You must set the boundary for your life. People are remarkably adaptive. If you don’t plan your life, others will.

Here is an example. I have the TYM project with a senior MG: the only thing he did was checking the grammars. Yet, because I willingly played the role of juniors, the normal game between seminar and junior workers, MG played his part to: they say they have great expectation of you. What they really mean is that  they will do almost nothing but want to slave you.

So I decided to reset the expectation by playing the game a bit differently. Here is the series of email exchanges in the last a few days.


Hey, L – just checking in. Paper was to go out in Nov, Dec, Jan… Almost March!

We going to get it under submission?


Dear M,

I have several pending revision going on. So I won’t be able to work on it for at least a month.

If you are in the hurry, here is the latest version that you can work on.




I haven’t heard from him since. Expectation reset?

A side note. In signaling games, actions must be credible in order to signal (often go with costs others are unwilling to bear). The purpose is to change/sharpen the belief of the rival. Maybe this is the underlying logic for my action.



It took me about a week. Finally I get the flow right.

We start the separating equilibrium part with the two necessary conditions. To characterize the equilibrium, we proceed in three steps. First, we specify the belief. This is an unusual treatment, but it is critical to simply the later exposition. Because R’s reactions— half of the story—depend on this belief.

The second step is to find the best retail and direct selling quantities. Here we use backward induction argument. We also lay out the optimization problem for each player.

The third step is to specify the equilibrium wholesale prices. The driving force here comes from the incentive compatibility constraints. Because type-L enjoys two advantages—the information advantage over R and the selling cost advantage over type-H—he has strong incentive to mislead. Therefore, for the equilibrium to be separating, type-L must prefer revealing than mimicking.

Got to finish the writing tomorrow.

2016-02-19 00.14.59

[Jasper, Canada, 10/2015]


[7:15pm-7:36pm, 21min]

Michael Spence is the first to formalize the signaling role of education.

The game plays out as follows. A firm wants to hire a worker for his ability, which can be high or low. Yet ability is the worker’s private information, unknown to the firm. For the same level of education, the more able the worker, the easier he can obtain that level.

The firm only knows that the proportion of the able workers is p. However, it believes that higher education attainment is more productive, because he can obtain the same level of education at lower effort/cost/discomfort.

To get high wage, the high ability worker may obtain more education to ‘signal’ his ability. The lower ability worker, however, would refrain from so doing, because the pain of get that high education outweighs the happiness the higher pay can brings to him. In response, the employer will pay high ability higher wage, and low ability lower wage. Hence, in the labor market, education serves as a signaling device to separate workers; it may not necessarily improve ability. In this sense, over education can be socially wasteful, akin to advertisement.

For education to work effectively, the proportion of high ability worker cannot be too high. Otherwise, the employer may shut out lower ability workers altogether, and pay high ability workers only.


[San Francisco, CA, 7/2009]