Tuesday, December 29, 2015

NAWRU IV The Long Term

Guest post by Marco Fioramanti explaining how the EC "reconciles" the NAWRU Phillips curve model with completely implausible long term implications with their long term forecasts. Brief summary -- they don't take the model seriously.

by Marco Fioramanti

At this link you can find an EC's WP in which they distinguish between NAWRU and Structural Unemployment.

  They say (pg 1 link above):

“(...) Importantly, this method [the Unobserved Component Model for NAWRU - mf] provides only a proxy for structural unemployment that might not remove fully the impact of all temporary shocks. In particular, persistent shocks are likely to contaminate the trend.”

  Macroeconomic projections for the forecast and the fiscal stuff in EC works this way:

  1. Country desks do the forecast up to t+1/t+2 depending on the season (t+1 in spring, t+2 in autumn).

2. The service in charge of the estimation of potential output and output gap takes the numbers from (1) applies the methodology (production function approach called PF below) and go to t+5. In doing this a set of rules (on top of the PF approach) is used to extend the projection horizon, one of which is that the output gap closes in t+5. These estimations are at the core of the SGP.

3. EC also has long term projections for long term sustainability of public finance (pension, health care etcAgeing-related expenditure) which goes from t+10 to t+50. This is a simplified PF approach. To link t+5 to t+10 they have introduced a modified version of PF(t+5) methodology in which they anchor NAWRU to converge to structural unemployment in t+10 (for Italy structural unemployment is 9.4% according to EC’s latest estimate).

They say (pg. 39 in this WP ):

            “The NAWRU framework incorporates economic rationale into the T+10 NAWRU forecast, relying on a set of four labour market economic indicators (i.e. unemployment benefit replacement ratio; tax wedge; active labour market policies; & union density) and a set of macro control variables (i.e. TFP; real interest rate; employment in construction; & the T+5 NAWRU) to guide the forecast beyond T+5. This approach allows for a decomposition of the NAWRU into structural drivers and medium term cycles & for a prediction between T+6 and T+10 which reverts the NAWRU back towards the long run structural unemployment rate. A simple convergence rule towards the T+10 NAWRU is applied & a so-called prudent rule is built into the approach in order to override the calculations in cases where the T+10 NAWRU forecast is deemed to be surrounded by a relatively high degree of uncertainty for a particular country.”

So, while EC recognizes the NAWRU is too cyclical, they keep using it.

Fioramanti, Padrini and Pollastri (2015) propose (among others methodologies) using structural unemployment instead of NAWRU in the PF approach also for the t+5 projections - so the estimate of the natural rate of unemployment used in calculating the output gap would be the natural rate of unemployment as originally defined by Friedman. Endnote: In contrast to the t+5 methodology (for which you can get all the necessary files and instructions to replicate EC’s results) the t+10 methodology is not fully public.

 -- Marco Fioramanti

Comment by Robert Waldmann

It is very important that the EC doesn't take the I(2) NAWRU model seriously.  They have a model used for long term forecasts and another model used to calculate the NAWRU and cyclical unemployment = unemployment - NAWRU.  The NAWRU is used only to calculate the output gap and allowed deficits under the stability and growth pact (SGP). The bad long term implications of policies with good effects in the short term is the justification for the SGP *and* the claim that the concept of the natural rate of unemployment has policy relevance.

The estimates of cyclical unemployment include a series of arbitrary fudge factors NAWRU minus structural unemployment called "medium term cycles". There is no explanation of the cause of "medium term cycles".  The model of "medium term cycles" is not a model of "cycles" as the word is used in macroeconomics, because NAWRU-structural unemployment is not stationary.

It is hard to justify the use of a model (the NAWRU accelerationist Phillips curve discussed in these posts) with absurd long term implications which are ignored when making long term forecasts in the application of the SGP. I think there are two possible explanations of this modelling strategy. First it is possible that the technicians were instructed to make the accelerationist Phillips curve work and overcome all challenges from the data. Second it is possible that DG EcFin is willing to allow deficts over 3% of GDP because of cyclical corrections but only so long as the cyclical corrections are small. This corresponds to dividing the cycle into a "medium term cycle" with huge variance which has no role in a short term cycle with small variance (based on which Treasuries are given some slack) .

In any case, I certainly agree with Fioramanti, Padrini and Pollastri's proposal to use unemployment minus structural unemployment when estimating output gaps.

No comments: