Let's-Flog-This-Dead-Horse-S'more Chronicles

Listen, I remember the criticisms from 2004 because I made a lot of them. When fraud claims came following the Recall Referendum then, they all seemed to be...

Listen, I remember the criticisms from 2004 because I made a lot of them. When fraud claims came following the Recall Referendum then, they all seemed to be based entirely on arcane mathematical theory. The opposition didn’t really have a coherent, compelling narrative about how the seemingly weird patterns in the election data might have come about, they just had some strange-seeming numbers. And strange seeming numbers aren’t really enough to hang a fraud claim on.

Fast forward nine years, and we’ve way over-learned that lesson. These days, the opposition does  have a coherent and compelling narrative about the how of fraud – witnesses were prevented from watching the audits in selected voting centers, and ballots were stuffed at those centers – but it has entirely foregone trying to show that with actual data, falling back on a generic demand that an obviously uncooperative CNE provide with documentation it will self-evidently never give up.

We’ve gone from bald to two-toupee’d in just nine years, and it’s unfortunate. Because while the 2004-era argument that numbers-without-a-political-logic do not a contestation strategy make still holds, it seems clear to me that if you really want to establish your case to the neutrals, you need to show the fraud in the data as well.

This is especially odd because the alleged rigging mechanism would clearly leave visible statistical traces. For ballot stuffing to make a difference, you’d have to do it on a scale that would be obvious in the data CNE did release. Ballot-stuffed centers should have both noticeably higher turnout and noticeably higher propensity to vote for Maduro.

In a scatterplot, that would look something like this:

Slide1With the visible cluster in the upper-right quadrant giving the game away. Instead, here’s what the actual graph looks like, courtesy of D. Kronick:

real data
Source: CNE. Lowess curve: degree one polynomial, tri-cube weight function, bandwidth=0.05.

So prima facie, there’s no there there.

Of course, this analysis might be too simple by half. You might need a more sophisticated test that matches the results from voting centers where you think ballots were stuffed with results in near-by, otherwise similar centers with a history of voting similarly, and then show the disparity. Or think up some other smart way to look into this that we haven’t thought of yet.

But looking back, it’s hard to find clear evidence that there were wild swings between voting behaviour back in October and then last April. In other words, if they stuffed ballots on April 14th, they did it in the exact same voting stations as the ones they stuffed last year:

Graph_Turnout_June3

Again, it’s possible that more sophisticated research would uncover a pattern of irregularities that match the opposition’s allegations that we’re not seeing here.

What’s troubling is that either nobody in the opposition has bothered to do this research or nobody has thought it germane to circulate it. One way or another, I sure haven’t seen it. (If you’ve done it, send it! I’ll gladly eat my post…)

Instead, what I’ve heard is a screechy pseudo-maracucho blathering about whether you’ve ever met anyone who voted for Maduro but not for Chávez.

Which makes for an ok soundbite, I suppose, if soundbites are the main thing you’re interested in. But it doesn’t get you very far if your main concern is figuring out, y’know, what actually happened.