On Ballot Stuffing, I’m a Glass-99.86%-Full Kind of Guy

0

On this blog, we’ve always stressed the difference between "Hard Rigging" an election – actual, Persian-style numerical fraud – and "Soft Rigging" it through an accumulation of unfair practices. In Venezuela, evidence for the former is extremely scant, while evidence for the latter is all around. That still leaves the possibility of an "In-Between Rig": stuffing ballots in tables with no opposition witness.

In principle, I have to accept this is possible: with no witnesses there to raise a stink, there’s nothing to stop chavistas in out-of-the-way rural voting centers from helpfully voting "in place of" their neighbors who "forgot" to turn out and vote for the government.

Concerned about this, I decided to do a quick bit of research into the possibility by looking at the Mesa-by-Mesa data from the 2009 referendum.

My conclusion?

Yes, ballots probably are getting stuffed in some places where the opposition has no witnesses. But, the practice isn’t widespread, it’s rare.

Perhaps 100-200 mesas are involved out of the more than 30,000 mesas used for the referendum.

Specifically, I could find just 108 mesas with Turnout higher than 85% where the "No" got less than 5% of the vote. (We’re talking about 0.3% of the mesas in the election.) That includes a handful of mesas where the Sí got 100% of the vote!

I interpret that as prima-facie evidence of ballot stuffing – poll workers voting in place of people who didn’t turnout, and for the government. In those 108 mesas, the government got 21,974 votes.

Remember that essentially all of those mesas are in rural areas where the government was easily getting 60% of the vote without cheating, we can say perhaps 40% of those votes were a "Ballot Stuffing Bonus" – additional votes the government got thanks to ballot stuffing and wouldn’t have gotten without it.

The ballot stuffing bonus, nationwide, was in the neighborhood of 9,000 votes. That’s about 0.14% of the government’s total nationwide vote.

If you’re a glass-half-empty type, you note that yes, there is evidence of ballot stuffing already in place in 2009. If you’re a glass-half-full type, you realize that it’s hard to do this in big enough numbers to really make a difference. 

Me? I’m a Glass-99.86%-Full kind of guy.

Faced with this, longtime Caracas Chronicles comment-section vago and leading ballot-stuffing theory proponent Kepler noted that this kind of really blatant ballot stuffing may only have affected 100someodd mesas, "but that does not mean there was no cheating in loads of other places where Chavismo won with 65% or 70%. We just cannot prove it."

Thinking about this, I realized that’s not quite true. If it was happening, we would be able to tell, because diffuse ballot stuffing would leave its tell-tale statistical mark on the results.

In effect, if the theory is that you would add a few hundred extra "Sí" votes to the government’s tally in Mesas with no opposition witnesses, then you would expect to see a systematic difference between turnout rates in otherwise very similar tables – including tables within the same voting center. In this scenario, some mesas would end up showing:

  1. Suspiciously high turnout, and
  2. A systematically higher share of pro-Chávez votes

That’s what you would expect to see if diffuse ballot stuffing was really happening wherever the oppo had no witnesses.

Do we see that? This time, the answer is an unambiguous "no".

Since this part of the analysis was a bit more technically demanding, I asked Dorothy, a reader with a good statistics background, to run the numbers for me. Her conclusions:

  1. Only 79 of 30,004 mesas had abstention rates that were at least two standard deviations below the mean within their centro.  But these 79 mesas had lower than average Sí shares.  
  2. Among all mesas with lower-than-mean abstention rates (about half of the mesas, given the distribution), Sí shares were significantly lower than the overall average or than the average among mesas with higher-than-mean abstention. Like, six to twelve percentage points lower. This is true whether you compare the abstention rate to the overall national mean or to the mean within a given mesa’s centro. In other words, high turnout favored the No vote.

If you’d like to check, Dorothy’s Stata program is in the comments section. Preliminary analyses of the 2006 presidential election showed pretty much the same thing.

This is not to say that they won’t stuff the ballots next time. It’s just to say that there’s no prima facie statistical evidence that it’s happened in the past, aside from a handful of extremely sketchy results concentrated in Delta Amacuro State that involve a few thousand votes out of over 10 million votes cast.

Leave a Reply