When the index of government spending is left out in Model C, the adjusted R2 value
dropped down for more than 50% from 0.489 in Model B to only 0.227 in Model C. One
suspicion could be that there is a Multicollinearity in the model. This occurs when there are two
or more explanatory variables that are correlated. However, the standard errors of the estimated
coefficients in the three models are relatively small and the t-statistics values are not small.
Therefore, Multicollinearity might not be the main problem and this drop in adjusted R2 might be
caused by the limited sample size which made the data to be relatively sensitive to slight changes
in the models.