Perceived productivity and the natural rate of interest Gianni Amisano and Oreste Tristani European Central Bank IRF 28 Frankfurt, 26 June Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June / 22
In ation and shifts in productivity growth The US experienced a productivity slowdown in the 97s and an acceleration in the past decade. Potential link with in ationary performance (Orphanides, 23). Edge, Laubach and Williams (25): How should policy respond to (imperfectly perceived) shifts in technology growth? We revisit the question within an empirical model of the euro area...... from the viewpoint of the natural rate of interest Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 2 / 22
Why is this relevant for the euro area? Labour productivity growth Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 3 / 22
Our results so far Productivity shocks are important drivers of EU in ation in the medium term Compared to the benchmark of a policy rule "tracking" the natural rate of interest, policy was too lax in the 98s and early 99s; OK, thereafter Key issue: natural rate estimates tend to be volatile in DSGE models. Our model is rigged towards producing smoother estimates Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 4 / 22
Euro area real interest rates Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 5 / 22
Outline Related literature Features of the model that shape our results: predetermination of decisions measurement errors persistence of the shocks imperfectly observed productivity shocks The role of productivity shocks Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 6 / 22
Related literature On monetary policy response to shifts in productivity growth: Edge, Laubach and Williams (25). In ation and productivity growth Galí, López-Salido and Vallés (23). We rely on a small model estimated using full information methods. Larger models Empirical estimates of the natural rate: Laubach and Williams (23) (for euro area Garnier and Wilhelmsen, 25; Benati and Vitale, 27) in a semi-structural approach Conditional dynamics of microfounded notion: Neiss and Nelson (23); Andrés, López-Salido and Nelson (28) Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 7 / 22
Model features: () predetermination As in Rotemberg and Woodford (997), Christiano, Eichenbaum and Evans (25), households decisions are determined d periods in advance h(c s (i) hc s (i)) e v (N s (i)) i γ E t d s=t β s t γ and similarly for rms pricing decisions. This generates transmission lags. Tracking the natural rate quarter-by-quarter is pointless. Expected natural rate d quarters ahead should be tracked (Woodford, 23). We extend the Giannoni and Woodford (25) framework to a growing economy with imperfectly perceived technology shocks (but without Calvo wages) Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 8 / 22
Predetermination: why? Transmission lags are arguably realistic. Milton Friedman. Applied frameworks such as Svensson (997) Predetermination represents a possible general equilibrium explanation Agents presumably lter out shocks from the observation of actual variables. Aggregate variables published with a lag: HICP in ation rate around 2 days after the reference month; GDP gures in the second quarter following the reference period. A general predetermination of 2 quarters is a rough way of capturing these lags. Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 9 / 22
2-period predetermination When γ = the model can be written as E t 2 ˇx t = E t 2 ˇx t+ eσe t 2 b i t bπ t+ br t n ˇπ t = βe t 2 ˇπ t+ + κ φe! t 2 x t + eσ E t 2 ˇx t where ˇx t = E t 2 ˇx t + (ğ t E t 2 ğ t ) (ˇy t n E t 2 ˇy t n ) ˇx t Ξ (Ξ! x t h! x t ) βh (ΞE! t 2 x t+ h! x t ) ˇπ t bπ t ιbπ t ( ι) E t 2 bπ t Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June / 22
Predetermination Given the IS equation E t 2 ˇx t = E t 2 ˇx t+ eσe t 2 b i t bπ t+ br n t a policy such that E t 2 bi t = E t 2 br t n + E t 2 bπ t+ bi t = E t 2 bi t can ensure that the output gap measure E t times 2 ˇx t is zero at all Because of the expectations term, E t 2 br n t smoother than br n t. Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June / 22
Predetermination In the Phillips curve, both! x t and ˇx t drive in ation ˇπ t = βe t 2 ˇπ t+ + κ φe! t 2 x t + eσ E t 2 ˇx t ˇx t Ξ (Ξ! x t h! x t ) βh (ΞE t 2! x t+ h! x t ) Tracking the natural rate can ensure that E t 2 ˇx t = at all times, but it does not imply E! t 2 x t =. A combined e ect of habits and predetermination Tracking the natural rate reduces, but does not eliminate, in ationary pressures. Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 2 / 22
Impulse responses to a persistent technology growth shock. Estimated rule.8 Natural rate rule.6 Impulse responses to a temporary technology growth shock..2 %.4.2 %.3.4 Estimated rule Natural rate rule.5.2 5 5 2.6 5 5 2 Impulse responses to a government spending shock.4 Estimated rule Natural rate rule.3.2 %.. 5 5 2
Model features: (2) measurement errors No structural shock can a ect in ation on impact. How do we account for in ation surprises? We add measurement errors to all observation equations, including the in ation rate Advantage: the identi cation of measurement errors is easier (by construction). Any high-frequency volatility in in ation must be erratic Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 4 / 22
.3 Quarterly inflation rate Model based inflation.25.2.5..5 98 985 99 995 2 25 2
Model features: (3) imperfectly perceived shocks Shifts in trend productivity growth tend to unfold slowly over time. The ongoing uncertainty about future productivity trends is an important element to take into account when evaluating monetary policy responses to productivity developments We assume that observed productivity growth is the combination of two shocks, one temporary and one highly persistent As in Edge, Laubach and Williams (25, 27), all agents in our model use a (steady state) Kalman lter to discriminate the two productivity growth shocks Filtering appears to improve the overall empirical performance of the model Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 6 / 22
Model features: (4) shocks persistence The means of the prior distributions of persistent technology growth and in ation target processes are set at high values; their standard deviations are very small: Mean StDev ρ ξ.98. ρ g.5.7 ρ π.98. σ ξ.3.3 σ a.5.35 σ g.5.35 σ π.3.3 Movements in these processes tend to be very smooth over time Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 7 / 22
Empirical results Estimation on quarterly euro area data over the 98Q-27Q2 sample. 3 variables: rate of growth of output per capita, the 3-month nominal interest rate and the GDP de ator Measures of t: priors and posterior distributions empirical cross-covariances Variance decomposition Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 8 / 22
Policy rule For estimation purposes, the system is closed with a simple monetary policy rule of the form i t = const. + ψ π! π t π t + ψy y t ξ + ρ I i t where! π t = (π t + π t + π t 2 + π t 3 ) /4 and π t = ( ρ π ) π + ρ π π t + ε π t is the central bank s in ation objective, which is assumed to vary stochastically around a long-run level π. Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 9 / 22
4 β 4 ψ π 2 ψ y 4 ρ i γ 2 2 2 5.95.5.5 2 2 2 h 5 φ ζ ι 5 5.5 2.5.5 2 Ξ 2 Π *.2.4 ρ ξ.2.4 ρ g ρ π * σ ξ σ a 4 4 4 4 4 2 2 2 2 2.8.9 σ g 4.5 σ * π 4.8.9 σ i 2 4 σ π x 3 4.2.4 σ y 4 2 2 5 2 2.2.4 2 4 x 3 2 4 x 3 2 4 x 3 2 4 x 3
i.8.8.8.8.8.6.6.6.6.6.4.4.4.4.4.2.2.2.2.2 2 3 2 3 2 3 2 3 2 3.8.8.8.8.8.6.6.6.6.6 π.4.4.4.4.4.2.2.2.2.2 2 3 2 3 2 3 2 3 2 3.8.8.8.8.8 y.6.4.6.4.6.4.6.4.6.4.2.2.2.2.2 2 3 ε g 2 3 ε * π 2 3 2 3 ε a 2 3 ε ξ meas.err.
9 x 5 i 6.5 x 5 π x 5 y r 8 7 6 r 6 5.5 5 4.5 r.5.5 5 5 5 4 5 5 5 5 7 x 5 i 7 x 5 π 4 x 6 y 6 6 2 5 5 pi 4 pi 4 pi 2 3 3 4 2 5 5 2 5 5 6 5 5 Dy 2 x 6 4 6 8 i Dy 2 x 6 3 4 5 6 π Dy 3 x 5 2 lower bound median upper bound sample 2 5 5 7 5 5 5 5
In ation Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 2 / 22
Conclusions Our model supports the hypothesis that shifts in trend productivity growth can explain a sizable portion of observed euro area in ation Our microfounded estimate of the natural rate of interest suggest that in recent years monetary policy behaved consistent with the objective of eliminating in ationary pressures from the economy to the largest possible extent The assumption of 2-quarters predetermination in private sector decisions is an important determinant of our results. While the overall empirical performance of our model is reasonably good, some robustness checks are needed.eg. d 6= 2, di erent across rms and households; less dogmatic prior distributions Amisano-Tristani (European Central Bank) Productivity and the natural rate IRF 28 Frankfurt, 26 June 22 / 22