Where the errors u_i are multiplicative and have mean one and constant variance. Then the variance of ε_i is proportional to (β_0+β_1 X_i )^2. The effect of nonconstant (heterogeneous) variances on least squares estimators is similar to that of correlated errors. The least squares estimators are no longer efficient and the variance formulae in equations 1.26 and 1.32 are not valid. Alternative methods are discussed in Chapter 11.