Estimating the Swedish Phillips Relationship in a Markov-Switching Vector Autoregression

Sammanfattning: The Swedish Phillips relationship was recently examined by Svensson (2015), who found that the long-run trade-off is downward-sloping. Hence, there is an unemployment cost of inflation. He argues that this has occurred because inflation expectations are anchored to the inflation target, while average inflation has deviated from the target. This empirical finding has a large implication on the estimation of the Phillips curve. We wish to examine the Phillips relationship in light of the debate on whether surveyed expectations are anchored, by using an econometric method that is robust to whether this assumption holds or not. We study the Swedish Phillips curve in a regime shifting framework using the same data as Svensson (2015) for the period 1997Q4-2011Q4. We estimate a bivariate Markov-switching VAR with inflation and the unemployment rate, with regime-dependent dynamics and heteroskedasticity. We confirm the existence of a non-vertical long-run Phillips curve and conclude that linear estimation of the inflation-unemployment relation seems to be robust with regard to expectations over this period.