There is a population consisting of N subjects, with data variables x and y. A simple regression equation can in principle be fitted by OLS to the population data: yi = a + bxi + ui, where N i=1 ui = N i=1 xiui = 0.

Although Harry does not have access to data on the full population, he can take a sample of size n

n is moderately large, but small relative to N. He will estimate the parameters a and b by running a regression of yi on xi for i in the sample. He will have an intercept in the equation.

(a) Are the OLS estimates biased or unbiased? Why? (Hint: is the true relationship linear?)

(b) Should he believe the standard errors printed out by the computer?

Discuss briefly.