I'm a little suspicious of most generational analyses, but leaving generic doubts aside there certainly is an interesting pattern in the data shown in the chart accompanying David Kirkpatrick's Times Week in Review article. You see a kind of cyclical pattern where an age cohort with lots of Democrats is replaced by a cohort with few Democrats. The interesting thing is that this isn't the sort of pattern where you have kids rejecting their parents' fashions. Instead, it looks like twentysomethings, our fiftysomething parents, and their parents are all unusually Democratic while the interstitial cohorts are the unusually Republican ones.
On the other hand, one thing the chart pretty clearly shows is that party identification doesn't actually tell us very much. Absolutely every age group except 35 and 36 year-olds contains more soi disant Democrats than Republicans. Nevertheless, Republicans clearly win plenty of elections.