Question

Given an output from the White Test to test for heteroskedasticity, how do you know if heteroskedasticity is present or not? What exactly is heteroskedasticity?

Answer 1

Heteroskedasticity should be written with "k" not with "c". Believe it or not, there is a paper about it.

Heteroskedasticity means that the variance of each error term is different. This change might depend on an explanatory variable or on time.

Here is an easy example for time varying heteroskedasticity:
During the recession times uncertainty increases a lot. A firm might do larger amount mistakes in their calculations for their returns, for instance, compared to good times. Hence, the errors in their calculations are changing over time. Large errors during recessions, small errors during booms.

Here is an example for heteroskedasticity due to an explanatory variable:
Imagine there is a machine that produces tables. You need to provide the measurements of the table you want. And imagine, first you want to produce a lot of small tables, e.g. 1 meter long. The machine cannot produce exactly 1 meter long tables. It will make mistake probably ±± 2 cm on average (note that on average the mistake is zero. But the absolute mistake will be positive on average). In some tables it might be little bit more, in some tables it might be less; but on average it is 2 cm of mistake. Next, imagine you want to produce much longer table, e.g., 100 meter long. This time the machine will make some mistakes too. But the mistake will not be 2 cm anymore, probably it will be ±± 1-2 meters of mistakes on average. Hence, the bottomline is that the variation in the errors depends on the measurement input you provide. For instance, we could have written for this example (even though it is not exactly correct to write this) Var(ei|measurementi)=measurementi/50Var(ei|measurementi)=measurementi/50.