When skill cannot be shown directly, a diploma speaks instead.
Studying is easier for the able worker.
So only the able can afford the costly degree.
For the less able, copying is just too painful.
Firms read the diploma and guess the skill.
Yet the education itself builds no skill.
So the signaling cost below is pure waste.
For example, a fast π§ runner easily wins a race medal.
A slow runner can hardly win one at all.
So a medal tells everyone who is fast.
But winning the medal makes nobody faster.
Two types with productivity ΞΈHβ=2, ΞΈLβ=1, and education cost c(e,ΞΈ)=e/ΞΈ.
When firms believe eβ₯1 signals high (wage 2) and e<1 signals low (wage 1), verify with both types' incentive constraints that eβ=1 is the least-cost separating equilibrium.
UHβ(e)=w(e)β2eβ,ULβ(e)=w(e)βe
w(e)={21βeβ₯1e<1β
eβ=1 β the least-cost separating equilibrium, where the low type's incentive constraint binds exactly.Show proof
At eβ=1 the high type's payoff is UHβ(1)=2β1/2=1.5. Deviating to e=0 earns wage 1 for a payoff of 1, so with 1.5>1 there is no deviation β the high type's participation constraint (eββ€2) holds.
The low type's payoff is ULβ(1)=2β1=1.
Choosing e=0 also gives 1β0=1, so there is no gain from mimicking β the low type's incentive constraint (eββ₯1) holds with equality.
That is why this is the least-cost equilibrium: any lower e would bring the low type along.
The signaling cost borne by the high type is eβ/ΞΈHβ=0.5 β pure waste, burned only to earn the wage 2 while contributing nothing to productivity.
In the same model, explain through the low type's mimicking incentive why e=0.5 cannot be a separating-equilibrium education level.
ULβ(0.5)=1β0.5=0.5,ULβ(1)=2β1=1
e=0.5 β the low type would mimic e=1 for the wage 2, so types do not separate.Show proof
At e=0.5 the wage is 1, so the low type's payoff is ULβ(0.5)=1β0.5=0.5. But choosing e=1 earns the wage 2 for a payoff of ULβ(1)=2β1=1>0.5.
So the low type strictly gains from mimicking, and e=0.5 cannot separate the types. Separation requires the low type's incentive constraint 2βeββ€1, i.e. eββ₯ΞΈLβ(ΞΈHββΞΈLβ)=1.