WA-05, Goldmark vs. McMorris: Analysis from Yesterday's Primary
First, a brief bit of news. Yesterday, Peter again spent some time at DailyKos, posting his fourth diary there and spending an hour or so in the comments section answering questions.
This was on occasion, as you've likely read here or elsewhere, of his race against Cathy McMorris being upgraded by the DCCC to "Emerging Race" status.
[Editorial note, my not-so-humble opinion: I'll admit I'm a bit cynical about the DCCC. Frankly, I don't trust them with my money and would much rather do the research myself and donate directly to the candidates who resonate with me.
Reading between the lines on their Emerging Races page, I get the sense that "Emerging Races" is a euphamism for "Gee, I guess we should have been supporting these folks all along!" Whatever the case may be, I'm glad they're seeing the light now. Welcome to the party, DCCC!]
And now, the main event.
As you're no doubt aware, Washington's primary election was yesterday. Although Washington's partisan primary system muddies the waters somewhat, we can still draw some conclusions about the general election from last night's results.
Here are the raw results, by county. Don't worry, you don't have to digest the raw results in their entirety. Feel free to gloss over them, I'll tell you what I think they mean afterwards. Note that all of this data comes from the Secretary of State's Primary Election Results page:
Total Votes Cast with Party Breakdown:
Here are the vote totals for Peter Goldmark and Cathy McMorris, again by county:
* Estimates; Secretary of State's election results website has not provided (D)/(R) split for the raw ballots in Asotin county yet, although they did provide (obviously) counts in the Goldmark/McMorris race.
I used the percentage from that race to estimate the (D)/(R) ballot split in the first table, although doing so obviously that means that we can derive no useful relative performance data for Peter in that county.
Observations
First, the bad news: As a straight vote-count, Peter loses this election 42.45 percent to 57.55 percent. But, of course, that isn't the whole story because it's a partisan primary where one's selection of who one can vote for is limited.
Now, some good news: Peter out-performed expectations. That is to say, the actual number of votes cast for Peter exceeds the number you would expect to see given the overall party split in the total number of ballots cast.
It's not a big amount, that +1.72% at the bottom of the second table, but it's there. Particularly heartening is that Peter's best relative performance is in Spokane county, which has the lion's share of the votes. That's not the whole game of course, but it's a good sign, and indicates that Peter's message and/or Cathy's poor job performance are having some effect.
Further, Peter's largest negative performances (in Garfield and Adams counties) were also in two of the smallest counties. This is not to say that Peter should think he can blow off campaigning there, but that poor performance there will certainly hurt him less than elsewhere.
[Fair warning: The rest of this analysis is to a degree speculative, because there isn't raw data that directly addresses the difference between our partisan primary system and the open general election. I am forced to make some assumptions and educated guesses along the way. I will tell you what those are when they occur, and you're entirely free to leave some comments telling me why my assumptions and guesses are no good.]
So what's it all about?
In a word: undervotes. A word we all learned about from Florida 2000, ballots in which no vote for a candidate was recorded.
Undervotes in this election are 4967 for Peter, 10632 for Cathy. Or in other words, 4967 ballots marked as (Democrat failed to indicate a vote for Peter, while over twice as many Republican ballots failed to indicate a vote for Cathy.
The obvious and hopeful conclusion we would like to draw from that is that ten thousand Republican voters across the district would prefer to vote for Peter, but can't do so in the primary.
While that would be nice, a) it's still not enough to overcome the 17819 vote difference between total Democratic and Republican ballots in the whole election, b) if we do that then we have to be fair and assume the converse position for the 4967 Democratic undervotes and assign them to McMorris. The truth is almost certainly somewhere in the middle.
I find it more instructive is to look at undervotes as a percentage of total D and R votes cast. Because it certainly is significant that McMorris had twice as many undervotes as Goldmark.
What we really want to do is estimate, out of those undervotes, the percentage of Democrat and Republican crossover voters. Let's do the math:
(D) undervotes: 4967 out of 39152 total ballots = 12.69%
(R) undervotes: 10632 out of 56972 total ballots = 18.66%
In an uncontested partisan primary, which both of these are, there are three main reasons why undervotes could occur.
If we make the generous (to Goldmark) assumption that every single Democratic undervote was of type 1 or 2, then we would expect a similar percentage of Republican undervotes due to those same reasons.
That leaves an excess difference of almost 6% in McMorris' undervotes, which we can conclude may represent the percentage of Republican crossover voters who would have preferred to vote for Peter. That assumption - 0% (D) crossover voters and 6% (R) crossover voters - yields a general election prediction of Goldmark 44.29% / McMorris 55.71%. Still a win for McMorris, but not by quite as much as in yesterday's raw results.
Of course even that generous assumption for Peter, that no (D)ocrat would vote against him, is almost certainly too generous.
Just to get a sense for what might be going on, let's imagine that a full one-half of the Democratic undervotes were of type 3. That's surely more than reality, and as such represents some sort of worst-case scenario. That would leave 6.35% of type 1 and 2 undervotes as the base undervote rate, meaning that McMorris now has an excess difference of 12.31% in her undervotes!
The revised prediction, sending the crossovers from both sides over to their preferred candidates, is actually slightly better for Peter: Goldmark 45.44% / McMorris 54.56%. This counter-intuitive result is because the assumption forces us to also conclude that the Republican crossover-rate is higher.
The Elephant in the Room
What we can't measure, of course, are voters who indicated no party preference. Those voters did not get to voice an opinion on Goldmark or McMorris. So you need to appreciate these results with that large grain of salt. We just have no idea, on the basis of the primary election results, what the heck those people are feeling or how they might vote come November 7th.
Conclusions
Any way you slice it, at this point McMorris is seemingly still winning this race. Goldmark and his campaign crew clearly have a bunch of work left to do between now and November 7th.
There are positive signs, in that his message seems to be resonating in Spokane, Pend Oreille, Lincoln, and Whitman counties.
However, something is clearly not working in Walla Walla County, and perhaps there the message needs to be tweaked for those voters' more specific concerns.
I don't live there, so I won't claim to have any idea what those are, but clearly Peter's people need to take a look. Personally, I think that Peter's core strategy of crafting a one-two punch out of farming concerns and homeland security, in the form his concept of "national security crops" is brilliant.
Tying those in with the more left-leaning concerns over global warming and the environment makes it a one-two-three punch. I do believe that if Peter can effectively communicate this message across his district, then he has a shot at winning. I still think it's a fairly long shot, but the target is visible.
One thing, though: even the most generous analysis of yesterday's election puts Peter 8700 votes behind in the general. It will be tempting to concentrate in Spokane county, where Peter is already doing better than the party split would indicate, and where the bulk of the voters are. But I think that would be a mistake.
To overcome that vote difference in Spokane County alone would mean taking that county 57% to 43%. That's a mighty tall order. For now, I think, Peter needs to stay on the road, getting his great message out there.
This was on occasion, as you've likely read here or elsewhere, of his race against Cathy McMorris being upgraded by the DCCC to "Emerging Race" status.
[Editorial note, my not-so-humble opinion: I'll admit I'm a bit cynical about the DCCC. Frankly, I don't trust them with my money and would much rather do the research myself and donate directly to the candidates who resonate with me.
Reading between the lines on their Emerging Races page, I get the sense that "Emerging Races" is a euphamism for "Gee, I guess we should have been supporting these folks all along!" Whatever the case may be, I'm glad they're seeing the light now. Welcome to the party, DCCC!]
And now, the main event.
As you're no doubt aware, Washington's primary election was yesterday. Although Washington's partisan primary system muddies the waters somewhat, we can still draw some conclusions about the general election from last night's results.
Here are the raw results, by county. Don't worry, you don't have to digest the raw results in their entirety. Feel free to gloss over them, I'll tell you what I think they mean afterwards. Note that all of this data comes from the Secretary of State's Primary Election Results page:
Total Votes Cast with Party Breakdown:
| County | (D). Votes | (R). Votes | Total | % (D). |
| Adams | 434 | 1534 | 1968 | 22.05 |
| Asotin | 2229* | 2811* | 5040 | 44.23 |
| Columbia | 278 | 870 | 1148 | 24.22 |
| Garfield | 300 | 417 | 717 | 41.84 |
| Lincoln | 539 | 2020 | 2559 | 21.06 |
| Okanogan | 1852 | 1744 | 3596 | 51.50 |
| Pend Oreille | 856 | 2119 | 2975 | 28.77 |
| Spokane | 26902 | 35055 | 61957 | 43.42 |
| Stevens | 2204 | 3524 | 5728 | 37.55 |
| Walla Walla | 2188 | 3639 | 5827 | 37.55 |
| Whitman | 1370 | 3239 | 4609 | 29.72 |
| TOTAL | 39152 | 56972 | 96124 | 40.73 |
Here are the vote totals for Peter Goldmark and Cathy McMorris, again by county:
| County | Goldmark | McMorris | % (D). | Relative Performance (%) |
| Adams | 172 | 750 | 18.66 | -3.40 |
| Asotin | 1599 | 2016 | 44.23 | 0.00* |
| Columbia | 214 | 680 | 23.94 | -0.28 |
| Garfield | 208 | 373 | 35.80 | -6.04 |
| Lincoln | 439 | 1461 | 23.11 | +2.04 |
| Okanogan | 1737 | 1557 | 52.73 | +1.23 |
| Pend Oreille | 737 | 1656 | 30.80 | +2.03 |
| Spokane | 24118 | 28834 | 45.55 | 2.13 |
| Stevens | 1968 | 3070 | 39.06 | +0.59 |
| Walla Walla | 1745 | 3245 | 34.97 | -2.58 |
| Whitman | 1248 | 2698 | 31.63 | +1.90 |
| TOTAL | 34185 | 46340 | 42.45 | +1.72 |
* Estimates; Secretary of State's election results website has not provided (D)/(R) split for the raw ballots in Asotin county yet, although they did provide (obviously) counts in the Goldmark/McMorris race.
I used the percentage from that race to estimate the (D)/(R) ballot split in the first table, although doing so obviously that means that we can derive no useful relative performance data for Peter in that county.
Observations
First, the bad news: As a straight vote-count, Peter loses this election 42.45 percent to 57.55 percent. But, of course, that isn't the whole story because it's a partisan primary where one's selection of who one can vote for is limited.
Now, some good news: Peter out-performed expectations. That is to say, the actual number of votes cast for Peter exceeds the number you would expect to see given the overall party split in the total number of ballots cast.
It's not a big amount, that +1.72% at the bottom of the second table, but it's there. Particularly heartening is that Peter's best relative performance is in Spokane county, which has the lion's share of the votes. That's not the whole game of course, but it's a good sign, and indicates that Peter's message and/or Cathy's poor job performance are having some effect.
Further, Peter's largest negative performances (in Garfield and Adams counties) were also in two of the smallest counties. This is not to say that Peter should think he can blow off campaigning there, but that poor performance there will certainly hurt him less than elsewhere.
[Fair warning: The rest of this analysis is to a degree speculative, because there isn't raw data that directly addresses the difference between our partisan primary system and the open general election. I am forced to make some assumptions and educated guesses along the way. I will tell you what those are when they occur, and you're entirely free to leave some comments telling me why my assumptions and guesses are no good.]
So what's it all about?
In a word: undervotes. A word we all learned about from Florida 2000, ballots in which no vote for a candidate was recorded.
Undervotes in this election are 4967 for Peter, 10632 for Cathy. Or in other words, 4967 ballots marked as (Democrat failed to indicate a vote for Peter, while over twice as many Republican ballots failed to indicate a vote for Cathy.
The obvious and hopeful conclusion we would like to draw from that is that ten thousand Republican voters across the district would prefer to vote for Peter, but can't do so in the primary.
While that would be nice, a) it's still not enough to overcome the 17819 vote difference between total Democratic and Republican ballots in the whole election, b) if we do that then we have to be fair and assume the converse position for the 4967 Democratic undervotes and assign them to McMorris. The truth is almost certainly somewhere in the middle.
I find it more instructive is to look at undervotes as a percentage of total D and R votes cast. Because it certainly is significant that McMorris had twice as many undervotes as Goldmark.
What we really want to do is estimate, out of those undervotes, the percentage of Democrat and Republican crossover voters. Let's do the math:
(D) undervotes: 4967 out of 39152 total ballots = 12.69%
(R) undervotes: 10632 out of 56972 total ballots = 18.66%
In an uncontested partisan primary, which both of these are, there are three main reasons why undervotes could occur.
- the voter just didn't bother to mark a vote for the sole option available to them.
- the voter made some kind of mechanical error in marking their ballot such that the ballot couldn't be counted.
- the voter would have preferred to vote for the candidate of the other party.
If we make the generous (to Goldmark) assumption that every single Democratic undervote was of type 1 or 2, then we would expect a similar percentage of Republican undervotes due to those same reasons.
That leaves an excess difference of almost 6% in McMorris' undervotes, which we can conclude may represent the percentage of Republican crossover voters who would have preferred to vote for Peter. That assumption - 0% (D) crossover voters and 6% (R) crossover voters - yields a general election prediction of Goldmark 44.29% / McMorris 55.71%. Still a win for McMorris, but not by quite as much as in yesterday's raw results.
Of course even that generous assumption for Peter, that no (D)ocrat would vote against him, is almost certainly too generous.
Just to get a sense for what might be going on, let's imagine that a full one-half of the Democratic undervotes were of type 3. That's surely more than reality, and as such represents some sort of worst-case scenario. That would leave 6.35% of type 1 and 2 undervotes as the base undervote rate, meaning that McMorris now has an excess difference of 12.31% in her undervotes!
The revised prediction, sending the crossovers from both sides over to their preferred candidates, is actually slightly better for Peter: Goldmark 45.44% / McMorris 54.56%. This counter-intuitive result is because the assumption forces us to also conclude that the Republican crossover-rate is higher.
The Elephant in the Room
What we can't measure, of course, are voters who indicated no party preference. Those voters did not get to voice an opinion on Goldmark or McMorris. So you need to appreciate these results with that large grain of salt. We just have no idea, on the basis of the primary election results, what the heck those people are feeling or how they might vote come November 7th.
Conclusions
Any way you slice it, at this point McMorris is seemingly still winning this race. Goldmark and his campaign crew clearly have a bunch of work left to do between now and November 7th.
There are positive signs, in that his message seems to be resonating in Spokane, Pend Oreille, Lincoln, and Whitman counties.
However, something is clearly not working in Walla Walla County, and perhaps there the message needs to be tweaked for those voters' more specific concerns.
I don't live there, so I won't claim to have any idea what those are, but clearly Peter's people need to take a look. Personally, I think that Peter's core strategy of crafting a one-two punch out of farming concerns and homeland security, in the form his concept of "national security crops" is brilliant.
Tying those in with the more left-leaning concerns over global warming and the environment makes it a one-two-three punch. I do believe that if Peter can effectively communicate this message across his district, then he has a shot at winning. I still think it's a fairly long shot, but the target is visible.
One thing, though: even the most generous analysis of yesterday's election puts Peter 8700 votes behind in the general. It will be tempting to concentrate in Spokane county, where Peter is already doing better than the party split would indicate, and where the bulk of the voters are. But I think that would be a mistake.
To overcome that vote difference in Spokane County alone would mean taking that county 57% to 43%. That's a mighty tall order. For now, I think, Peter needs to stay on the road, getting his great message out there.
