## _wp1516

## Source details

**Canonical URL:** [_wp1516](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1516.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1516.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1516.pdf.json)

---

### I. Purpose and main questions
- Research focus: interaction between financial markets and aggregate fluctuations since the dot.com bubble and the 2007-8 global financial crisis.
- Key policy question: whether central banks should move the policy rate in response to asset/stock price fluctuations.
- Benchmark: conventional New Keynesian view—explicit response to stock prices restricts determinacy of the REE and can generate sunspot-driven indeterminacy; requires a reinforced Taylor principle (sufficiently active response to inflation).
- Departure: introduces non-Ricardian households (Blanchard-Yaari/perpetual-youth) so that stock prices are non-redundant for business cycle fluctuations.
- Mechanism: Financial Wealth Channel (FWC) — turnover between long-time traders and newcomers creates a non-degenerate distribution of financial wealth and consumption; stock prices affect consumption through wealth effects and thus feed back into real activity.
- Tractable extremes of turnover:
  - As turnover → 0, model collapses to benchmark New Keynesian (infinitely-lived traders).
  - Turnover = 1 implies one-period-lived agents (rule-of-thumb consumers).
- Evaluation criteria for policy rules: equilibrium determinacy and learnability (E-stability of Evans and Honkapohja (2001)).
- Main qualitative result: with sufficiently positive turnover (strong FWC), adding a positive but not excessive response to stock prices in the interest-rate rule can enlarge the policy space for which the equilibrium is determinate and learnable; the Taylor principle is no longer strictly necessary.
- Additional amplifiers: effects stronger with (i) lower elasticities of substitution across differentiated products and/or (ii) more sluggish real wage adjustment.

### II. Model structure and key equations
- Population and turnover:
  - Perpetual-youth Blanchard-Yaari structure: cohorts survive with probability 1−; entry and exit equal; total population size one.
  - Turnover rate denoted ; newcomers enter with no financial wealth.
- Household preferences and parameters:
  - Lifetime utility: E_τ Σ_{k=0}^∞ ^k (1−)^k [ln C_{j,τ+k} +  ln(1−N_{j,τ+k})], with ∈(0,1), ∈[0,1), >0.
- Financial assets and wealth:
  - Agents hold state-contingent bonds B_{j,τ+1} and equity S_{j,τ+1}(i) with real price Q_τ(i).
  - Long-time traders’ nominal financial wealth: A_{j,τ} = [ B_{j,τ} + Π_τ ∫_0^1 ( Q_τ(i) + Δ_τ(i) ) S_{j,τ}(i) di ] for j<τ; newcomers j=τ have A_{j,τ}=0.
  - Aggregate scaled wealth Ω_{j,τ} ≡ A_{j,τ} / (1−) for j<τ.
- Key first-order conditions and identities:
  - Marginal rate:  C_{j,τ} / (1−N_{j,τ}) = Ω_τ / Π_τ. (equation (6))
  - Stochastic discount factor: F_{τ,τ+1} =  (C_{j,τ}/Π_τ) / (C_{j,τ+1}/Π_{τ+1}). (equation (7))
  - Equity pricing: Π_τ Q_τ(i) = E_τ [ F_{τ,τ+1} Π_{τ+1} ( Q_{τ+1}(i) + Δ_{τ+1}(i) ) ]. (equation (8))
  - Individual consumption linear relation: Π_τ C_{j,τ} = [1 − (1−)] ( Ω_{j,τ} + H_{j,τ} ). (equation (10))
- Aggregation and the Financial Wealth Channel:
  - Aggregate wealth: Ω_τ = Π_τ ( Q_τ + Δ_τ ); aggregate consumption: Π_τ C_τ = [1 − (1−)] ( Ω_τ + H_τ ). (equation (13))
  - Key aggregated identity: (1−)/(1−(1−)) C_τ =  Q_τ + (1−)/(1−(1−)) E_τ [ F_{τ,τ+1} Π_{τ+1} C_{τ+1} ]. (equation (18))
  - Interpretation: term  Q_τ is the FWC — stock market wealth affects intertemporal consumption profile when >0; with =0 stock prices do not affect consumption profile.

### III. Log-linearized blocks and policy rule
- Notation: lower case letters denote log deviations.
- Demand-side (aggregate IS curve):
  - y_τ = (1/(1+ψ)) E_τ y_{τ+1} + (ψ/(1+ψ)) q_τ − (1/(1+ψ)) ( r_τ − E_τ π_{τ+1} ). (equation (20))
  - ψ ≡  / [1−(1−)] × (1−)/(1+ρ) ρ > 0, with ρ steady-state real interest rate. (equation (21))
  - Properties: ψ(0) = 0 and ψ'() > 0; ψ strictly decreasing in .
- Supply-side (New Keynesian Phillips curve):
  - π_τ = ˜ E_τ π_{τ+1} + ● (1+Ω) ( y_τ − z_τ ), where ● ≡ (1−)(1− ˜)/, Ω ≡ N/(1−N), and ˜ ≡ /(1+ψ). (equation (23))
  - Positive  makes economy de facto less forward-looking (˜ < ).
- Financial-side (stock price equation):
  - q_τ = ˜ E_τ q_{τ+1} − λ E_τ y_{τ+1} − ( r_τ − E_τ π_{τ+1} ) + ξ z_τ,
    where ξ ≡ [1−˜] (−1)(1+Ω)_z and λ ≡ [1−˜][(−1)(1+Ω) − 1]. (equation (24) and (25))
- Monetary policy rule (instrumental forward-looking Taylor rule):
  - r_τ = φ_π E_τ π_{τ+1} + φ_q E_τ q_{τ+1}, with φ_π > 0 and φ_q > 0. (equation (26))
  - Terminology: φ_π>1 is active (Taylor principle); φ_π<1 is passive.
- Compact system:
  - x_τ = Γ E_τ x_{τ+1} + Θ z_τ, where x_τ = [ y_τ, π_τ, q_τ ]', Γ and Θ depend on structural and policy parameters. (equation (27))

### IV. Methodology: determinacy and learnability
- Determinacy: Blanchard and Kahn (1980) conditions — REE locally determinate iff eigenvalues of Γ meet conditions (predetermined vs non-predetermined count).
- Learnability: E-stability (Evans and Honkapohja (2001)).
  - MSV-FE (Minimal State Variable representation of a fundamental equilibrium) x_τ = N z_τ.
  - CF-SSE (Common Factor representation of stationary sunspot equilibria) x_τ = N z_τ + G ε_τ for indeterminate cases with real roots.
- Focus: parameter regions in (φ_π, φ_q) space where REE is determinate and MSV-FE is E-stable; identification of learnable sunspot equilibria when φ_q is very large.
- Key analytical implication: positive turnover  (stronger FWC) can make positive responses to stock prices beneficial for determinacy and learnability, potentially relaxing the Taylor principle.

### V. Main analytical results (Propositions and interpretations)
- Proposition 1 (case  = 0 — no FWC):
  - Assumption:  = 0.
  - Determinacy: REE locally determinate if and only if two inequalities stated in the source (expression (28)) hold.
  - E-stability of MSV-FE: MSV-REE E-stable iff φ_π > 1 + (/●) φ_θ. (expression (29))
  - E-stability of CF-SSE: conditions (30)-(31) specify when E-stable CF representations of SSE exist.
  - Interpretation:
    - With  = 0, an explicit policy response to stock prices (φ_θ > 0) can be destabilizing for determinacy and learnability.
    - There exists an upper bound φ*_θ ≡ 1+ 1+ on the response to stock prices above which equilibrium is always indeterminate for any policy rule.
    - Indeterminacy can coexist with learnable MSV-FE or learnable sunspot equilibria in parts of parameter space.
- Proposition 2 (sufficient condition for learnability when FWC present):
  - Definitions:  and  from equations (21) and (25). Let _* ∈ (0,1) be unique solution to  = , so  >  for  > _*.
  - If  > _*, then a sufficient condition for learnability of the MSV-FE is:
    - φ_π > Φ^ ≡ 1 + (1 − ̃)( − )
      
      (1 − ̃ + ) φ_θ  (equation (32))
  - Properties of Φ^:
    - Since  >  when  > _*, Φ^ is strictly decreasing in φ_θ.
    - Φ^ < 1, so any active response to inflation satisfies the condition; stock-price targeting is not necessarily a threat under sufficient FWC.
  - Economic intuition (long-run Taylor principle):
    - Rewriting yields condition (33): φ_π + (1 − ̃)( − )
      
      (1 − ̃ + ) φ_θ − 1 > 0.
    - The left-hand side equals the long-run change in the real interest rate following a permanent 1 percent increase in inflation under the rule. The factor (1 − ̃)( − )
      
      (1 − ̃ + ) is the permanent percent change in the stock price index following a 1 percent permanent inflation increase.
    - If  sufficiently large ( > _*), a positive φ_θ helps obtain the long-run real rate increase even if φ_π < 1.
  - Contrast with no FWC: when  = 0 the corresponding expression reduces to −  and the standard active response requirement applies.

### VI. Quantitative illustrations and thresholds
- One period corresponds to a quarter.
- Benchmark parameter choices for graphical displays:
  -  = 0.99.
  - Inverse Frisch labor elasticity  = 0.25 (elasticity = 4).
  - Phillips-curve coefficient  = 0.019.
  - Elasticity of substitution  = 2.3.
- Turnover examples and empirical context:
  -  = 0.05 used as an illustrative positive turnover (intermediate between low-end and high-end estimates).
  - Low-end turnover estimates: 0.005 to 0.015.
  - Kozora (2010): institutional investor horizons between 19 and 25 months (1990-2007).
  - Cella et al. (2013): quarterly portfolio turnover ~17% implying ~6 quarters ( ≈ 0.17? — note: source reports ~17% turnover).
  - Castelnuovo and Nisticò (2010) posterior mean  = 0.13 (planning horizon slightly shorter than 2.5 years).
- Examples of threshold and bounds:
  - _* can be as low as ~0.02 in benchmark calibration (implying a downward-sloping determinacy frontier for φ_θ).
  - For  < _* there exists an upper bound φ_θ^max on policy response to stock price given φ_π:
    - φ_θ^max ≡
      
      (1 − ̃ + )
      (1 − ̃)
      ( − )
      (φ_π − 1)  (equation (34))
    - Example ( = 2.3): φ_θ^max = 0.6 when  = 0.
    - φ_θ^max increases rapidly with  and tends to infinity as  → _* (~0.02 in that example).
- Plausible  range considered: [0,0.13].
- Example (, ) pairs used to generate similar effects: (3, 0.5), (4, 2/3), (5, 0.75).

### VII. Extensions and robustness
- Forward-looking rule responding to expected future output:
  - Rule: r_τ = φ_π E_τ π_{τ+1} + φ_θ E_τ q_{τ+1} + φ_ψ E_τ y_{τ+1} with φ_ψ = 0.5/4 (Taylor (1993) baseline).
  - Positive response to output improves determinacy with or without FWC; does not qualitatively change need for sufficiently positive  for stock-price responses to enlarge determinacy.
- Contemporaneous rules:
  - Rule: r_τ = φ_π π_τ + φ_θ q_τ.
  - Light gray regions (indeterminacy with learnable MSV-FE) disappear.
  - Two outcomes only: determinate REE with learnable MSV-FE, or indeterminate REE with non-learnable MSV-FE and CF-SSE.
  - If  > _* there is no upper bound on φ_θ: for every active policy rule the REE is determinate and MSV-FE is E-stable for any φ_θ ≥ 0.
- Real wage rigidity:
  - Real wage: ω_τ − π_τ = (1 − )[(1 + ) ψ_τ −  ζ_τ], with  ∈ [0,1] indexing rigidity.
  -  becomes:  ≡ (1 − ̃)[( − 1) (1 + )(1 − )^{-1}].
  - Higher  lowers  and thus lowers threshold _* needed for  > , enhancing scope for FWC to overturn conventional wisdom.
- Demand elasticity and real wage rigidity (Figures 4 and 5 implications):
  - _* strictly increasing in  for any .
  - Increasing  shifts _* lower, making  > _* more likely within plausible  range.
  - Numerical ranges where plausible  > _*:
    - For  = 0.5: non-empty for 2 <  < 4.
    - For  = 2/3: non-empty for roughly 2 <  < 5.5.
    - For  = 0.75: non-empty for 2 <  < 7.
  - Example: for  = 5, _* equals 0.05 if  = 2/3, and drops below 0.01 if  = 0.75.
- Overall robustness: results hold across alternative monetary policy specifications and are quantitatively stronger with lower  and/or higher .

### VIII. Policy implications and scenarios
- Conventional wisdom reaffirmed when  = 0:
  - Responding to stock prices increases equilibrium indeterminacy; larger φ_θ requires larger φ_π for uniqueness.
  - There exists an upper bound φ*_θ on acceptable stock-price response.
- When FWC present ( sufficiently large):
  - A positive response to stock prices in a forward-looking interest rate rule can enlarge policy space where equilibrium is determinate and MSV-FE is learnable.
  - The Taylor principle (φ_π > 1) is not strictly necessary for learnability if  > _*; a positive φ_θ can substitute partially for an active inflation response by raising the long-run real rate via wealth-channel feedbacks.
- Cautionary scenario:
  - If response to stock prices is too large, policy can generate a continuum of stationary sunspot equilibria; learnable CF-SSE may exist, implying potential macro instability driven by extrinsic uncertainty.
- Practical guidance:
  - For policy design, consider the strength of the FWC (turnover ) and structural parameters (, , ) when evaluating interest-rate rules that respond to stock prices.
  - In economies with measurable turnover (examples:  = 0.05,  = 0.13) and/or significant real wage rigidity, modest positive φ_θ may be stabilizing.
  - Upper bounds on φ_θ are relaxed as  increases; compute φ_θ^max from equation (34) when  < _*.

### IX. Suggested extensions
- Introduce Blanchard-Yaari structure with credit frictions and a financial accelerator to study interaction of demand- and supply-side asset price fluctuations.
- Study optimal monetary policy rules (second-order welfare approximations) that incorporate financial stability concerns.
- Explore consequences of IH learning (the authors plan to study IH learning in a separate project).

### X. Exact numeric examples and thresholds (preserved)
- Turnover rate examples:  = 0.05;  can be as low as 0.02 to produce a downward-sloping determinacy frontier; low-end estimates 0.005 to 0.015; Castelnuovo and Nisticò (2010) posterior mean  = 0.13.
- Benchmark elasticity:  = 2.3.
- Example upper bound: φ_θ^max = 0.6 when  = 0 for  = 2.3.
- Plausible  range considered: [0,0.13].
- Example (, ) pairs: (3, 0.5), (4, 2/3), (5, 0.75).

*Italic source: _wp1516 - section V we will discuss how a significant FWC can also obtain for much higher elasticities if combined*

### References .............................................................................................................

### _wp1516 - References .............................................................................................................

### I. Introduction — purpose and main questions
- Research focus: interaction between financial markets and aggregate fluctuations since the dot.com bubble and the 2007-8 global financial crisis.
- Key policy question: whether central banks should move the policy rate in response to asset/stock price fluctuations.
- Conventional wisdom (benchmark New Keynesian): an explicit response to stock prices restricts determinacy of the REE and can generate sunspot-driven indeterminacy; requires a reinforced Taylor principle (sufficiently active response to inflation).
- This paper’s departure: introduces non-Ricardian households (Blanchard-Yaari/perpetual-youth) into a New Keynesian DSGE model so that stock prices are non-redundant for business cycle fluctuations.
- Financial wealth channel (FWC) mechanism: turnover between long-time traders and newcomers creates a non-degenerate distribution of financial wealth and consumption; stock prices affect consumption through wealth effects and thus feed back into real activity.
- Tractable extremes of turnover:
  - As turnover → 0, model collapses to benchmark New Keynesian (infinitely-lived traders).
  - Turnover = 1 implies one-period-lived agents (rule-of-thumb consumers).
- Evaluation criteria for policy rules: equilibrium determinacy and learnability (E-stability of Evans and Honkapohja (2001)); both are desirable for policy relevance.
- Main qualitative result: with a sufficiently positive turnover (strong FWC), adding a positive but not excessive response to stock prices in the interest-rate rule can enlarge the policy space for which the equilibrium is determinate and learnable; the Taylor principle is no longer strictly necessary.
- Additional findings: effects are more prominent with (i) lower elasticities of substitution across differentiated products comparable to Broda and Weinstein (2006) estimates, and/or (ii) more sluggish real wage adjustment comparable to Blanchard and Gali (2007) and Uhlig (2007).
- Also highlights possibility of learnable sunspot equilibria when response to stock prices is so large that it induces indeterminacy.

### II. The Model — structure and key equations
- Population and turnover:
  - Perpetual-youth Blanchard-Yaari structure: cohorts survive between periods with probability 1−; entry and exit equal, total population size one.
  - Turnover rate denoted ; newcomers enter with no financial wealth.
- Household lifetime utility and parameters:
  - Lifetime utility: E_tau Σ_{k=0}^∞ ^k (1−)^k [ln C_{j,τ+k} +  ln(1−N_{j,τ+k})], with ∈(0,1), ∈[0,1), >0.
- Financial assets and individual wealth definitions:
  - Agents hold state-contingent bonds B_{j,τ+1} and equity S_{j,τ+1}(i) with real price Q_τ(i).
  - Long-time traders’ nominal financial wealth A_{j,τ} = [ B_{j,τ} + Π_τ ∫_0^1 ( Q_τ(i) + Δ_τ(i) ) S_{j,τ}(i) di ] for j<τ.
  - Aggregate form: Ω_{j,τ} ≡ A_{j,τ} / (1−) for j<τ; newcomers have A_{j,τ}=Ω_{j,τ}=0 for j=τ.
- Key first-order conditions and budget dynamics:
  - Marginal rate:  C_{j,τ} / (1−N_{j,τ}) = Ω_τ / Π_τ. (equation (6))
  - Stochastic discount factor: F_{τ,τ+1} =  (C_{j,τ}/Π_τ) / (C_{j,τ+1}/Π_{τ+1}). (equation (7))
  - Equity pricing: Π_τ Q_τ(i) = E_τ [ F_{τ,τ+1} Π_{τ+1} ( Q_{τ+1}(i) + Δ_{τ+1}(i) ) ]. (equation (8))
  - Individual consumption linear relation: Π_τ C_{j,τ} = [1 − (1−)] ( Ω_{j,τ} + H_{j,τ} ). (equation (10))
- Production:
  - Retail sector: Y_τ = [ ∫_0^1 Y_τ(i)^{(−1)/} di ]^{/(−1)} with elasticity >1.
  - Wholesale sector: monopolistic competition, Calvo pricing with probability 1− to reset price; production Y_τ(i)=Z_τ N_τ(i), with z_τ ≡ ln Z_τ following AR(1): z_τ = _z z_{τ−1} + _τ.
  - Real marginal cost: MC_τ = (1−) Ω_τ / (Z_τ Π_τ) where  is a labor subsidy.
  - Steady-state dividends Δ are decreasing in  and ; steady-state dividends equal to 1/[(1+)].
- Aggregation and the Financial Wealth Channel:
  - Aggregate wealth Ω_τ = Π_τ ( Q_τ + Δ_τ ); aggregate consumption relation Π_τ C_τ = [1 − (1−)] ( Ω_τ + H_τ ). (equation (13))
  - Key aggregated identity (after manipulation): (1−)/(1−(1−)) C_τ =  Q_τ + (1−)/(1−(1−)) E_τ [ F_{τ,τ+1} Π_{τ+1} C_{τ+1} ]. (equation (18))
  - Interpretation: the term  Q_τ is the FWC — stock market wealth affects the intertemporal consumption profile when >0; with =0 stock prices do not affect consumption profile.

### III. Log-linearized economy and equilibrium — reduced form blocks
- Linearization conventions: lower case letters denote log deviations; model reduced to demand-side, supply-side, and financial-side blocks plus an interest-rate rule.
- Demand-side (aggregate IS curve):
  - y_τ = (1/(1+ψ)) E_τ y_{τ+1} + (ψ/(1+ψ)) q_τ − (1/(1+ψ)) ( r_τ − E_τ π_{τ+1} ). (equation (20))
  - Composite coefficient ψ ≡  / [1−(1−)] × (1−)/(1+ρ) ρ > 0, where ρ denotes the steady-state real interest rate. (equation (21))
  - Properties: ψ(0) = 0 and ψ'() > 0. (equation (22))
  - Interpretation: ψ captures the strength of the FWC; higher  raises ψ and thus strengthens the FWC (term ψ/(1+ψ) q_τ).
  - ψ is strictly decreasing in : lower  (higher market power) increases the FWC because steady-state financial wealth is larger.
- Supply-side (New Keynesian Phillips curve):
  - π_τ = ˜ E_τ π_{τ+1} + ● (1+Ω) ( y_τ − z_τ ), where ● ≡ (1−)(1− ˜)/, Ω ≡ N/(1−N), and ˜ ≡ /(1+ψ). (equation (23))
  - Positive turnover  makes the economy de facto less forward-looking (˜ < ), increasing the role of current marginal costs in inflation.
- Financial-side (stock price equation):
  - q_τ = ˜ E_τ q_{τ+1} − λ E_τ y_{τ+1} − ( r_τ − E_τ π_{τ+1} ) + ξ z_τ, where ξ ≡ [1−˜] (−1)(1+Ω)_z and λ ≡ [1−˜][(−1)(1+Ω) − 1]. (equation (24) and (25))
  - Current stock price depends positively on E_τ q_{τ+1} and z_τ, negatively on expected future output and the real interest rate term.
- Monetary policy rule (instrumental forward-looking Taylor rule):
  - r_τ = φ_π E_τ π_{τ+1} + φ_q E_τ q_{τ+1}, with φ_π > 0 and φ_q > 0. (equation (26))
  - Terminology: rule with φ_π>1 is active (satisfies simple Taylor principle); φ_π<1 is passive.
- Model compact form:
  - x_τ = Γ E_τ x_{τ+1} + Θ z_τ, where x_τ = [ y_τ, π_τ, q_τ ]', Γ and Θ depend on structural and policy parameters. (equation (27))

### IV. Equilibrium determinacy and learnability — methodology and focus
- Objective: examine local determinacy and learnability of the REE with respect to policy parameters φ_π and φ_q.
- Determinacy approach: rely on Blanchard and Kahn (1980) conditions.
- Learnability approach: focus on E-stability (Evans and Honkapohja (2001)); analyze two representations:
  - MSV-FE: Minimal State Variable representation of a fundamental equilibrium (x_τ = N z_τ); may exist under determinacy or indeterminacy.
  - CF-SSE: Common Factor representation of stationary sunspot equilibria when equilibrium is indeterminate with real roots (x_τ = N z_τ + G ε_τ, where ε_τ is an autoregressive sunspot).
- The paper uses these frameworks to identify parameter regions where policy rules induce a determinate and learnable REE, including potential learnable sunspot equilibria when φ_q is very large.
- Analytical implication emphasized: positive turnover  (stronger FWC) can make a positive response to stock prices in the rule beneficial for determinacy and learnability, potentially relaxing the necessity of the Taylor principle.

* _wp1516 - References .............................................................................................................*

### Section V.

### Section V.

### Model setup and fiscal side
- The government levies lump-sum taxes to pay the labor subsidy to firms, balancing period-by-period its budget constraint; therefore the accumulation of public debt and the government budget constraint can be abstracted from in the determinacy and learnability analyses.
- Section V presents results for rules responding to expected future output, as well as for contemporaneous rules responding to current-period inflation and stock price levels.
- The benchmark New Keynesian model is obtained by setting  = 0 in the reduced form system described by equations (20)-(25); in this case, the FWC (financial wealth channel) is absent.

### Proposition 1 — determinacy and E-stability results (case  = 0)
- Assumption: no turnover in financial markets:  = 0.
- Determinacy.
  - The REE is locally determinate if and only if:
    -  < 1+ 1+
    - and 1+    <  < 1+ 2(1+)  − 2+    (expression (28) as stated)
- E-Stability of MSV-FE.
  - The MSV-REE is E-stable if and only if
    -  > 1+     (expression (29) as stated)
- E-Stability of CF-SSE.
  - There exist E-stable CF representations of SSE if and only if:
    -  > 1+ 2(1+)  − 2+   for  ≤ 1+ 1+  (expression (30) as stated)
    -  > 1+    for  > 1+ 1+  (expression (31) as stated)
- Proof: See Appendix C.1. (as referenced in the source)

### Interpretation and implications
- The proposition shows that, in the absence of the FWC ( = 0), an explicit policy response to stock prices ( > 0) can be destabilizing from the perspective of determinacy and learnability.
- There exists an upper bound ∗ ≡ 1+ 1+ on the response to stock prices above which the equilibrium is always indeterminate, for any active (and passive) policy rule.
- The analysis assesses learnability of both fundamental and non-fundamental solutions:
  - There exist policy parameterizations for which the equilibrium is indeterminate but the MSV-FE (fundamental equilibrium) is learnable — e.g., for  > 1+ 1+ and  > 1+   .
  - Such parameterizations can also induce learnable sunspot equilibria (E-stable CF-SSE), implying that macroeconomic stability can be at stake because of extrinsic uncertainty.
- The proposition extends previous results (e.g., Carlstrom and Fuerst (2007)) in two dimensions:
  - It considers a forward-looking specification, which is generally more prone to indeterminacy.
  - It evaluates learnability of both fundamental and non-fundamental solutions.

### Numerical calibration and graphical representation
- One period in the model is taken to correspond to a quarter.
- Parameter choices reported for the graphical display (left panel of Figure 1):
  - Discount factor  = 0.99.
  - Inverse Frisch labor elasticity,  = 0.25 (i.e., an elasticity equal to 4).
  - The elasticity of current inflation to marginal costs in the Phillips curve,  = 0.019 (chosen via the Calvo probability of price rigidity as in Carlstrom and Fuerst (2007)).
  - Elasticity of substitution across differentiated goods,  = 2.3 (mid-point between reported median elasticity values 2.5 and 2.1 from Broda and Weinstein (2006)).
- Rationale: monopolistically competitive firms enjoy a considerable degree of market power under  = 2.3, which in the set-up generates sufficiently large profits/dividends and gains from equity holdings.

*Source: _wp1516 - Section V.*

### section V we will discuss how a significant FWC can also obtain for much higher elasticities if combined

### _wp1516 - section V we will discuss how a significant FWC can also obtain for much higher elasticities if combined

### Determinacy and E-stability under a forward-looking rule (summary of Figure 1)
- An increase in the policy coefficient _θ diminishes the policy space where the equilibrium is determinate (white area) and where the MSV-FE is E-stable (white and light gray areas labeled ES-MSV-FE).
- Raising _θ enlarges the possibility of:
  - an indeterminate equilibrium (dark gray area), and
  - an E-unstable MSV-FE (dark gray area labeled EU-MSV-FE).
- Within regions of indeterminacy with real eigenvalues, E-stable CF-SSE (learnable common-factor sunspot equilibria) exist in the light gray area (where the MSV-FE is also learnable) but not in the dark gray area.
- With a positive turnover in financial markets (chosen =005 in the right panel of Figure 1):
  - A positive response to stock prices enlarges the policy space where the equilibrium is determinate (white area) and where the MSV-FE is learnable (white and light gray areas).
  - The lower determinacy/E-stability frontier becomes downward-sloping in _θ.
  - A rather mild FWC can overturn the conventional wisdom: responding to stock prices does not necessarily threaten determinacy/E-stability if  is sufficiently positive.
- Quantitative illustrative points highlighted in the discussion:
  - =005 is intermediate between low-end estimates (0005 to 0015) and high-end estimates from portfolio turnover.
  - Kozora (2010) reports institutional investor horizons between 19 and 25 months (1990-2007); Cella et al. (2013) report a quarterly portfolio turnover of about 17% implying ~6 quarters; Castelnuovo and Nisticò (2010) posterior mean =013 (planning horizon slightly shorter than 25 years).

### Proposition 2 — sufficient condition for learnability of the MSV-FE (key formulae)
- Definitions recalled:  and  from equations (21) and (25).
- Let _* ∈ (0,1) be the unique solution to  = , such that  >  for  > _*.
- If the turnover rate  is larger than _* (i.e., the FWC is sufficiently strong) then a sufficient condition for the learnability of the MSV-FE is:
  - _ > Φ^ ≡ 1 + (1 − ̃)( − )
    
    (1 − ̃ + ) _θ  (equation (32))
- Properties of the threshold Φ^:
  - Since  >  when  > _*, the threshold Φ^ is strictly decreasing in _θ.
  - Φ^ < 1, so condition (32) holds for any active response to inflation; stock price targeting is not necessarily a threat to aggregate stability under sufficient FWC.
- Comparison with the no-FWC case:
  - Without FWC the threshold is 1 +   _θ (always increasing in _θ).
  - With FWC and  > _*, the threshold becomes Φ^ (strictly decreasing in _θ).

### Economic intuition: long-run Taylor principle and the role of the FWC
- Rewriting (32) yields:
  - _ + (1 − ̃)( − )
    
    (1 − ̃ + ) _θ − 1 > 0  (equation (33))
- Interpretation:
  - The left-hand side corresponds to the long-run change in the real interest rate following a permanent 1 percent increase in inflation under the interest rate rule (26).
  - The factor (1 − ̃)( − )
    
    (1 − ̃ + ) is the permanent percent change in the stock price index following a 1 percent permanent increase in inflation.
  - When (33) holds, the rule induces a long-run increase in the real interest rate after a permanent inflation rise, correcting sunspot-driven increases in inflation over time.
- Role of FWC:
  - If  is sufficiently large ( > _*), a 1 percent increase in inflation implies a permanent (1 − ̃)( − )
    
    (1 − ̃ + ) percent increase in the stock price index, so a positive _θ helps obtain the long-run real rate increase even if _ < 1 (a passive response to inflation).
  - By contrast, when  = 0 (no FWC), the corresponding expression reduces to −  and the standard requirement that the response to inflation must be sufficiently active applies.

### Upper bound on _θ for given _ when  < _* (Figure 2 and equation (34))
- If  < _* (hence  < ), the inequality in (32) holds for:
  - _θ < _θ^max ≡
    
    (1 − ̃ + )
    (1 − ̃)
    ( − )
    (_ − 1)  (equation (34))
  - Interpretation: For a given active response to inflation (_ > 1), there exists an upper bound _θ^max on the policy response to the stock price.
- Properties of _θ^max (Figure 2):
  - _θ^max is monotonically increasing in .
  - Example (benchmark parametrization  = 23):
    - _θ^max = 06 if  = 0.
    - _θ^max increases rapidly as  becomes positive and tends to infinity as  approaches _* (~002 in that case).
  - Even when  is insufficient to overturn the conventional wisdom, a larger  relaxes the upper bound on _θ consistent with determinacy/E-stability.

### Extensions and robustness (section V)
- Responding to output:
  - Consider forward-looking rule also responding to expected future output: ρ_τ = _ E_τ _τ+1 + _θ E_τ θ_τ+1 + _ψ E_τ ψ_τ+1 with _ψ = 05/4 (Taylor (1993) baseline).
  - A positive response to output improves determinacy with or without the FWC, but does not qualitatively change the need for a sufficiently positive  for stock price responses to enlarge determinacy.
- Contemporaneous rules:
  - For contemporaneous interest rate rule ρ_τ = _ _τ + _θ θ_τ the light gray areas (indeterminacy with learnable MSV-FE) disappear.
  - Two possible outcomes: determinate REE with learnable MSV-FE, or indeterminate REE with non-learnable MSV-FE and CF-SSE.
  - For  > _* there is no upper bound on _θ: for every active policy rule the REE is determinate and MSV-FE is E-stable for any _θ ≥ 0.
  - Main message preserved: sufficiently positive  overturns conventional wisdom.
- Real wage rigidity:
  - Introducing real wage rigidity where the real wage is a weighted average of a notional wage (weight 1 − ) and a wage norm (weight ) yields:
    - (log) real wage: ω_τ − π_τ = (1 − )[(1 + ) ψ_τ −  ζ_τ]
    - Parameter  ∈ [0,1] indexes real wage rigidity;  = 0 fully flexible,  = 1 constant real wage.
  - The parameter  (from (25)) becomes:
    -  ≡ (1 − ̃)[( − 1) (1 + )(1 − )^{-1}].
  - A higher  lowers , which lowers the threshold _* above which  >  (needed to overturn conventional wisdom).
  - Thus, real wage rigidity enhances the scope for FWC to overturn conventional wisdom.

### The importance of demand elasticity and real wage rigidities (Figures 4 and 5)
- Figure 4 (threshold _* as function of  for different ):
  - The threshold _* is strictly increasing in  for any .
  - Increasing  shifts _* toward the lower-right, making  > _* more likely.
  - Numerical illustrations for plausibility ranges (restricted to [0,013] for ):
    - For  = 05, the range of plausible  above _* is non-empty for 2 <  < 4.
    - For  = 2/3, non-empty for roughly 2 <  < 55.
    - For  = 075, non-empty for 2 <  < 7.
  - Example: for  = 5, threshold _* equals benchmark 005 if  = 2/3, and drops below 0.01 if  = 075.
- Figure 5 (determinacy/E-stability with real wage rigidity):
  - For selected (, ) pairs:
    - a)  = 3 and  = 05
    - b)  = 4 and  = 2/3
    - c)  = 5 and  = 075
  - With no turnover (top panels) and with  = 005 (bottom panels), a positive turnover makes stock-price targeting beneficial for determinacy/E-stability.
- Overall conclusion from extensions:
  - The FWC can overturn the conventional wisdom in economies with:
    - lower demand elasticities (higher market power) and more flexible labor markets, and/or
    - higher demand elasticities but more rigid labor markets.

### Main conclusions (section VI)
- In the benchmark New-Keynesian model, explicitly responding to stock prices increases equilibrium indeterminacy: larger _θ requires larger _ for uniqueness.
- Introducing a financial wealth channel (FWC) via turnover among non-Ricardian heterogeneous portfolio holders changes this:
  - Under suitable conditions, a positive response to stock prices in a forward-looking interest rate rule enlarges the policy space where the equilibrium is determinate and the MSV-FE is learnable.
  - Thus, the FWC can overturn the conventional wisdom that central banks should not respond to stock prices.
  - However, if the response to stock prices is too strong, the policy can generate a continuum of stationary sunspot equilibria with learnable CF representations.
- Robustness:
  - Results hold under alternative monetary policy specifications (contemporaneous, forward-looking responding to output).
  - Quantitative importance increases with less competitive goods markets (lower ) and/or more rigid labor markets (higher ).
- Suggested extensions:
  - Introduce Blanchard-Yaari structure with credit frictions and a financial accelerator to study interaction of demand- and supply-side asset price fluctuations.
  - Study optimal monetary policy rules (second-order welfare approximations) that incorporate financial stability concerns.

- Key exact parameter values and thresholds mentioned:
  - Turnover rate examples:  = 005;  can be as low as 002 to produce a downward-sloping determinacy frontier; low-end estimates 0005 to 0015; high-end posterior mean  = 013 (Castelnuovo and Nisticò (2010)).
  - Benchmark elasticity:  = 23 (used in numerical illustration).
  - Example upper bound: _θ^max = 06 when  = 0 for  = 23.
  - Plausible  range considered: [0,013].
  - Examples of (, ) that generate effects similar to Figure 1: (3, 05), (4, 2/3), (5, 075).

*Italic source: _wp1516 - section V we will discuss how a significant FWC can also obtain for much higher elasticities if combined*

### conclusions one would get under the EE approach. We plan to explore the consequences of having IH learning in our

### C. Proofs

### C.1. ProofofProposition1
- Reduced-form equilibrium system (27) with turnover rate =0 (such that, =0 and ̃=) yields
  Γ ≡
  ⎡
  ⎢
  ⎣
  11−
  
  −
  
  +(1−
  
  )−
  
  −1−
  
  −
  
  ⎤
  ⎥
  ⎦
  (A.3)
  and Θ is conformable but not needed for the analysis.

- Determinacy:
  - All variables in ξt = [ψt, t, θt]' are non-predetermined. REE is locally determinate iff all eigenvalues of Γ are within the unit circle.
  - Characteristic polynomial Π(ε) = ε^3 − Tr(Γ) ε^2 + S2(Γ) ε − Δdet(Γ) = 0, where:
    - Δdet(Γ) = [ − θ(1 + )]
    - Tr(Γ) = 1 +  + (1 − ) +  − θ
    - S2(Γ) = (1 + ) +  − θ − θ − θ + (1 − )
  - One root equals  ∈ (0, 1). Write Π(ε) = (ε − ) ̃Π(ε) with ̃Π(ε) = ε^2 + α1 ε + α2 where:
    - α1 = θ − 1 −  − (1 − )
    - α2 =  − θ(1 + )
  - All roots of ̃Π(ε) within unit circle iff:
    a) ̃Π(1) = 1 + α1 + α2 > 0
    b) ̃Π(−1) = 1 − α1 + α2 > 0
    c) |α2| < 1
  - These conditions are equivalent to those spelled in (28) in the proposition.

- E-stability of MSV-FE:
  - Define M ≡ Γ − I. For MSV-FE E-stability, all eigenvalues of M must have negative real parts.
  - One eigenvalue of M equals  − 1 < 0.
  - Characteristic polynomial of M: Π(ε) = ( − 1) ̃Π(ε) with ̃Π(ε) = ε^2 + α1 ε + α2 where:
    - α1 = (1 − ) + ( − 1) + θ
    - α2 = [( − 1) − θ]
  - All roots of M have negative real parts iff α2 > 0 and α1 > 0.
  - α2 > 0 iff  > 1 + (/) θ. If α2 > 0 then α1 > 0 holds as well.
  - Therefore MSV-FE is E-stable iff  > 1 + (/) θ for any θ ≥ 0.

- E-stability of CF-SSE:
  - From determinacy, characteristic polynomial Π(ε) = (ε − ) ̃Π(ε) = 0. Define *θ ≡ (1 + )/(1 + ).
  - Determinacy conditions in (28) imply indeterminacy for:
    -  > 1 + [2(1 + )/] − [2 + )/] θ and for  < 1 + (/) θ when θ < *θ,
    - and for any  ≥ 0 when θ ≥ *θ.
  - Within these ranges ̃Π(1) < 0 or ̃Π(−1) < 0, so all eigenvalues of Γ are real.
  - From Appendix B, a E-stable Common Factor representation of a stationary sunspot equilibrium (CF-SSE) exists within the policy space where MSV-FE is E-stable as well.
  - The conditions in the proposition follow after noting the inequality ordering for θ ≤ *θ.

### C.2. ProofofProposition2
- Reduced-form equilibrium system (27) with Γ defined as:
  Γ ≡
  ⎡
  ⎢
  ⎣
  1−
  1+
  1−
  
  
  1+
  ̃−
  
  
  1−
  1+
  ̃+(1−
  
  )
  (
  1+
  ̃−
  
  ) −1−
  
  ̃−
  
  ⎤
  ⎥
  ⎦
  (A.4)
  with Θ conformable and not necessary for the analysis.

- Learnability (E-stability) of MSV-REE:
  - MSV-REE is learnable iff all eigenvalues of M ≡ Γ − I have negative real parts.
  - By Routh Theorem, all roots of M have negative real parts iff:
    - Δdet(M) < 0
    - Tr(M) < 0
    - S2(M) Tr(M) − Δdet(M) < 0
    where S2(M) is sum of 2x2 principal minors of M.
  - After algebra:
    - Δdet(M) =
      (̃ − 1)[
      (1 − ̃ + ) − θ( − )/(1 + )
      ]
      + (1 − )[1 − ̃/(1 + )]
      (A.5)
    - Tr(M) = − [/(1 + )](1 + ) + 2(̃ − 1) − θ + (1 − )
      (A.6)
    - S2(M) = ( − 1)[2 − ̃/(1 + )]
      + (1 − ̃ + ) − θ( − )/(1 + )
      + (̃ − 1)[ − /(1 + )(1 + ) + ̃ − 1 − θ ]
      (A.7)

  - Using (A.6)-(A.7) one gets:
    (̃ − 1) S2(M) = (̃ − 1)^2[Tr(M) − (̃ − 1)] + Δdet(M) − (1 − ) ̃^2 /(1 + )
    and therefore S2(M) Tr(M) − Δdet(M) can be written as expression (A.8).

  - Necessary inequality set (derived):
    -  > 1 − (1 − ̃)(1 − ̃ + )(1 − ̃ + ) + (1 − ̃)( − )(1 +  − ̃) θ ≡ Φβ(θ)
      (A.9)
    -  > 1 − /(1 + )(1 + ) + 2(1 − ̃) − θ ≡ Φδ(θ)
      (A.10)
    -  > 1 − (1 − ̃ + ) + (1 − ̃)(1 + )[/(1 + )(1 + ) + 1 − ̃][2(1 + ) − ̃] + [ − ε][2(1 + ) − ̃] θ ≡ Φε(θ)
      (A.11)
      where ε ≡  + (1 − ̃)(1 + ) > .

  - Sufficient condition from (A.9):
    - A sufficient condition is  > 1 + (1 − ̃)( − )/(1 +  − ̃) θ ≡ Φ(θ).
    - The slope of Φ(θ) depends on sign of  − .
    - Using definitions ( = (1 + )−1,  in (25), ̃ ≡ /(1 + )), one can show  >  iff condition (A.12) holds; there exists a threshold * ∈ (0, 1) such that  >  for  > * under mild parameter restrictions.
    - Assume turnover rate  > *. Then ε >  >  and Φ(θ), Φδ(θ), Φε(θ) are strictly decreasing in θ; and Φ(θ) > max{Φδ(θ), Φε(θ)} for any θ ≥ 0.
    - Hence  > Φ(θ) is sufficient for Δdet(M) < 0, Tr(M) < 0 and S2(M) > 0.

  - Completing the proof that S2(M) Tr(M) − Δdet(M) < 0:
    - Consider (A.8). Define equivalence:
      Tr(M) − (̃ − 1) < 0 ⇔  > 1 − /(1 + )(1 + ) + 1 − ̃  − θ ≡ Φφ(θ).
    - Φφ(θ) < Φ(θ) for any θ ≥ 0.
    - If  > Φ(θ) then Tr(M) − (̃ − 1) < 0 and S2(M) > 0, which imply S2(M)[Tr(M) − (̃ − 1)] < 0 and (̃ − 1)^2[Tr(M) − (̃ − 1)] < 0.
    - Extended algebra shows these are sufficient to imply S2(M) Tr(M) − Δdet(M) < 0 for any  > Φ(θ).

  - Conclusion:
    - For  > *, the condition  > Φ(θ) is sufficient for all roots of M to have negative real parts, and therefore for the MSV-FE to be learnable.

- Additional note:
  - The function Φ, Φδ, Φε satisfy that Φ has the highest intercept and is the flattest among the three.

*Italic source attribution: Excerpt from the PDF content unit _wp1516 - conclusions one would get under the EE approach. We plan to explore the consequences of having IH learning in our model in a separate project.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1516.pdf_
