Miles Kimball wrote an oddly titled post claiming that Janet Yellen is not a "dove" because she knows that the U.S. economy needs some inflation. The actual post argues that the U.S. economy needs some unemployment. Yellen supposedly knows the economy needs some unemployment because her husband, George Akerlof, played a key role in developing "efficiency wage" models of unemployment.
Kimball explains the intuition behind the theory well enough. In a perfectly competitive labor market, with all workers identical and all firms identical, each worker earns the same wage which is equal to the marginal value product of labor. The market clears, and there is no surplus or shortage of labor. Any worker who quits one job can immediately find another job.
The "efficiency wage" argument is that a firm will be motivated to pay its workers more than this equilibrium wage. This will provide each worker an incentive to work for them rather than some other employer. Such workers, then, will be motivated to work harder for fear of being fired.
With all firms and workers identical, they all think the same way. The result is a wage above the competitive equilibrium, so that there is a surplus of labor. The fear of becoming one of the surplus workers motivates each worker to obey orders, work hard, and avoid being fired.
The dual labor market hypotheses is that there are two technologies for producing goods. One is the low supervision technology. Some firms use that technology and pay their workers a high wage. The other technology is the high supervision technology. The firms using that technology pay their workers a low wage. The low supervision technology is only possible because employees fear being fired for not working hard. If they are fired, then they might end up with a job using the high supervision technique and earn lower wages.
If all the firms are identical, however, this dual labor market can hardly be stable. All the firms will want to use the more profitable technology. This "problem" is solved by just having the unemployed workers do nothing. I don't really see that as a solution if the high supervision technology is still profitable. That is, the wage the unemployed workers will accept is less than the marginal value product of labor using the high supervision technology.
Anyway, the solution to this problem is pretty simple. The same firm can use both technologies. They use the high supervision/low wage technology for some workers and the low supervision/high wage technology for other workers. Workers who do a poor job under low supervision lose their high wages and are shifted to high supervision/low wage jobs. Even if all firms are identical, they can all use this approach.
While this "solution" seems disconnected from reality, where workers are much more likely to be fired than receive pay cuts, the actual labor market institutions that allow this to occur are common. Workers initially are paid a low wage. They begin working in a high supervision/low pay job. After a time, they are shifted to the low supervision/high wage job. If they fail to perform, then they are fired. They can get a new job, but it will be with a different firm, and they will start in a job with the high supervision/low wage technology at the other firm. This provides the motivation to work hard without excessive supervision.
Now, let's suppose that new employees are not all that productive and need extra supervision. They are paid lower wages. Then, after a time, they learn the ropes and are more productive for a variety of reasons, one of which is that they need less supervision. Suppose the employer increases pay after a probationary period. Workers seek to prove that they are good workers when their pay is low, so that they will pass the probationary test. Workers who have passed the test and earn higher wages are motivated to do a good job because if they are fired, they will have to start again, proving themselves through probationary work at some other firm. Workers are paid more or less what they are worth now, and the "problem" that would be solved by "efficiency wages" never exists.
Of course, in a world where the demand for labor is growing faster than the supply of labor, the typical worker earns higher wages. With all workers identical, they all get the same wage increase. But in the real world, not all workers are identical. And so, employers really need to provide different raises to different workers. This also tends to solve the "problem" that efficiency wages supposedly solves.
Bryan Caplan wrote that the survey data from Bewley suggests that employers do not pay workers more so that they can threaten to fire them if they shirk. Employers instead describe that as bad management.
In my view, unemployment of labor is a disequilibrium phenomenon. If all employers and employees were the same, and they always produced the same amounts of the same things, then I think it likely that the unemployment rate would be zero. That is, unless there is some market intervention, like a minimum wage or universal unionism, that pushed wages above the market clearing level.
That said, a monetary regime that tries to drive the unemployment rate down to zero in a real world dynamic economy, is likely to result in a hyperinflationary disaster. If the reason Yellen will avoid such errors is because of wrongheaded efficiency wage theories of unemployment--well, better to do the right thing for the wrong reason than to do the wrong thing.