Theoretical sample and actual sample are two very different things. My wife was a marketing research analyst and she used to hate the inherent data collection problems with phone sampling. 9-5 M-F to lines where people actually answer after caller ID was 90% retirees. After business hours even fewer answered because of the telemarketing hours of 5-8. Weekends, fugedabowit. The data centers end up adjusting the calling parameters just to get a large enough sample and fill the responses by any means to meet a deadline.