Signaling vs Credible Threat
Signaling and Credible Threat are two Game Theory & Information concepts in AP Economics that students often mix up. Signaling is when an informed party credibly reveals private information to a less-informed party to overcome asymmetric information. A credible threat is one the threatening player would actually want to carry out when the time comes, which is why a rival believes it. Here is how they compare side by side.
A classic example is education as a signal of ability to employers, or a warranty signaling product quality. Effective signals are costly enough that low-quality types won't fake them.
In a sequential game, work out what the threatening player would do at the node where the threat is supposed to be carried out. If following through pays less than backing down, the rival can ignore the announcement, and the threat is empty no matter how loudly it is made. Threats become credible when a player changes their own future payoffs in advance, by sinking money into extra capacity, signing a contract with penalties, or handing the decision to a rule that removes their discretion. That is the counterintuitive part: reducing your own options can strengthen your position, because the rival now knows what you will do. A promise works the same way and has to be self-enforcing to be believed.
Signaling vs Credible Threat: Two Different Things Being Made Believable
| Signaling | Credible Threat | |
|---|---|---|
| What is being made believable | A claim about the sender's hidden type | A statement about what a player will do later |
| Why anyone believes it | The action costs the wrong type more than it could ever gain | Carrying it out is the player's own best move once that moment arrives |
| Information the idea needs | Only bites when one side knows something the other does not | Works even when both sides know everything |
| How you test it | Compare the action's cost with the reward, type by type | Work backwards from the end of the game tree |
| Role of cost | A high cost is what makes it work | A high cost is what breaks it |
| Failure mode | Pooling, where every type sends the same message | An empty threat the rival simply ignores |
| How you repair a failure | Make the action dearer for the type you want to exclude | Commit in advance so the threatened move becomes your best reply |
The cost of the action does opposite jobs in the two ideas
Both concepts answer the same question, which is why anyone should believe you, and both answer it by looking at payoffs rather than at words. The cost of the action then pulls in opposite directions. A signal is believable because it is expensive, and more precisely because it is more expensive for the type you want to be told apart from. If a credential costs a capable worker 15 and a less capable one 45, while the wage paid to credential holders is 30 higher, only the capable worker buys it. Make that credential free and it separates nobody, since every type sends it. A threat is believable for the reverse reason. If carrying out the threatened action would be expensive for the threatener, the rival works out that it will not happen and ignores the announcement. Talk is free to make and free to abandon, which is exactly why declaring a price war persuades nobody on its own. So the repairs run in opposite directions too. To rescue a failed signal, raise its cost to imitators. To rescue a failed threat, lower the cost of following through, so that the punishment becomes the move you would want anyway.
Working backwards kills the price-war threat, and a sunk investment revives it
Take an incumbent earning 100 alone and an entrant deciding whether to come in. Staying out leaves the entrant with nothing. If the entrant comes in and the incumbent shares the market, the entrant earns 20 and the incumbent 50. If the incumbent instead starts a price war, the entrant loses 10 and the incumbent drops to 30. The incumbent announces that any entry will be met with war. Test that by looking only at the instant the choice actually arrives: entry has already happened, so the incumbent compares 50 for accommodating against 30 for fighting, and it accommodates. The threat is empty, the entrant enters, and the announcement changed nothing. Now let the incumbent sink 25 into capacity that is worthless unless a war is fought. Gross payoffs after entry become 60 for fighting and 55 for accommodating, so net of the sunk 25 the incumbent gets 35 for fighting against 30 for accommodating. Fighting is now its best reply, the entrant anticipates its loss of 10, and it stays away. The incumbent ends with 100 minus 25, or 75, against the 50 it collected when its threat was hollow. Nothing was said differently. The payoff at the decision node changed, and that is the only thing credibility ever rests on.
Frequently asked questions
What makes a threat credible in game theory?
Credibility comes from the payoffs at the moment the threat would have to be carried out, not from how firmly it was stated. Work backwards through the tree: if the threatened move is the threatener's best response once that node is reached, a rational rival believes it, and if some other move pays more there, the threat is empty. Commitments such as sunk investments, binding contracts and public penalties all work by changing that one comparison. More at /glossary/backward-induction.
Can the same action be both a signal and a threat?
Sometimes, and the overlap is where students lose marks. A firm that sinks money into spare capacity commits itself, which makes a price war credible, and also reveals something about its costs, which is a signal. The distinction still holds: the commitment matters because it changes what the firm will want to do later, while the signal matters because it changes what rivals believe the firm already is. Decide which of those two the question is testing before you answer.
Why does a costly signal work when talk does not?
Cheap words can be copied by anyone, so they separate nobody and every type says the same thing. A signal separates only when its cost falls unevenly across types, so the reward is worth paying for if you are the high type and not worth paying for otherwise. The cost does not have to be productive, and in the pure model it produces nothing at all. Uneven cost is the entire mechanism. See /glossary/screening for the version where the uninformed side moves first.
Related comparisons
Get AP Econ exam tips in your inbox
Occasional emails with study tips, new interactive graphs, and exam-season reminders. Free, no spam.
No spam. Unsubscribe anytime. Read our privacy policy.
Keep track of what you have studied
A free EconLearn account adds progress tracking, your quiz history, and achievements. Studying here is free either way, and there is nothing to pay for as a student.
Create a free accountAlready have one? Sign in
Last updated