New poll results out today from WBUR and MassInc indicate a close race for the US Senate in Massachusetts between incumbent Republican Scott Brown and likely Democratic challenger Elizabeth Warren but how does an astute political observer determine which polls matter and which trends to watch for?
Both Real Clear Politics and TPM use a simple moving averages (TPM includes less trusted online surveys and neither differentiate between likely voter polls and registered voters polls) to demonstrate trends and leaders. The disadvantage here is that both methodologies ignore many of the points outlined in Silver’s critique, including factors such as sample size, pollster rating and freshness date. Yet the approach is widely appreciated by the public because the methodology is transparent and easily understood.
On the flip side Silver’s methodology is far more detailed and I would argue far more accurate, yet much more difficult to comprehend and thus harder to defend. Silver has not posted any 2012 US Senate polling forecasts.
Further none of the three methodologies provide crosstab data – only top line data – so it is difficult to determine electoral opportunities within the numbers. For example, in a tight race 45%-42%, no tracking data is available for key populations – party affiliation, gender or age.
Do you find polling averages helpful? What would you like in a #masen polling average/forecast?