My statement was aimed at the article; i`m not familiar with the whole study data, but assuming the article showed all of the important data, it`s interpretations are misleading. Further more data are lacking in a way that meaningful interpretation is not even possible; important reliability measures are not presented.
First of all, let`s take a look at how the referred "app" works. There are three steps to follow in order to complete a connection, as shown below:
1. The "app" suggests dating partners by vicinity and interest. (In the process of registration you have to enter the postal code where you are living and provide three interests; also select of your desired dating partner`s interests.)
2. To invoke a request you have to select from the suggestions the "app" has given you and send an initial request to show your interest in a person.
3. The person receiving the request decides if s/he accepts it or not. In the first case a notification is sent to the one initiating the contact (=> connection complete) and the two can proceed from there.
The whole process is not based on "race" and you don`t even have to provide this info. Only those who did were captured in the study.
Now, let`s take a closer look at the three steps leading to a completed connection:
Step 2 is depending on step 1, which means there is not a totally free choice of desired dating partners in the context of "race". (It is safe to assume that you will always be suggested "races" that dominate the neighborhood in numbers to the highest percentage. Assuming the one who sends an initial request has no preference regarding to "race" most requests sent will be to people of the "same race" in such a situation, because the highest percentage of both "app" users and suggestions can be assumed to reflect the highest percentage of inhabitants of the neighborhood.) This shows that dating interests (by "race") cannot be interpreted directly by counting and comparing the numbers of requests sent to groups of desired dating partners.
An analysis of initial requests (step 2) could have been done, though. For example by preselecting areas that have each two of the "race groups" of interest represented by approximately the same number of people and selectively analyzing the interactions of those two groups only. In a further step all those pairings could have been set into relation leading to a comparable result. (This technique is used, for instance, for rating the purchasing power of currencies, which can only be done by looking at the exchange rates of pairs of currencies first and then analyzing relations between each of those pairs on the world market in an iterative process, considering "chains" and "circular relations" between the currencies.)
In this study no analysis about the initial requests by the "dating app" users has been done which is a downside, because it would have been an important reliability measure for the study. (A user`s interest can not only be derived from the responses (step 3) to an initial request but also from the initial requests (step 2) themselves. The study could only be assumed to be reliable if the results from analyzing both initial requests and their responses would have represented approximately the same numbers. Otherwise the study would have proven worthless, and an interpretation would have been that "racial preference" does not play a major role for most of the "dating app" users.)
Now that we have discussed what has been left out of in this study let`s take a closer look at what actually has been done and how the results have to be interpreted (or if they even possibly can)...
First, let me say that relative numbers (percentages) only make sense in relation to absolute numbers they represent. For instance when you look at a mixture of two fluids, a percentage given for a fluid could represent volume, mass or molecule count, in general, each leading to different relative numbers. In a social study absolute numbers represent either the number of people within a group or within a relation. In this article no absolute numbers in any case are presented, which makes interpretation of the relative numbers provided impossible!
Looking at the above picture, what do those percentages mean? According to the article, take for instance the black women`s "highest response rate" (first line). It means that 9.3% of the initial requests to black women made by black men are answered positively. (Yet, we don`t know how many black men or other men(!) have made a request to black women!)
As an example of (mis)interpretation(?) of those numbers let us take a look at white women`s response rates which show the highest spread in the group of responses made by women.
Highest response rate by white women: 6.7% to white men;
lowest response rate by white women: 2.8% to black men.
This means a difference of 3.9% between highest and lowest response rates and the highest response rate being almost 2.4 times higher than the lowest. However we do not know anything about the numbers of requests made to white women by both white men and black men.
In a ficticious case where a white woman would receive 15 requests by white men and 37 requests by black men in the same period of time, she would have answered more requests done by black men than white men, even though "her" response rate to black men is lowest and that to white men is shown to be highest.
This "effect" is even more problematic when we look at the response rates by Asian women which appear to have the smallest spread:
Highest: 7.8% to white men; lowest: 6.7% to black men.
Here a ficticious number of 13 requests by white men versus 16 done by black men (which is a fairly possible distribution in "real life"!) would mean a higher number of responses to black men (that are designated "lowest response rate" by the study) than to white men (which appear as "highest response rate" in this relation)!
The absence of absolute numbers of initial requests yields the relative percentages to be useless and incomparable!
The same observation holds for the statement that "Men respond to women around [three] times more often than women reply to men’s messages."
In a situation where initial requests are done three times more often by men vs. women (which is not impossible as there are many studies about dating "tools" as well as "real life" dating that show that it is kind of cultural in many communities that mostly men make the first step, not women) positive responses given by both men and women would be about the same in absolute numbers!
Same here.
