Reserve Requirements and Optimal Chinese Stabilization Policy 1 Chun Chang 1 Zheng Liu 2 Mark M. Spiegel 2 Jingyi Zhang 1 1 Shanghai Jiao Tong University, 2 FRB San Francisco ABFER Conference, Singapore May 24, 217 1 The views expressed herein are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of San Francisco or the Federal Reserve System. 1 / 36
PBOC frequently adjusts reserve requirements (RR) China required reserve ratio Percent 25 2 15 1 5 4 5 6 7 8 9 1 11 12 13 14 15 16 17 Source:Bloomberg Since 25, adjusted RR over 4 times Between 26 and 211, RR rose from 8.5% to 21.5%
Active RR adjustments when global interest rates declined Chinese interbank rate vs. US Treasury rate 3-month maturities Percent 7 6 5 4 3-mo. SHIBOR 3 2 3-mo. US Treasury 1 27 28 29 21 211 212 213 214 215 216 217 Source: Bloomberg -1 Under capital controls, declines in US yields raised cost of sterilization (e.g., Chang, Liu, and Speigel (215)) Raising RR a cheaper alternative to sterilization
Introduction RR increases encouraged shadow banking activity Shadow bank lending increased over 3% per year between 29 and 213 Shadow banking facilitates financial intermediation but increases financial risks [Gorton and Metrick (21)] Tightened regulations on formal banking contributed to shadow bank expansion (Elliott, et al (215); Hachem and Song (216); Chen, Ren, and Zha (216)) binding loan/deposit caps (small/medium banks) Interest rate controls Increases in RR Large-scale fiscal stimulus in 28-9 fueled demand for shadow bank financing 4 / 36
Introduction Impact of RR on financing costs affects resource allocations RR act as a tax on commercial banks Disproportionately affects state-owned enterprises (SOEs) SOEs enjoy implicit government guarantees on loans SOEs have superior access to bank loans despite low productivity Shadow banking not subject to RRs Main source of financing for privately-owned enterprises (POEs) (Lu, et al. (215)) RRs reallocates resources from SOEs to POEs Reduces SOE activity relative to POE POEs have higher average productivity (Hsieh-Klenow, 29) Thus, raising RR increases aggregate TFP 5 / 36
Introduction Illustrative macro evidence of RR s reallocation effects Simple BVAR with RR, 3-mo deposit rate, log real GDP, SOE investment share Data 1995:Q1 to 213:Q4; 4-qtr lags with Sims-Zha priors Ordering implies RR responds to all shocks in impact period Impulse responses: positive shock to RR reduces SOE investment share Results robust to RR being ordered last 6 / 36
BVAR: RR reallocates investment away from SOEs #1-3 6.8831.68 Error Bands Required reserve ratio #1-3 1.9445 Interest rate -3.76-1.2972 #1-3 4.2445 Real GDP #1-3 5.7697 SOE investment share -4.625-1.7645 4 8 12 16
Introduction Corroborating micro evidence of RR s reallocation effects Do RR increases reduce SOE stock returns relative to POE? Consider regression model: H h= H R e j,t+h = a +a 1 RR t 1 +a 2 SOE jt RR t 1 +a 3 SOE jt +bz jt +ε jt where R e j,t+h = R j,t+h ˆβ j R m,t+h denotes risk-adjusted excess return, RR t 1 denotes changes in RR, and Z jt is a vector of controls (size, book-to-market, industry fixed effects, year fixed effects) Focus on relative effects on SOEs (a 2 <?) Daily data for non-financial firms listed on Shanghai/Shenzhen stock exchanges, 25-215 Identification: event study of RR announcement effects 8 / 36
Introduction RR announcements effects on stock returns Event window 1-day (H=) 3-day (H=1) 5-day (H=2) RR t 1.26.479.157 (7.2) (9.21) (15.74) SOE jt RR t 1 -.12 -.225 -.442 (-3.21) (-3.32) (-5.5) SOE jt -.7 -.26 -.41 (-2.6) (-5.29) (-6.47) Size jt -.34 -.99 -.155 (-27) (-43) (-53) BM jt.9.24.47 (2.22) (3.29) (4.96) Sample size 4,119,971 4,79,847 4,3,53 R 2.71.182.288 9 / 36
Introduction The RR announcements effects observed mainly after 29, with rise of shadow banking following fiscal stimulus Pre-stimulus (25-28) Post-stimulus (29-215) Event window 1-day (H=) 3-day (H=1) 1-day (H=) 3-day (H=1) RR t 1.1.3.29.81 (2.) (.31) (8.8) (12.57) SOE jt RR t 1.1.12 -.24 -.46 (.11) (1.3) (-4.78) -5.3 SOE jt.2.5 -.2 -.5 (2.9) (4.9) (-4.85) (-8.86) Size jt -.3 -.8 -.4 -.11 (-9) (-14) (-26) (-41) BM jt..1.1.4 (-.25) (-.56) (2.91) (4.5) Sample size 1,18,628 1,3,518 3,11,343 3,76,329 R 2.5.11.8.22 1 / 36
Introduction What we do Build a two-sector DSGE model with financial frictions and Chinese characteristics to study: 1. implications of RR policy for allocation efficiency, aggregate productivity, and social welfare 2. role of RR policy in stabilizing business cycle fluctuations 3. optimal RR under simple policy rules and interactions with interest-rate policy 11 / 36
Introduction Main findings Raising RR improves aggregate productivity Acts as tax on banking and SOE activity Diverts resources to more productive POEs But raising RR also increases bailout costs SOE funding costs rise More incidence of SOE bankruptcies Tradeoff between efficiency and bailout costs interior optimal RR RR rule and interest-rate rule complementary for stabilization Interest-rate rule effective for stabilizing inflation and output RR rule more effective for reallocating resources 12 / 36
The model Two sector DSGE model Representative household consumes, saves, and supplies labor Retail sector: use wholesale goods as inputs; monopolistic competition and sticky prices Wholesale goods a CES aggregate of intermediate goods produced by SOEs and POEs POEs have higher average productivity (Hsieh-Klenow, 29) External financing for working capital subject to costly state verification: financial accelerator (BGG, 1999) Banks provide working capital to firms in both sectors Loans to SOEs are subject to RR, but debt guaranteed by government (on-balance-sheet) Loans to POEs exempt from RR, but no government guarantees (off-balance-sheet) 13 / 36
The model Representative household Utility function U = E Budget constraints [ ] β t ln(c t ) Ψ H1+η t, 1 + η t= C t + I t + D t = w t H t + r k D t 1 t K t 1 + R t 1 + T t P t P t Capital accumulation with adjustment costs (CEE 25) [ K t = (1 δ)k t 1 + 1 Ω ( ) ] 2 k It g I I t, 2 I t 1 14 / 36
The model Retail sector Final good CES composite of differentiated retail products [ 1 ] ɛ/(ɛ 1) Y f = Y t (z) (ɛ 1)/ɛ dz Demand curve facing each retailer ( ) Pt (z) ɛ Y t (z) = Yt f Monopolistic competition in retail markets, with quadratic price adjustment costs (Rotemberg, 1982) ( ) Ω p Pt (z) 2 2 πp t 1 (z) 1 C t P t Optimal price decision Phillips curve 15 / 36
The model Wholesale and intermediate goods Wholesale good a CES composite of SOE and POE products ( ) σm σm 1 σm 1 σm 1 σm σm M t = φy + (1 φ)y st Intermediate good production function in sector j {s, p} [ ] α Y jt = A t Ā j ω jt K 1 α jt (Hjt) e 1 θ Hjt θ where ω jt F jt ( ) denotes idiosyncratic productivity shocks pt Ā j = is scale of TFP, with Ā s < Ā p Aggregate TFP: A t = g t A m t, where A m t follows the process ln A m t = ρ a ln A m t 1 + ɛ at, 16 / 36
The model Financial frictions Firms finance working capital with net worth N j,t 1 and external debt B jt (BGG) Working capital constraint satisfies N j,t 1 + B jt P t = w t H jt + w e jth e jt + r k t K jt where wjt e is the real wage rate of managerial labor Constant returns implies that revenue linear in net worth p jt Y jt = Ã jt ω jt N j,t 1 + B jt P t where Ãjt denotes rate of return on firm investment (in consumption units) 17 / 36
The model Defaults Firms default if realized productivity ω jt sufficiently low: ω jt < ω jt Z jt B jt à jt (N j,t 1 + B jt ) where Z j,t is contractual rate of interest If firm defaults, liquidated by lender with fraction m j lost output Government covers loan losses on SOE loans (but not POE loans) using lump sum taxes 18 / 36
The model Financial intermediaries Banks take deposits from household at rate R t On-balance-sheet loans to SOEs subject to RR RR drives wedge between loan and deposit rate RR acts as tax on SOE borrowing Government guarantees imply risk-free loan rate R st for SOEs (R st 1)(1 τ t ) = (R t 1). Off-balance-sheet loans to POEs not subject to RR Funding cost Rpt = R t No government guarantees on POE debt lender charges default premium over funding cost (i.e., credit spread) on private loans 19 / 36
The model Financial contracts Optimal financial contract is a pair ( ω jt, B jt ) that solves max Ãjt(N j,t 1 + B jt )f (ω jt ) subject to the lender s participation constraint à jt (N j,t 1 + B jt )g(ω jt ) R jt B jt where B jt denotes loan amount and ω jt is cutoff productivity for firm solvency Defaults socially costly: ωjt ωjt f (ω jt )+g(ω jt ) = 1 m j ωdf (ω)+l j [ω jt (1 m j )ω]df (ω) where l s = 1 and l p = are guarantees on SOE and POE lending respectively 2 / 36
The model Benchmark monetary policy Two instruments for monetary policy: deposit rate and RR Interest rate follows Taylor rule ln ( ) Rt = ψ rp ln R ( ( πt ) + ψ ry ln π RR stays constant at steady-state level τ t = τ GDP t GDP ) 21 / 36
The model Market clearing and equilibrium Final goods marke clearing Yt f = C t + I t + G t + Ω p 2 (π t π 1)2 C t + N j,t 1 + B jt à jt m j P t j {s,p} Capital and labor market clearing ωjt ωdf (ω) K t 1 = K st + K pt, H t = H st + H pt Credit market clearing B st = (1 τ t )ζ t D t, B pt = (1 ζ t )D t, where ζ t is share of deposit for on-balance-sheet activity 22 / 36
Quantitative results Calibration Model solved based on calibrated parameters Parameters calibrated to Chinese data where available µ =.5: match SOE employment share α =.5: labor income share (Zhu, 212) κ = 1.587 and ωm =.37: match TFP dispersion (Hsieh-Klenow, 29) Relative TFP of POE Āp /Ā s = 1.42: Hsieh-Klenow (29) ψ =.45: target SOE share in industrial output of.3 σ m = 3: substitutability b/n SOE and POE outputs, Chang, et al. (215) ξs =.97 and ξ p =.69: match SOE and POE bankruptcy ratios in data Other calibration parameters fit to US data See Calibration for details 23 / 36
Quantitative results Steady state impact of RR increase.44 SOE output/poe output.669 Output based TFP.43.668.42.667.41.666.4.665.39.664.38.2.4.6.8 1.663.2.4.6.8 1.25 SOE bankruptcy ratio 5 x 1 4 Welfare gains.2.15 5.1 1.5.2.4.6.8 1 15.2.4.6.8 1 Reallocation from SOE to POE improves TFP Higher funding costs increase SOE bankruptcies Tradeoff interior optimum τ =.34 under our calibration 24 / 36
Quantitative results Monetary policy rules for stabilization Two instruments for monetary policy: deposit rate and RR Consider two types of simple (Taylor-like) policy rules Interest rate rule ( ) Rt ( πt ) ln = ψ rp ln + ψ ry ln R π ( GDP t GDP ) Reserve requirement rule ln ( τt ) ( πt ) = ψ τp ln + ψ τx ln τ π ( GDP t GDP ) 25 / 36
Quantitative results Compare macro stability and welfare under 4 policy rules Benchmark policy: Taylor rule with ψ rp = 1.5 and ψ ry =.2 and constant τ =.15 Optimal interest-rate rule: ψ rp and ψ ry set optimally to max welfare, and τ kept constant Optimal reserve-requirement rule: ψ τp and ψ τy set optimally, Taylor rule coefficients kept at benchmark values Jointly optimal rule: Coefficients for both interest rates and reserve requirements set optimally 26 / 36
Quantitative results The financial accelerator mechanism Financial accelerator: recession default prob rises monitoring cost and credit spread increase firm funding costs rise more default and even higher credit spread... Financial accelerator muted for SOEs but operative for POEs SOE debt guaranteed by gov t no default premium POE debt not guaranteed financial accelerator operative POE firms more sensitive to macro shocks Default premium always countercyclical, but credit spread can be pro- or countercyclical, depending on strength of credit demand (Carstrom-Fuerst, 1997; Faia-Monacelli, 27) Overall macro stability can be enhanced by using RR and interest-rate instruments 27 / 36
Quantitative results Aggregate Responses to TFP Shock: Benchmark Impulse responses to TFP shock 1.2 GDP -.3 Inflation -.4 1.1 -.5 1 -.6 -.7.9 1 2 3 4 -.8 1 2 3 4 -.2 Deposit rate.1 Required reserve ratio -.4.5 -.6 -.8 -.5-1 1 2 3 4 -.1 1 2 3 4 28 / 36
Quantitative results Sectoral responses to TFP shock: Benchmark Impulse responses to TFP shock 2 SOE output 2 POE output 1 1 1 2 3 4 SOE leverage.2 -.2 1 2 3 4 SOE bankruptcy ratio 2-2 1 2 3 4 SOE credit spread.2 -.2 1 2 3 4 1 2 3 4 POE leverage 5-5 1 2 3 4 POE bankruptcy ratio 2-2 1 2 3 4 POE credit spread 1-1 1 2 3 4 29 / 36
Quantitative results Aggregate Responses to TFP Shock: Benchmark vs alternative policies Impulse responses to TFP shock 1.2 GDP.2 Inflation 1.8.6 Benchmark Optimal R rule Optimal = rule Jointly optimal rules.4 1 2 3 4 -.2 -.4 -.6 -.8 1 2 3 4.5 Deposit rate 2 Required reserve ratio 15 1 -.5 5-1 1 2 3 4 1 2 3 4 3 / 36
Quantitative results Sectoral responses to TFP shock: Benchmark vs alternative policies Impulse responses to TFP shock 2 SOE output 2 POE output 1-2 1 2 3 4 SOE leverage.2 1 2 3 4 POE leverage 5 -.2 1 2 3 4 SOE bankruptcy ratio 2-5 1 2 3 4 POE bankruptcy ratio 2-2 1 2 3 4 SOE credit spread.2-2 1 2 3 4 POE credit spread 1 -.2 1 2 3 4-1 1 2 3 4 31 / 36
Quantitative results Macro stability and welfare under alternative rules Variables Benchmark Optimal τ rule Optimal R rule Jointly optimal rule Policy rule coefficients ψ rp 1.5 1.5 7.42 5.18 ψ ry.2.2.7.12 ψ τp. 13.14. 11.67 ψ τy. 4.81. 15.96 Volatility GDP 8.618% 8.155% 5.279% 4.952% π 3.49% 3.231%.84%.136% C 6.118% 5.95% 4.388% 4.36% H 2.13% 1.835%.599%.416% R 3.412% 3.236%.398%.349% Y s 9.91% 6.999% 5.362% 3.415% Y p 8.132% 8.455% 5.552% 5.982% Welfare Welfare gains.2423% 1.1799% 1.181% 32 / 36
Quantitative results Jointly optimal rule allows for complementary use of policy tools Adjust R-rule to stabilize inflation and GDP Adjust τ-rule to achieve desired reallocation of resources across sectors Leads to higher welfare gains than each individually optimal rule the two policy instruments are complementary 33 / 36
Conclusion Conclusion Examine RR policy in DSGE model with BGG financial accelerator and Chinese characteristics Changes in RR incur tradeoff between allocation efficiency and SOE bailout costs RR and interest rates are complementary policy instruments Interest rate effective for macro stabilization RR more useful for improving allocation efficiency and welfare Caveats: Results are second-best Open-economy features not in model: RR policy may stem from sterilized intervention in FX market 34 / 36
Additional material Parameter calibration I Back Variable Description Value A. Households β Subjective discount factor.995 η Inverse Frisch elasticity of labor supply 2 Ψ Weight of disutility of working 18 δ Capital depreciation rate.35 Ω k Capital adjustment cost 1 B. Retailers ɛ Elasticity of substitution between retail products 1 Ω p Price adjustment cost parameter 22 C. Firms g Steady state growth rate 1.125 k Shape parameter in Pareto distribution of idiosyncratic shocks 1.587 ω m Scale parameter in Pareto distribution of idiosyncratic shocks.37 A s SOE TFP scale (normalized) 1 A p POE TFP scale 1.42 α Capital income share.5 θ Share of household labor.94 ψ Share parameter for SOE output in intermediate good.45 σ m Elasticity of substitution between SOE and POE products 3 C. Financial intermediaries m s SOE monitoring cost.15 m p POE monitoring cost.15 ξ s SOE manager s survival rate.97 ξ p POE manager s survival rate.69 35 / 36
Additional material Parameter calibration II Variable Description Value C. Financial intermediaries m s SOE monitoring cost.15 m p POE monitoring cost.15 ξ s SOE manager s survival rate.97 ξ p POE manager s survival rate.69 D. Government policy π Steady state inflation rate 1.5 τ Required reserve ratio.15 ψ rp Taylor rule coefficient for inflation 1.5 ψ ry Taylor rule coefficient for output.2 G Share of government spending in GDP.14 GDP l s Fraction of SOE debt guaranteed by the government 1 l p Fraction of SOE debt guaranteed by the government E. Shock process ρ a Persistence of TFP shock.95 σ a Standard deviation of TFP shock.1 36 / 36