It’s become the stuff of Latin American Legend, and a favorite Caprilista talking point of the last few months: despite trailing badly in most well-regarded polls, Nicaragua’s Violeta Chamorros trounced the Sandinistas in the 1990 Presidential Election by 54% to 40%. Lesson: in a highly conflict-ridden and polarized Latin American context, people lie to pollsters about voting intentions.
But how applicable are the lessons of Nicaragua in 1990 to Venezuela in 2012? Well, in this 1994 paper for the American Journal of Political Science, John A. Booth reviews the Nicaraguan experience in detail.
Here are the key factors he identifies for why so many pollsters got it so wrong:
1-Narrowness of Focus. The view that Daniel Ortega was a shoo-in for re-election in 1990 came if you looked at Presidential Voting Intention question in isolation. In fact, the same surveys were strewn with hints that sandinista support was soft. This came out in two ways: a high number of “Undecideds” in the sample, and Presidential Approval and Favorability ratings that were out of whack with the voting intention numbers. Too many people were telling pollsters that they didn’t approve of Ortega’s handling of his job, didn’t like the guy…but intended to vote for him anyway. Interestingly, there was no gap between favorability and approval ratings in polls that showed Chamorro ahead and in those that showed Ortega ahead.
Venezuela Similarity Quotient: Mixed. As discussed ad infinitum in these pages, many Venezuelan surveys really do show oddly high numbers of undecideds. However, Chávez’s approval and favorability numbers have not gone noticeably soft in the last 12 months.
2. Sampling Problems. Booth says, “in the two decades since the last census, the country had undergone two wars, a revolution, natural disasters, massive urbanization, heavy outmigration and sharply accelerated population growth… Not even for the capital, Managua, did up-to-date dwelling unit maps exist.” These data limitations wreaked havoc with pollsters efforts to construct a reliable sample.
Venezuela Similarity Quotient: Low. With as many as 10 national votes in the last decade, Venezuelan pollsters have plenty of experience constructing reliable samples with the data in hand.
3. Sample Comparability. Pollsters in Nicaragua weren’t always straightforward about whether they had polled only Managua, large urban areas or the whole country, all voters, only registered voters or those judged likely to vote.
Venezuela Similarity Quotient: Low. Serious Venezuelan pollsters always report characteristics of their sample, though some go further in polling small-town voters than others. (A C21 strong suit.)
4. Allocation of Undecided Voters. In Nicaragua, “not all the candidate preference results reported in the various polls were based straightforwardly upon respondents declared preferences.” Some pollsters went to town “assigning” undecided voters to one side or another on the basis of follow-up probes to pressure them to declare a leaning, or deleted undecideds from the results as if they did not exist.
Venezuela Similarity Quotient: Undetermined. I’m not sure if Consultores 21 and Varianzas, which show lower undecided tallies than the rest, are doing this.
Booth discusses field experiments by other researchers that suggest just how sensitive voters were to unconscious cues of pollster partiality. Something as subtle as the color of the pen a pollster was using to write down answers could, according to some researchers, be read as a sign of partisanship by respondents and skew responses.
Booth notes that one interpretation is that pollsters that respondents perceived as pro-Chamorro produced more accurate results than pollsters that sought to present themselves as neutral (perhaps because they were suspected to be pro-government). He estimates that 1-in-6 respondents who told neutral-seeming pollsters they would vote for Ortega actually went on to vote for Chamorro.
Interestingly, the best-performing pollster in 1990 was a Venezuelan outfit called DOXA, affiliated with COPEI, that was widely perceived as pro-Opposition in the Nicaraguan context. Some researchers have speculated that the perception of DOXA as a pro-Chamorro pollster was the key to its success. They’re still around, in one form or another, and they think Capriles is winning the battle of the NiNis.
So there’s a bit in that paper for everyone. To me, though, it suggests one key question to ask is whether there’s a systematic gap between IVAD, Datanalisis, Consultores 21 and Varianzas in terms of Chávez’s favorability and approval ratings, as well as his voting intention. So…I’m adding that to my To Do list.
[Hat Tip: Capablanca]