Skewed Business Cycles Sergio Salgado Fatih Guvenen Nicholas Bloom University of Minnesota University of Minnesota, FRB Mpls, NBER Stanford University and NBER SED, 2016 Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 1 / 25
Introduction What happens to firm and macro level risk in recessions? Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 2 / 25
Introduction What happens to firm and macro level risk in recessions? Uncertainty Shocks Symmetric Risk Dispersion of firm-level outcomes is countercyclical (sales growth, stock returns, TFP, etc.) Second moment shocks generate a recession (Arellano, Bai, and Kehoe (2012); Bloom et al. (2015); Christiano et al. (2014)) Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 2 / 25
Introduction What happens to firm and macro level risk in recessions? Uncertainty Shocks Symmetric Risk Dispersion of firm-level outcomes is countercyclical (sales growth, stock returns, TFP, etc.) Second moment shocks generate a recession (Arellano, Bai, and Kehoe (2012); Bloom et al. (2015); Christiano et al. (2014)) Disaster Shocks Macro Tail Risk Small probability of large negative aggregate shock (Rietz (1988); Barro (2006,2009); Barro and Ursua (2008,2009)) Increase in disaster risk generates a decline in output, investment, and employment (Gourio (2012,2014); Gabaix (2012); Killic and Wachter (2015)) Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 2 / 25
Introduction Open questions Macro and firm level tail risk? Does the probability of a firm-level disaster increase during recession? Firm-level risk in other countries? Most firm-level analysis for the US economy only Is the countercyclical dispersion a US-only phenomenon or seen more broadly across di erent countries? Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 3 / 25
Introduction This paper Firm and macro-level panel data for 30+ countries and 20+ years We study the cyclicality of the distribution of growth rates of firm and macro level outcomes Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 4 / 25
Introduction This paper Firm and macro-level panel data for 30+ countries and 20+ years We study the cyclicality of the distribution of growth rates of firm and macro level outcomes Main Results In recessions, the left tail stretches out while the right tail changes little Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 4 / 25
Introduction This paper Firm and macro-level panel data for 30+ countries and 20+ years We study the cyclicality of the distribution of growth rates of firm and macro level outcomes Main Results In recessions, the left tail stretches out while the right tail changes little Skewness is robustly procyclical! Asymmetric Risk Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 4 / 25
Introduction This paper Firm and macro-level panel data for 30+ countries and 20+ years We study the cyclicality of the distribution of growth rates of firm and macro level outcomes Main Results In recessions, the left tail stretches out while the right tail changes little Skewness is robustly procyclical! Asymmetric Risk Both at the micro (e.g. sales growth) and the macro level (e.g. GDP growth) Within di erent groups of countries, within di erent industries, and firm-size categories. Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 4 / 25
Introduction General equilibrium model General Equilibrium model: Firms face convex cost and partial irreversibility of adjusting capital Idiosyncratic productivity: time-varying variance and time-varying skewness Results (in progress) A negative skewness shock generates a drop and then an overshoot in investment and output Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 5 / 25
Introduction Related literature Uncertainty Shocks: Bloom (2009), Bloom et al. (2015), Bloom et al. (2007), Christiano et al. (2014), Gilchrist et al. (2015), Bachmann and Bayer (2013), Arellano et al. (2012) Disaster Risk: Barro (2006,2009), Barro and Jin (2009), Barro and Ursua (2008,2009), Farhi and Gabaix (2008), Gabaix (2012), Gourio (2012,2014), Killic and Wachter (2015), Rietz (1988), Tsai and Wachter (2015), Wachter (2013) Firm Dynamics: Bachmann et al. (2013), Bachmann and Bayer (2014), Bachmann and Ma (2015), Khan and Thomas (2003,2008), Clementi et al. (2014) Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 6 / 25
Empirical Results Empirical Results Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 6 / 25
Empirical Results Data Data sources Large firm-level panel data Cross Country: annual sales and employment (Osiris; 44 countries, 500M obs.) and daily stock prices (Global Compustat; 23 countries, 72MM obs.) for 1985-2013 US: quarterly and annual sales, inventories, profits, employment (CRSP, 266M obs.) and daily stock prices (CRSP, 31MM obs.) for 1964-2013 Country-level data Quarterly GDP (46 countries) and daily stock market index (29 countries) for 1970 to 2014 Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 7 / 25
Empirical Results Example: Sales Growth Distribution in the US Firm-level sales growth distribution in expansions Density 0.2.4.6.8 P10 P50 P90 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 162 expansion quarters (235M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 8 / 25
Empirical Results Example: Sales Growth Distribution in the US Firm-level sales growth distribution in expansions Density 0.2.4.6.8 P90 P10 1.61 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 162 expansion quarters (235M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 9 / 25
Empirical Results Example: Sales Growth Distribution in the US Firm-level sales growth distribution in expansions Density 0.2.4.6.8 P50 P10 0.77 P90 P50 0.84 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 162 expansion quarters (235M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 10 / 25
Empirical Results Example: Sales Growth Distribution in the US Firm-level sales growth distribution in expansions Density 0.2.4.6.8 P50 P10 0.77 P90 P50 0.84 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 162 expansion quarters (235M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 10 / 25
Empirical Results Example: Sales Growth Distribution in the US Firm-level sales growth distribution in expansions Density 0.2.4.6.8 P50 P10 48% P90 P50 52% 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 162 expansion quarters (235M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 10 / 25
Empirical Results Example: Sales Growth Distribution in the US Asymmetric risk: left tail is thicker in recessions Density 0.2.4.6.8 P50 P10 1.18 P90 P50 0.90 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 18 recession quarters (24M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 11 / 25
Empirical Results Example: Sales Growth Distribution in the US Asymmetric risk: left tail is thicker in recessions Density 0.2.4.6.8 P50 P10 57% P90 P50 43% 2.5 2 1.5 1.5 0.5 1 1.5 2 2.5 Annual Growth of Quarterly Sales Note: Panel of Compustat firms with 25+ yrs. of data from 1970 to 2014. There are 18 recession quarters (24M obs.). Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 11 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Panel of countries: graphical construct Skewness: Kelly (normalized to mean 0, SD 1) 3 2 1 0 1 2 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 12 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Panel of countries: skewness is procyclical Skewness: Kelly (normalized to mean 0, SD 1) 3 2 1 0 1 2 Sales Growth Stock Returns GDP Growth Index Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 12 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Panel of countries: dispersion is countercyclical Dispersion: P90 P10 (normalized to mean 0, SD 1) 1 0 1 2 3 Sales Growth Stock Returns GDP Growth Index Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 13 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Skewness is procyclical also in the US economy Skewness US only Skewness: Kelly (normalized to mean 0, SD 1) 2 1 0 1 2 Sales Growth GDP Growth Stock Returns Index Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Dispersion US only Dispersion: P90 P10 (normalized to mean 0, SD 1) 2 1 0 1 2 Sales Growth GDP Growth Stock Returns Index Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 14 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Panel of countries: regression results, firm-level Dependent Variable: (1) (2) (3) (4) (5) (6) (7) (8) Growth Rate of GDP within each country Distribution: Growth Rate of Sales Daily Stock Returns P9010 i,t -0.017** -0.243*** (0.008) (0.0845) P5010 i,t -0.048** -0.556*** (0.019) (0.184) P9050 i,t 0.004-0.366** (0.014) (0.141) KSK i,t 0.028*** 0.064*** (0.008) (0.018) R 2 0.510 0.523 0.506 0.523 0.717 0.721 0.713 0.714 N 760 760 760 760 1793 1793 1793 1793 Freq. Annual Annual Annual Annual Quarter Quarter Quarter Quarter Years 86-12 86-12 86-12 86-12 86-14 86-14 86-14 86-14 Under. Spl 505M 505M 505M 505M 72MM 72MM 72MM 72MM Note: Each column reports the results from a country-period OLS panel regression including a full set of country and time fixed e ects. Standard errors, shown in parentheses below the point estimates, are clustered at country level. denotes 1%, denotes 5%, and denotes 10% significance respectively. US Micro More Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 15 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Panel of countries: regression results, macro-level Dependent Variable: (1) (2) (3) (4) (5) (6) (7) (8) Growth Rate of GDP within each country Distribution: Growth Rate of GDP Daily Returns of Stock Price Index P9010 i,t -0.150*** -0.253*** (0.048) (0.081) P5010 i,t -0.194*** -0.620*** (0.059) (0.157) P9050 i,t -0.165* -0.362** (0.088) (0.159) KSK i,t 0.005** 0.047*** (0.002) (0.011) R 2 0.650 0.650 0.642 0.641 0.723 0.723 0.722 0.723 N 5746 5746 5746 5746 2878 2878 2878 2878 Freq. Quarter Quarter Quarter Quarter Quarter Quarter Quarter Quarter Years 70-14 70-14 70-14 70-14 70-14 70-14 70-14 70-14 Note: Each column reports the results from a country - quarter OLS panel regression including a full set of country and quarter fixed e ects. Standard errors, shown in parentheses below the point estimates, are clustered at country level. denotes 1%, denotes 5%, and denotes 10% significance respectively. US Macro Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 16 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Skewness is also procyclical within country-groups Developing Countries Skewness: Kelly (normalized to mean 0, SD 1) 3 2 1 0 1 2 Sales Growth GDP Growth Stock Returns Index Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Developed Countries Skewness: Kelly (normalized to mean 0, SD 1) 3 2 1 0 1 2 Sales Growth GDP Growth Stock Returns Index Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 17 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries Within industries US only Skewness of Quarterly Sales Growth Skewness: Kelly (normalized to mean 0, SD 1) 2 1 0 1 2 Ag/Mi Con FIRE Man Trade Ser Tran 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Skewness of Daily Stock Returns Skewness: Kelly (normalized to mean 0, SD 1) 2 1 0 1 2 Ag/Mi Con FIRE Man Trade Ser Tran 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 18 / 25
Empirical Results Micro and Macro Skewness in a Panel of Countries And within employment-size categories US only Skewness of the Quarterly Sales Growth Skewness: Kelly (normalized to mean 0, SD 1) 3 2 1 0 1 Small Medium Medium/Large Large 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Skewness of the Daily Stock Returns Skewness: Kelly (normalized to mean 0, SD 1) 2 1 0 1 2 Small Medium Medium/Large Large 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 19 / 25
Empirical Results Summary Summary We document strong procyclical skewness In recessions, the left tail stretches out while the right tail changes little At macro and micro level Within groups of countries Within di erent industries and firm-size categories Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 20 / 25
General Equilibrium Model General Equilibrium Model Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 20 / 25
General Equilibrium Model Firms Firms Production Function y = zek q l a, q > 0, a > 0, q + a < 1 Aggregate productivity, log z 0 = r z log z + h, h N 0, sh 2 Idiosyncratic productivity, log e 0 = r e log e + n, with time-varying variance, s 2 n, and skewness, g n Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 21 / 25
General Equilibrium Model Firms Firms The aggregate state space is W z, s 2 n, g n, µ, then, V (e, k; W) =max l,i ( p (W) hezk q l a w (W) l i AC k, k 0 i + be V (e 0, k 0 ; W 0 ) ) k 0 = (1 d) k + i, AC (k, k 0 )= I (i<0) (1 x) i + wk log e 0 = r z log e + n, W 0 = P (W) i 2, k Household Problem Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 22 / 25
General Equilibrium Model Calibration Stochastic process Assume s 2 n and g n follow a two-point Markov chain: s n 2 {s h, s l } where Pr s 0 n = s j s n = s k = pkj s, g n 2 {g h, g l } where Pr g 0 n = g j g n = g k = p g kj, Use SMM to find values for s 0 s,theg 0 s,andthep 0 s We match moments of the time series of P9010 and KSK of the distribution of sales growth of Compustat firms Fit Parameters RBC Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 23 / 25
General Equilibrium Model Results E ects of a skewness shock 0.5 0-0.5 A skewness shock generates a drop and overshoot of investment, labor, and output % Change -1-1.5-2 Output Investment -2.5 Consumption Employment -3-5 0 5 10 15 20 Quarters (Skewness shock in quarter 1) Note: Based on independent simulations of 1000 economies of 300 quarters length. We impose a skewness shock in period 1, allowing normal evolution of the economy afterward. We plot the cross-economy percentage deviation of each series of this economy with respect to an economy that evolves normally for the entire simulation length. Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 24 / 25
Conclusion and Next Steps Conclusions and current work Empirical Evidence We document procyclical skewness of the growth rates of micro and macro level outcomes In recessions the left tail stretches out Quantitative Model Skewness shock generates bust-boom in investment and output Current work Aggregate skewness and (downside) risk aversion Draft coming up soon. Stay tuned! Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix US: regression results, firm-level Back Dependent Variable: (1) (2) (3) (4) (5) (6) (7) (8) Growth Rate of GDP Distribution Growth Rate of Sales Daily Stock Returns P9010 t -0.056** -0.262*** (0.022) (0.095) P5010 t -0.157*** -0.706*** (0.033) (0.208) P9050 t 0.067** -0.263 (0.033) (0.183) KSK t 0.078*** 0.095*** (0.013) (0.023) N 200 200 200 200 200 200 200 200 Frequency Quarter Quarter Quarter Quarter Quarter Quarter Quarter Quarter Under. Spl 266M 266M 266M 266M 31MM 31MM 31MM 31MM Years 64-13 64-13 64-13 64-13 64-13 64-13 64-13 64-13 Note: Each column reports a di erent time series OLS regression including a constant an a linear trend. In each column, the dependent variable is growth rate of the GDP per capita between quarter t and the same quarter of the next year. Newey - West standard errors are applied in each column to control for auto correlation (max of 1 lag). denotes 1%, denotes 5%, and denotes 10% significance respectively. Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix US: regression results, macro-level Back Dependent Variable: (1) (2) (3) (4) (5) (6) (7) (8) Growth Rate of GDP Distribution Growth Rate of Sales Daily Stock Returns of Price Index P9010 t -0.141-0.771*** (0.0938) (-3.03) P5010 t -0.328** -1.888*** (0.131) (-4.08) P9050 t 0.092-0.874 (0.168) (-1.59) KSK t 0.011** 0.123*** (1.98) (4.83) N 200 200 200 200 200 200 200 200 Frequency Quarter Quarter Quarter Quarter Quarter Quarter Quarter Quarter Years 64-13 64-13 64-13 64-13 64-13 64-13 64-13 64-13 Note: Each column reports a di erent time series OLS regression including a constant an a linear trend. In each column, the dependent variable is growth rate of the GDP per capita between quarter t and the same quarter of the next year. Newey - West standard errors are applied in each column to control for auto correlation (max of 1 lag). denotes 1%, denotes 5%, and denotes 10% significance respectively. Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Households Back Choose consumption, C, labor N, and next period s shares, l 0, to solve, C 1 y W (l; z, µ, g) = max C,N,l 0 1 y k Nc c + be W l 0 ; z 0, µ 0, g 0, Z C + q(k 0, e 0 ; W)l 0 (d e 0 k 0 ) apple w(w)n K Z + r(k, e; W)l 0 (d e 0 k 0 ), K subject to the law of motion of µ and b is the discount factor. The solution of this problem delivers w (W) U C (C, 1 N) = U N (C, 1 N) be " # U 0 C r 0 (e 0, k 0 ; W 0 ) U C r 1 (e 0, k 0 ; W 0 = 1, ) Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Left and right tail dispersion Back Left Tail Dispersion: P50 P10 (normalized to mean 0, SD 1) 1 0 1 2 3 Left Tail Dispersion Micro Sales Macro GDP Micro Returns Macro Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Right Tail Dispersion: P90 P50 (normalized to mean 0, SD 1) 2 1 0 1 2 Right Tail Dispersion Micro Sales Macro GDP Micro Returns Macro Returns 1 2 3 4 5 6 7 8 9 10 GDP growth deciles Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Targeted moments: time series over Compustat firms Back Skewness Sales Growth Distribution Kelly Skewness.35.2.05.1.25.4 1970q1 1980q1 1990q1 2000q1 2010q1 Kelly Skewness GDP Growth.05 0.05.1 GDP Growth Dispersion Sales Growth Distribution P9010.3.4.5.6.7 1970q1 1980q1 1990q1 2000q1 2010q1 P9010 GDP Growth.05 0.05.1 GDP Growth Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Dispersion and Skewness Calibration Back Targeted Moments P9010 KSK Moment Data Model Moment Data Model µ 0.393 0.376 µ 0.051 0.052 s 0.075 0.022 s 0.118 0.025 r 0.757 0.770 r 0.163 0.170 Untargeted Moments P9050 P5010 Moment Data Model Moment Data Model µ 0.206 0.198 µ 0.187 0.178 s 0.043 0.013 s 0.047 0.011 r 0.744 0.691 r 0.416 0.675 Note: The moments are calculated from the annual growth rate of sales from a sample of Compustat firms with 25+ years of data using quarterly sales from 1970-2014. The model micro moments are calculated in the same fashion as data moments. Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Parameter calibration in the model Back Parameter Value Description Stochastic Process {s h, s l } {3.83, 3.04} High and low volatility states in % {g h, g l } {0.029, 0.776} High and low skewness state {p hh (s), p ll (s)} {0.97, 0.98} Transition Matrix, s n {p hh (g), p ll (g)} {0.80, 0.95} Transition Matrix, g n Preferences and Technology b 0.95 1/4 Annual discount factor of 95% k 2 Leisure preference, HHs spend 1/3 time working q 0.25 Capital share of 1/3 a 0.58 Labor share of 2/3 d 2.6% Annual depreciation of capital stock of 10% x 34.0% Resale cost of capital in %, Bloom (2009) w 0.0 Convex adjustment cost Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Business cycle statistics Back Output Investment Consumption Hours Data s(x) 1.6 7.0 1.3 2.0 s(x) s(y) 1.0 4.5 0.8 1.3 r(x, y) 1.0 0.9 0.9 0.9 Model s(x) 2.90 11.17 2.60 0.97 s(x) s(y) 1.00 3.85 0.89 0.33 r(x, y) 0.99 0.98 0.99 0.56 Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Sample selection Back Same sample criteria for Osiris and Global Compustat Firms with more than 8 year of data, Periods with more than 100 firms, Countries with more than 5 years of data, and countries with more than 10 countries. For the US sample Firms with more than 25 years of data Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25
Appendix Skewness shock to idiosyncratic productivity Back A skewness shock does not change mean or variance 2 0-2 % Change -4-6 -8-10 Mean -12 Variance Skewness -14-5 0 5 10 15 20 Quarters (Skewness shock in quarter 1) Note: Based on independent simulations of 1000 economies of 300 quarters length. We impose a skewness shock in period 1, allowing normal evolution of the economy afterward. We plot the cross-economy percentage deviation of each series of this economy with respect to an economy that evolves normally for the entire simulation length. Salgado Guvenen Bloom Skewed Business Cycles SED, 2016 25 / 25