The “Gambler’s Fallacy” at Work in Civil Service Law
Pop Quiz: You have flipped a coin five times and it has come up five heads five times in a row. If you were to bet on the next flip, what would you pick: heads or tails?
A lot of people would pick tails, believing consciously or subconsciously that the fact that the coin has previously landed on heads several times somehow is evidence that the next flip is more likely to be tails. And when doing so, they would become victims of “The Gambler’s Fallacy.” You see, there’s a tendency in humans to see patterns, to look for balance in an otherwise unbalanced world. The next time you’re in a big casino in Las Vegas [Note to self: a great location for an FELTG seminar!], look closely at the video screen next to the roulette wheel. When you do, you’ll see that a lot of casinos will track how many times red or black has come up in a row and display that number for all the future gamblers who might not otherwise bet on the spin of the wheel, but who start pulling out the Benjamins when they see that noir or rouge has shown itself five or six or ten times in a row.
And that’s why big casinos have a lot of your money; those guys paid attention during the first day of statistics while you were checking your Facebook page for the most recent updates by Justin Bieber. Statistically speaking, the act of flipping a coin at this very moment results in a 50/50 chance of either side coming up, regardless of which sides have come up previously. The past does not control the future of a random event such as coin tossing or roulette wheel spinning.
But not so for the decisions of judges. In one recent study, investigators reviewed over 150,000 decisions by immigration judges in cases in which the individual appearing before the judge was seeking asylum. The results show the subconscious application of the Gambler’s Fallacy to those decisions by those judges. If the immediately previous case resulted in the judge granting asylum, the next case was about one percent less likely to result in an asylum grant. If two cases in a row resulted in a judge granting asylum, the judge was five percent less likely to grant asylum in the third case. Daniel Chen, Tobias J. Moskowitz, and Kelly Shue, “Decision-Making under the Gambler’s Fallacy: Evidence from Asylum Judges, Loan Officers, and Baseball Umpires,” Fama-Miller working paper, March 2015.
You see, judges have memories whereas coins do not. Judges know that they are hired to make decisions, and they subconsciously apply the Gambler’s Fallacy to their decision-making, working to bring order to an otherwise disordered world. If some applicants deserve asylum, then there must be others who do not, or we wouldn’t need judges to make the distinction.
I first ran into this phenomenon way back in 1992. The Board had been issuing decisions for over ten years by that time, ironing out word by word what the “new” Civil Service Reform Act was directing agencies to do when firing bad employees. Upon reviewing MSPB’s annual reports for each of those preceding years, I noted that the Board was upholding removals at the rate of about 78% each year. In an article I wrote (for a lesser newsletter) that year, I opined that this trend didn’t make sense. Yes, during the early ’80s when the CSRA was initially being interpreted, you’d expect agencies to be making mistakes. There were a lot of new terms and procedures to be understood, and it’s only through trial and error that terms and procedures come to have a definitive definition in a business based on adjudicatory oversight such as our civil service. My surprise was that the appellate loss rate for agency removals did not drop each year during the decade as we all came to learn what the CSRA was all about.
And my surprise has not diminished in the quarter-century since that original epiphany. In this past year, and in fact every year since that time, the agency success rate on appeal hovers right around 78%. Why has it not improved since that time? Why are we no better at removing bad employees today that we were back in the early ’80s when the Board’s decisions could be read from first to last over a weekend? And if you’re a real purest, why does an agency EVER lose a case on appeal? Have any of you readers ever been involved in a case in which the agency fired someone with full knowledge that it had somehow removed the employee incorrectly?
Could it be … Gambler’s Fallacy? All labor relations specialists know that it is black letter law in the arbitration community that an arbitrator will not stay in the arbitration business very long if she always holds for management regardless of the merits of the cases brought to her. Is it possible that the Board and its judges are subconsciously prone to sometimes set aside agency removals because if it ALWAYS upheld the agency, it would be seen as not doing its job? It only makes sense that if some agency removals should be upheld, others should not or otherwise we wouldn’t need judges.
And it only makes sense that if a coin has come up heads five times in a row, the next flip is more likely to result in a tail. Of course, that’s undeniably wrong, but it still makes sense. As does a flat Earth, ghosts, and lucky charms.
There is a treatment for Gambler’s Fallacy among judges that has been shown to reduce the effects of its bias:
- Publicly acknowledge that the bias exists, thereby encouraging the judge not to succumb to it.
- Assign judges different types of cases in a dispersed order, thereby reducing the carry-over effect from one removal case to the next.
- Periodically review decisions to evaluate if the Gambler’s Fallacy is having an effect on decision-making and provide feedback to the judges as to how they are doing relative to the bias.
We are starting to learn that a LOT of our biases are subconscious, and that we have them even though in our hearts we believe we do not (yeah, I’m looking at you, Presidential Candidates). MSPB and EEOC would do the country a service by assessing their own judges for the tendency to give in to the Fallacy bias and to take steps to reduce it. Wiley@FELTG.com