Omitted variable bias vs multicollinearity

[Siewwen]

Hi I am running a regression and have a coefficient for x with a negative value which I know should be positive. I want to add another variable to the model that is negatively correlated with x, but due to the correlation between the variables I'm think that there will be multicollinearity. Is there a way to get around the problem? I need to interpret the coeff for x.

To be more specific,
Tourist arrivals = c + b1 real exchange rate + ... + u
Variable I want to add is GDP

Thanks!

[startz]

If you think that GDP belongs in the equation, then add it. Multicollinearity does not bias coefficients, omitted variable bias does.

[EViews Glenn]

Startz, once again on a never-ending crusade to stamp out excessive concern over multicollinearity...I suspect that the problem wouldn't be as bad if the textbooks discussed it in the context of (lack-of-)identification.

[Siewwen]

Thanks for the helpful reply!