Expected revenue in an auction

98 Views Asked by At

Consider an auction in which two independent buyers bid for an antique object. If each of them draws their valuations from $\mathcal{U}[0,1]$ and the auction lets the highest bidder take the object by paying the lowest bid, what's the expected revenue of the auction?

Denote the two buyers' bids by the r.v.s $X, Y \sim \mathcal{U}[0,1]$.

I think we have to calculate $\mathbb{E}(X | X \geq Y) + \mathbb{E}(Y | Y \geq X)$. When I calculate it, I get $$2 \left(\int_{0}^{1} \int_{0}^{x} x \cdot \frac{1}{1-y} dydx\right) = 1.5$$

This is incorrect. Where did I go wrong?

Please note that I know a possible way to solve this is to evaluate $\mathbb{E}(\min(X,Y))$. But I am trying to figure out where I went wrong with my above method.

1

There are 1 best solutions below

5
On BEST ANSWER

This is a funny auction but why not. The expected revenue is \begin{align} \mathbb E\big[\min(X,Y)\big]&=\textstyle\int_0^1\int_0^1\min(x,y)\,dx\,dy=\int_0^1\int_0^yx\,dx\,dy+\int_0^1\int_y^1y\,dx\,dy\\ &=\textstyle\int_0^1\frac{y^2}{2}\,dy+\int_0^1y-y^2\,dy=\int_0^1y-\frac{y^2}{2}\,dy\\ &=\textstyle\frac{1}{2}-\frac{1}{6}\\[2mm]&=\frac{1}{3}\,. \end{align} Alternatively one could use the law of total expectation by which \begin{align} \mathbb E\big[\min(X,Y)\big]&= \mathbb E\big[Y1_{\{X\ge Y\}}\big]+\mathbb E\big[X1_{\{Y>X\}}\big]\\[2mm] &=\underbrace{\mathbb E\big[Y\big|X\ge Y\big]}_{\textstyle=\frac{1}{3}}\,\underbrace{\mathbb P\big\{X\ge Y\big\}}_{\textstyle=\frac{1}{2}}+\underbrace{\mathbb E\big[X\big|Y> X\big]}_{\textstyle=\frac{1}{3}}\,\underbrace{\mathbb P\big\{Y>X\big\}}_{\textstyle=\frac{1}{2}}\\ &=\frac{1}{3}\,. \end{align} Proof.

Let's calculate the conditional expectations $$\mathbb E[Y|X\ge Y],\quad\mathbb E[Y|Y> X],\quad\mathbb E[X|X\ge Y],\quad\mathbb E[X|Y> X]\,$$ that were discussed in the comments. From $\mathbb P\{X\ge Y\}=\mathbb P\{Y>X\}=\frac{1}{2}$ we get \begin{align} \mathbb P\{Y\le z|X\ge Y\}&=\frac{\int_0^z\int_y^1\,dx\,dy}{1/2} =2\textstyle\int_0^z1-y\,dy=2z-z^2\,,\\ \mathbb P\{Y\le z|Y>X\}&=\frac{\int_0^z\int_0^y\,dx\,dy}{1/2} =2\textstyle\int_0^zy\,dy=z^2\,. \end{align} Taking the derivative w.r.t. $z$ the conditional PDFs of $Y$ are therefore, \begin{align} p(y|X\ge Y)=2-2y\,,\quad p(y|Y>X)=2y\,. \end{align} This leads to

\begin{align} \mathbb E[Y|X\ge Y]&=\textstyle\int_0^1y(2-2y)\,dy=2-\frac{2}{3}=\displaystyle\frac{1}{3}\,,\tag{1}\\ \mathbb E[Y|Y>X]&=\textstyle\int_0^12y^2\,dy=\displaystyle\frac{2}{3}\,,\\ \mathbb E[X|X\ge Y]&=\frac{2}{3}\,,\quad\text{ (by symmetry) }\\ \mathbb E[X|Y>X]&=\frac{1}{3}\,.\quad\text{ (by symmetry) }\tag{2}\\ \end{align} $$\tag*{$\Box$} \quad $$ Remarks.

The proposed solutions in the edited OP, resp. in a recent comment \begin{align} &\mathbb E[X|X\ge Y]+\mathbb E[Y|Y> X]=\frac{4}{3}\,,\\ &\mathbb E[Y|X\ge Y]+\mathbb E[X|Y> X]=\frac{2}{3} \end{align} are both wrong because we add expectations that are conditioned on mutually exclusive events. They are also different from $1/3$ which was mentioned as the correct answer in the OP that got overridden by some edits.

The other proposed solution (1) in a comment equals the correct solution (as well as (2) does) but I think this is a coincidence that is based on the independence of $X$ and $Y$ and on the uniformity of their distributions.