You're probably right in this specific case; this seems suspiciously one-sided. Do you have a link to the source where they explain their methods?

Generally something like this can happen though, especially if you do e.g. random dialing on the landline to survey people; mostly older people still use landlines and mostly retired people actually pick up during office hours. A good social scientist would obviously try to measure and control for those sampling errors though, not make them on purpose to get pre-determined results.