Scribd
Upload
129 views

Measurement Errors

This document discusses measurement errors in regression analysis. It explains that when explanatory variables are measured with error, it can lead to an attenuation bias in the OLS estimates towards zero. Measurement errors in the explained variable do not bias the estimates but increase their inefficiency. An empirical example shows how measurement error in income at the individual level helps explain why micro savings regressions perform poorly compared to macro versions.

Uploaded by

David Adeabah Osafo
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
129 views

Measurement Errors

This document discusses measurement errors in regression analysis. It explains that when explanatory variables are measured with error, it can lead to an attenuation bias in the OLS estimates towards zero. Measurement errors in the explained variable do not bias the estimates but increase their inefficiency. An empirical example shows how measurement error in income at the individual level helps explain why micro savings regressions perform poorly compared to macro versions.

Uploaded by

David Adeabah Osafo
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/ 16

Measurement Errors

Walter Sosa-Escudero
Econ 471. Econometric Analysis. Spring 2009

April 21, 2009

Walter Sosa-Escudero Measurement Errors


Motivation

One particular case of explained variable that is random arises


when it is measured with errors.

Walter Sosa-Escudero Measurement Errors


Explanatory variables measured with error

Consider the following simple case

Ci = 1 + 2 Xi + ui

For example, C can be consumption and Xi is permanent


income.
Now suppose we can only observe a noisy version of Xi :

Xi = Xi + i

i is a measurement error.
For example, Xi can be current income.

Walter Sosa-Escudero Measurement Errors


We will assume E(i ) = 0, V (i ) = 2 and uncorrelated with
Xi and ui .
Replacing Xi = Xi i , the regression model can be written as

Ci = 1 + 2 Xi + i
with = 2 i + ui

Walter Sosa-Escudero Measurement Errors


Now

Cov(Xi , i ) = E(Xi , i ) = E [(Xi + i ) (2 i + ui )]


= 2 2 6= 0

Then the OLS estimator that regresses Yi on Xi is biased, since Xi


is correlated with the error term i .

Walter Sosa-Escudero Measurement Errors


It can be shown that when n is large:

2
 
' 2 X 2
+ X
This is a very important result.
2 /( 2 + 2 ) is a number between 0 and 1.
X X

Then || tends to be smaller than : measurement errors
imply an attenuation bias.
Measurement errors shrink estimates towards zero.
In practice people blame non-significances to measurement
errors. Be careful!

Walter Sosa-Escudero Measurement Errors


Walter Sosa-Escudero Measurement Errors
Walter Sosa-Escudero Measurement Errors
Walter Sosa-Escudero Measurement Errors
Walter Sosa-Escudero Measurement Errors
Measurement errors in the explained variable

Y = 1 + 2 Xi + ui
Yi = Y + i
i uncorrelated with ui , V (i ) = 2
Replacing above

Yi = 1 + 2 Xi + i + ui
= 1 + 2 Xi + i

Walter Sosa-Escudero Measurement Errors


Measurement errors in the explained variable
Note that by our assumptions, i i + ui is uncorrelated
with Xi , then the presence of measurement errors in the
explained variable does not bias the OLS estimator
V (i ) = 2 + 2 > 2 , the variance of the new error term is
larger than the one in the model without measurement error:
mesurement errors in the explained variable makes OLS more
inefficient.

Walter Sosa-Escudero Measurement Errors


Walter Sosa-Escudero Measurement Errors
Empirical example: savings function

Saving models for individuals:

si = + yi + ui

where: si = savings; yi = permanent income (not observable)

We estimate:

si = + yi + i
yi = current income
Remember that the macro version (at the national level) had a
very high R2 . Now we estimate the micro version, with household
data.

Walter Sosa-Escudero Measurement Errors


Income is not relevant: measurement error (attenuation bias)

Walter Sosa-Escudero Measurement Errors


Why does the micro version perform so poorly compared to the
macro version?
In the macro verion Yt is current national income, which is like an
average income
1X
yt = nyit
n
i=1
1X 1X
= nyit + nit
n n
i=1 i=1

Note " #
1X 1
V nit = 2
n n
i=1

Then, the importance of the aggregate measurement error tends


to vanish: aggreate current incomme does a relatively good job
capturing aggregate permanent income, but not at the individual
level.
Walter Sosa-Escudero Measurement Errors

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