In light of bad polling during the election, a look at other bad research methods that will foul a study.
Last week, quantitative minded people everywhere cheered as analysis trumped gut. Nate Silver’s predictions about the Electoral College proved precise and far more accurate than those who insisted the election “felt” like a tossup. But, with Silver’s prescience on display, it was hard to overlook the many other “researchers” who got it wrong (looking at you Gallup). Bad research is driven by bad methods—and one method in particular has been on my mind lately:
When performing social science research in the public sector, identifying a sample of people to survey and study is essential. This identification is the first step in a long series of assumptions you need to make about your research. If you have ever read Freakonomics or Gang Leader for a Day you are already aware of the concept known as snowball sampling, and you likely know why it is controversial.
- Snowball sampling is a research method in which you identify a group that you wish to study, and then ask members of the group to identify acquaintances to also join the study. Sudhir Venkatesh, author of Gang Leader for a Day, made this tactic famous when he studied the inner-workings of gang life in the south side of Chicago. Venkatesh immersed himself in the gang and started by talking to a few crack dealers and prostitutes. They in turn referred him to others in the gang, and the sample “snowballed” from there. From his study, he was able to make some astounding assertions regarding the economics of gang life. On one hand, that is the beauty of snowball sampling. It gives researchers the opportunity to identify subjects they may have been unable to identify. On the other, it limits a sample to a select group of people in one large social web. Therefore, in snowball sampling, the sample group is no longer random, and no longer represents the population at large. Without a random sample, a researcher is unable to extrapolate their research to the population at large, and the value of the research is diminished.
- Snowball research can also be adapted to other aspects of the research process, such as using it to find sources for a report. By reading one article and looking at the author’s sources, you can continue to find more and more sources, and thus a snowball of information is created. However, this method has its drawbacks; while you may be able to find a large amount of sources, it can lead to serious gaps in your argument.
- A snowball point-of-view, also known as a confirmation bias, is the tendency for individuals to test their hypothesis with positive examples as opposed to negative ones. In other words, people tend to use sources of information that prove their argument (thus confirming it), instead of finding sources that disprove it. This tendency often leads to a snowballed perspective, and could lead to a flawed conclusion. This problem is often found in political writing.
Take sequestration for example, some writers rely on partisan groups to prove their points. Depending on their ideology, an author may rely on the Center for American Progress’s definition of how sequestration will impact the federal government, versus how the Cato Institute may define it. Therefore, these authors are just using evidence to confirm their assumptions, as opposed to finding evidence to the contrary-and disproving it with their own logic. This doesn’t just lead a biased perspective, but to an argument with some serious holes.
Where have you been seeing bad research lately? Have you seen any examples of snowballing?