June 29, 2007
Posted by Jay Livingston
Sometimes when I was an adolescent, I would voice some opinion about teachers or sex or drugs, and an adult would say, “Just wait till you get older and have kids.” And indeed, the opinions of the people in the room often divided along age lines. My peers and I were more liberal than the grown-ups.
In the previous post on age differences in political views, I mentioned the difficulty of knowing whether differences were a matter of age or of generation. If it’s age, then people’s opinions will change as they grow older. But if it’s generational, then the members of that generation will retain their views forever. Imagine today’s twentysomethings in fifty years, still thinking their tattoos are cool, while the youth near them on the beach shake their heads in disbelief.
A survey like the one done by the Times/CBS/MTV shows a cross-section of the population at one point in time. But what we’d really like is “longitudinal” or “time-series” data that can show us what happens over the course of time.
Statistician Howard Wainer has an example that illustrates the dangers of drawing longitudinal conclusions from cross-sectional data. If you did a cross-sectional survey on language development in Miami, you might be tempted to conclude that when Miamians are young, they speak Spanish. As they mature into their middle years, they change to speaking English. And when they get even older, they switch to Yiddish.
Hat tip to my brother Skip for relaying this example.
the miami example is beautiful, jay/skip. i'll definitely appropriate it when next i teach methods.
ReplyDelete