Question

Why are linear regression models limited in its power to project into the future? Why should one be very careful in using linear regressions to make projections?

Answer 1

One can quote number of reasons why linear regression models in forecasting can be limited. Some of them include

1) Anamolous Data

Sometimes the trend can be accounted by certain anamolous data for which linear regression cannot explain on the whole.

2) Treatment of cycles

It can be argued that one of the major criticisms of linear trends is when they are used over only a part of a cyclical pattern where there can be a chance that the upward trend over the rising part of a cycle is grossly misleading.

3) Linear regression looks only at the mean and variance of the dependent variable which can get to quote certain negative projections.

As the linear regression models thoroughly use projections which are sensitive to outliers and usage of anamolous data is out of reach, one must be careful in using linear regressions to make projections.