independent variable relationships

[Adrienne Wold]

Hi, I am trying to make sense of the following results from a regression model where I have vote share as the outcome and perceptions of bribery as the X. The result of a simple bivariate model is that the magnitude of the coefficient is sizable- -0.2 and significant. When I include another variable- perceptions of corruption in general- the bribery coefficient drops to -0.3 and loses its significance, while the perceptions of corruption variable magnitude is -.30 and significant.

Given that when I enter the general corruption variable into the model, the statistical significance of bribery drops out, would it be fair to say that since it takes over the explanatory role, that bribery is just a component of a more generalized corruption form, which is actually a stronger predictor of vote share?

[Nick Cox]

Plotting your two predictors against each other should help explain.

[Carlo Lazzaro]

Adrienne:
in my opinion, bivariate regressions were created by the devil during her/his leisure time.
Most of the data generating process cannot be explained in terms of y regressed onto x, because it's usually a matter of several predictors.
That said, I would first check whether your model is not misspecified.
As far as your question is concerned, my gut-feeling is that your result makes sense as the perception that some politicians might be corrupted is probably more widespread among voters than the awareness that some of them receive bribes.
As a sidelight, following the FAQ and posting what you typed and what Stata gave you back would be helpful. Thanks.

[Adrienne Wold]

Also forgot to say these are fairly correlated (~0.49)