It’s Thursday, May 19th.
The Wonk is a twice-weekly policy newsletter that delivers quick hit headlines you may have missed, and long-form analysis on an issue that matters. If you enjoy the Wonk, consider forwarding to a friend!
Since we last spoke, Texas Governor Greg Abbott has signed an order which will prevent local government officials (except those involved in healthcare, prisons, and a few other institutions) from passing / enforcing mask-mandates. In practice, this means that in Texas public schools and various localities, masks will no-longer be required once the order fully takes effect on June 4th.
In a bit of lighter news, President Biden took a trip to Michigan on Tuesday to lay the foundation for a policy push centered around electric vehicles & American manufacturing. We also got this fun tid-bit of his driving.
Below, enjoy an old issue with some new reader feedback. Look out for a special edition Wonk on Sunday covering the Israeli Government - Hamas conflict.
High-Stakes, For Keeps
The gambler - casino relationship is always one I’ve struggled with.
My qualms are not moral, both parties have agreed to the rules and subsequent odds of the game. Rather, the part of my brain married to rationality asks how the gambler can go to the Blackjack table or the Roulette wheel knowing full well that over the long-run, they are signing up for a losing proposition.
Make no mistake about it, on any given night there will be winners at these games, but those “wins” are mere distractions from a one-way ticket to Loserville (not a profitable place to be).
However, in the casino there is one game that is different. In Slots, BlackJack, Roulette, Baccarat, Craps, etc. there is a very simple odds calculation in which the casino has a greater than 50.1% chance to win and the gambler has a less than 49.9% chance to win. Meaning, over enough “rounds” of the game, the casino is a sure victor and the gambler a sure loser.
Win-Win-Lose is the glorious and unique proposition which distinguishes poker from all other casino games. First, the win for the casino is guaranteed by taking a percentage of each round of betting for themselves (i.e., “the rake"). Then, the players are left to their own devices and a combination of luck (in the short-term) and skill (over the long-run) to decide the winners and losers.
That last piece of the equation, poker skill, is a tricky beast that can take months to grasp, years to learn, and a lifetime to master. However, there is one relatively basic poker concept which I can teach in 2 minutes that can meaningfully explain many of the most intractable problems in public policy.
So let’s do just that. Let’s go to “Poker-Policy School”.
Expected Value
This is the bread-and-butter formula for any aspiring World Series of Poker champion. Expected value allows a poker player to navigate one of the most central quirks of the game; players are trying to make optimal decisions with incomplete information. Two big pieces of missing data in poker?
Who knows what cards our opponent has?
Who knows what community cards (cards in the middle of the table) the dealer will place that could change how strong or weak our hand is?
And thus, the formula that resolves all of this uncertainty is:
Expected Value = (Probability of Outcome A * Value if Outcome A Happens) + (Probability of Outcome B * Value if Outcome B Happens) + (Probability of Outcome C * Value if Outcome C Happens), and so on until you have exhausted all of the possible outcomes.
Congratulations, you’ve graduated Poker-Policy School! You now how the foundational knowledge to be a world champion poker player (I can take bank wires but would prefer ether when you want to give me some of your winnings).
If a life of poker-faces and smoke-filled casinos doesn’t tickle your fancy, I also think you could be a fantastic legislator with this formula under your belt.
Now, let’s apply theory to practice. I’ll try to make two things clear using the following case studies:
Policy failures occur when decision makers either don’t fully consider the possible outcome set or
When they have an approximation of an outcome’s value and/or probability which drifts really far from the “true value”
COVID & Schools
This is a topic that I wrote about in a previous Wonk earlier this month. The decision of whether or not to close schools during the various stages of the pandemic was not an easy one. There were a number of serious outcomes at play, but especially for the youngest students, I think this is an example where the expected value calculation could be quite helpful.
Probability A: Certainly there was an evolving answer here as we learned more about the virus but generally, in-person schooling wasn’t associated with higher levels of community spread. With the youngest students in particular (i.e. elementary school) a significant degree of spread in schools or back to the community was/is infrequent.
Outcome A: However, the potential outcome here was really bad. Imagine a situation where tens of thousands or hundreds of thousands of school-aged children get infected with COVID. Early-on in the pandemic we didn’t know the precise death rate for younger age brackets but it’s possible that there could have been thousands of deaths and/or tens of thousands of cases of “long-COVID” (i.e., lingering symptoms).
Probability B: This is an easy one. There is a 100% chance if you close schools that students will be forced to resort to remote learning.
Outcome B: Again, this is an area where our understanding evolved during the pandemic. After a few weeks turned into a number of months of online instruction, it became increasingly clear that remote learning was causing students to fall behind and was placing unique stressors on families (especially those where parents were essential workers).
Doing our approximate expected value math, this is a case where we have to weigh a very low risk of a truly bad outcome (the health of children) against a 100% chance of a pretty bad outcome (lost education especially in formative years could have life-long consequences).
The two missteps here from some policymakers who weren’t thinking with an expected value framework:
Some failed to update prior beliefs as science indicated the probability of transmission in schools was low & that the fatality rate among young people was also exceedingly low
Perhaps some assumed that “Outcome B” didn’t have a significant negative value attached to it (i.e., remote learning would largely be the same as in-person instruction)
Old Bridges
Infrastructure may be the policy area that is talked about the most and acted upon the least.
Republicans and Democrats always point to infrastructure as a site for potential bipartisan action but inevitably balk when talks break down over what to fund or how much to spend.
This particular policy area also helps to demonstrate a second way in which the expected value formula can be tremendously useful in decision-making.
To put it in the simplest possible terms, the cost of replacing an old bridge is immediately noticeable. Taxpayers can “see” the money come out of their pockets (via the Federal or State government) to fund the project. What is much less visible is the probability that the same old bridge could collapse which could slow commerce, make it harder for people to move around an area or, in a worst-case scenario, cause injury and death.
And so, for too long, some policymakers have been willing to attend to the visible (i.e. avoiding the large price tag on bridges and a host of other infrastructure priorities) while avoiding or ignoring what we can’t easily see.
Through expected value thinking, one would be tasked with thinking through the exhaustive outcome set and factoring in the full-range of possibilities (regardless of their initial visibility).
Think Like a Poker Player
Of course, the two case studies are just the beginning of how expected value thinking in policy analysis could be tremendously valuable. If you have other examples that come to mind I’d love to hear about them (you can always reach me by replying to this email).
And yet, expected value thinking is of course not the silver bullet.
There are a number of valid criticisms to applying this statistical/poker concept to the complex world of public policy. One good argument is how should someone compare outcomes that happen in different units of measure (i.e., years of educational loss versus potential health impacts)? Furthermore, even if expected value thinking encourages us to think about the full set of possible outcomes, how do we know that we are being truly exhaustive?
Both of those criticisms are fair and deserve serious thought, but I don’t think they are central enough to deny the argument that if policy makers in Washington thought even a bit more like poker players in Vegas, we might all be better for it.
Reader Feedback
This is an occasional section of The Wonk which I’ll continue to use as an opportunity to continue the conversation.
In response to this issue, Nick H. wrote in with the following:
The use of the poker analogy to explain the expected value calculations was useful, and I especially enjoyed the application to the issue of in-person schooling. I think another great topical example would be the current operation of Enbridge's Line 5 pipeline in the straits.
And what an example Nick chose!
To lay the groundwork. Enbridge Energy’s “Line 5” is a 645 mile pipeline that transports 22.7 million gallons of light crude oil and natural gas everyday. The line starts in Northern Wisconsin, runs across Michigan (critically under the Straits of Mackinac), and eventually to its terminus in Ontario.
The key controversy is that pipeline opponents fear that Line 5 is a disaster waiting to happen given Line 5’s imperfect safety record, the relatively high volume of fuel transported, and the critical element that a small portion of the pipeline runs under the Great Lakes (and more of the pipeline runs very close to the lakes). Enbridge and proponents of the pipeline emphasize that a major leak has never occurred and that the pipeline is a critical part of the Michigan (and broader) energy supply chain.
Using our “poker-policy” framework, we can break down this issue into the following two areas:
Likelihood of a major spill * the environmental consequences of that spill
Likelihood of alternative energy transport methods (e.g., railcars, trucks, etc.) being used without a pipeline in service (100%) * the economic & environmental consequences of those alternative methods
Sure, it’s still imperfect (how do you weigh environmental damage against economic consequences for a private company and/or energy consumers?) but Nick has provided yet another example where the framework can take an unbelievably complex issue, and give us places to start a fact-first conversation about the true probabilities and magnitudes of various possible outcomes.
Agree with Nick? Have a different take or another example? Hit reply and let’s talk!
Wonk Wrap
What you could read
A five minute piece in the Wall Street Journal which details how & why Colonial Pipeline’s CEO choose to pay the ransom demanded by the hackers who forced the pipeline offline in early May. (hint, hint: a 4.4 million dollar ask probably isn’t too big of a pill to swallow for a company which did north of 400 million dollars of net income last year).
See you on Sunday for a special edition!
With Gratitude,
- Sam