Home | Entries [Previous |Next]
14 Comments
I think that's a good point that we haven't really seen what is "baseline" early voting numbers to really tell if early voting is up or down. Still, we can at least emperically say early voting is up over 2002.
Posted by: Adam | November 3, 2006 08:40 AM
This is interesting and helpful - but your conclusions are misleading and in some cases inaccurate. You can't compare "rate-of-growth" numbers and decide that's the margin of early turnout advantage. Also, your decision to name all counties that did not exceed the state-wide GOP advantate as "blue" is mind-boggling. If nothing else, consider Hamilton, Corker's hometown! By this standard you are considering Hamilton a blue county? Some of the raw data here is interesting and helpful but doesn't lead to nearly the level of advantage you claim. and some of it--like naming Hamilton blue because it didn't exceed the Bush statewide margin is a bit perplexing.
Posted by: Don | November 3, 2006 10:53 AM
I'M NEW TO TENNESSEE. I DON'T UNDERSTAND WHY TENNESSEE DOESN'T WANT TO MAKE HISTORY THAT WILL ONLY BE GOOD FROM AN HISTORICAL PERSPECTIVE. IT WOULD MAKE ME VERY PROUD TO BE PART OF HISTORY.
FORD WOULD BE THE FIRST BLACK SENATOR ELECTED FROM THE SOUTH SINCE RECONSTRUCTION. THE DEPBATES CLEARLY SHOWED THAT FORD WAS MORE QUALIFIED THAN CORKER. I JUST DON'T UNDERSTAND WHY FORD ISN'T RUNNING AWAY WITH THIS ELECTION. IRAQ IS A BIG ISSUE WHICH IS ALL THE MORE REASON TO VOTE FOR FORD.
I HEARD THAT EAST TENNESSEE WOULD NEVER VOTE FOR A BLACK AND THAT THEY WOULD CONSIDER IT TREASON!
THIS REALLY SHOWS THAT TENNESSE IS BACKWARD AND OUT OF TOUCH WITH THE REST OF THE COUNTRY. I'M WONDERING IF TENNESSEE IS THE RIGHT PLACE FOR ME AND MY FAMILY. I THOUGHT THE SOUTH CHANGED. IT SADDENS ME TO THINK THE SOUTH HAS NOT CHANGED AT ALL.
Posted by: Mary | November 3, 2006 11:18 AM
Don, I don't know that I'm claiming "rate-of-growth" numbers is the margin of early turnout advantage. At least that was not my intent. My intent was to measure "rate-of-growth" in counties that have characteristically voted one way or the other - with the assumption that a growth in turnout in these core counties would help one party or the other.
As to your second point about Hamilton County, I don't know what you are talking about. The MOV in Hamilton County exceeded the statewide MOV in both 2002 and 2004, and thus it is included in the second standard above - as well as the first as a "red" county. Likewise, it is counted in the red totals in both the Clement and Kerry totals, because Hamilton Co. voted for Alexander and Bush respectively.
Also, it is included in the 25% standard (the strong red standard), even though it's MOV was not >25%, because as you say, it is Corker's home county.
So - again, Hamilton County is counted as red under every standard in this post.
Posted by: Adam | November 3, 2006 11:42 AM
I'm a died in the wool Democrat, so I like what you're implying.
But it's just as easy to assume that Republicans would try harder this year in counties voting Democratic in 2004 (and vice versa) as it is to assume that additional early voters this year vote the same as the rest of the county will later on, and the same as the county did in 2004. Ie, it could be that all the additional early voters in a Dem county are voting GOP. (That's not likely of course, but just for the sake of argument.)
Bottom line: aggregate (ie county-level) analyses like this are no substitute for actually polling early voters and seeing what they did. And Dem GOTV efforts shouldn't let up between now and 8pm 11/7.
Posted by: Thomas Nephew | November 3, 2006 11:53 AM
Thomas, your point - "it could be that all the additional early voters in a Dem county are voting GOP. (That's not likely of course, but just for the sake of argument.)" - is well taken ... perhaps I should caveat this post.
And I agree with you - I'm not trying to imply that the Democrats have a lead in the vote counts in early voting coming out of this period, simply that turnout is up at a higher percentage in Democratic leaning areas than Republican strongholds.
Posted by: Adam Groves | November 3, 2006 12:30 PM
I like the numbers, but I think the flaw in methodology is assuming that early voters are voting in roughly the same ratio as the final tallies of elections past. Young people, poor people, and people of color tend to vote less often than white senior citizens, so if there is an increase across the board, I would suspect that voters with Democratic sympathies are the biggest untapped pool of voters.
It would also be interesting to do a similar analysis by early voting location. In Nashville at least, certain areas (Bordeaux; Belle Meade; Green Hills; etc) scream one demographic or another.
Posted by: Halfdan | November 3, 2006 12:44 PM
"I like the numbers, but I think the flaw in methodology is assuming that early voters are voting in roughly the same ratio as the final tallies of elections past."
I'm not making that assumption - in fact this entire post is about how the number of early voters in each county have increased over the 2002 early voting period for the same time frame.
Posted by: Adam | November 3, 2006 12:50 PM
Adam - thanks for the clarification re Hamilton but what I am talking about is the early vote, which I assumed you would follow. Hamilton's MOV did not exceed Bush's MOV in early voting. It did in the final vote, as you say. Why not compare early vote with early vote?
But more importantly is the rate of growth flaw. Your change in language makes it more precise but still misleading in that it implies an advantage that may not in fact be real. To use an extreme example to demonstrate the flaw: if you defeat me 10-2 one race and in the next my turnout increases to 6 while yours increases to 15, I've increased 200%! and you only 50! I have a 150% advantage by your terms. But in reality you have increased your lead from 8 to 9. I don't deny that there is some increase in strong Dem counties (of which there are really only 2 of any size) compared to strong GOP counties. But it's nothing like what you imply. A 10% advantage in rate of growth is meaningless without considering the size of the counties in question, and the previous margins of victory. An increase in Shelby is more important for Ford than an increase in Davidson; and an increase in Williamson more damaging than an increase in other Rep. counties would be.
Posted by: Don | November 3, 2006 04:30 PM
Early voting totals for specific candidates do not exist as segregated totals - as I understand it -- at least not in election return data that I have seen. You can get the total number of early votes cast, but not the number of early votes cast for each candidate. Correct me if I am wrong.
To your second point, I realized this potential flaw, however, if we're looking at the strong red / strong blue standard above, the total voter share for the red counties approximately equals the blue counties. In other words, they are essentially the same in their percent share of the vote totals.
Perhaps I should add a column for voter share and then multiply the percent increase by the percent voter share to get a blended totals. That would probably fix the issue that you are raising.
However, it isn't necessarly my point to emphasize the exact percent turnout advantage Democrats have over Republicans, only to show that in Democratic areas across the state turnout improved more than in Republican areas.
Posted by: Adam | November 3, 2006 05:08 PM
You are wrong about that. The election commission past results by precinct include county total results that break down early and absentee votes by candidate. Not all counties include this info. But Hamilton does (they're the precincts marked z1, z2, z3, z8, z9).
And I agree that your point is "to show that in Democratic areas across the state turnout improved more than in Republican areas." But that's not what you showed! And that's the problem. Republican areas could have improved *more" than Democratic in terms of margin of vote but still less in terms of the rate of growth. Rate of growth is not the same as growth. My extreme example in the earlier comment magnified a flaw that is still in your analysis, even if your numbers start off closer together. There is no 10% advantage of any kind that is meaningful at all. If you multiplied the turnout by the past early vote in each county, then you'd have something. When I do that it offers Ford a very small increase over both Kerry's and Bredesen's models. So it's not like the news is bad for Ford, but it's not as good as some are using your analysis to claim. And it definitely doesn't show Dems with an early vote advantage (though you are certinaly not claiming that it does) I think it's also interesting to gauge whether GOP and Dem counties have *increased their influence* in the statewide totals. That lets you compare apples to apples a little easier. That's what I investigated in my DailyKos diary today here. And offered some other similar analysis here.
Posted by: Don | November 3, 2006 05:24 PM
Don, I realize that early vote totals are available in some counties and not available in others. I'm interesting in data from all 95 of TN's counties, but thank you for bringing the fact that Hamilton posts theirs to my attention.
To your point, I would encourage you to take a look at the third column in the charts above. This is the difference in change of number of votes in this area that you are asking about. That's "growth" as you call it. So you can see that "change" in the Republican and Democratic areas. You'll notice that this "growth" doesn't benefit the "blue" in some of the charts. I hope that addresses what you are raising.
Your analysis is of voter share, which is very relevant and interesting, but we're measuring different things. I'm measuring aggregate county's individual performance changes in early voting (without regard to share of the total vote) and you're measuring voter share of early voting shifts for only a few counties.
It probably would be a good exercise to carry out your voter share analysis to all the counties. I'll post that when I finish. Thanks for your comments.
Posted by: Adam | November 3, 2006 07:48 PM
Thanks for putting up with my grumbling--i like your update #2 especially. That seems to explain accurately what's happening.
Posted by: Don | November 3, 2006 09:30 PM
Analysis for November 02, 2006
Analysis: Early Voting - Who Benefits?
Early voting ended today - so now it's time to take a look at who benefited. In order to do this, we can do look at the percent changes in early voting county-by-county, but I figured it would be more interesting to look at early voting changes with a few different standards. But first, a disclaimer - these numbers do not include the last day of early voting because the data was not available at time of the analysis. For percentage changes figures, we have used the percentage changed based on the first 13 days of early voting in 2002.
The Strong Red / Blue Standard
Our first standard of analysis of the county early voting totals centers around the top 17 blue counties and the top 17 red counties. Why 17? That's the number of counties John Kerry won in 2004 in Tennessee and all of those counties (with the exception of Shelby Co.) also voted for Bob Clement in 2002. As such, we can predict that the following 17 blue counties will also go for Harold Ford, Jr: Benton, Clay, Davidson, Grundy, Haywood, Houston, Humphreys, Jackson, Lauderdale, Overton, Perry, Shelby (throwing it back in), Smith, Stewart, Tipton, Van Buren.
For the GOP, we must use a higher standard to determine the top 17 red counties. However, this is not hard. In the following counties, the margin of victory for the GOP in both the Senate race in 2002 and the Presidential race in 2004 was over 25%: Blount, Bradley, Carter, Greene, Hawkins, Jefferson, Johnson, Knox, Loudon, McMinn, Rhea, Sevier, Sullivan, Unicoi, Washington and Williamson. That's 17 counties - throwing in Bob Corker's home county of Hamilton (which gave the GOP 25 and 15% margins in 2002 and 2004 respectively). So among the top 17 red and blue counties here's the turnout:
| 2002 | 2006 | Change | % Change | |
| RED (17) | 150,206 | 228,973 | 78,767 | 52.44% |
| BLUE (17) | 143,643 | 233,852 | 90,209 | 62.80% |
| DIFF | 6,563 | (4,879) | (11,442) | -10.36% |
Using this standard, we can say early voting turnout grew at a higher percentage in the sample of seventeen Democratic leaning areas than in the sample of seventeen Republican leaning areas. Putting it another way, the Democrats-leaning base had a 10% stronger growth in turnout than the Republican base.
Red Counties Over Statewide MOV 2002/2004 (All Others Blue)
Looking at only a few of the counties may help in determining who drove core supporters to the polls (at least in the abstract, since we do not yet know who voters in these counties actually voted for), but we will need a different standard if we a standard that takes into consideration turnout for all counties in Tennessee. Using the following standard to define red and blue counties, we can do this. First, to find the red counties, we will use all counties where the GOP margin-of-victory (MOV) in that county was greater than the statewide margin-of-victory for the GOP in both the 2002 and 2004 elections. That yields a list of 56 of Tennessee's 95 counties. We'll assume that the other 39 counties go blue, even though they may have had narrow GOP victories in the past. That produces the following turnout comparison:
| 2002 | 2006 | Change | % Change | |
| RED (56) | 311,017 | 453,824 | 142,807 | 45.92% |
| BLUE (39) | 200,462 | 302,689 | 102,227 | 51.00% |
| DIFF | 110,555 | 151,135 | 40,580 | -5.08% |
Using this standard, Democratic-leaning counties still increased turnout at a higher rate than Republican-leaning counties, although it is less the extent because this takes into account a larger number of counties.
Clement Counties / Kerry Counties
Finally, we want to look at defining blue county turnout by looking just at the Clement counties and Kerry counties. Here are those tables, with Shelby Co. thrown in Clement counties:
BOB CLEMENT COUNTIES (PLUS SHELBY)
| 2002 | 2006 | Change | % Change | |
| RED (55) | 238,885 | 327,566 | 88,681 | 37.12% |
| BLUE (40) | 201,839 | 304,460 | 102,621 | 50.84% |
| DIFF | 37,046 | 23,106 | (13,940) | -13.72% |
JOHN KERRY COUNTIES
| 2002 | 2006 | Change | % Change | |
| RED (79) | 368,113 | 526,038 | 157,925 | 42.90% |
| BLUE (17) | 143,366 | 230,475 | 87,109 | 60.76% |
| DIFF | 224,747 | 295,563 | 70,816 | -17.86% |
Using the Kerry figures above, Democrats have a growth-in-early-voting turnout advantage over Republicans of around 18%. Averaging all the methods together, we can estimate that the growth-in-early-voting turnout among Democrats was up around 15% higher than Republican levels. Does this mean that the Democrats have registered more votes than Republicans during early voting? We don't know for sure, but the growth of early voting turnout in Tennessee's blue counties must be troubling for Republicans.
Updated 11/3 -- Edited out some language, hopefully to avoid confusion about what we are implying here. I want to emphasize again, we look at net-change growth in early-voting turnout, which has little to do with counts of actual votes.
Updated 11/3 II -- We've had a lot of discussion about whether or not these change figures are significant because they do not take into account the percentage total each country contributes to the total number of early voters. So, for the sake of discussion and more talking points - I'll add the following comparison of changes in voter share. First we determine a counties contribution to early voting in 2002 by dividing the county vote total by the statewide vote total and then do the same for the 2006 numbers. The difference between those two figures is as follows for my different standards of "red" and "blue."
Strong Red / Strong Blue -- Red +0.90%, Blue +2.83%
County MOV > State MOV -- Red -0.82%, Blue +0.82%
Clement Counties -- Red -0.78%, Blue +0.78%
Kerry Counties -- Red -2.44%, Blue +2.44%
Note the last three groups of changes are complements because we're taking the data from all 95 counties, whereas in the first example we only use a total of 14 counties.
Adam -
Nice analysis, but I feel that your data set is too small. It seems intuitive that such an analysis in, say, 2012, might be relevant because early voting should have a predicatable baseline by that time. It is hard to argue that it does now.
Still, excellent work.
Cheers,
Rob
Posted by: Rob Huddleston | November 3, 2006 07:21 AM