I am honored to have a version of this essay appearing in Bill Frank’s 97 Things about ethics everyone in data science should know. Pick up a copy now.
I have taught ethics in Applied Econometrics and Data Analysis for at least the last 20 years. But I rarely have used the word ethics, resorting to phrases such as data skepticism, and other attitudes that suggest acting ethically.
Nothing in the past 20 years has had as much impact on me and my classroom teaching and my ethics of data analysis as Peter Kennedy’s “Sinning in the Basement: What are the Rules? The Ten Commandments of Applied Econometrics.” This essay also appears in his Guide to Econometrics (Kennedy, 2008).
From the moment I read this paper, I was completely transformed and forever a disciple of his. I was fortunate to host him on my campus where he spoke of the misuse of econometrics and failure of research to make it past his editor’s desk at the Journal of Economic Education. One example, a paper is rejected because they did not acknowledge a problem in their analysis, ignored it, and probably hoped the editor would not notice. Being honest and transparent enough to acknowledge a problem of which the authors were aware, but unable to solve is sometimes enough Peter would point out. Hiding one transgression suggests other ethical abuses of data.
I used the word ethical, but Peter did not, preferring the oft used word sin and sinning. But the point is made. When I got my Ph.D. at The Ohio State University in 1980, I had taken nine separate statistics and econometrics courses over the 5 years that I was there. I learned classical estimation and inference from some of the best professors, and yes we had to “sin in the basement” (by pilgrimage to the mainframe computer in those days) because there was scarcely a day’s instruction of how to use the computer much less conduct what Peter would call the moral obligation of applied econometrics 20 years later.
Kennedy says “… my opinion is that regardless of teachability, we have a moral obligation to inform students of these rules, and , through suitable assignments, socialize them to incorporate them into the standard operating procedures they follow when doing empirical work.… (I) believe that these rules are far more important than instructors believe and that students at all levels do not accord them the respect they deserve.” (Kennedy, 2002, pp. 571-2)” I could not agree more and have tried to follow faithfully these rules and to teach my students and colleagues to do likewise.
Failing to follow these rules brings about ethical implications if not direct unethical behavior. To knowingly violate the rules is to be, or at least risk being, unethical in your handling of data. To unknowingly violate the rules would nevertheless lead to unintended consequences of poor outcomes that could be avoided.
The Rules of Applied Econometrics | |
Rule 1: | Use common sense and economic reasoning |
Rule 2: | Avoid Type III errors |
Rule 3: | Know the context |
Rule 4: | Inspect the data |
Rule 5: | Keep it sensibly simple |
Rule 6: | Use the interocular trauma test |
Rule 7: | Understand the costs and benefits of data mining. |
Rule 8: | Be prepared to compromise |
Rule 9: | Do not confuse statistical significance with meaningful magnitude |
Rule 10: | Report a sensitive analysis |
Failing to well and fully articulate the problem (rule 1) is so critical that to not spend time on the problem and the common sense and economic theoretical solution can lead to serious flaws in the study from the very first step. It might lead to a violation of rule 2 where the right answer to the wrong question is discovered? What if you fail to inspect the data (Rule 4), fail to clean the data and provide for necessary transforms, fail to control for selection bias, then you will have results based on assumptions that are not realistic and produce wrong results that are unduly influenced by the dirty data. The importance of this cannot be over emphasized. Recall that Griliches exclaims that if it weren’t for dirty data, economists wouldn’t have jobs. What if you violate Rule 7 and, knowingly or not, allow the data to lie to you. In the words of Nobel Prize winning economist, Ronald Coase (1995), “If you torture the data long enough, it will confess.” A violation of Rule 9 might lead you to worship R2 or participate in p–hacking. It might cause you to ignore a huge economic implication (large magnitude) only because it has a large p-value. Violations of Rule 10 may be the largest of all. Suppose in the spirit of discussing “sinning” you believe your model is from God (as suggested by Susan Athey) then why would you ever look at alternative specifications or otherwise validate the robustness of your findings. What you find is what you find and alternatives be dammed.
“Many data scientists make bad decisions – with ethical implications – not because they are intentionally trying to do harm, but because they do not have an understanding of how the algorithms they are taking responsibility for actually work. (Jennifer Priestly, August 2019, LinkedIn post).” Likewise many in the field who ignore the Rules for Applied Econometrics risk doing real harm, not out of intentionality, but out of ignorance or neglect. This later lack of motive is just as real and likely more widespread than the intentional harm, but harm occurs in any event.
The American Economic Association adopted ethical, code of conduct guidelines that states in part: “The AEA’s founding purpose of ‘the encouragement of economic research’ requires intellectual and professional integrity. Integrity demands honesty, care, and transparency in conducting and presenting research; disinterested assessment of ideas; acknowledgement of limits of expertise; and disclosure of real and perceived conflicts of interest.”
The AEA statement does not go directly to data ethics, but is suggestive since little economic research and no applied economic research can be conducted without data. The AEA statement is a beginning, but I suggest that those who do applied economic research would do well to hold to the rules for sinning in the basement. This is so important now since going to the basement is no longer the norm and many more analysts should be trying to avoid sinning wherever and whenever they have their hands on their laptop.
References:
American Economic Association (2018) Code of Professional Conduct, Adopted April 20, 2018, accessed at https://www.aeaweb.org/about-aea/code-of-conduct on October 7, 2019.
Coase, Ronald. (1995) Essays on Economics and Economists, University of Chicago Press,
Kennedy, Peter E. (2002) “Sinning in the Basement: What are the rules? The Ten Commandments of Applied Econometrics.” Applied Econometrics, Blackwell Publishers Ltd. Accessed at https://onlinelibrary.wiley.com/doi/abs/10.1111/1467-6419.00179.
Kennedy, Peter E. (2008) A Guide to Econometrics, 6th edition, Cambridge, MIT Press.
Gola, joanna. 10 Commandments of Applied Econometrics (series), Bright Data, SAS Blog, first in the series accessed here: https://blogs.sas.com/content/brightdata/2017/03/01/10-commandments-of-applied-econometrics-or-how-not-to-sin-when-working-with-real-data/
Priestley, Jennifer L. (2018) “The Good, The Bad, and the Creepy: Why Data Scientists Need to Understand Ethics.” SAS Global Forum, Dallas. April 28-May 1, 2019.
Further Reading:
Zvi Griliches (1985) “Data and Econometricians–The Uneasy Alliance.” The American Economic Review, Vol. 75, No. 2, Papers and Proceedings of the Ninety Seventh Annual Meeting of the American Economic Association, pp. 196-200 accessed at http://www.jstor.org/stable/1805595 on October 7, 2019.
DeMartino, George and Deirdre N. McCloskey, eds. (2016). The Oxford Handbook of Professional Economic Ethics, Oxford University Press, New York.
Author contact:
Steven C. Myers
Associate Professor of Economics
College of Business Administration
The University of Akron
myers@uakron.edu
https://econdatascience.com
https://www.linkedin.com/in/stevencmyers/