0
$\begingroup$

This is an old past paper question that I am struggling to understand, so any help or hints would be appreciated...

Consider the choice between two options, such as two product brands. Let $U_0$ denote the utility of outcome $y = 0$ and $U_1$ the utility of $y = 1$. For $y = 0$ and $1$, suppose that $U_y = \alpha_y +\beta_yx+\epsilon_y$, using a scale such that εy has some standard distribution. A subject selects $y = 1$ if $U_1 > U_0$ for that subject. If $\epsilon_0$ and $\epsilon_1$ are independent $N(0,1)$ random variables, show that $P(Y = 1)$ satisfies the probit model.

So far, I have that $P(y=1)=P(U_1>U_0)=P(\alpha_1 +\beta_1x+\epsilon_1>\alpha_0 +\beta_0x+\epsilon_0)=P(\epsilon_0-\epsilon_1<\alpha_1-\alpha_0+x(\beta_1-\beta_0))$

but I am not sure where to go from here, or if what I am doing is even correct. Any help would be appreciated.

$\endgroup$
1
  • $\begingroup$ Is this a question from a course or textbook? If so, please add the [self-study] tag & read its wiki. $\endgroup$ May 17 '16 at 15:08
1
$\begingroup$

By definition the Probit model is:

$P(Y=1|X)=\Phi(X^T\beta),$

where $\Phi$ is the CDF of the normal. You're basically done since $\epsilon_0-\epsilon_1$ is going to be normally distributed with mean 0 and some variance like $2\sigma^2$. Then $P(\epsilon_1-\epsilon_2<X^T\beta)$, which after rescaling $\beta$ will give you $\Phi(X^T\beta)$.

$\endgroup$
1
  • $\begingroup$ what do you mean by rescaling $\beta$ ? because the $\epsilon$s have variance 1, does that mean $\epsilon_0-\epsilon_1$ will have variance 2, so do I need to divide $\beta$ by 2? $\endgroup$
    – kat245
    May 17 '16 at 18:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.