It was a survey, not really a poll
The Jefferson Herald had a dose of Andy Warhol’s “15 minutes of fame” a few decades ago, stemming from a presidential preference survey of Greene County political junkies.
One day in late winter or early spring, I think about 20 months before the presidential election, Bev Lehman, Bee and Herald writer, and I, as editor-publisher, were musing about the Iowa caucuses.
Bev was a dyed-in-wool-Democrat (still is) and I was a Republican.
The caucuses were nearly a year away. We decided to try a survey of the county’s political junkies to determine which candidates were leading among that group in both parties.
How were we going to determine who was such a junkie?
The best way, we decided, was to check the lists of attendees at the most recent precinct caucuses.
Those would have been held in the previous off-year, a year when caucuses were held but not a presidential election year. (In Iowa, elections and caucuses are held in all years ending in an even number — in other words, every two years; but presidential caucus years occur once only every four years.)
Who would attend a political party caucus in an off-year?
Only those whose politics are so deeply embedded that they will turn out for a winter’s night meeting to elect local party officials and debate the wording of platform planks.
They would be true political junkies.
So we secured the lists of caucus-goers in the county from the county auditor’s office. I don’t think you could do that today for free; the parties maintain ownership of those lists now and sell them to potential candidates. But back then they were available at no charge from the auditor.
We picked them up, took them back to the office and looked them over. As luck would have it, each list contained exactly 50 names — 50 Democrats and 50 Republicans.
One hundred phone calls to make around the county.
Bev took the Democrats and I took the Republicans, and we started calling. We assured each of them that we were not going to print their names. We simply wanted to calculate percentages of support for each of the presidential candidates, and to publish those numbers.
We were pleased that nearly everyone was willing to give us their preference, or to tell us they were undecided if that was the case for them. We did the math and published our story.
Then we started getting phone calls from national media.
As I recall, the Baltimore Sun called, as well as the New York Times and the Washington Post. They had read a story from the Des Moines Register about our poll, and were hungry for political data from Iowa — any data.
They figured our poll could be spun as a bellwether from the first-in-the-nation caucus state.
I explained to the callers that the poll was definitely unscientific, even though Jefferson is the hometown of George Gallup. We had called only those who attended the previous off-year caucuses, I said.
What’s more, the evangelical wing of the Republican Party was just starting to take form with the Christian Coalition, and many of those activists would be voting for Pat Robertson. But most of them had not attended those off-year caucuses, and consequently Robertson’s strength was understated in our survey.
All that didn’t matter to the media reporters who called us. They simply wanted numbers, and no one else in Iowa apparently had them at that early date. So they took our data and reported it.
Several months later, in the fall of that same year, about three or four months before the approaching presidential caucuses in Iowa, Bev and I decided to take the survey again, to see if there had been any movement in presidential preferences among the county’s political junkies over the summer.
We made our calls again and published the results, explaining that it was simply a survey, with no real scientific validity.
Again we got calls from national media political reporters.
“I understand you’ve got data showing movement among the presidential candidates in your county,” each of them said.
I explained once again how we had taken our surveys, and that the results were emphatically not reliable as predictors. For instance, I said, a couple among those surveyed the first go-round had left for Arizona to visit a relative, and we couldn’t reach them by phone in the second survey. That alone skewed the results for their candidate by six percentage points or so.
And the Pat Robertson factor was still going to be miscalculated.
But all that still didn’t matter. What the media wanted was a story, and they thought we had one for them. Several of them did stories, with a “Jefferson, Iowa” dateline.
The episode taught me to read stories based on political polls suspiciously, first checking a poll’s methodology before taking its results seriously.
The Iowa Poll conducted periodically by the Des Moines Register is now the gold standard as far as I’m concerned; Nate Silver for the New York Times is also excellent.
The jury is still out for most of the others.