is good. You can think of Tinder as kind of the singles bar of today,
And Tinder, for those who you who don't know, it's a GPS-based phone app. It tells you who, within the radius that you specify, also has the Tinder turned on, and you can see their picture and a
pretty thin profile and you sort of - you flip in one direction if you want to meet them and in the other direction if you don't. And if they swipe on you and you swipe on them, then you can
communicate with them. So Tinder is like the singles bar, but with a couple of important differences. Difference number one is efficiency. There are more people on Tinder than at the singles bar. You
go to the singles bar, there might be 12 or 15 people there. On Tinder, you're talking about hundreds or thousands. The second difference is safety, okay? So the only people who can communicate with
you on Tinder are the people you want to communicate with. That's not true at the singles bar. You sit down at the bar, someone's going to sidle right up next to you, whether you want to talk to them
or not. You know, you're getting the whiskey breath, okay?
So the safety is a big advantage. Now, what bothers me about this photo is, you know, it seems clear, this is the photo from the Vanity Fair story. It seems clearly staged to show people at the bar,
young attractive people who all staring down at their cellphones, okay, they're not talking to each other. And this is part of a critique of cellphone and Internet use; that the cellphone is going to
interfere with and substitute for face-to-face interaction. But I don't actually think that this critique is fair. The Tinder uses that I've interviewed, when they're out on a date with somebody that
they've matched with on Tinder, they're not checking their phones. At least that's what they tell me. They're not checking their phones because they spent a good amount of effort getting together on
this date with somebody who that they have high expectations is going to sleep with them that night. And that person has their full attention, okay? So they're not checking their phone. They're
checking the person. And even if they don't like that person that much, they're still planning to sleep with them; they're listening to everything this person says. So in that sense, Tinder is
facilitating, rather than substituting, for face-to-face interaction.
Now, historically, social elites have always stigmatized and tried to regulate whatever media or popular culture is really popular with the youth or with working class people, okay, and this is part
of what sociologists refer to as "the moral panic," okay. There's a lot of good sociology on the history of moral panics; in the 1950s, there was the moral panic around comic books. Okay, comic books
were going to corrupt the children and instill ideas about communism and so on. I don't know if you remember the moral panics around Gangsta rap in the 1980s, right, it's going to teach everyone
about violence. And recently we've had moral panics about sexual profligacy of our young people, so there was a moral panic around sex bracelets in middle school, and rainbow parties. And there's
this terrific book by Joel Best and Kathleen Bogle, where they explain that there never were any rainbow parties or sex bracelets, that these were creations of a kind of hyperventilating media. Now,
just because there are examples of moral panics from the past doesn't mean that today's moral panic isn't - might not be based on reality, right? Just because something in the past wasn't true,
doesn't mean today's also isn't. It requires empirical study. So before this talk is over, I'm promising you some empirical evidence on the social impacts of Internet dating. So I'm going to get to
It's also worth noting that moral panic is not value-neutral. The moral panic assumes that whatever was going on in the past was better, that - and the preference of the past over the present is a
fundamentally conservative idea. So it's important to realize that there's an implicit political element to the moral panic.
So, briefly, we - there are critiques of the cellphone and the Internet that sort of allege that the cellphone leads to a limited attention span, displaces face-to-face interaction. Technology is
supposed to make us more superficial; that's according to some critiques. But in my view, what makes us superficial is us, right? We, you, me, maybe you more than me, but all of us, right - there's
an element of superficiality in how we see each other. And it's not clear to me that that's enhanced by the technology. The other thing that I want to talk a little bit about is this theory of choice
overload. In a lot of the more academic literature about online dating you see choice overload cited as an explanation for why the big choice set available online might undermine relationships. And
this theory works in several ways. One way it works is that, you know, if there's so many people that you could meet online, how are you going to pick one, how are you going to make sense of all that
variety? The idea is that the big choice set is demotivating. So that's one aspect of it. And another aspect of it is, maybe the fact that there's thousands or millions of people online doing the
online dating would diminish your resolve to stay with your current partner. You might think, well, there's so many fish in the sea, like, why would I stay married to this person when I could meet
somebody new, right? That's the theory.
Now, there's this - it's based in part by this classic experiment by Iyengar and Lepper, which took place in a grocery store, which I'm going to briefly describe. They sat in a grocery store with a
table and they had six jars of different jams on one day, and then another day they came back and they had a selection of 24 varieties of jam. And what they found is that on the day they were there
with only six jams, people bought more jam. So their interpretation was that more choice is demotivating, less is better. And the question is, is this theory really applicable to dating and online
dating? And I'm going to argue that it's not, for several reasons. One reason is that, well, you need sex, companionship and romance more than you need jam, okay? [LAUGHTER] And I'm not trying to put
down the aficionados of the fig preserves or whatever, but you can do without it for a long time. The second thing is that in order to make this experiment work, Iyengar and Lepper actually took out
of the choice set the jams that they thought people would like the most - that is, strawberry and raspberry, because those are people's most preferred jams So but the whole point of choice is that in
a big choice set, you're more likely to find the thing you're looking for, if you're looking for something specific. And I think in dating, people are usually looking for something specific, okay?
And the third aspect of this is that one of the indelible aspects of business in the Internet era is that the businesses with the big choice set, like Amazon, totally dominate the businesses with the
small choice set like your corner bookstore. And, in fact, the store that Iyengar and Lepper did the experiment in had 300 jams in stock. And if they thought they could sell more jam by reducing
stock to six, they probably would have done it.
Okay. So I promised data. Let me fulfill the promise a little bit. One of the - one of the ideas of choice overload is that in the Internet era, we might see - we might see more breakup. And if you
look at divorce data to the extent that we have consistent divorce data over time - this is mostly from the NSFG, National Survey of Family Growth, what you see is that we had a big rise in divorce
in the '60s and '70s, and then a kind of plateau in the '80s and '90s. This is not consistent with the idea that the modern technology era is totally undermining relationships, okay, because the
divorce rate was rising in the United States for 100 years, more or less, up to 1980.
Briefly, if you look at the rate at which people are victimized in intimate partner violence, you find that intimate partner violence has declined by something like 70 percent since 1995. Now, it's
true that almost all measures of crime have sharply declined since 1995, so intimate partner violence is not unique. Still, and nonetheless, some of the critics of the modern hookup culture, the lack
of commitment in relationships, have argued that people are less safe in this newer world and I don't think that that's true. The data, the evidence don't support that.
So now this is a graph that comes from the How Couples Meet and Stay Together project. This is a graph of transitions to marriage, and the red line is people who met offline, and that data goes back a
long way. And the blue line is people who met online. So here's something interesting. The people who meet online, and this is heterosexual couples only, because the legal climate and context for
same-sex marriage has changed dramatically over this era, and for the better, but that's a story for another day. The people who meet online actually transition to marriage faster. And this is
surprising if you think about the stigma that's associated with meeting online, and the critique that online relationships tend to be more superficial. No. Well, in fact, Tinder is only a small
corner of who meets online, and there are these Internet dating Websites that have not only millions of users, but thick descriptions and all kinds of interesting profiles that allow you to take
advantage of the broad choice set, okay, because if you're looking for something specific, let's say your idea of the perfect partner is the Hindi speaking, mountain climbing, vegetarian former
Catholic, OK, let's just say. You could be sitting next to a person who fits that description right now, but you wouldn't know it, if you don't know that person. But online, you can actually find
that person, right. So if you're looking for something specific, the Internet is really advantageous.
If we divide the people who met online into the people who met through online dating, which is the green triangle, and the people who met online but not through online dating, which is the X's, we see
that it's really the online dating that drives the greater transition to marriage. And this is because it's the online dating Websites - the Match, the eHarmony, the OkCupid, where people can take
advantage of the broad choice set and search in it and find the particular characteristics they're looking for. A lot of the other ways of meeting online are kind of haphazard, more like meeting in
real life - you know, people who meet through chat, who meet through gaming, and so on.
So this is a figure from a 2012 paper of mine that I wrote with Reuben J. Thomas. And this is a figure that describes the history of how people have met. And on the left, you have the heterosexual
couples and on the right you have the same-sex couples. And what you see is, there's a sharp rise in meeting online, that's sort of this red line, that starts around 1995, which is the beginning of
the graphical Web, okay? Meeting online starts to displace all the other ways of meeting. It hasn't caught up yet with meeting through friends, which is still number one for heterosexuals, leading to
the corollary, be nice to your friends, right? It's a time-tested truism. But you see this sharp rise in meeting online, starting in 1995, getting up to about 23 percent. About 23 percent of
recently-formed heterosexual unions met online. Now, on the right side, you see the similar graph for same-sex couples, but with a notable difference. For same-sex couples, meeting online constitutes
about 67 percent of all the same-sex couples in the United States. A dominant portion of the same-sex couples recently formed in the United States have met online. Why is there such a big difference?
Thomas and I argued that the reason for this difference is that people who are looking for a same-sex partner are usually in what we refer to as a "thin dating market." It's hard to know who, in a big
crowd of people, the other gay and lesbian people are. Even in Chicago or Los Angeles or San Francisco, most people are not gay, and about two percent of all couples in the United States are same-sex
couples. So it's hard to identify potential candidates. And this is where the advantage of large choice set and search really helps people, right? So there's a tremendous benefit, especially for some
sectors, in Internet dating in terms of leveraging the large choice set and the ability to search across it. And then as for the potential, the reputed negative social impacts of the Internet, these
prove to be mostly in the imaginary category. And I'm going to leave it at that. Thank you. [APPLAUSE]
ELIZABETH BRUCH: Hi, I also would like to thank Paula for organizing this session, and I would like to thank all of you for coming out. And last, but certainly not least, I'd like to thank the online
dating company that collected these data I'm going to talk about today. And before I begin, I also wanted to add that this is joint work with Mark Newman at the University of Michigan. So I'm here
today to talk to you about the structure of online relationship markets, and I want to start by thinking about the idea of what is a relationship market in the first place. The idea that
relationships are pursued within some kind of market is deeply ingrained in both academic and popular culture. The everyday language we use to talk about mate seeking implies scarcity and
competition, and a hierarchy of desirability. Think about phrases like, "she's a hot commodity," or, "he's out of your league." But these layman's views of markets are under-theorized and empirically
ill-defined. In social science we have classic works in economics and mathematics; they give us models of how people's preferences aggregate into observed matches. But these studies treat markets as
abstract undifferentiated entities. They don't tell us anything about how markets might be divided horizontally into submarkets, or vertically into leagues; or how market structure might vary with
the size and composition of the pool of potential mates.
The goal of this study that I'm talking about today is to describe the empirical structure of relationship markets as evidenced by messages exchanged on a major online dating site. We know very little
about the broad structure of romantic markets, or how the structure varies from place to place. And part of the problem is the difficulty of collecting data on a scale necessary for a truly
comprehensive study. Ideally, this would be millions of individuals or more. But the advent and subsequent rise of online dating, however, creates students to study relationship markets on an
unprecedented scale. The messages that are exchanged by people online can be represented as a network, and then they can be analyzed using powerful methods of graph theory and network analysis. I'm
going to use these methods to quantify vertical and horizontal stratification and romantic markets, and how they differ across two cities that vary in their relative numbers of men and women; New
York City and Seattle. Specifically, I identify distinct submarkets within each geographic area, and determined the extent to which those submarkets overlap. I'm also going to devote some effort to
conceptualizing and measuring leagues. The idea that someone might be "out of one's league" is deeply rooted in popular conceptions of dating and marriage, but has not been subject to scientific
So I want to start by giving you a framework with which to start thinking about romantic market structure. As I see it, there's two fundamental dimensions along which markets are organized. The first
one I'm going to call "market position," this is really submarkets. And the second I'm going to call "market desirability," or popularity. In the online dating context, a submarket is really a
cluster of individuals within which there's a high density of messages taking place; significantly more messages within submarkets than between them. Now, there's some subtly to how we think about
whether or not two users are in contact. In particular, let's distinguish between message initiation and message reply. Message initiation implies that two users are in contact; if one sends a
message to the other, regardless of whether that message receives a reply. This condition is important, because the vast majority of messages online don't receive replies, over 80 percent. Thus, that
network is going to primarily represent the interests and aspirations of the sender, rather than being an indicator of actual contact. Message reply, on the other hand, is a more practical view of
the market. Given who has written to me, who am I interested in being with. The network is more representative of mutual interest, and I'm going to present results focusing on that reply market.
Now, some markets identify who messages whom, but they don't actually identify who is most highly sought after. It's likely some people are more desirable than others, and those desirable people may
be clustered at the same or different parts of the submarket, so the other piece of the puzzle is figuring out who is desirable and where are they located in the market structure.
Now, I said I would tell you about leagues. And I want to give you some - one thing to think about as we go through the next set of results. For leagues to exist, we need two things to be true. One is
that we need there to be a hierarchy of desirability, in such that some people are more desirable than others. And second, what we need is a limit to reach; in other words, desirable people might be
sought after, but they're unavailable to less desirable others. And I'll show you some results on that in just a bit.
What I want to do first is tell you a little bit about the data and, more importantly, the measures that I'm going to be using to capture market structure in terms of market position and market
desirability. Let's start with market position. So there's actually a lot of different ways you can measure market position, but in the interest of time, I'm just going to focus on one, which is to
think about submarkets as being discreet. So in other words, they're not a continuous distribution of people throughout the market space, they're sort of clumps of people, as shown in this slide,
here. And the method I use is called, "degree corrected stochastic block models." And you don't need to worry about what that is technically. Intuitively, what it does is, it's an algorithm that
says, how do I slice up this network into a given number of groups so that I maximize the messaging within clusters, and minimize the messaging between them. So in the figure that you see in front of
you, we essentially have a messaging network. And if we tell it that we want to divide these people up into four groups, the slices might look something like this. It's going to find two groups of
men and two groups of women who end up messaging each other.
Now, just a side note to keep in mind here, we don't have to tell the algorithm that these are stray people, they'll reveal that by messaging patterns and they'll be put into the appropriate
submarket. When I show you submarkets later, this is the method that I'm going to be using.
Now let's talk a little bit about how you might measure market desirability. So a first approach might be to say, well, the people who get the most messages are the most desirable. And that's what's
shown in this figure here, it's called, "[indegree?]," essentially, the person with the highest indegree is the most desirable. The problem with that is that, well, you can be king of the losers,
right? You can get a lot of messages from highly undesirable people, and we don't want to necessarily call you the most desirable person in the group. So instead, what you'd rather do is, you want to
weigh the value of messages you're receiving by the desirability of the sender. In other words, messages should count for more if they're coming from more desirable people. And it turns out that
there's a readily-available way to do that with an algorithm called page rank, which you might be familiar with from Google's way of organizing Web pages. Essentially, what page rank is doing is,
it's weighting people in this way such that the most desirable people's messaging receives the greatest weight. And as you can see here, you get a somewhat different picture of who's most desirable.
Now, all of the desirability results that I'm going to show you are going to be based on this page rank measure. So now I want to tell you a little bit about the online dating data I used for this
analysis. It comes from one of the most popular sites in the United States. In total, in January of 2014, the site had over four million active users in the United States. By "active users," I mean
someone who logged onto the site within a three-month observation period. Now, this is a free site aimed at a general population, and it caters to a wide range of relationship types. On average, the
users tend to be coastal, young and urbane. For this particular analysis, I'm focusing only on straight users, and I'm looking at all of their messaging activity in January of 2014. Now, it wasn't an
accident that I chose January, 2014, because messaging actually fluctuates quite a bit over the course of the year. January is a time when there's a massive influx of new users. For me, that was
useful, because I was getting sort of a burst of messaging activity.
Now, as I said before, I'm going to focus on two cities in this analysis - New York and Seattle. And you see a couple of descriptive statistics about them below. First thing I want you to notice is
that on average, men and women in both cities are in their early 30s. And that's true across the board for a number of sites that I've looked at. On average, women receive more messages than men,
that's also a pattern that gets repeated across sites. Now, one thing to notice, of course, is that people in New York on average are getting more messages than people in Seattle, but that's partly
due to the fact that it's a larger site, and the mean is getting pulled up by a few very desirable users, which I'll show you later. Finally, I want to finally point out that New York is a much
larger market than Seattle. But most importantly for our purposes, the reasons why I chose these two markets is because they vary quite a bit in the ratio of men to women.
So New York is notorious for being a female-dominated market; there's more single women than single men. And that's indicative on this site, as well. In fact, you can't necessarily appreciate how
extreme the sex ratio is, unless I tell you that the average sex ratio in cities across the United States is about 45 percent women and 55 percent men. So the fact that this city has 53 percent women
is significantly skewed compared to the rest of the results. Seattle, on the other hand, is a male dominated market. It's 58 percent men and 42 percent women. And one of the things I'm going to show
you now is how those sex ratios translate into variation in market structure.
So I have a month's worth of messaging data for both sites. And I use those stochastic blog models I mentioned briefly earlier to identify the structure of submarkets, so that's going to be our next
set of results. Oh, it's like a laser pointer, though - just kidding.
So in order to make sense of these results, I want to show you the age distribution of both cities. So on one panel, you see the age distribution for men and women in New York, and on the other panel
you see the age distribution for Seattle. Now, a couple of things to notice. First of all, I mentioned before that New York is predominantly women, and you can see from the graph that that's true
more or less across all age groups; women outnumber men. It's a little more pronounced among younger women. But overall, it's pretty even. In Seattle, we see clearly that men outnumber women, and
that's most pronounced among men under 35. So then the questions is, okay, so these are the age distributions on these two markets. What kinds of submarkets do they form?
Now, I'm going to show you results allowing for four distinct submarkets, which turned out to fit data reasonably well. And I want to show you first the results for New York. So, a couple of things to
notice here. Thank you. These are the four submarkets graphed based on the age of the people who are in them. So submarket one, which is in the upper left-hand corner, is our younger submarket. You
can see people there are generally in their early 20s. Submarket two, which is in the upper right hand corner, is median age of around 28. And incidentally, this is our hottest submarket, they're the
people who are most in demand and most desirable, according to the page rank measure. Submarket three has an average age of 31, and a lot of the messaging with the men in that submarket overlaps with
the messaging for the women in submarket two. But what's distinctive is that the women in submarket two don't reply to the men in submarket three, which is what ends up keeping them in submarket
three. Finally, submarket four will be distinctive in that it's a very heterogeneous submarket, and essentially it contains every single person in New York City on the site who's over the age of 40.
But here's what I want to call your attention to. You can see that there's a slight shift in the women's age distributions than the men, such that women are younger in each market on average. But
there's something else going on here which I argue might be indicative of something about the overall sex ratio, which is that it looks like young women tend to shift into older submarkets in New
York City. And that's indicative by that massive surplus of women in submarket four. Now, notice one thing that's interesting is that older women often complain about the fact that there aren't very
many men in their dating pool. What this suggests, at least for New York, is there's actually relatively even numbers of older men to older women, but part of the issue is that there's younger women
moving into that older submarket, and that's having implications for the sex ratio.
Now let's talk about Seattle, which is predominantly men. Here we actually see an opposite situation. The submarkets have the same general shape, but the market shifting takes place in the opposite
direction. Here, older men are shifting back to younger markets. And that's why you're seeing the surplus of men in the younger markets. Now, of course, there is a surplus of men in Seattle, but this
market shifting is actually leading to greater numbers of younger men, or older men in those younger markets than there are actually younger men in those markets. And as I said before, these have
implications for the sex ratios. So I talked about the overall sex ratio, but now what I'm showing you is the sex ratio in each submarket. And what you can see is that the overall sex ratio gets
distributed very differently across the four submarkets. In Seattle, because of this market shifting, what we end up with is an overabundance of men in the younger submarkets, submarkets one and two,
and relative parity in submarkets three and four. In contrast in New York, because of women shifting into older submarkets, what we end up with is some of the surplus of men in the younger
submarkets, but then a surplus of women in the older submarkets, and significant surplus in submarket four.
So these are two cities, so it's not necessarily a definitive result, but it is suggestive of how overall sex ratios might translate into individual strategies and then have implications for these
submarkets. So from here, I want to turn - I've talked a little bit about market structure. Now I want to talk a little bit about the desirability, and in particular, I want to give you some results
on whether or not leagues exist. Recall earlier, I said what does it take to have leagues? Well, we need a hierarchy of desirability, and we need to have a limit to reach such that desirable people
are unavailable to less desirable others. In terms of a hierarchy of desirability, what this is is the number of messages received by people in terms of the total number of people who received those
messages. And due to the fact that this is a long-tail distribution, I've put it on a log scale. The big takeaway here, if you only listen to this sentence, is, yes, there is a hierarchy of
desirability. Essentially most people get no messages, and a few people get a lot of messages. In fact, we actually thought that we did something wrong when we got this result, because what it looked
like was over the course of January, one person was getting over 10,000 messages.
Now, it turns out that person is real. We call her our "movie star." She gets 10,000 messages a week. She lives in New York City. She's in her late 20s. Of the 10,000 messages she gets per week, she
replies to about two of them. So if you're messaging her, sorry. But she exists, and the messages appear to be legitimate. So, bottom line, there is a hierarchy desirability. Some are more desirable
than others, and what we're interested in now is, are the most desirable people pursuing the most desirable potential mates? Or are the not so desirable people pursuing the most desirable potential
So I'm just going to give you a sense of how I calculated what I'm going to refer to as "the desirability gap," okay, so the gap between the desirability of a message sender and the desirability of a
message receiver. Recall before that I said that way that I'm looking at desirability is using this page rank measure that ranks and weights by the desirability of message centers. This gives me a
score for every single person. So what I do with those scores is that for all men and for all women, I compute the percentile distribution of people within their own gender. So in other words, I can
say that you're the median woman or you're the median man within gender. Now, the reason why this is useful is, it rescales the page range to go from zero to 100, or zero to one, and also lets us ask
questions like, does the median woman message the median man? Or more importantly, does the median woman message the 80th percentile man?
Okay. So that gives us - thank you - that gives us information about the desirability gap. And what I'm going to do is, I'm going to look at - this gives us information about these desirability
percentiles. And now I'm going to look at the difference between the percentile of the message sender and the percentile of the message receiver. So if that difference is zero, that means the user is
messaging people who are of the same relative desirability as they are. If the number is greater than zero, it means that the receiver is more desirable than the sender. And if that number is less
than zero, it means that the sender is more desirable than the receiver. And then for each user, I calculate the average desirability gap for them, so on average, are they reaching up the
desirability food chain, or are they reaching people who are similar to them?
Well, it looks like most people are reaching up. So what this shows is on the X axis, it's the desirability gap. On the Y axis is the fraction of people, and it's broken out by city and by gender.
What you see here is that for both men and women in both cities, most people on average are messaging people who are more desirable than them. And we can quantify that. For Seattle, men and women are
messaging people who are on average 25 or 26 percent more attractive than them. In other words, if you're a zero, you're messaging someone in the 25th percentile. If you're in the 75th percentile,
you're messaging the person in the 100th percentile. In New York, the numbers are slightly different, but more or less the same. Men on average are messaging women who are 27 percent more desirable
than them. And women are messaging men who are 23 percent more desirable than them. So in general, we would say that people are reaching up.
So the question is, they're reaching up, which is good for leagues, but is there a limit to reach? Does the probability of getting a reply vary with the desirability gap? Well, once again, the answer
is yes. What this graph shows is the probability that a message receives a reply, given the desirability gap between the sender and the receiver. And here now, the sender and the receiver are
slightly modified, so a minus one here means the sender was extremely attractive and extremely desirable, and a one here basically means that the sender was the least desirable person in the
population. And what you see is for both men and women, this sharp negative slope. The probability that a man will reply to a woman when the desirability gap is huge and in his favor is a little over
.7, whereas the probability that he'll reply to a woman who is extremely undesirable relative to him is around .3 or .2. Same thing goes for women, but the scale is slightly different because women
tend to respond at lower rates than men. And we see some variation between cities, but overall the patterns are pretty similar.
So what I've done here is, I've given you a couple of ways to think about market structure, and a couple of tools for actually measuring it. And then I've applied them to two cities to give you a
sense of what the structure of those markets look like. I showed you the submarkets and showed you how the submarkets are organized in large part by age. So age is a major factor organizing
submarkets, but it's not actually the only factor. I don't have enough time to give you all the results today, but another interesting thing that comes out is that race is another factor organizing
submarkets. So that submarket four, that very diverse submarket, which tends to be disadvantaged in general in these markets also is disproportionately minority group.
The other thing that I showed you is that overall sex and balances get distributed unevenly across the markets. And I argued that this may be indicative of some kind of market shifting, so that an
excess of men leads to a surplus of men in younger markets, and an excess of women leads to a surplus of women in older markets. And finally, I've given you some preliminary evidence that leagues
exist. I showed you that there's a tendency for people to message up the desirability food chain. And the probability of a reply varies with the desirability gap. That's what I've got. Thank you!
JENNIFER HICKES LUNDQUIST: Hi, everyone. Thank you for being here tonight when you could be out in Chicago right now. I appreciate it. I would like to first start by acknowledging my collaborator,
who has been instrumental from the beginning of this project, Ken [Holen?], Assistant Professor at University of Texas, Austin, and more recently, Celeste Curington, a graduate student at University
of Massachusetts has entered the project with us, and has also been wonderful to work with.
So we got this data from an online dating Website a couple of years ago, and it was very exciting. We were among a small handful of scholars to first get their hands on this kind of data. And from my
perspective, it really enabled a unique perspective into interracial interactions well before couple formation, and what, until now, has really been a veritable black box. So for today's talk, I want
to talk a little bit about an overview of some of our findings. We asked, when white daters are privy to a large dating market, are they more open to contacting and responding to minority daters? And
how does this vary by gender and sexual preference? We focus today on white preference behavior, but I do want to note that much of our published work has also taken into consideration the preference
behavior of minority daters as well, which are equally interesting.
So why is this important? Why do we care about this? Well, some of the most significant boundaries between social groups are maintained through coupling patterns and understanding the extent to which
racial groups are open to intermixing as a telling barometer of social boundaries in any given society. So what do we know? Well, mostly what we know comes from census data about interracial
coupling, which applies really only to households. Couples that have formed at the household level. We know very little about what happens prior to household formation, for example, at the earliest
stages of couple formation. But the primary pattern is a homophyly, that is, groups prefer like other groups. And most interracial partnerships are mainly within their same-race groups. Interestingly
- well, and this also, of course, varies by the age of the households where interracial cohabitation and intermarriage is slightly higher. And there's also been some interesting data coming out now
that the U.S. census is collecting information on same-sex householding that LGBTQ households have higher rates of interracial cohabitation than different sex households, and there's a lot of
theories about what drives this.
So when interracial mixing does occur in the census data, basically the theories that generally support the patterns are these. The first is tri-racial hierarchy, where white - whites are much more
likely to intermarry with Hispanics and Asians, and much less likely to intermarry with black Americans. But there is an interesting gendered component to this, where white men are more likely to be
married to Asian women, much more so than the Asian men are to white women, and just the reverse among African American and white couples, where black men are much more likely to be married to white
men than are African American women to white men.
Here's what we don't know. Is it preference, or is it exposure? The assumption is often that it's about preference. So in survey after survey, Americans overwhelmingly say that they approve of
interracial dating, interracial marriage, that their child would marry interracially. And yet only four percent of households are interracial households, nine percent if cohabiting. If one were to
randomly sort American households into how they would fall out, it would be about 40 percent would be interracial households. So the question is, how much is this about exposure, given that we live
in a highly racial segregated society, it may be that individuals, it's not so much about preference, but, rather, being with only the same-race folks in school and in neighborhoods, church, et
cetera. So to get around this, a number of studies have looked at stated dating preferences; that is, when people sign up for an online dating Website, which, ironically, is the only place where it's
socially acceptable to say, "I prefer this race and I don't want this race" - scholars such as Feliciano have looked at this and find that there do seem to be racial preferences that match with the
sort of mating patterns that we see in the census. But clearly there are social desirability issues of what one says they prefer. So ideally we get not just what people say, but what they do. And the
ability to be able to access online data from online dating Websites is incredibly tantalizing, because we can get at not just what people say, but what they do.
So we argue that an online dating setting has far fewer partner market constraints than most social settings. And so we expect that the propinquity, or exposure, this year will be much reduced;
therefore allowing us to separate actual preferences from market constraints. And a number of others are in the process of using various data from different online dating sites.
So first, just a little bit about the sample. It's one of the top - largest top five online dating Websites in the nation. It's a relatively generic Website, relative to some of the niche dating
Websites. And at the time that we got the data, the data go up through 2010, this was prior to the proliferation of some of the apps, the geo-based apps that Michael mentioned, and some of the social
networking apps. A lot of people ask us how we got this data, and basically we thought, well, let's just email the dating company and say, "We have a wonderful idea, how you would like to co-author
an academic analysis with us?" And they replied that they weren't interested at all, but they would be willing to sell it to us. And they gave us a price, which at the time seemed expensive. We
raised the money, and since then I've heard that the rate has tripled in price, so I think we got there early.
The original data set was originally nine million daters with 200 million messages exchanged. And we ended up whittling it down to this size that you see here, after excluding people who were - did
not check off that they were only looking for a relationship, so they were looking only at daters who were seeking a relationship. And after weeding out the incomplete profiles, profiles that had
minimal activity, and this is where we are.
So the drawback to the data is that it's not nationally representative of daters, nor is it necessarily nationally representative of online dating Websites, whatever that might mean. We also - we
don't have photos and we don't have message content, so we don't have text. We're very simply looking at the full profile information as controls, and we're looking at the rate of exchange; so the
likelihood that someone will message someone else and will respond to someone else. We also are careful to note that online dater demographics may overestimate racial openness, given that those who
are online, our average age is around 30 and younger folks and slightly higher socio-economic status folks are more open to interracial dating.
Okay, so the strengths are, as I mentioned, we directly observed the dynamics in bounded space, and were able to separate choices from circumstance, what people say from what they do. We have a very
large N, which minimizes that propinquity issue that I talked about. And perhaps the most interesting thing is, there's minimal social desirability bias in that even though our daters are, for
example, four inches taller than the average American, which is probably an exaggeration, the daters that are responding to them - thank you - the daters that are responding to them are responding to
that information. So that is their truth, unlike survey data, right, where someone might say something, but someone interacting face to face with them would know otherwise. And there is some
unobserved variable bias, for example, the text and the photo data. But for the most part, we're seeing everything the daters are seeing.
And in the interest of time, I'm going to skip over the descriptive data, at the risk of making you feel dizzy. Okay, so a challenge of this project was the novelty of its data. What is the best way
to design models that jointly take into the account of characteristics of millions of daters within your same city who you may contact, but don't contact? And so this necessitated some subsampling
creativity, which I credit my co-author, Ken, with. We experimented with a number of other models before settling on general estimating equations which allowed us to control for dyad-dependents,
while still considering very, very, very large samples, a high degree of singleton data and complex, variable interactions. And essentially, very simply, what we're doing is modeling two different
types of interactions, as Elizabeth talked about. Sending, the likelihood of sending, and responding; controlling for all of these attributes that you see there in the control variables of both
members of the dyad. So the question is, what is the likelihood controlling for all characteristics that a white dater will contact a minority dater, compared to their probability of contacting a
white dater? And same with response; what is the likelihood one will respond to a different race dater relative to a same-race dater, holding everything else constant?
And I'm going to show you a figure from our paper that just came out in Social Forces. Focus first on the sending side. This is the sending model. And these are predicted probabilities. And what you
can see is - so, essentially, everything below one is a negative probability of contacting a different race dater relative to a same-race dater. And so what we see is that there is different group
variation between the likelihood of sending an initial message across gay women, gay men, straight women and straight men with straight men and gay women looking pretty similar, gay men slightly less
likely and straight women, unlikely. Now, when you switch to the responding models, we see something a little bit different. You see that the probability goes much higher. And part of what we're
grappling with, as others are who are researching this, is what, theoretically, does it mean to send an initial message? And how does that differ from a response? And the way that we think about
sending is, this is sort of, perhaps, a habituated behavior, an indicator of who one's ideal is, but also who one might expect to respond to them. So if you might - if you anticipate rejection,
perhaps you would not contact that person, whereas response, you know the risk of rejection is not there. And so we look at response as revealed preferences, and a lot of our work focuses on the
response models. But they're certainly utility, considering them both simultaneously.
So what you see with response here is - you know, first we were talking a lot about, okay, so what is that - why are gay women and straight men more likely to respond outside of their race relative to
gay men and straight women? And then we started thinking about actually, what's really important to think about is the target date. And so to what extent are - what we're really seeing is that
minority men are less likely to be responded to relative to minority women. Now, we also do a number of analyses that break this down into each specific race. And I will say that even though the
patterns are all very much the same, but there are some racial hierarchies that are introduced. So for example, even though gay women and straight men are more likely overall to respond to all races,
they are least likely to respond to African American women who contact them; whereas gay men and straight women are basically unlikely to respond to anyone who contacts them, except for whites, in
our analyses.
I also want to show you a quick overview of some results in a paper that is coming out in ASR, because we're very interested in, what is the extent to which multiraciality interacts with our findings?
And you can see here, we haven't done analyses yet of non-straight daters. So this - the first on top are, again, predicted probabilities; the probability of responding to straight white women
responding to a different race data. And what you can see here is that while straight white women are less likely to respond to Asians, blacks and Hispanics, that there seems to be a whitening effect
for multiracials, where they're more likely to respond to the - for example, a black multiracial dater than, say, a black dater. And interestingly, Asian white daters, they are equally likely to
respond to them as they are whites. Among men, you see a different pattern where, again, the same kind of racial hierarchy exists among mono-racial responses. But you see a bonus effect among Asian
whites, where white straight daters are more likely to respond to Asian multiracial women than they are white women. And then similar patterns as the women, where they're more likely to respond to
multiracial daters than their mono-racial counterparts.
Okay. So to return to the motivating question of this analysis, which white daters have access to a large dating market, do they behave more race inclusively than different sex daters? And the answer,
as it turns out, is highly dependent on the gender of the minority dater. So just to repeat here, this finding that minority men overall tend to be avoided on the online dating market by white
daters, and minority women less so, we root in the literature on gender formation theory, where black men are often stereotyped as dangerously hyper-masculine, and Asian men as effeminate. And in
contrast, Asian women are often depicted as, say, exotic and submissive, while other minority are depicted as sexually available to white men, as somewhat of an explanation for what we're seeing
here. In terms of the multiracial effects, this really occurs with an increasing media preoccupation with this racially ambiguous prototype of models and actors, in kind of a self-consciously
post-racial celebration of diversity. And we're finding a similar effect, but only for some multiracial groups. On the positive side, the bonus effect that we're seeing indicates that there's not so
much of a ridge white, non-white divide, at least in the online dating market, as there has been previously. And this calls into question the predominating honorary whiteness theories that
consistently identify whites at the top of the racial hierarchy. So thank you for listening! [APPLAUSE]
PAULA ENGLAND: And we're going to open things up for questions. If you want to ask a question, I'm going to ask you to come to one of the mikes on either side. You can ask a question to a specific
panelist, or you can ask a general question and they'll kind of confer about who's going to take it. So do I have any questioners? While we're waiting to get some people emboldened to ask questions,
anyone on the panel want to ask a question of each other? Yes, Jennifer?
JENNIFER HICKES LUNDQUIST: I would love to hear about some of the geographical differences in the race results, if you could talk a little bit about that.
ELIZABETH BRUCH: I haven't gotten to race yet.
ELIZABETH BRUCH: But I will say that the thing that I liked about the stochastic block models was that they weren't parsing out variation on a particular race. They were just saying, given the
messaging at work, what are the best cuts? And so age seemed to be a major stratifying factor there, but the submarkets themselves have other compositional differences. So, I mean, the one thing that
we found was that, for example, that submarket four is predominantly black. But more interestingly, black women are younger on average in every submarket. So black women in submarket four are on
average eight years younger than the white women in submarket four, which suggests, consistent with the fact that younger women seem to move into older submarkets when they're disadvantaged, there's
this other layer of it about race we don't fully understand. And that's kind of stacked across all of the submarkets. But beyond that, we haven't looked at the race effects, although that's something
that we want to do, going forward.
JENNIFER HICKES LUNDQUIST: Yeah, I wish that we had better data than just the MSAs, the metropolitan statistical areas, so that we could look at indices of dissimilarity of segregation, for example,
and see how specific areas might affect dating behavior in terms of racial contact. So while we're just controlling for geography, I think in the future, being able to hone in the way that you're
doing will be really effective.
ELIZABETH BRUCH: Yeah. My data are granular, and it was precisely because of looking at those kinds of geographic variation, even within a particular metro area. Yeah.
PAULA ENGLAND: Okay, we have a question.
>> So I have a question. I guess everyone could sort of speak to this, and it's going to be this obnoxious thing where I ask you to talk about something that this isn't about. But it's related. So all
of you talked about sort of Internet dating, and, like, sort of what happens when sort of people - assuming, like, people want to date, right, but we also know, I think this speaks to Michael, the
framing of his talk. People use the Internet for just sexing, right, we're, like, I don't actually want to date, I want to sex. And maybe dating happens or not. And I guess I'm just wondering, like,
do you think - are there similar kinds of patterns you see when you're sort of assuming dating and partnering? Does it play out the same way when I'm looking for something a little less long-term?
PAULA ENGLAND: So, like, an interesting question is, is there a desirability hierarchy in either case, I would think the answer would be yes. But maybe Elizabeth wants to comment. I don't know, have
you looked at the non-dating parts of your data? People who are more looking for casual -
ELIZABETH BRUCH: Yeah, so for this analysis, I restricted it to people seeking. But I'm also very interested in how people seeking different types of relationships might lead to different market
structures. And more importantly, I think it might be interesting to learn the extent to which the market for people seeking, say, marriage overlaps with the market seeking for one night stands,
either because the same people are seeking both, or because, say, the people seeking one night stands are messaging the people seeking marriages.
PAULA ENGLAND: Yeah. Michael, go ahead.
MICHAEL ROSENFELD: Well, it's also true that people start out with a one night stand and end up in a relationship, and vice versa. So and this is actually, you know, some people, more women than men,
describe the frustration of this, right, that the guy describes him as interested in a long-term relationship, and then as soon as she starts messaging him, then the question is, can I come over
right now? Right? You know, which is a behavior that you'd more associate with Tinder or Grinder or Craigslist Casual Encounters. I mean, I interviewed some interesting people who - I interviewed an
interesting woman who every friend that she has in the bay area started out as a Craigslist Casual Encounter, right? And she decided that she just wanted to hang out with people who she was attracted
to, and that she wanted to get that out of the way first. And so it's hard - it's interesting how what can start out as one thing can turn into something else. And people are a little bit cagey about
what it is they're looking for.
JENNIFER HICKES LUNDQUIST: The Website that we use, it's possible the daters check off whether they're interested in just a platonic relationship, or if they want a committed relationship, or whether
they're interested in just casual sex. And we played around with trying to compare some of these findings that we found so far by looking only at individuals who check off casual sex, compared to the
others who want a relationship. But we were quickly halted in our steps because there are so few people who only check off casual sex; they often check off both. If they're going to check off casual
sex, they also check off relationship. So we're going to go back and see what we can do with that maybe descriptively. But I expect that some of it, at least some of the hierarchies that we're seeing
are probably much lessened when people are thinking less long-term.
PAULA ENGLAND: Yes? Question over here.
>> Hi. I had a question about whether or not you guys considered the impact of occupation. So, for specifically about the market structures, if you look at Seattle and you look at New York, you're
looking at the tech boom in Seattle. And I'm from Seattle, and that's, you know, I've experienced online dating there. And I can tell you there's a predominance of a specific type, which is usually
people employed in the technology field. And I'm curious, I would say New York City would be very different. But have you considered controlling for the differences in occupation status in the two
populations, and if the results vary?
ELIZABETH BRUCH: So occupation is entering this story in two ways, right, so because of these technological geographical location issues, there is a proliferation of a certain type of person in
Seattle versus, say, New York. And that could contribute. And in fact, you know, the San Francisco Bay area is even more lopsided than Seattle. The San Jose metro area on my site is 70 percent men,
and I can't remember exactly what the San Francisco Bay area is, but it's significantly higher than Seattle. However, many of these men, despite logging on to the site appearing to be active users,
never send a message. So Seattle seemed to be lopsided enough, but not so demoralizing to the men that they were still active. And that's why I ended up going with Seattle. But that doesn't answer
your question.
In terms of occupation, we haven't looked at that yet. We've been focusing mainly on sort of documenting this market shifting around age. But it would be interesting to see, for example, if certain
occupations were devalued in certain places where there's a lot of that type, versus if there's only one or two of that type. I mean, you could think the same thing - I'm from Ann Arbor, I can
imagine dating a professor is a dime a dozen in Ann Arbor, whereas it might be more exciting in another city. So it's an interesting question.
>> So as a follow-up, would you consider impacting the hierarchy? Like, would you consider occupation -
ELIZABETH BRUCH: Well, one thing you could do is see, for example if that - I mean, the desirability hierarchy kind of is what it is, regardless of the attributes of the people, right? But you can
look at who's at the top and who's at the bottom. Without doing more sort of statistical analysis, it would be straightforward to look at, well, I haven't done this to look at, for example, are the
most desirable men, these men in the tech industry? Or do those men end up falling lower down the desirability food chain by virtue of the fact that they're so common, or there's stereotypes about
>> Thank you.
>> Good evening. I have two different questions. But first, I'm wondering if you could speak to - and I meant all of the panelists - the to the extent to which you're expecting Internet dating and
online dating to be something that increases exponentially over time, or if you're seeing some change over time in terms of decline or stabilization? The second question I wanted to present was, in
the last couple of presentations that included race, it's really important to think about the digital divide, right? And so to think about the proportional number of blacks, Hispanics, that have
access to the Internet in a way in which would facilitate them dating online. I mean, if you have to go to a public facility like a library and log on for 30 minutes to try to find a date, I mean,
that's a whole different thing. And I'm not sure if thinking about the digital divide might help understand those racial differences that you were describing. Thank you.
MICHAEL ROSENFELD: So I can try to speak to the question of Internet dating increasing over time. And I guess you guys also could, but in my data, which is now five years old, so it's old, up through
2009, there was kind of a plateau for heterosexual couples at around 22, 23 percent meeting online, whereas it seemed like same-sex couples meeting online was headed towards 100 percent. So I don't
know if there's more recent nationally representative data on that - I don't know if you guys -
ELIZABETH BRUCH: Is the Pew data nationally representative?
MICHAEL ROSENFELD: It would probably be.
ELIZABETH BRUCH: Because it definitely showed an increase up until - I think that was 2014 that that study came out.
MICHAEL ROSENFELD: Okay. Yeah. In terms of Internet access impeding people's ability to do Internet dating, I think that where there is a will, there is a way. I mean, I've definitely interviewed
people who are, you know, are living in the garage and barely scraping by, but they have the cellphone, right? The ubiquity of the cellphone is kind of a big deal and anything that, you know, the
cellphone is Internet-enabled, so people are doing the OkCupid and the Tinder, and all that kind of stuff. You don't need a lot of connectivity to do it. So I think that it's not that huge a barrier.
And the percentage of households that have Internet access at home is, you know, around 85 percent or more. There's certainly a class issue there for sure. But the barrier to getting the phone is not
as great.
PAULA ENGLAND: We're going to go to another question over here.
>> Okay. So I'm going to ask you a question, I have been thinking about it, not just from your presentations but ever since I heard the OkCupid guy the other night. So it turns out, like,
attractiveness matters, race matters, you know, people suck. Like, I would like to hear you reflect on what is meaningful about finding this out in this context? Do you hear what I'm asking? That
basically a lot of what I've been hearing in a lot of presentations related to this issue is very descriptive, that people discriminate, people have these hierarchies, there's a league or whatever.
That feels like a basic kind of sociological insight, but what I'm asking is, what do we get from hearing about that in this context?
JENNIFER HICKES LUNDQUIST: Well, to me, I think that one of the more optimistic findings is the fact that people are much more likely to go outside of their racial boundaries when they're responding
than when they're sending. And I think that that's, I think that is indicative of the fact that when people are more exposed to one another, then they're more likely to consider interacting with
others who are not like themselves. And so, you know, I think about, for example, the contact hypothesis, which is the idea that in highly racially integrated settings, people tend to have higher
rates of interracial marriage and more open attitudes towards racial mixing. And, you know, to a certain extent, while cyberspace is certainly not integrated, it enables more of that contact to
happen and I only see that increasing.
PAULA ENGLAND: Yes? Question over here.
>> Hi, thank you. Interesting. I wanted to just make a couple of quick comments from a different context, from the U.K. and some more qualitative research. I did a small study of people living on
their own in the age 25 to 44, some of whom were Internet dating. But there was a bit of critical commentary on Internet dating as well, so one of the things that women - some women said was the
repeat partner cycle that you have to go through became sort of tedious and something that men said was that their masculinity was sometimes undermined by having to Internet date, and that if they
couldn't get a partner by conventional means, then there was something about that that was troublesome. So, in a way, it was maybe reinforcing some conventional gender scripts, rather than subverting
or overturning. And I think that maybe reflects what the previous speaker was also asking. So I think some of the optimism that we had about Internet dating making a more level playing field doesn't
necessarily come to fruition, when you speak to people about their experience.
MICHAEL ROSENFELD: Well, so it's true that a lot of what's complicated and problematic about face-to-face everyday life, a lot of those things are replicated online and that shouldn't be totally
surprising. But I think it's also, maybe to amplify a point that Jennifer was making, that even though there may be a lot of preference bias and who messages who and so on, it's also true that if
you're looking for something really unusual, if you're interested in interracial dating, if you're interested in dating someone 20 years younger or 20 years older, or if you have a very particular
preference, it's also easier to satisfy that preference to find a partner who satisfies that preference. Now, the fact that fewer people have that preference is interesting, too. But I don't think we
should expect the Internet to undermine all of the social hierarchies that we ourselves spent so long building before, right, because who's making the Internet socialist? Just us, right? So it's just
a reflection of a lot of our own biases and prejudices, problems and hierarchies. But it also provides a window for people to get around that if they want to.
JENNIFER HICKES LUNDQUIST: Yeah, it's funny. Because I do this research, I'm often thought of as some kind of expert on online dating, so I often have particularly young sponsors come to me and ask
for advice. "What's the best way to get a good date?" I think there needs to be - you know, these technologies are moving along much faster than our protocols of dealing with them. And what we are
seeing from all of our data is that men are much more likely to send. They're taking on - it's kind of a replication of a conservative role. Women let me come to them, and they will respond only to
the men who contact them, much more so than one might expect when the rejection rate risk is a much lower stake situation. And so what I often recommend to women that, you know what, don't even pay
attention to the men who contact you. Think about - approach it very analytically, think about what you want in a partner. Come up with, you know, 20 good matches and you contact each of them, and
they will be so shocked that a woman is contacting them that they'll respond to you. In a very small - and of, you know, six professors at my university, it seems to be working. [LAUGHTER]
ELIZABETH BRUCH: Actually, I can make your end 20,000 -
ELIZABETH BRUCH: - because that is one of the bright spots of online dating, is that women hardly ever send messages. But the reply rates for men are at least four times as high as reply rates women
have for men. So, and it's funny, there's a game theoretic model of how preferences turn into matches. And basically what it shows is, you know, imagine there is a marriage market and only one sex
can propose. So if men propose to women, the end match is basically overall favor men's preferences. But if women propose to men, the end matches favor women's preferences. And so whenever I'm
teaching this stuff, I'm always, like, ladies, you got to get out there! So that's another piece of that story.
PAULA ENGLAND: Yes, we have a question over here.
>> Hi! So I wanted to talk about the topic of safety that came up earlier. As someone who met her partner of nearly two years on Tinder, I have very vivid memories of going, getting ready for the date
and insisting on driving my own car and packing my Taser because I was nervous. I didn't know if he really was the nice guy in his profile, or if he was some kind of serial killer. So I was wondering
if you have any data about rates of safety, because I know on Tinder it's consensual when people talk to each other on the messaging app. But beyond that, is there any way we know that it's actually
safer or not? Luckily it was for me.
MICHAEL ROSENFELD: I don't know of any data that would speak to that specifically. I don't know if Tinder would know, or - but my suspicion is that it's safer than everyday life. And part of this has
to do with research on where violence - where intimate partner violence comes from. It tends to come from the steady partner, and not from the person you only meet one time. And that's kind of a
reversal of the script about crime and violence that we see on TV, that it's all about the scary strangers. But in fact it's - you know, intimate partner violence actually comes more from
relationships. So, but that doesn't mean that if you're - you know, I mean, I think - I don't know about the Taser, but if people are cautious about meeting others - but at least from the people I've
interviewed, you know, the kinds of stories that people tell about the guy who was bothering them and wouldn't leave them alone, it's when they've let them into their everyday world and they can come
to the door, because it's easy to block somebody on your phone, right? So but - the people who I've interviewed, the women who I've interviewed, you know, the woman who was meeting everybody on
Craigslist Casual Encounters, right, which you would think - I mean, it's been described by someone people as, like, the bathroom door of the Internet, right, it's kind of like where stuff gets
scribbled and it's all about who wants to come over in the next half an hour. And you know, she found it to be perfectly safe.
JENNIFER HICKES LUNDQUIST: There was a high profile case of, I think, a woman in California who was sexually assaulted on a - maybe it was Craigslist. As a result, an attorney sued Match.com and
maybe one other online company, and they now do background checks. They have some kind of background check system. So there is some response to the perceived higher risk. But I tend to agree that the
risk is probably not as high as it's often portrayed in the media, often because the places where people are meeting are in highly public places, and often, you know, I think daters should be coached
about what would be the kind of the safest protocol for meeting someone for the first time. Someone was also telling me about an online dating app that I've never heard of, but I thought sounded like
it might respond to that. And that was where there are group dates; so it's two women and two men if they're straight daters. And they're all matched up in some way and they all meet as a group,
which seems to be somewhat of a safer scenario and maybe lower stakes.
ELIZABETH BRUCH: Well, and apps like Hinge are designed to put you with people who at least you have several degrees of separation so there's some accountability. I mean, I think it is a big issue,
and it's something that at least when I've talked to people at these companies, it's on their mind. I mean, I work, analyze data also from an affairs Website, and they think about it a lot because
about safety, I think people are more concerned on an affairs Website than necessarily Match.com. We talked about, I mean - you want something sort of like [INAUDIBLE] where people can write reviews.
But that would be completely inappropriate, so, you know, there's this challenge of, like, is there a way of putting community accountability into these sites in some ways so that people can have
some sort of sense of the person with external validating.
MICHAEL ROSENFELD: And just to get back to the question, I would say, you know, safe compared to what, right? Because there's no such thing as perfect safety. So in the older dating system where you
were going to meet somebody that your mom knew their parents, right, the idea - I think sociologists kind of assume that because your parents knew their parents, that this was going to be a match in
the community, that there would be a lot of information about that other person. But it turns out, I think, that your mom may not have actually known as much about other people as Google does, right?
So it might be easier for you in this day and age to glean information about this new person than it was for you to get valuable information from your mom about this new person, because what does
your mom know about this new person? Your mom knows that their mom is nice. Right? And that was kind of how information was transmitted in the old days. Well, you know, I was hanging out with them at
church and they seemed really nice. And they have a son just your age. Right? And it didn't really need to be more specific than that. But the quality of that information about the character of this
person might not have been all that great.
JENNIFER HICKES LUNDQUIST: We also have some data - we have a variable, then, to get whether our daters were blacklisted, which means that they were complained about to the dating company. It could,
you know, it could be harassment, it could be they're a spammer, or something like that, so we don't know how to tell. But that might be interesting to look at how that correlates with some of the
profile characteristics.
PAULA ENGLAND: At last we have come to the bewitching hour, so thank you so much for coming. And please join me in applauding our panel. I'm sorry we don't have time for another question. [APPLAUSE]