We are so busted!

 https://freakonomics.com/podcast/reasons-to-be-cheerful-rebroadcast/

Good but long podcast.  This caught my eye:

BAUMEISTER: …And incidentally, professors complain a whole lot. I remember visiting a university and I was having a conversation like this. I say, “This is a wonderful job,” and so on. And they looked at each other and said, “Well, we never say that out loud. You have to always be complaining. Otherwise, the administration won’t give us a raise. We always have to act like everything’s awful.” 

Economists vs. environmentalists

 Economics and environmentalism are belief systems that shape their adherent's way of thinking about the world.

                                                --Robert H. Nelson (link) 

When I earned my PhD, I started doing God's work (link to funny essay, "The Market as God") at the Justice Department, challenging anticompetitive mergers and putting price-fixers in prison.  My housemate was trying to do the same at the Environmental Protection Agency, using marginal analysis to design incentives to get polluters to face the consequences of their behavior.  If polluters produce up to the point where

    MR=MC

an output tax equal to the harm they cause, T=P, would bring pollution down to the point where the benefits of producing more are equal to its costs, including the costs of pollution.

    MR=MC+T.

There is legitimate debate about the magnitude of P but the principle seems obvious.  But not to [some/many?] environmentalists at the EPA.  They worship another God, and view all pollution as heresy.

In 2009, this debate made it up to the Supreme Court, where the economists prevailed, 6-3:

...the Supreme Court overturned [Sotomayor's earlier appellate decision that straightforward benefit-cost analysis was illegal] in a 6-3 ruling...

For the time being, benefit-cost analysis is OK.  

Training Artificial Intelligence to not discriminate

The top book review on Amazon reduced Prediction Machines to three propositions: 

• Artificial Intelligence (AI) is mostly about prediction 
• The cost and price of prediction is falling, 
• This will increase demand for complementary skills, like judgement and decision making.

As the cost of developing predictions has fallen, tremendous innovation has followed.  For example, AI can identify tumors with much greater accuracy than humans.  

However, AI can also identify race with "frightening" accuracy:

Whatever the algorithms were seeing, they saw it clearly. The software could still predict patient race with high accuracy when x-rays were degraded so that they were unreadable to even a trained eye, or blurred to remove fine detail.

To see how this could create problems, imagine training AI to diagnose and treat patients.  And imagine that the training set reflected the racial disparities sometimes associated with healthcare, i.e., minorities are under-treated or under-diagnosed relative to non-minorities.  With such a training set, the AI would "learn" to continue the disparate treatment.  

The director of medical imaging research at Royal Adelaide Hospital, calls AI's ability to recognize race “the worst superpower.”

House affordability

 

We have blogged extensively about the zoning restrictions that restrict the supply of new homes and raise the price.  

Historical Knockout Auctions

The Journal of Political Economy, one of the more prestigious economics journals, gets submissions of interesting economics related anecdotes for its back cover. The most recent issue contained this:

Collusive Bidding and Intermediary Profits in Congo a Hundred Years Ago

Five traders, who have hurried up in their cars, were waiting for the market to open. The region here has not been conceded; the market is free and the bidding began at once. We were surprised to see it stop almost immediately. But soon we understood that these five gentlemen were making a ring. The first carried off the whole crop for seven francs fifty a kilo, which probably seems a very fair price to the native, who only recently was selling his rubber at three francs; but at Kinshassa, where the traders resell it, it has fetched for some time past between thirty and forty francs, which leaves a very respectable margin. What about our gentlemen? As soon as the business is concluded with the native, they meet together privately in a little room, where another auction begins and they divide the spoil among them. The administrator is powerless against this secret auction, which, with every appearance of being illicit, does not, I am told, come within the power of the law.

[Andre Gide, Travels in the Congo (1927), translated by Dorothy Bussy (Hopewell, NJ: Ecco, 1994), p. 45. See also Daniel Graham and Robert Marshall, “Collusive Bidder Behavior at Single-Object Second-Price and English Auctions,” J.P.E., vol. 95, no. 6 (December 1987), 1217–39]

(Suggested by Laurent Lamy)

A Simple Way to Teach Regression

 A Simple Way to Teach Regression

Vanderbilt Owen Graduate School of Management Research Paper

15 Pages Posted: 10 Jan 2020 Last revised: 5 Aug 2021

Luke M. Froeb

Vanderbilt University - Owen Graduate School of Management

Date Written: August 05, 2021

Abstract

This paper introduces a simple free web app that can teach regression to anyone who can point and click. Originally designed to teach Justice Department attorneys enough about regression so that they could cross examine rival experts, the app ``inverts'' the usual pedagogy: instead of showing users how to run regressions on data, it asks them to click on a graph to ``create'' data to achieve a given outcome, like a statistically significant line. Successful completion of each task is rewarded with immediate feedback that reveals the principle behind the exercise. This paper describes short, intuitive exercises to teach: (i) hypothesis testing, statistical significance and confidence intervals, (ii) the difference between correlation and causality, and (iii) how to diagnose functional form mis-specification. These exercises can be completed in just a few minutes.

Keywords: Teaching Regression; statistical significance; correlation vs. causality

JEL Classification: A2 (Economic Education)

Suggested Citation:

Froeb, Luke M., A Simple App to Teach Regression (December 30, 2020). Vanderbilt Owen Graduate School of Management Research Paper, Available at SSRN: https://ssrn.com/abstract=3507142 or http://dx.doi.org/10.2139/ssrn.3507142

Why do 25% of people get so easily duped?

MarginalRevolution.com has the answer:  because they don't "look ahead and reason back," one of the maxims from our textbook.  In the simple game below, solved by backward induction (look ahead to the last move by the third player to see what the second and then the first player should do), 75% of fifth graders learned to look two steps ahead (as the first player) to correctly anticipate what the second and third players would do:

Player 3 is simply asked to match a shape. 

Player 2 earns the most by choosing the color chosen by Player 3. Of course, Player 2 doesn’t know what color Player 3 will choose and so has to reason about Player 3’s actions.

Player 1 earns the most by choosing the same letter as Player 2 but now must reason about Player 2 which involves reasoning about how Player 2 will reason about Player 3. 

Interestingly, women and those interested in STEM do better.  

What my daughter is learning in Rome

We have blogged about the labor market problems in Italy before:

This is why Southern Europe is a mess
Look ahead and reason back:  Italy
If you can measure absenteeism, you can control it

But I never realized how bad it was until I received this e mail from my daughter who is studying in Rome this summer:

Today, in my global practicum class we had a speaker who owns several McDonald's franchises in Rome and he enlightened us on the troubles of being an employer in Italy
 
First of all, he said, all employees who are hired, are hired for life.
They are allotted 6 weeks of paid vacation per year-- even working at Mcdonald's!!!
 
Each employer must pay a 100 percent tax of what he pays to the employee to the government.
 
Each employee has a 6 month paid sick leave per year that they can take with a valid doctors note. (which many tend to pull off with ease)
 
Employees are not allowed to be fired for poor behavior or work ethic, only if they steal or destroy company property.
If the employee is fired and sues the employer, the case is taken to civil court.  Most of these cases can take up to 3 years to be processed, 90% of the time the judge rules in favor of the employee and then the employer must pay 15 months salary for firing them, pay the salary they would have earned during the three years it took the case to be processed, and then must rehire the employee.

So basically no one wants to own or start a business in Italy, but everyone wants a freaking job.

Feel free to forward this to father, I am sure he would be interested, and I thought America was corrupt.

Her semester abroad seems like money well spent. 

Game theory: play the trust game online (Repeated free riding dilemma)

 Pretty good teaching implementation of a repeated prisoners' dilemma.  Very close to co-author Mike Shor's implementation of the same game.  

Professors:  Mike teaches one of the most innovative game theory classes, and has designed a good set of interactive apps.

Clorox is outbidding Vanderbilt for ad space

Following up on an earlier post showing Facebook, Instagram, Google, and Twitter show 20% fewer STEM ads to women than men:  as  Scientific America explains

Women are pricier to reach because they generally make more household purchasing decisions than men do. ...

...on Instagram it cost $1.74 to get a woman’s eyeballs on the ad but only 95 cents to get a man’s.

In other words, Clorox is outbidding Vanderbilt for ad space likely to be seen by women because Clorox places a higher value on the ad.  However, auctions are efficient, so both men and women end up seeing the highest-value ads.  

 OK, so there is no "disparate treatment" of men and women, but isn't it illegal to adopt practices that have a "disparate impact?" My understanding (I am NOT an attorney) is that Federal law prohibits both "disparate treatment" and "disparate impact" discrimination, even though adhering to one would violate the other. 

 For example, if you give a bidding advantage to STEM educational institutions when they bid for ad space likely to be seen by women, the algorithm is treating women differently than men.  This kind of disparate treatment is what the article recommends.

The same kind of tradeoff shows up in the current debate about equality (equal treatment) and equity (equal outcomes). If we want equity, we have to give up on equality.