A poll based on 160 respondents is useless. There are 350 million people in the country. Extrapolating the results from 160 to 350 million is stupid, and does not give you a reliable number.
The 8% margin of error assumes that the 160 were a very typical representative segment of the population, in no way different from a group composed of 160 people who are definitely going to vote, chosen completely randomly without regard to where they live, possession of a land line, willingness to talk to a pollster, without regard to time of day (many people work at night).
That doesn’t include the dead people in Chicago who will vote democrat.
Not entirely true. The quality of the samples count too. Now I’m not defending anything here, I’m just stating that there is nothing inherently wrong with a sample size of 160. Just that if you do under sample there are other things you have to do to normalize the results across a broad spectrum of the state’s demographic (national polls are virtually useless).