CA numbers changed *dramatically* toward end of precincts reporting

[republicansareevil]

This is a carry-over from the following thread:
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=104&topic_id=495211

In that thread I was trying to make two points -- first, that if you compared the "no" votes on the recall question to the votes for Arnold on the candidate question, that Arnold's "lead" was almost nonexistent. My second point was that even this "lead" was not valid because an unknown number of voters voted "no" on the recall and also voted for Arnold on the candidate question, which indicates that their first preference was to keep Gray Davis in office and it was only their second choice to have Arnold replace Davis. If that unknown number of voters (which might be estimated through exit poll data) was greater than the number of votes of Arnold's "lead" then that showed that a plurality of voters preferred keeping Davis to replacing him with Arnold and that it was only the inherently unfair format of the election that made it seem otherwise.

When I first crunched the numbers, there were 96-97% percent of precincts reporting (I didn't note the exact percentage but I recall it was 96 point something). This was at 5:58 Pacific Time. Arnold was leading "no" votes by only 38,000 votes, or 1%. It was certainly plausible at that time that Arnold's "lead" would disappear if you mentally subtracted the number of votes that overlapped with a "no" vote on the recall. Over 42,000 people voted for Huffington even though she dropped out of the race, so 38,000 voters opposing the recall but choosing Schwarzenegger as their next choice was not beyond believability.

Anyway, I wanted to keep my numbers updated as more precincts reported in. That's when I noticed that the numbers for Arnold were starting to change. (See post #12 in referenced thread.) By four hours later, his "lead" over "no" votes had become 3.3%. I subtracted the earlier numbers from the later numbers to determine the makeup of just those votes that had been reported in the four-hour time span. They differed dramatically from the data for the first 96-97% of precincts reporting. I made the following observations:

"Does it seem plausible that these later votes coming in would be so different from the first 96% or so of precincts reporting? This is a sincere question and not an accusation. I know just a little about statistics and even less about how votes are reported in elections. I realize that this is not a random sample and that there may be logical reasons for late-reporting precincts to differ from early-reporting precincts. Maybe they use different voting technologies that take different amounts of time to process votes. Or maybe larger precincts take more time than smaller. But it still seems like a very large difference to me. Because you are comparing a quarter-million votes against 7 or 8 million votes, you would expect the larger sample to be more "stable" (I don't know the statistical term). But this quarter-million was able to make a noticable difference in the later votes compared to the earlier votes. Anyone with any knowledge of statistics who can shed some light here? Also, does anyone know what determines the order of how precincts report? Are conservative areas traditionally the last to report?"

I'd still like to know if anyone can answer any of those questions. I know that Bev posted my data on the BBV forum. Anyway, I'm going to post the final numbers with 100% of precincts reporting, compared to the data with 96-97% reporting, and finally a look at just the votes for the last 3-4% of precincts reporting. I'll do that in a separate post so that I can use plain text and have the columns line up better...

[republicansareevil]

1. Here is the data...

Numbers with 96-97% of precincts reporting:

PART 1: RECALL
yes to recall...... 4,096,670...... 54.4% of votes on part 1 
no to recall....... 3,439,300...... 45.6% of votes on part 1
total (estimated).. 7,535,970

PART 2: CANDIDATE
Schwarzenegger..... 3,477,359...... 47.9% of votes on part 2
total (est.)*...... 7,259,630 (accurate within 7578 votes)

difference between "no" votes and Arnold votes:
38,059 votes (1% more Arnold votes)



Numbers with 100% of precincts reporting:

PART 1: RECALL
yes to recall...... 4,415,341...... 55.4% of votes on part 1
no to recall....... 3,559,400...... 44.6% of votes on part 1
total.............. 7,974,741   

PART 2: CANDIDATE
Schwarzenegger..... 3,743,393...... 48.7% of votes on part 2
total (estimated).. 7,686,646 (accurate within 7892 votes)  

difference between "no" votes and Arnold votes:
183,993 votes (5% more Arnold votes)



Numbers extracted for last 3-4% of precincts reporting:

PART 1: RECALL
yes to recall........ 318,671...... 72.6% of votes on part 1 
no to recall......... 120,100...... 27.4% of votes on part 1
total................ 438,771

PART 2: CANDIDATE
Schwarzenegger....... 266,034...... 62.3% of votes on part 2
total (estimated).... 427,016

difference between "no" votes and Arnold votes:
145,934 votes (122% more Arnold votes)

[Spazito]

2. Good work!

Between 96-97% reporting and 100% reporting, Arnold's vote went up 122%? Is that what this says? If so, WOW! Is this a normal voting pattern? It seems odd to me!

[punpirate]

3. No, it doesn't say that.... n/t

[Spazito]

4. Thanks, I am a neophyte on this issue, any help to understand is....

welcome.

[republicansareevil]

10. The 122% applies to the last 3-4% of precincts reporting

First let me repeat my caveat that there may be a legitimate reason for the last 3-4% of precincts reporting to be different from the first 96-97% reporting. That's one of the things I'm trying to figure out here.

The 122% is the percentage of Arnold votes greater than the "no" votes (e.g. if Arnold had 600,000 votes at some point and the "no" votes on the recall were at 500,000 at that same point, Arnold's "lead" over the "no" votes would be 20%). It sort of shows Arnold's support compared to Davis's support (but not exactly -- for the reasons I explained earlier). The statistic itself is not as important as the fact that it changed so quickly toward the end of reporting.

With 96-97% of precincts reporting, Arnold's vote tally was only 1% higher than the "no" votes. By the time 100% of precincts had reported, the Arnold tally was 3.3% greater than the "no" tally. If you look at the votes from just those precincts that reported after the first 96-97% of precincts had already reported (i.e. the last 3-4% of reporting precincts), you see that Arnold's "lead" over the "no" votes was 122%. So the Arnold tally was over twice the "no" tally for those precincts even though they were neck and neck for the first 96-97% of precincts. A big change. That's why the statistic went from 1% to 3.3%. Even though the precincts were only a small part of the total precincts, they had a big effect on that statistic.

So does that make sense or is it more confusing than ever?

[Spazito]

12. Thank you!

Yes, that does help me. At the very least, it is a very interesting anomaly that, imo, bears further exploration.

[republicansareevil]

24. I mean 5%, not 3.3%

The 3.3% figure was from when 99% of precincts had reported. 5% is the final statistic after 100% of precincts reported.

(Not sure how many absentee ballots have yet to be counted or how they will affect "final" numbers.)

[Spazito]

25. Regarding the absentee ballots...

I believe there is also the provisional ballots as well. There doesn't seem to be a firm # on how many of each there was and whether any have already been added to the already published totals.

[rucky]

28. I remember a thread

about Freepers forgetting to vote on the recall, and just voting for Arnold.

If that were a common error for the not-so-bright Arnold voters, that would explain the spread.

[republicansareevil]

29. Which spread do you mean?

Sorry. Can you clarify? It doesn't explain why later-reporting precincts differed so much from the earlier-reporting precincts. Oh, do you mean people who voted early forgot to vote on the recall whereas later voters didn't because the issue had been publicized by then? If that is what you mean, it doesn't make sense here because we're not distinguishing between early voters and late voters but rather between votes in early-reporting precincts and votes in late-reporting precincts.

[AP]

5. Orange Cty and Alameda were counted late.

Orange Cty would have been expected to be safe for Arnold, but turnout was low, I believe.

Alameda, a democratic stronghold, had low turnout too, and had a lot of people voting for the unknowns.

I believe that the republicans probably wanted to make sure that Arnold got more votes than no votes so to confer legitimacy.

And I think it's interesting that the Diebold counties came in so late and that Orange Cty came in late as well.

[republicansareevil]

8. That's what's fishy.

I, too, had heard that Diebold areas came in late. I remember there had been some hypotheses that Florida punch cards (in 2000) were spoiled en masse late in the game (one article referred to "Katherine Harris with tweezers") which resulted in the high number of overvotes. I wondered if some last-minute tweaking was done in CA as well?

So entire counties report at once? I wasn't sure whether precincts reported individually or by entire county. If entire counties report at once it makes the sample (of the last 3-4% of precincts reporting) even less "random", but I'd like to find out which areas were the latest to report. I wonder if Tulare (sp?) was one of them. That's the one mentioned on Bartcop as having a high number of votes for unknown candidates even in areas that are not where the candidates live (i.e. it's not a friends and family effect).

[AP]

14. No, counties will report precincts as the roll out. They don't wait.

Alameda didn't report it's first precincts until very late. Orange reported some precincts early, but didn't finish until late. LA also reported a few early, but didn't finish until late.

Tulare was very late too. I don't think they reported their first precincts until about the time Alameda started reporting, which might have been as late as 1 am.

You can get all this info, including when the first and last precincts reported, from the county status report which is one the SOS's web site.

[republicansareevil]

19. Thanks, AP!

I'll check that out.

[benderlane]

6. hmmm

Its also possible that a majority of the people who Voted for Arnold had jobs that got out late.

I know that in NYC a majority of the republican voters tend to show up after 6PM.

[creativelcro]

7. Hmmmm....

How about Democratic voters?... -C

[republicansareevil]

9. It has nothing to do with when people voted.

I'm not saying that the votes that were cast later differed from earlier votes. I'm saying that the votes in precincts that reported late differed from votes reported earlier.

[newyawker99]

11. Hi benderlane!!

Welcome to DU!! :toast:

[republicansareevil]

13. One other important point...

That 96-97% of precincts that reported earliest represents about 94-95% of the votes.

The number of votes that had been cast on the recall question at that point (5:58 am Pacific) is equal to about 94.5% of the votes cast on the recall question after 100% of precincts have reported.

The number of votes that had been cast on the candidate question at that point is equal to about 94.4% of the votes cast on the candidate question after 100% of precincts have reported.

So the last 3-4% of precincts reporting represents 5-6% of the total votes (not a random sample), and that 5-6% of votes was able to swing the Arnold vs. "no" statistic from 1% to 5%.

edit: corrected erroneous number -- 5%, not 3.3% (the 3.3% was with 99% reporting)

[NewYorkerfromMass]

15. what about the exit polls? are they not legitimate?

how many different exit polls were conducted?
By what organizations?
Any refutation of the official tallies will have to be supported by an exit poll with a contradictory result.

[noiretextatique]

16. i wonder: was ANYONE here interviewed by an exit poller?

no one i know was :bounce:

[arcane1]

21. not ME

nor anyone in the area where I voted

though Wolfie was reporting Arnold as the winner based on exit polls at 8:01 pm!!
:grr:

[Athletic Grrl]

34. They already know how Oakland will go....

I doubt I'll ever see an exit pollster.

[republicansareevil]

17. I don't know about the exit polls

I thought VNS stopped doing them, but I had seen an exit poll done by the LA Times and I'm sure there are others. But I don't know if they measure up to VNS or not. VNS was something like 99.9% reliable as I recall. Do you have any links to exit polls? As I said, I don't know that much about staistics and I'm not sure I'd be able to compare the data meaningfully.

[NewYorkerfromMass]

18. This is the key

I raised a hue and cry here last Nov. because the exit poll is a legitmate de-facto audit (or parallel cross-reference if you don't want to call it an audit) of the election.
This is what we have to organize for the election next year- A HUGE exit poll so that we may say: "O.K. here's our numbers- shows us yours".
We need to do this in all the key states: Florida, Georgia. Penn, Ohio., etc...
We need to organize now.

[DEMActivist]

20. www.pollwatch.org

There are some DUers connected with this group which is trying to organize some citizen exit pollers. Go sign up!

[morgan2]

22. delete me

oops wrong thread

[Oilwellian]

23. Check out what I captured for Colusa County

County Status As of Oct 7, 2003 at 11:44 pm

County Name Total Precincts Precincts Rpt'g % Rpt'g Reg'd Voters Ballots Cast % Turnout First Report Date-Time Latest Report Date-Time Report Type*
Alameda 1,010 706,056
Alpine 5 2 40 815 252 30.9 07-9:53pm 07-10:30pm R
Amador 57 57 100 19,245 14,811 76.9 07-9:03pm 07-10:36pm F
Butte 162 103 63.5 115,492 46,540 40.2 07-8:44pm 07-11:37pm R
Calaveras 30 8 26.6 25,154 9,418 37.4 07-8:49pm 07-11:23pm R
Colusa 14 14 100 7,718 7,718 100 07-10:08pm 07-11:37pm F


In essence, Colusa county reported 100% turnout at 11:44PM EST. Now look at what they reported less than an hour later:

County Status As of Oct 8, 2003 at 12:35 am

County Name Total Precincts Precincts Rpt'g % Rpt'g Reg'd Voters Ballots Cast % Turnout First Report Date-Time Latest Report Date-Time Report Type*
Alameda 1,010 618 61.1 706,056 129,502 18.3 07-11:55pm 08-12:01am R
Alpine 5 5 100 815 573 70.3 07-9:53pm 08-12:33am F
Amador 57 57 100 19,245 14,811 76.9 07-9:03pm 07-10:36pm F
Butte 162 140 86.4 115,492 57,934 50.1 07-8:44pm 08-12:07am R
Calaveras 30 20 66.6 25,154 13,994 55.6 07-8:49pm 08-12:34am R
Colusa 14 14 100 7,718 5,064 65.6 07-10:08pm 08-12:25am U

[republicansareevil]

26. Wow

Have you tried calling them (Colusa elections officials) to ask about this? I don't live in CA or I would. No place gets 100% turnout. Of course, it might just be a reporting error that they later corrected, but still...

[republicansareevil]

27. I'd like to see how the numbers looked over time

Does anyone know if there is a way I can go back and see how the numbers looked with, for example, 50% of precincts reporting? I only have these numbers because I saved them for another reason. Somebody says Google caches things, but I'm not sure how that works. The CA Sec of State web site was continually updated as numbers came in.

[Eloriel]

30. This needs a kick

How the heck did it slide off page 1

Eloriel

[chimpy the poopthrower]

31. kick (nt)

[republicansareevil]

32. one final kick...

...from me :evilgrin:

[RedEagle]

33. Back Up

Kick