With mechanism design
- Revelation principle: any mechanism that you can think of, can be described as “ask worker for his type \(n\) and then tell the worker to earn \(y^n\) and consume \(c^n\)”
- We will create two options \((y^l,c^l),(y^h,c^h)\) such that each type chooses the option that is meant for this type
- This we ensure by imposing incentive compatibility constraints:
\begin{align}
\label{eq:3} \tag{$IC_l$}
u(c^l)-y^l/w^l &\geq u(c^h)-y^h/w^l \\
\label{eq:4} \tag{$IC_h$}
u(c^h)-y^h/w^h &\geq u(c^l)-y^l/w^h
\end{align}
- Adding these two equations yields:
\begin{equation*}
y^h \left(\frac{1}{w^l} - \frac{1}{w^h} \right) \geq y^l \left(\frac{1}{w^l} - \frac{1}{w^h} \right)
\end{equation*}
- Hence in any mechanism we will find that the high type earns higher gross income than the low type
- Moreover, high type always has higher utility than low type
\begin{equation}
\label{eq:5}
u^h = u(c^h)-y^h/w^h \geq u(c^l)-y^l/w^h \geq u(c^l)-y^l/w^l = u^l
\end{equation}
where the last inequality is strict if \(y^l>0\)
- We guess that (\ref{eq:4}) is binding: high productivity type tends to be “lazy” and wants to mimic low type
- We ignore (\ref{eq:3}) and check at the end that indeed it is not violated
- Rewrite (\ref{eq:4}) as
\begin{equation}
\label{eq:6}
u(c^h)-n^h = u(c^l)-(y^l/w^l)*(w^l/w^h) = u(c^l) - \omega n^l
\end{equation}
with \(\omega = w^l/w^h < 1\)
- Finally, we have the government budget constraint
\begin{equation}
\label{eq:7} \tag{GBC}
\phi c^l +(1-\phi) c^h + E = \phi w^l n^l + (1-\phi)w^h n^h
\end{equation}
- Government needs to raise tax revenue \(E\) to finance a public good
- We write the government’s optimization problem as follows
\begin{split}
\max_{c^l,c^h,n^l,n^h} & \phi (u(c^l)-n^l) +
(1-\phi)(u(c^h)-n^h) \\
-&\lambda (\phi c^l +(1-\phi) c^h+E - \phi w^l n^l - (1-\phi)w^h n^h) \\
+&\mu (u(c^h)-n^h -u(c^l) + \omega n^l)
\end{split}
where \(\lambda\) denotes the Lagrange multiplier on (\ref{eq:7}) and
\(\mu\) on (\ref{eq:4})
- First order conditions can be written as
\begin{align}
\label{eq:FOCcl}
\phi u'(c^l) - \phi \lambda -\mu u'(c^l) & =0 \\
\label{eq:FOCnl}
-\phi + \phi \lambda w^l +\mu \omega & =0 \\
\label{eq:FOCch}
(1-\phi) u'(c^h) - (1-\phi) \lambda +\mu u'(c^h) & =0 \\
\label{eq:FOCnh}
-(1-\phi) + (1-\phi) \lambda w^h -\mu & =0
\end{align}
\begin{align}
\label{eq:ch}
u'(c^h) &= \frac{1}{w^h} \\
\label{eq:cl}
u'(c^l) &= \frac{1}{w^l} \frac{\phi-\omega\mu}{\phi-\mu} > \frac{1}{w^l}
\end{align}
- High type faces a marginal tax rate equal to 0 (no distortion at the top)
- Low type faces a positive marginal tax rate
- Explain that to the SP!
- Because \(u(c)\) is concave, government would like to redistribute from high to low type
- But high type wants to mimic low type
- What is best way to raise l-type’s utility such that h-type does not want to mimic it?
- Reduce \(y^l\) below its first best value: positive marginal tax rate
- We can solve first order conditions above for \(\lambda,\mu\):
\begin{align}
\label{eq:8}
\lambda &= \left( \frac{\phi}{w^l} + \frac{1-\phi}{w^{h}} \right) \\
\label{eq:9}
\mu &= \phi(1-\phi) \left(\frac{1}{\omega} -1 \right)
\end{align}
- Marginal cost of public funds, \(\lambda\), is weighted average of effort costs \(1/w^l,1/w^h\):
- asking each type to earn one unit more, gives government additional unit of budget and satisfies both IC constraints
- Shadow price of (\ref{eq:4}) equals 0 if \(\omega = 1\): both types are the same; no informational rents; first best can be implemented with lump sum taxes
- as \(\omega\) falls, \(\mu\) increases: government would like to redistribute more and (\ref{eq:4}) becomes “more binding”
- Use (\ref{eq:4}) and (\ref{eq:7}) to solve for \(n^l,n^h\) and hence for \(y^l,y^h\):
\begin{align}
\label{eq:10}
y^l &= \phi c^{l}+(1-\phi)c^h+E - (1-\phi)w^h(u(c^h)-u(c^l)) \\
\label{eq:11}
y^h &= \phi c^{l}+(1-\phi)c^h+E + \phi w^h(u(c^h)-u(c^l))
\end{align}