Do early impressions create lasting voting behavior? The answer, according to David Leonhardt at the New York Times's Upshot blog, is yes--up to a point.

He draws on research from political scientists Yair Ghitza and Andrew Gelman, on the basis of which his NYT colleague has prepared a terrific interactive graphic from which you can see how white Americans born in each year from 1937 to 1994 have been voting over the course of their lives. Blacks are excluded since their voting behavior has not changed significantly, at least since 1964. That covers almost everyone here, since only blacks born in 1943 or earlier could vote in almost every state by that date, and few blacks were allowed to vote in most Southern states before then.

It’s not clear to me whether the data points represent voting for president or a more generalized political preference. Probably the latter, since when you look at voting behavior by birth year you see narrower oscillations than occurred in presidential voting in some of these years, as when Americans voted 61 percent Democratic for president in 1964 and 61 percent Republican in 1972, eight years later. In addition, no birth year cohort came in as more than 56 percent Republican in 2012, even though the exit poll showed Mitt Romney carrying whites with 59 percent of the vote.

Nonetheless, Gitza and Gelman put a presidential focus on their findings: “Generations appear to be formed through a prolonged period of [perceived] presidential excellence.” This seems right, but again only up to a point.

To get a sense of how people born in different years have differed, I looked at the percentages Republican or Democratic for those both in each year in the election of 2012 (the graphs show a percentage only for the majority party, and I calculated percentage Republican, assuming that 1 percent did not vote for either party). Viewed that way, the Republican percentages by birth year varied from a high of 56 percent (birth years 1964, 1965, 1966, 1968) to a low of 45 percent (birth years 1984, 1985).

Party preference doesn’t change much from birth year to birth year, usually remaining the same or changing by just 1 percent. But you do see some flexion points. Republicans predominated in 2012 among those born in birth years 1937-48 and 1955-80, Democrats among those born 1949-54 and 1981-1994. These dates tend to support Gitza-Gelman’s theory that perceived presidential performance shaped attitudes, but not entirely: those born 1937-48 turned 18 not only during the second Eisenhower presidency but in the Kennedy-Johnson years, including time when those Democratic presidents had high job approval; those born 1949-54 turned 18 mostly during Richard Nixon’s years in office, including some when he had high approval as well.

The biggest flexion point—the biggest difference between one birth year and the next—is between those born in 1954 (49 percent Republican) and those born in 1955 (53 percent Republican). That’s double any other difference between adjacent birth years. What accounts for that? I think it’s this. There was a military draft in 1972, the year those born in 1954 turned 18. There was no military draft in 1973, the year those born in 1955 turned 18.

I also note that the 2012 Republican percentage slipped below 50 percent among those born in 1949, who turned 18 in 1967, as we were ramping up to the maximum troop numbers in Vietnam. The 2012 Republican percentages were lowest, 48 percent, among those born before the Millennial generation in those born in 1950, who turned 18 in 1968, the year of the Tet offensive and the maxmimum troop building; and those born in 1952, who turned 18 in 1970, the year of the controversial incursion into Cambodia. Birth year 1952ers have been Democratic ever since; birth year 1955ers have been Republican, with an exception around the time of Jimmy Carter’s election in 1976, ever since. It’s quite a startling difference between people born just three years apart — and in the middle of the Baby Boom generation, which is generally regarded as a homogeneous blob.

My hypothesis: The draft moved many of those subject to it to an antiwar position in those years and to a much greater likelihood of supporting Democrats ever since. That’s ironic, because the maximum draft calls and the escalation of U.S. military involvement in Vietnam came by the decisions and acts of Democratic presidents, and the draft was abolished under a Republican president (and with a big push by a young Republican congressman from Wisconsin, William Steiger, who also was the major force behind the cutting of the capital gains tax in 1978).

But it is also true that the Democrats emerged from the Vietnam years as the party more likely to oppose military interventions — the opposite of what they had been for the half-century starting in 1917—and the Republicans as the party more likely to favor military intervention. And if, as I am hypothesizing, the pendency of the draft led many young men — and women — to oppose military action, out of patriotic reflection or personal rationalization, that meant that they would be more likely to support the Democratic party in the years after 1975.

I doubt that many younger people appreciate the impact of the draft on young people’s lives in those years. The availability of college and professional deferments diverted many into career choices they might not have otherwise made, and the availability of deferments for fathers diverted at least some into fatherhood (and of course affected, perhaps even more, the women who became mothers). Thus a single public policy—the draft—may have had more effect on many voters than perceived presidential performance.

A final note. Those born in 1994, the last birth year eligible to vote in 2012, came in as 47 percent Republican. Not a good mark for that party, but above the party's 45 percent nadir among those born in 1986 and 1987. I would add this to the other evidence, like these Guardian blogposts by blogger Harry Enten -- not quite as fragmentary as Leonhardt suggests, but far from conclusive either -- that younger Millennials may be less influenced by negative assessments of George W. Bush than by negative assessments of Barack Obama.

Indeed, Leonhardt’s blogpost is headlined, “Why Teenagers Today May Grow Up Conservative,” though the text itself places more emphasis on the “may” than the headline does. Have the slow-growth economy and the failure of big government programs like Obamacare reduced Millennials’ faith in big government to an extent comparable to the impact of the military draft on the attitudes of Baby Boomers?

We’ll see. The lesson here, I think, is that particular public policies that have a profound personal impact can have as much effect on political attitudes — or even more — than assessments of presidential performance.