Stiglitz's important paper(1)

["Z.CUI" <zcui@xxxxxxx>]

Central Banking in an Democratic Society(Joseph Stiglitz)
Introduction
        It is a special pleasure for me to be here to give this lecture to
honor Professor Tinbergen, because his many interests coincide so closely
with my own. He spent much of his later life working on the economics of
income distribution, a subject with which I began my professional life in my
doctoral dissertation, and which has continued to be a focus of my concern.
Tinbergen's thesis that the relative wages of skilled and unskilled workers
depends on both supply and demand factors resonates throughout my work on
optimal taxation.2 Its importance
has been borne out dramatically in wage movements in the United States and
elsewhere during the past two decades.
        From my present vantage point, I am especially appreciative of his
devotion to the economics of development, which became the focus of his
concern in the mid-1950s, in order, as Tinbergen (1988) explained in
retrospect,"to contribute to what seemed to me the highest priority from a
humanitarian standpoint."
        But today, I want to focus on other aspects of his work: his
contribution to economic policy, particularly the problems of controlling
the economy, the relationship between instruments and objectives, and the
scope for
decentralization, which absorbed him and earned him international
recognition in his early days. I have reflected a great deal upon these
issues during the past four years, during which I served as economic adviser
to the
     President of the United States, and especially the last two, when I
served as Chairman of the Council of Economic Advisers.
        I had the good luck to serve at a time of rising prosperity - and
even improved income distribution, lowered  poverty, and increased inclusion
in our society of previously marginalized groups, such as minorities. The
President took much credit for these achievements, and I often quipped that
if some of the glory of what was, at least in some dimensions, the strongest
economy in three decades should rub off on the President, shouldn't at least
some of that rub off on his economic adviser? After all, as my staff
jokingly pointed out, while I was Chairman of the Council the misery index -
the sum of the inflation rate and the unemployment rate - was half of what
it was when Alan Greenspan was Chairman of the Council. Others suggested
that it was not theAdministration which should get the credit, but the
Federal Reserve Board.
        To Tinbergen, this debate might have seemed strange indeed.
Macroeconomic success depended on coordination of the monetary and fiscal
instruments. It was the two working together. Curiously enough,economic
policy making in the United States - and in many other countries - is
designed to inhibit this coordination and cooperation. We have created
independent central banks, who may, and indeed are instructed to, pursue
policies independently of the wishes of the elected officials. In the United
States, the deliberations of
     the open market committee which sets interest rates is kept secret -
even from the President of the United States.
     To be sure, in the past, Presidents have not been shy about expressing
to the Fed what they think it should do,
     but the Fed has not been shy about ignoring these messages. Early on in
the Clinton Administration, we adopted
     a policy of not commenting on Fed policy, not because we did not have
strong views - at certain critical stages,
     many in the Administration thought their policies were seriously
misguided - but because we thought a public
     debate would be counterproductive. We thought the Fed would not listen,
the newspapers would love the
     controversy, and the markets, worried by the uncertainty that such
controversy generates, would add a risk
     premium to long-term rates, thereby increasing those rates, which was
precisely what we did not want to happen.

     There is an irony in all of this. The President is held accountable for
how the economy performs - whether or not
     he has much control. Indeed, econometric models suggest that an
infallible predictor of the outcome of
     presidential elections is the state of the economy;3 just as the
weaknesses of the economy were largely
     responsible for Clinton's election in 1992, the strength of the economy
was largely responsible for his reelection
     in 1996. The Council's own econometric models in 1995 and 1996
corroborated the findings of others predicting
     an electoral outcome close to that which emerged - suggesting that
President Clinton really didn't need to do all
     that campaigning.

     While the President is held accountable, his major tools for affecting
the macroeconomy have been removed.
     Deficit stringency has removed the scope for discretionary fiscal
policy (though fiscal impacts played a role in the
     fine tuning of the 1993 deficit reduction plan)4; and the independence
of the Fed has removed the Executive
     Branch's influence over monetary policy. Members of the Administration
did communicate privately, in weekly,
     sometimes daily, conversations. We shared our views of what was
happening to the economy - but we did not
     always agree. And according to the rules of the game that we adopted,
we did not participate in the public
     debate on monetary policy.

     In a democracy, public discussion and debate about issues of central
importance, like the management of the
     economy, are essential. The Council of Economic Advisers did attempt to
contribute to this discussion - but
     obliquely, especially in the Annual Economic Report of the President.

     Today, I want to address two issues which I felt stifled from
discussing more openly during my tenure at the
     Council of Economic Advisers. The first issue concerns the principles
of monetary policy in a low inflation
     environment such has prevailed in the United States for the past decade
and half - how should it set its targets?
     Should it seek to take pre-emptive strikes against inflation? Is it
true that it cannot, or at least should not, wait to
     act until the "white of the eyes of inflation" can be seen? The second
issue is more fundamental: What should be
     the institutional arrangements by which monetary policy is set in a
democratic society? How independent should
     the central bank be? And if it is independent, what should be its
governance? Who should choose those who
     essentially control the economy, and what characteristics should these
decision makers have? Though I do not
     wish to give away my bottom lines, to pique your interest, let me hint
at the conclusions I shall draw:

        1.Monetary policy matters, and the successful conduct of
macro-policy in the post-war period has led to far
          greater stability of the economy. This is not to imply that
American economic policy has been perfect -
          major mistakes, some arising from an imperfect understanding of
the economy, have at times contributed
          to unnecessarily high unemployment or to the economy enjoying a
stronger boom than intended.
        2.In particular, I will argue that the strategies of opportunistic
disinflation or pre-emptive strikes are based on
          a set of hypotheses about the economy for which there is little
empirical support. I will argue, at least in the
          context of the American economy today, for an alternative, which I
call cautious expansionism.
        3.There is a rationale for a degree of independence of the central
bank, even in a democratic society. But
          the central bank must be accountable, and sensitive, to democratic
processes; there must be more
          democracy in the manner in which the decision makers are chosen
and more representativeness in the
          governance structure. The movement in the opposite direction in
some places is particularly disturbing.

     Monetary Policy Matters: The Stabilization of the Postwar Business Cycle

     Before answering the two questions which are the focus of my concern
today, I have to address a prior issue:
     Does monetary policy matter? For clearly, if monetary policy has no
effect, then the design of monetary
     institutions, the choice of monetary policy strategy, and the
coordination of monetary and fiscal policy do not
     matter. I believe strongly that monetary policy does matter - and it
was not just frustration with our inability to use
     discretionary fiscal policy combined with envy of the economic power of
those sitting along Constitution Avenue
     in the Federal Reserve Board building that led me to this conclusion.
This conclusion was based on theoretical
     work that I had done before entering government5 and recent empirical
work by Francis Diebold and Glenn
     Rudebusch (1992), which we have confirmed and extended. Their findings
have not received the attention they
     deserve; they shed light on a long standing controversy about whether
there are in fact business cycles or simply
     random economic fluctuations. Their somewhat surprising conclusion is
that there appear to have been cycles
     prior to the Great Depression, but that in the postwar period, these
cycles - in the sense of regular periodic
     movements in output - have been eliminated. Before turning to the
statistical results, let me comment briefly on
     the circumstances that led up to my work in this area.

     Though we did not control monetary policy it was important for us to
have views on where the economy was
     going and what we thought monetary policy should be. My friend Jacob
Frenkel (governor of the Central Bank
     of Israel) once quipped that central bankers have a fascination with
fiscal policy - they are always willing to
     comment on the appropriate size of the deficit (zero), though they
thought it inappropriate for the fiscal authorities
     to comment on monetary policy. By the same token, we had a fascination
with monetary policy - and wished we
     could comment on it.

     It is remarkable how little insight into these issues is shed by
current macro-economics. One major school of
     thought, Real Business Cycles, argues that there is no involuntary
unemployment. It was hard to tell that story to
     the President, who was elected on a platform of "Jobs, Jobs, Jobs!", or
to the voters in California, when
     unemployment - they didn't think it was just a superabundance of
leisure - exceeded 10 percent. Another major
     school, new classical economics, with its emphasis on rational
expectations, argues that monetary policy is
     ineffective, because the private sector would adjust its expectations
and actions to undo any systematic
     monetary policies.6 If correct, concerns about policy coordination are
not of much importance! And if correct,
     the myriad of economists, in government and business, and the multitude
of reporters, who were engaged in
     trying to figure out what the Fed was about to do, are all behaving
irrationally. While these schools of thought
     might have little sway in the real world of government or business,
they have had remarkable influence in
     academia over the past quarter of century, especially in America. Both
of these schools suggested that our
     difficulties in the Council of hiring macro-economists from academia
who knew something about the economy
     was of no consequence: they would be wasting their time in any case.
Needless to say, these were perspectives
     with which I had little sympathy.

     As the economy continued the robust recovery from the 1990-91
recession, I was asked by reporters with
     increasingly frequency, did I expect the recovery to end. Their view
was that the economy was perched on a
     knife-edge, ready to fall off into a recession on one side or rising
inflation on the other. Furthermore, they seemed
     to believe that the longer an expansion lasted, the more likely there
was to be a downturn. In contrast, I believed
     in Keynes' animal spirits, and believed that those animal spirits might
be driven, if ever so gently, towards a more
     favorable view of the economy, and hence stronger investment. The
fundamentals of the U.S. economy were
     clearly sound, but I wanted to make a further argument: that expansions
do not end of old age, a popular way of
     saying that there was no such thing as a business cycle. The results in
Figure 1 provide dramatic support for this
     argument as applied to the post-World War II U.S. economy - the
probability of an expansion ending appears to
     be independent of its length.7 This result should not come as surprise,
if one makes three assumptions: monetary
     policy seeks to maintain expansions, monetary policy is forward
looking, and monetary policy is somewhat
     effective. For if there were any systematic time dependency - or
dependency on time and other observable
     variables - the monetary authorities should seek to take offsetting
actions. The result does not require that the
     monetary authorities be perfectly efficient, only that any errors have
no systematic component to them.

     From this perspective, downturns come as a surprise, an unexpected
event not anticipated, or imperfectly
     anticipated, or whose consequences were not fully calculated, perhaps
because of misunderstandings about the
     structure of the economy. Monetary authorities seek to offset these
effects, to restore the economy to its
     potential. In the short run, there is a tendency of monetary
authorities to think of the downturn as a temporary
     deviation, which will correct itself shortly. Given the lags in the
effectiveness of monetary policy, expansionary
     policies might then complement the natural forces of recovery, leading
to inflation. Over time, if the downturn
     persists, political pressure - even on an independent monetary
authority - to do something mounts; the policy of
     doing nothing, or doing too little becomes hard to maintain. Moreover,
information about the true nature of the
     downturn becomes more apparent.

     This pattern is clearly evidenced in the series of pronouncements of
the Fed Chairman between 1991 and 1993.
     Even as the National Bureau of Economic Research was about to declare
that the economy was in recession in
     July 1991, the Fed Chairman's Humphrey-Hawkins testimony (which he is
required to give before Congress
     twice a year) did not indicate that the Fed was worried about
recession.8 To be fair, economic forecasters have
     almost always missed recessions. (Also, I should add parenthetically
that one of the responsibilities of Fed
     officials is to maintain confidence in the economy. Private views may
be more pessimistic than public
     pronouncements. Still, in this particular case, policy seemed to
conform remarkably closely to the public
     pronouncements. Moreover, the Fed Chairman is a master of Fedspeak -
some say a modern version of a
     Delphic oracle - which is designed to carefully calibrate what
information is revealed and what is obscured rather
     than to provide complete enlightenment. This provides plenty of
opportunity for him to make announcements that
     bolster confidence in the economy while being sufficiently vague so
that in retrospect they seem to provide keen
     insights into the workings of the economy regardless of what happens.)

     As the downturn persisted the Fed continued to see it as an unexpected
shock leading to a "normal" cyclical
     downturn that would respond to standard policies. This viewpoint is
evident in the Humphrey-Hawkins testimony
     from February 1991 which reads "[n]onetheless, the balance of forces
does appear to suggest that this downturn
     could well prove shorter and shallower than most prior post-war
recessions. An important reason for this
     assessment is that one of the most negative economic impacts of the
Gulf war - the run-up in oil prices - has been
     reversed. Another is that the substantial decline in interest rates
over the past year and a half - especially over the
     past several months - should ameliorate the contractionary effects of
the crisis in the Gulf and of tighter credit
     availability."

     It was not until the economy was on its way to recovery, in February
1993 that the Fed finally recognized the
     "economy has been held back by a variety of structural factors,"
[emphasis added] most notably fundamental
     weaknesses in the financial system.9

     As the nature of the problem became clearer, and as the political
pressure to do something mounted, monetary
     policy was eased 24 times, contributing to the recovery. The pattern
evidenced in our most recent recession is
     typical, as confirmed by the statistical data: Figure 2 shows that
there is a strong time dependency in recovery.

     These patterns are markedly different from those that prevailed before
the Great Depression. Since World War
     II, expansions are longer and recessions are shorter, as Table 1 shows.
Figure 3 shows, using data for the United
     States for the period 1854 to 1918 that prior to the Great Depression,
expansions did end of old age: the
     probability of an expansion ending increased markedly the longer the
expansion continued, with a probability of
     approximately one-third in the second year, increasing to two-thirds in
the fourth. By contrast, recovery form a
     downturn seems to have been largely a random event, as Figure 4 shows.
While some of these changes could
     have been accounted for by changes in the structure of the economy, I
suspect that it is improved macro-policy
     (including automatic fiscal stabilizers) that accounts for much of the
change.

     Incidentally, these results strongly rebut the claim of Christina Romer
(1986) that there is no evidence of
     improved macro-economic performance in the postwar period. Her argument
relies on adjustments in output
     series which are debatable. Our methodology only requires qualitative
assessments about whether the economy
     is expanding or contracting. Because it does not require measures for
every subcomponent of GDP and because
     it can utilize data from other sources, the timing of expansions and
downturns provides a far more robust way of
     assessing economic performance.

     These results, while they show convincingly that monetary policy
matters and has been used to improve the
     overall performance of the economy, do not require us to believe that
the monetary authority behaves perfectly
     or even that it is efficient. I already discussed one example of a
mistake: the Fed doing too little and acting too
     late to avert or minimize the depth and duration of the 1990-91
recession. A second illustration is the current
     expansion which can be thought of as also partially attributable to
mistakes, at least initially. There is a tendency
     to think of mistakes as one sided - always working to the detriment of
the economy. But mistakes, by their
     nature, should be random, and in at least some cases should work to the
benefit of the economy. In this case,
     there were in fact two errors on the part of the Fed, with one more
than offsetting the other.

     Throughout the earlier 1990s, the Fed continued to have an overly
pessimistic view concerning the NAIRU
     (non-accelerating inflation rate of unemployment), and the economy's
potential for reducing unemployment
     without inflation increasing. But they also continued to
underappreciate the role of financial markets and
     continued to fail to understand key aspects of banking behavior. Had
they better understood these factors, given
     their beliefs about the NAIRU and given their strong aversion to
inflation, they would have prevented the
     unemployment rate from declining below 6.0 percent to 6.2 percent. It
might have been a long time - possibly
     never - before we learned about the economy's real potential.10 It was
our good fortune that they did not see
     accurately where the economy was going!

     To understand what happened - and why the Fed failed (fortunately) to
see the strength of the recovery - we
     need to return to the early days of the Clinton Administration. When
the President took office in February 1993,
     he moved quickly to introduce a deficit-cutting budget. Eventually the
Congress enacted a plan to reduce the
     deficit by $500 billion over five years (in contrast, the 1997 balanced
budget legislation only cut the deficit by
     $200 billion over five years). Old-style Keynesians warned that deficit
reduction would undermine the fragile
     recovery. Those of us who believe that the markets were forward
looking, understood that credible,
     pre-announced deficit reduction would lower interest rates and thus
stimulate the economy. What took us all by
     surprise was just how much it was stimulated.

     Parsing out how credit should be divided up became a preoccupation. To
be sure, the strong economy played a
     role, but this just leads to the further question of how much of the
strong economy was attributable to deficit
     reduction? We were more willing to take credit for higher growth than
for the increasing profit share (which led
     to higher tax revenues), a result partly of the low interest rates, but
also partly a reflection of wage performance.
     But by any reckoning, the tax increases and expenditure reductions in
the Omnibus Budget Reconciliation Act of
     1993 (OBRA 1993) were directly responsible for more than half of the
deficit reduction that followed.

     But in spite of all of the rhetoric, the connection between deficit
reduction and economic recovery remained
     somewhat of a puzzle: Shouldn't the Fed be able to manage monetary
policy to maintain the economy at full
     employment, that is, at the NAIRU? Nothing in the modern theory of
monetary policy suggested that the Fed's
     ability to do that should be affected in any way by deficits or deficit
reduction, so long as the changes were
     appropriately anticipated and offsetting actions undertaken. If the
government ran a slightly larger deficit, then the
     Fed would have to run a slightly tighter monetary policy; the short-run
macroeconomic performance would be
     the same, but the composition of output would shift from private
investment to government spending, potentially
     impairing long-run growth.

     As I thought about it more, I finally recognized the connection, but it
was more subtle and based on the link
     between financial markets and economic activity. It is an interesting
story, and illustrates that while two wrongs
     do not make a right, in economic policy two mistakes can more than
offset each other, and result in an economic
     boom.

     In the 1980s, banks had significant holdings of long-term Treasury
bonds. This represented a gamble on falling
     interest rates. Banks were allowed to take this gamble because
accountants valued these bonds at face value and
     regulators judged risk by the chance of default - which was zero in
this case - not by the likely volatility of asset
     prices (interest rate risk).11 When interest rates rose in the late
1980s, the value of bank assets fell, which
     together with substantial losses on loans, forced many banks to curtail
their lending. Subsequently, with the 1993
     deficit reduction and the lowered inflationary expectations, interest
rates declined. The result was a major
     revaluation of bank assets at the same time that loans were again
becoming more profitable. Given their increased
     net worth and cash flow, banks were both willing and able to increase
their lending. And this is precisely what
     they did. Had this effect been anticipated by the Fed, it is unlikely
that they would have allowed the Federal funds
     rate to stay so low for long.

     Three Propositions About Monetary Policy

     The effective conduct of monetary policy is extraordinarily difficult.
It requires assessing the state of the economy
     today and in the near future. It requires a detailed knowledge of the
economy, so that the consequences of
     various actions - interest rate increases and decreases - can be
carefully assessed. Monetary policy is necessarily
     conducted in an environment with considerable uncertainty, and
therefore requires careful balancing of risks,
     including the risks of inflation, the chances that it might increase or
accelerate, and the costs of disinflation.

     One leading monetary policy strategy responds to this uncertainty by
recommending that policymakers act to
     eliminate projected rises in inflation. The argument for aggressive,
pre-emptive strikes against inflation is based on
     three premises. The most fundamental premise is that inflation is
costly. This provides the motivation for trying to
     avert or lower inflation. The second premise is that once inflation
starts to rise it has a tendency to accelerate out
     of control. This belief provides a strong motivation for erring on the
side of caution in fighting inflation. Finally, the
     third premise is that increases in inflation are very costly to
reverse. The implication of this premise is that even if
     you care much more about unemployment than inflation, you would still
keep inflation from increasing today in
     order to avoid having to induce large recessions to bring the inflation
rate down later on. All three of these
     premises are hypotheses that can be tested empirically.

     In many countries throughout the world, monetary policy seems to be
based on a belief by policymakers in these
     three premises, even when these beliefs are not fully incorporated in
the formal models that the staffs of the
     central banks employ. I would like to discuss the evidence underlying
each one.

     The Costs of Inflation

     Many people treat inflation as if it were something that was costly in
its own right. This, of course, is not true.
     Individual utility functions depend only on quantities; prices do not
enter because, by themselves, they do not
     make people better or worse off. The same is true of the social welfare
functions that politicians should use in
     guiding their thinking. Putting unemployment or output in the social
welfare function might be reasonable, although
     even here we must worry about a number of finer points, including the
valuation of leisure. Putting inflation into
     the objective function is, however, never justified. Instead, inflation
only matters in so far as it effects the two
     variables we do care about: output and its distribution. When
economists or commentators speak about
     balancing the costs of inflation against the costs of unemployment they
are implicitly mixing an objective function
     (which weights output and distribution) with a model of the economy
(which links inflation with these variables)
     and combining them into a reduced form. This shorthand is acceptable,
as long as we remember it is just that - a
     shorthand. All too often, however, this shorthand turns into a rigid
assumption.

     What then is the evidence concerning the costs of inflation? There is
an old theoretical literature that emphasizes
     menu costs, shoe-leather costs, tax distortions, and the increasing
noise introduced into the price system.
     Estimates of the deadweight loss imposed by these distortions in
countries like the United States have, for the
     most part, been disappointingly small from the perspective of
inflationary hawks.12 In the last decade, an
     increasingly sophisticated literature has attempted to measure the
costs of both the level and variability of inflation
     indirectly by examining their consequences for the level of output and
growth. Probably the most persuasive
     studies were done by Bruno and Easterly (1986) who found that when
countries cross the threshold of 40
     percent per year inflation they fall into a high inflation/low growth
trap. But below that level, their is no evidence
     that inflation is costly.13

     Others, like Barro (1997) and Fischer (1993), have used cross country
growth regressions in an attempt to
     quantify the impact of inflation on growth. They have confirmed that
high inflation is, on average, deleterious for
     growth, but again have failed to find any evidence for costs of low
levels of inflation. Fischer also found the same
     results for the variability of inflation.14 (The strength of the
nonlinearity in the relationship between inflation and
     social welfare is clear from the outcome of research conducted by the
United States Federal Reserve Bank.
     Despite the efforts of their minions of first rate economists - some of
them devoting much of their time to
     analyzing the costs of inflation - the Fed has still failed to find
definitive evidence of costs of inflation in the United
     States. Should they eventually succeed in finding such results, they
will only have proven that data mining does
     work, not that inflation does not.)

     Recent research by Akerlof, Dickens, and Perry (1996) has argued that
low inflation is actually beneficial. Some
     inflation, according to this view, helps maintain full employment by
facilitating the downward adjustment of real
     wages. Their simulation suggests that maintaining zero inflation would
be consistent with a 10 percent long-run
     unemployment rate. This is probably too high - surely people would
eventually become less resistant to nominal
     wage cuts after some experience with zero inflation. Still, their
research forcefully reminds us that we need to
     weigh the costs of low inflation against its benefits.

     The acceleration of inflation

     The second premise of many inflation hawks is that inflation is like a
genie, once you let it out of the bottle it will
     just keep on expanding. Stepping off precipices, sliding down slippery
slopes, and falling off the wagon are other
     metaphors that often dominate popular thinking about inflation. Again,
these metaphors can be subject to
     rigorous testing. Relatively few people have done this, in part because
most economic models assume that
     inflation does not accelerate, a position that is at variance with the
conventional journalistic wisdom. My own
     tests have provided no basis for believing the conventional wisdom.

     One test is to nest non-accelerationism as a special case in an
accelerationist model. You can, for instance,
     estimate empirical Phillips curves in which the change in inflation
depends not just on unemployment but also on
     the past change in inflation. In the United States, this coefficient is
insignificant.15 Alternatively, techniques like
     logit can be used to infer whether or not the likelihood of inflation
increasing depends on, among other variables,
     the level or the rate of change of inflation. These tests also find no
evidence whatsoever that changes in inflation
     are more persistent than can be accounted for by the persistence of the
unemployment rate. In fact, after
     controlling for the unemployment rate, we find just the opposite: when
inflation has been rising it is more likely to
     reverse course and start falling - the exact opposite of the
conventional wisdom (Figure 5). The difficulty of
     finding evidence - or casual experience - which supports the precipice
hypothesis probably explains why it is
     rarely a feature of standard economic models. However, this hypothesis
is still all too present in discussions about
     the proper stance of monetary policy.

     Inflation is costly to reverse

     The third premise is that inflation is costly to reverse. The standard
mantra is that even if we cared very little
     about inflation it might be worthwhile to endure a little extra
unemployment today in order to avoid increasing
     inflation leading to a recession down the line. This premise is based
on fallacious backward induction. It asks
     what should the Fed's reaction function be today given a fixed reaction
function in the future. Even framed in his
     way, this argument is based on two somewhat shaky premises.

     The first premise is that you cannot slow the economy down without
creating a recession, and worse, you cannot
     create a small recession. One rationale for this belief is that slowing
growth triggers inventory buildups that in turn
     further dampen production. Whatever the merits of this belief, as a
matter of logic there is no reason why it
     should apply any more strongly to a tightening in the future than to a
tightening today. If you really believe that fine
     tuning is impossible and even mild restraints risk triggering a major
recession, then there is little argument for
     pre-emptive tightening.

     To be sure, several economic downturns have been due to excessive zeal
by the monetary authorities in response
     to inflation. But one needs to interpret this evidence carefully. The
Fed has managed on many occasions "to trim
     the economy's sails" without inducing a recession. There will be
mistakes in monetary policy. Sometimes
     policymakers will act excessively, and this will cause a downturn. But
this does not mean that tightening
     necessarily must aim to overshoot, and must aim to overshoot by a lot.
The fact that occasionally the Fed steps
     on the brakes too hard does not mean that it always steps on the brakes
too hard; it only means that those are
     the cases which we are most prone to notice.16

     More generally, the theory of economic adjustment suggests that when
firms and households are given time to
     adjust, they can do so at far lower costs. If they are put on notice
that there will be a gradual slowing of
     economic growth, then the magnitude of the economic fluctuations
induced by excess inventory build-ups can be
     reduced.

     The second premise is that low unemployment increases inflation much
more than high unemployment lowers it,
     and that the costs of lowering inflation increase more than
proportionately with the magnitude of the inflation
     reduction - in other words, that the Phillips curve is convex. Prices
and wages, according to this view, are
     downwardly rigid. As a result, decreases in aggregate demand translate
more into falling output and employment
     than into lower inflation. Increases in aggregate demand, however, have
just the opposite effect, raising inflation
     with relatively little gain for output.

     The logic behind this belief is not terribly compelling. We often see
falling nominal prices and nominal wages. And
     furthermore, in a world with positive inflation the issue is not
whether there is downward price and wage rigidity
     but whether there is resistance to downward adjustments in the rate of
increase of prices or wages. Casual
     observation provides very little reason to believe this is the case.

     Other theories suggest the possibility that the Phillips curve might be
concave. One intuition comes from strategic
     price setting in oligopolistic competition. Suppose two competitors,
say Coke and Pepsi, both faced kinked
     demand curves. If Pepsi raised its prices, Coke might stand pat in
order to gain market share. But if Pepsi
     lowered its prices, Coke would have to match them in order to avoid
losing market share. The consequence is
     that prices adjust downward together more easily than they adjust
upward. This reverses the logic I just
     described: expansions in aggregate demand will go more into output than
inflation, and vice versa for
     contractions.

     I see no reason for an a priori belief in one shape of the Phillips
curve over another when different theories are
     consistent with such different shapes. Instead, the question is
empirical. Unfortunately, it is much easier to assess
     the sign or even the average magnitude of the relationships between
variables than it is to assess the shape of the
     function relating them. One study conducted by the Council of Economic
Advisers found that if you allow a kink
     in the Phillips curve, the data "chooses" a concave shape.17 When I
went to the World Bank I thought it would
     be interesting to run the exact same regression for other countries.
Amazingly, in all but one country we found
     that the Phillips curve was statistically significantly concave.18

     These statistical findings corroborate the experiences of case studies.
Some of the most dramatic episodes of
     inflation reduction - most recently, the experiences of Brazil and
Israel - show that very large reductions in
     inflation and inflationary expectations can be achieved at very low costs.

     I would caution against drawing overly strong conclusions from the
evidence I just discussed. The power of the
     tests I described is very uncertain. The premises that inflation is not
costly or that inflation does not accelerate, for
     instance, are typically treated as null hypotheses. Failing to reject
them may have as much to do with the difficulty
     of drawing sharp inferences from the data as it does with their
substantive merits. And, the concavity result, like
     all nonlinear regressions, may be sensitive to the choice of functional
form. The evidence does, however,
     decisively refute the extreme versions of all three premises. And in
doing so, the arguments underlying the policies
     of aggressive, pre-emptive strikes against inflation, a stance that is
the basis of the rhetoric if not the practice of
     so many central banks, are undermined. These policies are based on
articles of faith, not on scientific evidence.

     If monetary policy were conducted under perfect certainty both about
the state of the economy and the
     consequences of policy, the three propositions I just discussed would
not matter very much. But because
     monetary policy is a process of sequential decision making under
uncertainty, all of these questions are very
     relevant. Consider the problem facing U.S. monetary authorities
throughout the 1990s. There is uncertainty about
     the NAIRU. There is a chance that not engaging in a pre-emptive strike
might lead to an unemployment rate
     below the NAIRU. But if it is true that keeping the unemployment rate
below the NAIRU only raises inflation by
     a small amount, if the costs of this added inflation itself are
negligible, and if this mistake is easy to correct by
     raising the unemployment rate slightly above the NAIRU, then this is a
risk worth taking: the economy enjoys a
     lower unemployment rate and higher output, marginalized groups are
brought into the labor force, and through
     the reverse hysterisis effect, the level of the NAIRU itself may be
lowered. We should not follow a policy of
     pre-emptive strikes, but rather one of cautious expansionism.
Zhiyuan Cui
617-253-2951(p)
617-258-6164(f)
http://web.mit.edu/polisci/www/faculty/Z.Cui.html