[Epistemologist]

Well, the sample size itself is indicative of a statistical error, because it is not adequately representative of the population, which is the U.S. It's too small for any hypothesis test. That mistake right there is enough to stop and say it's messed up. Also, these were interviews, and questioning bias is always present in a large amount; it's just a matter of how big it is. Also, the poll was conducted on a majority of women, and no attempts at stratification or anything other than SRS were mentioned; in fact, it's not even said to be random. There are so many other possible confounding factors that could be alluded to in the published study.

Also, the people who conducted the poll themselves lack credibility here. They're a private firm that has an obvious incentive to form that conclusion since they're against abstinence-only education. Their non-objective nature is also a seed for bias. These pseudo-statistics have become a mere tool to aid propaganda.