The Trafalgar Effect

By Jim Ellis

Oct. 15, 2020 — The Trafalgar Group is the polling firm that came to national political notoriety four years ago when they correctly predicted a Donald Trump victory in both Michigan and Pennsylvania and were the only survey research firm to do so. Since that time, they have forecast at least four other wins when the active polling community was arriving at opposite conclusions.

Yesterday, Trafalgar released its latest Pennsylvania data (Oct. 10-12; 1,034 likely Pennsylvania voters) and finds former vice president Joe Biden leading President Trump 47.4 – 45.1 percent — just over a two-point spread. In October, not counting the Trafalgar number, we see 12 other pollsters returning Pennsylvania data and they average a pro-Biden forecast of just under seven points.

Routinely, Trafalgar’s data shows President Trump in better position than most pollsters because they attempt to quantify what is termed the “shy Trump voter,” i.e., those who are actually voting for the incumbent but won’t admit it to a pollster. In most cases, the Trafalgar calculations, derived from a proprietary algorithmic formula, have been reliably accurate.

From 2016, we remember that, generally, the polling community missed badly in the Trump-Clinton presidential race. While their national count was accurate – predicting a tight plurality for Hillary Clinton (final result: 48.2 – 46.1 percent) – many state projections were off, particularly those in the Great Lakes region.

In the previous presidential election cycle, a total of 62 surveys were conducted in the state of Pennsylvania, and only three found a lead for President Trump, including the Trafalgar pre-election survey. In Michigan, 45 polls were publicly released, and Trump led in just two, one of which was Trafalgar’s final 2016 study. In Wisconsin, 33 polls were taken, and none found President Trump running ahead. Yet, in all three cases, he won the state.

The Great Lakes/Mid-Atlantic region was not the only area where 2016 polling missed the mark. In North Carolina, the margin average looked to be dead even heading into the election, but President Trump won with a 3.6 percent spread. The cumulative polling missed Arizona by two points, and Florida by 1.2 percent. In all of these instances, the Republican voted was under-estimated.

In the North Carolina race, Trafalgar missed on the high side, predicting a five-point win for President Trump. It was their poll, however, that brought the cumulative polling average closer to the even mark. In Florida, they also missed on the high side, coming in with a four-point projection for President Trump but correctly predicted his victory.

Polling miscalculations also were detected at least to a degree in 2018. Cumulatively, the research firms had a big miss in the Ohio governor’s race. The last three pollsters predicted Democrat Richard Cordray to be holding an average lead of 4.7 percentage points going into the election, but Republican Mike DeWine actually won with a 4.3 percent margin. No poll from Sept. 1 to Election Day forecast DeWine with a lead, yet he won the race rather comfortably.

The 2016 North Carolina Senate race was also missed. The Real Clear Politics national news site calculated Sen. Richard Burr (R) with lead projection of 2.0 percent, yet his actual winning margin reached 5.7 percent.

The 2017 GA-6 special election that featured current Georgia Democratic US Senate candidate Jon Ossoff, who raised an eye-popping $36 million for that contest, resulted in him losing despite the pollsters projecting him holding a one-point lead heading to Election Day. In that election, as Trafalgar correctly predicted, Republican Karen Handel won the race with a 3.8 percent victory spread.

Finally, the pollsters have missed on the last three major Florida statewide races. We have already covered their relative miss in the 2016 presidential campaign, though only falling 1.2 percentage points short of complete accuracy is certainly acceptable, but they generally failed to identify the winner. The 2018 Senate and gubernatorial outcomes were larger misses.

In the Senate race, a contest that then-Gov. Rick Scott (R) won by the smallest of margins (50.1 – 49.9 percent), the pollsters forecasted an average 2.3-point lead over election weekend for then-senator Bill Nelson (D). Then-congressman Ron DeSantis (R) was predicted to lose by an average of 4.7 percentage points, yet also managed to record a tight half-point victory.

Once more, Trafalgar was one of only two pollsters in the field within six days of the election to correctly call a Scott victory. In the governor’s race, Trafalgar was the only pollster among 15 others surveying within the last two weeks to accurately predict a DeSantis victory.

Though the overwhelming preponderance of data is predicting a clear victory for Biden in the 2020 presidential race, it is important to note that President Trump is actually running slightly ahead of his 2016 pace in the three Great Lakes states that he barely carried in 2016: Michigan, Pennsylvania, and Wisconsin.

Whether or not the actual votes fall in line for the president the way they did in 2016 remains to be seen, and while several underlying factors suggest they won’t, it is clear, however, that the ensuing Trafalgar Group data will be viewed in a more serious manner.

One comment

Leave a Reply

Your email address will not be published. Required fields are marked *