When estimating the NAWRU the European Commission staff imposes the assumption that it shifts exogenously. The assumption is that the NAWRU is I(2) that is that the first difference of the first difference of the NAWRU is stationary. This implies that the NAWRU will almost certainly *not* remain in the interval from 0% to 100%. This makes no sense, but, in this post I will focus on the assumption that the disturbance terms which shift the level and the trend of the NAWRU are independent of all other variables. This means that the NAWRU is assumed to be genuinely exogenous -- not just exogenous to the model but exogenous to the economy, and, in particular, not affected by monetary or aggregate fiscal policy. This means that when unemployment is decomposed by total unemployment = NAWRU + cyclical unemployment, the conditional distrubtion of the NAWRU depends on past NAWRU but is independent of current and past cyclical unemployment.
The assumption that these shifts are exogenous is extremely important. An alternative view (discussed in 1960 in the second major article on the Phillips curve) is that structural unemployment is not exogenous because, with time, cyclical unemployment can become structural. This possibility was discussed and named "hysteresis" by Blanchard and Summers in 1986. They and Eugenio Cerutti have returned to the topic and presented massive evidence that it is important in (pdf warning) 2015.
The assumption that the NAWRU is not affected by aggregate demand is not easily reconciled with the fact that, in the 21st century, the estimate NAWRU closely tracks total unemployment. This must be true of estimated NAWRUs as inflation has remained stable in Eurozone countries (warning same pdf to which I linked in an earlier post).
The hypothesis that the NAWRU is exogenous can be tested by embedding the model used by the Commission in a more flexible model. One possibility (actually explored by the commissions DG EcFin) is to allow the acceleration of wage inflation to depend on lagged shifts in total unemployment. I have estimated both the official model and this more flexible model by maximum likelihood. My impression (that is the output of a program I wrote but don't swear by) is that the null of the official model is strongly rejected against the alternative. In any case, Blanchard, Cerutti and Summers present strong evidence that recessions often appear to have long lasting effects. The general pattern of European unemployment and wage inflation from 1980 on is of huge long lasting shifts in unemployment and a single huge decline in wage inflation in the 80s.
The assumption that the NAWRU is exogenous and must simply be accepted by fiscal authorities is critically important. If austerity causes an increase in the NAWRU, it can be self defeating. There is now overwhelming evidence that, especially when the safe short term interest rate is near zero, reduced government spending causes lower GDP and higher unemployment (pdf warning). If cyclical unemployment becomes structural, this implies a long lasting reduction in revenues and a long lasting increase in social insurance transfers. As noted by DeLong and Summers (pdf warning), this can imply that reduced government spending causes a higher long run debt to GDP ratio. The technical assumption that the NAWRU is exogenous has fundamentally important policy implications. It can be tested.
The econometricians at the Commissions DG EcFin are placed in a difficult position. They are required to look at data to apply the stability and growth pact, but they are not allowed to estimate parameters which suggest that the stability and growth pact causes instability and stagnation.