What went wrong with the polls in the last Catalan elections?

by Marc Guinjoan, Associate Professor, Pompeu Fabra University

What is the story?

The 25th of November, Catalans elected their new regional parliament. In my previous blog post on the subject, I pointed out that there was little uncertainty about the results. All opinion polls indeed showed a strong convergence towards a clear repartition of vote share. In The main uncertainty laid in the decision of the new government to hold or not a referendum on self-determination.

I must acknowledge that I was wrong. The actual results differed drastically from poll estimates. Figure 1 shows these discrepancies. The 2007 electoral results are also presented. The right-wing nationalist party CiU maintains as expected its leadership position. But it did not gain a majority of parliamentary seats nor massive popular support. The CiU was expected to gain between 62 and 66 deputies and eventually obtained only 50 out of the 135 seats of the Catalan parliament. The discrepancy is partly explained by the important gains obtained by the left-wing secessionist party ERC, which became the first opposition party, over-passing the left-wing federalist party PSC (which was still ruling the Catalan government only 3 years ago). The right-wing unionist party PP experienced a small gain, as well as the eco-socialist pro-referendum party ICV. The unionist C’s also increased their electoral support, while the far left-wing secessionist party CUP entered the Parliament with 3 deputies.

Why did the opinion polls fail to anticipate the electoral results?






Figure 1. Electoral results and polls in the 2012 Catalan elections

First explanation: Quality of the polls

Since the election, the quality of opinion polls in Spain has been largely discussed (see for example this post, in Spanish). Newspapers typically conduct opinion polls on small non-representative samples For instance,

The proportion of citizens choosing to answer surveys in Catalan is systematically higher than the actual proportion of citizens talking this language at home.

Although national and regional opinion polls institutes usually correct for these problems, there remains some doubt about the overall validity of the polls.

Second explanation: An increase in turnout

As shown in Figure 2, the turnout in Catalonia tends to be lower in regional than in national elections. This phenomenon of differential abstention has been pointed out by some scholars. They argue that some voters only identify themselves with the national Spanish institutions and therefore tend to abstain in regional elections.

Due to the large focus put on the possibility of holding a referendum about the independence of Catalonia during the campaign, those usual abstainers may have decided to show up at the polling booths the 25th of November.

There was indeed an increase in the participation rate of almost 11 percentage points, compared to the 2007 elections.






Figure 2. Participation in the Catalan and general elections, 1977-2012

In a following blog post, I will perform an ecological inference analysis to test whether this hypothesis is supported by empirical evidence. Additionally, I will analyze the consequences of these electoral results with regard to the debate on independence and the likelihood of a referendum.

Follow us on Facebook and Twitter.

This entry was posted in Blog. Bookmark the permalink. Both comments and trackbacks are currently closed.

One Comment

  1. Carol Galais
    Posted December 13, 2012 at 4:21 PM | Permalink

    It is definitely a shameful year for polls (if we exclude maybe the US). I would add some tips to Marc’s reflections
    1) The language in which the survey is conducted may play a role, but CEO let people choose between Catalan and Spanish. Nevertheless, the language of the first contact is always catalan, and this may still condition the answers. Thus, they are about to conduct some experiments to test if contact language affects the results.
    2) The quality of the polls includes sampling problems. I have the feeling that pollsters have become “lazy” when they have to sample zones that are (in)famous for their levels of abstention.
    3) I would also draw attention to administration modes. How many young people have landline telephones (CEO’s administration mode)? If you combine this with quotes, things get even worse. You’ll find at the end of the day a young working class male between 25 and 30 years old at home during working hours that will pick up his landline telephone and answer the survey, but he won’t probably be representative of this population group.
    4) Another problem is the amount of people that decided their vote at the very last minute, and probably –to some extent- on the basis of these very wrong surveys…