Polls are usually right, but when they miss, wow!

The recent surprising Conservative victory in Great Britain was a reminder that political polls are fallible and only an estimate of what voters will do.

Often, however, they are wrong. Sometimes, like this one, they can make the public believe one thing only to find out on Election Day the polls were off by a wide margin.


The first time I heard about polls that were drastically wrong was in a political science class in 1965. Back then the teacher was talking about the famous 1948 polls that led the Chicago Daily Tribune to print an election-night edition with the headline: "Dewey Defeats Truman."

The next day it turned out vice-president Harry Truman had won the election by more than 3.5 points.

The three main polling companies at the time — Gallup, Roper and Crossley — all predicted Thomas Dewey, the Republican candidate, would defeat Truman by a significant margin. Just the opposite happened.

In his next column, Elmo Roper wrote: "We were wrong. We couldn't have been more wrong. We're going to find out why."

They were wrong for two reasons. They had stopped polling a week before the election, and each of the data collectors was assigned a quota of voters to call. They didn't do random-sample polling and continue the research until the night before the election.

Polls are, more often than not, accurate. Still, when they are wrong as they were in Great Britain, it can be a doozy.

Back in 1990 when I was running a media news organization, we did a poll for the presidential elections in Nicaragua. Our poll showed Sandinista leader Daniel Ortega would defeat Violeta Chamorro, the widow of the publisher of the country's most important newspaper, with ease.

We had been running polls on Central American elections for several years and had won national recognition for their accuracy.

This time, however, we were wrong — and boy, were we wrong. Chamorro defeated Ortega handily. I remember the president of the company calling me and demanding that we offer an explanation to the public.

We did in our newscast that next day.

The explanation was simple. Nicaraguans were afraid of Ortega, who had been the leader of the Sandinista movement, which had recently defeated the Somoza dictatorship. We couched it in prettier language saying that once again the Mayan Indians in the region had worn their masks to hide their true feelings.

Ortega, a Marxist, ran a tightly controlled police state and had elections only because everyone assured him of an overwhelming victory.

But everyone was wrong.

Back then, as pollsters had done back in 1948, they learned from their mistakes. But give them time and they will make new ones.


In South Florida, polls continuously tell us that Cuban-Americans are changing. That the new arrivals from Cuba want closer ties to the island, that the older generation of exiles is slowly dying and their sons, daughters and grandchildren do not care as much about the Castro regime. That they approve of President Barack Obama's policy of re-establishing relations with the communist government in Cuba.

In general, the polls are right. My generation is disappearing. The younger Cuban-Americans care less about Cuba than we do. But these polls are also misleading.

Most polls ask all Cubans in the country, and not just those who are American citizens, their feelings on the issue of re-establishing diplomatic relations between Cuba and the United States.

Thus, even though the different polls say that Cuban-Americans favor Obama's policy, local Cuban-American Republican representatives in Congress on Election Day are all favored to be re-elected.

The lesson is simple. Polls depend on the information you ask. If one puts trash in, the likely outcome is also going to be trash.

As we approach 2016 and the presidential election, it is worthwhile to keep these ideas in mind. Read the polls for what they are: a snapshot of the election at that instant. And remember the result of the polls can be affected by what questions are asked and how they are asked.

Guillermo I. Martinez lives in South Florida. His email address is