Im familiar with sequential games in game theory. Specifically Im interested in if one can forecast using such a method?
what specific models I should know about if I can forecast using game theory?
$\begingroup$
$\endgroup$
5
-
1$\begingroup$ What exactly do you mean by "if one can forecast"? Specifically what would you like to forecast and how much do you care about accuracy? If these are not important one can also forecast using tea leaves... $\endgroup$– GiskardNov 14, 2017 at 18:29
-
$\begingroup$ There's a sub-field in behavioral economics called behavioral game theory that tries to predict how people behave in games. There are also research programs in what's called "predictive game theory" both within economics and outside economics. $\endgroup$– Herr K.Nov 14, 2017 at 18:34
-
$\begingroup$ @denesp I care about accuracy. specifically im interested in forecasting a workers strike duration. $\endgroup$– EconJohn ♦Nov 14, 2017 at 18:39
-
$\begingroup$ I think - but cannot source - that is widely believed the assumption of rationality does not hold. This alone would make game theory a shaky modeling tool. Unfortunately it is also quite hard to accurately gather data about the payoffs, and about people's beliefs on unknown factors. As such I think that game theory is not useful as a forecasting tool, unless you mean it in a very general sense. It calls attention to some interesting phenomena, but can only give broad answers. $\endgroup$– GiskardNov 14, 2017 at 19:00
-
1$\begingroup$ I also enjoy the following interview with Ariel Rubinstein, an eminent game theorist: Why Study Game Theory. $\endgroup$– GiskardNov 14, 2017 at 19:02
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
As @denesp has pointed out making real precise predictions with game theory is difficult because it is hard to gather data on payoffs and behavior under uncertainty, not to mention the fact that game theory typically assumes individual rationality (although the behavioral game theory referenced by @HerrK above loosens some of those requirements)
Of course that has never stopped people from trying. At least in environmental and resource economics there are plenty of papers looking into dynamic games that are calibrated to make some general (but perhaps not very accurate) predictions. This overview paper from Hannesson includes a lot of references to such papers in fisheries, a similar but older overview paper on transboundary polution was written by Finus.
Given what you're interested in, I'd look into repeated games with sequential moves. In a very simple set-up it could work out like this: each period the employers make an offer, and the unions then decide to either go on strike (hurting the employers and employees and moving to another round) or accept (which ends the game). As said, I don't think it will give you an accurate prediction, but it would perhaps give you some insights in the motives at work.
The problem with such a game set-up is that
- I'd expect both the employers and unions to fall for sunk cost arguments (can't quit now, we've gone so far...), which a rational player wouldn't do
- Repeated games in the presence of uncertainty often have multiple equilibria that can be supported by trigger strategies. Although these trigger strategies are subgame perfect, they are also non-renegotiation proof, and that is probably exactly what you're interested in.
Hope that helps.
