Driven to Dispersion

Over on YVPolis, Iñaki Sararzazu has a last word on Venezuelan public opinion research in 2012, and that word is: appolling!

Getting beyond the ulcer-inducing guessing game about which one is right and which one is wrong, Iñaki gets at the bigger problem: pollsters just don’t release enough information for us to make an educated guess about the source of the problem.

Until Venezuelan pollsters start to be forthcoming about exactly their samples are constructed and weighted, how exactly their questions are phrased and sequenced and the exact percentage of respondences to each individual question in their questionnaire (including a per-question NS/NC number) we’re going to be in the dark.

What’s clear is that normal statistical variation cannot account for the crazy dispersion of the polling data we’ve been fed this time. Someone is seriously screwing this up. And we don’t have the information we’d need to figure out who it is.

In such a weird setting, it’s foolhardy to place too much faith on any one poll.

25 thoughts on “Appolling!

  1. I’m done thinking about the polls. I simply have no mental space for it anymore – unless they show Capriles leading big.


  2. This is the nature of statistics and forecasting. In a sample of 1000, only twenty vote intentions are needed to switch the polls 2 percent from one candidate to another. However, the polls that show Chavez leading by 10 to 20 points are just wrong.


    • But those things are taken into account when we talk about confidence intervals and error margins. There are solid tools that produce acceptable results if applied correctly.

      What Iñaki is pointing out is that there’s something deeply wrong with Venezuelan pollsters. In a normal electoral scenario you would expect these two things:

      1. Some consensus on trends and forecasts among pollsters.
      2. Decreasing variance as election day approaches.

      And neither is happening. How and why? Well, we could speculate about the multiple causes, but in my opinion “the nature of statistics and forecasting” is at the bottom of the list. To give you an example from Iñaki’s blog, look at the results from GIS XXI — they reported a highly unlikely and sudden upsurge for Capriles after placing him way below the average for months. It’s ridiculous.

      I really, really recommend Iñaki’s series of posts “combinando las encuestas” and, if you are versed in mathematical statistics, the paper that describes his method (here).


      • I recommend, as I have for years, not to bother paying attention to pollsters. Here’s why.

        1. When these polls are commissioned ($$$) by Fulano, that very act compromises objectivity, taints methodology, and produces questionable outcomes. So, is it any wonder why the pollsters don’t fully reveal their methodology used on any given sample population?

        2. In a highly dynamic race, as this one has been, the lag between the questions asked and the poll results produced yield expired results, insofar as ‘reading’ voter moods and intentions.

        3. The fear of losing one’s job, given repeated threats experienced by a large swath of voters in government (and military) posts, many of whom have families to feed, will result in skewed answere, either for Chávez or as a Ni-ni.

        To-date, I’ve not seen any pollster, or any analyst, make reference to the above variables. For all these reasons, NOT ONE poll should be taken seriously. Perhaps some day, when Venezuela’s political process gains maturity and stability, as a fully fledged and intelligent democracy, then number 3, at leasst, will be eliminated as a deviation to the polling results, and perhaps, so will number 2.


        • This is fine advice for anyone looking for closure or reassurance. The thing is that Iñaki’s points would be valid even if we weren’t amidst an election cycle. Something as simple as having pollsters publish their datasets (not just the methodology, but the actual answers from each and every respondent in the sample) would largely solve problem (1) because anyone with a computer can run his own tests and determine if bias is present and if the pollster’s conclusions follow from the data. You will know who used tendentious questionnaires, who fudged responses and who failed to apply the methods correctly.


  3. An authority once said, “It’s difficult to make preditions, especially about the future”, which in the Venezuela of today, Friday, October 5, would seem to sum it up. Is anyone taking bets on how much Monday-morning quarterbacking there’ll be on….Monday morning?


    • I was just about to write that Monday morning’s going to be fun, when a cold sweat broke over my brow, as I thought in horror of the possibility that the cancer-ridden concrete block in office might remain, beyond January 2013..


    • I hate sounding pessimist but watching closer the level of the Chavista parafernalia and how it’s been managed with no scruples it remains a gigantic feat to defeat the incumbent.
      Bonds are rallying already, but the bet is not so much whether Chavez losses but instead, that even if he wins he’s done within 2 years. They probably don’t know this is the biggest masterpiece of G2 in decades.


      • Dunno, M. The way they been running the campaign tells me that the guy is sick. How sick is he, I cannot tell. But the guy wouldn’t risk such a low profile campaign when so much is at stake. Were Chavez healthy, he’d have given Capriles a run for his money in the campaign trail. But he did not. He stayed put most of the time in Miraflores and did the bare minimum. Yes, he invested lots of resources, but it was not the same.
        That does not mean that the guy is gonna die tomorrow. Just look at the SoB of Fidel. The guy’s been sick for several years and he’s still there. He may even outlast some of us…


      • The bond market is extremely short-sighted. It is NOT betting on events more than a few days in the future. “Capital is a scared rabbit”, as the saying goes, scurrying for a safe hiding place. Post-election violence, even if Capriles wins, as he should, will see the bond market go down.


        • “Venezuelan bonds are posting the biggest rally among major developing nations as investors bet that President Hugo Chavez’s tenure will end soon even if he wins the closest election he’s faced in 14 years.” Bloomberg “Chavez Losing Invincibility Tag Spurs Election Bond Rally”


  4. Asesinan a dos personas dentro de discoteca en principal de Las Mercedes
    Un grupo de afectos al oficialismo que venía del acto de cierre de campaña de Chávez disparó contra empleados y asistentes al local nocturno porque no los dejaban entrar, ya que se encontraban en estado de ebriedad. El hecho ocurrió la madrugada de este viernes.

    2 people were killed & between 15 & 18 injured when a group associated with Chavismo arrived at the disco at 2 AM as it was closing demanding entry. As they were drunk they were refused so they opened fire.


    • 6 PM Friday until 6 PM Monday however anyone with half a brain has already stocked up for the big party on Monday.

      I guess that means most of the Chavistas will be sober while the Caprilistas will all be drunk. :-)


Comments are closed.