Tag Archives: opinion poll

El régimen de Maduro solo tiene 10% de respaldo

El estudio Assessment of Voter Intent Using Three Different Polling Methods in Venezuela October 2015: Evidence for Fear Bias in Traditional Polls and Media Bias in Online Polls concluye que el respaldo actual para el partido oficialista en Venezuela es de tan solo 10% ±5%. El estudio se basa en tres encuestas paralelas, o mejor dicho, las mismas preguntas por tres métodos distintos para poder medir y corregir por sesgo. La evaluación de las respuestas revela:

    • Encuestas tradicionales (los que vez en los medios) tienen un sesgo, conocido como “factor miedo”, y por eso subestiman el respaldo por la oposición mucho
    • Encuestas en Facebook tienen un sesgo que se puede llamar “factor medio”, y por eso subestiman el respaldo por el oficialismo algo
    • El respaldo real por el PSUV en las parlamentarias es de 10% ±5%
Graph of intention of vote
El respaldo al PSUV es 10% ±5% dejando 90% ±5% para la oposición

Abstract

Las encuestas de opinión en Venezuela no son fiables, y tampoco lo son los resultados de las elecciones. Se ha argumentado que existe un “factor miedo” en las encuestas tradicionales. Por otro lado, ahora es fácil de realizar encuestas en Facebook, pero las respuestas en línea son muy diferentes tanto de las encuestas como de los comicios. Este estudio se realizó para evaluar la precisión de los diferentes métodos de encuesta, y para estimar el verdadero respaldo para el régimen. Para evaluar los factores que influyen en la respuesta, se hizo una serie de preguntas sobre diferentes temas en la puerta de domicilio, en la calle, y en Facebook. En cada caso se les aseguró el anonimato, pero el análisis sugiere que los encuestados desconfían del encuestador en diversos grados, dependiendo de método de encuesta. Un análisis de la correlación entre el uso de los medios de comunicación y las respuestas revela una correlación en ciertos temas que sesga los resultados de Facebook. Encuestas tradicionales suelen subestimar severamente el apoyo a la oposición, y las encuestas de Facebook suelen subestimar el apoyo al régimen. Sobre la base de este estudio, se estima que el verdadero apoyo al régimen en un 10% ± 5%.

Quiero enfatizar que los que participaron en ejecutar la encuesta dentro de Venezuela, violando las leyes de la dictadura, son héroes de la democracia. No puedo mencionar sus nombres porque pueden ir presos o peor, pero un día su Patria les agradecerá, Dios con ellos.

El manuscrito ha sido enviado a un periódico cientifico internacional con “peer review”. Como su evaluación no termina antes de las elecciones, y el resultado puede ser de interés público, respetando las normas el manuscrito como presentado ha sido “self-archived” en la página personal del autor. 

Flawed Opinion Poll in NarcoNews

The website “Narco News” has made available an opinion poll made in Honduras in August regarding the political crisis. According to Narco News it shows that the majority of Hondurans want Zelaya to be reinstated, and they are against the military coup. Let us look at the report.

First, in the introduction the author explains that the purpose of the poll is “to know, in a trustworthy way, the political situation in the country after the coup d’état against president Manuel Zelaya Rosales on June 28th. As is known, Roberto Micheletti Bain is presiding over a de facto government, which, since the moment it took the power, has received forceful rejection from the majority of the Honduran population, which daily expresses itself by means of demonstrations by the National Resistance directed by prestigious popular leaders and honest politicians”.

Let us stop right there. Is this report biased or what? Only a complete fool can believe that anything that follows that statement can be unbiased, let alone trustworthy. That sentence is a strong declaration of support for one side in the poll.

From now on, we are no longer looking for signs of bias. We know it is there. The only thing left to look for is how they allowed their bias to influence the results.

They interviewed 1470 persons, one per household, in 54 neighbourhoods distributed over 16 out of the 18 “departamentos” in the country. The neighbourhoods were allegedly selected randomly, but the households within each neighbourhood was selected as every 4th or 5th. About 11% had university eduction, 12% had no education whatsoever, the ages ranged from 18 years to above 60, and about half were presently employed. These types of questions (including sex, religion) suggest that there is, at least to me, no obvious bias in the sample.

However, the devil is in the questions.

The first question is “¿Está en favor o está en contra del Golpe de Estado del 28 junio pasado al Presidente Manuel Zelaya Rosales?” (Are you for or against the coup d’état of June 28 against president Manuel Zelaya Rosales?) This is a very biased question, designed to get a specific result. It is like asking a jury “Do you think the defendant was right or wrong in killing his wife?” instead of asking “Do you think the defendant killed his wife?”

Furthermore, many are of the opinion that it was ex president Manuel Zelaya Rosales who planned to do a coup d’état on June 28th. Thus, a small minority might actually have misunderstood the question altogether, and replied that they were “against the coup”, when the correct answer as the question was posed would have been “for”. In Spanish the difference between the two questions is “pasado al” (against), and “pasado del” (of), so when read out they sound virtually the same.

Again, there is no credibility in the unbiasedness of the poll. The fact that only 17.4% responded “for” does not mean much when the question is so obviously biased, and many surely did not dare to be honest in the violent environment that existed, when they knew what the opinion of the pollster was. Most of the 29.9% who did not respond must therefore be suspected to be sympathetic to the democratic institutions who deposed of Zelaya.

The second question was if the Micheletti government should stay in power or leave power. Here 60.1% said that he should leave, and 22.2% said he should stay. Given that he has to leave by January 27th, it is not clear how the respondents would have interpreted the question. There are quite a few people who wish Micheletti could stay beyond January 27th. Was that how they understood the question?

On the third question it gets more substantive: Do you support or not support the return of Manuel Zelaya Rosales to the presidency?

51.6% support, 33.0% do not support it, and 15.4% did not answer. It is mostly the youngest and the oldest who are in favor of his return, and men more than women. But then something strange appears: People with the highest education level are most in favor of his return. This runs contrary to all my experience so far, and it suggests to me that there is something fishy about the data that goes beyond biased sampling, biased questions, and biased attitudes in asking the questions.

The sixth question is, “Do you agree with the repression or do you condemn the repression that the military and police have carried out against the National Resistance?” Again, a clearly biased question. The unbiased version would have been, “Do you agree with the way in which the security forces have strived to uphold the order in the face of street protests?”

In the seventh question the respondents were asked to point out which group was behind the coup: politicians, business interests, military, foreign interests, all of the above, none, or don’t know/no answer. Again, what is the respondent to reply who does not view it as a coup, but a constitutional succession? It’s like asking, “Have you stopped beating your wife?”

In question eight they ask for the favorability rating of Ramon Custodio, ombudsman for human rights, CONADEH. They find it to be 21.4% good or very good, and 39.3% bad or very bad. However, in question eleven, where they ask for favorability of a number of persons, they again include Ramon Custodio, who now, strangely, only has 12.4% good or very good rating.

In the questions on from which radio and TV channels the respondents get their news, the bias in the sample becomes clear. In both cases the pro-Zelaya channel got the highest percentage, even though they are not the largest channels in the country. The pollsters spin this to mean that people have changed viewing habits since they know that the pro-government channels are just lying.

On the question if the general and constitutional elections should be held or canceled, 66.4% said they should be held, and 23.8% that they should not.

Conclusion

It is evident from the questions that the respondents are not a random representation of the population. Those who are favorable to Zelaya are seriously over-represented. We see this from what radio and TV channels they watch.

How can that be, if they selected the respondents randomly?

The simple answer is, they didn’t. Remember that they selected the neighborhoods randomly, but not the houses and persons in the houses. They took about every 4th to 5th home, they wrote. They had 1500 surveys filled out and discarded 30 for being filled out erroneously. However, a figure that does not appear anywhere is how many persons they approached in total. It is impossible that they only approached those 1500, as any pollster can tell you. Most people refuse to participate, which is the main reason for the bias in the results in any poll.

In this case those who designed the poll had built the bias into it on purpose. Remember the first question? Many who got that as the first question would most likely slam the door – or something worse – on the pollster. By starting out with such biased language they assured themselves a seriously biased result in their favor.

And that is all there is to it. Lies, damned lies, and opinion polls.