New Hampshire Suffolk Tracking Poll

Let’s be a little cautious on the role of Suffolk’s tracking poll on setting Romney expectations. The plot above shows that the Suffolk track for Romney differs quite a bit from the trend estimate based on all other polls, excluding Suffolk.

Suffolk’s final track has Romney at 33. The trend estimate without Suffolk is 40.6, or 39.2 for the sensitive estimator. That’s quite a difference. None of the other candidates are so substantially different between Suffolk and the other polls trend.

Interpreting this is a dilemma. Suffolk has a good record. In 2008 the Suffolk poll had the smallest error on Clinton’s surprising win among all active NH pollsters (though like the rest it underestimated Clinton, by 5, and had Obama winning as did everyone.) But past record is not a good predictor of future accuracy. Year to year variation is pretty random.

So the concern is that without Suffolk, we see Romney as holding flat at about 40. Suffolk alone would predict a substantial decline, into the low 30s.

My standard estimate, shown in the previous post here, puts Romney at 38.8, while considering all polls. I think we might want to stick to that balancing of the evidence rather than embrace the tracker alone or the most sensitive estimators. At the least, acknowledge there is enough difference among polls to raise our uncertainty about Romney’s standing.