It’s been 6 years since our mentor, colleague, and friend’s death. Warren Mitofsky was a clear thinker and major innovator of the public polling community. Beyond his methodological rigor, he communicated long-lasting, yet, simple messages to the profession. His thoughts remain vital through the 2012 election cycle.
Despite this year’s successful scientifically based public polls, the road was rocky, beset by a drum-beat of critics. Yet, Warren’s frequently uttered message, now ably echoed by Joe Lenski, remains a guide. “Believe your numbers!”
If your methods are scientifically sound, and
…you uncover unique results which pin the tag “outlier” on your findings, believe your numbers.
…you have a wider than expected spread in party identification, that brings a cascade of unwarranted criticism about weighting to party, believe your numbers.
…you are labelled a “newcomer” to Florida polling when you have Obama +2 and other long-standing polls have Romney +6, it isn’t a “house effect”. Believe your numbers.
…you detect a changing demography… an increase in minorities… in your likely voter models, it may simply reflect changing demography. Believe your numbers.
…more voters are telling your interviewers that they have already voted than are being reported by state tallying sources, it may reflect a time delay in mailing and recording early votes. Believe your numbers.
And, if you are being hammered for belonging to a conspiracy of pollsters who are cooking numbers and skewing results, stay focused.
Yes, it was “shoot the messenger” time and public pollsters were definitely in season.
Warren also advised us to always, always, always, poll right up to Election Day, even if you opt, to avoid confusion with Election Day exit polls, not to release the poll. Recognizing that campaigns don’t stop when you finish your “final” survey, sometimes a week out, there just might be something to be learned for future elections about the electorate and your likely voter models with this “exercise.”
We forgot his sage advice on the eve of the 2008 New Hampshire presidential primary when Hilary Clinton “upset” Barack Obama. It would have saved us re-calling our respondents all week to ascertain the late movement among women to Clinton.
This year, the initial impact of Hurricane Sandy was picked up in our pre-weekend NBC/WSJ/Marist Polls of FLOHVA — Florida, Ohio, and Virginia. But, was there any late movement on the eve of the election? We decided to invest, as per Warren’s dictum, in one last poll, bringing the grand 13 month total to 53 surveys. Sunday and Monday, we conducted a national survey and found Obama +3 among registered voters and +2 among the likely electorate.
There were many juicy poll nuggets in this survey including information about independent voters, approval ratings, the electorate’s view of the direction of the nation and the economy, minority participation, and where undecided voters were likely to end up. This all provided a context for Tuesday’s official tally and will guide our polls, especially our likely voter models, in future election cycles.
So, Election 2012 is now comfortably in our rear view mirror. Thanks, Warren, for being the lead driver once again.