On account of the underestimate of the difference between SYRIZA-ND (New Democracy Party) given by gallups and exit polls in the recent elections of 20th September 2015 (which rose to 7.4% instead of the predicted 1.5-3% on average), the discussion regarding the credibility of such research and, in extent, their use emerges again.
Public opinion research is conducted by companies upon a client’s request (usually the media but often enough political parties as well); the aim is to examine the electoral body’s voting tendencies and to analyze more aspects of such tendencies (demographics, age, shifting, unity, motives, etc.) in order to deduct further useful conclusions (which demographic categories are expressed by which party, where a given party is winning/losing, etc).
However, in the context of relentless TV antagonism between privately owned channels during hours of high viewing (7-10pm) on election night, exit polls are the subject of confrontation and, therefore, the deciding factor on media profits. Gallups, on the other hand, almost monopolize the media’s pre-election interest.
Like any scientific tool, a political behavior research should be conducted based on prerequisites such as reliability and lack of bias. This means:
- A minimum sample of 1,000 people.
- Weighting based on previous preferences.
- Sample should be representative both in terms of age and demographics.
Like many other things in modern Greek society, public opinion researches were first deified and then demonized. But be calm. Let’s put things in perspective.
Gallups/exit polls are, as it is commonly said, “instant photos”. In practice, this means that they imprint absolutely instantly citizens’ fluid (by nature) intentions. This demonstrates the arbitrariness of published “vote estimate”, which differs from voting intention on this significant point: unspecified votes (invalid, blanks, abstaining, undecided) have been removed and, usually, it is considered that undecided voters will behave in a similar way as those who have reached a decision.
However, this cannot be proved. Undecided voters may be influenced by a number of factors, such as which party is favored, negative votes, angry votes, other people, etc. Proof of the arbitrary reducing of voting intention to vote estimate is the fact that, in the last elections, gallups gave SYRIZA an average estimate of 31.2% while exit polls gave the party 33.0% – in the end, the party ended up gathering 35.5% (it is possible that this percentage is the result of withheld guilty vote, a historically recorded phenomenon due to various reasons).
Moreover, in the last few years, research companies have four significant methodological problems to face:
- The large number of people who deny answering questions (in the context of lacking the mood or patience to do so, and the inclusion of such researches in the general apathy towards politics).
- The false declaration of preference (a characteristic example being the referendum of 5 July 2015).
- The fact that many voters, who are otherwise willing to partake in research, do not remember what they voted in the last elections (based on which the weighting of the sample is predominantly made), due to the many elections in recent years.
- The partiality in choosing a sample (a characteristic case was the overestimate in difference between Rena Dourou and Giannis Sgouros in the prefecture of Attica, in the regional elections of May 2014 – where, instead of a predicted 56-44%, the actual numbers were 51-49%; expert Elias Nikolakopoulos explained this by proposing that company supervisors were not adamant in checking their employees about not accepting the participation of “proud-to-be-SYRIZA” volunteers).
Research companies have a long way to go before they adequately solve these problems.
Greek media use research of this kind not only to manipulate the public (according to whatever their business interests dictate each time) but also to achieve viewership (TV-channels) or sales (newspapers) increase. One does wonder how meaningful it is for the credibility of such research to be at stake because of the way it is deified in order to satisfy the needs of the three-hour hysteria on election night, when the Ministry of the Interior has managed (using SRT) to provide us with a safe estimate of the results for the last few years; an estimate that usually does not deviate more than 0.2% from each party’s final results.
In conclusion – especially on account of the experience we gained from the recent national elections of 20 September 2015, we come to the following “lessons learnt” regarding reading such research:
- We read the findings with coolness and do not pay too much notice to decimals. A 0.1% in a sample of 1,000 people is one person.
- We follow the general trends (e.g., who is in the lead and what margins there are for further unification), as well as the most important shifts.
- We show an interest for the origins of undecided voters, who often return to their original “source”.
- The only historically irrefutable pointer is the “victory show”, meaning the prediction of a winner (what was once called “the current”).
- When there is a limit to parliamentary representation (like, for instance, a 3% in Greece), we do not consider safe the parliamentary representation of a political party that is not systematically recorded as safely above said percentage. For example, in the last elections, The River (Potami) had an average of 5.5% in gallups and finally received 4.1% – while Popular Unity (Laiki Enotita) was recorded at 4.3% and received, in the end, 2.9%.
- Finally, we take into account the sum of published researches, as well as their clients’ interests.
If all that had been achieved in the last elections, we wouldn’t have to deal with the hysterics of certain political leaders against the “easy prey” of research, which is called to trace the tendencies of a society that is overall discrediting towards politics. Between deification and demonization, there is the middle ground of coolness when dealing with such research.