Page 25 - FINAL CFA II SLIDES JUNE 2019 DAY 3

READING 8: MULTIPLE REGRESSION AND ISSUES IN REGRESSION ANALYSIS
Correcting Heteroskedasticity – 2 Methods!

MODULE 8.6: ASSUMPTIONS: HETEROSKEDASTICITY
Method 1:
Calculate robust standard errors (White-corrected standard errors or heteroskedasticity-consistent standard errors).
(In the exam, use robust standard errors (instead of SE) to compute t-statistics if data shows heteroskedasticity is present).

Method 2:
Use generalized least squares -it attempts to eliminate heteroskedasticity by modifying the original equation.

EXAMPLE: Using White-corrected standard errors: An analyst runs a regression of annualized Treasury bill rates (the dependent
variable) on annual inflation rates (the independent variable) using monthly data for 10 years. The results of the regression are shown in
the following table.

He determines using the BP test that heteroskedasticity is
present, so he also estimates the White-corrected standard error
for the coefficient on inflation to be 0.31. The critical two-tail 5% t-
value for 118 df is 1.98. Is inflation statistically significant at
the 5% level?

1.98 Meaning?

After correcting for heteroskedasticity, we Fail to Reject the Ho that the inflation coefficient is zero!

Therefore, inflation is NOT statistically significant

This means the inflation coefficient estimate of 0.60 was not affected by heteroskedasticity:
• Had we not used ‘white corrected SE’ we could have stuck with a very low SE of 0.28, inflating t-statistics to a very high 2.14

• Now 2.14 > 1.98: So what? We risked Rejected Ho when it is true (Type 1 Error)

After using the higher White-corrected standard error of 0.31, the t-statistic fell to 1.94.