Friday, February 29, 2008

Sources of Democratic Voting Gaps

Following up on my last post, I thought it might be interesting to look "under the hood" of the Democratic voting gaps presented earlier. While it takes the behavior of two groups to make a "gap" there is no reason to assume that those groups favor their preferred candidates to the same degree. Indeed this is borne out by exit poll data from the 2008 Democratic primaries and caucuses:

The pattern so far in the season is pretty clear: groups supporting Barak Obama break to him much more decisively than Hillary Clinton's groups break to her.

Note: the data in this figure are based on averaging the voting gaps across states. The normal caveats about exit polls (earlier posts) apply.

Monday, February 25, 2008

Gapology: The Democrats

As promised, I've put together some data on voting gaps in the Democratic primary electorate. I'll present the same for Republicans at a later date.

Again, a couple of caveats are in order. First, these data are only for those primary/caucus states for which exit/entrance poll data are available. Second, in some instances, the gaps may seem substantial but are based on relatively few cases for at least one of the categories. For instance, Iowa had a large race gap (below) but the Democratic electorate was only 4% African-American. In these cases, the overall impact of the "gap" may not be as important as it appears.

First, the oft-referenced gender gap.


Here we see a consistent pattern of Clinton faring better among female voters and Obama faring better among male voters. It is important to point out, though, that this does not mean that Clinton won the female vote, nor that Obama won the male vote, in every case. It also doesn't mean that large gender gaps always help Senator Clinton. In fact, the largest gender gap thus far was in Wisconsin (from whence this blog is written), where Clinton and Obama split the female vote and Obama won the male vote by thirty-six points.

As expected, there is also a substantial race gap.


The figure above shows a persistent and quite large gap in candidate support between White and African-American voters. While Whites(on average, across the states where both sides of the gap can be measured) favor Clinton by approximately 17 percentage points, African-Americans favor Obama by 63 percentage points.

The generation gap is also very real, with Clinton faring best among older voters and Obama drawing greater support from younger voters.


I should mention here that the age groups used above are not always the same. In a couple of states, the young group was 17-29 and in a couple of others, the older group was 65+.

Finally, we turn to the income gap.


As per media accounts we see Clinton running best among lower income groups and Obama running best among higher income groups, with just a couple of state showing the opposite pattern. Again, though, this does not mean that Clinton consistently won among low income voters or that Obama consistently won among high income voters; just that they tended to do better among those groups.

The figure below summarizes the average size of the voting gaps:


The race and generational gaps are clearly most substantial, followed gender and income. I'm not sure how this fits with the media accounts of the race. Any thoughts?

Thursday, February 21, 2008

A changing electorate?

One of the oft-repeated themes of media coverage of the campaign this season is the changing complexion of the electorate, especially on the Democratic side. A couple of groups--independents and young voters--have received a lot of attention, and there have also been claims about increased participation of women and African- Americans.

It's difficult to assess the actual turnout rates of various groups, in part because it is difficult to assess the both numerator (how many young people voted?) and the denominator (what is the size of the18-29 year-old Democratic--and independent?-electorate?).

We can, however, examine the share of Democratic voters who are 18-29 years old and then compare that number to the 18-29 year-old share from the previous election. An increase in this share indicates a greater increase in turnout among 18-29 year-olds, relative the change in turnout among other age groups. And, of course, the same method can be used for other demographic groups.

The data posted below come from an analysis of the 2004 and 2008 Democratic primary and caucus exit polls. Because I am using exit polls, the analysis is limited just to those states in which the 2004 and 2008 exit polls were administered.

Here's a look at the change in the youth vote:














Generally, across the states, young voters constitute a larger share of the Democratic electorate in 2008 than they did in 2004. Interesting that New Hampshire, site of Senator Clinton's rebirth, is the only state that saw a decline in young voters as a share of the electorate.

Now, what about the other much-talked-about group, political independents?














Not so much, huh? Sure, there was an increase in some states, but not in most states.

What about female voters?














Pretty much an across-the-board increase in women as a share of the Democratic electorate.

Just "eyeballing" the data I don't see any clear relationship between changes in participation and candidate success.

Changes in the African-American share of the electorate:














Increases in most states, though not across the board.

Again, I don't see a clear pattern here between changes in the electorate and candidate support.

Finally, let's take a look at the overall patterns. The figure below shows the average change in groups as a % of the electorate, across states, from 2004 to 2008.














On balance,the greatest changes in participation have occurred among young voters and women, then among African-Americans, and there is virtually no average increase in independents as a share of the electorate.

An important caveat: these data include all (18) states for which comparable exit poll data are available, so we don't know what the differences are in the excluded states. Also, the data do not distinguish between open/closed/primary/caucus status, which no doubt plays some role, especially among independents. Finally, the data include Florida and New York, which are unique for obvious reasons.

Up next: Gaps, gaps, gaps!