Wednesday, October 04, 2006

Don't Take it Personally (Bogus Polls Again)

Roby Brock paid to have a poll conducted. The results seemed bogus to me, and I said so. It is nothing personal. I did not say that Mr. Brock himself was bogus. He just paid to have some work done, and I questioned the quality of the work they did for him. I did not mean it as an insult, it happens to all of us. It is like telling an associate whose car was worked on but not fixed that he "ought to get his money back". I did not expect Mr. Brock to jump to the defense of the poll so strongly. He put an article on his site aimed right at me, which Max Brantley quickly put on the Arktimes site, though he mislabeled me as a "Republican".

Roby did not allow any comments on the thread he aimed right at me, and you have to register at Arktimes so they can delete your comments when you start winning the debate with them, so I guess I will just have to respond right here.

The issue is the credibility of his most recent poll. It showed Hutchinson ahead 60-20 in the third congressional district last month. This month it showed him behind 46-45. In a Republican stronghold that sent him to Congress for years! Did anything happen to justify a 41 point swing in the space of a month? No. I think the pollsters have messed up somewhere.

But Mr. Brock choose not to focus on that, rather he took issue with my complaint that the growing third CD is under-polled.

(continued- to see Brock's case of why I am wrong and my rebuttal, click WEDNESDAY below and scroll down)


Blogger Mark Moore (Moderator) said...

First I will give you Roby's points...

Our Constituent Dynamics poll and others have been criticized by some for supposedly undersampling the 3rd Congressional district. The words “flawed” and “bogus” are usually associated with the myriad of polls, including ours, I presume because the outcomes are not agreeable to the critics.

I’ve read comments from would-be political experts that claim the 3rd district should account for 32-33% of the vote count, while the 1st and 4th districts should each be reduced to close to 20% of voter turnout.

We’ve allowed a 24% weight in the 1st district, 26% in the 2nd district, with the 3rd and 4th districts accounting for 25% apiece. Why?

Because it’s scientifically sound.

To set the record straight and to negate the arguments from some who say that the Republican-heavy 3rd Congressional district should be given more persuasion, I present to you the historical numbers of voter turnout from every county of every Congressional district in Arkansas during the last three general election cycles.

I used the 2004 U.S. Senate race, the 2002 Governor’s race, and the 2000 Presidential election because these races had the highest voter turnouts in Arkansas in those election years.

In 2000, the four Congressional districts represented the following vote totals:
1st – 220,189 (24%)
2nd –235,612 (26%)
3rd –233,413 (25%)
4th – 232,573 (25%)

In 2002, the vote totals and percentages by Congressional district were as follows:
1st – 195,742 (24%)
2nd – 208,883 (26%)
3rd – 199,155 (25%)
4th – 201,906 (25%)

And in 2004, these were the cumulative vote returns by Congressional district:
1st – 245,900 (24%)
2nd – 277,278 (26%)
3rd – 271,459 (26%)
4th – 244,630 (24%)

Any pollster who uses anything but historical turnout numbers is conducting a “flawed” or “bogus” poll.

Let’s put this myth to bed right now: the 3rd Congressional district is not going to account for 32% percent of the vote or a third of the statewide vote or even 28% of the vote. It will be doing well to pull 26% of the statewide vote.

When the statistics start backing up some of the wild claims about the influence of the 3rd district, I’ll suggest a less “flawed” polling methodology to our pollster. In the meantime, I’ll trust our numbers until voters prove us wrong.

7:36 PM, October 04, 2006  
Blogger Mark Moore (Moderator) said...

That means he will be "trusting the numbers" for 37 more days.

You can't use the 2000 numbers to show a trend for a very simple reason: There was a redistricting that year that moved the boundaries of the districts. The "3rd district" and the "4th District" ceased being the same patches of ground in 2002 that they were in 2000.

That leaves us the 2002 results to bounce off of the 2004 results. But before I show you just why I think these pollsters (and Brock too since he is so anxious to take ownership of what was originally not his mistake) let's first get clear what each of us is claiming. That way we know in 37 days who is right.

Let me clarify what I have claimed: I never said that the 3rd district would get a third of the vote, I have said it would get "close to a third" or "closer to a third of the vote than a quarter of the vote". A poster came on and said he crunched some numbers with the 3rd getting 32% of the vote, but that was not me.

I did say the 1st and or the 4th would be closer to 20% of the turnout. They are going to be closer to 20% than a quarter. That is to say, one or likely both of them will be less than 22.5% of the vote on election day.

Mr. Brock sticks by his claim that the 4th will have essentially the same turnout as the 3rd (25%), and that the 1st will lag behind the 3rd by only 1%. He also says that the 3rd district "Is not going to account for.... 28% of the vote. It will be doing well to pull 26% of the statewide vote"

I have said that the third Congressional District will be closer to a third of the vote than it is a quarter or that it will be "close to a third". So the midpoint between a third and a quarter is 27.5%. I feel confident that the third district will exceed this proportion of the vote. In fact, he says that it will not pull 28% and I predict it will pull between 28 to 30% of the statewide vote.

So let's lay out our predictions, so you can see after election day who is more credible.....

Moore....... 22%.....22%.....29%

7:58 PM, October 04, 2006  
Blogger Mark Moore (Moderator) said...

So now let's deal with the numbers and why it is a reasonble guess. Brock says he is using historical numbers, but polling is about projecting historical data into the future.

I already explained why redistricting compels us to exclude the 2000 data to predict future turnout. That leaves us with a comparison from 2002 to 2004. Let's compare the 3rd and 4th districts, which Brock says will have equal turnout in 06.

The 4th district went from 202,000 voters in 02 to 244,000 voters in 04, so the 02 vote was about 0.83 of the 04 vote.

The 3rd district went from 199,000 voters in 02 to 271,000 in 04, the 02 vote was only 0.73 of the 04 vote.

Now I understand that this is an off year election, and that turnout will not continue to increase and may well drop in 06. Don't be confused by that when I do the next step. I am only interested in the PROPORTIONS of voters in 06, not the absolute numbers. Still, in order to get the relative changes in proportions in those districts I can measure the rate of change in one district to the rate of change in the other.

As long as I measure both districts by their RATE OF CHANGE I can get the PROPORTIONS right, even if the vote total is going to be wrong.

So to get an estimate of the proportions- not a predicted number of voters- I can extrapolate the gains in 04 to each base number in 04. That is, I can multiple the 3rd district base in 02 by .73 to get close to the vote in 04, and so I will multiply the 04 base by this same factor to get an estimated proportion factor of 371K.

That is 271K / 0.73 = 371.

Now for the 4th congressional district proportion estimate. That is found the same way, by taking the vote in 04 and multiplying it by the growth in the vote from 02 to 04. 244K / 0.83 = 294K.

In both cases I "multiply" by dividing by the inverse, which yields the same number.

So the projected proportions are 294/371 for the 4th and 3rd districts respectively. 294/371= 0.79. That is to say the 4th district vote should be around 79% the size of the 3rd district vote.

So let's say the 4th does get 25%, what would that make the 3rds' share? 0.25 / 0.79 = 31.6%

Ok, but that is close to the 32% that even I took care to note that it was not my claim. And those numbers would add up to over 100% anyway if you did them for four districts. We need two numbers that add up to 50% or 51% of the vote coming from those distircts to get a realistic estimate.

How about 22% for the 4th?

That is 0.22 / 0.79 = 0.278

So that is close to 28% for the third CD, and the district totals would be 49.8%. The other point for the 3rd is going to come from the 1st district, which should have even lower turnout than the fourth. That would give you 28.8%, basically the 29% that is my call for the 3rd distict.

Is this estimation perfect? No. Is it even any good? Well, it is better and more accurate than pretending the voter trends from 02-04 do not continue from 04-06. They do. NWA is still growing double digits every year.

In 37 days we will see if a hobby-blogger is or is not more reliable than the establishment media. See ya at the polls!

8:38 PM, October 04, 2006  
Anonymous Anonymous said...

I see a flaw in your analysis, Mark, if I'm understanding you correctly. The percentage share of the 3rd District voting patterns will not continue to grow at the same rate as it has in the past, like compound interest. If that were the case, then by 2010 it would have something like 35-40% of the votes in Arkansas, which is plain silly. It will level off -- there's some indication that economic growth in NWA is already leveling off. What I also see lacking is any data on the actual population changes in the various CDs over the past 2 years. Most areas of Arkansas have not actually lost population, even while NWA has outpaced growth elsewhere.

Maybe NWA will have something like 27% of the vote in 2006. Maybe the 1st District will have 23%. Even with those numbers, based on everything I've seen, Beebe's still going to win big. Asa needs a silver bullet or a knockout punch, and he doesn't have it. (At least, he sure hasn't shown it in the debates or in advertising.) The only polls that even show it close are the Zogby on-line polls, which are highly suspect. (I followed Zogby closely in 2004, right up until election night, when he precited a solid Kerry victory.)

6:43 AM, October 05, 2006  
Anonymous Mark M said...

Why wouldn't it grow at a compound rate like interest? That is exactly how it will grow. If the 3rd has a seven percent growth rate for ten years it will double in population in that time. I don't say that it will, only that it has from 2000 to 2006. Meanwhile, the first and fourth districts will be basically unchanged in population.

Please don't make me look that up. You can probably google it and find it in five minutes.

And yes, that would mean that over time the third would represent 35 or even 40 percent of the vote- that is why we have REDISTRICTING every decade, so that imbalances like that can be corrected. My guess is that after 2010 the 3rd will be as compact as the 2nd or nearly so.

6:57 AM, October 05, 2006  
Anonymous Mark M said...

And I am not saying Beebe is going to lose, just that he very well might lose. The election will be close and Asa or Beebe could win it. There is a good chance the winner will have less than 50 percent of the vote, that is how close I think it will be.

All I am saying is that the folks who say Beebe has this locked up and we can all go home are mistaken.

7:00 AM, October 05, 2006  
Anonymous Anonymous said...

The flaw in your reasoning is that the 3rd CD will have a 7% growth rate, above and beyond the growth rates of the 3 other CDs, for 10 years in a row. That simply isn't happening and won't happen. I took your advice and googled growth rates in various counties in Arkansas. Huge growth in Washington and Benton counties from 2000 to 2005 -- 22% and 14% respectively. Together, they have added about 60,000 people since 2000. Outside those 2 counties, it gets more modest -- neighbors Madison and Carroll have solid growth of 5% and 6.5% respectively. Crawford's growing at a solid 8.2%. But Sebastian County's 3.2%, Marion's 3.7%, Franklin's 2.5% and Newton's -1.8% are on part with other parts of Arkansas. Growth rates of other areas in Arkansas include Greene and Craighead Counties at 5.6% and 5.5%, Miller County at 6.7% and Pulaski at 1.2%. Many small Delta counties have negative growth.

What's my point? That growth in the two shinging stars of NWA is great, but it's not going to much affect this political season or the race for governor. Roby Brock is correct -- the last 4 years should be a decent guide to this year. In order to make the sort of difference you predict, with 3rd CD getting "close to" a third of the vote, the population increase throughout the 3rd would have to outstrip the rest of the state, and I don't see it based on the numbers.

From what I've seen Roby Brock's analysis makes more sense than Mark's. The actual numerical growth in the CDs won't be much different than in 2002 and 2004, certainly not enought to affect this election cycle. I guess we'll see who's right on November 7.

7:47 AM, October 05, 2006  
Anonymous Mark Moore said...

Hey, it is the polling company that has sold Roby that bill of goods. He will know better next time.

I see the same data that you are giving and I come to the opposite conclusion. First about the growth in the 3rd. Bear in mind that Newton counties' -1.8% growth rate does not cancel out Benton counties 22% growth rate because they are not "equal slices" of the 3rd disticts population pie. The loss in Newton county only represents 160 people. The gain in Benton represents 40,000 or more.

If you added Benton and Newton county together, they would probably still have a growth rate of 20%. That is because Benton county has 186,000 folks while Newton has only 8,600.

The only parts of the 3rd that are not growing at a healthy clip are those that are tiny to begin with. Sebastian is the only big county that is kind of low at 3.2% but isn't that still faster growth than the rest of the state outside the 3rd?

As for the other districts, yes you can give me a county or two where they are doing well, but those disticts have a lot more counties, the counties that are doing well are not as large a slice of the pie as Benton and Washington are, and those districts have a lot more counties with flat or negative growth.

All I have done is extrapolate the relative growth from 02 to 04 into the 04 to 06 numbers. The only assumption that makes is that the growth patterns from 02 to 04 persist from 04 to 06.

I want to add one more factor that has been in my mind but I don't think I have put in writing before. The "motivation index" in the poll showed that Republicans were more motivated to vote than Democrats or independents. I think that will contribute to a higher proportion of the vote coming from the area of the state that is most Repbulican leaning.

IN CONCLUSION: The Biztalk poll makes one assumption, that you should average past numbers to get future numbers. My assumption is that you should extrapolate past trends to get future numbers. We will see which model more closley reflects reality in 36 days.

8:50 AM, October 05, 2006  
Anonymous Anonymous said...

If you actually believe, as you suggested, that any county in the 3rd CD has a growth rate of 7% per year, you've got bad facts. The fastest-growing, Benton County, grew at 22% total over five years -- a compounded growth rate of something like 4% per year. That's the best county in the district. The flaw in your logic is that the % growth rate will continue to remain the same every year. It will fall, almost certainly. If Benton County adds 5,000 people per year, it has a declining growth rate, because the % of the population growing per year is falling. For the % to remain the same, it has to add 5,000 the first year, 5,500 the next, 6,000 the next, etc. If Pulaski County adds 10,000 in one year, and Benton County adds 7,000, Pulaski has added more people, but Benton County's growth rate is higher.

Simply taking the % growth from 02 to 04, and assuming the % will be the same from 04 to 06, is therefore not reasonable. The % will slow down, and probably already has.

9:37 AM, October 05, 2006  
Anonymous Mark Moore said...

Oh the 2nd district is going to gain voter-share too. To clarify: I did not use a 7% growth rate in my calculations. I only used that as an EXAMPLE of how fast exponential growth can cause a doubling of population and I knew the 7/10 rule and I used that as an example. Sorry if that caused any confusion.

When I calculated the numbers I did it just as I described. I took the relative increase from 02 to 04 and applied that same increase to the 04 base. That method allows for the growth in the other counties.

We just seem to have a disagreement about what is more "reasonable", averaging past numbers to get future numbers or extrapolating past trends to get future numbers.

I agree you can't extrapolate too far out into the future, for example I agree with you that the NWA growth rate of the last six years won't continue for another ten. But that is not what I am doing.

I am taking TWO years, that have already occurred (05 and 06) and asking myself, "did the trends from 02 to 04 continue from 04 to 06 or did they change? I concluded that the trends did not change, so I used the trends to make a prediction. How is that unreasonable?

10:53 AM, October 05, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home