4 obtain additional or new data since

4. Obtain additional or new data.
 Since multicollinearity is a sample feature, it is possible that
in another sample involving the same variables collinearity
may not be so serious as in the first sample.
 Sometimes simply increasing the size of the sample (if
possible) may attenuate the collinearity problem.
 An often suggested remedy for multicollinearity is simply
increasing the size of the sample, when possible. This remedy
make sense from the perspective that increasing the sample
size will improve the precision of OLS estimators; thus,
reducing the adverse effects of multicollinearity.
5. Transform the variables.
 For example one could estimate a per capita version of the
equation.
 You can use “ changes ” instead of “ levels ” . The advantage is
that “ changes ” may not be as highly correlated as their
“ levels ” .
6. Other methods of remedying multicollinearity.
 Multivariate statistical techniques such as factor analysis and
principal components or techniques such as ridge regression
are also employed to “solve” the problem of
multicollinearity.

ECON 301 - Introduction to Econometrics I May 2013
METU - Department of Economics
Instructor: Dr. Ozan ERUYGUR e-mail: [email protected]
Lecture Notes
18
5. Summary
The effects of multicollinearity and the means of identifying them are
summarized below:
 Estimates of parameters are still unbiased, consistent, and BLUE.
 The standard errors of the coefficients might be high, making
individual coefficients insignificant.
 The model might exhibit high R
2
but low t -values:
 A joint F-test might reject the null hypothesis that several
coefficients are zero, but individual t -tests might accept them,
leading to an apparent contradiction.
 Coefficients may change considerably when variables are added or
deleted, thus making the interpretation of individual coefficients
difficult.