The following are the primary and empirical reasons why I don't agree with the views of the president of a company who wrote to the Associated Press questions the utility of the survey methodology:

• Typically, well-reputed agencies such as the Associated Press, use a probability random sampling method for selecting participants for a nationwide survey. In a random probability sampling, all individuals in the population have equal chance of being included in the survey. Therefore, the data collected are free from bias and subjectivity.
• Survey data collected from a random probability sample are likely to fairly duplicate the population from which the sample was drawn. In other words, data collected from random samples are likely to follow a normal distribution and likely contain the true population parameters.
• Typically, a nationwide survey of over 1200 subjects will produce a small and empirically verifiable margin of error. So the survey, such as the aforementioned one from the Associated Press, will admit that there is some probability that the views and opinions of those surveyed do not echo the views and opinions of the wider population. However, as long as the margin of error is small we can asserts that the views and opinions of the sample represent the views and opinions of the wider population.

Caveat:

• While surveys that select individuals from random probability samples and produce small margins of error are a great utility in making assertions about the wider population, because of number of methodological risks and limitations, the samples are prone to inaccuracies and inconsistency on how they represent the general population. It is useful to externally validate the results from one survey with one or more similar surveys or with data from real-world situations before accepting the trustworthiness of the survey results.