Professional Documents
Culture Documents
International Finance
Princeton Initiative 2023
Topics include:
Amplification, peculation of shocks, resilience, financial cycle
Financial stability, spillovers, systemic risk measures
(Un)conventional central bank policy and balance sheet, maturity structure, CBDC
Capital flows
Macro-Finance & Related Fields
FINANCE
Public
Finance
Corporate
DD81, AG
MACRO Lucas HT
Finance
KM97
BGG99 BruSan, HK Asset
Aiyagari
Safe Pricing
HANK Asset I-Theory Merton
of Money FF
Woodford
Intermediary
KW
Finance
MONETARY
Heterogeneous Agents
Lending-borrowing/insuring since agents are different
Poor-rich Rich-poor
Productive Less productive
Less patient Limited direct lending
due to frictions
More patient
Less risk averse More risk averse
More optimistic More pessimistic
Incomplete markets
“natural” leverage constraint (BruSan)
Costly state verification (BGG) state 2
+ Leverage constraints
(no “liquidity creation”)
Exogenous limit (Bewley/Ayagari) state 1
Collateral constraints
Next period’s price (KM) Debt limit
Next periods volatility (VaR, JG) can depend on
prices/volatility
Current price
Risk premia, price of risk
Risk premia = price of risk * (endogenous + exogenous risk)
Precautionary savings
Role for money/safe asset
A L A L
Net worth
Capital Debt Loans
𝑁𝑡
18
Financial Frictions and Distortions
Belief distortions
Match “belief surveys”
Incomplete markets
“natural” leverage constraint (BruSan)
Costly state verification (BGG)
state 2
+ Leverage constraints
(no “liquidity creation”)
Exogenous limit (Bewley/Ayagari)
state 1
Collateral constraints
Next period’s price(KM)
Debt limit
can depend on
Next periods volatility (VaR, JG)
prices/volatility
Current price
Two Sector Model Setup
Expert sector Household sector
Output: Output:
Consumption rate:
Consumption rate:
Investment rate:
Investment rate:
Risk-free debt
Equity, but most hold 21
Two Sector Model Setup
Expert sector Household sector
Output: Output:
Consumption rate:
Consumption rate:
Investment rate:
Investment rate:
𝜅 =1 allocation of
A L 𝜒 =1
accumu- Debt
physical assets
lation
Outside
… risk equity
y
ti lit
drift
drift growth
consumption + investment 𝜂 volatility
ith expectations
net worth
distribution 23
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions g-utility
backward equation
Lo
a. Value fcn. as fcn. of individual investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and prices of risks
4. Numerical model solution 24
a. Transform BSDE for separated value fcn. into PDE
0. Postulate Aggregates and Processes
Individual capital evolution:
Same Brownian
Capital share:
, volatility of process is
Ito product rule: stock price exchange rate
27
The Big Picture
𝜅 =1 allocation of
A L 𝜒 =1
accumu- Debt
physical assets
lation
Outside
… risk equity
y
ti lit
drift
drift growth
consumption + investment 𝜂 volatility
ith expectations
net worth
distribution 30
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions g-utility
backward equation
Lo
a. Value fcn. as fcn. of individual investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and price of risk
4. Numerical model solution 31
a. Transform BSDE for separated value fcn. into PDE
1a. Individual Agent Choice of , ,
Choice of is static problem (and separable) for each
32
1a. Method 3: Martingale Approach – Cts. Time
Expected return:
For risk-free asset, i.e. :
33
1a. Optimal Portfolio Choice - back to our model
Using for capital return (instead of generic asset A)
ut equity issuance
37
2. Law of Motion of Wealth Share
Method 1: Using Ito’s quotation rule
38
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions backward equation
a. Value fcn. as fcn. Lofogindividual
-utility investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and price of risk
4. Numerical model solution 39
a. Transform BSDE for separated value fcn. into PDE
The Big Picture
𝜅 =1 allocation of
A L 𝜒 =1
accu- Debt
physical assets
mulation
Outside
… risk equity
y
ti lit
drift
drift growth
consumption + investment 𝜂 volatility
ith expectations
net worth
distribution 40
3a. CRRA Value Function Applies separately for each type of agent
Martingale Approach: works best in endowment economy
Here: mix Martingale approach with value function (envelop condition)
for individuals
For CRRA/power utility
increase net worth by factor, optimal for all future states
increases by this factor -ratio is invariant in
value function can be written as
Investment opportunity/ “net worth multiplier”
-function turns out to be independent of
41
Change notation from -function to -process
3a. CRRA Value Function: relate to
Next step:
a) Special simple cases
b) replace with
something scale invariant 42
3a. CRRA Value Function: Special Case utility
Consumption choice:
does not matter income and substitution effect cancel out
⇒ 𝑞 𝑡 =𝑞 , ∀ 𝑡
Hence
Capital market
45
4b. Model Solution
Using , , for
46
Numerical example
𝑎=.11 , 𝜌 =5 % , 𝜎 =.1 ,Φ ( 𝜄 )=
log ( 𝜙𝜄+1 )
, 𝜙=10
47
𝜙
Observation of Basak-Cuoco Model
fluctuates with macro shocks, since experts are levered
Price of risk, i.e. Sharpe ratio, is
Debt Loans
Capital Net worth
𝑒 Outside Equity
𝑁 𝑡 equity Capital
𝑒
≥ 𝛼 𝜅𝑡
+ Leverage constraints
(no “liquidity creation”)
Exogenous limit (Bewley/Ayagari)
Collateral constraints
Next period’s price (KM)
𝑒 h
𝜌 ≥𝜌
Friction: Can only issue
Risk-free debt
Equity, but must hold
64
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions backward equation
a. Value fcn. as fcn. of individual investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and price of risk
4. Numerical model solution 65
a. Transform BSDE for separated value fcn. into PDE
The Big Picture
𝜅 allocation of
A L 𝜒≥𝜒
accu- Debt
physical assets
mulation
Outside
… risk equity
y
ti lit
drift
drift growth
consumption + investment 𝜂 volatility
ith expectations
net worth
distribution 66
2. GE: Markov States and Equilibria
Equilibrium is a map
Histories of shocks prices
+ (lots of algebra)
68
2. Amplification Formula: Loss Spiral
Recall
By Ito’s Lemma on
Total volatility
Loss spiral
Market illiquidity (price impact elasticity)
69
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions backward equation
a. Value fcn. as fcn. of individual investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and price of risk
4. Numerical model solution 70
a. Transform BSDE for separated value fcn. into PDE
Solution
Price of capital Amplification
𝑒 𝑒 𝑒 𝑒
𝜅 𝑡 <1 𝜅 𝑡 =1 𝜅 𝑡 <1 𝜅 𝑡 =1
n
regio
n
n
r eg i o
regio
n
al
r eg i o
Norm
is
al
Cris
Norm
is
Cris
Parameters:
71
Volatility Paradox
Comparative Static w.r.t. .01,.05,.1
72
Risk Sharing via Outside Equity
Comparative Static w.r.t. Risk sharing .1,.2, .5
(skin the game constraint)
73
Market Liquidity
Comparative static w.r.t.
74
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions backward equation
a. Value fcn. as fcn. of individual investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and price of risk
4. Numerical model solution 75
a. Transform BSDE for separated value fcn. into PDE
From & to Stationary Distribution
Drift and Volatility of
𝑒 𝑒
𝜂 =𝛼 𝜅
“Steady state” ,
Case 1a
Case 2a
Case 2a
Case 1b
Case 1b
𝜂 𝜂
Experts’ skin in the game Experts’ skin in the game
constraint binds, 76
constraint binds, 76
5. Kolmogorov Forward Equation
Given an initial distribution
the density diffusion follows PDE
77
Solving MacroModels Step-by-Step
0. Postulate aggregates, price processes & obtain return processes
1. For given -ratio and SDF processes for each finance block
a. Real investment + Goods market clearing (static)
Toolbox 1: Martingale Approach, HJB vs. Stochastic Maximum Principle Approach
b. Portfolio choice + Asset market clearing or
Asset allocation & risk allocation
Toolbox 2: “price-taking social planner approach” – Fisher separation theorem
c. “Money evaluation equation”
Toolbox 3: Change in numeraire to total wealth (including SDF)
2. Evolution of state variable (and ) forward equation
3. Value functions backward equation
a. Value fcn. as fcn. of individual investment opportunities
Special cases: log-utility, constant investment opportunities
b. Separating value fcn. into
c. Derive -ratio and price of risk
4. Numerical model solution 78
a. Transform BSDE for separated value fcn. into PDE
5. Stationary Distribution
Stationary distribution of
𝑒 𝑒
𝜂 =𝛼 𝜅
81
5. Fan chart and distributional impulse response
Starts at stationary distribution
Simulate a shock at quantile of original Brownian
shock () for a period of .
Converges back to stationary distribution
82
5. Density Diffusion
Starts at stationary distribution
Simulate a shock at quantile of original Brownian shock ()
for a period of .
Converges back to stationary distribution
83
5.Density Diffusion Movies
84
5. Distributional Impulse Response
Difference between path with and without shock
Difference converges to zero in the long-run
𝜎 =0.15
85
Desired Model Properties
Normal regime: stable around steady state
Experts are adequately capitalized
Experts can absorb macro shock
Endogenous risk and price of risk
Fire-sales, liquidity spirals, fat tails
Spillovers across assets and agents
Market and funding liquidity connection
SDF vs. cash-flow news
Volatility paradox
Financial innovation less stable economy
(“Net worth trap” double-humped stationary distribution)
87