One of the problems with trying to do age sampling from the exit polls is that no two polls use the same age brackets. I have combined things in ways that make sense to me, but be sure to examine the actual charts on the MSNBC site. Jerome Armstrong has done an analysis of age representation in the primaries, along with a few other trends, and notes interesting things. First, the youngest age cohort, 18 - 29 year olds, have declining participation levels. He doesn't include Michigan in the mix, but the age turn out there is in keeping with his findings, namely that it is less than New Hampshire and greater than Nevada. He also notes that people self-identified as "Very Liberal" are now voting more for Clinton than for Obama, and thinks there is a correlation between fewer younger voters and fewer Obama-aligned Very Liberal voters. There is no doubt a connection between fewer young voters and Obama's declining vote share, but the self-identification is a bit of a red herring.
One trend he doesn't note about the youth vote is that HRC has tripled her % claim on it, evidently pulling from the other candidates. In IA, Obama got 57% of the 29 and under vote, while HRC garnered only 11%. By NV, HRC had raised her proportion to 33%, while Obama had increased by only 2, to 59%. The share of youth voters as a percentage of the whole is going down, while Hillary's share of that slice is going up. I suspect that what we're seeing is attrition from prospective Obama voters. Compared to the percentages in IA, Obama lost over 5,000 votes from the youngest age cohort in Nevada - they just weren't there.
In the other age groups, much attention has been given to the elderly vote, but in truth the numbers for voters over the age of 40 have been increasing since Iowa. Here is a little counterintuitive information. Obama is increasing his support among senior voters. The trouble is he isn't substantially increasing his percentage of any other age cohort. The table below shows, once again, that HRC is the one benefittting from John Edwards' loss of support. She is steadily increasing her percentages from Iowa:
I have asterisked the Michigan line because Edwards and Obama support is merged there. In truth, I suspect Michigan shows that Edwards supporters easily moved over to HRC's column, and that most "Uncommitted" votes cast were from Obama supporters. I also think that the numbers in Michigan were lower than Obama would have earned had his name been on the ballot directly. Edwards was definitely the big loser for having taken his name off. It may have been the second most fatal mistake of his campaign, the worst being joining ranks with Obama to go after Hillary, which damaged Hillary but did not benefit Edwards at all.
Without a solid set of fully comparable primaries (and I'm afraid we won't get that until Super Tuesday itself), this is more than a bit reading tea leaves. What I think the numbers do show is, again, how anomalous Iowa was compared to the other contests. Without Edwards in the running, I think Hillary would have trounced Obama soundly in Iowa. She did not lose votes to Obama (though she did lose second round support from Richardson, Biden and Dodd voters, probably due to Obama aggression on the caucus floors - remember, that's totally legitimate in a caucus situation), she lost them to Edwards.
What I'm also not able to really show in these tables is the big difference between the number of voters in one age cohort within a given contest. For example, (more counter-intuitiveness), the 65 and over crowd had their second greatest proportional representation and their largest turn out in raw numbers in Iowa, something that listening to the main stream media would not be obvious. Had they voted for HRC in the same numbers there as they voted for her in Nevada, for example, she would have won Iowa. Where Hillary pulled out her victory in New Hampshire was among working age people 40 to 64. They were more than half of all voters and she beat Obama by 10% with them. While her percentage with retirees is high, the bulk of the votes come from middle-aged voters. Nevada is the only location where seniors outnumbered the middle-aged cohort, and only by 4600 voters out of 115,000+. That was less than her margin of victory.
The big picture here is that Obama is increasing his margins a little bit from his Iowa showing, but nowhere near as much as Hillary has done, particularly among young voters. I think South Carolina will look a lot like Iowa, with Edwards taking away a significant number of HRC votes, and Obama's showing among AA voters comparable to his margins in Iowa.