Monte Carlo power comparison of seven most commonly used heteroscedasticity tests

Assumption of the classical linear regression model states that the disturbances should have a constant (equal) variance. When this requirement is not met, the loss in efficiency in using ordinary least squares may be substantial and the biases in estimated standard errors may lead to invalid inferences. This problem is known as heteroscedasticity. There are many tests for heteroscedasticity, we want to know which test is more powerful. We use Monte Carlo simulation to compare the power of seven most commonly used tests for detecting heteroscedasticity, namely, Breusch-Pagan test, Glejser test, Goldfeld-Quandt test, Harvey-Godfrey test, Harrison-McCabe test, Park test, and White test for six common types of heteroscedasticity. Simulation results show that the Harrison-McCabe test has generally the most power in all of the six common types of heteroscedasticity and the White test has generally the least power in all of the six common types of heteroscedasticity.

Automated summary

A comparison study on seven tests for detecting heteroscedasticity revealed that the Harrison-McCabe test generally has the highest power, while the White test generally has the lowest power across all six common types of heteroscedasticity.