Introductory Econometrics: A Modern Approach (7e)

Chapter 8
Heteroskedasticity

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 1
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (1 of 18)
• Consequences of heteroskedasticity for OLS
• OLS still unbiased and consistent under heteroskedastictiy!
• Also, interpretation of R-squared is not changed

• Heteroskedasticity invalidates variance formulas for OLS estimators

• The usual F tests and t tests are not valid under heteroskedasticity
• Under heteroskedasticity, OLS is no longer the best linear unbiased estimator
(BLUE); there may be more efficient linear estimators

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 2
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (2 of 18)
• Heteroskedasticity-robust inference after OLS estimation
• Formulas for OLS standard errors and related statistics have been developed
that are robust to heteroskedasticity of unknown form.
• All formulas are only valid in large samples.
• Formula for heteroskedasticity-robust OLS standard error.

• Using these formulas, the usual t test is valid asymptotically.

• The usual F statistic does not work under heteroskedasticity, but
heteroskedasticity robust versions are available in most software.

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 3
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (3 of 18)
• Example: Hourly wage equation

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 4
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (4 of 18)
• Testing for heteroskedasticity
• It may still be interesting whether there is heteroskedasticity because then OLS
may not be the most efficient linear estimator anymore.

• Breusch-Pagan test for heteroskedasticity

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 5
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (5 of 18)
• Breusch-Pagan test for heteroskedasticity (cont.)

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 6
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (6 of 18) Hprice1.dta

• Example: Heteroskedasticity in housing price equations

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 7
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (7 of 18)
• The White test for heteroskedasticity

• Disadvantage of this form of the White test

• Including all squares and interactions leads to a large number of estimated
parameters (e.g. k=6 leads to 27 parameters to be estimated).

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 8
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (8 of 18)
• Alternative form of the White test

• Example: Heteroskedasticity in (log) housing price equations

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 9
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (9 of 18)
• Weighted least squares estimation
• Heteroskedasticity is known up to a multiplicative constant

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 10
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (10 of 18)

• Example: Savings and income

• The transformed model is homoskedastic

• If the other Gauss-Markov assumptions hold as well, OLS applied to

the transformed model is the best linear unbiased estimator.
© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 11
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (11 of 18)

• OLS in the transformed model is weighted least squares (WLS)

• Why is WLS more efficient than OLS in the original model?

• Observations with a large variance are less informative than observations with
small variance and therefore should get less weight.
• WLS is a special case of generalized least squares (GLS)

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 12
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (12 of 18)

• Example: Financial wealth equation
Independent Variables (1) (2) (3) (4)
OLS WLS OLS WLS
inc .821 .787 .771 .740
(.104) (.063) (.100) (.064)
(age – 25)2 - - .0251 .0175
(.0043) (.0019)
male - - 2.48 1.84
(2.06) (1.56)
e401k - - 6.89 5.19
(2.29) (1.70)
intercept -10.57 -9.58 -20.98 -16.70
(2.53) (1.65) (3.50) (1.96)
Observations 2,017 2,017 2,017 2,017

R-squared .0827 .0709 .1279 .1115

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 13
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (13 of 18)

• Important special case of heteroskedasticity
• If the observations are reported as averages at the city/county/state/-
country/firm level, they should be weighted by the size of the unit

• If errors are homoskedastic at the individual-level, WLS with weights equal to firm size mi should
be used. If the assumption of homoskedasticity at the individual-level is not exactly right, one can
calculate robust standard errors after WLS (i.e. for the transformed model).

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 14
 Introductory Econometrics: A Modern Approach (7e)

Weighted Least Squares: Why and How

• In survey data, individuals are often sampled from population with
different probabilities
• For example, the most poor or rich individuals often has a higher probability to
be sampled.
• If we are computing the mean income, an unbiased estimator is the Horvitz-
Thompson estimator:

1
where 𝜋𝑖 is the sampling probability, and 𝑀 = σ𝑖
𝜋𝑖
• Principle: weight = inverse probability, i.e. Inverse Probability Weighting(IPW)

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 15
 Introductory Econometrics: A Modern Approach (7e)

Weighted Least Squares: Why and How

• Similar weighting strategy can be used for survey data in

regression analysis
• Inverse of probability can be used as the weights in regression so
that the regression is more representative.
• Note that the inverse of probability is equivalent to number of
individuals represented by a specific sample.
• For example, if we have 1000 samples in Shanghai, then every
single individual in Shanghai is sampled with probability
1000/24million=1/2400, which means each sample represents
2400 individuals in Shanghai.
• In Stata: [w=swgt]

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 16
 Introductory Econometrics: A Modern Approach (7e)

Weighted Least Squares: Why and How

• If the observations are reported as averages at the city/county/state/-
country/firm level, each observation can be view as a single sample
from the city/county/state/-country/firm
• So according to the IPW principle, the weights = size of the unit,
which is equivalent to the weights above.
• But here, the motivation is different: representative v.s. Heteroscedasticity
• Since we don‘t konw whether the homoscedasticity of individuals‘ error term is
satisfied, robust standard errors should be use.
• For example, if the dependent variable if averages at the city level, then
weight=city population.

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 17
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (14 of 18)

• Unknown heteroskedasticity function (feasible GLS)

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 18
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (15 of 18)

• Example: Demand for cigarettes
• Estimation by OLS SMOKE.dta

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 19
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (16 of 18)

• Estimation by FGLS

• Discussion
• The income elasticity is now statistically significant; other coefficients are also
more precisely estimated (without changing qualitative results).

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 20
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (17 of 18)

• What if the assumed heteroskedasticity function is wrong?
• If the heteroskedasticity function is misspecified, WLS is still consistent under
MLR.1 – MLR.4, but robust standard errors should be computed.
• WLS is consistent under MLR.4 but not necessarily under MLR.4‘

• If OLS and WLS produce very different estimates, this typically indicates that
some other assumptions (e.g. MLR.4) are wrong.
• If there is strong heteroskedasticity, it is still often better to use a wrong form of
heteroskedasticity in order to increase efficiency.

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 21
 Introductory Econometrics: A Modern Approach (7e)

Heteroskedasticity (18 of 18)

• WLS in the linear probability model

• Discussion
• Infeasible if LPM predictions are below zero or greater than one.
• If such cases are rare, they may be adjusted to values such as .01/.99.
• Otherwise, it is probably better to use OLS with robust standard errors.

© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 22