Last week, I posited that David Alvarez lost the mayoral election  in January, well before Election Day on Feb. 11. He emerged from the primary very far behind his rival Kevin Faulconer and nothing changed by the time people started sending in mail-in ballots.
Alvarez actually lost it the moment that former Mayor Bob Filner resigned and set in motion a special election. The people he needed to vote simply would not vote in an election like this.
According to Kogan’s analysis, 64 percent of voters who supported President Obama in San Diego in 2012 did not come out to vote in the special 2014 runoff election for mayor last week. In the same analysis, only 23 percent of people who supported Mitt Romney failed to vote in this special election.
And that’s at the heart of Alvarez’s loss. The coalition behind Alvarez spent millions trying to avoid this and failed.
Here are Kogan’s conclusions in two tables (click on to enlarge):
Kogan shared the analysis with me. He used a technique called ecological inference regression and he consolidated the precinct data to determine how many Obama voters came out and how voters for the 2012 mayoral candidates split this time. He said the conclusion may change when all the provisional ballots are counted (but likely not).
Kogan argues that much of the punditry that says Faulconer won because he was a moderate does not play out in the data.
“Almost all of Faulconer’s votes came from DeMaio voters and almost all of Alvarez’s votes came from former Filner voters. Actually, marginally more DeMaio voters supporting Alvarez than Filner voters supported Faulconer,” Kogan told me.
Thus, the difference in the election was almost completely determined by differences in turnout.
Would a different candidate emerging from the primary, say Nathan Fletcher, have mattered? Kogan says maybe.
“Fletcher may have had a different base of support. Perhaps some people would have turned out who ended up staying home. But the gap was so large, it would have had to be a very big difference and a really good mobilization effort. It’s hard to imagine the difference being that huge,” Kogan said.
Kogan, a former writer at Voice of San Diego and co-author of Paradise Plundered , is now a professor at The Ohio State University. He helped U-T San Diego do a similar analysis after the 2012 election . You can bug Kogan about his findings at firstname.lastname@example.org.
Here’s a deeper explanation from Kogan about how he parsed the data considering that the registrar had different precincts in 2012 than it did for the 2014 special vote:
I basically disaggregated the 2012 election results into the home registration precincts, with votes from each consolidated precinct allocated to the home precinct in proportion to the percent of registered voters who were from that home precinct. I then re-aggregated the home precincts into the 2014 consolidated precincts. All of this is possible with the home-to-consolidated-precinct mappings that the registrar of voters provides.
With the 2012 election results now compiled into the 2014 precincts, I calculated for each precinct the following quantities: % of registered voters who supported each candidate; % of registered voters who stayed home in 2012. I then repeated the same calculations using the results from 2014. With these percentages, one is then able to run a statistical procedure called “ecological inference,” in which one uses aggregate (precinct-level) data to estimate individual (voter-level) behavior. There are quite a few statistical assumptions involved in this step and the math is way above my head. The method was invented by Gary King, a statistician at Harvard who wrote a book  on it.