Aren't grandparents adorable? They're sweet and kind, they've been married for decades, and they've got wonderful archaic 1920s names like Edward and Edwina. Last week, based on the anecdotal evidence of my own grandparents and a couple from an NPR report, we speculated that couples from that older generation were more likely to have similar names than couples from the current generation.
It seems plausible, but is it really true? We invited readers to give us the names of their own grandparents, as well as their current significant others, and they responded with over 3,000 names. Then we took a look at data from the U.S. Social Security Administration (SSA) about the distribution of names in the population. Do people pair up with people with the same name more frequently than if they were paired randomly?
We picked two years of SSA data to compare with our results: 1982 (about the "average" year our readers were born in) and 1932 (our guess as to when their grandparents would have been born). If you haven't checked out the SSA names database, you should give it a look -- it's quite interesting. We considered the top 1,000 given names for both males and females in each year. Right away, we noticed an interesting pattern:
If you were born in the U.S. in 1932, you were significantly more likely to be given a name in the top 1,000 than in 1982. This was especially true for females. To put it another way, more different names are used today than half a century ago. You might think this would make it more likely for couples to have similar names if they were married half a century ago compared to today. But our results suggest otherwise. Let's take a look at the data from the grandparents' generation first:
In a random pairing of 1932 names (weighted for popularity), about 5.7 percent of pairs start with the same letter. The actual couples from our survey shared the same first letter 6.3 percent of the time -- not significantly different from the random sample. We did several other tests of similarity and each came up with the same result. Even when we combined these tests into a overall score, there was no difference between our sample and a random sample.
But what about current-day couples? Here are those results:
Based on these responses, today's couples are significantly more likely to share both the first letter of their names and the first two letters, whether you compare them to their grandparents, or to a random sample of names from 1982. They're almost three times as likely as a random sample to share the first two letters of their names! When we combine several tests of similarity, we get the same result. Our readers are more likely to pair up with people who share similar names than you would expect based on chance.
Why might this be?
Here's one possibility: Perhaps our readers are in shorter, more superficial relationships than their grandparents were. Maybe a similar name acts as an initial attractant, but those couples are more likely to break up over the long haul. We could test for this possibility by seeing if relationship length was correlated with name similarity. It should be inversely correlated if the couples who were attracted only by similar names eventually saw the light and split up. We found no correlation, positive or negative, for either age group.
Another possibility which we can't test for: perhaps readers were more likely to participate in our survey if their names were similar to their partners. I think that's unlikely, since we framed the survey to focus on the grandparent's generation, but it is certainly a possibility.
Any other thoughts?
I've never seen the SSA names database, but can you search by zip code or state? Perhaps our names are not randomly distributed across the population, as a random pair sampling assumes. We are (presumably) more likely to be in a relationship with someone who is in the same geographic area where we spend most of our time, and the names of people in each geographic area may be heavily correlated to each other - say, if our parents were choosing the most popular names in the area when we were born.
Here's a link to the database. You can look up by state. Another issue is the international nature of our readership -- we have readers from other parts of the world besides the US.
It might be true that similar names are initially attractive, and the difference between generations is due to our generation having more opportunities to find such mates, thanks to increased mobility and the internet.
What about the changing cultural profile of the US from 1932 to 1982? To simplify, say there were a lot of Hispanic people in 1982 as opposed to 1932. Then say there were disproportionate Hispanic-Hispanic pairings, and that the subset of Hispanic names is more likely to be similar to each other. (This could be true of various groups, of course, in 1982 and 1932 to different degrees.) Then you would expect the greater name similarity than random in 1982 more so than 1932.
That's a possibility, but I think if that was happening among our readers, it would also be reflected in their grandparents, who would also be Latino and have Latino names.
That's a very intriguing suggestion. I'd also suggest that regardless of physical mobility, there's more cultural mobility today; e.g. pairing different races, religions, and so on. So even if you don't leave your immediate community, you have more options for pairings within your community.
Then say there were disproportionate Hispanic-Hispanic pairings, and that the subset of Hispanic names is more likely to be similar to each other. (This could be true of various groups, of course, in 1982 and 1932 to different degrees.) Then you would expect the greater name similarity than random in 1982 more so than 1932.
I think your survey was most likely biased. People who have grandparents with similar names would have been more likely to have participated in your study. For instance, I don't have grandparents with similar names, so I didn't feel much motivation. It seems like this method of survey is even less reliable than a self-reported kind, because your data is only based on the people who happened to have the time or whim to respond. Just my thoughts.
Also concerned about sampling bias here. Norman's critique does not hold, since grandparents' names were no more similar than chance. However, your results could be explained if your readers are interested in themselves, not in their grandparents. That is, the readers most likely to respond were those to whom name similarity was a salient issue - i.e., those with similar-named partners.
I'd always heard that the attraction was to someone with a family member's name, not one's own name. A good friend's father and husband have the same name. Another friend's sister and sister-in-law have the same name. And so on. Perhaps the initial attraction is to the familiar?
Several of my friends share _very_ similar last names, since they got to know each other in an introductory university class which was split into groups based on last name (i.e. the top 1/nth part of the list of students was group A, etc).
Since people nowadays are more mobile, their partner-choices are less likely to be influenced by geography (which is presumably fairly uncorrelated with names within a single country). Whether some psychological preference for a similar name is a factor - whose to say? On the other hand, the such a preference is almost certainly not solely responsible for the patterns in the graphs above.
I would have expected the opposite actually, given that people in 1930's lived in more closed communities. At the same time, a possible explanation is that since there are so many possible groups, communities, and subcultures around us today, people choose to stay within the confines of only one of those.
It may also have something to do with culture specifically.... an a related note for instance, other cultures sometimes have similar first and last names (my grandmother's maiden name was Sara Sarafanov).
Basing this study on the letters of names will yield a lot of false-positives and false-negatives. Alliteration is mostly about how words sound, not how they're spelled. Especially in English, which is such a mishmash of influences, sound is more indicative.
So, pairings like Geoff and Janet, Karl and Christine, Kent and Cathy, Eileen and Ivan would be missed, while pairings like Charles and Christine, Evan and Eileen, and Gene and Gordon would count even though they don't alliterate when spoken.
Cute study, but there is a potential large flaw in using how things are spelled rather than how they sound.