Multicollinearity

[Mark van Goor]

Dear all,

I am struggling with multicollinearity in my regression model. I omitted the variable in question, however, I cannot figure out why the multicollinearity exists.

The independent variables included are:

Share of consumer goods exports (CSH)
KOF index
Total Factor Productivity (TFP)

Dependent variables:

Manufacturing employment share
Real value-added share
Nominal value-added share

The situation is as follows:

I wanted to create an interaction term between CSH and the KOF index. However, when testing for multicollinearity, a correlation of 1.0 exists between the interaction term and CSH.

When I started to investigate the problem, I discovered that CSH and the interaction term correlate 1.0, regardless of which variables I included in the place of the KOF index for the interaction term (CSH and TFP, CSH and manufacturing employment share, etc.)

Is anyone familiar with such a situation and knows why this happens?

The data for the CSH is sourced from the Penn World Table 10.01 Trade Detail (Feenstra et al., 2015).

[Jeff Wooldridge]

Is it possible KOF doesn’t change over time?

As per FAQ, show your data, commands, and output.