More Intrigue in Clark County Early Voting, Day 5

Clark County Day 5 early voting results have enough in them for every party to spin them to their advantage. The comparative performances by the two major parties were even with the prior day’s results. Democrats cast 14,934 ballots while Republicans cast 9725 ballots resulting in  the Republican early vote accounting for 65% of the Democrat early vote for the second straight day.  Although this matched the prior day’s comparative performance it was a slight improvement for the Democrats relative to my model.  Democrat turnout was +3.4% higher than expected while Republican turnout was -2.3% below expectations.  Not a horrible performance but something the GOP needs to meaningfully counteract if they hope to close the Clark County Democrat advantage sufficiently enough to carry the state. Democrat’s aggregate lead in early ballots cast stands at 28,699, which is -9373 less than the 38,072 ballot lead at the same point in 2008.

More interesting though is daily turnout over the last 3-days has held fairly steady but yesterday both party’s % of the overall vote dropped day-over-day. Democrats were down -0.6 percentage points while Republicans were down -0.2pp.  These deficits were made-up by the steady increase in the Independent/Other parties which accounted for 19.3%, higher than any level achieved by this segment other than their 2008 peak of 19.5% on the final day of early voting. While the two major parties duke it out with results bouncing up-and-down depending on the day, the Independent/Other group has made a steady march higher in each of the four days of early voting.  The Republicans comparative gain on Democrats is a big story thus far in early voting but the rise of he Independents could end up being story of early voting. At this rate, if one candidate can decisively carry this vote it will likely swing the election.

The Clark Model

After the new assumptions were added into the model yesterday I think it performed pretty well based on the above results.  The one thing I want to constantly tweak is the growth rate of Republican’s % of the Democrat vote because this can have a disproportionate impact on the overall difference.  So every day the model will reflect a new growth rate based only on the updated 2012 actual results.  This risk with employing such a daily change is the forecasted results will jump around a lot but because that is what is really happening on the ground I’d rather have each day be a fair estimate of where the vote would end up based only on the actual growth rate achieved and my best guestimate on what I think the growth rate will be (I’ll leave that to your state pension managers).  In this case it brings the growth rate down since the GOP failed to increase the day-over-day turnout relative to Democrats.  The new growth rate is 3.2% versus the 4.2% used yesterday.  We see what a sizable impact this change can have on forecasting in that the expected vote margin between the two parties heading into election day is now 56,034 whereas yesterday the model had this number at 40,820.

Note: the inverse applies to my Washoe model where I have been hesitant to change the growth rate because the GOP is killing it up there and the actual growth rate is nearly double what I modeled. In that model the final result is balanced out by an overestimate of Democrat turnout but  I’ll be changing the Washoe growth rate and expected turnout after tonight’s data though so we’ll see what happens.


For reference purposes, here are links to each of the previous day’s analysis.

  • Day 1: Nevada Early Voting is Starting and Day One Results Are Out
  • Day 2: Republican Early Vote Continues to Gain in Clark County, Nevada
  • Day 3: Clark County, Nevada Early Voting Day 3
  • Day 4: Republicans Continue to Make Waves in Clark County, Nevada Early Voting



  1. kyle
    Posted October 25, 2012 at 10:10 am | Permalink | Reply

    Yikes!! not a good trend! However, if we can keep them under 60k in Clark…I think we have a chance!

    • Posted October 25, 2012 at 11:03 am | Permalink | Reply

      You have to remember, McCain lost Nevada by 120,909 votes; Clark Co by 123,687 votes.

    • WillBest
      Posted October 25, 2012 at 3:08 pm | Permalink | Reply

      Clark needs to be kept under about 25-30k to have a decent shot at winning NV

  2. Dennis
    Posted October 25, 2012 at 10:12 am | Permalink | Reply

    Keith. Any thoughts about Mark Halperin’s new piece on why the Obama pollsters are still absolutely certain Obama will win the critical swing states, including Ohio?

    • Posted October 25, 2012 at 10:21 am | Permalink | Reply

      The Obama campaign assumes a racial make-up of the electorate that almost certainly will not occur this year. If you adjust down the % of Whites and adjust up the % of Hispanics it makes it increasingly hard for Republicans to win. My sense is the Democrats are putting all their marbles in the demographics bucket and hoping White voters don’t show up. I think they are delusional.

      • No Tribe
        Posted October 25, 2012 at 10:29 am | Permalink

        Right, white voters only composing 72% of the electorate is an Obama pipe dream. It was 74% last time, and likely to go up a digit or two.

      • William Jefferson
        Posted October 25, 2012 at 11:22 am | Permalink

        I think Obama is on track to lose even if white voters comprise 72% of the electorate. It’s hard to win when your white support is 35%.

      • Posted October 25, 2012 at 11:35 am | Permalink

        The funny thing is everyone uses 74% for the 2008 election (even me sometimes) but the real # is 76.3 according to the official after-election review. This makes the Democrats 72% all the more ludicrous.

    • Dennis
      Posted October 25, 2012 at 10:54 am | Permalink | Reply

      Thanks, Keith. As usual, your analysis and comments are superior. I really appreciate this web site and your most impressive dedication to a truthful presentation of the real state of the election.

    • zang
      Posted October 25, 2012 at 12:48 pm | Permalink | Reply

      The fact that Obama’s pollsters are talking to the media about how good their internal poll numbers are is good sign — for Romney.

  3. No Tribe
    Posted October 25, 2012 at 10:24 am | Permalink | Reply

    The biggest variable in this is the unaffiliated voter, which you can see is trending up. Both parties are going after less than 4/4 voters, more like 1/4 or 2/4 voters, the low propensity voters. And a lot of them will tend to be unaffiliated or Independent. But, the campaigns have identified and confirmed them as a voter and will bug the crap out of them until they vote. And whom is turning those out, we don’t know. But that’s rising as a percentage, which is what I would expect. The tell would be whose hide it comes out of as an expense, as a sign of party motivation.

  4. Adam
    Posted October 25, 2012 at 10:24 am | Permalink | Reply

    So here’s something to think about Keith… 3 of the 4 Nevada polls report the independant numbers, (not rassmussen.) I averaged them out, and we have Romney at 48-44. Assign those numbers to the rising independant early votes, where are we sitting when you add R+I(48) and D+(I44)?

    • Posted October 25, 2012 at 11:12 am | Permalink | Reply

      That’s good info. I’ve been saying Romney is trailing with Independents in NV and another commenter disagreed with me. Happy to be proven wrong. I hope he keeps it up!

      Under your scenario Obama would beat Romney in the early vote in Clark County 57 to 41 — a 16pt advantage. In 2008 at the same point Obama was probably leading 64 to 36 — a 28pt advantage. Shave 12pts off Obama lead and suddenly you flip the state. To be clear my assumptions are slightly favorable to Romney. In Nevada most polls give Obama a greater crossover vote and higher base support. My quick and dirty calculations assumed the same base support and crossover vote. By even excepting that, it does reveal a compelling case for Romney thus far.

      • Posted October 25, 2012 at 2:59 pm | Permalink

        Early voters will likely make up a larger share of the total vote this year than they did four years ago. It looks to be at least 70% of the vote. A good Dem lead in Clark County amongst early voter turnout, even if smaller than 2008, will be substantial… with even fewer election day voters to potentially shift things.

    • William Jefferson
      Posted October 25, 2012 at 11:53 am | Permalink | Reply

      Romney won Indies in Rasmussen’s poll 59-37.

  5. Ranger375
    Posted October 25, 2012 at 10:33 am | Permalink | Reply

    Sorry to inject a question that is off topic for this thread but so as not to assume anything and because my wife and I are trying to learn from you all – how is the Margin of Error determined?

    Thank you for tolerating our primary school questions.

    • Posted October 25, 2012 at 11:31 am | Permalink | Reply

      that’s a stats question and I hated my stats classes. I’ll have to leave that for someone else.

      • Ranger375
        Posted October 25, 2012 at 11:35 am | Permalink

        Fair enough 🙂

    • RhodyKev
      Posted October 25, 2012 at 1:21 pm | Permalink | Reply

      Simplest margin of error is soley based on the poll sample size and the assumption of a normal or standard distribution. (classic Bell Curve)
      The equation is Error = zα/2/(2√ n)

      Zalpaha/2 is parameter taken from normal distibution Z tables given a level of confidence desired – (less than 100%) 95% confidence level is most common.
      N – is the sample(poll) size.

      Alpaha = 1-confidence. So alpha is usually = 1-.95 = .05
      Z(alpha/2)= Z(.025) = 1.96 (normal distirbution z table)

      Example: Poll sample size =1000, confidence desired 95%

      +/-Error = 1.96 / (2*√ 1000) = 1.96/(2*31.6) = 3.1%

      There are more complex error calculations. A reference to learn more is

      • Posted October 25, 2012 at 1:29 pm | Permalink

        Thank you.

      • Ranger375
        Posted October 25, 2012 at 4:05 pm | Permalink


        Thank you! The link you provided was very interesting and explained exactly what I was looking for.

        It was the paragraph under The “Advanced Margin of Error” Class that really went to the heart of my question.

        I see conclusions all the time using the “within the margin of error” to say that even though one candidate is up by some number it is still tied.

        The writer at the link states. “It is a favorite pastime of bloggers and pundits to take a poll, say, with a margin of error of ±4% and results such as the following:

        Romney – 48%
        Obama – 41%

        and to declare rather haughtily, “What this actually means is not that Romney is ahead, but that the race is tied.”

        It is this area I was interested in and I got my answer thanks to you!

        All I can say is this polling thing can be used to make many points — the same poll used in opposite directions for instance.

        Now I understand why that can be.

      • RhodyKev
        Posted October 26, 2012 at 10:27 am | Permalink

        Gald it helped! That’s the frustration of statistic’s – the spin is too easy. Takes some knowledge to understand the shortfalls. That’s what make’s this blog a great resource.

  6. housebroken dad
    Posted October 25, 2012 at 10:35 am | Permalink | Reply

    In your estimation, Keith, what realistic number do we need to keep Clark Cty under to have a shot on 11/6? Right now, we are at 29k. Is a 75k advantage too much to overcome if we narrowly win Washoe and counting the rest of the state?

    • Posted October 25, 2012 at 11:30 am | Permalink | Reply

      75k is likely too much to overcome. The GOP should win Washoe by ~10-15k overall and the rest of the state by ~30-35k overall. At 50k in Clark you’re in the ballgame. At 40k you’re feeling decent about your chances.

  7. Kent Ostby
    Posted October 25, 2012 at 11:22 am | Permalink | Reply

    For this and Washoe, can we get 2008 lines under the net votes ahead on the 2012 polls so we can see the comparison more clearly.

  8. housebroken dad
    Posted October 25, 2012 at 11:34 am | Permalink | Reply

    Interesting tidbit. Not sure how they can measure this but it’s somewhat comforting:

  9. No Tribe
    Posted October 25, 2012 at 3:00 pm | Permalink | Reply

    Rasmussen would have to be terribly wrong here for Obama to win. But you can look at it, and its been strikingly right with ’10, ’08, ’06 and ’04

    It predicts a huge Romney landslide ala what happened in ’10

  10. rcl_in_va
    Posted October 25, 2012 at 3:56 pm | Permalink | Reply

    First time post; thanks for all the hard work and insight! I’m plagiarizing your Clark spreadsheet for some personal “what if’s”; particularly I/O assumptions and noticed in 2008 the third EV day (Monday, 10/20) IO vote count outlier (6429). I’m wondering if you can explain it. The Columbus Day holiday was the previous Monday wasn’t it?

    • Posted October 25, 2012 at 4:04 pm | Permalink | Reply

      Columbus day did not affect early voting it was the prior Monday.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: