However, Internet polls are invariably biased - since only those with an interest in the outcome (or pranksters) tend to vote on them. In that case, it's probably not a reliable indicator.
Response bias is always taken into account, assuming the person reading the poll knows enough about statistics. Considering the people at beth all went to college, and its a required course for many degrees, I imagine they would know to take it into account.
It also only represents this population, but if you add up the means of each sample and then average those means you should get the mean of the population.
There is of course bias, sample size could be better, and it could be a simple random sample (thats hard for this poll).
however the sample size of 300 is large enough to minimize the effect of outliers (pranksters).
The real proportion could be 1/10, but that doesn't matter much. What does matter is that if they made it steam exlusive they would alienate a large portion of their consumer base. 10% is still a lot of consumers to lose. And even with bias, this isn't here to prove that 1/3rd of the PC population considering buying Skyrim will refuse to buy it, its here to prove that they would alienate a significant portion of their consumers, which is bad no matter what. Its just to help Bethesda to rethink their decisions.