Today's edition of "📈📉 papers you should know" is Dezsö Szalay's @reveconstudies.bsky.social paper, "The Economics of Clear Advice and Extreme Options". The paper concerns itself with information acquisition choices under delegated decision making. Let's go through an example... - ThreadSky

elliotlip.bsky.social • 100 days ago

Today's edition of "📈📉 papers you should know" is Dezsö Szalay's @reveconstudies.bsky.social paper, "The Economics of Clear Advice and Extreme Options". The paper concerns itself with information acquisition choices under delegated decision making. Let's go through an example...

Comments

maxkwiek.bsky.social•100 days ago

📌

elliotlip.bsky.social•100 days ago

A state is uniformly distributed on [-3,3]. First, the principal (P) chooses a subset D of [-3,3]. Then, the agent (A) decides whether to bear a private cost c to observe the state. Next, A (having seen the state or not) chooses an action from D.

elliotlip.bsky.social•100 days ago

Both P and A face a loss from the agent mismatching the state. Say this loss is equal to 6* the absolute error (so when the error is uniformly distributed on [0,x] the expected loss will be 3x). Both want A to guess the state correctly, but only A bears acquisition costs.

elliotlip.bsky.social•100 days ago

Since guessing the state correctly helps everyone, making D bigger is clearly better, if you held fixed the agent's information acquisition choices. Does this mean P should always offer full flexibility? Interestingly, no!

elliotlip.bsky.social•100 days ago

What happens if P chooses D=[-3,3]? If A acquires info, he makes a perfect decision, facing total loss c. If he doesn't acquire, he'll guess 0, and his error will be uniform on [0,3], for an average loss of 9. So if c>9, the agent will "shirk" and just guess 0.

elliotlip.bsky.social•100 days ago

Can P do better? Suppose P offers D=[-3,-1]U[1,3], that is, forbids actions in (-1,1). If A stays uninformed, he'll now guess 1 or -1, and so his error will be uniform on [0,4], for a loss of 12.

elliotlip.bsky.social•100 days ago

If A acquires info, he makes a perfect decision 2/3 of the time, and otherwise (choosing 1 or -1) has an error that's uniform on [0,1]. A's resulting total expected loss c+1. So if c<11, then A wants to acquire information. By making it harder to "wing it", P induces A to invest.

o.simardcasanova.net•100 days ago

📌

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply