Online Appendix: “Revisions in Utilization-Adjusted TFP Robust Identification of News Shocks”∗ Andr´e Kurmann

Eric Sims

Drexel University

University of Notre Dame & NBER

January 31, 2017

Abstract This online appendix provides supporting material and documentation for our paper titled “Revisions in Utilization-Adjusted TFP and Robust Identification of News Shocks.” It is not intended for publication.

∗ Kurmann:

Drexel University, LeBow College of Business, School of Economics, 3220 Market Street, Philadelphia, PA

19104 (email: [email protected]); Sims: University of Notre Dame, 710 Flanner Hall, Notre Dame, IN 46556 ([email protected]).

1

Introduction

This appendix provides additional details and supporting material for “Revisions in Utilization-Adjusted TFP and Robust Identification of News Shocks,” which can be found here. Section 2 provides details on the DSGE model used in the paper for Monte Carlo simulations. Section 3 shows impulse responses for an eight variable VAR with newer and more dated vintages of Fernald’s adjusted TFP series using the Barsky-Sims identification.

2

A Medium Scale DSGE Model

For the Monte Carlo experiments considered in the paper, we use as a laboratory a conventionally specified medium scale DSGE model. The model features sticky prices and wages, capital accumulation, habit formation in consumption, an investment adjustment cost, variable capital utilization, and a central bank which implements monetary policy according to a Taylor rule. The model is very similar to Christiano, Eichenbaum, and Evans (2005), Smets and Wouters (2007), and Justiniano, Primiceri, and Tambalotti (2010). As such, we only list the full set of equilibrium conditions here, rather than fully laying out the decision problems of each type of agent in the model. The full set of equilibrium conditions are listed below. A brief discussion of each expression follows. −1

λt = (Ct − bCt−1 )

  It κ λt = ξt Zt 1 − 2 It−1

f1,t = ϕt ψ

(2)

λt Rt = ξt (δ1 + δ2 (ut − 1))

(3)

ξt = βEt [λt+1 Rt+1 ut+1 + (1 − δ(ut+1 ))ξt+1 ]   2     It It It+1 It+1 − gI − κ − gI + βEt ξt+1 Zt+1 κ − gI It−1 It−1 It It

!w (1+χ) L1+χ t

wt#

f2,t = λt

(1)

λt = β(1 + it )Et λt+1 (1 + πt+1 )−1

wt# = wt

−1

− βbEt (Ct+1 − bCt )

wt wt#

+ βθw Et

!w Lt + βθw Et

w f1,t w − 1 f2,t # wt+1

!w (1+χ) 

!w 

wt#

(5) (6)

wt# # wt+1

(4)

(1 + πt )ζw 1 + πt+1

(1 + πt )ζw 1 + πt+1

−w (1+χ) f1,t+1

(7)

1−w f2,t+1

(8)

e tα−1 Nt1−α Rt = αmct At K

(9)

e α Nt−α wt = (1 − α)mct At K t

(10)

1 + πt# p x1,t = 1 + πt p − 1 x2,t

(11)

1

x1,t = λt mct Yt + βθp Et (1 + πt )−ζp p (1 + πt+1 )p x1,t+1

(12)

x2,t = λt Yt + βθp (1 + πt )ζp (1−p ) Et (1 + πt+1 )p −1 x2,t+1

(13)

e tα L1−α Yt vtp = At K − Xt F t

(14)

  p vtp = (1 + πt )p (1 − θp )(1 + πt# )−p + θp (1 + πt−1 )−p ζp vt−1

(15)

" Kt+1 = Zt

κ 1− 2



It It−1

2 # It + (1 − δ(ut ))Kt

− gI

δ(ut ) = δ0 + δ1 (ut − 1) +

δ2 (ut − 1)2 2

(16)

(17)

Yt = Ct + It

(18)

e t = ut Kt K

(19)

(1 + πt )1−p = (1 − θp )(1 + πt# )1−p + θp (1 + πt−1 )ζp (1−p )

(20)

wt1−w = (1 − θw )wt#,1−w + θw



(1 + πt−1 )ζw wt−1 1 + πt

1−w (21)

  it = (1 − ρi )i + ρi it−1 + (1 − ρi ) φπ (πt − πt∗ ) + φy (ln Yt − ln Yt−1 − ln gY )

(22)

ln Zt = ρZ ln Zt−1 + sZ εz,t

(23)

ln ϕt = ρϕ ln ϕt−1 + sϕ εϕ,t

(24)

ln St = ρS ln St−1 + sS εS,t

(25)

ln Γt − ln Γt−1 = (1 − ρΓ ) ln g + ρΓ (ln Γt−1 − ln Γt−2 ) + sg εg,t−q

(26)

µt = mc−1 t

(27)

At = St Γt

(28)

In these equations λt is the Lagrange multiplier on the flow budget constraint of a household and ξt is the Lagrange multiplier on the capital accumulation equation. Ct denotes consumption, Yt output, It investment, Kt physical capital, and Lt aggregate labor input. wt is the aggregate real wage, and Rt is the e t = ut Kt denoting capital services. real rental rate on capital services. ut denotes capital utilization, with K πt is the net inflation rate and it is the net nominal interest rate. vtp is a measure of price dispersion across firms. wt# is the reset real wage for a household given the opportunity to adjust its wage in a given period, while πt# is the reset inflation rate for firms given the opportunity to adjust their price. f1,t and f2,t are auxiliary variables related to optimal wage-setting, and x1,t and x2,t are auxiliary variables related to pricesetting. mct is real marginal cost, the inverse of which is the price markup, µt . St is a stationary technology shock, while Γt is a non-stationary technology shock. At is technology, which is the product of these two

2

terms. Zt is a shock to the marginal efficiency of investment and ϕt is a shock to the disutility from labor. Xt is a trend factor, to be discussed below. gI is the steady state gross growth rate of investment, and gY is the steady state gross growth rate of output. (1) defines λt , the shadow value on the flow budget constraint facing a household. The parameter b measures internal habit formation and β is a discount factor. (2) is the Euler equation for bonds, which prices the nominal interest rate, it . (3) is the first order condition for capital utilization. The cost of capital utilization is faster depreciation of capital, which is governed by (17). δ0 is the steady state depreciation rate, while δ1 and δ2 are parameters governing the linear and quadratic terms of the utilization adjustment cost function. The optimality condition for the choice of future capital is given by (4), while the first order condition for investment is given by (5). An investment adjustment cost is captured by the parameter κ. Each period, there is a 1 − θw probability that a household can adjust its wage. Optimal wage-setting for updating households is characterized by (6)-(8). w is a parameter governing how much market power a household has in setting its wage, χ is the inverse Frisch labor supply elasticity, and ψ is a fixed parameter governing the disutility from labor. ζw is measures how much non-updated wages may be indexed to lagged inflation. Cost-minimization by firms gives rise to factor demand curves for capital and labor in (9)-(10). α is the exponent on capital services in the production function, with 1 − α the exponent on labor. Because firms face the same factor prices, they all have the same real marginal cost and hire capital and labor in the same ratio. Each period, firms face a 1 − θp probability of being able to adjust their price. Optimal price-setting for updating firms is characterized by (11)-(13). p measures the extent of monopoly power in price-setting. ζp is a parameter measuring how much non-updated prices are indexed to lagged inflation. The aggregate production function is given by (14). F is a fixed cost of production, scaled by Xt , which measures the economy’s trend growth. vtp is a measure of price dispersion, the evolution of which is given by (15). Physical capital accumulates according to the law of motion given in (16). The aggregate resource constraint is given by (18). The evolution of inflation and the aggregate real wage are governed by (20) and (21), respectively. Monetary policy is governed by a Taylor rule, (22), which features interest rate smoothing governed by the parameter of ρi , a reaction to inflation, φπ , to deviations from a long run target, π ∗ , and a reaction to deviations of output growth from its steady state level, φy . The exogenous processes for the marginal efficiency of investment shock, the labor supply preference shock, and the stationary technology shock are given by (23)-(25). Each follows a stationary AR(1) process with steady state levels normalized to unity. The permanent productivity process is a stationary AR(1) in the growth rate and is given by (26). g denotes the steady state growth rate. The innovation is dated t − q, for q ≥ 0. q = 0 means that the technology improvement materializes immediately. q > 0 means that agents observe the shock before it impacts productivity. The price markup is defined as the inverse of real marginal cost in (27), and composite technology is the composite of the stationary and non-stationary terms, given in (28). Many of the variables in the model inherit the stochastic trend from Γt . It is straightforward to show 1

that the stochastic trend factor is Xt = Γt1−α . Re-scaling trending variables by this factor renders the model 3

stationary, and permits solution of the model using standard techniques. We solve the model via linearization about the non-stochastic steady state in the re-scaled variables. The parameters of the model are set to values listed in Table 1 below: Table 1: Calibrated Parameters Parameter

Value

Description

β α

0.99 1/3

Discount factor Capital’s share

δ0 δ1 δ2 ψ π∗

0.025 u∗ = 1 0.025 L∗ = 1/3 0

Steady state depreciation Utilization linear term Utilization squared term Labor disutility Steady state inflation

φπ φy ρi p w χ

1.5 0.3 0.85 11 11 1

TR inflation TR output growth TR smoothing Elasticity sub goods Elasticity sub labor Inverse Frisch

F θp θw ζp ζw b

Π∗ = 0 0.75 0.75 0.00 0.00 0.8

Fixed cost Calvo prices Calvo wages Price indexation Wage indexation Habit formation

κ g ρΓ ρA ρZ

1 1.0025 0.6 0.85 0.80

Investment adjustment cost SS growth of productivity AR productivity growth AR surprise productivity AR investment

ρϕ sg sA sZ sϕ

0.90 0.0031 0.0065 0.0121 0.0645

AR labor supply SD growth shock SD surprise productivity SD investment shock SD labor supply shock

These parameter values are largely drawn from the literature and are therefore quite standard. The frequency of time is a quarter. The discount factor is set to 0.99 and the parameter α in the production function is 1/3. The steady state depreciation rate is set to 0.025. The linear parameter in the utilization adjustment cost function is set to be consistent with a normalization of steady state utilization to unity. The coefficient on the squared term in the utilization cost term is set to δ2 = 0.025. The disutility of labor parameter is chosen to be consistent with steady state hours worked of 1/3. The parameters p and w are chosen to be consistent with steady state price and wage markups of 10 percent. The Calvo price and wage parameters are both set to 0.75, implying that the average duration between price or wage changes is four quarters. We assume no backward indexation of prices or wages to lagged inflation. The habit formation parameter is set to b = 0.8, and the investment adjustment cost parameter is set to κ = 1. We assume that there is no trend inflation, so π ∗ = 0. The Taylor rule features significant interest smoothing and strong 4

responses to both inflation and output growth. The steady state gross growth rate of productivity is set to 1.0025, which implies output growth of 1.5 percent per year in steady state. The fixed cost of production, F , is chosen so that profits are zero in steady state or the fixed cost is set to 0. We have less to guide us in terms of the parameterization of the shock processes. Our model features four stochastic shocks, which coincides with the number of variables included in our estimated VARs. We set the autoregressive parameters in the shock processes to conventional values. We then choose the standard deviations of the shocks to generate a standard deviation of output growth of 0.01, a contribution to the unconditional variance share of output growth of 0.50 owing to the investment shock (this is based on Justiniano et al. 2010), a contribution to the unconditional variance share of output growth due to the growth shock of 0.25, and a contribution to the unconditional variance share of output of the labor supply shock of 0.125 (which means that the surprise productivity shock accounts for the remaining 0.125 of the variance of output growth). These standard deviations are chosen for the specification in which q = 0 but differ little when we instead consider q = 1.

3

Barsky-Sims Identification with Larger VAR

In this appendix, we show results for the eight variable VAR considered in Barsky and Sims (2011) and document that the different vintages of adjusted TFP matter in meaningful ways. Like in the four variable VAR, there are important differences between the estimated responses depending on which vintage of adjusted TFP data one uses. With the 2007 vintage of data, output, investment, and hours all decline on impact in response to a favorable news shock. Inflation falls significantly, as does the Federal Funds Rate. With the 2016 vintage of data, output and investment increase on impact, and the impact decline in hours is significantly smaller than with the 2007 vintage of adjusted TFP. The impact declines in inflation and the Federal Funds Rate are smaller with the 2016 vintage of data compared to the 2007 vintage, and cease to be statistically significant.

5

Figure 1: Empirical Responses, BS Identification, Eight Variable VAR, 2007 vs. 2016 Vintage of Adjusted TFP

Consumer confidence percent

percent

Adjusted TFP 0.4 0.2 0 5

10

15

20

25

30

35

4 2 0

40

5

10

15

quarters

0.5 0 10

15

20

25

30

35

40

5

10

15

25

30

35

5

35

40

10

15

20

25

30

35

40

Inflation

Federal Funds rate

-0.4 15

30

quarters

-0.2 10

25

0.4 0.2 0 -0.2 -0.4

40

0

5

20

quarters

percent

percent

20

40

Hours percent

percent

Investment

15

35

quarters

2 1 0 -1 10

30

0.8 0.6 0.4 0.2 0

quarters

5

25

Gross domestic product percent

percent

Consumption 1

5

20

quarters

20

25

30

35

40

0

2016 TFP vintage 2007 TFP vintage

-0.2 -0.4 5

quarters

10

15

20

25

30

35

40

quarters

Notes: Solid black lines are the posterior median estimates from the VAR system estimated with the 2016 vintage of adjusted TFP. The gray bands correspond to the 16 to 84 percent posterior coverage intervals. The red dash-dotted lines are the posterior median estimates for the system estimated with the 2007 vintage of adjusted TFP. The red dashed lines correspond to the 16 to 84 percent posterior coverage intervals.

6

References Barsky, R. and E. Sims (2011). News shocks and business cycles. Journal of Monetary Economics 58 (3), 273–289. 3 Christiano, L. J., M. Eichenbaum, and C. L. Evans (2005). Nominal rigidities and the dynamic effects of a shock to monetary policy. Journal of Political Economy 113 (1), 1–45. 2 Justiniano, A., G. Primiceri, and A. Tambalotti (2010). Investment shocks and business cycles. Journal of Monetary Economics 57 (2), 132–145. 2, 2 Smets, F. and R. Wouters (2007). Shocks and frictions in US business cycles: A bayesian DSGE approach. American Economic Review 97 (3), 586–606. 2

7

Revisions in Utilization-Adjusted TFP Robust Identification of News ...

Jan 31, 2017 - *Kurmann: Drexel University, LeBow College of Business, School of ... 19104 (email: [email protected]); Sims: University of Notre ...

432KB Sizes 0 Downloads 260 Views

Recommend Documents

Revisions in Utilization-Adjusted TFP and Robust Identification of ...
Mar 7, 2017 - news shocks for business cycle fluctuations remains hotly debated, the main .... run output while deviations of output from trend are mostly driven by ..... In contrast, the 2014 and 2016 vintages of adjusted TFP growth are less ...

Revisions in Utilization-Adjusted TFP and Robust Identification of ...
Mar 7, 2017 - basic idea behind this estimation is that cost-minimizing firms simultaneously vary .... However, nothing about the illustration .... idiosyncratic measurement error is orthogonal to the news shock, making it irrelevant for the.

Robust Temporal Processing of News
Robust Temporal Processing of News ... measure) against hand-annotated data. ..... High Level Analysis of Errors ... ACM, Volume 26, Number 11, 1983.

Robust Detection and Identification of Sparse ...
are alternations of DNA of a genome that results in the cell having a less or more than .... can be regarded as continuous when coverage of the genome is sufficiently high, for .... and is compared with the performances of LRS and CBS.

Efficient and Robust Music Identification with Weighted Finite-State ...
of Mathematical Sciences, New York, NY USA, and Google Inc. e-mail: {mohri ... tion II presents an overview of our music identification ap- ...... he worked for about ten years at AT&T Labs - ... M.Eng. and B.S. degree in Computer Science from.

Efficient and Robust Music Identification with Weighted Finite-State ...
be used to give a compact representation of all song snippets for a large .... song during the course of three iterations of acoustic model training. mpx stands for ...... transducers in speech recognition,” Computer Speech and Language, vol.

Efficient and Robust Music Identification with Weighted Finite-State ...
a database of over 15 000 songs while achieving an identification ... our system is robust to several different types of noise and ..... can be distributed along a path. ...... [9] D. Pye, “Content-based methods for the management of digital music,

Efficient and Robust Music Identification with Weighted Finite-State ...
that it helps speed up the construction of the weighted suffix automaton in our task by a ... In a practical setting, the test recording provided to a music identification ...... to be r (line 19) and that of r is set to be s[q′] (18), and l[r] set

Efficient and Robust Music Identification with Weighted Finite-State ...
large-scale music identification system can be constructed for a database of over 15 .... In a second step we apply ordinary k-means clustering to refine the clusters ... 3) Measuring Convergence: In speech recognition, each utterance is usually ...

ROBUST COMPUTATION OF AGGREGATES IN ...
aggregate statistics (aggregates) amid a group of sensor nodes[18,. 29, 14]. .... and broadcasts a group call message, GCM ≡ (groupid = i), to all its neighbors. 1.3 The ...... [15] J. M. Hellerstein, P. J. Haas, and H. J. Wang, ”Online Aggregati

Dress Code Revisions table format.pdf
student is attired in a manner which is likely to cause disruption or interference with the operation of the 3 ... logos that depict death or violence; gang related.

Approved ACE Revisions 2015.pdf
There was a problem previewing this document. Retrying... Download. Connect more ... Approved ACE Revisions 2015.pdf. Approved ACE Revisions 2015.pdf.

Identification of novel proteins affected by rotenone in mitochondria of ...
Aug 16, 2007 - Email: Jinghua Jin - [email protected]; Jeanne Davis ..... ASAP Ratio: Automated Statistical Analysis of Protein abundance ratio. The results are ...

Summary of Revisions for the 2008 Clinical Practice ...
and under 60 years of age. (E). Diabetes care: ○ Components of the comprehensive di- abetes evaluation revised. ○ Continuous glucose monitoring may be.

Robust Estimation of Edge Density in Blurred Images
Electronics and Telecommunications Research Institute (ETRI). Daejeon 305-700, Korea. {jylee, ywp}@etri.re.kr. Abstract—Edge is an ... of Knowledge and Economy (MKE) and the Korea Evaluation Institute of. Industrial Technology (KEIT). (The Developm

Identification of PTSD in Cancer Survivors
definition, patients do not meet criteria for the disorder if the trauma identified is medical ill- ness. ... PSYCHOSOMATICS descriptive, ... predictive factors of PTSD. There is also evidence that when a traumatic event is particularly ..... analysi

Identification and in vitro production of Lactobacillus ...
Mar 5, 2010 - program was 95°C for 2 min, 35 cycles of 95°C for 30 s,. 55°C for 1 ..... that the Lactobacillus species is also an important factor since women ...

short report: identification in triatomine vectors of ...
mained undetermined (10.5%) and 11 (6.4%) showed a com- plex pattern that could be explained by multiple blood sources and sequencing reactions were unreadable. Figure 1 shows the different HDA patterns observed among T. longi- pennis blood meal samp

identification of pacemaking region in zebrafish ... - Semantic Scholar
In previous work, voltage dynamics have been visualized and activation maps have been manually drawn based on the visualization [8, 29, 47]. Manual methods are ... How do we perform computer assisted manual identification on optical mapping data? ...