This interactive from the New York Times shows you very clearly how people have voted in terms of U.S. presidential elections by birth cohort. Confusingly, it shows voter's simulated preferences based on statistical models even before they're 18, which is slightly confusing at glance--just pay attention to what's going on around the 20 year line. An important note is that this model is for American-born, White people only. As you can see, most cohorts do in general have a higher tendency of voting Republican when in middle age and beyond, but that also complex things happen. This particular model (and this is statistical inference, not exactly how people expressed preferences in surveys) argues that most people's preferences are most malleable between 14-24, and that little that happens after age 40 affects how cohorts vote. One of the more interesting things is that the cohort born roughly during the 1970's has gotten slightly but consistently less likely to vote Republican as they age. These people came of age during a period where Republicans were triumphant, and chilled a little over time. You can see a slightly similar things in the 1940's, for people who came of age in the Eisenhower era.

Set the dial to, say, 1974 and notice that this cohort of white Americans is less likely to vote Republican today than they were when they came of age in 1992. It's a hard chart to see the ideological splits by age at any particular point in time (since they are comparing people by at specific ages rather by election), but it shows you how cohorts can change over time. What it doesn't show you is how much recent voting have been effected by demographic changes, such that Bush in 2000 was able to beat Al Gore among White people 55% to 42%, but only barely win the election. Compare that to the last close election without a strong third party candidate (Ross Perot, you make discussing presidential voting trends more difficult), the 1976 election, where Carter won 50% to 48%, and Ford won White people only 52% to 48%. Hispanics and Blacks went very strongly for Carter, there were just many fewer Hispanics voting. In 1976, Hispanics made up 1% of voters and Blacks made up 9%. In 2000, Hispanics made up 7% of voters, Blacks made up 10%, and Asians made up 2%. I know you're not asking about age, not race, but I feel like it's important for understanding the above model which is only based on White voters, otherwise the numbers wouldn't seem to add up (As a note to this note, Bush also did particularly well with Hispanic voters, compared to other recent Republican candidates, getting close to 40% of the vote. In a three-man race in 1996, 73% of Hispanics voted for Clinton).

If you look at the Roper Center data (an easy way to look at voting in each election by age group), you see that 18-24 were more likely to vote for Reagan than 25-29 or 30-49 year olds, though just as likely as 50-64 year olds and less likely than people 65+. According to the above theory, they came of age during Reagan's tremendously popular presidency and so were permanently effected by his high approval ratings, but the effect dissipated slightly over time. You see little peaks and valleys like that--Dukkakis and Clinton were most popular with the young and very old, for example (perhaps in part due to old Southerners continuing to vote Democrat as younger cohorts moved more fully into the Republican fold--if so, this would be a measure of partisanship rather than ideology--but I am not sure exactly what explains that), with a distinct dip in support among the 45-59 year olds.

But yes, for the modern polling era, in America, at the presidential level, the general trend is that the young as a whole vote to the left of the old as a whole, even if there are fluctuations within the subgroups. Whether you want to say the polling era began in the 1960's or the 1970's or even in the 1950's (Gallup proved the success of telephone polling using random digit dialing in the 1948 "Dewey defeats Truman" election, but I don't think they collected or at least released subgroup data for that election), in the second half 20th century we start to get good data about how subgroups of the population vote. Before that period, I'm not sure it would be easy to tell which way the youth as a whole were voting, though we could see which movements had more youth activists, which is likely a decent proxy, though harder to collect and interpret than already published poll results.