RDP 2012-01: Co-movement in Inflation Appendix B: Feasible Least Squares Estimation

As discussed in Appendix A, for σ2 > 0 the error term in the restricted SUR form of the model is heteroskedastic, with variance given by σtΩ where Inline Equation. In the version of the model where σ2 was estimated, the posterior mean estimate of σ2 was found to be small but away from zero, which suggests there is some evidence of (minor) heteroskedasticity in the errors. Therefore as a robustness check we also estimated the restricted SUR form of the model using feasible least squares and calculated robust standard errors. Table B1 presents t-statistics for some of the key coefficients in the model. While there are some differences compared with the Bayesian estimates, qualitatively the results are similar. In particular, the common inflation indicator was still found to be significant in both the headline and core inflation versions of the model and oil price inflation was found to be a significant explanator of headline inflation.

Table B1: Feasible Least Squares Estimation Results
Coefficient t-statistics
Model with headline inflation Model with core inflation
Inflation equations
Common inflation indicator 4.95 5.51
Common real activity indicator −0.16 −0.06
Oil price inflation 4.58 0.32
Non-fuel commodity price inflation 1.62 0.75
Real variable equations
Common real activity indicator 2.07 1.90
Oil price inflation −0.13 −0.16
Non-fuel commodity inflation 2.65 2.75
Notes: Estimation sample is 1981:Q2 to 2011:Q1; t-statistics calculated using robust standard errors; numbers in bold indicate significance at the 10 per cent level or lower