It’s easy enough to get more than 1,500 people to answer your survey. But every pollster is plagued by the fact that people who voluntarily answer surveys are usually not representative of the broader population on the relevant issue.
Basically any method for sampling the population introduces bias. The sample of people who pick the phone, who respond to texts or emails, who come to the door for a stranger, are all subpopulations with their own biases. A pollster’s ability to correct for these biases is what separates the good from the bad.
I’m not sure I agree. Windows still has 95% market share in Steam’s hardware report. And even the modest gains that Linux has seen in this space have been as much a result of the efforts of a few small companies who have financial interests in undermining Microsoft’s monopoly as it has been a display of grassroots enthusiasm for open ecosystems.
Basically any method for sampling the population introduces bias. The sample of people who pick the phone, who respond to texts or emails, who come to the door for a stranger, are all subpopulations with their own biases. A pollster’s ability to correct for these biases is what separates the good from the bad.