ba-job: Bayesian Bad Job Simulator

A "discrete choice" model of labour turnover and matching

This doesn't work yet - I mean, this is version 0.01.

The labour market is a two-sided market for lemons: there are many crap jobs, many crap candidates, and highly expensive and inaccurate measurement of quality on each side.

Questions this simulation will answer, given heroic assumptions:

• What fraction of vacancies are bad?
• How many bad workers will an average organisation hire?
• How many bad jobs does an average worker take?

Bayesian approach: with an imperfect test (HR, Glassdoor) and huge sampling, how many false positives?

Limitations, to be relaxed later:

• Binary measure of quality: a job is good or job is bad, likewise candidates.
• Otherwise undifferentiated jobs and candidates. Equal pay, status. No licensing.
• Point estimate of turnover time
• De novo market: all jobs start unfilled
• Static size: No job destruction or creation.
• Accuracy of firms' vetting process is uniform and 0
• Accuracy of candidates' vetting process is uniform and 0.
• Disjoint quit/rehired period
• Good workers stay in good jobs.
• No experience variable
• No referral