China's Model of Managing the Financial SystemChina's Model of Managing the Financial System Markus K. Brunnermeier Michael Sockin Wei Xiong Feb, 2017 ... University of Chicago Booth

JRCPPF Escalating Risks

China's Model of Managing the Financial System

Markus K. Brunnermeier Michael Sockin Wei Xiong

Feb, 2017

Discussion by Lin William Cong University of Chicago Booth School of Business

Overview Model Comments Summary

Contribution and Significance

• Government Intervention and Impact on Price Dynamics

• Context of China’s Financial SystemStock market crash in 2015.

• Real, Important, and Timely.

Slide 1/24 — Discussion by Will Cong (Chicago Booth) — China’s Model of Managing the Financial System














What Happened in China in 2015?

• A taxi-driver’s tale.



What Happened in China in 2015?

• A taxi-driver’s tale.



Fund Managers’ Speculation

• All about interventions:



Closed-door Round-Table Discussion

• “Anomalous Movements” , Reducing Volatility.


From "Saving China's Stock Market" by Yi Huang, Jianjun Miao, and Pengfei Wang


Important

• China’s Economy and Financial Markets.

• By the end of September 2015, 42 trillion RMB, dailyvolume of August 2015 was 514.8 billion RMB



Important

• New Investors Dynamics



Important

• International Ramifications

• Currency Devaluation August 2015Daily reference rate down 1.9 percent, largest since Jan1994

• Implications for other countriesUnconventional monetary policy in OECDs.Introduction of Euros (Gopinath et al. (2015))Economic stimulus after the global recession.

• Our understanding of business (credit) cycles andallocational efficiency.

• Caballero et. al (1994), Cooper et al (1993), Kiyotaki &Moore (1997), Ramey & Watson (1997)



Important








Important








Timely

• Impact and unintended consequences of marketinterventions.Brunnermeier, Sockin, & Xiong (2017), Bai, Hsieh, & Song (2016), Hachem & Song (2016), Huang, Pagano, & Panizza (2016), Cong & Ponticelli (2017), Chen, He, & Liu (2017)

• Ongoing reforms of financial markets• “Reform Bull”: SOE reforms and transition from

export-based to consumption-based• “Liquidity Bull”: China’s version of QE• “Capital Bull”: abundance capital and excessive shadow

leverage• Registration-based IPO, Microstructure Reforms (T + 1,

down limit, circuit breaker), internationalization of RMB

• Ongoing reforms and interventions in other countries.



Timely

• Impact and unintended consequences of marketinterventions.Brunnermeier, Sockin, & Xiong (2017), Bai, Hsieh, &Song (2016), Hachem & Song (2016), Huang, Pagano, & Panizza (2016), Cong & Ponticelli (2017), Chen, He, & Liu (2017)








Timely

• Impact and unintended consequences of marketinterventions.Brunnermeier, Sockin, & Xiong (2017), Bai, Hsieh, & Song (2016), Hachem & Song (2016), Huang, Pagano, & Panizza (2016), Cong & Ponticelli (2017), Chen, He, & Liu (2017)








Model Setup

• Grossman-Stiglitz REE with a large player and OLGagents

• Dividends Dt = θt + σDεDt

persistent component (fundamental) AR(1)θt = ρθθt−1 + σθε

θt

• Excess Payoff: Rt+1 = Dt+1 + Pt+1 − R f Pt

• Participants: noise traders, rational traders (OLG),government



Model Setup

• Noise traders: Nt = ρNNt−1 + σNεNt

• Investors:

• i ∈ [0, 1], live for one period

• Endowed with wealth W̄

• CARA utility U it = E[− exp(−γW i

t+1)|Ft ]

• Government

• Minimizes price vol and deviation from fundamental

• min γσVart [∆Pt ] + γθVart[Pt − 1

R f −ρθθt+1

]• γσ risk aversion for price vol; γθ risk aversion for

deviation.



Model Setup

• Noise traders: Nt = ρNNt−1 + σNεNt

• Investors:

• i ∈ [0, 1], live for one period

• Endowed with wealth W̄

• CARA utility U it = E[− exp(−γW i

t+1)|Ft ]

• Government

• Minimizes price vol and deviation from fundamental

• min γσVart [∆Pt ] + γθVart[Pt − 1

R f −ρθθt+1

]• γσ risk aversion for price vol; γθ risk aversion for

deviation.



Key Results

• Benchmark: perfect information

• No intervention, price vol explode, and marketbreakdown when σN is large.

• Noise trading + Investor myopia

• Role for the government to reduce market vol andstabilize the market.

• Intervention to reduce price volatility consistent withimproving price efficiency (always the case?).



Key Results

• With information frictions: unobservable fundamentals

• No intervention, reducing price volatility consistent withimproving price efficiency.

• Intervention introduces additional noise.

• Speculation on gov noise.• Price vol reduced at the expense of price efficiency.• Gov-centric equilibrium has less gov trading• Gov pre-commitment and equilibrium multiplicity.



Comment 1: Government’s Objective

• Reducing both asset price deviation from fundamentalsand price vol.

• Alternative 1: preventing crash




• Alternative 2: Favoring State-connected Firms

• e.g., Song, Storesletten, & Zilibotti (2011); Cong &Ponticelli (2017)

• Alternative 3: Social Stability• Agents with imperfect signal xit• Prior of economy fundamental θt• Government intervention mt

• Market breaks down if too many run Arunt > mt + θ∗t

• Government minimizes Z (mt , z , θt)Astayt L + K (m) (or

threshold), L is loss upon failure.• Learning or info acquisition about z• Government-centric vs Fundamental-centric




• Alternative 2: Favoring State-connected Firms

• e.g., Song, Storesletten, & Zilibotti (2011); Cong &Ponticelli (2017)

• Alternative 3: Social Stability• Agents with imperfect signal xit• Prior of economy fundamental θt• Government intervention mt

• Market breaks down if too many run Arunt > mt + θ∗t

• Government minimizes Z (mt , z , θt)Astayt L + K (m) (or

threshold), L is loss upon failure.• Learning or info acquisition about z• Government-centric vs Fundamental-centric



Comment 2: Minimization of Pricing Error


• The Bias-Variance Trade-off (Decomposition):E (y0 − f̂ (x0))2 = Var(f̂ (x0)) + [Bias(f̂ (x0))]2 + Var(ε)





• The Bias-Variance Trade-off (Decomposition):E (y0 − f̂ (x0))2 = Var(f̂ (x0)) + [Bias(f̂ (x0))]2 + Var(ε)




• Penalty for Pricing Errors

E

[(∑t

Dt − Pt(υN̂,t))2|FMt−1

]= Var(

∑t

Dt) + E [∑t

Dt ]2 + Var [Pt(υN̂,t)]

+E [Pt(υN̂,t)]2 − 2E [(∑t

Dt)Pt(υN̂,t)]

= Var(∑t

Dt) + Var [Pt(υN̂,t)] + E

[Pt(υN̂,t)− E [

∑t

Dt ]

]2

• Natural interpretation of the government’s objective.• For equilibrium existence, Bias-Variance are qualitatively

equivalent. But unclear in general.



Comment 3: Information Design and Intervention Costs

• Information Design vs Mechanism Design (Bergemannand Morris (2016))

• Information Structure Design & Symmetric Learning• Intervention impact uncertain, effectiveness less than

expected.• Alteration of the informational environment.• Goldstein and Huang (2016), Cong et al (2017)

• Asymmetric Information and Learning• Footnote 11;

Signaling (Angeletos et al (2006)); Reputation (Huang(2017))

• Multiplicity and commitment






















• Intervention Costs

• Taxpayers’ money, and investment risks in interventions.• Political capital• Disrupted IPO reform and currency internalization• Firing of head of CSRC, Xiao Gang; imprisonment of

Citic officials.

• Impact of Intervention Costs

• Interior optimal intervnetion absent cost of intervention.• What is the welfare gains and losses in gov-centric vs

fundamental-centric equilibrium?• Cost matters when there is dynamic information design.• Relevant for multiple interventions: Cong et al (2017).






Citic officials.









Citic officials.









Citic officials.









Citic officials.






Comment 4: Cum Dividend Return and Variance

• By p7

Rt+1 = Dt+1 + Pt+1 − R f Pt

= (θt+1 + σDεDt+1) + (pθθt+1 + pNNt+1)− R f Pt

= (1 + pθ)(ρθθt + σθεθt+1) + σDε

Dt+1

+(pN(ρNNt + σNεNt+1))− R f Pt

• Conditional Variance

Vart(Rt+1) = (1 + pθ)2σ2θ + σ2D + p2Nσ2N (1)

where pθ = 1/(R f − ρθ). which differs from theequation on p9.

• Conditional mean:(1 + pθ)ρθθt + pNρNNt − R f (pθθt + pNNt)= ((1 + pθ)ρθ − R f pθ)θt + pN(ρN − R f )Nt



Comment 4: Cum Dividend Return and Variance

• cum dividend?

• Dt + Pt linear in θt and noise?

• No qualitative changes to the results.

• But should be consistent.



Other Comments

1 Uniqueness of pN

• Focus on less negative root of pN because it convergeswhen σN → 0

• What about σN →∞? Is the other root more stable?• Not an issue in Wincoop (2006, 2008)

2 Microfoundation of government noise and investorlearning.Footnote 11: Gov possesses private signals.



Other Comments

1 Uniqueness of pN

• Focus on less negative root of pN because it convergeswhen σN → 0

• What about σN →∞? Is the other root more stable?• Not an issue in Wincoop (2006, 2008)

2 Microfoundation of government noise and investorlearning.Footnote 11: Gov possesses private signals.



Other Comments

1 Role of leverage and deleveraging-driven firesale.

2 Empirical predictions or evidence that the govintervention is indeed a priced factor.

3 coordination failures among regulatory agencies. StateCouncil, PBOC, CBRC, CSRC, and CIRC.

4 Online Appendix.



Other Comments




4 Online Appendix.



Other Comments




4 Online Appendix.



Other Comments




4 Online Appendix.



Recap

1 Realistic, important, and timely.

2 More research on the effectiveness and unintendedconsequences of interventions.

3 Empirical studies and policy applications.



Recap






Recap





Documents

China's Model of Managing the Financial SystemChina's Model of Managing the Financial System Markus K. Brunnermeier Michael Sockin Wei Xiong Feb, 2017 ... University of Chicago Booth