Phillips Curve bashing and immaculate inflation

A few points prompted by the recent exchange between David Andolfatto and Paul Krugman, and the recurrent phenomenon of comment pieces on the demise of the Philips Curve.  These pieces are hot lately because of the surprising coincidence of falls in unemployment with a failure of inflation to pick up materially, particularly in the UK and the US.

First, fluctuations in the correlation between unemployment and inflation – which have been a thing ever since I can remember paying attention [see, for example, this paper by Luca Benati] – don’t tell us much on their own about the existence or not of the aggregate supply relations that comprise the Philips Curve in modern macro models.  A shifting mixture of shocks to demand and supply can change this correlation straightforwardly in model world. Shifts in the data are therefore not themselves sufficient to refute the model and its Philips Curve.

Second – a point Nick Rowe reminded me to make – is that the existence of the Phillips Curve nonetheless does not entitle us to say that falls in unemployment ’cause’ a rise in inflation.  The proximate cause is the underlying demand shock, in this case.

Third, if you want to attack the Philips Curve, it seems to me you have to contend that neither prices, nominal wages nor information sets are sticky.  With any one of those present in a model, a monetary shock that constituted a loosening would cause a positive output gap.

The empirical macro data concludes, from what I recall, that nominal – eg monetary policy – shocks do have real effects.  So this is counterfactual for those who believe in flexible prices/information.  One can certainly debate how well such shocks are constructed and identified.  But the result is remarkably robust.  And then there is the data from surveys and micro price data, which it would take a fairly extreme position to reconcile with a flexible price spot market.

Fourth. The other thing that can generate shifts in the Philips Correlation is a shift in model parameters, including those that define the natural rates. This may be seen as symptomatic of the fact that the model is incomplete and wrong in some respect. But, to re-emphasise, if you subscribe either to sticky prices, wages or information sets, or all three, then whatever model you later find that explains this initial apparent parameter shift is going to have embedded within it a set of aggregate supply relations that we would recognise as a modern Philips Curve.

Fifth, one occasionally witnesses interlocutors sceptical of the PC conceding these points but nonetheless claiming that the PC or the natural rate of unemployment is ‘not a useful’ concept for policy, given uncertainties about its building blocks.  This argument makes little sense.  Policymakers have to take models and the uncertainty about their constituents as they find them and set policy accordingly.  If they see that the natural rate of unemployment is uncertain, then they consult the literature on monetary policy and natural rate uncertainty and act accordingly.  Optimal policy in the presence of such uncertainty is different from that in its absence.  But natural rate uncertainty doesn’t imply that one should simply drop the idea and look for a new policy ingredient.