Sunday, October 25, 2009

Diversity in aesthetic tastes over the last 130 years: Evidence from names


Virginia Postrel argues in The Substance of Style that we have entered a new aesthetic age, where design objects are available to a mass audience, where those consumers no longer consider just the price but also the look and feel of the stuff they buy, and so where producers don't compete with each other only on price but also on how aesthetically pleasing their products are. (Think of how many stylish toilet brushes you can buy at Wal-Mart or Target.)

She wrote this in 2003 and claims it began sometime in the mid-to-late 1990s. That trend seemed to take off even more during the recent financial euphoria, and although it's surely taking a beating during the recession, it'll be back once we feel safe spending again.

One thing I kept asking myself was, "Is this really so new?" As a general rule, you should be skeptical of all stories about how we're "entering a new age." After all, what about the 1920s and '30s, which Postrel admits could have been a previous age of aesthetics? She counters that design wasn't enjoyed by nearly as large a fraction of the population as it is today, and that the range of products that were designed with aesthetics in mind also was more limited than today. What about the more mass-market design of the 1950s and '60s? She says that the choices available weren't nearly as varied and driven by customization as the ones we have today.

I've decided to take a different approach by looking at baby names, a program pioneered by Stanley Lieberson to investigate fashions (summarized in his excellent book A Matter of Taste). Unlike toilet brushes, cars, or paintings, names you give your baby don't cost money. They carry non-monetary costs -- you wouldn't name your kid Adolf, and it's not because you'd have to plunk down big bucks to register that name. Still, they are a lot closer to the ideal test of someone's desires where you ask them to "pretend that price isn't an issue."

Postrel sees the choice to endow an object with aesthetic qualities as "making it special." That has two consequences: it will tend to create turnover, as people start to make things special by making them different from what they were like yesterday; and it will tend to make things more diverse, as individuals try to look different from each other today. Lieberson has already provided graphs from many European countries on the turnover in names over time, going back centuries. The take-home message is that roughly before the industrial era, there was little to no turnover in names. (He measures turnover as the number of names in this year's Top 10 that were not present in last year's Top 10.) With industrialization, the fashion in names took off. The turnover rate really shoots up sometime in the mid-20th C., but it's not clear if it's something different or just the rapid-growth phase of a single, sustained exponential increase.

That's not too surprising, and it underscores the importance of not believing in "new era" stories. To a first approximation, there was the transition to agriculture, and then the transition to industrial capitalism -- and that's it. That's basically all we see in Lieberson's turnover data -- there is a pre-industrial stage with no turnover and an industrial stage with high turnover, and that's about all.

What about the diversity of names? Even if there were no turnover in the Top 10 names, we could still see people introducing more and more new names in order to make their children special. Recall that a big part of Postrel's story is that in our new aesthetic age, we really value standing out from others at the same point in time, not simply moving from one universal style to another over time. Lieberson does have some graphs on how highly concentrated names are, but the data are much more regional than national, they only cover some of the 20th C (although a decent chunk of it), and the measure isn't as precise as it could be. (He uses the percent of all names that are held by the Top 20 names, as well as the percent of all names that are given to only one person.) The rough pattern is that diversity of names seems somewhat static from roughly 1920 through either 1945 or 1960, and by no later than 1960 we see a tendency toward greater diversity.

To solve these problems somewhat, I turn to the Social Security Administration's archive of popular baby names. These are national data, so the sample size is a lot larger than Lieberson's regional data, and the regional differences are smoothed out. They also cover more time -- 1880 to 2008. And I've used a more standard measure of variation when the data are not quantitative (like names), as well as develop a measure of my own that is more flexible.

The standard measure I use is one of many ways to measure "qualitative variation," where the data are not numbers. It is thus impossible to compute how far a given datum deviates from the mean, and thus impossible to compute the variance. For data that are numbers -- height, IQ, etc. -- the variance tells us how spread out vs. how similar the values are. In the Appendix, I discuss what the two measures I use are, the standard one and my own. The only thing you need to know is that bigger values mean greater diversity.

With that out of the way, let's see what the historical pattern of conformity vs. diversity has been in aesthetic tastes. There are separate data for males and females, so I also show what the female - male gap has been:

First, note that the female line is always above the male line, a pattern that shows up in Lieberson's turnover data too. Parents give greater fashionableness to their daughters' names, and are less enthusiastic about their sons standing out.

There certainly has been a sharp increase during the past generation, starting around 1983 for boys and 1987 for girls. But that's nothing new. There was a more modest upturn from just after WWII until the early-mid 1980s. Moreover, there was another pronounced increase going back at least to 1880 (and probably before -- remember that the turnover data tracked industrialization), which fizzled out around 1905 for girls and 1910 for boys. From then until the post-WWII era, diversity actually stagnated or slightly declined. This trend obviously goes against what we'd expect from our view of the 1920s and '30s as a previous age of aesthetics. At least in the sense of "making special" as distinguishing things from one another in the same time period, the turn-of-the-century through the post-WWII era was conformist compared to the dizzying changes of the earlier industrializing stage or the turbulence that would follow.

In looking at the female - male gap, we are looking at another facet of culture -- namely, how egalitarian-minded people were with respect to sex (and perhaps in general). We see a narrowing of the gap from at least as far back as 1880 until 1915. From then until 1969, we see a general widening of the gap, although there are noticeable egalitarian dips during the depth of the Great Depression in the early-mid 1930s and during WWII. From 1915 until WWII, the widening gap is caused by boys receiving more similar names, with little change among girls. To me, that suggests traditionalism -- don't make girls' names any more fashionable, and drive boys names away from being fashion symbols.

From WWII to 1969, though -- i.e., during the Baby Boom -- the widening gap is caused by girls receiving more diverse names at a faster rate than boys are. This is a compromise between tradition and change -- clearly it's a break with the traditional values by making boys names more subject to fashion, but you don't notice it so much because girls' names are becoming even more subject to fashion.

Once the earliest Baby Boomers (born around 1945) reach child-bearing age (around 25), or in 1970, we see a steady reversal. Boomers wanted to equalize the outcomes of their own babies, and this was mostly due to slowing down the fashionableness of girls' names. That's what they tried elsewhere -- recognizing that females are more compliant than males, you should force females to be more like males rather than vice versa. Boomers were more likely to favor "bring your daughter to work day" than "give your son a dollhouse to play with."

There is a brief widening of the gap from 1987 to 1996, and this corresponds to the generation I've elsewhere called the disco-punk generation. They are too young to be canonical Boomers and too old to be Generation X-ers. I've estimated that they were born between about 1958 and 1964, so the typical member born in the early '60s would have started having kids in the late 1980s. This generation is very different from Boomers and X-ers, not having been ideological or attention-whoring when they were coming of age around 1980.

However, once the prototypical Gen X-er, born in 1971, starts having kids in 1997, the gap starts to close again. Remember, they were the ones who made Third Wave feminism a success.

So, the female - male gap shows us that our aesthetic preferences reflect larger social and cultural changes, such as how ideological we are about sex equality.

Even the overall data, which don't show such sensitivity to smaller-scale social changes, still reflect social change, although at a much larger level. The industrializing and globalizing stage was completed around the turn of the century. The change in the connectedness of the global economy was far greater from 1830 to 1920 than from 1920 to today. And most of the major innovations that came with industrialization were in place by then too. Those allowed design objects to reach a wider market, as well as greater customization with each passing year. We see this saturation of the industrial trends by the flat lines from roughly 1910 to 1945.

After WWII, we did start another round of changes away from manufacturing and toward services, and toward consumption of more "frivolous" goods and services than more basic ones, as the former became cheaper. What Postrel calls the new age of aesthetics really began in the mid-1980s, perhaps reflecting our greater taste for fashion items as we began to borrow more to buy more.

To conclude, we've found hard data showing that Postrel's hunch was right about the last generation being more and more aesthetically minded. But contrary to her "first time ever" theme, there was fantastic change during the industrializing stage too. Unfortunately the data don't go back to the beginning of that stage, which would probably show the trend even more clearly. But like I said before, that's the expected picture -- there's the shift to agriculture and the shift to industrial capitalism, and the rest are just hiccups.

Update: The historical pattern of rising diversity from the late 19th C through WWI, a stagnation or slight decline from then until WWII, and then another stage of sharp increase, mirrors the pattern of American openness to international trade. We became more integrated from the late 19th C through WWI, became more isolationist and protectionist from then until WWII, and then began reducing barriers to global trade. Perhaps there is some abstract "openness to differences" that characterizes the zeitgeist -- wanting to trade with people all over the world vs. economic nationalism, as well as wanting to explore a wider range of baby names vs. having more parochial tastes.


Here are the graphs using the standard statistic. The patterns are harder to see because the lines have to obey a ceiling of 1, whereas the statistic I used lets them go wherever they want as long as they're not negative. The female - male gap doesn't reveal the influence of the disco-punk generation, but the other patterns are there.

The standard statistic is:

(N / (N - 1)) * (1 - sum((share)^2))

Here, N is the number of names. The Social Security Administration gives the top 1000 names, so it's 1000. A name's "share" is what fraction of the sample has that name. We square each name's share, add them up, subtract from one, and multiply by N / (N - 1). This gives us the probability that two randomly chosen individuals will have different names. If it's 0, everyone has the same name, while if it's 1, everyone has a unique name. I include the graphs based on this statistic as an appendix since they show roughly the same patterns but not as clearly, given how constrained the possible values are.

I don't like this statistic because it's forced to lie between 0 and 1, whereas the variance of a distribution just has to be non-negative. We want to let the measure grow infinitely large in order to capture cases where the data are that spread out. I made up my own statistic, but there are probably others like it out there already; not having a PhD in statistics, I don't know what they're called. The idea is similar to the first one. Start with the shares, and see what would happen at the two extremes of "everyone is the same" vs. "everyone is unique."

Like the inventors of the first statistic, I hit on the idea of squaring the shares and summing them up. (It's the obvious way to go if you start down this path.) With an infinitely large number of names, the extremes are 0 for all-unique and 1 for all-same. By taking the log of this, we get new extremes of negative infinity for all-unique and 0 for all-same. Multiply by -1, and they all become non-negative values that increase as the diversity increases, just like a variance is supposed to be:

- ln( sum ( (share)^2))

The sum of the squared shares is the probability that a randomly chosen pair share a name, and we'll just label that p. Then my statistic, which I call the name diversity index, is:

ln( 1 / p)

Again, the key is that it's 0 when everyone is the same and tends toward infinity when everyone is unique, just like variance.

No comments:

Post a Comment