While such approaches can be done better or worse, the trend has been toward lower quality due to, as Cohn reports, machine responses (bots), bogus respondents, internet segmentation, and other factors. It's reached the point that most professional associations of pollsters and market researchers have launched efforts to combat the rising tide of bad data and invalid results.

Cohn reports that there have been, and still are, highly inexpensive convenience polls that can be relatively accurate. One major challenge, however, is a lack of transparency in the methods by which many of these polls are conducted. Without transparency, there is no way to judge the validity of the research, and therefore, no way to know whether we can trust the findings. And it seems the more quality declines, the less transparent organizations are about how they conducted their research. In the end, quality matters, and quality simply cannot be assessed without transparency.

And at the end of the day, it is the quality of the research that matters most. Research that is prone to bots and bogus respondents will always struggle to be valid and reliable. If anything, random sampling is even more important today than it was a half-decade ago, given how segmented society has become and how specialized our interests are. Now more than ever, we need tools that capture as much of the full diversity of Americans’ attitudes, opinions, and behavior as possible.