Articles and Analysis


Shifts in Vote and Turnout in New Jersey and Virginia


The shifts in outcomes between the 2008 presidential and 2009 gubernatorial elections in New Jersey and Virginia were driven far more by shifts in voting preferences among groups than by changes in turnout across those groups. Only age groups show consistently substantial changes in relative share of the electorate. Vote preference, in comparison, shows quite large shifts between election years. While one narrative of the 2009 election was changing turnout motivation, this turns out to be substantially false. Instead, changes in candidate preference drove the Republican wins in both New Jersey and Virginia.

The chart above shows the direction and size of change in vote preference for nine categories and 27 groups measured by exit polls in both years. The arrows start at the 2008 vote and point to the 2009 vote. The length of the arrow shows the amount of change and the arrow shows the direction of change. The colors code the shift in majority vote from 2008 to 2009. Blue indicates a Democratic majority for the group in both years. Red represents a Republican majority both years. Purple shows the groups that switched from a Democratic majority in 2008 to a Republican majority in 2009. None of the 27 groups switched from Republican to Democratic majorities.

In Virginia, large shifts in preference came among 18-29 year olds, those without a college degree, independents, rural voters and males. Smaller but still interesting changes came among lower and middle income voters, both of which shifted from majority Dem to majority Rep.

The most talked about shifts are among partisans and ideological groups. The large 16 point shift from 49 to 33 percent Dem among independents has justifiably received a lot of attention. But perhaps as interesting is the similarity of partisan loyalty among Dems and Reps. Neither shifted by enough to make the length of the arrow stand out. Virginia Democrats actually increased their Dem support by a point, while Republicans came home by a small 4 percentage points more than in 2008. Clearly the independents drove the dynamics of the outcomes.

Among conservatives, there was a modest shift of 9 points more Republican support in 2009, and a 5 point shift among moderates. Liberals moved a single point more Democratic.

By contrast, the shifts in share of the electorate were quite modest, as seen below.


By far the largest shifts are among the various age groups. The 18-29 year olds dropped 11 points, from 21 to 10 percent of the electorate. Those 30-44 also declined a bit, from 30 to 24 percent. These were matched by gains of 9 points among 45-64 year olds and of 7 points among those 65 and older. Age is one of the most potent predictors of turnout, and as this chart shows, one of the most dynamic from 2008-09.

The other two groups with interesting shifts are the rise in share of the electorate among conservatives (up from 33 to 40 percent of voters) and the similar decline in turnout among Democrats, from 39 to 33 percent.

Not only are these shifts substantial, but they also stand out against the very modest shifts in share of the electorate for all other groups. Many of the arrows have invisible lengths, indicating very small changes of one or two percent.

In New Jersey, we also see large preference shifts and even smaller turnout shifts.


The giant preference change in New Jersey is among independents, the same as Virginia. NJ independents took a massive 21 point shift from 51 percent for Obama to just 30 percent for Corzine. Also as in Virginia, Republicans came home to their party a bit, from a 14 percent defection rate for Obama to just 8 percent defection to Corzine. Democrats meanwhile barely budge, down from 89 to 86 percent Dem.

There were other substantial movements in vote preference, among 30-44 year olds, moderates, whites, hispanics and males. In short, many groups in New Jersey made substantial movements away from Democratic votes.

By contrast, the makeup of the New Jersey electorate changed a bit among age groups but hardly at all for virtually all other groups.


As with Virginia, there were declines in share of the electorate among 18-29 and 30-44 year olds and compensating increases among those 45 and above.

No other group comes close to such large changes in size. Several change by exactly zero (the open circles in the chart) and most others have lengths too small to see in the plot. The nearest exceptions are a decline of moderates of five percent and a corresponding five point rise among conservatives.

The bottom line for both states is that turnout changes were mostly about the age structure of the electorate. Younger voters are more responsive to short term stimulation, and in 2008 that translated to relatively large turnout, while in the absence of that stimulus in 2009 the more stable commitment to voting among those over 45 advantaged that group.

The shifts in preference in both states were significantly larger for many more groups. Preferences are driven by candidates and issues and those were the primary drivers of the change in outcomes from 2008 to 2009.

Below are alternative looks at the data, comparing the share and the vote for each state. These give a better look at the entire set of groups, but it is harder to compare magnitude of changes along the diagonal line in these charts than in the arrow plots above.







I'm confused about why the data shows changes in vote preference "more than" changes in turnout. For example, if the turnout shifted (independent of other groups) from 40% conservative to 60% conservative, for example, wouldn't that shift show up across all the voter preference arrows except the liberal/conservative dimension, as seen here? I would have thought that the greater explanatory power lies in the model that has the fewer larger arrows (much like a principal component analysis).



Charles, you're provably wrong on this regarding Virginia, and you need to review actual turnout data and walk back this post and its conclusions. I can't speak to New Jersey because I don't live there, but I live in Virginia and have studied actual turnout myself to see that depressed Democratic turnout is provable and was a major issue in the election. Your big mistake is relying on exit polls without reviewing voter turnout in Virginia.

Statewide, voter participation dropped from roughly 44% of eligible voters in 2005 to roughly 39% in 2009, and that is the lowest voter participation in over 40 years. Further, the total statewide vote in the Governor's race increased by only 1,110 votes out of nearly 2 million cast in both years, a fraction of a percentage point increase in total votes while the state's eligible voting population and actual voter registration total both increased dramatically. Normal voter participation growth adjusted simply for population increases, and even adjusting for the lower participation rates of racial minorities who are a disproportionate share of the population increase the past 4 years, would show a considerably higher number of statewide voters.

So it is an established fact, beyond debate, that total voter turnout was depressed.

Secondly, it is established when studying precinct-level turnout data that high Democratic performing precincts consistently showed a drop in voter turnout, even in the total number of votes compared to 4 years ago which is far more dramatic and eye-opening than merely a drop in the percentage of registered voters participating. Meanwhile, high Republican performing precincts consistently showed higher numbers of votes compared to 4 years ago. As an example, in my Virginia House District 34, 11 of the 12 precincts won by Republican Barbara Comstock showed more total voters in the Governor's vote compared to 4 years ago, while 6 of the 7 precincts won by Democratic incumbent Margi Vanderhye showed a drop in total number of votes cast in the Governor's race compared to 4 years ago. This pattern repeats itself across the state.

A big mistake political pundits make is relying too much on exit polls, taking the data as sacred. And in this case, Charles, you're drawing inferences that aren't clearly supported by the exit poll data itself and are clearly contradicted by actual turnout data.

Please revisit your commentary here in light of a more complete examination of turnout data.



Let me add 2 follow-up points that I wish I would have noted in my first comment had I not been hasty to click "post."

First, besides looking simply at Virginia exit poll data and not also the state's actual turnout data, another mistake you make is comparing 2009 with 2008, rather than with 2005. Comparing to 2005 is the apples-to-apples comparison that reveals in stark relief the drop in turnout. It's the 2009-to-2005 comparison I actually used above, but I neglected to point out the difference between my comparison and yours.

Second, to the extent you're arguing merely that a "more normal" Democratic turnout would not have been enough for Deeds to win, I don't dispute that conclusion. Deeds did, indeed, underperform badly with independents compared to Kaine in 2005 and Warner in 2001, and he would, indeed, have lost without better performance with those groups.

But the political reality is that depressed Democratic turnout and poor performance with independents actually shared the same cause: a poor campaign where Deeds presented no positive and clear public image of himself and also failed miserably in the nuts-and-bolts of running a campaign, specifically in field organization. A better campaign on those two fronts would have simultaneously improved Democratic turnout and Deeds' vote share with independents. He still might not have won, but the final margin would have been close. So to the extent that one segregates Democratic turnout from performance with independents, that segregation is a mistake.




I disagree in dismissing turnout out of hand. Turnout is probably the reason for the victory, opinion change the margin.

If you gave Deeds Obama's numbers by Party Affiliation according to the 2008 exit polls, Deeds would still have only taken around 48%.

The turnout drop was quite stark, there was approximately a 75% drop in turnout among those under 30 and 57% drop in those between 30 and 45. That has real consequences in terms of ideology and partisan preference.


Dotty Lynch:

One of the reasons for using 2008 as the point of comparison is that one of the big questions going into 2010 is wither the "obama electorate." Thus the 2008/2009 exits are helpful. The retrospective vote question which was asked only in Virginia indictes a "McCain electorate" in 2009 overall and for independents. In 2008 Independents split 49-48 for Obama; in 2009 the independents who went to the polls said they had picked McCain 53-37 over Obama. I think this supports the thesis that key parts of the Obama constituency did not come to the polls in 2009.

I think both 2005 and 2008 comparisons for exits and turnout are helpful. It depends which questions you are trying to resolve.


Alan Abramowitz:

This is highly questionable, especially regarding Virginia. It is clear from the exit poll data that Republicans and conservatives made up a substantially larger share of the electorate in both states in 09 vs. 08 and especially in Virginia. In fact, the VA exit poll data show that McCain won among those voting in 09 by 8 points after losing the state by 6 points in 08. (Apparently the 08 vote question was not asked in New Jersey.) So it seems clear that there was a fairly big shift in turnout patterns here with Republicans/McCain voters turning out at a substantially higher rate than Democrats/Obama voters. Whites also made up a substantially larger share of the Virginia electorate this year (78 percent vs. 70 percent). These shifts in the partisan, ideological, and racial composition of the electorate clearly had a lot to do with the GOP swing, especially in Virginia. So while the share of the electorate from other demographic groups may not have changed much, that could be misleading because what matters is which types of partisans turned out within those demographic groups. For example, it is possible that much of the swing in independent preference between 08 and 09 was a result of higher turnout among Republican-leaning independents vs. Democratic-leaning independents rather than any shift in the prefrences of independents voting in both elections. According to my calculations, if the 08 presidential preference of the 09 electorate in VA had been identical to the actual 08 VA electorate, 53-47 Obama rather than the 54-46 McCain preference seen in the exit poll (excluding the small group of 09 voters who didn't vote in 08 or voted for someone other than Obama or McCain), McDonnell's margin of victory would have been around 4 points rather than 17 points. So the difference in the composition of the electorate between the two elections, which is clearly a turnout effect, accounted for the large majority of the swing between the two elections.


Mark W:

More rehashing of what's been said, but for people who haven't seen the question that we would like Charles to address, it runs as follows:

In the 2008 election for president, did you vote for:
John McCain - 51%
Barack Obama - 43%

Considering that Obama won in '08 with a comfortable 6 point lead, this seems to suggest that dem voter turnout in '09 was depressed.

And the link for the exit poll: http://www.nytimes.com/interactive/2009/11/04/us/politics/1104-va-exit-poll.html



This analysis is completely at odds with what PPP found in their polling. They were one of the few pollsters that got NJ right.


I'm glad to see a vigorous disagreement on the issue of turnout. I'll stand by my conclusions, but hopefully add a bit to the discussion.

Let's start by agreeing that I'm looking only at the exit poll data and that exits are imperfect despite how some treat them in other contexts. But let's also acknowledge that the exits are the only source of individual level data on how particular groups behaved, something that aggregate precinct data cannot do. (Pre-election polls can also give us individual data but with the added problem of not knowing who actually voted.)

I take all the comments about precinct returns seriously but that isn't what I'm analyzing. That is valuable information and I don't dismiss it and perhaps will take a look at it when it is available. It just doesn't provide the information on different groups that the exits provide.

I'll also agree that for many purposes a comparison of 2005 with 2009 is appropriate. But that wasn't the question I wanted to address, which was posed by the coverage and discussion: how did two Obama states become two states with Republican governors a year later. That's my focus and why I compare 2008 with 2009.

Finally, this isn't about decline in turnout from 2008 to 2009 but rather RELATIVE decline across groups. In 2008 3,723,260 votes were cast for president in VA versus 1,878,017 cast for governor in 2009, a decline of almost exactly half. The question is whether the relative proportions of groups in the 2009 electorate changed substantially from their share of the 2008 electorate. The margin swung from +6.3 Dem in 2008 to +17.3 Rep in 2009, a whopping 23.6 point swing in the margin.

The two most prominent arguments are that Democrats stayed home and that liberals also dropped off disproportionately. There is some of that, but not a lot. Democrats were 39% in '08 and 33% in 09. Meanwhile Reps rose from 33% to 37% and Independents rose from 27% to 30%. Party ID is the single most powerful political identification and predictor of the vote so this is critical information from the exit poll.

What would have happened to the statewide Dem vote if voting behavior remained exactly the same in 2009 as it had been in 2008 and only changes in turnout had driven the election? The Dem vote would have dropped from 52.2% to 48.02% based on the exits (the exit doesn't perfectly match the 52.6 actual vote due to rounding). The actual swing was 52.6 to 41.3, an 11.3 point swing, of which 4.2 points can be attributed to the changes in partisan composition.

So differential composition of the partisan electorate can account for 4.2 of an 11.3 point shift, or just over a third of the swing. That's not nothing, but it leaves 2/3 of the story to preference changes, which is what I said in the post.

Now flip it around. Suppose the 2008 electorate had voted as they did in 2009, but had the same composition as was actually the case in 2008. Obama would have won just 46.9% instead of the 52.6% he actually got. That's all due to changes in preferences, not changes in composition.

The changes due to shifts among ideological groups tells a similar tale. Liberals went from 21% to 18%, moderates from 46% to 42%, and conservatives from 33% to 40%. If the 2009 electorate had voted the same way they did in 2008, the Democratic vote would have fallen from 51.5 to 47.8, or 3.7 points out of a total actual shift of 11.3, or again 1/3 due to composition and 2/3 due to preference. (Again the exits don't quite match the 52.6 actual '08 vote. Add that in and we'll say 52.6 to 47.8 is the most possible due to composition, so 4.8/11.3=42% of the shift is the most possible due to composition.)

There is also little evidence that the Dems who did show up were disgruntled or weaker partisans than the Dems of 2008. In 2009, 91% of Democratic voters went for Deeds while 90% went for Obama. This wasn't an unraveling of Dem support. There was a bigger shift among Reps, from 82% for McCain to 91% for Deeds. And, as I say in the post, the most massive shift was among independents, from 49% Dem to 33% Dem. Sure the make up of those who turned out might be different-- but the Dems voted the same way, and in the face of pro-Republican forces in 2009. And Reps became more loyal by a bit. But the major change among independents surely reflects the difference in candidate appeal between the two elections, rather than primarily a shift of composition.

Now party id and ideology are the most powerful predictors of vote, so their composition is the most politically relevant of all of these groups. And those two contribute about a third of the shift in vote due to composition. The other 2/3 is because groups didn't vote the same way as before, a shift in preference. That is my point and the data support it.



Here are the calculations:

2008 composition/prop Dem:
Dem 39.3 / .92
Ind 27.3/ .49
Rep 33.3/ .08

2009 composition/prop Dem:
Dem 33/.93
Ind 30/.33
Rep 37/.08 **CORECTION: This should be 37/.04. The calculations below use the right value, but the value originally listed here is a typo.**

If PID proportions were the same as in 2009, but with 2008 proportion Dem votes, then Dem gets 48.02% instead of Obama's 52.6%, a 4.3 point shift. But Deeds actually got only 41.3%, so no more than 4.6% of the 11.3 point shift can be due to composition.

What if Obama had the 2008 electorate but they voted like they did in 2009? Obama would have gotten only 46.9%, a worse outcome due to preference changes than due to 2009 composition which would have netted him 48.02%.

Or looking at Deeds:
Actual 41.3,
2008 composition, 2009 vote proportions: 46.9%
2009 composition, 2008 vote proportions: 48.0%
2008 composition, 2008 vote proportions: 52.2% ---IF DEEDS HAD BEEN OBAMA with OBAMA's Electorate.

Given a choice, Deeds would prefer the 2009 electorate if he could have the 2008 vote proportions rather than the 2008 electorate but the 2009 vote proportions.





Your analysis assumes that the support for Obama among Independents who voted in 2008 was identical among Independents who voted in 2009. We saw that drop in turnout was not even among partisans, I would expect that it also would not be even among the partisan leanings of Independents. Without xtabs we can't confirm this at this point though.

Based on partisan voting habits from 08 and 09, Independents in the Obama vs. McCain question would give an 8 to 16 point advantage to McCain. That level of support for McCain would mean that about 49% to 60% of the swing would be accounted for by the change in the composition of the electorate. Take this with a very big grain of salt though.


Alan Abramowitz:

I think that Charles's calculations are incorrect. When I calculate Deeds' vote based on his actual performance in an 08 electorate I get the following: (.93*.393)for Democrats + (.33*.273)for independents + (.08*.333)for Republicans which adds up to .482 or 48.2 percent of the vote, not 46.9 percent.


No, my calculations above are correct. Alan used the Rep vote percent for Deeds from an email I sent him, which had a typo. The actual value is .04 but my email had it as .08, which is what Alan used above. I ALSO repeat the typo in my comment above (since the email was just copied from the comment.) But the typo was NOT used in the calculations or in the figures above, just in the comment. Still-- my bad.

Here are the calculations all laid out:

> share08 > pctd08 > share09 > pctd09 > share08
[1] 39.3 27.3 33.3
> pctd08
[1] 0.92 0.49 0.08
> share09
[1] 33 30 37
> pctd09
[1] 0.93 0.33 0.04
> sum(share08*pctd09)
[1] 46.89
> sum(share09*pctd08)
[1] 48.02
> sum(share08*pctd08)
[1] 52.197
> sum(share09*pctd09)
[1] 42.07

So Deeds got 41.3 actual vote. The calculations from the exits put it at 42.07 thanks to rounding off both shares and pctd (and perhaps exits weighted to less than final vote.) And exits put 2008 at 52.2 rather than the actual 52.6.

And repeating the results from before and confirmed above:

2008 composition, 2009 vote proportions: 46.89%. Definitely better than 41.3. We agree on that.

2009 composition, 2008 vote proportions: 48.02%. Better still. **Deeds would prefer the 2009 distribution of party id if he could have the vote proportions from 2008.**

The composition makes less difference than the shift in vote proportions. When you combine BOTH different proportions AND different composition, you get the big change-- 10.12 based on exits, 11.3 based on actual votes. Not surprisingly both together have a bigger impact that either alone.

Something none of us has said: The shift in preferences and the shift in composition are certainly not independent of one another. Fewer people (of whatever partisanship) preferred Deeds than preferred Obama. That surely also affected their turnout. So we shouldn't look at turnout shifts as ENTIRELY due only to motivation unconnected to preference. That means my calculations above shouldn't be taken as the separable independent effects of composition and of preference, because the two correlate-- we can't separate the two effects cleanly because they are correlated. That is a complication not addressed here.



The formatter ate some of my R code in the previous comment.

Ignore the line

> share08 > pctd08 > share09 > pctd09 > share08

The first set of three numbers

39.3 27.3 33.3

are "share08", the percent Dem, Ind and Rep in 08. The rest of the code formatted correctly.



Alan Abramowitz:

Correction accepted. But I believe that using the retrospective presidential vote results will still produce a bigger shift based on change in composition. A shift from a +6D electorate to a +8R electorate is clearly very significant. Vote recall is not a perfect measure, of course, but recall of presidential vote only one year later is likely to be pretty accurate. In any event, I think we can agree that both a shift in composition and a shift in preferences made significant contributions to the outcomes of these two elections. And part of the shift in composition was probably a reflection of shifting preferences. Democrats didn't turn out, in part, because they didn't like the Democratic candidates.



Here a slightly different approach on the Virginia question. I'm looking at the exit polls for VA for 2008 and 2009, from

which match Franklin's post at 6:23EST, except that the NYT gives the 2009 Reps for D vote as 4%, not 8%.

To look at the relative impact of change in proportions versus change in preference without getting into full-blown alternate history scenarios, you can decompose [prop2009*pref2009 - prop2008*pref2008] (ie, the change in total votes) for each party into:

prop2009*pref2009 - prop2008*pref2008 =
(prop2009-prop2008)*(pref2009+pref2008)/2 + (pref2009-pref2008)(prop2009+prop2008)/2

ie, the change in the proportion of the parties times the mean preference, plus the change the preference of the parties, times the mean proportion. If you do this for each party, you get a good sense of the relative contributions of the change in proportion and the change in preference between the two years -- the first product on the right-hand-side is the contribution due to change in proportion, and the second is the contribution due to change in preference.

The results are (where column A is the points lost between 2008 and 2009 by the Dems due to turnout change, and column B is the points lost between 2008 and 2009 by the Dems due to preference change):

party: prop....pref
D: -5.829.....0.362
I: 1.107.....-4.584
R: 0.222....-1.406

This illustrates why it's a bit tricky to assign blame to one change or the other. For the Dem share of the change, most of the change appears to be due to a change in turnout: a loss of 5.8 points. For the Independent share, on the other hand, most of the change appears due to a change in preference -- 4.6 points -- but that's slightly counterbalanced by the increased turnout of Independents giving a slight boost to Ds as well as Rs (1.1 points). And the R share isn't much of the story, though as with the Independents, what change there is is more due to preference shift than turnout shift.

But of the three parties, the biggest loss appears to be via the Dem change, and the lion's share of that is due not to preference shifting, but to depressed turnout (relative to the other parties, of course). True, the Ind preference change does play a role, but it's not as large as the D turnout change, and it's somewhat counterbalanced by the overall increased Ind turnout. This seems to contradict Franklin's conclusions, but I'm just dabbling here, and open to explanations of how the two analyses diverge.

Oh, and you can also do this with age, say:

age: prop....pref
18-29: -5.72.....-2.48
30-44: -2.85.....-1.89
45-64: 4.14.....-4.25
65+: 3.01.....-0.87

Here the story appears that changes in the youth vote were devastating for the Dems: a loss of 5.7 points due to depressed turnout, and a loss of 2.5 points due to changed preference. Losing a total of more than 8 point on the 18-29 group is doom for any future Dem candidate. As for the middle groups, it's mostly a wash, but for the old, although there was a slight change in preference against Dem, it was actually more than made up for in the huge (and slightly mystifying; nicer weather?) increase in elderly turnout, which actually gave the Dems a bit of a boost overall.

So to summarize, based on this approach, it seems like turnout played a substantial role, especially among Democrats and the young. Changing preferences also played a role, but much smaller, among Independents and the young. If I were a campaign for 2010 looking at one or the other though, I'd focus on turnout -- particularly among young Dems.

(And yes, I'm aware that this may be rife with ecological inference error, and that it's silly to do this when cross-tabs do exist. But this is a casual forum, I hope...)


Alan Abramowitz:

A couple of more, and hopefully final thoughts on this issue:

1. The shift in preferences in Virginia is almost entirely concentrated among independents. Deeds did essentially as well as Obama among Democrats and only slightly worse among Republicans, but he did far worse among independents, 33 percent vs. 51 percent. But this could very well be a function of a shift in the partisan composition of the independent vote. We know that the large majority of independents are leaning partisans--only 7 percent of voters in 2008 were pure independents according to the ANES--and that these leaning partisans vote almost identically to regular partisans. I strongly suspect that the independents who voted in Virginia in 2009 included a substantially larger contingent of independent Republicans compared with those who voted in 2008. Unfortunately the exit poll data don't permit us to test this hypothesis.

2. Because of this issue with independents, I think that using the recalled presidential vote to estimate the effects of changes in composition vs. changes in preference makes even more sense. To do this, I assume that the revised 2009 electorate would consist of 50% Obama voters, 44% McCain voters and 6% voters who didn't vote in 2008 (that is identical to the 6% in the actual 2009 electorate). Multiplying these proportoions by Deeds' actual share of each of these groups (88% of Obama voters, 5% of McCain voters, and about 33% of the new voters (I had to estimate this since the NYT exit poll results don't give vote shares for this small subgroup), I calculate that Deeds would have ended up with 48.2% of the vote. On the other hand, if you assume an electorate consisting entirely of Obama and McCain voters with no new voters, one exactly identical in presidential preference to the 2008 electorate, Deeds would have ended up with 49% of the vote. Either way, the results indicate the a shift in the composition of the electorate was responsible for most of the shift in the election results.


Brackdurf:-- Thanks-- that's very interesting and more fine-grained approach.

Something missing from the "turnout vs preference" debate is that these are not constant and equal effects across groups. Your comment does a great job of pointing this out.

Implicit in that is that the groups are themselves somewhat heterogeneous, and we don't get to observe the within-group variability.

Thanks for commenting.




Something we can agree on: "The shift in preferences in Virginia is almost entirely concentrated among independents. Deeds did essentially as well as Obama among Democrats and only slightly worse among Republicans, but he did far worse among independents, 33 percent vs. 51 percent." One of my points in the post was that Dems were as loyal in 09 as in 08, with large change among Inds and small change among Reps. Preference shifts among independents were vital for the outcome.

The rest of point 1 is untestable with the available exit data, so perfect for speculation but of limited empirical value.

On 2. there is abundant evidence that recall of past vote is less reliable and less stable than are reports of current partisanship. Substituting a less reliable measure of partisan composition for a more reliable one is not helpful.


Alan Abramowitz:

Actually the evidence shows that recall of past presidential vote is highly accurate even over a considerably longer than one year. According to data from the NES 2000-2002-2004 panel, 95% of Bush and Gore voters correctly recalled their 2000 presidential vote in 2002 and 94% correctly recalled their presidential vote in 2004 a full four years after the 2000 presidential election. There was a small bias in recall with Bush voters recalling their vote more accurately than Gore voters. Nader voters were also somewhat less accurate in both years at around 75% accuracy. But I think these data indicate that while the presidential vote recall question is not perfect, it has a high degree of validity. If anything one would expect the 2009 recall question to result in an over-estimate of Obama support in 2008.


William Ockham:

The fundamental flaw in Mr. Franklin's analysis is to treat independent voters as a homogenous group. That's clearly wrong. Using Mr. Franklin's figures for the changes in partisanship in the Virginia electorate from 2008 to 2009, I (temporarily) excluded self-reported independents to calculate the Democratic vote share of the self-reported partisan electorate:

2008 ((39.3*0.92)+(33.3 *0.08))/(39.3+33.3) or 53.47%

2009 ((33*0.93)+(37*0.04))/(33+37) or 45.95%

That's a difference of 7.51%. Is there any reason NOT to expect at least that much change in the composition of the independents? The actual change among independents was substantially higher (16%), but we are only talking about an unexpected change of 9.49% of the 30% of an electorate that was half the size of 2008.

Mr. Franklin's data doesn't support his conclusion.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.