Lipstick on a Pig
I have never been very into magic or illusions. Even as a child I did not really care. Magic is the original (or at least a much older version of) "fake news". Even if I don't understand how the trick was done, I know it is a trick. While I can appreciate -- I guess -- the skill in pulling off the trick, I don't care enough to really watch. And, I'm certainly never convinced that magic is real.
While magic is not real, how does it work? The most common "trick" employed is distraction. The magician needs to distract your eyes. Look here! No, not there! While your eyes are focused where the magician wants, the magician is free to manipulate things elsewhere without scrutiny.
I have been thinking about the process of magic a little bit lately ... when I have a free brain cell or two to devote to econometrics. How does magic connect to econometrics, you ask? Well, I'm sure there are plenty of potential answers, but the connection that has been floating in the back of my brain for a while now is due to data visualization.
Combine this gnawing in my brain with the guilt of having not posted a new blog in a while and the serendipitous timing of a Paul Goldsmith-Pinkham tweet and we have arrived at this point.
Because I am anything but an expert in data viz (but, I do know the lingo), let us turn to Wikipedia. It offers the following definition:
"Data visualization is an interdisciplinary field that deals with the graphic representation of data. From an academic point of view, this representation can be considered as a mapping between the original data (usually numerical) and graphic elements (for example, lines or points in a chart)."
In short, data viz encapsulates anything to do with providing a graphical representation of data. In applied econometric research, this can entail graphical representations of the raw data or graphical representations of the product(s) of statistical analysis. In the latter case, new quantities are obtained by filtering the raw data through some statistical model or procedure and these new quantities are then represented in graphical form.
In research, there is typically much to be learned from visually examining the raw data prior to undertaking any statistical analysis. However, this has become some what of a lost art. Instead, the focus on data viz these days -- by applied econometricians -- seems primarily geared at visual representation of the results obtained from statistical analysis.
While a picture is worth a 1000 words, I am concerned that researchers are devoting too much time and mental energy to data viz. And, at the end of the day, data viz is -- perhaps -- the modern day, research equivalent of (black) magic.
The Wikipedia entry on data viz goes on to state:
"[B]ecause both design skills and statistical and computing skills are required to visualize effectively, it is argued by some authors that it is both an Art and a Science."
Although I am loathe to give too much credence to Wikipedia, in this case it is spot on in that it emphasizes the role of statistical skills in data viz. My concern is that researchers, particularly researchers being in trained in graduate school today, are being nudged into putting too much focus on the visual part of data viz at the expense of focus on the statistical skill part of data viz.
My concern is relatively uninformed. I hope I am wrong. Time will tell.
Apparently, this is already playing out in the legal arena. We all saw a lawyer using the latest data viz techniques to distract a judge from what was really going on.
Magic.
If my concern is not unfounded, then empirical research will become of a war of competing magicians, each coming up with a new way, more flashy than the last, to captivate the audience and distract them from what it really going on. Ensuring that we do not look at what else the researcher is doing, namely the statistical analysis.
Will we become so enamored with the magic that we won't care if it is a trick, hiding shoddy statistical analysis from view? If we do, then science is in trouble. Imagine the bad policy decisions and the replication crisis that will emerge if publication decisions are made -- implicitly or explicitly -- based on beauty of the data viz and not the quality of the statistical analysis. Because, as the saying goes, you can put lipstick on a pig, but it is still a pig.
Don't get me wrong. If you know me, you know I appreciate all animals; well, not those crazy Australian spiders and deadly snakes. Thanks, Alfredo! But, as much as I actually do love pigs, I wouldn't want to rely on them for scientific advancement.
As I said, I hope I am wrong. I certainly don't consider myself an expert on the pulse of applied econometric research. But, I do think there is cause for concern. It is the same cause for concern that led me to start this blog; see here. I think that purveyors of statistical analysis do not spend sufficient time understanding the nuances of the analysis they are performing.
Before the data viz revolution, the nitty-gritty statistical details got short shrift even though the only other things competing for the attention of researchers were grappling with the scientific importance of a research question and understanding the institutional details of the specific topic at hand. Now, we have ggplot2 and a bunch of other things that I don't understand.
And while I don't understand them, I do understand the amount of time currently being devoted to learning and coding in order to produce the newest, flashiest graphical representation. I just hope it's not a trick meant to deceive the audience.
At the end of the day, I want researchers to be most excited and devote the most time to the statistical analysis.
