Measure J Passed — and By More Than 15 Votes; Recount Would Be Hella Expensive, Likely Futile

The North OC Community College District’s Measure J bond apparently  — albeit barely — made its nut.  (Photo from Wikipedia.)

[Update, 12/4, corrected from 12/3]

County certification of election results comes due tomorrow.  The LA Registrar of Voters site shows its results as finally having ticked over from “semi-final,” to “final”; its results have not changed since Nov. 25.  So, pending any recount, Measure J has passed.

What Happens Now?

What happens now is that I tell you, from memory, about a conversation I had yesterday afternoon with OC Registrar of Voters Neal Kelley, who was kind enough to return my call from the road.  Under the circumstances, I didn’t ask him follow-up questions, but here’s the gist of what he had to say.

A recount of any or all precincts in either or both counties can be requested as early as Monday, December 8.  The deadline to request a recount would be Friday, December 12.  The period to request a recount in Orange County races ended a couple of weeks ago — but thanks to the presence of 16 precincts in Los Angeles County (mostly in La Habra Heights), this is not strictly an Orange County race.  The inclusion of those counties may turn out to be expensive for someone — or, if a recount by the “No” side is successful, for everyone, as the County would then cover the tab for labor.

The NOCCCD district received 163,667 of Orange County’s 640,358 ballots cast — just over ¼ of the total.  (I’m getting my numbers from the “Statement of the Vote” — an almost 2,500-page PDF at this link which I don’t necessarily suggest you read on your own.  Might be wise to trust me on this one.)  Neal Kelley estimated, if I understood him correctly, that they would quickly need to put together six tables of vote counters — which if I recall correctly is six people per table — who would need to start almost immediately and would continue on daily until they finished.  As a ballpark estimate, this might be $6000 to $7000 for around eight to ten days.  (In other words: it could bump into Christmas.  And I’m not sure that the crew gets to take off weekends and holidays.)  If I understood him correctly, after the recount, the entire county would have to be subject to another 1% canvass, so that adds another 6,404 ballots to the tab, bringing it to about 170,000.  (They won’t count only the 150,171 votes actually cast in this race; non-votes have to be veried just as votes do.)

The Register’s contention (see the first link below) that cost estimates for a recount in Los Angeles County begin at $5,054 per day seems unlikely, given that we’re only talking about a little under 4,000 votes in 16 precincts.  (That is: I don’t know that courts would uphold such a large “booking fee” on top of the actual cost of what would be 4-6 person-hours of labor.  Then again, the cost of a lawsuit to pull down the price of an LA recount may take up a large amount of what could otherwise be saved.)

Beyond that, the notion that an Orange County recount would be as little as anywhere near only $600 per day seems even less likely, given that 150,171 ballots were cast in OC in this race (and the ballots in the areas that left this race off of the ballot also need to be counted, so we’re actually talking about recounting 163,667 ballots.)  This compares to the 19,598 votes cast in Garden Grove, where Bruce Browadwater called it off after one day.  And this would require a whole lot more people than the OCROV had to hire to recount Broadwater’s race.  Those proposing a recount should budget for the prospect of paying $70,000.

Now, if they wanted to do it on the cheap, on the 8th they could request a recount of Los Angeles to begin on the 9th, see what happens, and have a recount of Orange County scheduled to begin thereafter.  They could stop on any day, after all; if they pull ahead at some point, the burden would fall on NOCCCD (or other “Yes” proponents) to continue the recall.

They could also do something potentially a little more sketchy: calling for a machine count of machine ballots and a hand-count of paper and provisional ballots.  This starts to get into some of the “equal protection” issues raised in Bush v. Gore, especially given that those casting provisional ballots are more likely to be racial and ethnic minorities — and that could mean litigation.  Neal Kelley has already made clear that he does not cotton to challenges of signatures on absentee ballots, rejecting 64 of 64 such challenges in Broadwater’s recount.  But the real trench warfare would come in provisional ballots — where No on J forces would try to prove that individual voters were not qualified members of the electorate.  This can be a very ugly and intensive form of hand-to-hand combat, with individual voters being located and potentially brought into court to prove their legitimate residency.

Batches containing provisional votes came in at about 8-9% more favorable to Measure J than did earlier ones.  If we figure that it takes 10 challenged ballots to yield one vote reduction in the 30.21 vote margin (see below), we’re talking about their coming up with 300 provisional ballots that can be successfully challenged, presuming that only about 70-85% of them would affect this race.  (And, remember — provisional ballot voters won’t necessarily go all of the way down the ballot.)  Orange County had 35,108 provisional ballots overall, so we’d expect maybe 9,000 of them to come from this race.  I’m guessing that no more than 8,000 of them cast cotes in this race.  So we’re talking between 4% and 5% of those votes — between 320 and 400 — being successfully challenged in order to have about an even chance of overturning the outcome.

That may not seem like much — but it’s a lot.  How successful a recount would be at eliminating ballots would depend on the rules that the individual judge placed in charge of the matter would apply — and, of course, on any appellate rulings that might take place during the process as well.  If it’s not clear by now why I’ve been obsessing about the exact vote margin in this race, by now it should be.

A Discussion of Marginal Interest

The Register says that Measure J won by a margin of 15 votes.  The Register is wrong.  The margin was about 30¼ votes, given the 55% margin.  The “Yes” side exceeded 55% by a little over 15 votes and the “No” side trailed “>45%” by a little over 15 votes; combined, that’s just over 30.2 votes.

Here’s an exchange I had in the Register’s comments section of their previous article on the topic with Connie Lanzisera (the bond opponent quoted in the article linked above) from my hometown:

Sure, Connie Lanzisera, anyone can say “recount” — but that’s not the same as doing it, which means paying for it.  Unless you’re going to ask provisional voters how they voted and then challenge only the ones that voted yes — in which event we will have a serious equal protection problem — every provisional ballot you eliminate will have one of three outcomes: Yes, No, and Neither.  When you eliminate a “Yes” vote, you eliminate one from the numerator of the fraction that has to be above 55% and one from the denominator.  When you eliminate a “No” vote, you eliminate one from the denominator, but nothing from the numerator.  When you eliminate a “Neither” vote, you affect neither numerator nor denominator.  Anyone who thinks that it’s a matter of eliminating just 15 provisional ballots is very mistaken.  Those ballots are almost as likely to push the percentage of “yes” votes up as they are to pull it down — and every time you subtract a vote from the numerator of the Yes vote, you subtract one from the denominator as well.

I’ll recopy the final vote totals from below:

Bonds – Yes 82,751 + 2,029 = 84,780
Bonds – No 67,420 + 1,918 = 69,338
Total votes – 154,118

Or, in percentage terms:

Bonds – Yes 84,780 = 55.0097976875%
Bonds – No 69,338 = 44.9902023125%

If you were able to surgically remove nothing but Yes votes from the total, you’d need to remove the following numbers to get at these percentages — and I’m giving you a range of them so that you can get the flavor of how much each vote matters:

30 (84,750/154,088) = 55.0010383677%

33 (84,747/154,085) = 0.55000162245%

34 (84,746/154,084) = 54.9998702007%

35 (84,745/154,083) = 54.9995781494%

40 (84,740/154,078) = 54.9981178364%

So removing 33 votes isn’t enough; you need 34.

But, of course, you can’t surgically remove only “yes” votes; you challenge the ballots while they are still “in the wrapper.”  (You can fight your hardest to eliminate a ballot and then find out that, surprise!, it would have favored you.)  To get a net reduction of one vote, on average (if the provisionals are 10% more favorable towards Measure J than are other votes — 60% vs. 50%) you may need to challenge about 6 of them for each vote you need.  So, for example, successfully challenging 240 votes means that you’ll eliminate 40 votes for Neither, and of the remaining 200 you’d eliminate 120 Yes votes and 80 No votes — just enough to win.  But slight changes in the expected percentages of how many vote this way, (or vote at all) among the population of successfully challenged ballots, as well as random fluctuations, can still leave you behind.

I think that, for someone with $100,000 or more to spend, a recount might be worth trying.  But, I don’t expect it to change the result unless the luck of the draw brings forth a judge who dislikes provisional voters and disrespects Neal Kelley.

[Update, 11/25]

Contrary to the hypothesis in the previous update, we were not done.  Another 22 votes have trickled in — 15 of them Yes on Measure J and 7 of them No votes.  As with that last batch, that is over 2/3 support when only 55% is needed, so unless LA has some huge “Damaged Ballot Reversal” like that which stopped Jay Humphry’s surge cold in Costa Mesa, the outcome remains determined.  Is such a reversal possible, thought?  Until they’re done done counting, the answer remains “yes.”  But they say that they’re now down to their 1% canvass, so only a handful of votes might change in LA County, and those could change in either direction.

The new results from LA County are: 2029 Yes and 1918 No.  That means that 15 Yes Votes and 7 No votes were added — that’s 68.18% of the new batch, much like the supposedly statistically shocking 63% and 64% that late updates to the OC figures brought (when, most likely, some provisional ballots got added to the mix.)

So our new totals are:

Bonds – Yes 82,751 + 2,029 = 84,780
Bonds – No 67,420 + 1,918 = 69,338
Total votes – 154,118

Or, in percentage terms:

Bonds – Yes 84,780 = 55.0097976875%
Bonds – No 69,338 = 44.9902023125%

The Register, by the way, misstates the margin as being 15 votes.  If 15 votes were to switch from Yes to No, the measure would have 55.0000648853% of the vote — still enough to win; it would take switching 16 votes to pull the total below 55.000…%.  But “how many votes switching would it take to reverse the result?” is not the right measure.  If it were, we would say that Bao Nguyen won by 8 votes rather than by 15, because a switch of only 8 votes would have reversed the result.

The right way to do this, I believe, is to add up the number of votes by which the Yes vote exceeds the “exactly 55%” margin and add to that the number of votes by which the No vote trails the “exactly 45%” margin.  This is essentially what we do in Garden Grove when we add together the 7.5 votes by which Bao exceeds the average — the 50% mark — between him and Broadwater with the 7.5 votes by which Broadwater trails that 50% threshold of the two-candidate count to get 15.  (This matches the margin between the votes because we’re using a 50% threshold of the number of votes cast for those two candidates.  The trick when it’s not a 50% margin is that we have to calculate the leader’s lead and the trailer’s deficit separately with respect to the number that each had to hit.)

So: 55% of 154,118 is 84764.9 votes.  The “Yes” vote exceeds that threshold by 15.1 votes.

And: 45% of 154,118 is 69,353.1 votes.  The “No” vote trails that threshold by 15.1 votes.  They would still have lost if they hit exactly 45, though, so let’s add .01 to that.

Combine them, and Measure J  wins by 30.21 votes.  I believe that we would have to round up, though, and say that it wins by 31 votes.  But I’m open the disagreement on that point.

In any event: it’s not “15 votes” unless you’re going to say that Bao won by 8 — which he did not.  You have to count both the surplus and the deficit.  Only Chris Nguyen and I may care about this, and I’m not even sure about him.

[Update, 11/22]

We may be done.  It’s not clear whether Friday’s update is the final one, or what we’re supposed to read into the term “semi-final,” if it isn’t.

The new results from LA County are: 2014 Yes and 1911 No.  That means that 41 Yes Votes and 20 No votes were added — that’s 67.21% of the new batch, much like the supposedly statistically shocking 63% and 64% that late updates to the OC figures brought (when, most likely, some provisional ballots got added to the mix.)

So our new totals are:

Bonds – Yes 82,751 + 2,014 = 84,765
Bonds – No 67,420 + 1,911 = 69,331
Total votes – 154,096

Or, in percentage terms:

Bonds – Yes 84,765 = 55.0079171426%
Bonds – No 69,331 = 44.9920828574%

How close is this, presuming that it’s over?  We can ask and answer that question in at least three four ways:

1) How many votes would have to switch from Yes to No for it to lose?

That would be 13.  An 84,753 to 69,343 margin (12 votes switching) would be 55.0001297892%.  An 84,752 to 69,344 margin (13 votes switching) would be 54.9994808431%.

2) How many Yes votes would have to be subtracted for it to lose?

You’d have to subtract 28 Yes votes.  Subtracting 27 would leave a fraction of 84,738/154,069 =  55.0000324530%.  Subtracting 28 would leave a fraction of 84,737/154,068 = 54.9997403744%.

3) How many No votes would have to be added for it to lose?

You’d have to add 23 No votes.  Adding 22 would leave a fraction of 84,765/154,118 =  0.550000648853%.  Adding 23 would leave a fraction of 84,765/154,119 = 54.9997080178%.

4) What is the sum of the margin by which the Yes votes exceed 55% and the No votes trail 44%+1/∞?

This is actually the correct method, and I show how it’s done with the next update above.  With a 50% threshold, this calculation is simplified to just subtracting the “No” total from the “Yes” total, which is the same as calculating (Yes – 50%) plus (50% – No).

So, however you slice it, this race falls in between Costa Mesa’s 47-vote margin and Garden Grove’s 15-vote margin — a little closer to the latter.  (The difficulty of figuring all of this out is a good argument for requiring only 50%+1 of the vote, because these calculations with a 55% margin are pretty nasty.)  Maybe some people on the No side will be willing to gamble about 6 times as much as will be required for the Garden Grove recount to try to shoot it down, but I doubt it — and if they try it, I doubt that it will reverse the results.

[Update, 11/19]

Where did we leave off with Measure J, which needs 55% support — not rounded off, but meaning more than 54.9999999999…% — to pass?  Let’s see: before those last 50 paper ballots, the score was this:

Bonds – Yes 82,733, 55.10316900001%
Bonds – No 67,409, 44.89683099999%

Add 18 yes votes and 11 no votes — 62.0689655172%, if anyone’s wondering — and we get this distribution of our 150,171 votes:

Bonds – Yes 82,751, 55.1045141872%
Bonds – No 67,420, 44.8954858128%

But that, of course, does not settle the issue.  The issue will be settled in La Harbra Heights (and a couple of blocks to its west.)  And that means that we’re waiting on the final returns from Los Angeles.

As faithful OJB readers know, the Los Angeles results from last weekend were this:

Bonds – Yes 1.946, 51.0627131986%
Bonds – No 1,865, 48.9372868013%

Which, combined with the previous OC total, was:

Bonds – Yes 82,733+1,946=84,679, 55.0031503121%
Bonds – No 67,409+1,865=69,274, 44.9968496879%

For some of us — me, Chris Nguyen, and probably at least several others — this sort of thing is exciting.

Well, I have good news for those few of us who care: LA updated its count yesterday!  It’s now:

Bonds – Yes 1.973, 51.0627131986%
Bonds – No 1,891, 48.9372868013%

That’s an addition of 27 “Yes” votes and 26 “No” votes.  In other words, 50.9433962264% to 49.0566037736%.  That’s as close to the previous percentage as you can get with a mere handful of votes.

[Note: this next section has been corrected]

So, in real time as a type, I’m going to calculate the new percentage across the two counties:

Bonds – Yes 82,751 + 1,973 = 84,724
Bonds – No 67,420 + 1891 = 69,311
Total votes – 154,035

From which we calculate:

Bonds – Yes 84,724, 55.0030837147%
Bonds – No 69,311, 44.9969162853%

Measure J is still passing.  And, if LA continues to come in at exactly 50% of the vote for Yes, it will continue to pass unless LA gets at least 96 more votes, at which point it would lose with 54.9999675601%.  (Actually, I only presume that that would be a loss.  Unless there’s a statute that specifies that you only go to seven or fewer significant digits past the decimal point, at least it loses in my courtroom.)  If only 94 votes come in at a 50% approval rate, it wins with 55.0000324404%.

We’ll update this post as we become aware that new LA returns have come in — or stopped doing so.

* * * * *

Here’s an interesting question, though.  The clock on the five days allotted to request a recount is ticking.  (And remember, this is about a quarter of the county, so it’s going to be really expensive.)  Let’s say that LA hasn’t finished counting before then.  Do either proponents or opponents have to call for a recount proactively, just in case LA puts them on the losing end — or can they wait?

I think that they can wait.  But if I had a lot riding on the answer, I’d be calling the ROV’s office pretty soon — if I hadn’t already.

* * * * *

As the question of how NOCCCD marketed the ballot measure has become a source of controversy here, I went to their page to find what they said there.  Here’s their description of what it will do:

Measure J

Now, did they market it that way, or in the way most advantageous to its passing?  Let’s just say that they did the same as pretty much everyone else tried to sell voters on their candidates or positions, so it one wants to call people out for that sort of thing, they’re not going to be anywhere near the front of the line.  Campaigning on one of one’s top three bullet points is pretty good by comparison to most efforts.