Question: When using OLS, we assume that the errors are independent (i.e., not serially correlated) and normally distributed with constant variance (i.e., homoscedastic). These assumptions allow us to use the F and t distributions to perform significance tests on the estimated coefficients in the model. However, we often wish to estimate models in which we doubt the validity of these assumptions. For the following scenarios, explain why these assumptions may fail:

a. Estimating a model of stock prices as a function of changing market conditions

b. Estimating one's expenditure on housing as a function of one's income using cross-sectional data

c. Estimating the effect of median income on the per-capita level of public education spending using data from a sample of 100 cities.