Comment on The Central Bank Balance Sheet as a Commitment Device By Gauti Eggertsson and Kevin Proulx Luca Dedola (ECB and CEPR) Banco Central de Chile XIX Annual Conference, 19-20 November 2015 Disclaimer: The views expressed are mine own, and do not reflect those of others in the European Central Bank or the Eurosystem.
Overview How does quantitative easing affect macroeconomic outcomes at the lower bound? Conceptually (not?) well-understood that QE works by changing supply and maturity of public liabilities held by private investors, as a policy lever to support aggregate demand. Signalling vs portfolio-balance effects: Lowering expectations of future short-term rates vs term premia see e.g. Hanson, Greenwood and Vayanos conference paper. Broad research agenda to which paper belongs, offering a useful theoretical framework for thinking about monetary policy and government debt. Key insight is that QE can act as a commitment device for rates forward guidance.
The debate Forward guidance powerful but not credible Central bank promise to allow overshooting of inflation and output gap once the fundamental shock dissipates, and thus mitigate the declines in inflation and output at the ZLB is time-inconsistent (Eggertsson and Woodford (2003) and others.) Alternatively, central bank can(not) commit to permanently increase monetary base (Krugman (1998)). QE early recognized as commitment mechanism for forward guidance Bernanke, Reinhart and Sack (2004): QE may strengthen the promise to keep short-term rates low, as an increase in official interest rates may lower the health of CB s balance sheet.
But CB s losses due to an increase in short-term interest rates may be too small to convince the public (Rudebusch 2011)... Others argue that the central bank is not a private institution for which negative equity matters.
Main results and policy messages QE credibly signals lower rates as it raises short-term public liabilities Large stock of CB and Govt short-term debt provides an incentive to keep future policy rates low. Cost of rolling over short-term debt is low when rates are low, reduce distortions from taxes, inflation. However it seems to matter quantitatively if short-term liabilities finance tax cuts or reduction in long term debt or real asset purchases as in this paper. Reminiscent of results on optimal public debt composition, e.g. Buera and Nicolini (2004).
Outstanding series of contributions, frustrating to discuss, all bases covered (in different papers) One-period government bond in Eggertsson (2006); One-period and long-term government bond, also with independent CB in Bhattarai, Eggertsson and Gafarov (2014); One-period government bond and real assets in this paper, no separate CB explicitly modeled. In my comments I will raise some quibbles, and talk about some evidence on QE and forward guidance.
Comments and questions 1. Story and intuition as arbitrage argument 2. Some quibbles on solution approach, some modeling choices 3. Empirical evidence: Is forward guidance ineffective because of imperfect credibility?
Story and intuition: An arbitrage argument? When ZLB not binding, discretionary government has an incentive to lower interest rates to reduce outstanding nominal liabilities against future tax, inflation distortions. When ZLB binding, incentive to accumulate nominal debt (with tax cuts/spending and asset purchases) to have lower future interest rates and raise inflation expectations. Incentives apparent from government budget constraint a t + b t = (b t 1 + a t 1 π t ) + i t q t ψτ t
It looks to me this is a lot more than an arbitrage argument (e.g. no fiat money in paper): Asset prices are endogenous and the government manipulates them to improve allocation. Aggregate demand externality: Agents take asset prices as given and respond to suboptimal high real rates at ZLB by saving more. To contrast self-enforcing increase in savings, current government manipulates asset prices, inflation. How? Influencing future governments through nominal debt accumulation by cutting taxes and buying what it can.
Some quibbles on solution approach Solution of MPE in dynamic economies with endogenous state variables usually not trivial due to Generalized Euler Equation (GEE). First order conditions depend on unknown first derivatives of policy functions, tricky even to solve for steady state. Paper makes enough assumptions to pick steady state with zero debt, assets and inflation (Not a theory of the size of government, CB balance sheet). Then log-linearize around steady state to get response to (zero probability) deflationary preference shock with binding ZLB. This approach greatly simplifies things, but there are some disadvantages.
Does the steady state matter? It would be nice to know whether full non-linear solution implies steady state unique or not Not clear given that SS is distorted. F. Martin (2009) finds positive interest rates in unique stable SS with nominal debt under discretion Even though Friedman rule optimal under commitment. Nakata-Schmidt (2014) find multiple equilibria under discretion and occasionally binding ZLB in standard NK model But in paper possible multiplicity not related to ZLB.
FOCs in SS from paper T t : φ BC ( 1 + g G s (T ) ) = 0 π t : φ BC [ 1 bπ 2 ] [ φ 4 C σ d (π) π ] = 0 [ i t : φ 2 C σ (1 + i) 2] φ BC [ 1 b (1 + i) 2 ] = 0 [ q t : φ 3 C σ (1 + q) 2] φ BC [ ( ) 1 a 1 ψ (α) (1 + q) 2 ] = 0 where φ 2 = φ 3 = φ 4 = 0 when functions ft E, SE t But with b,a non zero, also φ i 0. don t matter in SS.
Does the steady state matter? Setting the long run level of government debt (and assets) to zero allows to ignore complications due to GEE. However, I suspect that the level of debt at the beginning of the liquidity trap, the key state variable, matters in determining the outcome. Look at other distorted steady states with non-zero debt or non-distorted steady state where debt indeterminate.
Unanticipated ZLB Taking into account that ZLB occasionally binds may also be important Burgert-Schmidt (2014). It would pin down debt in stochastic steady state. For instance, if government debt is low relative to its steady state, then the policymaker may refrain from lowering the nominal interest rate all the way to zero, which exacerbates the fall in output and inflation. Another interesting question is whether there is a trade-off between ZLB and normal times. Dealing with ZLB compensates for potential inflationary bias outside of ZLB or possibility to use balance sheet policies always superior to conventional monetary policy?
Modeling choices Tax collection costs with lump-sum taxes are the key distortion in determining whether the debt increase will be undone by lower rates. Reasonable asset purchases only with huge collection costs But then taxes fluctuate a lot in experiment without purchases. With distortionary taxes, the tax base is endogenous and distortions affect inflation dynamics like a cost-push shock: π t = βe t π t+1 + κ ( y t φ σ + φ g t + (σ + φ) 1 1 τ Paper assumes g t = τ t = 0, shadow cost (φ BC 1t ) of Gov t resources only depends on reduced-form collection costs φ BC 1t + g G ( s (T t ) ) = 0 τ t ).
Forward guidance, QE and imperfect credibility Is forward guidance ineffective because of imperfect credibility, in contrast with QE announcements? At odds with surveys and financial market responses, as argued by Hanson et al. Del Negro, Giannoni and Patterson (2015): FG and QE both affect rate expectations, but differently Estimate change in forecasts in one-month window around FMC announcements on FG, QE: F (k) t+h = α + β 1 News + β 2 AP Changes + β 3 Conditions + γannouncementdummy + ε t
What are the effects of forward guidance? Evidence from blue chip financial forecasters Both QE and FG raise near term GDP growth forecast But FG lowers expected short rates, QE raises them
The Forward guidance puzzle Del Negro, Giannoni and Patterson (2012): Have we overestimated the power of forward guidance? Counterfactual with medium scale DSGE with good forecasting performance: FFR kept at ZLB through 2015Q2
Excessive response attributed to misspecified Euler equation or Phillips curve, and rational expectations.
Conclusion The paper has identified a set of conditions and assumptions under which a (very) large central bank balance sheet could be useful, and its conclusions are not likely to be sensitive to many of the details of model specification and solution. But some considerations the paper ignores may be more important than those it addresses. If the main benefit of QE is not via signalling the future policy stance, and if it may instead result in persistent amplification of quasi-fiscal effects of central bank decisions, then the argument for a much larger balance sheet is less strong.