When I want to find out how a particular characteristic has changed over the generations, I turn first to data collected over time. That way, we can be sure the differences are not due to age or to people misremembering what they were like when they were young (how many parents have fudged a detail or two about their own teenage years?) This approach makes Generation Me unique among books that discuss generations, because it summarizes psychological data — and a very large amount — collected at various times. I'm not looking at the generations as they are now, with Boomers middle-aged and retired and GenMe in youth and rising adulthood. Instead, I've found data on what Boomers were like when they were young in the 1960s and 1970s and compared it to data on young people from the 2000s and 2010s. Most surveys covered in the media — even those from great organizations such as Gallup and Pew Research — are one-time surveys that can't separate the effects of age and generation. Sometimes those surveys have data going back a few years, but they rarely go back to the 1960s or 1970s. Thus it's difficult to say whether differences in attitudes are due to age or generation from these polls.
Without a time machine, though, finding over-time data is difficult. There are usually two choices. The first choice is large surveys that have been conducted over time. These have recently become much easier to access, with the data files or research reports posted online. The three I draw from most often are Monitoring the Future (of 8th, 10th, and 12th graders), the American Freshman Survey (entering college students), and the General Social Survey (adults over 18). These surveys are also very large, with several thousand participants each year. That's an improvement over Pew and Gallup polls, many of which have only a few hundred respondents. The downside is that these surveys have measured a limited number of topics, and can only measure traits and attitudes recognized when the survey began. For example, they don't include measures of narcissism, as that trait wasn't studied much until the 1980s.
The second technique is called cross-temporal meta-analysis. I begin by searching computer databases for journal articles, master's theses, and dissertations that used a particular scale. I keep only those that used a normal population of a specific age — usually children or college students. Then I search to find the data sources at the library or in full text databases online, since only the entire article or thesis will have what I'm looking for: the average score of the sample on the questionnaire. Once I find all of the data, I can then graph those scores by the year the data were collected. Because the samples are roughly the same age (say, college students), this shows how young people differ from one generation to the next. No one had ever done this type of analysis before, so I started from scratch developing a way to find and analyze the data. The resulting analysis usually has two main numbers: the correlation between the average scores and the year the data were collected, and the effect size based on the individual-level standard deviation. (Which means that the effect size is NOT based on ecological correlations).
Other researchers have used cross-temporal meta-analysis as well, finding generational differences in empathy, belief in a just world, and attachment styles. All of these studies — and mine — are based on U.S. samples, but researchers in China have started to see generational differences there, too — for example, increases in anxiety among adolescents.
You might wonder if the changes in the questionnaires happen because people now have fewer qualms admitting to problems. However, the questionnaires across all of these studies were given on paper and not in interviews, and they're anonymous — respondents don't put their names on them. Most ask about specific symptoms ("Some unimportant thought runs through my mind and bothers me") rather than asking point-blank something like "Are you anxious and depressed?" The responses to all of the symptoms are added up to form the score, so the respondent only admits to small parts of a problem at a time. In addition, the changes described in the book are diverse. Some of them are in "good" traits (like self-esteem), but others are in "bad" traits (like anxiety), and some (like who controls your fate) have questions worded so there is no obvious "good" or "bad" answer. If people were more comfortable admitting to bad things, we'd expect to see change only in traits that are considered undesirable, but the changes show up in all kinds of characteristics. In two studies, we also directly accounted for defensive and socially desirable responding, but the generational differences still appeared. Learn about study 1 and study 2.