Bad information flows and its consequences

A psychologist
describes why people lie at work and the damage it does (WSJ, April 2020).

Maybe Screening is Overrated

This firm hires the first person who can answer "yes" to three simple questions (Fast Company, Feb. 2020).

Incentives for Public School Teachers in the D.C.

"screening mitigated both the adverse selection and the moral hazard problems and not by small amounts" (https://managerialecon.blogspot.com/2013/10/michelle-rhees-tenure-as-washington-dcs.html).

Some of the comments fear that cheating by teachers may account for some of the increase in measured performance.

Using AI to Screen Job Applications

The WSJ (Dec. 2019) describes how employers are using AI to screen applicants and offers advice. Some money quotes follow.

• "Spice up your résumé with specific on-the-job results, use meaningful job titles and tailor your choice of words to match companies’ requirements.
• "Rock Brouwer has hired many candidates ZipRecruiter has brought to his attention. 'When I get one of those, it just makes my day,' says Mr. Brouwer
• "About 60% of employers admit such tools cause them to miss some qualified candidates,
• "Most vendors refuse to tell employers how their algorithms work. And most employers lack deep, accurate performance data.
  The systems risk magnifying managers’ prejudices if those biases are reflected in the makeup of the employer’s current workforce
  ... High performers may share traits that have nothing to do with job performance, skewing outcomes." 
• "Even if employers and vendors aren’t trying to reject female or minority applicants, they still risk doing so if they train algorithms on data gleaned from a current workforce that lacks diversity." 

Hart and Holmstrom on aligning interests (and employment decisions)

1. This (re)post contains an excellent summary of what two Nobel laureates say about the best way to align interests of owners and employees. It contains a link to a more detailed summary in MRUniversity that is also excellent. The Nobel laureates are Oliver Hart and Bengt Holmstrom. Here are the basic elements.
2. The premise is that supervisors want to hire employees who work hard and reward them for doing so.
3. Output of an employee depends on how hard the employee works and luck. For example, a salesperson may have a great year when she works hard or when she is lazy and lucky. She can have a bad year even when she works hard because she is unlucky. 
4. Supervisors can observe signals of how hard the employee works. A signal contains information and noise. For example, output may be a signal. When output is a signal the supervisor knows that, on average, salespeople with high output work hard. The supervisor also knows that an individual salesperson may be lazy and have high output because she is lucky.  
5. The best compensation scheme uses all of the signals available to the supervisor to determine the reward to a worker.
6. The best compensation scheme places more weight on the signals that have the least noise. As noise decreases, the signal becomes more reliable. For example, suppose that output depends only on how hard someone works and that luck plays no role. In this case, the supervisor should measure output and use it and it alone to determine the reward. 
7. The best compensation scheme compensates risk averse employees with a higher base salary.
8. The best compensation scheme uses relative performance metrics ("tournaments", rankings) when employees have similar abilities.
9. The best compensation scheme uses absolute performance metrics when employees do not have similar abilities.

The post critiques compensation schemes that reward CEOs when the firm's stock does well because much variation is noise. Changes in the stock market affect the price of all stocks. Therefore, rewarding the CEO for appreciation often is a reward for being lucky, that is, being the CEO during a bull market. 

A better signal is the difference in return on the stock market between the firm and its competitors. This difference is more closely tied to what the CEO does and less affected by noise created by bulls and bears. In other words, a tournament may be best for CEOs.

Here is a key takeaway. When designing a compensation scheme, think about what you want to reward and what you can measure that is a signal. Identify the strongest, least noisy signal and put more weight on it in the compensation scheme. Daryl Morey spend years refining how to measure expected productivity of basketball players. He discovered that points per minute is a better signal than points per game and that points per possession is even better. Even then, noise beset him. He passed on drafting one player because a photo of the player without a shirt revealed man boobs and another player whose statistics were low because the player hated his college coach. 

Here is my final thought. Much of the analysis applies to measuring qualities of employees to hire. Think about the qualities you seek in the applicant and what you can measure that is a signal. Identify the strongest, least noisy signal and put more weight on it in the selection process.

Can you trust what a job applicant reports?

A recent study reports that a substantial number of CVs embellish the accomplishments of the candidate. Verifying information is costly and many employers probably accept what is written without question.

FYI: CV = Curriculum Vitae, sort of a résumé for people seeking teaching or research jobs.

Why do people who graduate from college earn more than people who do not?

Real Clear Books interviews an economist who thinks that college serves as a signal but does little to increase productivity