Two current debates about generations and what they mean. First, Siva Vaidhyanathan’s “recent article”:http://chronicle.com/free/v55/i04/04b00701.htm in the _Chronicle of Higher Education_, expressing skepticism about the concept of “Digital Natives”:https://crookedtimber.org/2008/09/22/at-berkman/.
Gomez writes. “For this generation — which Googles rather than going to the library — print seems expensive, a bore, and a waste of time.” When I read that, I shuddered. I shook my head. I rolled my eyes. And I sighed. I have been hearing some version of the “kids today” or “this generation believes” argument for more than a dozen years of studying and teaching about digital culture and technology. … Every class has a handful of people with amazing skills and a large number who can’t deal with computers at all. A few lack mobile phones. … almost none know how to program or even code text with Hypertext Markup Language (HTML). Only a handful come to college with a sense of how the Internet fundamentally differs from the other major media platforms in daily life. College students in America are not as “digital” as we might wish to pretend. And even at elite universities, many are not rich enough. All this mystical talk about a generational shift and all the claims that kids won’t read books are just not true.
Second, Matt Yglesias on whether it’s important that the “kids love Obama”:http://yglesias.thinkprogress.org/archives/2008/10/youth_decay_2.php.
I used to sometimes think that the relatively left-wing views of the under-30 generation were basically just a reflection of the fact that the under-30 cohort contains many fewer non-hispanic whites than does the over-30 cohort. This new report from Amanda Logan and David Madlan makes it clear that’s not right — young whites have substantially more progressive views on a whole range of key issues than do older whites … if you hunt down a copy of the current issue of The Atlantic you should find … a piece by yours truly observing that the present day conservative coalition seems to mostly be stuck with the shrinking slices of the demographic pie. This data shows us one of the major driving factors behind that.
I’m quite skeptical about the ‘digital generation’/’digital natives’ argument. First, I’m skeptical because of the reasons that Siva outlines. Second, I’m skeptical because I and my contemporaries were supposedly members of ‘Generation X,’ a purported cohort which seemed to have been whistled out of a gaseous combination of bad sociology, mediocre novels and marketing concepts, and which certainly had nothing that I could identify as relevant to my own life experience.
I’m also uncertain as to whether Matt is right in suggesting that young voters’ identification with Obama is necessarily evidence of a secular shift. Political scientists who study party identification often try to distinguish between _cohort_ effects, _age_ effects, and _period_ effects – this “short paper”:http://www.stat.columbia.edu/~gelman/research/unpublished/apc.doc by Andrew Gelman is a good introduction. In Gelman’s words:
!http://www.stat.columbia.edu/~cook/movabletype/archives/27-4.gif!
If we were to make a graph, similar to Figure 1, but based on data from 2002, would the red and blue lines simply be shifted four years to the left, so that the party identification of 40-year-olds in 2006 matches that of the 36-year-olds in 2002? If such a pattern happened consistently over time, this would support the hypothesis of a cohort effect.
Another possibility that could be revealed by additional data is an age effect: Suppose that the equivalent to Figure 1, constructed four or eight or twelve years earlier, looked identical to the 2006 pattern with no age shift. Then we would be inclined to believe that party identification is associated with age, rather than cohort, with new voters starting out as strong Democrats, moving toward the Republican Party in their middle age, and then moving back to the Demcorats. It would not be difficult to construct a story consistent with this pattern (were it to in fact appear in the data), possibly associated with life-course changes involving marriage, children, and participation in the workforce.
Finally, the 2006 pattern may be part of a period effect. Figure 1 shows, overall, a strong Democratic advantage, but these polls were taken during a period when the Republicans have been on the defensive. Did the Democrats have such an advantage five or ten years ago? Maybe not. In this particular example, we are less interested in period effects—our primary goal here is to understand the big difference between today’s young and middle-aged voters—but we certainly have to be aware of the possibility of period effects, if only to adjust for them in estimating age and cohort effects.
The point is as follows: what Matt (and others) are suggesting is that there is an important _cohort effect_ – because young people are in a cohort that is disproportionately attracted to Obama, they are likely to start to identify as Democrats, and perhaps to continue to do so over their lifetimes, to the continued disadvantage of the Republican party. This is a plausible argument, and one that is quite compatible with “Bayesian models”:http://www.jstor.org/sici?sici=0190-9320(200206)24%3A2%3C151%3APSARPI%3E2.0.CO%3B2-H of party ID formation, which seem to me at least to be intuitively attractive. But it isn’t one that we _know to be correct_ and there is no very good way to disentangle these various effects from each other. High levels of Democratic identification among young people may also be the result of age effects (perhaps they will get crankier and more conservative as they get older) or period effects (this is clearly a terrible moment for the GOP). Disentangling these three effects is (as best as I understand the literature) damn near impossible given the kinds of data we have, and even tentative efforts to disentangle these relations require both lots of good longitudinal data and high powered statistical analysis. So _perhaps_ there will be a long term generational shift towards the Democratic party – but only perhaps.
{ 24 comments }
Gaucho Politico 10.07.08 at 8:53 pm
a couple things. Regarding the digital generation narratives i do believe that relative to previous generations the way my peers and i consume media is different than the way previous generations do. Print media outside of a campus newspaper is a very rare read. given the choice between picking up a print copy of a paper and the ecopy online the ecopy wins every time. I do not know that you can say that because everyone is not encoding in html that we are not sufficiently digital. I take it as more of an orientation in the way people were once oriented towards horses and then became oriented towards cars. My peers just are not going to be looking towards print. Of course levels of commitedness vary. i do doubt that you will find many people in my age group who cannot use a computer. Also the way we browse the internet might be different in some way do to the lack of cross contamination of print methodology and ideology and the time spent on the net would also be greater, in my estimation, than that of older americans. finally, i dont know anyone in my age group who does not own a cellphone — with texting feature.
As far as the long term shift towards the dem party for my generation as a result of a cohort effect, i posit that there will be a statistically significant effect going forward but that it may not be the huge effect people expect. The 8 years of Bush and the events therein will have a lasting impact on the country as a whole and the GOP brand has been tarnished for people who have just come into their formative political understanding. So this means that they have a softer but higher support for the dems as a baseline that will probably be eroded by the age effects and by subsequent period effects. expecting people to hold the same ideology of their 20’s in their 40’s without regard to intervening effects would be absurd.
Pete 10.07.08 at 9:11 pm
I’ll be sure to refer the next person who comes out with that tired line about socialism, over/under thirties and hearts/heads to that paper.
Generation X was a great novel, though.
JoshR 10.07.08 at 9:34 pm
If you read that Matt links to, you can see that it compares young people’s attitudes to the attitudes of the same age demographics in the past. This seems to rule out an age effect, although a period effect.
lt 10.07.08 at 10:05 pm
Upon a cursory glance, the graph seems to support at least an element of the cohort effect: the Republican spike in the fortysomethings corresponds to folks who came of age in the 80s; where as the fifty and sixtysomethings did so in the 60s and 70s. It’s plausable that mid-career parents/homeowners are a more conservative group than either the older or the young, but it would seem difficult to find another reason for such a sharp distinction around age 50.
sunship 10.07.08 at 11:14 pm
seems that your question is just a reformulation of hypotheses from Zaller’s (1992) RAS model. i can’t recall the tests coming out of that book or subsequent studies based on it.
lemuel pitkin 10.07.08 at 11:33 pm
Is it really so difficult?
The Washington Post has a nice table of party ID by cohort and year. I’ve simplified it below by giving just the net Dem number (hopefully this formats ok.)
What you see here pretty clearly is that there are no significant age effects — every age range is both the most and the least Democratic at some point between 1972 and 2006. Seems like we can reject the hypothesis that there is a stable relationship betwen party ID and age, e.g. for life-cycle reasons.
On the other hand, it does look like there might be a significant cohort effect. here the high-powered statistics would be helpful but it’s hard not too notice that the most Democratic age group is 18-29 in 1972-74, 30-44 in 1984-86, and 45-59 in 1996-98. Similarly, the most Republican age group is 18-29 in 1986-90 and 30-44 in 2002-04.
Is there some reason this naive reading of the data is wrong?
18-29 30-44 45-59 60+
1972 21% 13% 12% -8%
1976 20% 15% 14% 13%
1980 17% 13% 17% 10%
1982 14% 12% 17% 18%
1984 0% 6% 5% -1%
1986 0% 10% 6% 2%
1988 0% 3% 1% 7%
1990 -5% 4% 3% 5%
1992 -1% 1% 4% 11%
1994 -3% -4% 3% 5%
1996 11% 1% 6% 4%
1998 1% 1% 9% -2%
2000 1% 2% 5% 7%
2002 -2% -8% -1% 4%
2004 2% -6% 3% -2%
2006 12% 0% 2% -1%
lemuel pitkin 10.07.08 at 11:34 pm
Ooops, let’s try again:
Year 18-29 30-44 45-59 60+
1972 21% 13% 12% -8%
1976 20% 15% 14% 13%
1980 17% 13% 17% 10%
1982 14% 12% 17% 18%
1984 0% 6% 5% -1%
1986 0% 10% 6% 2%
1988 0% 3% 1% 7%
1990 -5% 4% 3% 5%
1992 -1% 1% 4% 11%
1994 -3% -4% 3% 5%
1996 11% 1% 6% 4%
1998 1% 1% 9% -2%
2000 1% 2% 5% 7%
2002 -2% -8% -1% 4%
2004 2% -6% 3% -2%
2006 12% 0% 2% -1%
lemuel pitkin 10.08.08 at 1:18 am
OK, one more. This one gives the difference between net Dem identification for that age group, and net Dem identification for all voters. The patterns here are awfully close to what you would expect to see from a cohort effect, given the 12-year gap between the first two age groups and the 15-year gap between the next two.
1972 is the most Democratic year for the youngest group, 1984-86 are the most Democratic for the 30-44 group, and 1998 is the most Democratic for the 45-59 group. The youngest cohort shifts from relative Democratic to relatively Republican in 1982, the 30-44 group shifts in 1992, and the 45-59 group shifts to Republican in 2006. The most Republican year for the youngest group is 1990; the most Republican year for the 30-44 group is 2002. And so on.
I don’t claim this little exercise proves anything, exactly, but you’ve got to admit this sure *looks* like a cohort effect…
Year 18-29 30-44 45-59 60+
1972 12% 4% 3% -18%
1976 5% 0% -2% -3%
1980 3% -1% 3% -4%
1982 -1% -3% 2% 3%
1984 -3% 4% 3% -4%
1986 -5% 6% 2% -3%
1988 -3% 0% -2% 4%
1990 -7% 2% 1% 3%
1992 -5% -3% 0% 7%
1994 -3% -4% 3% 5%
1996 6% -5% 0% -2%
1998 -1% -1% 7% -4%
2000 -3% -2% 1% 3%
2002 0% -6% 1% 6%
2004 3% -5% 4% -1%
2006 9% -3% -1% -4%
sunship 10.08.08 at 5:11 am
lemuel, the problem is that you’re trying to make individual level inferences based on aggregate data. you have no idea what happens to the positions of individuals within age cohorts. there could be event effects, life cycle effects, etc.
Steve Casburn 10.08.08 at 5:21 am
Henry: Gelman’s chart estimating party identification by age cohort indicates that the most Republican cohort is that born in the early 1960s.
Barack Obama was born in 1961.
A possible explanation for why Obama speaks so well across the ideological divide: He grew up as a liberal in a time of conservative reaction, and he chose to learn how to modulate his rhetoric.
lemuel pitkin 10.08.08 at 6:06 am
lemuel, the problem is that you’re trying to make individual level inferences based on aggregate data.
Well, yes. What’s wrong with that?
you have no idea what happens to the positions of individuals within age cohorts. there could be event effects, life cycle effects, etc.
Well, no.The aggregate data are enough to reject the idea of life cycle effects. There clearly is not a relationship between age and party ID, at least not one that is strong & stable enough to matter. As far as whether there is a cohort effect, you’re right, this is just suggestive. With more detail, it’s possible the apparent cohort effect would disappear. But it still tells us something.
By the way, the Gelman paper Henry links to discusses doing exactly the same analysis: estimating age, cohort and period effects using just annual breakdowns of party ID by age. So obviously he (and presumably Henry) don’t agree that such an analysis leaves you with “no idea” about the underlying dynamic.
John Quiggin 10.08.08 at 6:11 am
I had a go at the generation game here and cohort effects here
novakant 10.08.08 at 6:44 am
I’m quite skeptical about the ‘digital generation’/’digital natives’ argument. First, I’m skeptical because of the reasons that Siva outlines. Second, I’m skeptical because I and my contemporaries were supposedly members of ‘Generation X,’ a purported cohort which seemed to have been whistled out of a gaseous combination of bad sociology, mediocre novels and marketing concepts, and which certainly had nothing that I could identify as relevant to my own life experience.
Your scepticism is warranted, teens read more books than ever, cf. e.g.:
Teens buying books at fastest rate in decades
Who says teens don’t read?
Generation R (R Is for Reader)
The assumption that a new medium necessarily inhibits the consumption of older media is fundamentally flawed, because most people consume several media simultaneously and media usage as such has increased immensely. For example TV hurt the film industry for a while in the fifties and early sixties, but it didn’t have any negative long term effects.
novakant 10.08.08 at 7:00 am
Also, one has to be very careful with ascribing attitudes to “generations”, unless one is able to provide a very fine-grained analysis: my brother is barely 4 years older than myself, but when we were teenagers, he was still very much influenced by the culture and values of the 70s, while I regarded all of this as terribly outdated hippy stuff and was fully a child of the 80s.
terence 10.08.08 at 7:24 am
the other thing is that, even if there is a cohort effect, it doesn’t mean we will necessarily entering an age of Democrat dominance. As likely is that we’ll simply see a shifting of the political centre as the Republicans move left (relatively speaking) to stay in contention. Triangulation, in other words.
cosma 10.08.08 at 11:53 am
If you allow for age and period and cohort effects, then your model isn’t identifiable, because any one of those variables is a function of the other two. See Fienberg and Mason.
c.l. ball 10.08.08 at 1:08 pm
The “digital literacy” of youth cohorts is over-stated. Many cannot even figure out how to insert footnotes into papers, let alone the simple mark-up code to insert a hyper-link into a discussion-forum post. I had immense difficulty getting many students to understand that accessing JSTOR from the on-campus network would be different from accessing it from an off-campus ISP. The rise of clunky and ugly course (mis)-management software like WebCT is testament to continued, widespread digital illiteracy among students and faculty.
Mikey in Plano 10.08.08 at 2:27 pm
The British Library and JISC released a report earlier this year on the “Digital Natives” generation. Or lack thereof.
lemuel pitkin 10.08.08 at 3:45 pm
Cosma,
Interesting, thanks. But you can still ask if cohort or age effects (or neither) explain the differences in party ID between different age groups, right? Especially if, as Gelman discusses, you constrain the effects to some space of the “reasonable”.
Would you agree that the patterns in the data from the Washington Post site offer some (weak) support for a cohort effect, and (stronger) grounds for rejecting a meaningful age effect?
arbitrista 10.08.08 at 5:15 pm
The work on political psychology seems to suggest that there are “impressionable periods” that shape cohorts with period effects, but that age tends to lead to a hardening of political loyalties – not a conservative trend.
ElectionWatcher 10.08.08 at 9:09 pm
The 2008 election is particularly generational in nature, and one of the key componenets of this is the emergence of Generation Jones; born 1954-1965, between the Boomers and Xers. GenJones, which is 29% of the electorate, is the most GOP-leaning generation, so is important to factor into a discussion 0f the changing demographics of conservatism. Here’s a compelling 5 minute video which discusse these issues while looking at the key role GenJones is playing in this election, the video has a bunch of high profile pundits and experts discussing this issue, and can be found here: http://www.youtube.com/watch?v=1Ta_Du5K0jk
clew 10.09.08 at 12:58 am
“Many cannot even figure out how to insert footnotes into papers, let alone the simple mark-up code to insert a hyper-link into a discussion-forum post. I had immense difficulty getting many students to understand that accessing JSTOR from the on-campus network would be different from accessing it from an off-campus ISP.”
But when we had to type them, many students couldn’t do footnotes correctly. (Remember working out the line count? And recasting sentences to make room? Ow.) Of course, your average person resents having to do footnotes, which probably slows mastery.
On the other hand, I’d expect GenP2P to understand being on one side or the other of a host, so the ISP problem is more confusing…. Unless ‘I couldn’t get to JSTOR’ is the new ‘dog ate it’.
Zephi Friel 10.09.08 at 1:40 am
I am fascinated by fields that attempt to deduce why people act the way they do. Statistics, economics, and psychology all have opened my mind to look at information in a different way and to think about causation. In someways, the “average” college kid is kind of predictable- we’re ideological, impulsive, quixotic. Most would like to think about the world the way they think it should be, because they haven’t yet had to fully support themselves. The political atmosphere during the time one is growing up does have a huge effect on the developing mind and may create more democrats for the future- but I wouldn’t “bank” on it. I think these trends have more to do with when one is in their life. Reality will hit sooner or later, and college kids will then enter the job market and see how much of their pay check is taken out by the government. They’ll feel disenfranchised when they realize what fighting the system really is like.
I agree with Henry’s skepticism on the two articles. These analyses examining the identity perspective of different demographics do bring up interesting points, yet I am not convinced that there will be more democrats in the future that still avoid computers. I love reading, and I don’t think books are going to go extinct- but I do think that it will be necessary for everyone to use the internet and computers. Computers are necessary for anyone that has a hope for a decent job.
Bryan 10.09.08 at 8:00 am
“Only a handful come to college with a sense of how the Internet fundamentally differs from the other major media platforms in daily life. College students in America are not as “digital†as we might wish to pretend. And even at elite universities, many are not rich enough. All this mystical talk about a generational shift and all the claims that kids won’t read books are just not true.”
yeah that would be a really good argument if they showed that the people that are not ‘rich enough’ and the people that don’t understand how the internet fundamentally differs from other major media platforms are also the parts of the population most likely to read book. What I remember from statistics I have seen there is a correlation between reading and being relatively well to do, and furthermore in my personal experience the people who understand that the internet fundamentally differs are also the people most likely to read. For these reasons I think the contrarian argument that the young will continue reading books despite the internet is uh, bullshit.
Comments on this entry are closed.