Spain: the last polls

The Spanish election is less than a week away. In the last few days, several polls coming out from several different sources. Here's the updated graph, for the three major parties, since Rubalcaba became PSOE's candidate (smoother 25% bandwidth):

The smoother seems to capture a slight decline for PP and a rise for IU, but that could be a function of the particular mix of companies that have done polls recently and the kind of "house effects" attached to their polls. A better way to ascertain this is to regress scores on dummies for companies (leaving CIS as the reference category) and dummies for time periods (months, in this case, considering the small number of polls). The results suggest that:

1. After their peak in April, PSOE lost about 3 points until today and has remained mostly stable since September.
2. Estimates clean of house effects for PP show a very small decline from October to November (.6%) but that only leaves them at the point they already were in September. Effect of the debate? An hypothesis, but impossible to say just on the basis of these data, really.
3. IU is the only party experiencing significant changes recently, up more than 1 p.p. since September.

Starting tomorrow, publication of polls in the media is not allowed in Spain.

P.S.- For estimates resulting from a Bayesian state-space model for pooling poll data first developed by Simon Jackman, see here.