Major Implications:
The important implications of multicollineam'ty are given below. In the presence of high, though not perfect, multicollinearity:
1) The OLS coefficient estimates have large variances.
2) Because of large variance the confidence intervals will be very large, which in turn means that there is a high probability of accepting the null hypothesis of zero coefficient, even when the actual parameter is positive.
3) Overall the regression may do very well, i.e., the R2 may be quite high despite not being able to reject the hypothesis of one or more parameters being equal to zero.
4) The OLS estimates and their standard errors can be verysensitive to small changes in the data.
Note that fiom a practical standpoint this is an extremely serious problem since most empirical studies try to estimate the impact of one or more economic variables on a particular variable, such as the income consumption relationship discussed above.
So if income and wealth are highly correlated then although the regression gives the true coefficient estimates, we cannot reject the null hypothesis that the impact of income on consumption expenditure is zero. This means that the whole purpose of the study islost, as nothing can be conclusively (statistically speaking) proved.