A NOTE ON PRIMARY DATA
Just as it's wrong each time the GOP says "Americans voted for Trump," because they didn't—the Electoral College elected him, the raw vote favored Clinton—you can't talk about "who Nevada voters voted for" using delegate counts rather than raw vote totals.
Just as it's wrong each time the GOP says "Americans voted for Trump," because they didn't—the Electoral College elected him, the raw vote favored Clinton—you can't talk about "who Nevada voters voted for" using delegate counts rather than raw vote totals.
PS/ A lot of people, in and out of media, are playing games with data, and it's tiresome. If you want to talk about Sanders' delegate haul in Nevada, the number is 46.8%. If you want to talk about *how people voted*, the number for Sanders is 34.0%. Anything else is a lie—period.
PS2/ I have people coming on my feed claiming that "progressives received [56.6%] of the vote in Nevada"—a number they fraudulently arrive at by adding up *delegate counts*. But when you say "vote," you're referring to *raw votes*—for which the "Sanders + Warren" figure is 46.9%.
PS3/ Bad data leads to terrible political analysis. I'm a progressive—so I want the "Sanders + Warren" vote total to be over 50% as much as anyone—but I'm sorry, it's just not (yet). What we saw Saturday was 46.9% of Nevadans showing up at the polls to vote for Sanders or Warren.
PS4/ Bad data is *also* being used to misstate how much more popular Sanders was than other candidates when Nevadans showed up at the polls. For instance, using *delegate counts*, people wrongly say Sanders was 37% more popular than Warren. Nope—the figure is 18.7% (still a lot).
PS5/ Media is to blame for this confusion—not voters. The delegate percentages are relevant for *one night*—to determine how many delegates each candidate gets from Nevada (a miniscule number, compared to what's needed for the nomination)—while raw vote totals matter *urgently*.
PS6/ *Any* conversation about "how [Candidate X] is doing in the Democratic primary" *must* be launched using the raw vote totals in a caucus state—*not* the byzantine processes caucuses use to determine SDEs ("State Delegate Equivalents") or CCDs ("County Convention Delegates").
PS7/ Take Warren, for instance. There's one analysis for how many delegates she will/won't get from Nevada, and it matters only a bit—as Nevada offers an infinitesimal percentage of the delegates a candidate needs to get the nomination. The raw vote totals tell a different story.
PS8/ In the raw vote totals, we see Warren 4.7% out of second place in Nevada. We also know that 75% of Nevadans voted early—and something like that number made up their minds on who to vote for weeks (not days) before the caucus. The "days before" data is dramatically different.
PS9/ So if someone is a well-paid contributor at CNN or MSNBC and they're giving viewers *delegate counts* as a sign of a candidate's *current* popularity—or information from *people who made up their mind when the race was in the state it was weeks ago*—they're misinforming you.
PS10/ Because everyone has their preferred candidate, however, CNN and MSNBC get *rewarded* for misinforming us because there will always be a large bloc of voters who *like* the data they're hearing from CNN and MSNBC, even if it's not measuring what the analysts say that it is.
PS11/ I use reliable NYT/WP data that is *actually measuring what it says it's measuring*. So yes, the raw vote totals from Nevadans who decided who to vote for in the week before the Nevada caucus (rather than before) give you the *best* sense of the *current* state of the race.
PS12/ Since we don't have raw vote totals for late deciders, the *best* we can do to see the *current* state of the race is to (a) use *overall* raw vote totals, and then (b) note that those figures must likely be *amended* in the *direction* of *entrance polls* on late deciders: 
PS13/ But even those numbers, which would suggest that (say) raw-vote-total late deciders went "between 33% and 25% for Sanders" and "between 13% and 19% for Warren"—skewing toward the latter number in each case—aren't fully predictive, as media reporting must also be figured in.
PS14/ That is, because media reports wrongly equated delegate counts with candidate popularity—then compounded the error by not distinguishing between early and late deciders in discussing *a current snapshot of the race*—future voters may be influenced by such erroneous reports.
PS15/ Unfortunately, Twitter is the *worst* place to point all this out, as it is—we all agree—such a cesspool of inter-party and intra-party bias that anything I say, even if *100% factually correct*, will appear to be nothing but bias in favor of my preferred candidate, Warren.
