Campaigns and elections

Forecasting Spanish Elections: 2011 Spanish Pre-Election Report

Sep 21 '11

We are pleased to welcome back “Pedro C. Magalhães”:, currently a FLAD Visiting Professor at Georgetown University and a previous contributor of two Monkey Cage election reports on Portugal (see “here”:http://tmc.local/blog/2011/05/04/june-2011-portuguese-parliamentary-elections-pre-election-report/ and “here”:http://tmc.local/blog/2011/05/04/june-2011-portuguese-parliamentary-elections-pre-election-report/) for the following guest post on Spain’s forthcoming “November parliamentary elections”:,_2011. The post is based on collaborative research with “Luis Aguiar-Conraria “: and “Michael Lewis-Beck”:

Last July, “Prime Minister José Luis Zapatero announced early legislative elections in Spain”:, which will take place in November 20th, 2011 (instead of the early Spring of 2012, when expected). There seems to be little doubt that the incumbent Socialists of PSOE will lose a significant share of the vote and that Mariano Rajoy’s Partido Popular (PP) will win. But what will be the magnitude of the Socialists’ punishment? The latest polls, measuring voting intentions two months before the election, point to a result of no more than 32% for PSOE, and suggest that the nomination of Alfredo Pérez Rubalcaba as the Socialist candidate may actually be hurting the party’s prospects (see “here”: or “here”: If this is indeed what awaits the Socialists, one would have to go back very far to find such a dismal performance for PSOE. Only before 1982, i.e., before the major electoral realignment that wiped out the Unión de Centro Democrático (UCD) from the party system, would we be able to find similar results for the Socialists. And this would also mean that only in 1982 would an incumbent party have ever performed worse.

Speculations about the fate of PSOE in these 2011 elections have almost exclusively been based on poll results, rather than on any other means of forecasting elections. But voting intention polls conducted a few months or even a few days before the election are not necessarily good forecasting instruments on their own. In 1996, while polls conducted just before the Spanish election suggested a PP lead over PSOE of about 10 points, the actual margin of victory ended up being a single percentage point. In 2000, the very latest polls awarded an average lead to PP over PSOE of about 5 points, but on election day that advantage reached 10 points and PP obtained its first absolute majority ever (see “here”: for an analysis of the two cases). In 2004, the terrorist attack of March 11th and its aftermath rendered pre-election polls understandably less useful and, in 2008, most media polls did end up converging on values close to the 44% of the vote and 4 point lead that PSOE would ultimately obtain. But just two months earlier, the reputed Centro de Investigaciones Sociológicas (CIS) estimated 40% of the vote for PSOE and a lead of just 1.5 points over PP.

In the United States, first, and now in many Western democracies, statistical models to forecast election results have become increasingly common, but not in Spain. This may seem odd, considering that voting behavior research in Spain has burgeoned, including the part of it that treats a central aspect of most election forecasting models: the role of the economy. However, there are perhaps two reasons for this absence of forecasting models in Spain. First, since Spain’s democratization, only 9 legislative elections have taken place, and only 8 of them (starting in 1979) can be said to represent a judgment over an elected incumbent. Second, studies about the economy and the vote in Spain show a relative uneasiness with the classic reward-punishment view. This includes past findings such as a “positive relationship between unemployment rates and government popularity”:, “shifts through time in the role of unemployment and inflation in shaping incumbent support”: as measured in polls, and the finding, using individual-level survey data, that the potentially negative effect of unemployment in the fate of the Socialists’ vote tends to neutralized or even reversed by the kind of welfare policies adopted by the PSOE in government and mistrust of the Partido Popular’s stance on social policies (see “here”: or “here”:

In a “paper”: we (Michael Lewis-Beck, Luís Aguiar-Conraria and myself) wrote this past June, we take both these issues into account, presenting a model and conditional forecasts for the Spanish election. We recently “updated the paper”: to take into account the scheduling of the election for November, which gives us the ability to make an actual forecast. On the one hand, rather than using the observations concerning the 8 relevant legislative elections, we increase our sample size by including also 6 European Parliament elections, expecting that, as the “second-order elections scholarship”: has long suggested, they should be marked by a dependable loss for the incumbent. On the other hand, we assume that the incumbent’s vote share should be a function of both economic performance and government popularity. Popularity is measured by surveys conducted by CIS since 1979, which provide an evaluation of the “political situation” of the country (actual government approval questions have only been asked since 1993). For economic performance, we focus on the “big two”, unemployment and inflation. All are measured with a 6-month lag in relation to the election. Focusing on these two macroeconomic variables also allows us to disentangle between different theories regarding how the economy affects political behavior:

* a simple “reward-punishment” model, assuming that any incumbent is affected by negative realizations of the unemployment and the inflation variables;
* Hibbs’s “partisan theory”, assuming that left-wing incumbents are mostly hurt by increases in unemployment while right-wing incumbents are negatively affected by increases in inflation;
* “policy-oriented” voting (Kiewiet 1983), assuming that different parties may actually benefit from a degradation of the economic problems to which they are seen as attaching highest priority;
* and finally, the kind of hypothesis that derive from the investigations by Maravall and Fraille using individual-level data in Spain: a “partisan neutralizing” effect, whereby PSOE’s stance in social policies in Spanish democratic history has resulted in incumbent Socialist governments remaining immune to increases in unemployment.

Our results suggest that incumbents in Spain are punished in European Parliament elections and that their electoral performance is also significantly affected by leading (6-month) popularity measures. Furthermore, we find little support for a simple “reward-punishment” or even for partisan or policy-oriented models. Instead, while all kinds of incumbents seem to be negatively affected by increases in inflation, only right-wing incumbents (but not PSOE) seems to be punished by increases in unemployment. The final forecasting model for the proportion of incumbent valid vote we estimate is the following:

VOTE = .368 – .039*European – .013*Inflation – .011*Unemployment*UCD/PP Incumbent + .008*Political Situation + u

The R^2^ (r-squared) is .95, all variables are significant with at least p<.05, and diagnostics show no evidence of linear dependence between the variables or non-normality. Out-of-sample forecasts show that the largest out-of-sample error takes place in the transformative 1982 election (when the UCD was nearly wiped out). For the remaining elections, the mean absolute out-of-sample error is 2.8 percentage points. Jackknife procedures also suggest great stability for the estimated coefficients.

Since the election takes place in November, we have all the data we need for an ex ante forecast. In the CIS Barómetro, the percentage of respondents who declared the political situation to be “good” or “very good” last May comprised was only 3%. The inflation rate was 3.5%. Since the values for European and UCD/PP Incumbent are 0, this results in a forecast for the incumbent (PSOE) valid vote of 34.5% (with a forecast standard error of 3.3%). The figure below places this forecast in comparison with past PSOE and incumbent party past electoral results.

Incumbent party and PSOE vote shares in the Spanish legislative elections (%)

A realization of this forecast would mean that PSOE would lose close to 10 percentage points in relation to the 2008 election, reaching the second lowest result ever for an incumbent in legislative elections. At the same time, such a result would still be above the lowest scores PSOE obtained before the 1982 election, similar to PSOE’s result in the 2000 elections (when PP triumphed with an absolute majority), and somewhat above current poll estimates.
Besides the usual sources of uncertainty, two merit special mention. First, the 2011 election will be just the third time an incumbent party has presented itself for a legislative election without being led by the incumbent prime minister. The first time was in 1982. Second, as “Marta Fraille”: conveyed to us in a personal communication, one open question is whether the recent austerity measures have affected PSOE’s reputation as party committed to mitigating the social consequences of unemployment and its actual ability to target the relevant segments of the electorate. But this is, at the moment, unknown.