Really Uncertain Business Cycles
Nick Bloom (Stanford, CEP & NBER)Max Floetotto (Stanford)Nir Jaimovich (Stanford & NBER) Very Preliminary
Kellogg, May 19th 2009
Policy makers believe that uncertainty matters, 1/4
FOMC (April 2008)“Several participants reported that uncertainty about the economic outlook was leading firms to defer spending projects until prospects for economic activity became clearer.”
Policy makers believe that uncertainty matters, 2/4
Olivier Blanchard (January 2009)"Crises feed uncertainty. And uncertainty affects behavior, which feeds the crisis. Were a magic wand to remove uncertainty, the next few quarters would still be tough, but the crisis would largely go away.”
Policy makers believe that uncertainty matters, 3/4
Larry Summers (March 2009)“…unresolved uncertainty can be a major inhibitor of investment. If energy prices will trend higher, you invest one way; if energy prices will be lower, you invest a different way. But if you don’t know what prices will do, often you do not invest at all.”
Policy makers believe that uncertainty matters, 4/4
Christina Romer (April 2009)“Volatility has been over five times as high over the past six months as it was in the first half of 2007. The resulting uncertainty has almost surely contributed to a decline in spending.”
We model uncertainty as a new type of shock
First moment shocks in the literature● Neutral technology shocks● Investment-specific technology shocks● Oil price shocks● Labor supply shocks● Monetary policy shocks● Financial shocks● News shocks
We want to consider a second moment (uncertainty) shock● For simplicity focus on technology shocks
The paper examines empirical evidence and a model on uncertainty and the business cycle
Paper has three main parts:
● Empirical evidence that uncertainty is counter-cyclical
● DSGE model of the impact of time varying uncertainty:
● Uncertainty shocks lead to business cycles
● Uncertainty substantially reduces the impact of policy
● Examine Census micro-data, to investigate further predictions we get from the model for the impact of uncertainty
What this paper does not (currently) do
● Attempt to endogenize uncertainty
● Modeled as exogenous, like first moment shocks
● If uncertainty is endogenous could think of as a propagation and amplification mechanism
● Include/analyze other potentially important uncertainty channels:
● Consumer durables
● Credit
● Risk
Measuring Uncertainty
Model
Testing the model on Census micro data
Uncertainty over the business cycle
● Uncertainty is hard to measure and the concept is vague
● Build on prior literature and use different types of proxies:
● Cross-industry, firm and plant evidence
● Time-series aggregate data
● Cross-forecaster disagreement evidence
● Combine these into an aggregate uncertainty index
● This reduced index rises by 48% during recessions
.01
.02
.03
.04
.05
.06
1970 1980 1990 2000 2010Year
Cross industry output growth spread
Inter-quartile range of the 3-month growth rates of industrial production. Covers all 196 manufacturing NAICS sectors in the Federal Reserve Board database.
until 2009 Q1
-.4
-.2
0.2
1970 1980 1990 2000 2010Year
1st, 5th, 10th, 25th, 50th, 75th, 90th, 95th and 99th percentiles of 3-month growth rates of industrial production within each quarter. All 196 manufacturing NAICS sectors in the Federal Reserve Board database.
99th percentile,2.2% higher in recessions
1st percentile7.4% lower in recessions
50th percentile,1.3% lower in recessions
Cross industry output growth distribution
.2.3
.4.5
1970 1980 1990 2000 2010Year
Cross firm sales growth spread
Interquartile range of sales growth (Compustat firms). Only firms with 25+ years of accounts, and quarters with 500+ observations. SIC2 only cells with 25+ obs.
Across all firms(+ symbol)
Across firms in a SIC2 industry
until 2008 Q2
.05
.1.1
5.2
1970 1980 1990 2000 2010Year
Cross firm stock returns spread
Interquartile range of stock returns (CRSP firms). Only firms with 25+ years of accounts, and quarters with 500+ observations. SIC2 only cells with 25+ obs.
Across all firms(+ symbol)
Across firms in a SIC2 industry
until 2008 Q4
Cross establishment sales & labor productivity growth spread
ASM data on 60,000 manufacturing establishments 1974-2006
Found two stylized facts:
• Cross-sectional spreads strongly counter cyclical
• Increase both overall and within SIC 4-digit category
.005
.01
.015
.02
.025
For
eca
st s
tan
dard
-dev
iatio
n, fr
om
AR
CH
mo
del
1965 1970 1975 1980 1985 1990 1995 2000 2005Year
Industrial production growth volatility
Monthly industrial production conditional heteroskedasticity, from a GARCH(1,1) auto-regression with 12 lags.
until 2009 Q1
1020
3040
5060
1965 1970 1975 1980 1985 1990 1995 2000 2005Year
Stock market volatility
S&P 100 implied volatility (the VXO, which is very similar to VIX) from 1987, and normalized realized volatility of actual S&P100 daily stock returns prior to 1986.
until 2009 Q2
0.0
5.1
.15
.2IQ
R o
f une
mp
gro
wth
rat
e fo
reca
st
1970 1975 1980 1985 1990 1995 2000 2005Year
Forecaster dispersion for unemployment
Interquartile range of year ahead unemployment rates / mean unemployment rates. From Survey of Professional Forecasters. Average of 41 forecasts per quarter.
until 2009 Q2
0.0
2.0
4.0
6
1970 1975 1980 1985 1990 1995 2000 2005Year
Forecaster dispersion for productionuntil 2009 Q2
Interquartile range of year ahead production/mean production. From Survey of Professional Forecasters. Average of 41 forecasts per quarter.
.51
1.5
22.
5
1965 1970 1975 1980 1985 1990 1995 2000 2005Year
Uncertainty index – average of last 7 measuresuntil 2009 Q1
Mean of the 7 prior indicators after they have all been normalized to an average of 1 during non-recessionary quarters. Only reported when 5+ indicators present.
1968:41969:11969:2
1969:31969:4
1970:1
1970:2
1970:31970:4
1971:11971:2 1971:3 1971:4
1972:11972:2 1972:31972:41973:1
1973:2
1973:3
1973:4
1974:11974:2
1974:3
1974:4
1975:1
1975:2
1975:31975:4
1976:11976:21976:31976:4
1977:11977:21977:31977:4
1978:11978:21978:3
1978:41979:1
1979:2
1979:31979:4
1980:11980:2
1980:31980:4
1981:1
1981:2
1981:3
1981:41982:1
1982:2
1982:3
1982:4
1983:1
1983:21983:3
1983:41984:11984:2
1984:31984:4
1985:11985:2
1985:3
1985:4
1986:11986:2
1986:31986:41987:1
1987:2
1987:3
1987:41988:1
1988:2
1988:31988:41989:1
1989:21989:3
1989:41990:11990:2
1990:31990:4
1991:1
1991:21991:3
1991:4 1992:1
1992:21992:31992:41993:11993:2
1993:31993:4
1994:11994:2
1994:31994:41995:11995:21995:31995:4
1996:11996:21996:31996:41997:11997:21997:3
1997:41998:1
1998:21998:3
1998:41999:11999:21999:31999:42000:1
2000:2
2000:3
2000:4
2001:12001:22001:3
2001:4
2002:12002:22002:32002:4
2003:12003:22003:3
2003:42004:1
2004:22004:3
2004:42005:12005:2
2005:3
2005:42006:12006:22006:32006:42007:12007:2
2007:32007:4
2008:12008:2
2008:3
2008:4
2009:1
.51
1.5
22.
5U
nce
rtai
nty
inde
x
-10 -5 0 5Quarterly industrial production growth (annualized in %)
Uncertainty index and industrial production growth
Are recessions also conditionally associated with recessions?
● So far only shown unconditional correlation between recessions and uncertainty
● Use VAR analysis to investigate the conditional correlation of uncertainty with a recession, noting this does not imply causality
VAR analysisUse standard VAR framework from Christiano, Eichenbaum and Evans (2005) that includes the following variables (in order):
● Real GDP (logs)● Real consumption (logs)● GDP deflator (logs)● Real investment (logs)● Real wage (logs)● Labor productivity (logs)● Federal Funds rate● Real profits (logs)● Growth rate of M2
Add aggregate uncertainty index, but ● Check robustness to change in ordering (first, last)● TFP to control for first moment shock (Basu, Kimball & Fernald)
VAR analysis – uncertainty firstShock calibrated to increase uncertainty 48% during recessions
Cholesky orthogonalized on quarterly data from 1968:4 to 2006:4 using 4 lags. Dotted lines are 95% confidence intervals
VAR analysis – different experiments
Cholesky orthogonalized on quarterly data from 1968:4 to 2006:4 using 4 lags. Dotted lines are 95% confidence intervals
Shock calibrated to increase uncertainty 48% during recessions
Results for Germany (from Frank Smets)
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
Response of UNCERTAINTY to UNCERTAINTY
-.010
-.005
.000
.005
.010
.015
1 2 3 4 5 6 7 8 9 10
Response of TFP to UNCERTAINTY
-0.8
-0.4
0.0
0.4
0.8
1.2
1 2 3 4 5 6 7 8 9 10
Response of IR3M to UNCERTAINTY
-.8
-.4
.0
.4
1 2 3 4 5 6 7 8 9 10
Response of CPI to UNCERTAINTY
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
Response of WAGE to UNCERTAINTY
-.02
-.01
.00
.01
.02
1 2 3 4 5 6 7 8 9 10
Response of GDP to UNCERTAINTY
Response to Cholesky One S.D. Innovations ± 2 S.E.
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
Response of UNCERTAINTY to UNCERTAINTY
-.010
-.005
.000
.005
.010
.015
1 2 3 4 5 6 7 8 9 10
Response of TFP to UNCERTAINTY
-0.8
-0.4
0.0
0.4
0.8
1.2
1 2 3 4 5 6 7 8 9 10
Response of IR3M to UNCERTAINTY
-.8
-.4
.0
.4
1 2 3 4 5 6 7 8 9 10
Response of CPI to UNCERTAINTY
-1.0
-0.5
0.0
0.5
1.0
1 2 3 4 5 6 7 8 9 10
Response of WAGE to UNCERTAINTY
-.02
-.01
.00
.01
.02
1 2 3 4 5 6 7 8 9 10
Response of GDP to UNCERTAINTY
Response to Cholesky One S.D. Innovations ± 2 S.E.
Impact of a one SD impulse in uncertainty. Prepared by creating a German Uncertainty Index over 10 years and running the same VAR specification.
Results for US consumption (from Mark Doms)
Source: Mark Doms (SF Federal Reserve Board), figure used for Board of Governors briefing work
Taking stock
● Uncertainty - however measured - is strongly countercyclical
● An increase in uncertainty robustly associated with a significant drop and rebound in output in a VAR framework
● Well known identification problems in VAR, so results are only suggestive
● Model allows us to study a possible mechanism further and provides additional micro-predictions to test in Census data
Measuring Uncertainty
Model
Testing the model on Census micro data
Model conforms as much as possible to the standard frictionless RBC
● Main deviations are:
● Second moment shocks
● Non-convex adjustment costs in both capital and labor
● Firm-level heterogeneity
Mechanism is linked to Ss investment thresholds arising from non-convex adjustment costs
Disinvest Invest
Productivity / Capital
Den
sity
of
un
its
Disinvest Invest
Productivity / Capital
Den
sity
of
un
its
Mechanism is linked to Ss investment thresholds arising from non-convex adjustment costs
Technology
● Large number of heterogeneous firms
● “Productivity” follows an AR process with time variation in the variance of innovations
● Uncertainty (σA and σZ) follow a 2-point markov chain
Capital and labor adjustment costs● Capital and labor follow the laws of motion:
where i: investment δk: depreciation
s: hiring δn: attrition
● Allow for the full range of adjustment costs
● Fixed – lump sum cost for investment and/or hiring
● Partial – per $ disinvestment and/or per worker hired/fired
● Quadratic – to invest/disinvest and/or hire/fire more rapidly
● To match micro data paid on all investment and hiring (even replacement investment and hiring)
Firm’s value function
Households
● Representative agent who works, consumes and owns the firms
● We assume the functional form for household utility
● Separability of preferences yields a simple SDF:
● The FOC for hours worked
General equilibrium solution overview
● We have a recursive competitive equilibrium
● Solve numerically as no analytical solution
● Numerical solution approximates μ (the firm-level distribution over z, k and n) with moments, building on Krusell and Smith (1998)
● Follow Kahn and Thomas (2008) and Bachman, Caballero and Engel (2008) in using two tricks to simplify the numerical solution
Simplifying the problem
Calibration
Calibration of the uncertainty process
Simulation of a shock to uncertainty
0%
20%
40%
60%
80%
100%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
Share of economies in high uncertainty state (in 1000 simulations)
Results not driven by a first moment shock
Average firm times macro productivity in the simulation
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
3%
4%
5%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
The effect of an increase in uncertainty on employment: 3 phases
-3.0%
-2.5%
-2.0%
-1.5%
-1.0%
-0.5%
0.0%
0.5%
1.0%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
Drop Rebound
Overshoot
Deviation from steady state (%)
The effect of an increase in uncertainty on investment
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5%
10%
15%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
Deviation from steady state (%)
The effect of an increase in uncertainty on output
-2.5%
-2.0%
-1.5%
-1.0%
-0.5%
0.0%
0.5%
1.0%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
Deviation from steady state (%)
The effect of uncertainty on measured TFP
-0.8%
-0.6%
-0.4%
-0.2%
0.0%
0.2%
0.4%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
Deviation from steady state (%)Measured TFP = output/(capitalαlaborν)
-3.0%
-2.0%
-1.0%
0.0%
1.0%
2.0%
3.0%
-4 -2 0 2 4 6 8 10 12 14 16
Quarter
Bad fit? The effect of uncertainty on consumption
Deviation from steady state (%)
0.80
0.85
0.90
0.95
1.00
1.05
1.10
1.15
-4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Quarter
Investment threshold
Disinvestment threshold
90th
10th
50th
Cross-sectional distribution of firm TFP/capital
Thresholds & percentiles of firm distribution over z/k (for fixed k & l)
Uncertainty alters the impact of policy
● Uncertainty widens firms’ Ss bands for investment and hiring, thereby reducing the impact response of any given stimulus
● But, once uncertainty falls firms will start to respond again
Illustrate with an investment credit from Mars
● Example of a 1% investment credit from Mars for 3 quarters
● From Mars so not GE (much simpler to model)
● Again for simplicity assume it’s a complete surprise to agents – they just find investment is 1% cheaper for 3 quarters
● Evaluate during a normal period and after an uncertainty shock
-2.5%
-2.0%
-1.5%
-1.0%
-0.5%
0.0%
0.5%
1.0%
1.5%
-2 0 2 4 6 8 10 12
Quarter
Impact of the 1% investment tax credit
(2) Low uncertainty + investment credit
(1) Low uncertainty
(4) Uncertainty shock + investment credit
(3) Uncertainty shock
Quarter
Out
put
(% d
evia
tion
from
low
-unc
erta
inty
sta
te)
-0.1%
0.0%
0.1%
0.2%
0.3%
0.4%
0.5%
0.6%
0.7%
0.8%
0.9%
1.0%
-2 0 2 4 6 8 10 12
Quarter
Uncertainty reduces and delays the impact
(2) – (1): Investment credit impact in low uncertainty
(4) – (3): Investment credit after uncertainty shock
Quarter
Out
put
(% d
evia
tion
from
low
-unc
erta
inty
sta
te)
Implications for policy impact of an uncertainty shock
● Suggests that stabilization policy to address the impact of an uncertainty shock would ideally be relatively:
● Rapid – to minimize creating additional policy uncertainty
● Large – you need a big stimulus given low responsiveness
● Temporary – want to avoid overshoot once uncertainty falls
Measuring Uncertainty
Model
Testing the model on Census micro data
Reduced response in periods of high uncertainty
High uncertainty (recession)
Low uncertainty (boom)
ΔZ / Z
ΔL
/ L
Use micro data to test differences in establishment response to
TFP during periods of low and high uncertainty
Conclusions and next steps
● Uncertainty appears strongly counter cyclical
● Realistically calibrated DSGE model shows:
● Uncertainty can lead to moderate business cycle fluctuations in output, investment, hiring and TFP growth
● Suggests micro rigidities are important
● Policy impact different at high uncertainty
● Next steps to:
● Improve numerical simulations and run parameter tests
● Develop policy in presence of uncertainty
● Investigate model predictions in micro data
BACKUP
Sketch of the numerical solution
Increase in uncertainty during a recession
Uncertainty index and GDP growth
1968:41969:11969:2
1969:31969:4
1970:1
1970:21970:3
1970:4
1971:11971:21971:31971:4
1972:11972:21972:3 1972:4 1973:1
1973:2
1973:3
1973:4
1974:1 1974:2
1974:3
1974:4
1975:1
1975:2
1975:31975:4
1976:11976:21976:31976:4
1977:11977:21977:31977:4
1978:11978:2
1978:3
1978:41979:1
1979:2
1979:31979:4
1980:11980:2
1980:31980:4
1981:1
1981:2
1981:3
1981:41982:1
1982:2
1982:3
1982:4
1983:1
1983:21983:3
1983:41984:11984:2
1984:31984:4
1985:11985:2
1985:3
1985:4
1986:11986:2
1986:31986:41987:1
1987:2
1987:3
1987:41988:1
1988:2
1988:3 1988:41989:1
1989:21989:3
1989:4 1990:11990:2
1990:31990:4
1991:1
1991:21991:31991:4 1992:1
1992:21992:31992:41993:11993:2
1993:31993:4
1994:11994:2
1994:3 1994:41995:11995:2 1995:31995:4
1996:11996:21996:31996:41997:1 1997:21997:3
1997:41998:1
1998:21998:3
1998:41999:11999:21999:3 1999:42000:1
2000:2
2000:3
2000:4
2001:12001:22001:3
2001:4
2002:12002:22002:32002:4
2003:12003:2 2003:3
2003:42004:1
2004:22004:3
2004:42005:12005:2
2005:3
2005:4 2006:12006:22006:32006:42007:1 2007:2
2007:32007:4
2008:12008:2
2008:3
2008:4
2009:1
.51
1.5
22.
5U
nce
rtai
nty
inde
x
-10 0 10 20dgdp
Policy makers believe that uncertainty matters, 5/5
Yoda (May 2009)“Uncertainty is the path to the dark side. Uncertainty leads to anger. Anger leads to hate. Hate leads to suffering.”
Tobin’s Q spread