For policy analysis and research studies, such as our Consumer Pulse Survey, ACI always uses a telephone survey conducted by a national survey research firm. These surveys are usually designed and tested (or weighted) to be representative of the nation and never have less than 800 responses.

However, the figures used in this blog were from an online survey. First, online surveys like these are NEVER representative of the nation for two reasons — one, the respondents were self-selected; and two, there is always a bias favoring respondents that have access to the Internet. Instead, what this survey depicts is a representation of the ACI’s readership.

While the piece points out that this was “not a large sample,” in this particular case, the small sample size is perfectly adequate. Let me explain why. For this small sample, we normally would have expected a large confidence interval (exceeding plus or minus 10%) if the results had been spread evenly among the survey choices. However, because no respondent chose “wireline telephone,” the actual confidence interval (when taken from statistical tables) is close to plus or minus 0%.