The Impact of Uncertainty Shocks: Firm-Level Estimation and a 9/11 Simulation Nick Bloom Stanford and Centre for Economic Performance March 2006 Monthly US stock market volatility Cambodia, Franklin Kent State National JFK assassinated Cuban missile crisis Monetary turning point OPEC I Afghanistan OPEC II 9/11 Enron Russia Gulf & LTCM War II Asian Gulf War I Crisis 0 30 10 20 40 (%) deviation standard Annualized 50 Black Monday* 1960 1965 1970 1975 Actual Volatility 1980 1985 Year 1990 1995 2000 2005 Implied Volatility Note: CBOE VXO index of % implied volatility, on a hypothetical at the money S&P100 option 30 days to expiry, from 1986 to 2004. Pre 1986 the VXO index is unavailable, so actual monthly returns volatilities calculated as the monthly standard-deviation of the daily S&P500 index normalized to the same mean and variance as the VXO index when they overlap (1986-2004). Actual and implied volatility correlated at 0.874. The market was closed for 4 days after 9/11, with implied volatility levels for these 4 days interpolated using the European VX1 index, generating an average volatility of 58.2 for 9/11 until 9/14 inclusive. * For scaling purposes the monthly VOX was capped at 50 affecting the Black Monday month. Un-capped value for the Black Monday month is 58.2. Monthly stock market levels September 114 JFK assassinated Vietnam Cuban missile crisis Russian & LTCM Default Cambodia, Kent State Asian Crisis Monetary cycle turning point OPEC I, ArabIsraeli War Franklin National financial crisis WorldCom & Enron Black Monday3 Gulf War II Gulf War I Afghanistan 0 50 OPEC II 1960 1965 1970 1975 1980 1985 Year 1990 1995 2000 2005 Note: S&P500 monthly index from 1986 to 1962. Real de-trended by deflating by monthly “All urban consumers” price index, converting to logs, removing the time trend, and converting back into levels. The coefficient (s.e.) on years is 0.070 (0.002), implying a real average trend growth rate of 7.0% over the period. The FOMC discussed uncertainty a lot after 9/11 Frequency of word “uncertain” in FOMC minutes 9/11 0.6% 0.5% 0.4% 0.3% 0.2% 0.1% Ja n Fe b Ma r Ap r Ma y Ju n Ju l Au g Se p Oc t No v De c Ja n Fe b Ma r Ap r Ma y Ju n Ju l Au g 0.0% 2001 2002 Source: [count of “uncertain”/count all words] in minutes posted on http://www.federalreserve.gov/fomc/previouscalendars.htm#2001 The FOMC also believed uncertainty mattered “The events of September 11 produced a marked increase in uncertainty ….depressing investment by fostering an increasingly widespread waitand-see attitude about undertaking new investment expenditures” FOMC minutes, October 2nd 2001 “The heightened degree of uncertainty and risk aversion following the terrorist attack seemed to be having a pronounced effect on business and household spending” FOMC minutes, November 6 2001 “Because the attack significantly heightened uncertainty it appears that some households and some business would enter a wait-and-see mode….They are putting capital spending plans on hold” FOMC member Michael Moskow, November 27th and even the Brits believed this mattered too “A general increase in uncertainty could lead to a greater reluctance to make commitments……Labour hiring and discretionary spending are likely to de deferred for a while, to allow time for the situation to clarify” Bank of England minutes, October 17th 2001 Motivation • Major shocks have 1st and 2nd moments effects • Policymakers believe both matter – is this right? – Lots of work on 1st moment shocks – Much less work on 2nd moment shocks • Closest work probably Bernanke (1983, QJE) – Predicts wave like effect of uncertainty flucatuations • I confirm, quantify & extend this work Summary of the paper Stage 1: Build and estimate structural model of the firm • Standard model augmented with – time varying uncertainty – mix of labor and capital adjustment costs • Estimate on firm data by Simulated Method of Moments Stage 2: Simulate 2nd moment shock • Generates rapid drop & rebound in – Hiring, investment & productivity growth • Confirm robustness to GE, risk-aversion, and AC estimates Stage 3: Compare to one example – 9/11 • Fits 9/11 data pretty well in magnitude and duration – Especially with additional 1st moment shock • Consistent with FOMC (and other central bank) comments Model Estimation Results Shock Simulations Base my model as much as possible on literature Investment • Firm: Guiso and Parigi (1999), Abel and Eberly (1999) and Bloom, Bond and Van Reenen (2005), Chirinko (1993) • Macro/Industry: Bertola and Caballero (1994) and Caballero and Engel (1999) • Plant: Doms & Dunn (1993), Caballero, Engel & Haltiwanger (1995), Cooper, Haltiwanger & Power (1999) Labour • Caballero, Engel & Haltiwanger (1997), Hamermesh (1989), Davis & Haltiwanger (1992) Labour and Investment • Shapiro (1986), Hall (2004), Merz and Yashiv (2004) Simulation estimation • Cooper and Ejarque (2001), Cooper and Haltiwanger (2003), and Cooper, Haltiwanger and Willis (2004) Real Options & Adjustment costs • Abel and Eberly (1994), Abel and Eberly (1996), Caballero & Leahy (1996), and Eberly & Van Mieghem (1997) • MacDonald and Siegel (1986), Pindyck (1988) and Dixit (1989) Firm Model outline Net Revenue Function, R Model has 3 main components Labor & capital “adjustment costs”, C Stochastic processes, E[ ] Firms problem = max E[ Σt(Rt–Ct) / (1+r)t ] Revenue function (1) Cobb-Douglas Production Q AK ( L H ) A is productivity, K is capital L is # workers, H is hours, α+β≤1 Constant-Elasticity Demand P BQ B is the demand shifter 1/ e Gross Revenue PQ Y 1 a b K (L H ) a b Y is “demand conditions”, where Y1-a-b=A(1-1/e)B a=α(1-1/e), b=β(1-1/e) Revenue function (2) Firms can freely adjust hours but pay an over/under time premium wages( H ) w1 (1 w2 H ) W1 and w2 chosen so hourly wage rate is lowest at a 40 hour week Net Revenue = Gross Revenue - Wages R(Y , K , L, H ) PQ w1 (1 w2 H ) L “Adjustment costs” “Adjustment Cost” Factor Concept Partial Irreversibility (PI) Labor Capital hiring/firing cost per person cost per unit capital resold Quadratic (QD) Labor Capital “rapid” hiring/firing more costly “rapid” investment more costly Fixed (FC) Labor Capital lump sum hire/fire cost lump sum investment cost C (Y , K , L, H ) PI QD FC Stochastic processes (1) “Demand conditions” evolve as a Random Walk • Hall (1987), Evans (1987), Dunne et al. (1989) for larger/older firms Yi ,t Yi ,t 1 (1 μ σ i ,t X i ,t ) X i ,t ~N( 0,1 ) 1st MOMENT SHOCK Stochastic processes (2) Uncertainty is comprised of a firm and macro component σ 2 i ,t 2F i ,t σ σ 2M t Firm level uncertainty is auto-regressive • Poterba and Summers (1986) σ 2F i ,t σ 2F i ,t 1 ρ (σ F σ 2* F σ 2F i ,t 1 ) σ Zi,t Zi ,t ~N( 0,1 ) F σ Macro level uncertainty has jumps • From initial graph σ 2M t σ 2M t 1 ρ (σ M σ 2*M σ 2M t 1 ) σ St St ~{0,1} M σ 2nd MOMENT SHOCK The optimisation problem is tough Value function V (Y , L, K , F , M , p K ) max R(Y , L, K , H ) C (Y , L, K , H , I , E , p K ) I ,E ,H 1 F M K E V (Y dY , ( L E )(1 L ), ( K I )(1 K ), , , p ) 1 r Note: I is gross investment, E is gross hiring/firing and H is hours Simplify by solving out 2 state and 1 control variables – Micro and macro uncertainty similar half-life (≈ 2 months), so assume σσM= σσF, and define σ2=σ2 F+σ2 M – Homogenous degree 1 in (Y,K,L) so normalize by K – Hours are flexible so pre-optimize out Simplified value function ~ ~ Q( y, l , , p ) max R ( y, l ) C ( y, l , i, e, p K ) K i ,e (1 K )(1 i ) E Q( y dy, (l e)(1 L ), , p K ) 1 r Solving the model • Analytical methods for broad characterisation: – Unique value function exists – Value function is strictly increasing and continuous in (Y,K,L) – Optimal hiring, investment & hours choices are a.e. unique • Numerical methods for precise values for any parameter set “Demand Conditions”/Capital: Ln(Y/K) Example hiring/firing and investment thresholds Invest Fire Inaction Hire “Real options” type effects Disinvest “Demand Conditions”/Labor: Ln(Y/L) “Demand Conditions”/Capital: Ln(Y/K) High and low uncertainty thresholds Larger “Real options” at higher uncertainty Low uncertainty High uncertainty “Demand Conditions”/Labor: Ln(Y/L) Taking the model to real data • Model predicts many “lumps and bumps” in investment and hiring • See this in truly micro data – i.e. GMC bus engine replacement – But (partially) hidden in plant and firm data by cross-sectional and temporal aggregation • Address this by building cross-sectional and temporal aggregation into the simulation to consistently estimate on real data Including cross-sectional aggregation • Assume firms owns large number of units (plants or markets) • Units demand process combines firm and unit shock Yt Yt Yt U F where YtF is a firm-level process as earlier Yt Y (1 i,tUt ) U U t 1 U Ut ~N( 0,1 ) ΦU relative unit uncertainty • Simplifying to solve following broad approach of Bertola & Caballero (1994), Caballero & Engel (1999), and Abel & Eberly (1999) – Assume unit-level optimization (managers optimize own “P&L”) – Links across units in same firm all due to common shocks Including temporal aggregation • Shocks and decisions typically at higher frequency than annually • Limited survey evidence suggests monthly frequency most typical • Model at monthly underlying frequency and aggregate up to yearly Model Estimation Results Shock Simulations Estimation overview • Need to estimate all 20 parameters in the model – 8 Revenue Function parameters • production, elasticity, wage-functions, discount, depreciation and quit rates – 6 “Adjustment Cost” parameters • labor and capital quadratic, partial irreversibility and fixed costs – 6 Stochastic Process parameters • “demand conditions”, uncertainty and capital price process • No closed form so use Simulated Method of Moments (SMM) – In principle could estimate every parameter – But computational power restricts SMM parameter space • So estimate 6 adjustment cost parameters & pre-determine the rest from the data and literature Pre-determined parameters Parameter: Value: Source: α (capital coefficient) 1/3 Prod function estimation β (labor coefficient) 2/3 Prod function estimation δK (capital depreciation) 10% Depreciation estimates δL (labor quit rate) 10% Matched to capital w1 (wage parameter) 1/3 10 employees per unit w2 (wage parameter) 7e-06 40 hour working week γ (wage parameter) 2.5 Overtime share 27% μ (demand drift) 5% Compustat average growth ε (demand elasticity) -3 50% mark-up pk* (capital price process) 1 Normalized to unity ρpk (capital price process) 0.12 NBER 4-digit industry data σpk (capital price process) 0.27 NBER 4-digit industry data σ* (uncertainty process) 0.29 Firm level share returns vol Fσ (uncertainty process) 0.16 Firm level share returns vol ρσ (uncertainty process) 0.42 Firm level share returns vol Simulated Method of Moments estimation • SMM minimizes distance between actual & simulated moments A S A S ˆ min [ ()]' W [ ()] actual data moments simulated moments weight matrix • Efficient W is inverse of variance-covariance of (ΨA - ΨS (Θ)) • Lee & Ingram (1989) show under the null W= (Ω(1+1/κ))-1 – Ω is VCV of ΨA, bootstrap estimated – κ simulated/actual data size, I use κ=10 Data is firm-level from Compustat • 10 year panel 1991 to 2000 to “out of sample” simulate 9/11 • Large continuing manufacturing firms (>500 employees, mean 4,500) – Focus on most aggregated firms – Minimize entry and exit • Final sample 579 firms with 5790 observations Note: This methodogly enables use of public firm data, avoiding the need to access the LRD and allowing match to firm financial data etc. Model Estimation Results Shock Simulations TABLE 2 Actual SMM Estimate “Adjustment cost” estimates Labor estimation moments Capital estimation moments Labor hire/fire costs (PI) 4.9 weeks wages Labor fixed costs (FC) 2.4 weeks revenue Labor quadratic costs (QD) 0 Capital resale cost (PI) 42.1% price capital Capital fixed costs (FC) 0.3 weeks revenue Capital quadratic costs (QC) 4.74 of K*(I/K)2 Std (ΔL/L) 0.197 0.234 Skew (ΔL/L) 0.213 0.437 Corr (ΔL/L)t, (ΔL/L)t-2 0.111 0.106 Corr (ΔL/L)t, (I/K)t-2 0.102 0.152 Corr (ΔL/L)t, (ΔS/S)t-2 0.137 0.174 Std (I/K) 0.141 0.146 Skew (I/K) 1.404 1.031 Corr (I/K)t, (ΔL/L)t-2 0.139 0.207 Corr (I/K)t, (I/K)t-2 0.305 0.318 Corr (I/K)t, (ΔS/S)t-2 0.210 0.325 Closer match between left and right columns of moments means a better fit Results for estimations on restricted models Capital “adjustment costs” only • Fit is only moderately worse • Both capital & labor moments reasonable • So capital ACs and (σt,pK) dynamics approximate labor ACs Labor “adjustment costs” only • Labor moments fit is fine • Capital moments fit is bad (too volatile & low dynamics) • So OK for approximating labor data Quadratic “adjustment costs” only • Poor overall fit (too little skew and too much dynamics) • But industry and aggregate data little/no skew and more dynamics • OK for approximately more aggregated data Robustness - measurement error (ME) • Labor growth data contains substantial ME from – Combination full time, part-time and seasonal workers – Rounding of figures – First differencing to get ΔL/L • Need to correct in simulations to avoid bias • I estimate ME using a wage equation and find 11% – Hall (1989) estimates comparing IV & OLS & finds 8% • Then build 11% ME into main SMM estimators – Also robustness test without any ME and find larger FCL Robustness – volatility measurement • Volatility process calibrated by share returns volatility – But could be concerns over excess volatility due to “noise” • Jung & Shiller (2002) suggest excess volatility more macro problem • Vuolteenaho (2002) finds “cash flow” drives 5/6 of S&P500 relative returns • Use 5/6 relative S&P500 returns variance and results robust – Find slightly higher adjustment costs Model Estimation Results Shock Simulations Simulating 2nd moment uncertainty shocks Structurally simulate shocks hard because big and unprecedented To recap combined micro and macro uncertainty process is as follows σ σ 2 i ,t 2 i ,t 1 2* ρσ(σ σ 2 i ,t 1 ) σ Zi ,t σ St F σ M σ where σ2*=σ2F*+σ2M* and ρσ=ρσM=ρσF St ~{0,1} Macro uncertainty shock Simulation of macro shock sets St=1 for one period • σMσ = σ* so shock doubles average σ2i,t • Calibrated from doubling macro share volatility after large shocks Auto-regressive σt approximated by Markov-chain Tauchen & Hussey (1991) to define 5-point space and transition matrix - Normal times (St=0) calibrated from firm share returns volatility σ=8% σ=17% σ=25% σ=38% σ=76% σ=8% 0.645 0.249 0.084 0.020 0.002 σ=17% 0.249 0.361 0.255 0.115 0.020 σ=25% 0.084 0.255 0.321 0.255 0.084 σ=38% 0.020 0.255 0.255 0.361 0.249 σ=76% 0.002 0.020 0.084 0.249 0.645 - Shock period (St=1) calibrated to double uncertainty σ=8% σ=17% σ=25% σ=38% σ=76% σ=8% 0.001 0.008 0.033 0.132 0.825 σ=17% 0.000 0.000 0.000 0.007 0.993 σ=25% 0.000 0.000 0.000 0.001 0.999 σ=38% 0.000 0.000 0.000 0.000 1.000 σ=76% 0.000 0.000 0.000 0.000 1.000 Simulation uncertainty macro “impulse” Average Uncertainty (σi,t) uncertainty shock Run model monthly with 100,000 firms for 5 years to get steady state then hit with uncertainty shock Month Simulation percentiles of firm uncertainty uncertainty shock Uncertainty (σt) 90th Percentile uncertainty shock shifts distribution of σit upwards 75th Percentile 50th Percentile 25th Percentile Month 10th Percentile Actual percentiles of firm volatility after 9/11 100 9/11 Actual Compustat firm level data Real 9/11 shock did actually shift distribution of returns volatility upwards 50 90th Percentile 75th Percentile 0 50th Percentile 25th Percentile 10th Percentile 2001.5 Monthly data 2002 Year Calculated from CRSP daily share returns volatility within each month of balanced panel of 1,052 firms in CRSP-Compustat matched sample with over 500 employees and full daily trading data from 1990 to 2003. 9/11 month volatility taken from the first trading day after the attack until the end of the month (the 9 trading days from 9/17/2001 until 9/28/2001). sd10 sd25 Aggregate net hiring rate (%) Net hiring rate uncertainty shock Month Percentiles of firm net hiring rates (%) Net hiring rate 99th Percentile 95th Percentile 5th Percentile 1st Percentile Month Macro gross investment rate (%) Investment rate uncertainty shock Month Firm percentiles of gross investment rates (%) Investment rate 99th Percentile 95th Percentile 5th Percentile 1st Percentile Month Productivity growth rate (%) Productivity growth uncertainty shock Total Between Within Cross Productivity & hiring, period of shock Gross hiring rate Gross hiring rate Productivity & hiring, period before shock Month Productivity (logs) Productivity (logs) GDP loss from 2nd moment shock Estimate rough magnitude of GDP loss, noting • Only from temporary 2nd moment shock (no 1st moment effects) • Ignores GE (will discuss shortly) so only look at first few months GDP loss from an uncertainty shock (% of annual value) First 2 months First 4 months First 6 months Input Factors 0.30 0.74 1.16 TFP (reallocation) 0.07 0.11 0.14 Total 0.37 0.85 1.30 Reasonable size – uncertainty effects wipes out growth for ½ half year Investment rate After a 1st moment shock expect standard U-shape downturn, bottoming out after about 6-12 months Hiring rate After a 2nd moment shock everything drops – just like a 1st moment shock - but then bounces back within 1 month Prod. growth Highlights importance identifying 1st & 2nd moment components of shocks To distinguish try using: (i) volatility indicators; (ii) plant spread; to help distinguish Month Robustness – Risk aversion • Earlier results assumed risk-neutrality • But can model discount rate (r) as a function of uncertainty • Re-simulate with an “ad-hoc” risk aversion correction – Calibrated so that increases average (r) by 2.5% Investment rate uncertainty shock risk-neutral risk-averse Month Robustness – Adjustment costs estimation • Need some non-convex costs - nothing with convex ACs only • Robust to type non-convex ACs (Dixit (1993) and Abel & Eberly (1996) show thresholds infinite derivate AC at AC≈0 ) PI=10%, all other AC=0 Aggregate Hiring Hiring Distribution Productivity Hiring Distribution Productivity FC=1%, all other AC=0 Aggregate Hiring Robustness - General Equilibrium effects • Could build in GE using approximation for the cross-sectional distribution of firms – But need another program loop, so much slower - thus choice between: (i) estimating ACs, or (ii) doing GE – Estimate ACs as more sensitive to this and do GE later • Less sensitive to GE for two reasons – Uncertainty shocks very rapid and big, but wages and prices “sticky” at monthly frequency and interest rates bounded at zero • Uncertainty shock adds 6% to 10% to hurdle rates, but after 9/11 interest rates fell by only 1.75% – Drop and rebound optimal with GE anyway as correct factor allocation unclear, expensive to change so a pause is good Robustness – Combined 1st and 2nd moment shock • Earlier results 2nd moment shock only ~ thought experiment • But shocks typically have 1st and 2nd moment component Investment rate • Re-simulate assuming – 2nd moment shock (doubles uncertainty as before) – 1st moment shock (-5% ≈ 1 years growth) 2nd moment shock 1st & 2nd moment shock Month How does the simulation fit against actual data? Look at 9/11 because • Large 2nd moment shock with relatively small 1st moment effect – So “cleaner” test of the model • Recent, so can match up data for – Central Bank minutes (FOMC from 1996, BOE from 1997) – Consensus forecasts (from 1998) 9/11 did generate a rapid drop and rebound Quarterly Net Hiring (total private, thousands) 1 9/11 0 Forecast of 23rd August 20013 Lowest quarterly value since 1980 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Year Quarterly Investment (% contribution to real GDP growth) 2 forecast dempq dempq1 0 5 Forecast of 23rd August 20013 -5 Lowest quarterly value since 1982 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Year 1 BLS Current Employment Statistics survey, Total private employees (1000s), seasonally adjusted, quarterly net change, from series CES0500000001 forecast z BEA National Income and Product Accounts, Contributions to % change in real Gross Domestic Product, seasonally adjusted at annual rates, from Table 1.1.2 Gross private domestic investment 3 Federal Reserve Bank of Philadelphia “Survey of Professional Forecasters” average of 33 economic forecasters, www.phil.frb.org/file/spf/survq301.html 2 9/11 did generate a rapid drop and rebound Quarterly Net Hiring (total private, thousands) 1 9/11 Forecast of 23rd August 20013 Forecast of 14th November 2001 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Year Quarterly Investment (% contribution to real GDP growth) 2 forecast1 Forecast of 23rd August 20013 0 5 forecast dempq -5 Forecast of 14th November 2001 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Year 1 BLS Current Employment Statistics survey, Total private employees (1000s), seasonally adjusted, quarterly net change, from series CES0500000001 forecast forecast1 BEA National Income and ProductGross Accounts, Contributions % change in realinvestment Gross Domestic Product, seasonally adjusted at annual rates, from Table 1.1.2 private to domestic 3 Federal Reserve Bank of Philadelphia “Survey of Professional Forecasters” average of 33 economic forecasters, www.phil.frb.org/file/spf/survq301.html 2 THE POLICY VERDICT Looks like the FOMC did the right thing after 9/11 • Pumped in liquidity to reduce uncertainty • Did not cut interest rates much – Cut Federal Funds Rates by 1.75%, but this was already falling (2-year market rates fell be less than 1%) Congress on the other hand was not so perfect… • “A key uncertainty in the outlook for investment spending was the outcome of the ongoing Congressional debate relating to tax incentives for investment in equipment and software. Both the passage and the specific contents of such legislation remained in question” FOMC Minutes, November 6th 2001 A QUICK HISTORICAL DIGRESSION (not really part of the paper) The Great Depression was notable for very high volatility 60 90 The Great Depression 0 30 9/11 1880 1890 1900 1910 1920 Year 1930 1940 1950 1960 Note: Volatility of the daily returns index from “Indexes of United States Stock Prices from 1802 to 1987” by Schwert (1990). Contains daily stock returns to the Dow Jones composite portfolio from 1885 to 1927, and to the Standard and Poor’s composite portfolio from 1928 to 1962. Figures plots monthly returns volatilities calculated as the monthly standard-deviation of the daily index, with a mean and variance normalisation for comparability following exactly the same procedure as for the actual volatility data from 1962 to 1985 in figure 1. Did uncertainty play a role in the Great Depression? • Romer (1990) suggests uncertainty played a role in the initial 19291930 slump, which was propagated by the 1931 banking collapse “during the last few weeks almost everyone held his plans in abeyance and waited for the horizon to clear”, Moody’s 12/16/1929 • In the model a GD sized persistent increase in uncertainty would also generate persistently slower productivity growth • TFP “inexplicably” fell by 18% from 1929-33 (Ohanian, 2001) • Output “oddly” not shifted to low-cost firms (Bresnahan & Raff, 1991) GNP growth in the Great Depression Fall in Rise in volatility volatility Banking panics Source: Romer (1992, JEH) END OF DIGRESSION Conclusions • Uncertainty spikes after major economic & political shocks • Estimation and simulation predicts rapid drop & rebound – For 9/11 appears to roughly match actual data • This time profile looks different from a levels shock • Suggests policy makers try to distinguish levels & uncertainty effects – Financial volatility (VXO) and compression of firm activity Current extension in progress Build GE model by approximating cross-sectional distribution. Should help with a number of business-cycle issues, in particular: • Lack of negative TFP shocks - 2nd moment shocks mimic these (especially after detrending) • Drop on impact for TFP shocks - 1st moment shocks raise uncertainty when the shock first hits (dynamic inference) • Instability of VARs without 2nd moment controls Also model link between volatility and growth – less reallocation (which drives about ½ to ¾ of TFP growth) at higher uncertainty Approximating cross-sectional distributions Number of ways to approximate cross sectional distributions, i.e. – Moments (Krussell and Smith) – Characteristics functions (Caballero and Engel) I use bins exploiting the fact agents know distribution is bounded, i.e: Actual distribution Bin approximation Capital/Demand (K/Y) BACK-UP “Adjustment costs” (2) • 1 period time to build • Exogenous quit rate δLand depreciation rate δK • Relative capital price is AR(1) stochastic p p ρ p K (p p ) σ p K U t K t K t 1 K* K t 1 U t ~N( 0,1 ) Impact of a levels shock looks different 1st moment shock (3%) Hiring 1st moment shock (3%) Investment Month Month Total Between Within Cross 99th percentile 95th percentile Hiring percentiles Productivity 5st percentile 1st percentile Month Month Robustness- general equilibrium effects (2) • Thomas (2002) and Veracierto (2002) suggest GE important – In particular they find under GE dM t d( ) dYt 0 dNC Mt is a BC variable like labor, or capital Yt is aggregate productivity/demand NC is some non-convex cost – But I look at dM t d t σt is uncertainty • So correctly highlight importance of GE, but on a different issue Also need to deal with aggregation Structures Equipment Vehicles Total Firms 5.9 0.1 n.a. 0.1 Establishments 46.8 3.2 21.2 1.8 Single plants 53.0 4.3 23.6 2.4 Small single plants 57.6 5.6 24.4 3.2 Aggregation across lines of capital standard deviation/mean of growth rates (US firm data) Quarterly Yearly Sales 6.78 2.97 Investment 1.18 0.84 Aggregation across time Aggregation across units % annual zero investment episodes (UK Firm and Plant data) 8 Interest rates Federal Funds rate 2-year rate (T-Bill) 2 4 6 9/11 2000 2001 2002 2003 2004 2005 Year 3-year T-Bills ir Fiscal position ≈ flat 2001-02 excluding personal tax cuts % GDP 01 Q1 01 Q2 01 Q3 01 Q4 02 Q1 02 Q2 02 Q3 02 Q4 Budget surplus 1.1 0.5 -1.8 -1.3 -3.3 -3.7 -3.7 -4.3 …exc. personal tax -11.8 -12.5 -12.7 -13.4 -13.6 -13.7 -13.6 -13.9 Source: Federal Reserve Board Statistical Release - http://www.federalreserve.gov/releases/H15/data.htm Employment quits, layoffs and 9/11 Month Source: Hall (2005a) “Adjustment costs” (1) Active literature with range of approaches, e.g. Labor or capital Labor and Capital Convex1 Traditional Euler and Tobin’s Q Shapiro (1986); Hall (2004), models Merz and Yashiv (2003) Convex1 and Non-Convex2 Abel & Eberly (1999); Cooper & Haltiwanger (2003); Cooper, Haltiwanger and Willis (2004) I look at convex & non-convex adjustment costs for both labor and capital 1 Convex typically quadratic adjustment costs 2 Non-convex typically fixed cost or partial irreversibility