.1. Polls measure quantity of preference but not intensity of preference.

Whether what each of us wants turns into a vote may depend on whether our intensity of preference is high enough so we take time to vote and are willing to stand in the rain or heat or cold to vote. Also, if we’re a racial minority, whether our intensity of preference makes us willing to run the risk of violence against us for voting.

.2. Many who are polled may not reveal their true preference.

Many voters may be reluctant to admit favoring a candidate who seems to be selfish or immoral but who promises–more than the other candidate does—to support a policy the voter strongly favors. People conceal their own flaws, as when asked if they sometimes worry about drinking too much, and favoring a flawed candidate despite the flaws can seem like admitting their own shortcomings.

.3. The perceived preference of who is asking may affect the answers.

If a pollster identifies herself as being with a newspaper or an ethnic group known to favor one of the candidates, many who answer may not tell the truth that they favor the other
candidate.

.4. Answers may depend on when people are polled.

Right after a convention when speaker after speaker has praised a candidate, public approval of that candidate may be higher than before the convention or weeks afterwards.
Right after the opposition candidate has made a big speech, sentiment may be lower for the speaker’s opponent. Also we should remember that Mrs. Clinton lost by a hair a few days after the FBI chief said she was being investigated.