Reflections on a Recount

Daniel Borenstein wrote this weekend about how John Perez does not deserve all the attacks he has drawn for taking advantage of the state's recount rules.  I agree with Borenstein's premise that much of the criticism is unwarranted.  However, I differ in that much of the criticism could have been avoided had the recount request been structured differently to better reflect both the letter of the law and the potential for public backlash.

So where did the recount go wrong?  I'm not talking about the campaign.  Anytime you lose by a tiny fraction of the vote you can think of a million decisions you might have made differently.  I'm talking specifically about the period between waiting for Lake County to (finally) report until conceding the election and how small changes might have reduced criticism and limited and damage to the Perez “brand name”. This post is a reflection of the mantra “first do no harm.”

My guess is that even with these changes the recount still would have failed. It is just too hard to make-up votes in a recount with current voting technology in a state that defaults towards counting versus rejecting ballots. 

1) ASK FOR A STATEWIDE RECOUNT: Note, I did not say pay or even commit to a statewide recount.  I said: ask for a statewide recount.  In 2000 Gore lost significant ground in the public relations debate because he only asked for some  counties to be counted.  Indeed, that strategy was later exploited by the U.S. Supreme Court with their "creative" ruling against the Vice President. 

A statewide recount request could have still have specified which counties and precincts be counted first.  When asked why prioritize the order for the recount, the answer would be not wanting to waste local election officials’ time and expend money until a more limited count showed whether the results were legitimately in doubt.  This would be defensible and understandable reasoning.  If Perez had then started making up ground (at the same time the  money ran short to finish the entire state) editorial boards would have been focused on criticizing state law not the candidate.  If there was no ground being made up and he pulled the plug, his message that he wanted to be sure that EVERY vote was counted might have had more resonance.

2)  OR REQUEST WHOLE COUNTIES BE RECOUNTED: At a minimum, the recount should have requested whole counties to be counted.   Indeed, California requires an entire county be recounted to change the final outcome.  So what was the good of doubling-down on the flawed Gore strategy by not only requesting only some counties to be counted, but only some parts of some counties? 

Again one could have still requested certain precincts and certain counties be done first.  As the Kern/Imperial counts showed, it would have taken a long time before any Betty Yee precincts would be reached.  The trick is you want to hand count your precincts and machine count your opponent's.  The former is more likely to change the result than the latter (and a machine count is far cheaper than a hand count).  The memo requesting the recount stated, "For each of the precincts specified in my list, I am seeking a manual recount.  For each of the remaining precincts in those counties, I reserve the right to recount them manually or by machine as the recount proceeds."  The specificity of the request made the selectiveness too transparent. 

Compare instead, "I am seeking a manual recount of all precincts in the specified order.  I reserve the right to request a change to a machine count for remaining precincts as the recount proceeds."  Then,  when the first Yee precincts were reached, the request could have been updated to switch to a machine count in that county.  When a reporter asked why, limited resources  could have been used as a shield to explain the shift.

Instead the "cherry picking" seemed so obvious that it handed Parke Skelton (Yee’s consultant) a talking point.  And you just don't empower Parke like that.

3) FOCUS MORE ATTENTION ON REJECTED BALLOTS: There were really only three places that Perez could have made up enough votes to close the gap.   The first was if something systematic was happening in “his” counties to cause Democratic voters generally and his voters specifically to have their ballots disproportionately rejected.  That is why his team made a wise choice to start with Kern County to see if Democratic (and Latino) ballots were disproportionately being thrown out in Republican counties. 

The second was to look at statewide systemic problems, but only seek redress in his best counties.  Perez's best results were on Day 2 in Kern when he added four votes.  All of these came from ballots initially counted as overvotes (i.e. someone the machine said voted for two people).  One can see how this would happen.  A voter marks Candidate A.  They meant to vote for Candidate B.  They don't understand  (or care enough) to get a new ballot, particularly if they are a mail ballot voter. So they cross out Candidate A, mark Candidate B, and trust or hope that the machines can figure out how they voted. 

As Borenstein points out, part of the reason the request was so selective (and thus so open to criticism) was because the state requires those requesting a recount to pay their way and Perez had less money for a recount than most assumed.  When resources are limited, you focus on the low hanging fruit.  Back during the days of punchcards, the low hanging fruit was indeed a full count in your best areas.  For example punchcards would give a different count each time you ran the ballots through the machines as more and more chads would be knocked off. 

But punchcards are no more.  Optical scan equipment is VASTLY more reliable.  Now the low hanging fruit consists of those ballots not counted before, either because the mail or provisional ballot was tossed before being opened or because the ballot was marked as an over or under vote. 

Focusing on these rejected ballots would have produced more potential votes more quickly.  Fundamentally, public opinion was going to turn within a week unless enough ballots were changing fast enough to throw the results into doubt.  Thus the recount request needed to be designed to maximize the odds of making up ballots as quickly as possible. 

4)  PUT MORE PRIORITY ON LOS ANGELES:  The other place Perez could have picked up ballots was in Los Angeles.  And that is not because he won Los Angeles or even because of Los Angeles' size.  Well not only for that reason.  Rather, as discussed before, while most of the state radically changed their voting systems after the 2000 election, Los Angeles made much more modest changes.  LA's InkaVote is really just a modified punchcard where the pin to punch a hole has been replaced with a pen to mark a target.   

This means that LA's voting system is more prone to differences between a machine count and a hand count.  For this reason if the recount had included a handful of LA precincts to be counted in the first week, he may well have picked up a few votes.  Due to LA's aforementioned size, even small margins from these precincts would have made it much easier to argue to the skeptical public, press and donors that the results of the election really were in doubt. 

If you count five of Imperial's 159 precincts and make up one vote, at best you can say you are on pace to pick-up around thirty votes.  But if you count five precincts of Los Angeles' 4,870 and make up one vote, you can say you are on pace to make up more than a thousand votes from just that one county.  This would have insulated Perez from the Yee press operation and allowed more time for the recount to play out (and raise more money to pay for the remaining areas). 

5) ANTICIPATE TIMELINE QUESTIONS: A 39 page recount request just begged the question: How is it possible for this to get done?  The request could have been much simpler.  We want you to count all 58 counties (or at least all 15).  Please start with these three (adding LA to their initial two).  For these counties start with these precincts.  After these initial results we will provide direction for the remaining counties. 

This would have eliminated the stories about the count going past November.  It would have maintained the ability to ask for a concurrent (rather than consecutive) count in the remaining counties.  And again, if anyone asked what was the plan to finish in time the response would have been simple, "We are starting with the areas that will tell us whether the election results are truly in doubt before asking for an expanded recount.  We wish we could have requested a statewide recount to start but unfortunately state law and our limited resources forced us to make a more narrow request."

BOTTOM LINE- Perez was not at fault requesting a recount nor was its failure the fault of those interpreting the election results.  However, it did not fully integrate this statistical information with the nuances of California election law and knowledge of inevitable press and public reaction.