Quants will speciously compare the 2016 and 2020 election cycles and claim analytical insight. But 2016 and 2020 are qualitatively different elections -- i.e., they are incomparable (as are all elections).
This Election Cycle Shows the Folly of Quant Political Science
-
The fact that I can’t say with certainty that a particular individual will die next year but that I can predict within a small margin how many people will die is a real and perplexing paradox for some people.
Your reasoning through metaphor (not science), which demonstrates the bankruptcy of quant political science.
-
Quants will speciously compare the 2016 and 2020 election cycles and claim analytical insight. But 2016 and 2020 are qualitatively different elections -- i.e., they are incomparable (as are all elections).
I stand with GAG's view of social science as a way to make generalisable claims about politics. That's why he made a sweeping generalisation when he said that the state is a function only of energy politics. Every state ever, wars, etc. All a result of one and only one factor: energy politics. Clearly he thinks that states can be compared across space and time, just like elections can be compared across space and time.
-
I've never met someone with a poorer understanding of statistics.
That's the key problem. Quants may know stats, but they don't know science. No personal knowledgeable of science would be a quant.
Human behavior is incomparable to anything in the physical/biological sciences. This is why stats have no analytical value with regard to human behavior. This is what Stapel concluded -- and why he had to turn to fraud.
-
Eh, I’m an Americanist teaching in the methods sequence and I basically agree with OP.
Forecasting presidential elections is a prediction problem with huge domain shift. There are only a dozen data points or so to train on. And we can’t evaluate the model’s ability to forecast in the new domain (presidential elections in 2020) since there is only one.
Presidential election forecasting is the Myers Briggs of polisci.
-
Those same problems make qualitative research into elections similarly flawed.
The reason that quant is ignorance is because it specifically seeks to identify patterns across cases (e.g., elections). Ontologically, such patterns are analytically meaningless -- in most cases outright fraudulent.
-
The reason that quant is ignorance is because it specifically seeks to identify patterns across cases (e.g., elections). Ontologically, such patterns are analytically meaningless -- in most cases outright fraudulent.
If your version of political science doesn't seek to identify patterns across cases, then you are probably doing historical description rather than political science.
The issue with saying qual is superior in any real sense, is that you're basically suffering from the same fundamental threats to inference, but you're generally doing it with fewer cases and and far more explanatory variables (even if vaguely defined) than you have degrees of inferential freedom.
I use qualitative evidence in my research every day -I think most of us lean more or less toward mixed-methods at this point. Process tracing and case studies are helpful to illuminate potential mechanisms (or at least to provide examples of them), but it does little to tell you about the significance or prominence of those mechanisms, which are two bits of information that are critical in my research.
-
The issue with saying qual is superior in any real sense, is that you're basically suffering from the same fundamental threats to inference, but you're generally doing it with fewer cases and and far more explanatory variables (even if vaguely defined) than you have degrees of inferential freedom.
Again, stats are use.less in analytically understanding human behavior. Worse yet, it imposes ignor.ance by falsely positing the idea that human behavior is akin to the behavior matter and energy.