View
217
Download
3
Category
Preview:
Citation preview
Compe&&on, Uncertain Mo&ves, and Strategic Communica&on
Jonathan Woon (University of Pi>sburgh) William Minozzi (Ohio State University)
Lobbyists influencing legisla&on
Interest groups guiding agency rules
Li&gants seeking court rulings
Candidates vying for office
Elites persuading the masses
How do experts cra. messages?
Do decision makers learn the truth?
Nature draws Target and ShiMs Receiver: Target T ∈ [-‐100, 100] LeM Sender: ShiM SL ∈ [-‐50, 0] Right Sender: ShiM SR ∈ [0, 50]
Senders observe T and privately observe own Si
Simultaneously send messages mL, mR ∈ [-‐150, 150]
Senders observe T and privately observe own Si
Simultaneously send messages mL, mR ∈ [-‐150, 150]
Receiver observes messages, chooses ac&on A ∈ [-‐150, 150]
Payoffs = 100 -‐ |A – (T + Si)|
How would you play this game?
1. Nature chooses Targets and ShiMs 2. Senders both observe Target, but privately observe ShiMs,
then send Messages 3. Receivers observe both Messages (but not Targets or ShiMs)
and choose an Ac&on Payoffs decreasing in distance from Ac&on to own ideal point (Target + ShiM)
Messages in babbling equilibria
Le. Sender Right Sender
Messages in par&&on equilibria
Le. Sender Right Sender
Messages in jamming equilibria
Le. Sender Right Sender
Equilibrium predic&ons
• Mul&ple equilibria: Babbling, Par&&on, Jamming
• Any equilibrium depends on all players having mutually correct expecta&ons of other players’ strategies
• Messages convey only limited informa&on – Senders must be careful not to reveal too much – Receivers must commit to ignoring some informa&on – Equilibria vary in the ways in which senders limit their informa&on
• Predic=on: Messages and ac=ons will only be weakly correlated with targets
Observed messages
-150
-100
-50
050
100
150
-100 -50 0 50 100 -100 -50 0 50 100
Left Sender Right SenderM
essa
ge
TargetGraphs by Type
Ac&ons and Messages
-150
-100
-50
050
100
Actio
n
-150 -100 -50 0 50 100Midpoint between Senders' Messages
Observed ac&ons
-100
-50
050
100
Action
-100 -50 0 50 100Target
Limited strategic sophis&ca&on
• Instead of equilibrium, suppose subjects a>empt to reason about how to play best responses
• Conjecture that Receivers split the difference
• Form beliefs or expecta&ons about the other sender’s message (through iterated reasoning/level-‐K or experience)
Limited strategic sophis&ca&on
T
mL = T -‐ E
T + SR
Limited strategic sophis&ca&on
T T + SR
mR = T + 2SR + E mL = T -‐ E
Limited strategic sophis&ca&on
T T + SR
mL = T -‐ E mR = T + 2SR + E
Limited strategic sophis&ca&on
T T + SR
mL = T -‐ E mR = T + 2SR + E
Summary
• Equilibrium theory predicts behavior poorly
• Senders engage in exaggera&on consistent with limited strategic sophis&ca&on (but they also underexaggerate)
• Offseing exaggera&on by compe&ng Senders allows the Receiver to learn more about the Target than in any equilibrium
Recommended