The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Odyssean forward guidance in monetary policy: A primer Jeffrey R. Campbell Introduction and summary The Federal Open Market Committee’s (FOMC) monetary policy statement from its September 2013 meeting reads in part: In particular, the Committee decided to keep the target range for the federal funds rate at 0 to 1/4 percent and currently anticipates that this exceptionally low range for the federal funds rate will be appropriate at least as long as the unemployment rate remains above 6-1/2 percent, inflation between one and two years ahead is projected to be no more than a half percentage point above the Committee’s 2 percent longer-run goal, and longer-term inflation expectations continue to be well anchored.1 This extended reference to the conditions determining the FOMC’s future interest rate decisions is an example of forward guidance. Although participants in FOMC meetings have long used speeches and congressional testimony to discuss the Fed’s possible responses to economic developments, the Committee has only issued formal and regular forward guidance since February 2000, when it began to include in its statement a “balance of risks.” The first one read as follows: “Against the background of its long-run goals of price stability and sustainable economic growth and of the information currently available, the Committee believes the risks are weighted mainly toward conditions that may generate heightened inflation pressures in the foreseeable future.”2 Less than two years later, the Committee’s August 21, 2001, statement noted that “... the risks are weighted mainly toward conditions that may generate economic weakness in the foreseeable future.”3 Between the FOMC’s first statement of risks and the financial crisis that began in August 2007 and intensified in September 2008, the Fed experimented 130 with making its internal decision-making process more transparent and therefore more forecastable. In this, they followed several foreign central banks that had already adopted explicit inflation targets. (See Bernanke and Woodford, 2005, for a review of inflation targeting and its implementation outside the United States.) The financial crisis dramatically accelerated the transition to greater openness, and the FOMC’s Jeffrey R. Campbell is a senior economist and research advisor in the Economic Research Department at the Federal Reserve Bank of Chicago and an external fellow at CentER, Tilburg University. The author is grateful to Marco Bassetto, Charlie Evans, Jonas Fisher, Alejandro Justiniano, and Spencer Krane for many stimulating discussions on forward guidance and to Wouter den Haan, Alejandro Justiniano, and Dick Porter for helpful editorial feedback. This article is being concurrently published in Wouter den Haan (ed.), 2013, Forward Guidance— Perspectives from Central Bankers, Scholars and Market Participants, Centre for Economic Policy Research, VoxEU.org. © 2013 Federal Reserve Bank of Chicago Economic Perspectives is published by the Economic Research Department of the Federal Reserve Bank of Chicago. The views expressed are the authors’ and do not necessarily reflect the views of the Federal Reserve Bank of Chicago or the Federal Reserve System. Charles L. Evans, President; Daniel G. Sullivan, Executive Vice President and Director of Research; Spencer Krane, Senior Vice President and Economic Advisor; David Marshall, Senior Vice President, financial markets group; Daniel Aaronson, Vice President, microeconomic policy research; Jonas D. M. Fisher, Vice President, macroeconomic policy research; Richard Heckinger, Vice President, markets team; Anna L. Paulson, Vice President, finance team; William A. Testa, Vice President, regional programs; Richard D. Porter, Vice President and Economics Editor; Helen Koshy and Han Y. Choi, Editors; Rita Molloy and Julia Baker, Production Editors; Sheila A. Mangler, Editorial Assistant. Economic Perspectives articles may be reproduced in whole or in part, provided the articles are not reproduced or distributed for commercial gain and provided the source is appropriately credited. Prior written permission must be obtained for any other reproduction, distribution, republication, or creation of derivative works of Economic Perspectives articles. To request permission, please contact Helen Koshy, senior editor, at 312-322-5830 or email Helen.Koshy@chi.frb.org. ISSN 0164-0682 4Q/2013, Economic Perspectives forward guidance became more elaborate and detailed. After lowering the federal funds rate from 5.25 percent in early August 2007 to 0–25 basis points in midDecember 2008, the Committee’s statement read: “In particular, the Committee anticipates that weak economic conditions are likely to warrant exceptionally low levels of the federal funds rate for some time.”4 “Extended period” replaced “some time” in March 2009, adding specificity. This phrase remained in the statement until the August 2011 meeting, when it was replaced with the even more specific “at least through mid-2013.” The January 2012 statement pushed this date back to “late 2014.” By this point, these statements had become known as calendar-based forward guidance. Campbell et al. (2012) discuss the confusion this language had engendered among the public and market participants as of early 2012. Was “late 2014” a forecast that the economy would remain weak until then or a reassurance that the Committee would keep interest rates low through that date regardless of economic developments? The Committee’s September 2012 statement somewhat clarified this by stating that the Committee expects “that a highly accommodative stance of monetary policy will remain appropriate for a considerable time after the economic recovery strengthens.”5 Also, in that statement, “late 2014” became “mid-2015.” In its December 12, 2012, statement, the FOMC changed the nature of its forward guidance to reduce confusion by explicitly tying increases in the federal funds rate to unemployment and inflation outcomes, using language nearly identical to that from the September 2013 meeting quoted previously.6 It might seem paradoxical that at a time when the FOMC has done so little with its policy interest rate, it has talked so much about its plans. Even in normal times, a policymaker promising particular future actions constrains her future behavior and concomitantly loses flexibility. However, such forward guidance (sometimes called “open-mouth operations”) can substantially improve current economic performance when households’ and businesses’ current decisions depend on their expectations of future macroeconomic outcomes. If the FOMC’s assurances that rates will remain low raise private individuals’ expectations for future inflation and growth, then they will wish to consume more today, thereby lifting current aggregate demand and closing the output gap (the gap between actual and potential economic output). Although this benefit might indeed come at the cost of future flexibility, poor enough current macroeconomic performance might merit this sacrifice. When the zero lower bound (ZLB) on interest rates makes further conventional accommodation infeasible, the Federal Reserve Bank of Chicago exchange of future flexibility for current macroeconomic performance becomes especially attractive. Future policy actions only have impact if credible In general, statements of future policy intentions have no impact (benign or otherwise) when the public does not find them credible. This problem is particularly acute for a central bank, because a central bank seeking to improve households’ current and future welfare will be tempted to renege on past interest rate promises. The interest rate that is currently optimal might not be consistent with promises that improved past economic performance, and breaking those promises now does nothing to the past and improves present and future outcomes. If the public anticipates that monetary policymakers will apply such logic in the future, then promises of low future interest rates will not be believed and, therefore, will have no beneficial effect in the present. This conundrum is one example of the time-consistency problem, for the discovery of which Kydland and Prescott (1977) received a Nobel Prize in 2004. Since this kind of beneficial forward guidance requires the policymaker to keep past promises, even when sorely tempted to do what seems best at the moment, Campbell et al. (2012) label this Odyssean forward guidance. Like Odysseus bound to the mast of his ship, a monetary policymaker must forswear the siren call of the moment and stick to plans laid in the past. Odysseus achieved this with ropes for himself and earwax for his crew. Research into the analogous tools available to monetary policymakers is ongoing. Of course, not every pronouncement by a monetary policymaker is a promise. Some statements merely forecast the evolution of the private economy. Campbell et al. (2012) label such forecast-based statements Delphic forward guidance. Like the pronouncements from the oracle of Delphi, they forecast but do not promise. While Delphic pronouncements undoubtedly contribute positively to the execution of monetary policy, I ignore them in this article to develop instead a primer on the economic theory of Odyssean forward guidance. This primer’s basic framework is the minimal New Keynesian model, in which the central bank chooses the interest rate to achieve the best feasible trade-off of output and inflation. First, I discuss this model, develop key results, and present some simple calculations of optimal monetary policy paths that start with the economy at the zero lower bound. Although I review the model’s two linear equations, one inequality, and quadratic social welfare function in the text, I present the main results in figures for simplicity. I conclude the primer with a brief discussion of current monetary policy examined through the lens of this theory. 131 Forward guidance in the New Keynesian model Effective forward guidance requires the central bank to communicate its intentions and the public to believe that the bank is committed to their execution. The potential contribution of communication and commitment to improved monetary policy can be most easily appreciated in the canonical New Keynesian model that summarizes the behavior of producers, households, and a central bank with a Phillips curve, an intertemporal substitution (IS) curve, the zero lower bound on interest rates, and a central bank loss function. 1) πt = κyt + βπt +1 + mt , 2) 1 yt = − σ ( it − πt +1 − rtn ) + yt +1 , 3) it ≥ 0, 4) L = ∑ βt ∞ t =0 1 2 (π 2 t + λyt2 ) . More advanced versions of this model incorporate uncertainty about future macroeconomic outcomes. For the sake of simplicity, this primer abstracts from this complication and presumes that, conditional on the central bank’s policy choices, future macroeconomic outcomes can be calculated with certainty. In equation 1, πt is the rate of price inflation in year t and ỹt is that year’s output gap, defined to be the percentage deviation of actual output from its potential. (In New Keynesian models, producers can only adjust their dollar-denominated prices infrequently. It is this sluggish price adjustment that drives output away from its potential.) The influence of future inflation on its current level reflects the forward-looking behavior of producers choosing their prices. Woodford (2003) and Galί (2008) present derivations of equation 1 from the optimal pricing decisions of producers who can only adjust their nominal prices infrequently. In those derivations, the coefficient β is the discount factor producers apply to their future profits. The Phillips curve’s slope, κ, is an increasing function of the frequency of price adjustment. Perfectly flexible prices lead to a vertical Phillips curve, so that κ = ∞, while perfectly rigid prices set κ to zero. The output gap influences producers’ prices because it reflects their current marginal costs of production. The markup shock finishes the right-hand side of equation 1. It evolves exogenously and embodies changes in producers’ prices that are unrelated to changes in their marginal costs. For example, an exogenous decline in competitive price pressures due to leniency 132 in antitrust enforcement or innovations in market segmentation can show up as a positive mt. Because the Phillips curve reflects producer decisions, it is often labeled the economy’s “supply side.” Equation 2 reflects households’ split of current income between saving and consumption. The model’s households can invest in a one-year risk-free bond at the nominal interest rate it. This choice yields the inflation-adjusted return it – πt+1. Individual households can buy and sell this bond in unlimited amounts, but I keep the model simple by assuming that it is in zero aggregate supply. The economy has no capital or other means for real wealth accumulation, so total consumption must equal total income. Therefore, the output gap ỹt also equals the percentage deviation of actual consumption expenditures from their potential. From this perspective, the IS curve relates the current consumption gap to the interest rate and the consumption gap in the next period. The parameter σ is called the inverse absolute intertemporal elasticity of substitution. It is typically positive, so that increases in the interest rate induce households to increase saving and delay consumption. On the other hand, high future consumption reduces the incentive to save and increases current consumption. The final term requiring explanation in equation 2 is rt n, the natural rate of interest. This term is an exogenously evolving sequence that embodies changes in households’ relative valuations of current and future consumption. If rt n drops but it – πt+1 remains the same, then the household wishes to reduce current expenditures to save more now and, thereby, allow more consumption in the future. In this sense, a relan tively low value of rt indicates that the household is unusually patient. However, this household-based interpretation of rt n is probably at best a convenient fiction. In practice, many economists interpret low measured levels of rt n since the onset of the financial crisis as arising from the crisis itself and the resulting desire of both households and financial firms to remove both debt and risk from their balance sheets.7 The IS curve can be thought of as the economy’s “demand side.” The ZLB in equation 3 seems natural, because negative nominal interest rates are rarely, if ever, observed. It also has empirical appeal, because investors can move their portfolios into cash (which has a zero interest rate by construction) rather than holding bonds with negative rates.8 In this article, I follow Eggertsson and Woodford (2003) and Christiano, Eichenbaum, and Rebelo (2011) and make the zero lower bound relevant with a large negative value of the natural rate of interest. The central bank controls the nominal rate of interest; and its choices influence inflation and the output gap through the Phillips and IS curves. The Federal 4Q/2013, Economic Perspectives Reserve Act mandates that the FOMC use this influence “to promote effectively the goals of maximum employment, stable prices, and moderate long-term interest rates.” The model’s central bank fulfills such a mandate by choosing interest rates to minimize the loss function in equation 4. It penalizes current and future deviations from zero of inflation and of the output gap.9 The coefficient λ gives the central bank the relative weight on its output stabilization objective.10 The central bank uses the firms’ discount factor, β, to evaluate the tradeoff between current and future losses. Woodford (2003) and Galί (2008) both give derivations of this loss function as quadratic approximations of households’ welfare. Under this interpretation, both inflation and deflation distort the relative prices of goods; and positive and negative output gaps move households away from their desired allocation of time between labor and leisure. The central bank’s choice of it directly influences the current output gap through the IS curve and, thereby, indirectly influences inflation through the Phillips curve. However, this traditional static view of monetary policy is incomplete because producers and consumers base their decisions not merely on current policy, but also on their expectations for future inflation and output. It is this channel that makes forward guidance potentially useful. Discretionary monetary policy One cannot appreciate the value of commitment without understanding outcomes in its absence, so I begin with a review of monetary policy under discretion. By discretion, I mean that the central bank can set the current interest rate but has no direct influence over future rates until the future itself arises. As discussed earlier, a discretionary central bank takes no account of how expectations of its current actions influenced past behavior because those bygones are just that, bygones. There is little room for central bank communication to alter macroeconomic outcomes, because the only credible forward guidance simply describes what the central bank will find to be optimal when the time comes. Campbell et al. (2012) place such statements in the category of Delphic forward guidance. Since future interest rates determine future inflation rates and output gaps, the only terms in the central bank’s loss function under its current control give the current loss, 12 (π02 + λy 02 ). The discretionary central bank’s optimal interest rate minimizes this current loss by taking as given ỹ1, π1, m0, and r0n . The divine coincidence I begin consideration of this choice with the very special case in which mt = 0 and rt n ≥ 0 always. If Federal Reserve Bank of Chicago fortuitously both ỹ1 and π1 also equal zero, then the IS curve allows the central bank to achieve a zero output gap by simply setting it to rt n . Since βπ1 + m0 = 0, the Phillips curve translates a zero output gap into zero current inflation. That is, if future inflation and the costpush shock both equal zero and the natural rate of interest is positive, then the central bank can achieve the minimum possible loss by completely stabilizing both the output gap and inflation. Blanchard and Galί (2010) have referred to a similar result in a more complicated model as a “divine coincidence.” The Phillips curve, which determines which inflation and output gap combinations are feasible, passes through the best possible such combination, no inflation and no output gap. One might object that this superior outcome merely reflects the good fortune of inheriting expectations of price and output stability, but the fact that the central bank wishes to achieve such stability gives one reason to believe that it will occur. Indeed, if both ỹ2 and π2 equal zero, then the central bank can and will achieve complete macroeconomic stability in period 1. Continuing in this fashion yields the following result: If mt = 0 always and rt n is never negative, then the interest rate rule it = rt n is feasible and achieves complete macroeconomic stabilization. To prove the result to yourself, simply note that the sequences ỹt = 0 and πt = 0 satisfy both the Phillips and IS curves if rt n = it always. Furthermore, this interest rate choice minimizes the current loss, so households and businesses should expect the central bank to follow it. The output-inflation trade-off When βπ1 + m0 differs from zero, the central bank cannot achieve complete stabilization because the Phillips curve no longer passes through the origin. In this case, the discretionary central bank faces a classic output-inflation trade-off. Panel A of figure 1 illustrates this trade-off with a familiar indifference curve budgetset diagram. Here, the Phillips curve (in red) plays the role of the budget constraint. The central bank can choose any inflation-output gap combination on the curve. Its slope equals κ, and it crosses the vertical axis at βπ1 + m0. The family of indifference curves comes from the central bank’s loss function. Each one gives the inflation-output gap combinations that yield a constant value for the current loss function. If λ equals one, each indifference curve is a circle. In general, the curves are ellipses, but I have drawn only their portions in the northwest quadrant. The points on an indifference curve that lie inside of another give a lower total loss. If the central bank were to choose an inflation-output gap combination with an indifference curve that crosses the Phillips curve, then it could achieve a lower loss by sliding away from the closest axis along the Phillips 133 insert source and footnotes figure 1 The inflation-output gap trade-off labelpolicy without the ZLB A. Optimal B. Optimal policy with the ZLB βπ 1 + m 0 re n Ph rv illi e ps cu rv e Chosen y˜ 0 , π0 Inflation rate, π 0 ffe cu Inflation rate, π0 I i nd ce βπ 1 + m 0 i0 < 0 i0 = 0 Output gap, ỹ0 0 C. Forward guidance without the ZLB r 0n + π1 σ Output gap, ỹ0 + ỹ 1 0 D. Forward guidance with the ZLB ↑ ỹ1 & ↑ π1 βπ 1 + m 0 ↑ π1 End ↓ π1 End ↑ π1 Output gap, ỹ0 0 Inflation rate, π0 Start Inflation rate, π0 ↓ π1 r 0n + π1 σ Start Output gap, ỹ0 + ỹ 1 0 Note: ZLB indicates zero lower bound. curve. Therefore, the Phillips curve must be tangent to the best possible point’s associated indifference curve. This is marked in the figure with the red point labeled “Chosen ỹ0, π0.” The central bank tolerates both higher-than-desired inflation and lower-thandesired output as the best feasible outcome. The exact inflation-output gap chosen balances the loss from increasing inflation slightly with the loss from slightly deepening the recession. The nominal interest rate is notable in this standard analysis of the output gap-inflation trade-off only by its absence. The Phillips curve alone determines the outputinflation trade-off. So long as the desired output gap is not below what can be achieved by setting i0 to zero, the IS curve merely determines the nominal interest rate that guides the private sector to the central bank’s 134 favored outcome. The IS curve becomes more relevant to the problem when the ZLB on i0 constrains the central bank. To see how, isolate i0 on the left-hand side of equation 2, substitute the resulting right-hand side into the ZLB in equation 3, and arrange the result to put ỹ0 on the lower side of the inequality, y 0 ≤ y1 + r0n + π1 . σ That is, the ZLB and IS curve together put an upper bound on the output gap. When this upper bound is a negative number, it can be interpreted as a lower bound on the size of a recession. If this lower bound is high enough, then conventional interest rate policy cannot mitigate a recession. Panel B of figure 1 depicts the central bank’s choice in this case. The dashed 4Q/2013, Economic Perspectives vertical line indicates the location of the upper bound on ỹ0. Without the ZLB, optimal monetary policy would guide the economy to the tangent point marked “i0 < 0.” The ZLB moves the actual outcome southwest along the Phillips curve to the point marked “i0 = 0,” where the Phillips curve intersects the vertical line. Since the central bank’s indifference curve is steeper than the Phillips curve, it would like to reduce the current output gap at the expense of higher inflation. However, the ZLB prevents it from doing so. This illustrates how conventional monetary policy at the ZLB is “too tight.” Monetary policy with commitment and communication Both the Phillips curve and IS curve are forward looking, so each of them can serve as a channel for forward guidance to influence current macroeconomic outcomes. Panels C and D of figure 1 illuminate these channels. Suppose that the central bank could credibly influence private expectations about inflation in year one. Lowering π1 directly shifts the Phillips curve down and, thereby, expands the set of possible current output gap-inflation outcomes. Panel C illustrates this situation, in which forward guidance moves inflation and the output gap toward their desired levels. Economically, a credible promise of future disinflation lowers producers’ current desired prices and, thereby, allows the central bank to achieve a given level of current inflation with a smaller output gap. Of course, the promised deflation and its accompanying output gap also cost the central bank. The size of the cost depends on the initial values for π1 and ỹ1. If a substantial deflationary recession was already anticipated, then fighting current inflation with forward guidance might be too costly. On the other hand, if both π1 and ỹ1 begin at zero, then slight changes to them have very, very small costs. Since the IS curve is irrelevant for discretionary monetary policy away from the ZLB, it should be no surprise that forward guidance works through the IS curve only when the ZLB constrains policy. Panel D of figure 1 shows how forward guidance can influence outcomes in this case. The upper bound for ỹ0 derived from the IS curve and the ZLB constraint increases in both π1 and ỹ1, so this lower bound shifts to the right if the central bank’s promises of low future interest rates increase expectations of inflation, the output gap, or both in year one. If this were the end of the story, the forward guidance would slide the inflation-output gap outcome along a fixed Phillips curve. However, the increase in promised inflation also shifts the Phillips curve up. As drawn, the cost of the additional current inflation is less than the benefit from the reduced output gap. (The indifference Federal Reserve Bank of Chicago curve running through the point marked “End” is interior to the one passing through “Start.”) Just as in the case displayed in figure 1, whether this improvement in current outcomes is worth the required change in π1 and ỹ1 will depend on their initial levels. If the central bank inherits expectations of future macroeconomic stability, then the cost of forward guidance is small. Optimal monetary policy as a path The same constraints that limit the central bank’s actions in year zero also apply to future years, so this discussion of forward guidance would be incomplete if it stopped at figure 1. To bring future years’ Phillips curves and IS curves into the picture, consider the problem of a central bank in year zero choosing values for πt, ỹt, and it from year zero into the infinite future. The central bank chooses these to minimize the loss function in equation 4, but the chosen sequences must satisfy the Phillips curve, IS curve, and ZLB in equations 1, 2, and 3 for all years. This dynamic formulation of the monetary policy problem is necessary for the full consideration of forward guidance, because it allows the central bank to quantitatively compare the current gains from forward guidance with the future costs of following through on promises made. Because Ramsey (1927) first conceived of economic policy as choosing a vector of economic outcomes to achieve the lowest social cost possible subject to the constraints imposed by private decision-making, economists call this a Ramsey problem and its policy prescription a Ramsey solution. In this particular context, the central bank’s loss function determines the social cost of specific sequences for the output gap and inflation, and the constraints imposed by private decision-making are the Phillips curve, IS curve, and ZLB. The Ramsey outcome can be best appreciated by studying an example calculated from a particular parameter configuration. To impose a neutral interest rate of 4 percent, the example set β = exp (−0.04). Evans (2011) discusses the numerical values for λ consistent with the Fed’s dual mandate of promoting maximum employment with stable prices, and the example uses his preferred value λ = 0.25. The absolute intertemporal elasticity of substitution σ equals one; so a 1 percent reduction in the natural interest rate lowers the output gap’s upper bound by 1 percent. Figure 2 shows the sequence of output gaps and inflation rates that minimize the central bank’s loss function with these parameters when a temporarily negative natural rate of interest drives the economy to the ZLB in year zero. That is, r0n = −0.01 and rt n = 0.04 for t ≥ 1. (The markup shock that placed the analysis of figure 1 into the northwest quadrant 135 insert source and footnotes figure 2 Optimal policy with one year at the zero lower bound label A. Flat Phillips curve: κ = 0.04 B. Steep Phillips curve: κ = 1.00 πt πt 0.30 0.01 t –0.04 t –0.07 –1.00 ∼ ∼ yt yt 0.50 0.35 t t –0.35 –0.47 –1.00 –1.00 it it 3.52 0 136 3.54 t 0 t 4Q/2013, Economic Perspectives equals zero here.) The figure reports results for two values of κ, 0.04 and 1.00. The smaller “flat” value of κ is of the magnitude favored by Eggertsson and Woodford (2003). It requires a 20 percent decrease in the output gap to lower inflation by 1 percent. One might judge such a large sacrifice ratio to be unrealistic, because actual disinflations (such as that engineered by Paul Volcker in the early 1980s) have not generated such large output declines. The relatively larger value for κ addresses this possibility. In figure 2, the black arrows pointing to the vertical axes indicate each variable’s value in year zero without forward guidance. (In all future years, the discretionary values of πt, ỹt, and it are zero, zero, and 0.04, respectively.) By construction, discretionary monetary policy can do nothing to mitigate the effects of hitting the ZLB. The negative 1 percent natural interest rate drives ỹ0 to −1 percent, irrespective of the Phillips curve’s specification. The Phillips curve’s slope determines the size of the associated disinflation. With the flat Phillips curve, this equals only ‒4 basis points, but with the steep Phillips curve, inflation falls 1 full percentage point. When the central bank instead employs forward guidance, the decline in the output gap is substantially reduced, to ‒47 and ‒35 basis points with the flat and steep Phillips curves, respectively. To achieve such moderation of the initial recession, the central bank engineers a future inflationary expansion. In year one, the output gap equals 50 and 35 basis points with the flat and steep Phillips curves, respectively. With the flat Phillips curve, inflation in year one is hardly noticeable, but it equals 30 basis points with the steep Phillips curve. More noticeable is the effect of forward guidance on year zero inflation when the Phillips curve is steep. It rises from ‒1 percentage point to ‒7 basis points. The experiments with both slopes feature very small deviations from steady state after year one, and they have nearly identical associated paths for the interest rate. By construction, i0 = 0. The interest rate equals about 3.54 percent in year one and thereafter stays very close to the natural rate. These numerical results illustrate two principles emphasized by Eggertsson and Woodford (2003). First, optimal monetary policy at the ZLB resembles the prescriptions of price-level targeting (PLT). Under PLT, the central bank announces targets for a relevant price index, such as the deflator for consumer expenditures excluding food and energy goods, for several dates. The central bank then chooses policy in order to come as close as possible to these targets. If inflation falls short of its expected value, then the central bank deliberately tolerates a later overshooting of inflation, which brings the price level closer to its stated target. Federal Reserve Bank of Chicago Qualitatively, this policy can be seen in the optimal inflation path with a steep Phillips curve. The deflation of 7 basis points is followed by an inflation of 30 basis points. Recall that even if the ZLB does not bind, a central bank facing an output-inflation trade-off resulting from an inflationary markup shock would like to promise deflation in the future to move the Phillips curve back toward the origin. The inflation followed by deflation also resembles the PLT outcome. Eggertsson and Woodford (2003) provide a more extensive but similar argument that PLT should always be followed, both at and away from the ZLB. The second principle can be seen in the accommodative interest rate in year one: Optimal forward guidance promises to maintain an expansionary monetary policy after the conditions that initially warranted it have passed. Conclusion Since economic growth remains below potential, inflation is running below the FOMC’s target of 2 percent, and the ZLB prevents further conventional monetary accommodation, the FOMC has turned to two nontraditional monetary policy tools, quantitative easing and forward guidance. This article has shown how the latter, through “open mouth operations,” can improve current macroeconomic outcomes by altering current expectations of future inflation and output. In the Ramsey problem, the central bank’s ability to manipulate expectations is assumed to be perfect. Campbell et al. (2012) review the considerable evidence that FOMC members did indeed influence private expectations before the financial crisis, and they expand upon it by showing that FOMC statements continued to move asset prices in the post-crisis period. Such influence is undoubtedly helpful for implementing forward guidance, so it seems reasonable to assume that FOMC participants have built up enough influence with the public to credibly commit to forward guidance. This primer reviewed the theory of such guidance, but the question of how well the FOMC’s current guidance matches that of the theory remains open. In the simple model I used to solve the Ramsey problem, the natural interest rate follows a simple predetermined path and there are no markup shocks. In practice, both the FOMC and the public face considerable uncertainty about the path of the natural interest rate. Furthermore, shocks to supply (through the markup shock) and demand (through the natural interest rate) continue to impact the economy even though they are more pedestrian than those that caused the financial crisis. Mimicking the Ramsey solution in such circumstances would require the FOMC to specify a comprehensive 137 rule for its interest rate decisions and associated forecasts for inflation and the output gap. In such a complex world, where the possible sources of future economic turbulence cannot even be reliably listed (not to mention quantified), such a complete solution is unrealistic. What the FOMC has done instead is provide threshold-based guidance. The Committee expects the current interest rate of approximately zero to remain appropriate at least as long as the unemployment rate remains above 6.5 percent and medium-term inflation expectations remain below 2.5 percent. This guidance can be consistent with the “overshooting” prescription of the Ramsey solution. Of course, the simple model presented here gives just a qualitative guide to optimal forward guidance. The more sophisticated model of Eggertsson and Woodford (2003) differs from it only by randomizing the time at which the natural rate of interest permanently returns to its long-run value, so that provides hardly more quantitative guidance for the current situation. Extending this policy framework to include a more realistic random evolution of rt n and ongoing markup shocks is the subject of current research. NOTES 1 The full press release from the September 18, 2013, FOMC meeting is available at www.federalreserve.gov/newsevents/press/monetary/ 20130918a.htm. 2 See www.federalreserve.gov/boarddocs/press/general/2000/ 20000202/default.htm. See www.federalreserve.gov/boarddocs/press/general/2001/ 20010821/default.htm. 3 4 See www.federalreserve.gov/newsevents/press/monetary/ 20081216b.htm. See www.federalreserve.gov/newsevents/press/monetary/ 20120913a.htm. 5 6 See www.federalreserve.gov/newsevents/press/monetary/ 20121212a.htm. One might object that the simple model economy at hand has no cash, only one-period bonds. Woodford (2003) asserts that adding cash to the model leaves its basic economics unchanged. This article uses the cashless version of the New Keynesian model to maintain simplicity. 8 9 Virtually by definition, bringing the output gap closer to zero improves social welfare. However, zero inflation is not necessarily the socially optimal definition of “price stability.” Reifschneider and Williams (2000) discuss this in more detail. For simplicity, this primer abstracts from this issue by defining “price stability” with a zero inflation rate. One might object that the output gap appears in equation 4 rather than an analogously defined employment gap. Since Okun’s law connects these two gaps, the stabilization of the output gap is indeed consistent with the Fed’s dual mandate. See Evans (2011) for a discussion of this issue. 10 Since it corresponds to no specific market interest rate, r n cannot t be directly observed. However, it can be inferred from observations of actual interest rates and households’ consumption and savings decisions. See Justiniano and Primiceri (2010) for a review of this procedure. 7 138 4Q/2013, Economic Perspectives references Bernanke, B. S., and M. Woodford (eds.), 2005, The Inflation-Targeting Debate, Chicago: University of Chicago Press. Galί, J., 2008, Monetary Policy, Inflation, and the Business Cycle: An Introduction to the New Keynesian Framework, Princeton, NJ: Princeton University Press. Blanchard, O., and J. Galί, 2010, “Labor markets and monetary policy: A New Keynesian model with unemployment,” American Economic Journal: Macroeconomics, Vol. 2, No. 2, April, pp. 1‒30. Justiniano, A., and G. E. Primiceri, 2010, “Measuring the equilibrium real interest rate,” Economic Perspectives, Federal Reserve Bank of Chicago, Vol. 34, First Quarter, pp. 14‒27, available at www.chicagofed.org/ webpages/publications/economic_perspectives/2010/ 1q_justiniano_primiceri.cfm. Campbell, J. R., C. L. Evans, J. D. M. Fisher, and A. Justiniano, 2012, “Macroeconomic effects of Federal Reserve forward guidance,” Brookings Papers on Economic Activity, Spring, pp. 1‒54. Christiano, L., M. Eichenbaum, and S. Rebelo, 2011, “When is the government spending multiplier large?,” Journal of Political Economy, Vol. 119, No. 1, February, pp. 78‒121. Eggertsson, G. B., and M. Woodford, 2003, “The zero bound on interest rates and optimal monetary policy,” Brookings Papers on Economic Activity, Vol. 2003, No. 1, pp. 139‒211. Evans, C. L., 2011, “The Fed’s dual mandate responsibilities and challenges facing U.S. monetary policy,” speech at the European Economics and Financial Centre Distinguished Speaker Seminar, London, UK, September 7. Federal Reserve Bank of Chicago Kydland, F. E., and E. C. Prescott, 1977, “Rules rather than discretion: The inconsistency of optimal plans,” Journal of Political Economy, Vol. 85, No. 3, June, pp. 473‒492. Ramsey, F. P., 1927, “A contribution to the theory of taxation,” The Economic Journal, Vol. 37, No. 145, March, pp. 47‒61. Reifschneider, D., and J. C. Williams, 2000, “Three lessons for monetary policy in a low-inflation era,” Journal of Money, Credit and Banking, Vol. 32, No. 4, November, pp. 936‒966. Woodford, M., 2003, Interest and Prices: Foundations of a Theory of Monetary Policy, Princeton, NJ: Princeton University Press. 139 A history of large-scale asset purchases before the Federal Reserve Benjamin Chabot and Gabe Herman Introduction and summary The Federal Reserve’s preferred policy instrument— the overnight federal funds rate—approached zero at year-end 2008. With the zero lower bound constraining additional policy accommodation through traditional channels, the Federal Reserve began a series of largescale asset purchases (LSAPs). The goal of the Fed’s LSAP strategy is to place downward pressure on yields of a wide range of longer-term securities, foster mortgage markets, and encourage a stronger economic recovery.1 While a consensus has emerged that LSAPs have lowered yields on U.S. Treasury bonds and other long-maturity, high-duration assets (thus increasing their prices),2 considerable uncertainty remains as to the magnitude of these yield changes and the exact channels by which central bank purchases influence yields.3 Much of this uncertainty stems from the fact that researchers have only a few examples of large open market purchases of government-guaranteed bonds to study. Central banks have traditionally preferred to implement monetary policy by altering short-term interest rate targets rather than utilize their balance sheets as a policy tool. Most of what we know about the effectiveness of LSAPs and the magnitude of their effects, therefore, come from evaluations of the small number of episodes when central banks wished to stimulate their economies but the traditional tool— the short-term policy rate—was constrained by the zero lower bound of nominal interest rates.4 In this article, we assemble a new historical database of monthly U.S. Treasury bond prices, contract terms, and amounts outstanding between 1870 and 1913. These new data allow us to look beyond the traditional empirical sample of LSAPs by examining the numerous large open market operations conducted by the U.S. Department of the Treasury during this pre-Fed era. During this period, the Treasury engaged in many refundings and open market sinking fund purchases5 140 that resulted in dramatic changes in the quantity and duration of aggregate Treasury bonds outstanding. These refundings and sinking fund purchases provide us with an opportunity to measure the effects of changes in the amount and duration of Treasury bonds on equilibrium yields.6 We compare the price response of high- and low-duration bonds to changes in the amount and aggregate duration of Treasury bonds outstanding, and find purchases of Treasury securities made by the U.S. Treasury Department narrowed the yield spread between Treasury bonds with high interest rate risk (the risk of an investment’s value changing on account of interest rate changes)7 and those with low interest rate risk. Benjamin Chabot is a financial economist and Gabe Herman is an associate economist in the Economic Research Department at the Federal Reserve Bank of Chicago. © 2013 Federal Reserve Bank of Chicago Economic Perspectives is published by the Economic Research Department of the Federal Reserve Bank of Chicago. The views expressed are the authors’ and do not necessarily reflect the views of the Federal Reserve Bank of Chicago or the Federal Reserve System. Charles L. Evans, President; Daniel G. Sullivan, Executive Vice President and Director of Research; Spencer Krane, Senior Vice President and Economic Advisor; David Marshall, Senior Vice President, financial markets group; Daniel Aaronson, Vice President, microeconomic policy research; Jonas D. M. Fisher, Vice President, macroeconomic policy research; Richard Heckinger, Vice President, markets team; Anna L. Paulson, Vice President, finance team; William A. Testa, Vice President, regional programs; Richard D. Porter, Vice President and Economics Editor; Helen Koshy and Han Y. Choi, Editors; Rita Molloy and Julia Baker, Production Editors; Sheila A. Mangler, Editorial Assistant. Economic Perspectives articles may be reproduced in whole or in part, provided the articles are not reproduced or distributed for commercial gain and provided the source is appropriately credited. Prior written permission must be obtained for any other reproduction, distribution, republication, or creation of derivative works of Economic Perspectives articles. To request permission, please contact Helen Koshy, senior editor, at 312-322-5830 or email Helen.Koshy@chi.frb.org. ISSN 0164-0682 4Q/2013, Economic Perspectives LSAP channels Theory suggests LSAPs can lower equilibrium bond yields through three channels, which we label scarcity, duration, and signaling. The scarcity channel, which is sometimes referred to as the portfolio-balance channel, is associated with the preferred habitat literature pioneered by James Tobin, Franco Modigliani, and Richard Sutch.8 Because of differences in asset risk characteristics or regulation, some investors prefer to hold certain assets and are reluctant or unable to hold alternative assets. For example, regulatory restrictions force money market funds to hold short-maturity assets, while insurance companies may prefer to hold longmaturity assets that match the duration of their liabilities; moreover, a bank that wishes to hedge the duration and negative convexity9 embedded in its mortgage portfolio will prefer to short sell10 long-maturity Treasury securities. Therefore, different classes of financial assets are not perfectly substitutable in investors’ portfolios and changes in the relative supply of preferred assets may alter their equilibrium prices and yields. There is every reason to believe that pre-1913 investors also preferred to hold certain bonds because of differences in asset risk characteristics or regulatory restrictions. The National Banking Acts of 1863 and 1864, for instance, provided regulatory incentives for national banks located outside “reserve cities” to deposit a portion of their reserves in reserve city national banks.11 These deposits could then be lent in the secured overnight call market12 where U.S. Treasury bonds were considered premium collateral, which required a smaller haircut13 than other securities. While most data on overnight lending are unavailable, the New York Superintendent of Insurance required insurance companies doing business in the state to report the collateral accepted and haircuts demanded to secure overnight loans appearing on their year-end balance sheets. An 1872 sample of this year-end insurance data confirms the favored status of government bond collateral (and especially low-duration government bonds): Insurance companies lent overnight against Treasury bond collateral with an average haircut of 11.2 percent, which was much lower than an average haircut of 23.2 percent on all collateral.14 National banks were also required to hold Treasury bonds as collateral against bank note issuance or government deposits. The funding needs of each bank and the price and risk characteristics of each Treasury bond issue determined how appealing a given Treasury bond was as collateral. For example, a high-coupon Treasury bond trading above face (par) value would secure more funding when it was pledged as collateral in the interbank call market (where bonds were haircut from Federal Reserve Bank of Chicago market value) than when it was pledged as collateral for bank note issuance or government deposits (for which legal requirements valued all government bonds at the minimum of par or market value). Therefore, a high-coupon, low-duration bond was better collateral for a bank that funded in the wholesale call market than a bank that funded via bank note issuance and government deposits. These differences were indeed reflected in the use of outstanding Treasury bonds as collateral. If we define high-duration Treasury bonds as bonds of this type with durations above the median duration of all Treasury bonds outstanding, 44 percent of the market value of all Treasury bonds was in highduration bonds at year-end 1872. Despite accounting for four-ninths of all Treasury bonds outstanding, highduration bonds accounted for only 18.3 percent of Treasury bonds posted as collateral for overnight loans from insurance companies.15 However, highduration bonds accounted for 82.2 percent of Treasury bonds posted to secure bank note issuance.16 Legal differences in circulation privileges, taxes, and option-induced17 convexity likely resulted in lessthan-perfect substitutability among bonds in the portfolios of certain investors during the pre-Fed era. When bonds are not perfectly substitutable, the scarcity channel implies that LSAPs can raise the prices of the purchased assets and similar assets but will have a limited effect on the prices of dissimilar assets. However, there are reasons to believe that LSAPs’ effect on asset prices through the scarcity channel is not monotonic. If open market purchases remove too much supply from a segmented basket of assets, the resulting decrease in liquidity (the ease at which an asset can be converted into cash) may make the remaining assets unattractive to investors who previously preferred to hold them. This concern is reflected in the current Federal Reserve policy to limit aggregate system open market account holdings of each Treasury bond issue to no more than 70 percent of outstanding issuance.18 The duration channel also arises from the existence of preferred habitat investors, and it likely existed during the pre-Fed era. Unlike the scarcity channel where asset purchases should only affect the prices of the purchased assets and similar assets, the removal of interest rate risk or duration risk19 from the market via LSAPs should affect the risk premium of all assets in proportion to their sensitivity to interest rate changes. In the model of Vayanos and Vila (2009), for example, the presence of preferred habitat investors who are willing to accept lower returns to hold assets in a preferred maturity neighborhood creates profitable trading opportunities for other risk-averse investors—called arbitrageurs—who are willing to trade assets of any 141 maturity. These willing traders can earn excess profits by holding assets that are out of favor with preferred habitat investors and short selling the assets that are in favor with those investors. The preferred habitat investors’ trading strategy exposes arbitrageurs to aggregate interest rate risk for which they must be compensated. In this framework, LSAPs can lower the equilibrium risk premium embedded in bond yields by removing aggregate duration risk from the portfolios of arbitrageurs. The final channel, which we call the signaling channel, is based on the insight that increases in central bank open market purchases can be interpreted as a signal of a more accommodative policy stance.20 If this signal results in a lowering of investors’ expectations for the future path of policy rates, open market purchases can lower the yield on longer-term assets by lowering expectations of future short-term interest rates. The United States had no central bank during our period of study, and the entity that conducted open market sales and purchases—the Treasury—did not (and perhaps could not) target any policy rate. Because the Treasury did not target a policy rate, our time period is an ideal laboratory to identify the effects of altering the duration and scarcity of Treasury bonds outstanding without a signaling channel confounding our measurements. Data: The market price and amount outstanding of U.S. Treasury debt in 1870–1913 We document the large-scale asset purchases of the pre-Fed era by collecting information about the amount outstanding and cash flow characteristics of each Treasury bond in existence between 1870 and 1913. Our main source is the U.S. Department of the Treasury’s Monthly Statement of the Public Debt (MSPD) database.21 The statements in this database report the amounts outstanding of each bond issue on or near the last day of the month. Also included in the statements are a number of bond characteristics necessary for specifying each bond’s promised cash flow— such as the coupon rate, the month(s) in which the interest payments are made, the schedule of final maturity payments, and the terms of any embedded options. During our period of study, a majority of United States bonds contained call option clauses granting the government the right, but not the obligation, to retire all or part of the issue outstanding for a particular price after a vesting date but before the bond’s maturity date (if it had one). For most bonds, the MSPD database includes the date on which the government’s call option vests. In cases where information from the 142 MSPD database was unclear, we determined option characteristics by locating the contract language of the bonds in De Knight (1900). We collect price data from the New York Stock Exchange (NYSE) closing bid and ask prices reported in the Commercial & Financial Chronicle, the New York Times, and the New York Tribune. Because some debt issues were not regularly quoted on the NYSE, we were able to find NYSE price quotations for only 77 percent of the monthly bond listings that appeared in the MSPD. We replaced the missing 23 percent of bond prices with model-generated prices by fitting a term structure of interest rates (yield curve)22 and implied volatility23 to the observable bonds via the Hull–White model described in the next section. Measuring the duration of U.S. Treasury debt in 1870–1913 Using the market prices, amounts outstanding, and cash flow characteristics of U.S. Treasury bonds from the data sources described in the previous section, we compute the Macaulay duration of each bond each month to form a monthly time series of the aggregate duration risk of U.S. Treasury bonds during our sample period. Option-free bonds with observable market prices We begin by computing the Macaulay duration of each option-free bond. When a bond has an observable market price and a nonstochastic cash flow, Macaulay’s duration is as follows: ( t ) CFt 1) T (1 + ytm ) t =1 P Dur = Σ t , where P is the bond’s price, CFt the bond’s cash flow at time t, and ytm is the bond’s yield to maturity measured at the same frequency as the coupon payments. Bonds with embedded options or no market price Fifty-eight percent of the bonds in our database have embedded options, which grant the Treasury the right, but not the obligation, to retire the bond at par after a vesting date but before the bond’s final maturity date. Options alter the duration of a bond by transforming the bond’s cash flow into a function of stochastic future interest rates. We use the Hull–White model to compute the option-adjusted durations of bonds with embedded options.24 The Hull–White model is a single-factor noarbitrage model of the term structure of interest rates in which the short-term rate is assumed to evolve via 4Q/2013, Economic Perspectives a stochastic differential equation with mean reversion. Given an initial zero-coupon yield curve (zero curve),25 we can use the model to compute the value and duration of an option-embedded bond as a function of the volatility of the short-term rate and the degree of mean reversion. To implement the Hull–White model, we require an initial zero-coupon yield curve and the volatility and degree of mean reversion for the shortterm rate. None of these initial parameters are directly observable during our time period,26 but we can calibrate each by fitting a zero curve, implied volatility, and mean-reversion parameter to best match observable bond prices. We select a time-invariant coefficient of mean reversion and, for each month in our data set, an implied volatility and a linear zero curve (level and slope coefficients) to best fit the observable bond price data.27 The result is a monthly time series of estimated zero curves and implied volatilities. With these parameter estimates in hand, we use the Hull–White model to compute the option-adjusted duration for each bond with an embedded option and generate the modelimplied price and duration for the option-free bonds with missing price data. Aggregate duration risk of U.S. Treasury debt in 1870–1913: Ten-year-equivalent debt outstanding With duration estimates and amounts outstanding in hand, we compute a monthly time series of ten-yearequivalent U.S. Treasury debt outstanding. A ten-year equivalent is a common measurement for interest rate risk in a portfolio. To express a portfolio’s interest rate risk in ten-year-equivalent units, we first compute how much the portfolio’s dollar value will change for a given change in yields and then compute how many bonds with ten-year duration one would have to hold to experience the same change in portfolio value. Thus, the ten-year-equivalent U.S. Treasury debt outstanding is Durn 2) 10 yearEq = Σ nN=1 MVn , 3650 where N is the number of bonds in the portfolio, Durn is the duration in days of the nth bond, and MVn is the market value of the nth bond. Figure 1 graphs the outstanding U.S. Treasury bonds’ par value (the size) and ten-year-equivalent value (representing the aggregate interest rate risk) over the period 1870–1913. As we will discuss in more detail in the next section, when the ten-year-equivalent value rises, it indicates that aggregate interest rate risk borne by holders of Treasury bonds has increased. Aggregate Federal Reserve Bank of Chicago interest rate risk can rise because the Treasury issues more risky bonds in total that must be held by the public (for example, during the 1893–99 period in figure 1) or because the existing bonds become more sensitive to interest rate changes (for example, during the 1876–79 period in figure 1). LSAPs: 1870–1913 The refunding of the U.S. Civil War debt in the late 1870s, the bond issuance associated with the return to the gold standard in 1879, the sinking fund open market purchases of the 1880s, and the deficit funding of the 1890s all provide examples of dramatic changes in the duration or amount of U.S. Treasury bonds outstanding.28 The refunding of Civil War debt replaced lowduration bonds with an almost equal amount of highduration bonds (a reverse Operation Twist29) and more than doubled the ten-year-equivalent size of outstanding U.S. Treasury bonds held by the public (see figure 2, panel A). The Civil War was largely financed by the issuance of legal tender notes (greenbacks) and the flotation of a series of bond offerings known to the market as the 5-20s and 10-40s. The 5-20s and 10-40s paid a 6 percent coupon maturing in 20 and 40 years, respectively, with embedded government call options vesting after five and ten years, respectively. By 1876, long-term interest rates on U.S. Treasury bonds had fallen well below 6 percent, and the Treasury took advantage of the lower prevailing rates by issuing option-free 4.5 percent, 15-year bonds in 1876 and 4 percent, 30-year bonds in 1877 at prices above par.30 The Treasury used the proceeds from these bond sales to retire the high-coupon 5-20s and 10-40s by exercising the call options embedded in them. Because the 5-20s and 10-40s had coupons well above both prevailing and forward interest rates, their embedded call options were deeply in the money31 and these bonds had lower durations compared with the new 15- and 30-year bond issuances that replaced them. The funding act that passed on February 25, 1862, instructed the Secretary of the Treasury to set aside the annual surplus from custom revenues for the establishment of a sinking fund to retire at least 1 percent of outstanding U.S. debt per annum by making open market purchases or by exercising embedded call options. There was no attempt to comply with the law during the Civil War and subsequent Treasury Secretaries used their own interpretations to largely ignore the sinking fund provision during the Depression of 1873 and the periods of bond issuance associated with the resumption of the gold-standard convertibility in 143 FIGURE 1 The size and interest rate risk of U.S. Treasury bonds, 1870–1913 billions of dollars 2.5 2.0 1.5 1.0 0.5 0 1871 ’74 ’77 ’80 ’83 ’86 ’89 Debt at face value ’92 ’95 ’98 1901 ’04 ’07 ’10 ’13 Debt in ten-year-equivalent units Notes: The data are month-end values. A ten-year equivalent is a common measurement for interest rate risk in a portfolio. This figure only includes positive duration obligations of the U.S. Treasury. Zero-duration liabilities, such as Treasury notes (cash in circulation), coinage, and pension funds, appear on the Treasury’s Monthly Statement of the Public Debt but have no effect on the ten-year-equivalent size of the U.S. portfolio and are excluded from the face value calculations. See the text for further details. Sources: Authors’ calculations based on data from the U.S. Department of the Treasury, Monthly Statement of the Public Debt database; De Knight (1900); and New York Stock Exchange quotations in the Commercial & Financial Chronicle, New York Times, and New York Tribune. January 1879. The issuance of new debt in the first half of 1879 resulted in a dramatic increase in the tenyear-equivalent duration (and, hence, the aggregate interest rate risk) of outstanding debt (see figure 2, panel A).32 Between 1879 and 1890, however, the Treasury enjoyed large fiscal surpluses and Treasury Secretaries regularly employed the sinking fund to retire outstanding Treasury debt.33 The Treasury retired debt by making open market purchases or exercising call options in 79 percent of the months over the period August 1879 through July 1891. By July 1891, the cumulative effect of making these purchases and exercising the call options had reduced the par value of Treasury bonds in the hands of the public by 68 percent and the ten-year-equivalent duration of outstanding Treasury debt by 59.4 percent. With one exception, the sinking fund purchases resulted in a nearly monotonic decline in both the duration and face value of Treasury bonds held by the public during the 1880s. The exception was a refunding of maturing debt in 1881, which increased the aggregate duration of outstanding Treasury debt by almost 15 percent while leaving the supply of Treasury bonds 144 virtually unchanged. A 5 percent bond matured on May 1, 1881, and three separate 6 percent coupon bonds matured on either June 30 or July 1, 1881. With interest rates on both secured overnight loans and longterm Treasury bonds close to 3 percent, the Treasury offered holders of these maturing bonds the choice of converting their bonds into 3.5 percent perpetual bonds callable at the pleasure of the government (that is, bonds without maturity dates that may be retired by the Treasury Department when it exercises their call options).34 Because of the uncertainty about how many bondholders would accept the conversion offer, the Treasury issued short-term refunding certificates that were redeemed in September 1881. These certificates added zero duration risk but accounted for the twomonth spike in the face value of outstanding Treasury bonds, as shown in figure 2, panel B. The conversion offer was accepted by approximately 90 percent of bondholders. While this conversion had virtually no effect on the total amount of Treasury debt outstanding, the introduction of callable perpetual bonds with coupons only 50 basis points above market rates dramatically increased the interest rate risk held by the public. 4Q/2013, Economic Perspectives FIGURE 2 The size and interest rate risk of U.S. Treasury bonds, by subperiods A. January 1877–July 1879 billions of dollars 2.5 2.0 1.5 1.0 0.5 1877 ’78 ’79 B. August 1879–July 1891 billions of dollars 2.5 2.0 1.5 1.0 0.5 1880 ’81 ’82 ’83 ’84 ’85 ’86 ’87 ’88 ’89 ’90 ’91 C. August 1891–December 1900 billions of dollars 1.6 1.4 1.2 1.0 0.8 0.6 0.4 1892 ’93 ’94 ’95 Debt at face value ’96 ’97 ’98 ’99 1900 Debt in ten-year-equivalent units Notes: The data are month-end values. A ten-year equivalent is a common measurement for interest rate risk in a portfolio. This figure only includes positive duration obligations of the U.S. Treasury. Zero-duration liabilities, such as Treasury notes (cash in circulation), coinage, and pension funds, appear on the Treasury’s Monthly Statement of the Public Debt but have no effect on the ten-year-equivalent size of the U.S. portfolio and are excluded from the face value calculations. The refunding of Civil War debt is covered in panel A. The refunding of 1881 and sinking fund purchases are covered in panel B. The deficit funding of the 1890s and the refunding of 1900 are covered in panel C. See the text for further details. Sources: Authors’ calculations based on data from the U.S. Department of the Treasury, Monthly Statement of the Public Debt database; De Knight (1900); and New York Stock Exchange quotations in the Commercial & Financial Chronicle, New York Times, and New York Tribune. Federal Reserve Bank of Chicago 145 By October 1891, sinking fund purchases had reduced the face value of Treasury bonds outstanding to $585.4 million. However, the recessions of 1890 and 1893 eliminated the budget surpluses the Treasury had relied upon for debt purchases. The Treasury was forced to reenter the market in 1894 and float bonds to replenish its rapidly dwindling stock of gold. Between 1894 and 1898, the Treasury issued option-free ten-year and 30-year bonds and a 20-year bond callable after ten years. By November 1898, the cumulative effect of these issues had raised the face value of Treasury bonds outstanding by 78 percent and the ten-yearequivalent duration of outstanding Treasury debt held by the public by 59 percent (see figure 2, panel C). By 1900, long-term interest rates had declined to less than 2 percent. The Treasury took advantage of these low rates by issuing a 30-year, 2 percent coupon bond at an initial yield to maturity of 1.82 percent. The majority of this bond offering was issued in a voluntary exchange for outstanding 5 percent coupon bonds due in 1904, 4 percent coupon bonds due in 1907, and callable 2 percent perpetual bonds. By replacing these low-duration bonds with a high-duration bond, the 1900 refunding acted like a reverse Operation Twist, which effectively increased the ten-year-equivalent interest rate risk held by the public by 39.5 percent in just three months. Constructing a test portfolio of highduration bonds minus low-duration bonds A number of open market operations are apparent in figure 1 (p. 144). Periods without new issuance of Treasury debt resulted in a decline in the duration of Treasury bonds held by the public with no corresponding change in the amount of bonds outstanding that could serve as collateral. Likewise, the federal government’s refunding of maturing debt with new longterm bond issuance resulted in jumps in the duration of Treasury debt held by the public without a corresponding change in bonds outstanding. Finally, the federal government’s open market sinking fund purchases resulted in a decrease in both the amount outstanding and duration of Treasury bonds held by the public. These operations allow us to disentangle changes in duration risk of outstanding Treasury debt from changes in the amount of Treasury bonds available as collateral. We evaluate the relative importance of changes in the aggregate duration and total supply of Treasury bonds by constructing a portfolio of high-duration bonds minus low-duration bonds. We construct this portfolio by sorting all bonds with observable market prices in each time period into a basket of high-duration bonds 146 or one of low-duration bonds based on whether the bond’s duration is above or below the median duration of the set of bonds in existence on that date. The portfolio of high-duration bonds minus low-duration bonds is then formed by computing the difference between the holding-period returns of the basket of equally weighted high-duration bonds and the basket of equally weighted low-duration bonds. We refer to this test portfolio as the high-minus-low (HML) portfolio. We can use the return on the HML portfolio to measure the relative importance of changes in aggregate duration and local supply of Treasury bonds outstanding. If there is a duration-risk premium, it should be apparent in the return of the HML portfolio. Highduration bonds expose their investors to more interest rate risk. Therefore, the price of high-duration Treasury bonds should be more sensitive to changes in the aggregate duration of Treasury bonds held by the public. That said, because of the more volatile market price of highduration Treasury bonds, they often require larger haircuts and are considered worse collateral than low-duration Treasury bonds. We would expect, all else being equal, high-duration Treasury bonds to be less sensitive than low-duration Treasury bonds to changes in the total supply of Treasury bond collateral available to the market.35 Figure 3 plots the index of cumulative holdingperiod returns of the HML portfolio. Consistent with a positive term premium,36 the high-duration bonds outperformed the low-duration bonds by an average of 22 basis points per year over our sample period. However, there were long periods where high-duration bonds dramatically outperformed or underperformed low-duration bonds, and these swings in their relative performance often coincided with changes in the amount (as represented by the face value measure in figure 3) or duration (as represented by the ten-year-equivalent measure) of outstanding Treasury bonds. For example, high-duration bonds outperformed low-duration bonds during the 1880s—when both the supply and aggregate duration risk of Treasury bonds held by the public dramatically declined—and underperformed lowduration bonds in the 1890s—when the supply and aggregate duration risk of Treasury bonds held by the public increased. This is evident in figure 3, which shows an increase in the HML portfolio index during the 1880s and a decrease in the 1890s. Measuring the effects of historical LSAPs The numerous refundings and sinking fund purchases between 1870 and 1913 provide us with a unique laboratory in which to measure the sensitivity of bond prices (and yields) to changes in the quantity or duration of outstanding Treasury debt. Moreover, 4Q/2013, Economic Perspectives FIGURE 3 The size and interest rate risk of U.S. Treasury bonds and the HML portfolio index, 1871–1913 billions of dollars 2.5 index, Jan. 28, 1871=100 130 2.3 125 2.1 120 1.9 115 1.7 110 1.5 1.3 105 1.1 100 0.9 95 0.7 0.5 Jan. 28, 1871 90 Jan. 20, 1877 Jan. 13, 1883 Jan. 5, 1889 Jan. 5, 1895 Dec. 28, 1900 Dec. 21, 1906 Dec. 20, 1912 Debt at face value (left-hand scale) Debt in ten-year-equivalent units (left-hand scale) High-duration bonds minus low-duration bonds (high-minus-low, or HML) test portfolio (right-hand scale) Notes: The data are collected every 28 days because of the HML test portfolio, which relies on prices that are collected from the Commercial & Financial Chronicle (a weekly publication) that published Friday prices. A ten-year equivalent is a common measurement for interest rate risk in a portfolio. This figure only includes positive duration obligations of the U.S. Treasury. Zero-duration liabilities, such as Treasury notes (cash in circulation), coinage, and pension funds, appear on the Treasury’s Monthly Statement of the Public Debt but have no effect on the ten-year-equivalent size of the U.S. portfolio and are excluded from the face value calculations. See the text for further details. Sources: Authors’ calculations based on data from the U.S. Department of the Treasury, Monthly Statement of the Public Debt database; De Knight (1900); and New York Stock Exchange quotations in the Commercial & Financial Chronicle, New York Times, and New York Tribune. the magnitude of the change in the size or duration of outstanding Treasury bonds due to pre-1913 LSAPs dwarfs that of the change in the size or duration of them due to the modern LSAPs. Pre-1913 interest rates were not constrained by a zero lower bound, and the Treasury did not target a short-run policy rate. Together, these facts make the period between 1870 and 1913 an ideal era for studying the sensitivity of bond prices to changes in the aggregate duration of Treasury bonds or their quantity available as collateral without the confounding effects of policy rate signaling by a central bank. However, the small number of Treasury bonds in existence during the pre-1913 era makes it difficult, if not impossible, to identify price changes due to the traditional scarcity channel of bond purchases. Recall that the effects of the scarcity channel should only be reflected in the prices of similar, substitutable securities. Researchers studying modern LSAPs measure the scarcity channel by carefully selecting bonds with cash flow characteristics Federal Reserve Bank of Chicago that are practically identical to those of the bonds purchased by the central bank.37 During the period between 1870 and 1913, the Treasury seldom had more than a handful of different bonds outstanding at any given time, and these bonds differed greatly with respect to embedded options, maturity, and coupon rate. As a result, close substitutes are hardly ever available. We can nonetheless infer the existence of preferred habitat investors by examining the effects of changes in the total supply and aggregate duration of Treasury bonds on the holding-period returns of high- and low-duration bonds. We measure the effects of changes in the total supply and aggregate duration of U.S. Treasury bonds by estimating the following monthly regression: 3) Ret HML_Dur = α + β1 (%∆AggDur ) + β2 (%∆FV ) + β3 (∆% HD) + β4 ( Retall _ bonnds ) + ε , 147 where RetHML_Dur is the holding-period return on the HML portfolio; %DAggDur is the percentage change in aggregate ten-year-equivalent duration outstanding; %DFV is the percentage change in the aggregate face value of U.S. Treasury bonds outstanding; D%HD is the change in the proportion of high-duration Treasury bonds outstanding relative to all Treasury bonds outstanding, with the proportion defined as (FVHDbonds /FV), where FVHDbonds and FV are the face value of bonds with durations above the median and the face value of all bonds, respectively; and Retall_bonds is the holding-period return on the equally weighted portfolio of all U.S. Treasury bonds outstanding. The a and β coefficients are free parameters to estimate; the βs measure the sensitivity of the HML portfolio return to changes in our variables of interest. And ε is the error term. The variable %DAggDur is our measure of the duration channel. The coefficient on %DAggDur tells us the difference in sensitivity of high-duration bonds and low-duration bonds to changes in the aggregate duration risk of Treasury bonds held by the public. If pre-1913 investors required compensation for holding duration risk in Treasury bonds in proportion to the quantity of duration in the bonds held by the public, we would expect high-duration bonds to be more sensitive to increases in the ten-year-equivalent size of Treasury bonds outstanding than low-duration bonds and also anticipate the coefficient on %DAggDur to be negative. The variables %DFV and D%HD are our measures of the scarcity channel. To the extent that lowduration bonds are preferred for collateral purposes, increases in the aggregate amount of Treasury bond collateral outstanding should decrease the price of low-duration bonds more than that of high-duration bonds; we would, therefore, expect the coefficient on %DFV to be positive. Likewise, if the total supply is fixed, an increase in the relative proportion of highduration bonds should decrease the price of plentiful high-duration bonds and raise the price of scarce lowduration bonds; we would, therefore, expect the coefficient on D%HD to be negative. Finally, we include the holding-period return on the market portfolio of all bonds to control for the fact that the high-duration bonds are more sensitive, by construction, to shifts in the yield curve. Our regression estimates can be found in table 1. The results in table 1 confirm that changes to the aggregate duration and total supply of Treasury bonds outstanding altered equilibrium prices. The coefficients on both aggregate duration and face value have the predicted sign and are economically and statistically 148 TABLE 1 Regression results 1 a 0.0003 (1.10) %ΔAggDur – 0.0336** (– 2.43) 2 0.0001 (0.12) – 0.0353*** (– 2.59) %ΔFV 0.0470*** (2.91) 0.0466*** (2.93) Δ%HD 0.0074* (1.86) 0.0073* (1.86) Retall_bonds 0.1235*** (3.23) R2 0.0095 0.0171 *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level. Notes: The two regressions take the following form: RetHML_Dur = α + β1(%∆AggDur ) + β2 (%∆FV ) + β3 (∆%HD) + β4 (Retall _ bondss ) + ε . See the text for details on the variables. The results in the first column are for the regression run without the return on all bonds. The results in the second column are for the regression run with the return on all bonds. The Newey–West t statistics are in parentheses. Sources: Authors’ calculations based on data from the U.S. Department of the Treasury, Monthly Statement of the Public Debt database; De Knight (1900); and New York Stock Exchange quotations in the Commercial & Financial Chronicle, New York Times, and New York Tribune. significant. Our point estimates suggest that removing 10 percent of the aggregate duration risk held by the public increased the price of interest-rate-sensitive highduration bonds by 35 basis points relative to that of low-duration bonds (table 1, regression 2, %ΔAggDur coefficient). Likewise, a 10 percent decrease in the face value of all Treasury bond collateral available to the market raised the price of the low-duration bonds (serving as good collateral) by 47 basis points relative to that of high-duration bonds (table 1, regression 2, %ΔFV coefficient). Both of these results are consistent with a model featuring preferred habitat investors who value low-duration bonds for collateral purposes and arbitrageurs who require compensation for bearing duration risk in proportion to the aggregate amount of duration risk held by the public. The coefficient on D%HD, however, is not consistent with the preferred habitat model. With the total supply and aggregate duration of Treasury bonds outstanding held constant, a 10 percent increase in the proportion of bonds with above-median duration 4Q/2013, Economic Perspectives actually increased the price of high-duration bonds relative to that of low-duration bonds (table 1, regression 2, Δ%HD coefficient). This result is inconsistent with the theory. In our opinion, the most likely explanation is that while our sorting procedure does a good job of identifying which bonds have more duration risk or are likely to be substitutes for collateral purposes, our procedure is too coarse to capture the effects of the scarcity channel, where changes in the local supply of Treasury bonds will only manifest themselves in the prices of the purchased assets and close substitutes. Unlike today’s market where the breadth of Treasury offerings assures us that very similar Treasury bonds always exist, the pre-1913 Treasury market seldom had more than four U.S. Treasury bond issues trading at any given time. Therefore, when a refunding alters the proportion of Treasury bonds with high duration, we are not measuring the relative change in price of a Treasury bond very similar to the new Treasury bond issue; rather, we are looking at the relative change in price of a bond that most likely differs in coupon rate, convexity, years to maturity, and the terms of its embedded option. Because changes in total Treasury bond collateral or aggregate duration of Treasury debt outstanding would affect the prices of all Treasury bonds, these sorting constraints are less likely to affect the coefficients on face value of total collateral or aggregate duration. In light of the face value and duration results, we think the most likely explanation of our result for the coefficient on D%HD is that the pre1913 Treasury bond offerings were too sparse for us to measure the scarcity channel. Conclusion There are few examples of central banks employing their balance sheets for policy purposes in the past 50 years or so. If one looks at periods before the Federal Reserve, however, large-scale asset purchases and operations like Operation Twist are far more common than previously thought. Between 1870 and 1913, the U.S. Department of the Treasury engaged in a number of refundings and sinking fund purchases, which altered Federal Reserve Bank of Chicago the duration risk and amount of Treasury bond collateral in the hands of the public. While the pre-1913 purchases were not conducted with an eye toward stimulating the economy through reaching new equilibrium bond prices, their effects on the size and duration risk of the aggregate portfolio of Treasury bonds held by the public were very similar to those of modern central bank LSAPs. The changes in Treasury bond yields due to Treasury bond purchases suggest a duration channel was present in the pre-1913 bond market. Sinking fund purchases or refundings that removed duration from the aggregate portfolio of Treasury bonds held by the public resulted in a narrowing of the yield spread between high- and low-duration Treasury bonds, consistent with a decrease in the term premium on highduration bonds. While open market purchases of Treasury bonds lowered equilibrium bond yields through the duration channel, the price effect of the removal of bonds from the portfolios of the public was not unambiguously positive. Models of segmented markets where quantities affect equilibrium prices imply that Treasury bonds provide a service that cannot be replicated by privately produced assets. Most likely, this service is the provision of a safe and liquid asset to serve as collateral. While all Treasury bonds are safe in terms of default risk, high-duration bonds that expose their owners to more interest rate risk are less valuable for collateral purposes. With the amount of aggregate duration held constant, a decrease in the face value of aggregate Treasury bonds outstanding was associated with a decrease in the price of high-duration bonds relative to their less risky lowduration counterparts. The behavior of bond prices between 1870 and 1913 is consistent with a segmented bond market in which participants valued safe and liquid bonds and required a duration-risk premium to hold high-duration assets. In such a setting, open market purchases that alter the amount or interest rate risk of Treasury bonds held by the public can stimulate the economy by generating changes in equilibrium bond yields. 149 NOTES For further explanations of LSAPs and the Federal Reserve’s rationale in making them, see www.federalreserve.gov/faqs/what-arethe-federal-reserves-large-scale-asset-purchases.htm. Recent empirical studies that examine LSAPs’ effectiveness include Krishnamurthy and Vissing-Jørgensen (2011, 2012), Gagnon et al. (2011), D’Amico and King (2013), Hamilton and Wu (2012), Hancock and Passmore (2011), Joyce et al. (2011), Neely (2013), Christensen and Rudebusch (2012), and Bauer and Rudebusch (2014). 1 For more on reserve cities and the National Banking System as a whole, see Champ (2011). 11 A call market is an overnight lending market where borrowers pledge collateral for a secured loan repayable on demand (which is called a call loan). See Griffiss (1923) for a description of the U.S. call market before the Federal Reserve. 12 In a collateralized loan, the haircut is the percentage by which the collateral asset’s market value is reduced to provide a cushion against the possibility that the collateral will decline in value before the loan can be repaid. For example, if a security with a $100 market value can secure an $80 loan, the asset has a 20 percent haircut. 13 See Krishnamurthy and Vissing-Jørgensen (2011), Fuster and Willen (2010), Hancock and Passmore (2011), and Wright (2011). In this article, a bond’s duration (that is, its Macaulay duration) is a measure of its sensitivity to changes in interest rates (equation 1 shows how the Macaulay duration is calculated). Moreover, duration is an approximation of the percentage change in a bond’s price for a 100 basis point change in its yield to maturity. The greater an asset’s duration is, the higher its sensitivity to interest rates changes— meaning that the asset’s price fluctuations due to interest rate changes will be more pronounced. Hence, duration risk is a measure of interest rate risk—which is the risk that an investment’s value will be altered because of a change in the absolute level or shape of the yield curve (that is, the line plotting the interest rates of assets of the same credit quality but with differing maturity dates at a certain point in time). 2 For the differences in magnitude of yield changes, see Williams (2013, table 1). For details about the channels by which central banks’ LSAPs influence asset yields, see Krishnamurthy and Vissing-Jørgensen (2013). 3 These episodes include the Federal Reserve’s large open market purchases of bonds since 2008 and the Bank of England’s large open market purchases since 2009; the Bank of Japan’s large open market purchases since 1987; and the Federal Reserve’s Operation Twist in the 1960s, which involved the sale of short-maturity bonds and purchase of long-maturity bonds (for details, see Alon and Swanson, 2011). 4 A refunding is the process of redeeming an outstanding bond issue at its maturity with the proceeds of a new debt issue. A sinking fund is a fund set up by a government agency (or corporation) for the purpose of periodically acquiring outstanding bonds via redemption or open market purchases (to retire debt). 5 Equilibrium values (for bond yields, prices, etc.) are the values that equalize a bond’s supply with its demand. 6 For more on interest rate risk, see note 2. Authors’ calculations based on a sample of 86 percent of all call loans appearing on insurance balance sheets reported in the State of New York, Insurance Department (1873). 14 15 Ibid. Authors’ calculations based on data from the Office of the Comptroller of the Currency (1872). 16 An option is a contract giving its owner the right, but not the obligation, to buy or sell a particular asset at a specified price on or before a specified date. In the case of our sample Treasury bonds, many granted the Treasury the option to buy back the bonds at face value. 17 18 See www.newyorkfed.org/markets/lttreas_faq.html. For a definition of duration risk (which is related to interest rate risk), see note 2. 19 20 Bauer and Rudebusch (2014). The MSPD database is available at http://treasurydirect.gov/govt/ reports/pd/mspd/mspd.htm. 21 22 For a definition of yield curve, see note 2. The implied volatility is a level of volatility that sets the modelimplied price of an option equal to the observed market price. 23 24 See Hull and White (1996, 2000). 7 See Tobin (1965, 1969) and Modigliani and Sutch (1966). 8 A bond’s convexity is a measure of the sensitivity of the bond’s duration to changes in its yield to maturity. A negative convexity indicates that the duration of a bond rises as its yield to maturity increases (and its price decreases); a positive convexity indicates that the duration of a bond rises as its yield to maturity decreases (and its price increases). 9 10 Short selling, or shorting, is the selling of a security that the seller does not own but has promised to deliver later (usually to the party from which the seller borrowed it). It is motivated by the belief that a security’s price will decline—which would enable the short seller to make a profit when the security is bought back at a cheaper price. 150 A zero-coupon interest rate is the yield to maturity on a bond with a single cash flow payment at its maturity. The zero-coupon yield curve is the line plotting the interest rates of zero-coupon bonds with differing maturity dates at a certain point in time. 25 In the modern era, the zero-coupon yield curve is directly observable from the market prices of zero-coupon STRIPS (Separate Trading of Registered Interest and Principal of Securities), and only the volatility and mean-reversion parameters of the Hull–White model are calibrated to the observable bond price data. However, there was no STRIPS market during our period of study, and the number of existing coupon bonds was always too sparse to identify a unique zero curve. 26 Specifically, we search over a grid of level, slope, and volatility to find the values that minimize the Euclidean distance between the prices implied by the Hull–White model and the observable market prices. 27 4Q/2013, Economic Perspectives Unless otherwise indicated, all the numerical values related to Treasury bonds reported in this section are from authors’ calculations based on data from the sources in figure 1. Similarly, unless otherwise indicated, the historical details provided here are based on the authors’ interpretations of the information from those sources. 28 29 For more on Operation Twist of the 1960s, see note 4. The 4.5 percent, 15-year bonds due in 1891 began trading at a price of 111.25 percent of par; and the 4 percent, 30-year bonds due in 1907 began trading at a price of 105.5 percent of par. The NYSE overnight call rate on loans backed by government bond collateral was in the range of 2.5–3 percent in the month before the conversion, and the secondary market yields on optionfree Treasury bonds maturing in ten and 16 years were 2.85 percent and 3.06 percent, respectively. 34 Griffiss (1923) and Chabot (2011) discuss the importance of the collateralized lending market before 1913. 35 30 An option is considered in the money when the option grants the option holder the right to sell an asset at a price above current market value or to buy it at a price below market value. 31 See Ross (1892, pp. 79–85) for a history of the sinking fund of 1862. 32 33 The term premium is the excess return that investors require to hold a long-term bond rather than a series of short-term bonds (this compensation is required because long-term bonds face higher interest rate risk than short-term bonds). 36 See, for example, the CUSIP (Committee on Uniform Securities Identification Procedures) level matching of D’Amico and King (2013). For details on CUSIP, see www.sec.gov/answers/cusip.htm. 37 Ibid. REFERENCES Alon, Titan, and Eric Swanson, 2011, “Operation Twist and the effect of large-scale asset purchases,” FRBSF Economic Letter, Federal Reserve Bank of San Francisco, No. 2011-13, April 25. De Knight, William F. (preparer), 1900, History of the Currency of the Country and of the Loans of the United States from the Earliest Period to June 30, 1900, 2nd ed., Washington, DC: Government Printing Office. Bauer, Michael D., and Glenn D. Rudebusch, 2014, “The signaling channel for Federal Reserve bond purchases,” International Journal of Central Banking, forthcoming. Fuster, Andreas, and Paul S. Willen, 2010, “$1.25 trillion is still real money: Some facts about the effects of the Federal Reserve’s mortgage market investments,” Public Policy Discussion Papers, Federal Reserve Bank of Boston, No. 10-4, November 18. Chabot, Benjamin, 2011, “The cost of banking panics in an age before ‘too big to fail,’” Federal Reserve Bank of Chicago, working paper, No. WP 2011-15, November 28. Champ, Bruce, 2011, “The National Banking System: A brief history,” Federal Reserve Bank of Cleveland, working paper, No. 07-23R, revised May 2011, available at www.clevelandfed.org/research/ workpaper/2007/wp0723r.pdf. Christensen, Jens H. E., and Glenn D. Rudebusch, 2012, “The response of interest rates to US and UK quantitative easing,” Economic Journal, Vol. 122, No. 564, November, pp. F385–F414. D’Amico, Stefania, and Thomas B. King, 2013, “Flow and stock effects of large-scale treasury purchases: Evidence on the importance of local supply,” Journal of Financial Economics, Vol. 108, No. 2, May, pp. 425–448. Federal Reserve Bank of Chicago Gagnon, Joseph, Matthew Raskin, Julie Remache, and Brian Sack, 2011, “The financial market effects of the Federal Reserve’s large-scale asset purchases,” International Journal of Central Banking, Vol. 7, No. 1, March, pp. 3–43. Griffiss, Bartow, 1923, The New York Call Money Market, Johns Hopkins University, dissertation, June. Hamilton, James D., and Jing Cynthia Wu, 2012, “The effectiveness of alternative monetary policy tools in a zero lower bound environment,” Journal of Money, Credit and Banking, Vol. 44, No. 1 supplement, February, pp. 3–46. Hancock, Diana, and Wayne Passmore, 2011, “Did the Federal Reserve’s MBS purchase program lower mortgage rates?,” Journal of Monetary Economics, Vol. 58, No. 5, July, pp. 498–514. Hull, John C., and Alan D. White, 2000, “The general Hull-White model and super calibration,” New York University, working paper, No. FIN-00-024, August. 151 __________, 1996, “Using Hull-White interest rate trees,” Journal of Derivatives, Vol. 3, No. 3, Spring, pp. 26–36. Joyce, Michael A. S., Ana Lasaosa, Ibrahim Stevens, and Matthew Tong, 2011, “The financial market impact of quantitative easing in the United Kingdom,” International Journal of Central Banking, Vol. 7, No. 3, September, pp. 113–161. Krishnamurthy, Arvind, and Annette VissingJørgensen, 2013, “The ins and outs of LSAPs,” paper at Federal Reserve Bank of Kansas City Economic Policy Symposium, Global Dimensions of Unconventional Monetary Policy, Jackson Hole, WY, August 22–24, revised September 16, 2013. __________, 2012, “The aggregate demand for Treasury debt,” Journal of Political Economy, Vol. 120, No. 2, April, pp. 233–267. __________, 2011, “The effects of quantitative easing on interest rates: Channels and implications for policy,” Brookings Papers on Economic Activity, Vol. 43, No. 2, Fall, pp. 215–265. Modigliani, Franco, and Richard Sutch, 1966, “Innovations in interest rate policy,” American Economic Review, Vol. 56, Nos. 1–2, March 1, pp. 178–197. Neely, Christopher J., 2013, “Unconventional monetary policy had large international effects,” Federal Reserve Bank of St. Louis, working paper, No. 2010-018D, revised August 2013. Office of the Comptroller of the Currency, 1872, Annual Report of the Comptroller of the Currency, Washington, DC: Government Printing Office, December 2, available at http://fraser.stlouisfed.org/ docs/publications/comp/1870s/1872/compcurr_1872.pdf. 152 Ross, Edward Alsworth, 1892, Sinking Funds, Publications of the American Economic Association, Vol. 7, Nos. 4–5, Baltimore: Press of Guggenheimer, Weil & Co., July and September. State of New York, Insurance Department, 1873, Fourteenth Annual Report of the Superintendent of the Insurance Department, Part 2, Life & Casualty Insurance, Albany, NY: Weed, Parsons & Co. Tobin, James, 1969, “A general equilibrium approach to monetary theory,” Journal of Money, Credit and Banking, Vol. 1, No. 1, February, pp. 15–29. __________, 1965, “The monetary interpretation of history,” American Economic Review, Vol. 55, No. 3, June, pp. 464–485. Vayanos, Dimitri, and Jean-Luc Vila, 2009, “A preferred-habitat model of the term structure of interest rates,” National Bureau of Economic Research, working paper, No. 15487, November. Williams, John C., 2013, “Lessons from the financial crisis for unconventional monetary policy,” panel discussion by president and CEO of the Federal Reserve Bank of San Francisco at the NBER Conference, Boston, October 18, available at www.frbsf.org/our-district/ press/presidents-speeches/williams-speeches/2013/ october/research-unconventional-monetary-policyfinancial-crisis/. Wright, Jonathan H., 2011, “What does monetary policy do to long-term interest rates at the zero lower bound?,” National Bureau of Economic Research, working paper, No. 17154, June. 4Q/2013, Economic Perspectives