RDP 2013-11: Issues in Estimating New-Keynesian Phillips Curves in the Presence of Unknown Structural Change 6. Conclusion

Structural change has been conjectured to lead to an upward bias in the estimated forward expectations coefficient in NKPCs. We have presented a simple New-Keynesian model that enables us to assess this proposition. The model enables us to distinguish the effects of specification error caused by structural change from small-sample biases that simply arise due to weak instruments. Experiments suggest that the latter dominates the former.

Imposing the restriction that the forward and backward coefficients sum to unity seems to be a useful thing to do, as it generally produces better instruments at the expense of a small specification bias. We also find that biases are relatively small when the coefficient on the forward-looking inflation variable is high and that structural change can improve the quality of instruments, so it may actually aid estimation.

We looked at an empirical study of the euro area Phillips curve by Castle et al (2010), who concluded that the large coefficient on the forward-looking inflation variable was due to structural change. Our analysis suggests that this is not true. It may be that the large coefficient reflects some misspecification, but it is not due to structural change. It seems that the estimators used by Castle et al (2010) were probably subject to weak instrument bias.