I'll have to go with my self as primary source, I'm going to bet the census was the original source's source though, they compile data on it, try googling census with P20-562, but the orginal was years back. I confirmed it from my own data (I crunch data for the local GOP) and spot checked against a few other places in my state and a few random other places elsewhere. You could actually compile everywhere but I don't know if anyone keeps a giant master database, usually just each board of elections and most of those do let you download those from their webpages. We can run the figures on the Dems for my own county but I've never had a specific interest in it. Educated guess off my own county and our demographic tilt toward older democrats, would be that the average Dem voter in the country during a presidential year would probably be about 50-51.

Reminder by the way for you or anyone else reading this, median vs average, median means half the people who voted were over that age, average would be the sum divided by the number. In context, if you feel like downloading your own area, find the column that list birth year and the one that lists votes for the primary year you're interested, paste those into a new sheet unless you've got a real small county or a real fast processor, sort by birth year then sort by letter for that primary, and just find the start and end points for the r's and the d's, and see what year's in the middle, like entry #5000 if you had 10,000 D votes, then subtract that birth year from the primary date - median voter age. Only since they're in the spring you might want to subtract one from your tally if that median year was close to the next year as opposed to the earlier one, birthday's and all. Simple enough, but damn time consuming in mass.

I can tell you what the median 2008 and 2010 age for my own county was for Dems since I already have that data. For Dem 2010 it was also 1950, in context 59 years, but in 2008, way more votes, and that dropped to 1954 bordering on 55, from 2008 as opposed to 2010, or 53 years old. Same year same county GOP, it was 1951 or 57 as opposed to 60 in 2010. So, unsurprisingly, voter age drops in presidential years. Again that's just my county, but I would expect that to hold steady. Two other fun bits, median voting age 2010 general election, 57, med voting age gen 2008, 50 - and a jump from lower 30 thousands to mid 40's in terms of voter count, again just my county, and that was normal about 50% higher turnout in 2008 over 2010. IIRC though the median voter age for 2008 for the whole country was 44, but I believe that was probably by average registered voter age, as opposed to average or median age of people who voted, so grain of salt. Note the value we got for my county's dem 2008 was higher than my dem 2008 nationwide guesstimate, but that probably is as it should be, old rustbelt area with union/factory background, we do have more older conservative dems locally.

One fun fact because I had the data up again and was bored, again specific to my county but probably good plus or minus a year to the country, subtracting registration data from birth year got me an average of 33, and a median of 30. So most people, of those who did register, didn't register until they were 30. Depressing huh?

One last note of caution, party affiliation in my data is determined only by which ballot they picked if they voted in the primary, and that's normally the case, but a lot of younger people don't vote in primaries even though many have very partisan views or would self-describe as one. If you boil it out, your youngest affiliation will tend to be independents. Long term analysis is also a touch tricky, it fluctuates by presidential, even, and odd years a lot and four years is more than enough time for our average life expectancy to jump a year and the demographic effects of aging boomers to factor in. So massive grain of salt.

Doubtless that was more info then you possibly could ever want on the subject, and if not then you are likely a masochist.