How to tackle endogenity?

Question

I need to estimate following regresssion:

$$ Price_{i} = \alpha_{i} + \beta_{i}Wage_{i} + \sum_k\gamma_{i}^{k}Z_{i}^{k} $$

where $Price_{i}$ is a price of a product, $i$ is a region number, $Z$ is a vector of exogenous variables, $k$ is an id of variable in $Z$.

$Wage_{i}$ is endogenous variable.

If I use $Wage_{i}$ divided by $Price_{i}$ in the regression, will be the problem of endogenity solved?

can you explain your question more? it is not clear what are $Z_i$ and $k$ as well. what do you mean by 'number of regressor from Z'. is that meaning that k is a random number? also,is the left hand side just price or is that $price_i$? — TPArrow, Aug 03 '15 at 20:43
Hamed, I corrected description. $Z_{i}$ is a vector of variable for region $i$, $k$ is id variables from $Z$. — Pythia, Aug 04 '15 at 07:05
I found that aproach in question is wrong as follows from economic theory, and it will not solve problem of endogenity. Wage will influenced by price regardless on selected measure (in the idea above it is number of the product with price equal to $Price$ that can be bought) — Pythia, Aug 04 '15 at 07:08
@Pyhia: do you mean that you have a system of equations consisting of (1) $p= \beta_1 + \beta_2 w + \beta_3 z^{(1)} + \beta_4 z^{(2)} \dots $ and (2) $w = \alpha_1 + \alpha_2 p$ ? — , Aug 04 '15 at 08:45

score 0 · Accepted Answer · answered Aug 04 '15 at 08:38

Endogeneity in simple terms means that explanatory variable is influenced by dependent variable. Dividing non-endogenous explanatory variable by dependent variable is a sure way of introducing endogeneity, not the other way around. If you suspect that you have endogenous variables in the model you should change your estimation method from OLS to two stage OLS with instrumental variables.

How to tackle endogenity?

1 Answers1