Conducting Monetary Policy at Very Low Short-Term Interest Rates

Conducting Monetary Policy at Very Low
                           Short-Term Interest Rates
                           By BEN S. BERNANKE            AND       VINCENT R. REINHART*

   Can monetary policy committees, accus-                          the costs and benefits of very low interest rates,
tomed to describing their plans and actions in                     an issue that bears on the question of whether
terms of the level of a short-term nominal in-                     the central bank should take the policy rate all
terest rate, find effective means of conducting                    the way to zero before undertaking alternative
and communicating their policies when that rate                    policies.
is zero or close to zero? The very low levels of
interest rates in Japan, Switzerland, and the                           I. Shaping Interest-Rate Expectations
United States in recent years have stimulated
much interesting research on this question and                        The pricing of long-lived financial assets,
have led some central banks to make changes in                     such as equities and mortgages, depends in part
their operating procedures and communications                      on the entire expected future path of short-term
strategies. In this paper, we will give a brief                    interest rates, as well as on the current short-
overview of current thinking on the conduct of                     term rate. Hence, a central bank can affect asset
monetary policy when short-term interest rates                     prices and economic activity by influencing
are very low or even zero.1                                        market participants’ expectations of future
   Monetary policy works for the most part by                      short-term rates. Recent research has examined
influencing the prices and yields of financial                     this potential channel of influence in fully artic-
assets, which in turn affect economic decisions                    ulated models (see Lars Svensson, 2001; Gauti
and thus the evolution of the economy. When                        Eggertsson and Michael Woodford, 2003;
the short-term policy rate is at or near zero, the                 Woodford, 2003 Ch. 6). This literature suggests
conventional means of effecting monetary ease                      that, even with the overnight nominal interest
(lowering the target for the policy rate) is no                    rate at or near zero, additional stimulus can be
longer feasible. However, it would be a mistake                    imparted by offering some form of commitment
to think that monetary policy was impotent. We                     to the public to keep the short rate low for a
discuss three strategies for stimulating the econ-                 longer period than previously expected. This
omy at an unchanged level of the policy rate:                      commitment, if credible, should lower yields
these involve (i) providing assurance to inves-                    throughout the term structure and support other
tors that short rates will be kept lower in the                    asset prices.
future than they currently expect, (ii) changing                      In principle, the central bank’s policy com-
the relative supplies of securities in the market-                 mitment could be either unconditional or con-
place by altering the composition of the central                   ditional. An unconditional commitment is a
bank’s balance sheet, and (iii) increasing the                     pledge by the central bank to hold short-term
size of the central bank’s balance sheet beyond                    rates at a low level for a fixed period of calendar
the level needed to set the short-term policy rate                 time. In this case, additional easing would take
at zero (“quantitative easing”). We also discuss                   the form of lengthening the period of commit-
                                                                   ment. However, given the many uncertainties
                                                                   that affect the outlook, a policy-making com-
   * Board of Governors of the Federal Reserve System,
                                                                   mittee might understandably be reluctant to
Washington, DC 20551. We have benefited from the re-               make an unconditional promise about policy
search of, and many discussions with, numerous colleagues,         actions far into the future. An alternative is to
as well as the comments of our discussant Carl Walsh.              make a conditional policy commitment that
However, the views expressed here are our own and are not          links the duration of promised policies not to
necessarily shared by anyone else in the Federal Reserve
System.                                                            the calendar, but to economic conditions. For
     See Bernanke (2002) and James Clouse et al. (2003) for        example, policy ease could be promised until
earlier discussions of these issues.                               the committee observes sustained economic
86                                        AEA PAPERS AND PROCEEDINGS                                   MAY 2004

growth, a decline in resource slack, or inflation            the government debt is large and not indexed to
above a specified floor.                                     inflation, and (ii) the central bank values the
   In practice, central banks appear to appreciate           reduction in fiscal burden that reflationary pol-
the importance of influencing market expecta-                icies will bring (for example, because it may
tions about future policy. For example, in May               reduce the future level of distortionary taxa-
2001, with its policy rate virtually at zero, the            tion). Ultimately, however, the central bank’s
Bank of Japan promised that it would keep its                best strategy for building credibility is to build
policy rate at zero as long as the economy                   trust by ensuring that its deeds match its words.
experienced deflation—a conditional policy                   Hence, the shaping of expectations is not an
commitment. More recently, the Bank of Japan                 independent policy instrument in the long run.
has been more explicit about the conditions
under which it would begin to raise rates; for                 II. Altering the Composition of the Central
example, it has specified that a change from                              Bank’s Balance Sheet
deflation to inflation that is perceived to be
temporary will not provoke a tightening.2 In the                Central banks typically hold a variety of as-
United States, the August 2003 statement of the              sets, and the composition of assets on the cen-
Federal Open Market Committee (FOMC) that                    tral bank’s balance sheet offers another
“policy accommodation can be maintained for a                potential lever for monetary policy. For exam-
considerable period” is another example of a                 ple, the Federal Reserve participates in all seg-
commitment by policymakers. The close asso-                  ments of the Treasury market, with most of its
ciation of this statement with the FOMC’s                    current asset holdings of about $650 billion
expressed concerns about “unwelcome disinfla-                distributed among Treasury securities with ma-
tion” implied that this commitment was condi-                turities ranging from four weeks to 30 years. As
tioned on the assessment of the economy. The                 an important participant in the Treasury market,
conditional nature of the commitment was                     the Federal Reserve might be able to influence
sharpened in the FOMC’s December statement,                  term premiums, and thus overall yields, by
which explicitly linked continuing policy ac-                shifting the composition of its holdings, say,
commodation to the low level of inflation and                from shorter- to longer-dated securities. In sim-
the slack in resource use.3 More generally, in               ple terms, if the liquidity or risk characteristics
recent years central banks have worked hard to               of securities differ, so that investors do not treat
improve communication with the public; a key                 all securities as perfect substitutes, then changes
objective of this effort is better alignment of              in relative demands by a large purchaser have
market expectations of policy with the policy-               the potential to alter relative security prices. The
making committee’s own intentions.                           same logic might lead the central bank to
   Of course, policy commitments can influence               consider purchasing assets other than govern-
future expectations only to the extent that they             ment securities, such as corporate bonds or
are credible. Various devices might be em-                   stocks or foreign government bonds. (The Fed-
ployed to enhance credibility, including trans-              eral Reserve is currently authorized to purchase
acting in financial markets in ways that makes it            some foreign government bonds but not most
costly to renege, such as writing options                    private-sector assets, such as corporate bonds or
(Clouse et al., 2003). An objection to this strat-           stocks.)
egy is that it is not obvious why a central bank,               An extreme example of a policy keyed to
which has the power to print money, should be                the composition of the central bank’s balance
overly concerned about financial gains and                   sheet is the announcement of a ceiling on some
losses. Eggertsson and Woodford (2003) point                 longer-term yield below the prevailing rate.
out that a government can more credibly prom-                This policy entails (in principle) an unlimited
ise to carry out policies that raise prices when (i)         commitment to purchase the targeted security at
                                                             the announced price. (To keep the overall size
                                                             of its balance sheet unchanged, the central bank
     For a recent appraisal of monetary policy options in
Japan, see Bernanke (2003).                                  would have to sell other securities in an amount
     FOMC statements are available at 具www.federalreserve.   equal to the purchases of the targeted security.)
gov/boarddocs/press/monetary典.                               Obviously, this policy would signal the policy-
VOL. 94 NO. 2                      POLICIES TO DEAL WITH DEFLATION                                    87

makers’ strong dissatisfaction with current mar-      constraint. In contrast, the Federal Reserve’s
ket expectations of future policy rates.              relative holdings of longer-dated Treasury notes
   Whether policies based on manipulating the         and bonds fell over the period, although the rate
composition of the central-bank balance sheet         ceilings at these longer maturities were not
can be effective is contentious. The limited em-      breached until inflation pressures led to the
pirical evidence suggests that, within broad          Fed–Treasury Accord and the abandonment of
classes, assets are close substitutes, so that        the pegging policy in 1951. The conventional
changes in relative supplies of the scale ob-         interpretation is that long-run policy expecta-
served in U.S. experience are unlikely to have a      tions must have been consistent with the ceil-
major impact on risk premiums or term premi-          ings at the more distant points on the yield
ums (Reinhart and Brian Sack, 2000). If this          curve. Less clear is the extent to which the
view is correct, then attempts to enforce a floor     pegging policy itself influenced those policy
on the prices of long-dated Treasury securities       expectations.
(for example) could be effective only if the             Probably the safest conclusion about policies
target prices were broadly consistent with in-        based on changing the composition of the cen-
vestor expectations of future values of the pol-      tral bank’s balance sheet is that they should be
icy rate. If, to the contrary, investors doubted      used only to supplement other policies, such as
that short rates would be kept low, the central       an attempt to influence expectations of future
bank could end up owning all or most of the           short rates. This combined approach allows the
targeted securities. Moreover, even if large pur-     central bank to enjoy whatever benefits arise
chases of, say, a long-dated Treasury security        from changing the relative supplies of outstand-
were able to affect the yield on that security, the   ing securities without risking the problems that
policy may not have significant economic ef-          may arise if the yields desired by the central
fects if the targeted security becomes “discon-       bank are inconsistent with market expectations.
nected” from the rest of the term structure and
from private rates, such as mortgage rates. Yet        III. Expanding the Size of the Central Bank’s
another complication affecting this type of pol-                      Balance Sheet
icy is that the central bank’s actions would have
to be coordinated with the central government’s          Besides changing the composition of its bal-
finance department to ensure that changes in          ance sheet, the central bank can also alter policy
debt-management policies do not offset the at-        by changing the size of its balance sheet, that is,
tempts of the central bank to affect the relative     by buying or selling securities to affect the
supplies of securities.                               overall supply of reserves and the money stock.
   Despite these potential pitfalls, policies based   Of course, this strategy represents the conven-
on changing the composition of the central            tional means of conducting monetary policy, as
bank’s balance sheet have been tried in the           described in many textbooks. These days, most
United States. From 1942 to 1951, the Federal         central banks choose to calibrate the degree of
Reserve enforced rate ceilings at two and some-       policy ease or tightness by targeting the price of
times three points on the Treasury yield curve.       reserves—in the case of the Federal Reserve,
This objective was accomplished with only             the overnight federal funds rate. However, noth-
moderate increases in the Federal Reserve’s           ing prevents a central bank from switching its
overall holdings of Treasury securities, relative     focus from the price of reserves to the quantity
to net debt outstanding; moreover, there is little    or growth of reserves. In particular, even if the
evidence that the targeted yields became “dis-        price of reserves (the overnight rate) becomes
connected” from other public or private yields.       pinned at zero, the central bank can still expand
The episode is an intriguing one, but unfortu-        the quantity of reserves beyond the level re-
nately the implications for current policy are not    quired to hold the overnight rate at zero, a
entirely straightforward. We know that, by            policy sometimes referred to as “quantitative
1946, the Federal Reserve System owned al-            easing.” Some evidence exists that quantitative
most nine-tenths of the (relatively small) stock      easing can stimulate the economy even when
of Treasury bills, suggesting that, at the short      interest rates are near zero (see e.g., Christina
end, the ceiling on the bill rate was a binding       Romer’s [1992] discussion of the effects of
88                                   AEA PAPERS AND PROCEEDINGS                               MAY 2004

increases in the money supply during the Great        replaces a direct tax, say on labor, with the
Depression in the United States).                     inflation tax.) Alan J. Auerbach and Maurice
   Quantitative easing may affect the economy         Obstfeld (2003) have analyzed the fiscal and
through several possible channels. In particular,     expectational effects of a permanent increase in
if money is an imperfect substitute for other         the money supply along these lines. Note that
financial assets, then large increases in the         the expectational and fiscal channels of quanti-
money supply will lead investors to seek to           tative easing, though not the portfolio substitu-
rebalance their portfolios, raising prices and re-    tion channel, require the central bank to make a
ducing yields on alternative, non-money assets.       credible commitment not to reverse its open-
In turn, lower yields on long-term assets will        market operations, at least until certain condi-
stimulate economic activity. The possibility that     tions are met. Thus this approach also poses
monetary policy works through portfolio substi-       communication challenges.
tution effects, even in normal times, has a long         Japan once again provides the most recent
intellectual history, having been espoused by         case study. In the past two years, current-
both Keynesians (James Tobin, 1969) and mon-          account balances held by commercial banks at
etarists (Karl Brunner and Allan Meltzer, 1973).      the Bank of Japan have increased about five-
Recently, Javier Andres et al. (2003) have            fold, and the monetary base has risen to almost
shown how these effects might work in a               30 percent of nominal GDP. While deflation
general-equilibrium model with optimizing agents.     appears to have eased in Japan, it is difficult
   Quantitative easing may also work by alter-        to know how much of the improvement is due
ing expectations of the future path of policy         to monetary policy and, of the part due to
rates. For example, suppose that the central          monetary policy, how much is due to the zero-
bank commits itself to keeping reserves at a          interest-rate policy and how much to quantita-
high level, well above that needed to ensure a        tive easing.
zero short-term interest rate, until certain eco-
nomic conditions obtain. Theoretically, this ac-              IV. Sequencing and the “Costs”
tion is equivalent to a commitment to keep                         of Low Interest Rates
interest rates at zero until the economic condi-
tions are met, a type of policy we have already          The forms of monetary stimulus described
discussed. However, the act of setting and meet-      above can be used once the overnight rate has
ing a high reserves target is more visible, and       already been driven to zero or as a way of
hence may be more credible, than a purely             driving the overnight rate to zero. However, a
verbal promise about future short-term interest       central bank might choose to rely on alternative
rates. Moreover, this means of committing to a        policies while maintaining the overnight rate
zero interest rate will also achieve any benefits     somewhat above zero. For example, an attempt
of quantitative easing that may be felt through       to influence market expectations of future short
non-expectational channels.                           rates by means of a policy commitment or to
   Lastly, quantitative easing that is sufficiently   affect term premiums by changing the compo-
aggressive and perceived to be long-lived may         sition of the central bank’s balance sheet does
have expansionary fiscal effects. So long as          not require that the policy rate be at zero.
market participants expect a positive short-term      (Quantitative easing, of course, is not compati-
interest rate at some date in the future, the         ble with a positive overnight rate.) The appro-
existence of government debt implies a current        priate sequencing of policy actions depends on
or future tax liability for the public. In expand-    the perceived costs associated with very low or
ing its balance sheet by open-market purchases,       zero overnight interest rates, as well as on op-
the central bank replaces public holdings of          erational considerations and estimates of the
interest-bearing government debt with non-            likely effects of alternative combinations of pol-
interest-bearing currency or reserves. If the in-     icies on the economy.
crease in the monetary base is expected to               What costs are imposed on society by very
persist, then the expected interest costs of the      low short-term interest rates? Observers have
government and, hence, the public’s expected          pointed out that rates on financial instruments
tax burden decline. (Effectively, this process        typically priced below the overnight rate, such
VOL. 94 NO. 2                      POLICIES TO DEAL WITH DEFLATION                                   89

as liquid deposits, shares in money-market mu-        impression that monetary policy is ineffective.
tual funds, and collateralized borrowings in the      As we have stressed, that would indeed be a
“repo” market, would be squeezed toward zero          misimpression, as the central bank has means of
as the policy rate fell, prompting investors to       providing monetary stimulus other than the con-
seek alternatives. Short-term dislocations might      ventional measure of lowering the overnight
result, for example, if funds flowed in large         nominal interest rate. However, it is also true
amounts from money-market mutual funds into           that policymakers’ inexperience with these al-
bank deposits. In that case, some commercial-         ternative measures makes the calibration of pol-
paper issuers who have traditionally relied on        icy actions more difficult. Moreover, given the
money-market mutual funds for financing               important role for expectations in making many
would have to seek out new sources, while             of these policies work, the communications
banks would need to find productive uses for the      challenges would be considerable. Given these
deposit inflows and perhaps face changes in           difficulties, policymakers are well advised to act
regulatory capital requirements. In addition, li-     preemptively and aggressively to avoid facing
quidity in some markets might be affected (for        the complications raised by the zero lower
example, the incentive for reserve managers to        bound.
trade federal funds diminishes as the overnight
rate falls, probably thinning brokering in that                       REFERENCES
   In thinking about the costs associated with a      Andres, Javier; López-Salido, J. David and
low overnight rate, one should bear in mind the         Nelson, Edward. “Tobin’s Imperfect Asset
message of Milton Friedman’s classic essay on           Substitution in Optimizing General Equilib-
the optimal quantity of money (Friedman,                rium.” Unpublished manuscript presented at
1969). Friedman argued that an overnight inter-         the JMCB/Federal Reserve Bank of Chicago
est rate of zero is optimal, because a zero op-         James Tobin Symposium, 14 –15 November
portunity cost of liquidity eliminates the              2003.
socially wasteful use of resources to economize       Auerbach, Alan J. and Obstfeld, Maurice. “The
on money balances. From this perspective, the           Case for Open-Market Purchases in a Liquid-
costs of low short-term interest rates can be seen      ity Trap.” Working paper, University of
largely as adjustment costs, arising from the           California–Berkeley, 2003.
unwinding of schemes designed to make hold-           Bernanke, Ben. “Deflation: Making Sure ‘It’
ing transactions balances less burdensome.              Doesn’t Happen Here.” Remarks before the
These costs are real but are also largely transi-       National Economists’ Club, Washington,
tory and have limited sectoral impact. More-            DC, 21 November 2002.
over, to the extent that the affected institutions           . “Some Thoughts on Monetary Policy
have economic functions other than helping              in Japan.” Address to the Japan Society of
clients economize on money balances (for                Monetary Economics, Tokyo, 31 May 2003.
example, if some money market mutual funds            Brunner, Karl and Meltzer, Allan. “Mr. Hicks
have a comparative advantage in lending to              and the ‘Monetarists.’ ” Economica, Febru-
commercial-paper issuers), there is scope for           ary 1973, 40(157), pp. 44 –59.
repricing that will allow these services to con-      Clouse, James; Henderson, Dale; Orphanides,
tinue to be offered. Thus, there seems to be little     Athanasios; Small, David H. and Tinsley, P. A.
reason for central banks to avoid bringing the          “Monetary Policy When the Nominal Short-
policy rate close to zero if the broader economic       Term Interest Rates Is Zero.” Topics in Mac-
situation so warrants.                                  roeconomics, 2003, 3(1), article 12.
   Perhaps the more important argument for en-        Eggertsson, Gauti and Woodford, Michael. “The
gaging in alternative monetary policies before          Zero Bound on Interest Rates and Optimal
lowering the overnight rate all the way to zero is      Monetary Policy.” Brookings Papers on Eco-
to ensure that the public does not interpret a          nomic Activity, 2003, (1), pp. 139 –233.
zero reading for the overnight rate as evidence       Friedman, Milton. The optimum quantity of
that the central bank has “run out of ammuni-           money and other essays. Chicago: Aldine,
tion.” That is, low rates risk fostering the mis-       1969.
90                              AEA PAPERS AND PROCEEDINGS                          MAY 2004

Reinhart, Vincent and Sack, Brian. “The Eco-      Economic Studies, February 2001, 19, Spe-
  nomic Consequences of Disappearing              cial Edition, pp. 277–312.
  Government Debt.” Brookings Paper on          Tobin, James. “A General Equilibrium Ap-
  Economic Activity, 2000, (2), pp. 163–209.      proach to Monetary Theory.” Journal of
Romer, Christina. “What Ended the Great De-       Money, Credit, and Banking, February 1969,
  pression?” Journal of Economic History, De-     1(1), pp. 15–29.
  cember 1992, 52(4), pp. 757– 84.              Woodford, Michael. Interest rates and prices:
Svensson, Lars E. O. “The Zero Bound in an        Foundations of a theory of monetary policy.
  Open Economy: A Foolproof Way of Escap-         Princeton, NJ: Princeton University Press,
  ing from a Liquidity Trap.” Monetary and        2003.
You can also read