Scribd
Upload
11 views1 page

Handout - 6 - Omitted Variable Bias Summary Table

This document discusses how omitting a variable (Z) that is correlated with the explanatory variable (X) and also affects the outcome variable (Y) can bias estimates of the coefficient on X. It presents a table showing how the direction of the bias depends on the sign of the correlation between X and Z and whether Z positively or negatively affects Y. It provides an example where omitting a variable that negatively determines GDP and negatively affected European settlement leads to an upward bias in estimating the effect of European settlement on GDP.

Uploaded by

mkevane
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
11 views1 page

Handout - 6 - Omitted Variable Bias Summary Table

This document discusses how omitting a variable (Z) that is correlated with the explanatory variable (X) and also affects the outcome variable (Y) can bias estimates of the coefficient on X. It presents a table showing how the direction of the bias depends on the sign of the correlation between X and Z and whether Z positively or negatively affects Y. It provides an example where omitting a variable that negatively determines GDP and negatively affected European settlement leads to an upward bias in estimating the effect of European settlement on GDP.

Uploaded by

mkevane
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

ECON 41

Outline #6: Understanding possible omitted variable bias (OVB) in single explanatory variable regression

Suppose we estimate Yi = β0 + β1 X i + ui but we think we have an omitted variable Z that is correlated with X and
^
also determines the outcome Y. What will be the direction and magnitude of the bias in our estimate of β1 ? We
^
know that the bias depends on the correlation between X and the error term, β1 → β1 + ρX ,u (σ u / σ X ) . But the
correlation ρX ,u is not obvious. The following table can help.

Suppose Suppose Suppose Z Then sign of So then if the The unbiased


underlying correlation affects Y… ρX,u will be… biased estimate estimated
coefficient β1 between X and (from estimating (obtained by
is… Z is… the equation) including Z in
were… the regression
and re-
estimating)
would be
positive positive positively + 6.4 4.4
negatively - 6.4 8.4
negative positively - 6.4 8.4
negatively + 6.4 4.4
negative positive positively + -6.4 -8.4
negatively - -6.4 -4.4
negative positively - -6.4 -4.4
negatively + -6.4 -8.4

So we can imagine estimating a relationship where the outcome variable is “current GDP per capita” and the
explanatory variable is “Extent of European settlement in region during era of imperialism.” We estimate a positive
coefficient. Did European settlement lead to higher GDP? We might think an omitted variable is “Index of tropical
climate” with higher numbers of the index indicating a more tropical climate. This variable negatively determines
GDP, and also negatively affected European settlement. So we see the estimated coefficient is biased upwards, and
the “actual” or unbiased effect of European settlement is smaller than what we have estimated.

The trickiest situation is when the bias is large enough ρX ,u (σ u / σ X ) that the sign is reversed! So in the above
example, European settlement could actually have had a negative effect on subsequent GDP, but the biased
coefficient is positive because Europeans strongly avoided tropical areas which might generally have lower GDP
than temperate areas. In other words, European settlement reduced GDP growth (from what it could have been), just
not enough to make subsequent GDP actually lower than tropical GDP levels.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy