Econometric Theory/Heteroskedasticity

From Wikibooks, open books for an open world
Jump to: navigation, search

One of our CLR assumptions for linear regression is that our disturbance terms are homoscedastic, meaning they have equal scatter (Var(\epsilon) = \sigma^2 ). However, there are times that regressions end up with heteroscedastic disturbance terms, meaning the scatters are unequal ( Var(\epsilon) = \sigma_i^2).

Hetroscedastic, Cone shape


Causes of Heteroscedasticity[edit]

Heteroscedasticity are more common in cross-sectional data than in time series. It is usually due to a scale or size factor.

Example: In basic Keynesian economics, we assume that savings and income are determined by wealth and income. Agents that have more wealth and income are more likely to save, this will produce a hetroscedastic relationship.

Consequences of Heteroscedastity[edit]

1) OLS Coefficients are still unbiased for true value.  E(\hat{\beta}) = \beta

Unbiased coefficients depend on  E(\epsilon) = 0, cov(x_i,\epsilon_i) = 0

So the regression is safe from hetroscedasticity. on this assumption.

2) OLS Coefficients are not efficient. There exists an alternative to the OLS Coefficant that has a smaller variance than the OLS one.  \exists_\tilde{\beta} st. var(\tilde{\beta}) < var(\hat{\beta}^{OLS}) where E(\tilde{\beta}) = \beta and therefore, is more efficient.

3) Hypothesis tests, in the presence of hetroscedasity of an OLS coefficient, based on the standard error of the coefficient are invalid.


Bias in estmated variance of OLS Estimators causes ineffiancies.

Recall:  \hat{\beta}^{OLS} = \beta +  \frac{\sum (x_i + \bar{x})\epsilon_i}{\sum(x_i + \bar{x})^2}

When beta-hat is unbiased, the second term goes to zero. However, with hetroscedasicity, the second term is not zero.

Derive  var(\hat{\beta}^{OLS}) when  var(\epsilon) = \sigma_i^2  var(\hat{\beta}^{OLS}) = var(\frac{\sum (x_i + \bar{x})\epsilon_i}{\sum(x_i + \bar{x})^2})       = (\frac{1}{\sum (x_i - \bar{x})^2})^2 var(\sum(x_i - \bar{x})\epsilon)       = (\frac{1}{\sum (x_i - \bar{x})^2})^2 \sum var((x_i - \bar{x})\epsilon)       = (\frac{1}{\sum (x_i - \bar{x})^2})^2 \sum x_i - \bar{x})^2 var(\epsilon)       = (\frac{1}{\sum (x_i - \bar{x})^2})^2 \sum x_i - \bar{x})^2 \sigma_i^2


We can represent the differences in variances in the error term by the matrix Z

 \sigma_i^2 = Z \sigma^2