FROM

Has the Fed Been a Failure?

George Selgin Department of Economics Terry College of Business University of Georgia

William D. Lastrapes Department of Economics Terry College of Business University of Georgia

Lawrence H. White Department of Economics George Mason University

Cato Institute

1000 Massachusetts Avenue, N.W.

Washington, D.C. 20001

The Cato Working Papers are intended to circulate research in progress

for comment and discussion. Available at www.cato.org/workingpapers.

*Corresponding author. We thank David Boaz, Christopher Hanes, Jeff Hummel,

Arnold Kling, Jerry O’Driscoll, Scott Sumner, Dick Timberlake, and Randy Wright

for their helpful suggestions, while absolving them of all responsibility for our

paper’s arguments and conclusions.

ABSTRACT

As the one-hundredth anniversary of the 1913 Federal Reserve Act

approaches, we assess whether the nation‘s experiment with the Federal Reserve

has been a success or a failure. Drawing on a wide range of recent empirical

research, we find the following: (1) The Fed‘s full history (1914 to present) has been

characterized by more rather than fewer symptoms of monetary and macroeconomic

instability than the decades leading to the Fed‘s establishment. (2) While the Fed‘s

performance has undoubtedly improved since World War II, even its postwar

performance has not clearly surpassed that of its undoubtedly flawed predecessor,

the National Banking system, before World War I. (3) Some proposed alternative

arrangements might plausibly do better than the Fed as presently constituted. We

conclude that the need for a systematic exploration of alternatives to the

established monetary system is as pressing today as it was a century ago.

"No major institution in the U.S. has so poor a record of performance over so

long a period, yet so high a public reputation." Milton Friedman (1988).

I. Introduction

In the aftermath of the Panic of 1907 the U.S. Congress appointed a National

Monetary Commission. In 1910 the Commission published a shelf-full of studies

evaluating the problems of the post-bellum National Banking system and exploring

alternative regimes. A few years later Congress passed the Federal Reserve Act.

Today, in the aftermath of the Panic of 2007, and as the one-hundredth

birthday of the Federal Reserve System approaches, it seems appropriate to once

again take stock of our monetary system. Has our experiment with the Federal

Reserve been a success or a failure? Does the Fed‘s track record during its history

merit celebration, or should Congress consider replacing it with something else? Is

it time for a new National Monetary Commission?

The Federal Reserve has, by all accounts, been one of the world‘s more

responsible and successful central banks. But this tells us nothing about its

absolute performance. To what extent has the Fed succeeded or failed in

accomplishing its official mission? Has it ameliorated to a substantial degree those

symptoms of monetary and financial instability that caused it to be established in

the first place? Has it at least outperformed the system that it replaced? Has it

learned to do better over time?

We address these questions by surveying available research bearing upon

them. The broad conclusions we reach based upon that research are that (1) the full

Fed period has been characterized by more rather than fewer symptoms of

monetary and macroeconomic instability than the decades leading to the Fed‘s

establishment; (2) while the Fed‘s performance has undoubtedly improved since

World War II, even its postwar performance has not clearly surpassed that of its

(undoubtedly flawed) predecessor; and (3) alternative arrangements exist that

might do better than the presently constituted Fed has done. These findings do not

prove that any particular alternative to the Fed would in fact have delivered

superior outcomes: to reach such a conclusion would require a counterfactual

exercise too ambitious to fall within the scope of what is intended as a preliminary

survey. The findings do, however, suggest that the need for a systematic

exploration of alternatives to the established monetary system, involving the

necessary counterfactual exercises, is no less pressing today than it was a century

ago.

As far as we know the present study is the first attempt at an overall

assessment of the Fed‘s record informed by academic research. 1

Our conclusions

draw importantly on recent research findings, which have dramatically revised

economists‘indicators of macroeconomic performance, especially for the pre-Federal

Reserve period. We do not, of course, expect the conclusions we draw from this

research to be uncontroversial, much less definitive. On the contrary: we merely

hope to supply prima facie grounds for a more systematic stock-taking.

In evaluating the Federal Reserve System‘s record in monetary policy, we

leave aside its role as a regulator of commercial banks. Adding an evaluation of the

latter would double an already large task. It would confront us with the problem of

distinguishing areas where the Fed has been responsible for rule-making from those

in which it has simply been the rule-enforcing agent of Congress. It would also

raise the thorny problem of disentangling the Fed‘s influence from that of other

regulators, because every bank the Fed regulates also answers to the FDIC and a

chartering agency. Monetary policy, by contrast, is the Fed‘s responsibility alone. 2

Although Martin Feldstein (2010, p. 134) recognizes that ?[t]he recent financial crises, the

widespread losses of personal wealth, and the severe economic downturn have raised questions about

the appropriate powers of the Federal Reserve and its ability to exercise those powers effectively,?

and goes on to ask whether and in what ways the Fed‘s powers ought to be altered, his conclusion

that the Fed ?should remain the primary public institution in the financial sector?(ibid., p. 135)

rests, not on an actual review of the Fed‘s overall record, but on his unsubstantiated belief that,

although the Fed ?has made many mistakes in the near century since its creation in 1913…it has

learned from its past mistakes and contributed to the ongoing strength of the American economy.?

Blinder (2010) argues that, given the premise that the Fed as presently constituted will continue to

be responsible for conducting U.S. monetary policy, it ought also to have its role as a supervisor of

?systematically important?financial institutions preserved and even strengthened. Goodhart and

Schoenmaker (1995) review various arguments for and against divorcing bank regulation from

monetary control.

II. The Fed’s Mission

According to the preamble to the original Federal Reserve Act of 1913, the

Federal Reserve System was created ?to furnish an elastic currency, to afford means

of rediscounting commercial paper, to establish a more effective supervision of

banking in the United States, and for other purposes.?In 1977 the original Act

was amended to reflect the abandonment of the gold standard some years before,

and the corresponding increase in the Fed‘s responsibility for achieving

macroeconomic stability. The amended Act makes it the Fed‘s duty to ?maintain

long-run growth of the monetary and credit aggregates commensurate with the

economy's long run potential to increase production, so as to promote effectively the

goals of maximum employment, stable prices, and moderate long-term interest

rates.?On its website the Board of Governors adds that the Fed also contributes to

?better economic performance by acting to contain financial disruptions and

preventing their spread outside the financial sector.?

These stated objectives suggest criteria by which to assess the Fed‘s

performance, namely, the relative extent of pre- and post-Federal Reserve Act price

level changes, pre- and post-Federal Reserve Act output fluctuations and business

recessions, and pre-and post-Federal Reserve Act financial crises. For reasons

already given, we don‘t attempt to address the Fed‘s success at bank supervision.

III. Inflation

The Fed has failed conspicuously in one respect: far from achieving long-run

price stability, it has allowed the purchasing power of the U.S. dollar, which was

hardly different on the eve of the Fed‘s creation from what it had been at the time of

the dollar‘s establishment as the official U.S. monetary unit, to fall dramatically. A

consumer basket selling for $100 in 1790 cost only slightly more, at $108, than its

(admittedly very rough) equivalent in 1913. But thereafter the price soared,

reaching $2422 in 2008 (Officer and Williamson 2009). As the first panel of Figure

1 shows, most of the decline in the dollar‘s purchasing power has taken place since

1970, when the gold standard no longer placed any limits on the Fed‘s powers of

monetary control.

The highest annual rates of inflation since the Civil War also occurred under

the Fed‘s watch. The high rates of 1973-5 and 1978-80 are the most notorious,

though authorities disagree concerning the extent to which Fed policy was to blame

for them. 3

Yet those inflation rates, in the low =teens, were modest compared to

annual rates recorded between 1917 and 1920, which varied from just below 15% to

18%, with annualized rates for some quarters occasionally approaching 40% (see

Figure 1, third panel). Significantly, both of the major post-Federal Reserve Act

episodes of inflation coincided with relaxations of gold-standard based constraints

on the Fed‘s money creating abilities, consisting of a temporary gold export embargo

from September 1917 through June 1919 and the permanent closing of the Fed‘s

gold window in 1971. 4

Although the costs of price level instability are hard to assess, the reduced

stability of prices under the Fed‘s tenure has certainly not been costless. As the

Board of Governors itself has observed (Board of Governors, 2009),

[s]table prices in the long run are a precondition for maximum sustainable

output growth and employment as well as moderate long-term interest rates.

When prices are stable and believed to remain so, the prices of goods,

services, materials, and labor are undistorted by inflation and serve as

clearer signals and guides to the efficient allocation of resources ….

Moreover, stable prices foster saving and capital formation, because when the

Because these were episodes not merely of inflation but of stagflation, they are frequently said to

have depended crucially on adverse aggregate supply shocks triggered by OPEC oil price increases.

This ?traditional?explanation has, however, been cogently challenged by Robert Barsky and Lutz

Kilian (2001) (see also Ireland 1999 and Chappell and McGregor 2004), who concludes ?that in

substantial part the Great Stagflation of the 1970s could have been avoided, had the Fed not

permitted major monetary expansions in the early 1970.?Blinder and Rudd (2008) have in turn

written in defense of the ?traditional?perspective.

World War II was also a period of substantial inflation, though this fact is somewhat obscured by

standard (BLS) statistics, which do not fully correct for the presence of price controls. Friedman and

Schwartz (1982, p. 106) place the cumulative distortion in the wartime Net National Product deflator

at 9.4%, while Rockoff and Mills (1987, pp. 201-3) place it between that value and 4.8%.

risk of erosion of asset values resulting from inflation—and the need to guard

against such losses—are minimized, households are encouraged to save more

and businesses are encouraged to invest more.

More specifically, as Ben Bernanke (2006, p. 2) observed in a lecture several

years ago, besides reducing the costs of holding money, stable prices

allow people to rely on the dollar as a measure of value when making long-

term contracts, engaging in long-term planning, or borrowing or lending for

long period. As economist Martin Feldstein has frequently pointed out, price

stability also permits tax laws, accounting rules, and the like to be expressed

in dollar terms without being subject to distortions arising from fluctuations

in the value of money.

Feldstein (1997) had in fact reckoned the recurring welfare cost of a steady inflation

rate of just 2%—costs stemming solely from the adverse effect of inflation on the

real net return to saving—at about 1% of GNP. 5

As Bernanke‘s remarks suggest, unpredictable changes in the price level have

greater costs than predictable changes. Benjamin Klein (1975) observed that,

although the standard deviation of the rate of inflation was only a third as large

between 1956 and 1972 as it had been from 1880 to 1915, inflation had also become

much more persistent. The price level had consequently become less rather than

more predictable since the Fed‘s establishment. Robert Barsky (1987) reported in

the same vein that, while quarterly U.S. inflation could be described as a white-

noise process from 1870-1913, it was positively serially correlated from 1919 to 1938

and from 1947 to 1959 (when the Fed was constrained by some form of gold

Lucas (2000), in contrast, put the annual real income gain from reducing inflation from 10% to zero

at slightly below 1 percent of GNP. The difference stems from Lucas‘s having considered inflation‘s

effect on money demand only, while overlooking its influence on effective tax rates, which play an

important part in Feldstein‘s analysis. Leijonhufvud (1981) and Horwitz (2003) discuss costs of

inflation, including those of ?coping?with high inflation environments and those connected to

inflation‘s tendency to distort relative prices. These costs, being very difficult to measure, are

overlooked by both Feldstein and Lucas.

standard), and has since become a random walk. These findings suggest that, as

the Fed has gained greater control over long-run price level movements, those

movements became increasingly difficult to forecast.

Our own estimates from an ARMA (1,1) model yield conclusions similar to

Klein‘s. Although the standard deviation of inflation was greater before the Fed‘s

establishment than it has been since World War II, the postwar inflation process

includes a large (that is, above 0.9) autoregressive component, whereas that

component was small and negative before 1915 (see Table 1). 6

Relatively small

postwar inflation-rate innovations have consequently been associated with

relatively large steady-state changes in the price level (see Figure 2). A GARCH

(1,1) model of the errors from the ARMA model accordingly reveals a stark

difference between the conditional variance of the inflation process before and since

the Fed‘s establishment, with almost no persistence in the variance of inflation

prior the Fed‘s establishment, and a very high degree of persistence afterwards, and

especially since the closing of the Fed‘s gold window (Table 1, second panel). 7

Lastly, by treating six-year rolling standard deviations for quarterly inflation and

price-level series as proxies for the uncertainty associated with each, we confirm

Klein‘s finding that, while the rate of inflation has tended to become more

predictable as inflation has become more persistent, forecasting future price levels

has generally become more difficult, with the degree of difficulty increasing with the

These findings are based on Balke and Gordon‘s (1986) quarterly GNP deflator estimates spliced to

the Department of Commerce deflator series in the fourth quarter of 1946. Hanes (1999) argues that

pre-Fed deflator estimates understate somewhat the serial correlation of pre-Fed inflation, while

overstating the volatility of pre-Fed inflation, owing to their disproportionate reliance upon

(relatively pro-cyclical) prices of ?less-processed?goods.

The coefficient on the ARCH(1) term for the pre-Fed period is not significantly different from zero.

In the event that it is indeed zero, the GARCH(1) coefficient is not identified.

Although Cogley and Sargent (2002) and several other researchers reported a decline in the

persistence of inflation coinciding with the beginning of the Great Moderation, Pivetta and Reis

(2007, p. 1354), using a more flexible, non-linear Bayesian model of inflation dynamics and several

different measures of persistence, find ?no evidence of a change in [inflation] persistence in the

United States?since 1965, save for ?a possible short-lived change during the 1982-1983 period.?

forecast horizon (Figure 3). The conditional variances implied by the GARCH model

are shown in Figure 4. 8

The last panel of Figure 4 makes it especially easy to appreciate why

corporate securities of very long (e.g. 100-year) maturities, which were common in

decades just prior to the passage of the Federal Reserve Act, have become much less

common since. To the extent that its policies discouraged the issuance of longer-

term corporate debt, the Fed can hardly be credited with achieving ?moderate long-

term interest rates.?9

IV. Deflation

While it has failed to prevent inflation, the Fed has also largely succeeded,

since the Great Depression, in eliminating deflation, which was a common

occurrence under the pre-Fed, post Civil War U.S. monetary system. Between 1870

and 1896, for example, U.S. prices fell 37%, or at an average annual rate of 1.2%

(Bordo et al. 2004, and Figure 1, panel 2).

The postwar eradication of deflation would count among the Fed‘s

achievements were deflation always a bad thing. But is it? Many economists

appear to assume so. But a contrasting view, supported by a number of recent

studies, holds that deflation may be either harmful or benign depending on its

underlying cause. Harmful deflation—the sort that goes hand-in-hand with

depression—results from a contraction in overall spending or aggregate demand for

goods in a world of sticky prices. As people try to rebuild their money balances they

Concerning the difficulty of forecasting inflation in recent years especially see Stock and Watson

(2007).

For more recent and international evidence of the negative effect of inflation on firm debt maturity

see Demirg��-Kunt and Maksimovic (1999). As one might expect, the post-1983 ?Great Moderation?

(discussed further below) revitalized some previously moribund markets for very long term corporate

debt. Thus Disney‘s 1993 ?Sleeping Beauty Bonds?became the first 100-year bonds to be issued

since 1954. The more recent decline in U.S. Treasury bond yields has also added to the

attractiveness of very long-term corporate debt. Indeed, on August 24, 2010, Norfolk Southern

managed to sell $250 million worth of century bonds bearing a record low yield of just 5.95 percent,

despite the risks involved. Still many investors remained skeptical. As one portfolio manager

opined (Financial Times August 24, 2010), ?You are giving a company money for a long period of

time with no ability to foresee the conditions in that period of time and for a very low interest rate.?

spend less of their income on goods. Slack demand gives rise to unsold inventories,

discouraging production as it depresses equilibrium prices. Benign deflation, by

contrast, is driven by improvements in aggregate supply—that is, by general

reductions in unit production costs—which allow more goods to be produced from

any given quantity of factor and which are therefore much more likely to be quickly

and fully reflected in corresponding adjustments to actual (and not just equilibrium)

prices. 10

Historically, benign deflation has been the far more common type. Surveying

the 20 th

-century experience of 17 countries, including the United States, Atkeson

and Kehoe (2004, p. 99) find ?many more periods of deflation with reasonable

growth than with depression, and many more periods of depression with inflation

than with deflation.?Indeed, they conclude ?that the only episode in which there is

evidence of a link between deflation and depression is the Great Depression (1929-

1934).?This finding stands in stark contrast with the more common view

exemplified by Ben Bernanke‘s (2002a) assertion, in a speech aimed at justifying

the Fed‘s low post-2001 funds target, that ?Deflation is in almost all cases a side

effect of a collapse in aggregate demand—a drop in spending so severe that

producers must cut prices on an ongoing basis in order to find buyers.?

Atkeson and Kehoe‘s arresting conclusion depends on their having looked at

inflation and output growth statistics averaged across five-year time intervals and

over a sample of 17 countries. There have in fact been other 20 th

-century instances

in which deflation coincided with recession or depression in individual countries

over shorter time intervals. In the U.S. this was certainly the case, for example,

during the intervals 1919-1921, 1937-1938, 1948-1949 (Bordo and Filardo 2005, pp.

814-19), and, most recently, 2008-2009. It remains true, nonetheless, that taking

both 19 th

and 20 th

-century experience into account, it is, as Bordo and Filardo (ibid.,

Selgin (1997) presents informal arguments for permitting benign (productivity-driven) deflation,

while Edge, Laubach, and Williams (2007), Schmidt-Groh� and Uribe 2007, and Entekhabi (2008)

offer formal arguments. For the history of thought regarding benign deflation see Selgin (1996).

p. 834) observe, ?abundantly clear that deflation need not be associated with

recessions, depressions, and other unpleasant conditions.?

Although the classical gold standard made deflation far more common before

the Fed‘s establishment than afterwards, episodes of ?bad?deflation were actually

less common under that regime than they were during the Fed‘s first decades (ibid.,

p. 823). Benign deflation was the rule: downward price level trends, like that of

1873-1896, mainly reflected strong growth in aggregate supply. Occasional

financial panics did, however, give rise to brief episodes of bad deflation. We take

up below the question of whether the Fed has succeeded in mitigating such panics. 11

Taking these findings into account, the Fed‘s record with respect to deflation

does not appear to compensate for its failure to contain inflation. It has, on the one

hand, practically extinguished the benign sort of deflation, replacing it with

persistent inflation that masks the true progress of productivity. On the other

hand, it bears some responsibility for several of the most severe episodes of harmful

deflation in U.S. history.

V. Volatility of Output and Unemployment

If the Fed has not used its powers of monetary control to avoid undesirable

changes in the price level, has it at least succeeded in stabilizing real output? Few

claim that it did so during the interwar period, which was by all accounts the most

turbulent in U.S. economic experience. 12

In fact, according to the standard

(Kuznets-Kendrick) historical GNP series, thanks to that turbulent interval the

cyclical volatility of real output (as measured by the standard deviation of GNP

from its Hodrick-Prescott filter trend) has been somewhat greater throughout the

full Fed sample period than it was during the pre-Fed (1869-1914) period.

The predominance of benign over harmful inflation appears to have been still more marked in the

UK and Germany, owing perhaps to those countries‘less crisis-prone banking systems (Bordo, Lane,

and Redish 2003).

On the volatility of macroeconomic series during the interwar period see especially Miron (1989),

who, comparing the quarter centuries before and after the Fed‘s founding, finds that stock prices,

inflation, and the growth rate of output all became considerably more volatile, while average growth

declined, and concludes that ?the deterioration of the performance of the economy after 1914 can be

attributed directly to the actions of the Fed.?

The same data also support the common claim (e.g. Burns 1960; Bailey 1978;

De Long and Summers 1986; Taylor 1986) that the Fed has made output

considerably more stable since WWII than it was before 1914 (Table 2, row 1 and

Figure 5, first panel). Christina Romer‘s (1986a, 1989, 2009) influential work has,

however, cast doubt even on this more attenuated claim. According to her, the

Kuznets-Kendrick pre-1929 real GNP estimates overstate the volatility of pre-Fed

output relative to that of later periods, in part because they are based on fewer

component series than later estimates and because they conflate nominal and real

values, but mainly because the real component series are almost exclusively for

commodities, the output of which is generally much more volatile than that of other

kinds of output. From 1947 to 1985, for example, commodity output as a whole was

about two and a third times more volatile than real GNP.

According to Romer‘s own pre-1929 GNP series, which relies on statistical

estimates of the relationship between total and commodity output movements

(instead of Kuznets‘na�ve one-to-one assumption), the cyclical volatility of output

prior to the Fed‘s establishment was actually lower than it has been throughout the

full (1915-2009) Fed era (Table 2, row 2 and Figure 5, second panel). More

surprisingly, pre-Fed (1869-1914) volatility (as measured by the standard

deviations of output from its H-P trend) was also lower than post-World War II

volatility, though the difference is slight. 13

Complementary revisions of historical unemployment data by Romer (1986b)

and J.R. Vernon (1994a), displayed here in Figure 6, likewise suggest that the post-

1948 stabilization of unemployment apparent in Lebergott‘s (1964) standard series

is an artifact of the data. Because Vernon‘s revised unemployment series is based

on the Balke-Gordon (1986) real GNP series, which is more volatile than Romer‘s

GNP series, and because his series includes the relatively volatile 1870s, Vernon

By looking at standard deviations of output after applying the Hodrick Prescott filter, rather than

simply looking at the standard deviation of the growth rate of output, we allow for gradual changes

in the sustainable or ?potential?growth rate of real output, and thereby hope to come closer to

isolating fluctuations in output traceable to monetary disturbances. Concerning the general merits

of the Hedrick-Prescott filter relative to other devices for isolating the cyclical component of GNP

and GDP time series see Baxter and King (1999).

finds a somewhat larger difference between 19 th

century and postwar

unemployment volatility than that reported by Romer. Nevertheless he finds that

his estimates ?indicate depressions for the 1870s and 1890s which are appreciably

less severe than the depressions perceived for these periods by economists such as

Schumpeter and Lebergott?(ibid., p. 707).

Romer‘s revisions have themselves been challenged by others, however,

including Zarnowitz (1992, pp. 77-79) and Balke and Gordon (1989). 14

The last-

named authors used direct measures of construction, transportation, and

communication sector output during the pre-Fed era, along with improved

consumer price estimates, to construct their own historic GNP series. According to

this series, the standard deviation of real GNP from its H-P trend for 1869 to 1914

is 4.27%, which differs little from the standard–series value of 5.10%. Balke and

Gordon‘s findings thus appear to vindicate the traditional (pre-Romer) view (Table

2, row 3, and Figure 5, third panel).

More recent work helps to resolve the contradictory findings of Romer on one

hand and Balke and Gordon on the other. Rather than rely on conventional

aggregation procedures to construct historic (pre-1929) real GDP estimates, Ritschl,

Sarferaz and Uebele (2008) employ ?dynamic factor analysis?to uncover a latent

common factor capturing the co-movements in 53 time series that have been

consistently reported since 1867. According to their benchmark model, which

assumes that the coefficients (?factor loadings?) relating individual series to the

latent factor are constant, there was in fact ?no change in postwar volatility relative

to the prewar [that is, pre-World War I] period?(ibid., p. 7). Allowing instead for

Although Zarnowitz (1992, p. 78) agrees that, because they are based on ?cyclically sensitive?

series, the standard (Kuznets-Kendricks) GNP estimates ?exaggerate the fluctuations in the

economy at large,?he claims that, in deriving her own estimates by ?simply imposing recent patterns

on the old data,?Romer ?precludes any possibility of stabilization, thus making her conclusion

inevitable and prejudging the issue in question.?Rhode and Sutch (2006, p. 15) repeat the same

criticism. But Romer‘s method does not rule out the possibility of stabilization any more than that

used in deriving the standard series does: both approaches take for granted a constant ratio of

commodity output volatility to general output volatility. The difference is that, while Romer

estimates the constant, Kuznets implicitly assumed a value of one. That Romer‘s estimate

necessarily reflects postwar structural relationships hardly renders her approach more restrictive

than, much less inferior to, Kuznets‘s.

time-varying factor loadings (and hence for gradual structural change), Ritschl et al.

find that post-WWII volatility was a third greater than pre-Fed volatility (ibid., p.

29, Table 1). These findings reinforce Romer‘s conclusions. 15

But Ritschl et al. are

also able to reproduce Balke and Gordon‘s postwar moderation using a common

factor based on their non-agricultural real time series only, which resemble the

series Balke and Gordon rely upon for their GNP estimates. Here again, the

moderation vanishes if factor loadings are allowed to vary. Balke and Gordon‘s

finding of a substantial reduction in post-WWII output volatility relative to pre-Fed

volatility thus appears to depend on their focus on industrial output and implicit

assumption that the relative importance of different components of that output

hasn‘t changed.

Even if one accepts the Balke-Gordon GNP estimates, it does not follow that

the Fed deserves credit for (belatedly) stabilizing real output. It may be that

aggregate supply shocks, the real effects of which monetary policy is unable to

neutralize, were relatively more important before 1914 than they have been since

World War II. The effects of this reduced role for supply shocks might then be

misinterpreted as evidence of the Fed‘s success in limiting output variations by

stabilizing aggregate demand.

Using the Balke-Gorden output series, John Keating and John Nye (1998)

estimate a bivariate vector autoregression (VAR) model of inflation and output

growth for the U.S, over the periods 1869 to 1913 and 1950 to 1994. They then

identify aggregate demand and supply shocks by assuming, in the manner of

Blanchard and Quah (1989), that supply shocks alone have permanent real effects,

which allows them to decompose the variance of output into separate supply- and

demand-shock components. Doing so they find that aggregate supply shocks were

of overwhelming importance in the earlier period, accounting for 95% of real

output‘s conditional forecast error variance at all horizons (Keating and Nye, Table

3, p. 246). During the post-World War II period, in contrast, the fraction of output‘s

The findings are, as one might expect, robust to the exclusion of nominal time series from the

study.

forecast error variance attributable to supply shocks has been just 5% at a one-year

horizon, rising to only 68% after a full decade (ibid., Table 2, p. 240).

Keating and Nye (1998) themselves, however, question the validity of these

findings because, according to their identification scheme, a positive pre-Fed

?supply?shock causes the price level to increase rather than to decline. But this

seemingly ?perverse?comovement may simply reflect the tendency, under the

international gold standard regime, for supply shocks involving exportable

commodities, such as cotton, to translate into enhanced exports and thus into

increased gold inflows (see Davis, Hanes, and Rhode 2009). A more recent study by

Michael Bordo and Angela Redish (2004) allows for this possibility by extending the

Keating-Nye model to include a measure of the pre-Fed money stock and by

assuming that the price level is uninfluenced in the long run by either aggregate

supply or aggregate demand shocks at the national level—an assumption consistent

with the workings of the international gold standard. According to their estimates,

which again rely upon Balke and Gordon‘s quarterly output data, aggregate supply

shocks accounted for 89% of pre-Fed output variance at a one-year horizon and for

almost 80% of such variance after ten years. These findings differ little from

Keating and Nye‘s for the pre-Fed period.

Bordo and Redish examine the pre-Fed era only, and so do not offer a

consistent comparison of it with the post-World War II era. To arrive at such a

comparison, while shedding further light on the Fed‘s contribution to postwar

stability, we constructed a VAR model allowing for four distinct macroeconomic

shocks—to aggregate supply, the IS schedule, money demand, and the money

supply—which are identified using different and plausible identifying restrictions

for the pre-Fed and post-World War II sample periods. Using this model (and

relying once again on the Balke-Gordon GNP estimates) we find that aggregate

supply shocks account for between 81 and 86 percent of the forecast variance of pre-

Fed output up to a three-year horizon, as opposed to less than 42% of the variance

after World War II (Table 3). 16

In terms familiar from recent discussions of the

causes of the post-1983 ?Great Moderation?in output volatility (discussed below),

our findings suggest that the post-WWII period taken as a whole enjoyed better

?luck?than the pre-Fed period. Our model also shows no clear improvement in the

dynamic response of either output or the money stock to aggregate demand shocks.

On the contrary: it suggests that the output effects of IS shocks have been more

persistent, and those of money demand (velocity) shocks more adverse, since World

War II than they were before 1914, and that the differences are connected to less

appropriate monetary responses (Figure 7).

Fiscal stabilizers, whether ?automatic?or deliberately aimed at combating

downturns, are also likely to have contributed to reduced output volatility since the

Fed‘s establishment, when combined state and federal government expenditures

constituted but a fifth as large a share of GDP as they did just before the recent

burst of stimulus spending (Figure 8). Thus DeLong and Summers (1986) claim

that the decline in U.S. output volatility between World War II and the early 1980s

was due not to improved monetary policy but to the stabilizing influence of

progressive taxation and countercyclical entitlements. Subsequent research (e.g.

Gali 1994; Fatas and Mihov 2001; Andres, Domenech and Fatas 2008; and Mohanty

and Zampolli 2009) documents a pronounced (though not necessarily linear)

relationship between government size and the volatility of real output. According to

Mohanty and Zampoli, a 10% increase in the government‘s share of GDP was

associated with a 21% overall decline in cyclical output volatility for 20 OECD

countries during 1970-1984. 17

For details see Lastrapes and Selgin (2010). Numerous other studies employing a variety of

identification schemes, also find that demand shocks have been of overwhelming importance during

the post-World War II period. See for example, Blanchard and Watson (1986), Blanchard and Quah

(1989), Hartley and Whitt (2003), Ireland (2004), and Cover, Enders, and Hueng (2006). A notable

exception is Gali (1992) who, using a combination of short- and long-run identifying restrictions,

finds that supply shocks were more important. None of these studies examines the pre-Fed period.

While government size is generally negatively correlated with the volatility of output growth, it

also appears to be negatively correlated with output growth itself. Thus Afonso and Furceri (2008)

find, based on estimates for the period 1970-2004, that for the OECD countries a one percentage

point increase of the share of government expenditure to total GDP was associated with a .12

Fiscal stabilizers appear, on the other hand, to have played no significant

part in the post-1984 decline in output volatility (as well as in both the average rate

and the volatility of inflation) known as the ?Great Moderation.?Consequently that

episode seems especially likely to reflect a genuine if belated improvement in the

conduct of monetary policy. We next turn to research concerning that possibility.

VI. The "Great Moderation"

The beginning of Paul Volcker‘s second term as Fed Chairman coincided

with a dramatic decline in the volatility of real output that lasted through the

Greenspan era. Annual real GDP growth, for example, was less than half as

volatile from 1984 to 2007 as it was from 1959 to 1983. The inflation rate, having

been reduced to lower single digits, also became considerably less volatile. Many,

including Blinder (1998), Romer (1999), Sargent (1999) and Bernanke (2004), have

regarded this ?Great Moderation?of inflation and real output as evidence of a

substantial improvement in the Fed‘s conduct of monetary policy—a turn to what

Blinder (1998, p. 49) terms ?enlightened discretion.?18

Bernanke, conceding that

the high inflation in the 1970s and early 1980s was largely due to excessive

monetary expansion aimed at trying to maintain a below-natural rate of

unemployment, argues similarly that Fed authorities learned over the course of

that episode that they could not exploit a stable Phillips curve, while Romer (1999,

p. 43) claims that after the early 1980s the Fed ?had a steadier hand on the

macroeconomic tiller?(Romer 1999, p. 43).

The ?enlightened discretion?view has, however, been challenged by

statistical studies pointing to moderating forces other than improved monetary

policy. 19

A study by Stock and Watson (2002, p. 200; see also idem. 2005) attributes

between 75% and 90% of the Great Moderation in U.S. output volatility to ?good

luck in the form of smaller economic disturbances?rather than improved monetary

percentage point decline in real per capita growth. To this extent at least automatic stabilizers

appear to be a poor substitute for a well-working monetary regime.

See also Clarida, Gali, and Gertler (2000).

Bernanke himself offered his thesis as a plausible conjecture only, without attempting to test it

against alternatives.

policy. Subsequent research likewise tends to downplay the contribution of

improved monetary policy, either by lending support to the ?good luck?hypothesis

or by attributing the Great Moderation to financial innovations, an enhanced

?buffer stock?role for manufacturing inventories, an increase in the importance of

the service sector relative to that of manufacturing, and other kinds of structural

change. 20

As usual, there are exceptions, prominent among which is the study of

Gali and Gambetti (2009), which finds that improved monetary policy, consisting of

an increased emphasis on inflation targeting in setting the federal funds target, did

play an important part in the Great Moderation.

Most authorities do attribute the substantial decline in both the mean rate of

inflation and in inflation volatility since the early 1980s to improved monetary

policy. Yet even here the contribution of enlightened monetary policy may be less

than it appears to be: according to Barro and Gordon‘s (1983) theory of monetary

policy in the presence of a time-inconsistent temptation to improve current-period

real outcomes using surprise inflation, the higher the natural rate of

unemployment, the greater the inflationary bias in the conduct of monetary policy,

other things equal. According to Ireland (1999) and to Chappell and McGregor

(2004), both the actual course of inflation in the 1970s and afterwards and the

arguments on which the FOMC based its decisions conform with the implications of

the theory of time-inconsistent monetary policy. 21

See, among many other works on the topic, McConnell and Perez-Quiros (2000), Ahmed, Levin,

and Wilson (2004), Alcala and Sancho (2004), Irvine and Schuh (2005), Dynan, Elmendorf, and

Sichel (2006), Sims and Zha (2006), Arias, Hansen, and Ohanian (2007), Leduc and Sill (2007), Davis

and Kahn (2008), Moro (2010), Liu, Waggoner, and Zha (2009), and Fern�ndez-Vallaverde, Guerr�n-

Quintana, and Rubio-Ramirez (2010). Besides attributing the Great Moderation to a ?fantastic

concatenation of [positive output] shocks?rather than to improved policy the last of these studies

reaches the more startling conclusions that ?there is not much evidence of a difference in monetary

policy among Burns, Miller, and Greenspan,?and that, had Greenspan been in command in 70s, a

somewhat greater rate of inflation would have been observed (ibid., pp. 4 and 33).

According to King and Morley‘s (2007) recent estimates, the natural rate of unemployment, having

peaked at over 9% in 1983, fell to less than half that level by 2000. Earlier estimates of the natural

rate show a similar pattern, though with smaller amplitude.

The argument summarized here is complemented by that of Orphanides and Williams (2005)

and Primiceri (2006) to the effect that a combination of a heavy emphasis on activist employment

stabilization and mistakenly low estimates of the natural rate of unemployment informed monetary

policy decisions that led to double digit inflation in the 70s and early 80s. In the later 80s, in

In the presence of supply shocks, moreover, the time-inconsistency

framework implies that higher inflation will be accompanied by a more marked

?stabilization bias,?and hence by greater inflation volatility. Richard Dennis

(2003; see also Dennis and S�derstr�m 2006) explains:

to damp the inflationary effect of the adverse supply shock, central bankers

have to raise interest rates more today, generating more unemployment than

they would if they could commit themselves to implement the tight policy

that they promised. In this scenario, the effect of the time-inconsistency is

called stabilization bias because the time-inconsistency affects the central

banker's ability to stabilize inflation expectations and hence stabilize

inflation itself. The stabilization bias adds to inflation's variability, making

inflation more difficult for households, firms, and the central bank, to predict.

As Chappell and McGregor observe (2004, pp. 249-50), to the extent that the

Great Moderation conforms with the predictions of the theory of time inconsistency,

that moderation supplies no grounds for complacency about the Fed:

Policy-makers may have greater appreciation for the importance of

maintaining price stability, but the fundamental institutions by which

monetary policy decisions are made have not changed, nor has the broader

political environment. Shocks similar to those that emerged in the 1970s

could do so again. While Blinder (1997) would comfort us with the argument

that the time inconsistency problem is no longer relevant, a more

troublesome interpretation is possible. The current time-consistent

equilibrium is more pleasant than the one prevailing in the 1970s, not just

contrast, the natural unemployment rate was overestimated or at least no longer underestimated.

See also Surico (2008). Of course these arguments don‘t by themselves rule out the possibility of

negative cyclical movements in inflation independent of changes to the natural rate of

unemployment, such as are likely to accompany a financial crisis like the recent one.

because the Fed is more enlightened, but also because of a fortunate

confluence of exogenous and political forces.

Recent experience has, of course, made it all too evident that prior reports of

the passing of macroeconomic instability were premature. According to Todd

Clarke (2009, p. 5) statistics gathered since the outbreak of the subprime crisis

reveal ?a partial or complete reversal of the Great Moderation in many sections of

the U.S. economy?(ibid., p. 7). Clarke himself, in what amounts to the flip-side of

the Stock-Watson view, characterizes the reversal as a ?period of very bad luck,?

asserting (ibid, p. 25) that ?once the crisis subsides …improved monetary policy

that occurred in years past should ensure that low volatility is the norm?(ibid., p.

27; compare Canarella et al. 2010). Those who believe, in contrast, that ?luck?was

no less important a factor in the moderation as it has been in the recent reversal, or

who (like Taylor 2009a) see the subprime crisis itself as a byproduct of irresponsible

Fed policy, are unlikely to share Clarke‘s optimism.

VII. Frequency and Duration of Recessions

Some of the hazards involved in attempting to compare pre- and post-Federal

Reserve Act measures of real volatility can be avoided by instead looking at the

frequency and duration of business cycles. Doing so, Francis Diebold and Glenn

Rudebusch (1992, pp. 993-4) observe, ?largely requires only a qualitative sense of

the direction of general business activity?while also allowing one to draw on

indicators apart from those used to construct measures of aggregate output.

The conventional (NBER) business cycle chronology suggests that

contractions have been both substantially less frequent and substantially shorter on

average, while expansions have been substantially longer on average, since World

War II than they were prior to the Fed‘s establishment. Because it is based on

aggregate series that avoid the excessive volatility of conventional pre-Fed output

measures (Romer 1994, p. 582 n.28), and because it only classifies contractions of

some minimum duration and amplitude as business cycles, the chronology does in

fact avoid some of the dangers involved in comparing pre-Fed and post-WWII

output volatility.

The NBER‘s chronology has nonetheless been faulted for seriously

exaggerating both the frequency and the duration of pre-Fed cycles and for thereby

exaggerating the Fed‘s contribution to economic stability. According to Christina

Romer (ibid., p. 575), whereas the NBER‘s post-1927 cycle reference dates are

derived using data in levels, those for before 1927 are based on detrended data.

This difference alone, Romer notes, results in a systematic overstatement of both

the frequency and the duration of early contractions compared to modern ones. 22

The NBER‘s pre-1927 indexes of economic activity, upon which its pre-Fed

chronology depends, are also based in part on various nominal time series which

(for reasons considered above) are a further source of bias (ibid., p. 582; also Watson

1994).

Using both the Fed‘s and an adjusted version of her and Jeffrey Miron‘s

indexes of industrial production (Miron and Romer 1990), Romer arrives at a new

set of reference dates that ?radically alter one‘s view of changes in the duration of

contractions and expansions over time?(ibid., p. 601). According to this new

chronology, although contractions were indeed somewhat more frequent before the

Fed‘s establishment than after World War II (though not, it bears noting, more

frequent than in the full Federal Reserve sample period), they were also almost

three months shorter on average, and no more severe. Recoveries were also faster,

with an average time from trough to previous peak of 7.7 months, as compared to

10.6 months. Allowing for the recent, 18-month-long contraction further

strengthens these conclusions. And while the new dates still suggest that

expansions have lasted longer since World War II than before 1914, that difference,

Decades before Romer, George W. Cloos (1963, p. 14) observed, in the course of a considerably

more trenchant evaluation of the NBER‘s business cycle dating methods, ?that the gross national

product and the Federal Reserve Board‘s industrial production index are usable measures of general

business activity and that peaks and troughs in these series are to be preferred to the Bureau‘s

peaks and troughs.?

besides depending mainly on one exceptionally long expansion during the 1960s

(ibid., p. 603), is also much less substantial than is suggested by the NBER‘s dates.

Because the Miron and Romer industrial production series begins in 1884,

Romer does not attempt to revise earlier business cycle dates. That project has,

however, been undertaken more recently by Joseph Davis (2006) who, using his own

annual series for U.S. industrial production for 1796 to 1915 (Davis 2004), finds no

discernible difference at all between the frequency and average duration of

recessions after World War II and their frequency and average duration throughout

the full National Banking era. Besides suggesting that the NBER‘s recessions of

1869-70, 1887-88, 1890-91, and 1899-1900 should be reclassified as growth cycles

(that is, periods of modest growth interrupting more pronounced expansions)

Davis‘s chronology goes further than Romer‘s in revising the record concerning the

length of genuine pre-Fed contractions, in part because it goes further in

distinguishing negative output growth from falling prices. The change is most

glaringly illustrated by the case of the recession of 1873. According to NBER‘s

chronology, that recession lasted from October 1873 to May 1879, making it by far

the longest recession in U.S. history, and therefore an important contributor to the

conclusion that recessions have become shorter since the Fed‘s establishment.

According to Davis‘s chronology, in contrast, the 1873 recession lasted only two

years, or just six months longer than the subprime contraction. 23

In comparing pre- and post-Federal Reserve Act business cycles we have

again tended to set aside the interwar period, as if allowing for a long interval

during which the Fed had yet to discover its sea legs. Nevertheless the Fed‘s

interwar record, and especially its record during the Great Depression, cannot be

overlooked altogether in a study purporting to assess its overall performance. And

that record was, by most modern accounts, abysmal. The truth of Friedman and

Some experts go even further than the NBER in confusing deflation with depression. For example,

FRB Dallas President Richard Fisher refers during a February 2009 CSPAN interview to the ?long

depression?of 1873-1896 (http://www.c-span.org/Watch/watch.aspx?ProgramId=Economy-A-40471).

Concerning the myth of a ?Great Depression?of 1873 to 1896 see Shields (1969) and, for Great

Britain, Saul (1969).

Schwartz‘s (1963, pp. 299ff.) thesis that overly restrictive Fed policies were

responsible for the ?Great Contraction?of the early 1930s is now widely accepted

(e.g. Bernanke 2002b; Christiano, Motto, and Rostagno 2003), as is their claim that

the Fed interfered with recovery by doubling minimum bank reserve requirements

between August 1936 and May 1937. Romer (1992) has shown, furthermore, that

although monetary growth was, despite the Fed‘s interference, the factor most

responsible for such recovery as did take place between 1933 and 1942, that growth

was based, not on any expansionary moves on the part of the Fed, but on gold

inflows from abroad prompted first by the devaluation of the dollar and then by

increasing European political instability. 24

Some economic historians, most notably Barry Eichengreen (1992), have

blamed the Great Depression in the United States on the gold standard rather than

on the Fed‘s misuse of its discretion, claiming that the Fed had to refrain from

further monetary expansion in order to maintain the gold standard. But Elmus

Wicker (1996, pp. 161-2) finds that gold outflows played only a minor role in the

banking panics that were the proximate cause of the monetary collapse of 1930-

1933, while Bordo, Choudri, and Schwartz (2002) show that, even had there been

perfect capital mobility (which was far from being the case), open market purchases

on a scale capable of having prevented that collapse would not have led to gold

outflows large enough to pose a threat to convertibility. Hsieh and Romer (2006),

finally, draw on both statistical and narrative evidence to examine and ultimately

reject the specific hypothesis that the Fed was compelled to refrain from

expansionary policies out of fear that expansion would provoke a speculative attack

on the dollar. Instead, they conclude (ibid., p. 142), ?the American Great Depression

According to Robert Higgs (2009), despite the gold inflows of the =30s and unprecedented wartime

government expenditures the U.S. private economy did not fully recover from the Great Depression

until after World War II.

was largely the result of inept policy, not the inevitable consequence of a flawed

international monetary system.?25

VIII. Banking Panics

If the Fed has not reduced the overall frequency or average duration of

recessions, can it nonetheless be credited with reducing the frequency of banking

panics and hence of the more severe recessions that tend to go along with such

panics? A conventional view holds that the Fed did indeed make panics less

common by eliminating the currency shortages and associated credit crunches that

were notorious features of previous panics; and Jeffrey Miron‘s research (1986)

appears to support this view by showing how, in its early years at least, the Fed did

away with the seasonal tightening of the money market, and consequent spiking

interest rates, that characterized the pre-Fed era.

However, more recent and consistent accounts of the incidence of banking

panics suggest that the Fed did not actually reduce their frequency. Andrew Jalil

(2009) concludes, on the basis of one such new reckoning, ?that contrary to the

conventional wisdom, there is no evidence of a decline in the frequency of panics

during the first fifteen years of the existence of the Federal Reserve?(ibid., p. 3).

That is, there was no reduction between 1914 and 1930, and hence none until the

conclusion of the national bank holiday toward mid March of 1933. Jali‘s findings

agree with Elmus Wicker‘s conclusion, based on his comprehensive analyses of

financial crises between the Civil War and World War II (Wicker 1996, 2000), that

previous assessments had exaggerated the frequency of pre-Fed banking panics by

counting among them episodes in ?money market stringency coupled with a sharp

break in stock prices?or collective action by the New York Clearinghouse but no

widespread bank runs or failures?(ibid. 2000, p. xii). In fact, Wicker states,

In particular, the 1930s Fed has been faulted for having regarded low nominal interest rates and

high bank excess reserves as proof that money was sufficiently easy (Wheelock 1989). Scott Sumner

(2009) argues that the Fed repeated the same mistake in 2008.

there were no more than three major banking panics between 1873 and 1907

[inclusive], and two incipient banking panics in 1884 and 1890. Twelve years

elapsed between the panic of 1861 and the panic of 1873, twenty years

between the panics of 1873 and 1893, and fourteen years between 1893 and

1907: three banking panics in half a century! And in only one of the three,

1893, did the number of bank suspensions match those of the Great

Depression (ibid.)

In contrast, Wicker (1996) elsewhere reports, the first three years of the Great

Depression alone witnessed five major banking panics. No genuine post-1913

reduction in banking panics, or in total bank suspensions, took place until after

1933; and most of the credit for that reduction belongs, not to the Fed, but to the

FDIC and (while it lasted) the FSLIC (Figure 8).

Besides supplying a more accurate account of the frequency of banking panics

before and after the Fed, Jalil‘s chronology of panics allows him to revise the record

concerning the bearing of panics on the severity and duration of recessions.

Whereas DeLong and Summers (1986), employing their own series for the incidence

of panics between 1890 and 1910, conclude that banking panics played only a small

part in the pre-Fed business cycle, Jalil (2009, p. 34) finds that they were a

?significant source of economic instability.?Nearly half of all business cycle

downturns before World War II involved panics, and those that did tended to be

both substantially more severe and longer-lasting than those that didn‘t: between

1866 and 1914, recessions involving major banking panics were on average almost

three times as deep, with recoveries on average taking almost three times as long,

as those without major panics (ibid., p. 35). 26

This evidence suggests that, by

serving to eliminate banking panics, deposit insurance also served, for a time at

least, to reduce the frequency of severe recessions. This fact in turn points to the

The precise figures are: average percentage decline in output, 12.3% for recessions involving

major panics, 4.5% otherwise; average length of recovery, 2.7 years for recessions involving major

panics, 1 year otherwise. The length of recovery is the interval from the trough of the recession to

recovery of the pre-downturn peak.

need for a further, downward reassessment of the Fed‘s post-1933 contribution to

economic stabilization.

Finally, those banking panics and accompanying, severe recessions that did

occur before 1914 were not inescapable consequences of the absence of a central

bank. Instead, according to Wicker (2000, p. xiii) and Eugene White (1993), among

others, banking panics both then and afterwards were fundamentally due to

misguided regulations, including laws prohibiting both statewide and interstate

branch banking. Besides limiting opportunities for diversification, legal barriers to

branch banking, together with the reserve requirement stipulations of National

Banking Act, encouraged interior banks to count balances with city correspondents

as cash reserves. The consequent ?pyramiding?of reserves in New York, combined

with inflexible minimum reserve requirements and the ?inelasticity?of the stock of

national bank notes (which had to be more than fully backed by increasingly

expensive government bonds, and which could not be expanded or retired quickly

even once the necessary bonds had been purchased owing to delays in working

through the Office of the Comptroller of the Currency) all contributed to frequent

episodes of money market stringency, some of which resulted in numerous bank

suspensions, if not in full-blown panics.

Other nations‘experience illuminates the role that misguided regulations,

including those responsible for the highly fragmented structure of the U.S. banking

industry, played in making the U.S. system uniquely vulnerable to panics. Michael

Bordo (1986) reports that, among half a dozen western countries he surveyed (the

others being the U.K., Sweden, Germany, France, and Canada), the U.S. alone

experienced banking crises; and Charles Calomiris (2000, chap. 1), also drawing on

international evidence, attributes the different incidence of panics to differences in

banking industry organization.

Given its proximity to and economic integration with the U.S., Canada‘s

experience is especially revealing. Unlike the U.S., which had almost 2000 (mainly

unit) banks in 1870, and almost 25,000 banks on the eve of the Great Depression,

Canada never had more than several dozen banks, almost all with extensive branch

networks. Between 1830 and 1914 (when Canada‘s entry into WWI led to a run on

gold anticipating suspension of the gold standard), Canada experienced few bank

failures and no bank runs. It also had no bank failures at all during the Great

Depression, and for that reason experienced a much less severe contraction of

money and credit than the U.S. did. Although the latter outcome may have

depended on government forbearance and implicit guarantees which, according to

Kryznowski and Roberts (1993), made it possible for many Canadian banks to stay

open despite being technically insolvent for at least part of the Great Depression

period, 27

the fact remains that Canada was able to avoid banking panics without

resort to either a central bank or explicit insurance. 28

IX. Last-Resort Lending

That the Federal Reserve System was not the only solution to pre-Fed

banking panics, that it may in fact have been inferior to deregulatory reforms aimed

at allowing the U.S. banking and currency system to develop along stronger,

Canadian lines, and that credit for the absence of panics after 1933 mainly belongs

not to the Fed but to deposit insurance, doesn‘t rule out the possibility that the Fed

has occasionally contributed to financial stability by serving as a lender of last

resort (LOLR).

The traditional view of the lender of last resort role derives from Walter

Bagehot (1873). In Bagehot‘s view a LOLR is a second-best remedy for a banking

system weakened by legal restrictions (first-best to Bagehot was a minimally

Kryznowski and Roberts (1993) claim that nine out of Canada‘s ten banks were insolvent on a

market-value basis for most of the 1930s. Wagster (2009), in contrast, concludes based on a different

approach they were insolvent only during 1932 and 1933.

The Bank of Canada was established in 1935, not in response to the prior crisis but, according to

Bordo and Redish (1987), to appease an increasingly powerful inflationist lobby.

Canadian banks‘relative freedom from restrictions on their ability to issue banknotes also

contributed to their capacity to accommodate exceptional demands for currency. In the U.S., in

contrast, national banks were unable to issue notes at all after 1935, and were severely limited in

their ability to do so before the onset of the Great Depression. State bank notes had been subject to

a prohibitive tax since 1866. Concerning the politics behind the decision to suppress state bank

notes, and the economic consequences of that decision, see Selgin (2000).

restricted and hence stronger system like Scotland‘s). 29

The LOLR can help prevent

financial panics, without creating serious moral hazard, by supporting illiquid but

not insolvent banks. Bagehot‘s classical rules for last-resort lending instructed the

Bank of England to extend credit ?freely and vigorously,?but only to borrowers that

passed a solvency test (Bagehot‘s was posting ?good banking securities?as

collateral), and only at a higher-than normal-rate of interest. As Brain Madigan,

Director of the Federal Reserve‘s Division of Monetary Affairs, has noted,

?Bagehot‘s dictum can be viewed as having a sound foundation in microeconomics?:

Specifically, lending only to sound institutions and lending only against good

collateral sharpen firms‘incentives to invest prudently in order to remain

solvent. And lending only at a penalty rate preserves the incentive for

borrowers to obtain market funding when it is available rather than seeking

recourse to the central bank (Madigan 2009, p. 1).

In Bagehot‘s day the solvency requirement was intended to protect the then-

private Bank of England‘s shareholders from losing money on last-resort loans.

Today it serves to protect taxpayers from exposure to public central bank losses.

Judged from a Bagehotian perspective, how well has the Fed performed its

LOLR duties? According to Thomas Humphrey (2010), a former Federal Reserve

economist and an authority on classical LOLR doctrine, it has performed them very

badly indeed, honoring the classical doctrine ?more in the breach than in the

observance?(ibid., p. 22). While Humphrey does identify episodes, including the

October 1987 stock market crash, the approach of Y2K, and (in some respects) the

aftermath of 9/11, in which the Fed seems to have followed Bagehot‘s advice, he

notes that this has not been its usual practice. 30

Why, then, did Bagehot recommend that the Bank of England serve as a LOLR instead of

recommending removal of its monopoly privileges? Because, as he put it at the close of Lombard

Street (1873, p. 329), ?I am quite sure that it is of no manner of use proposing to alter [the Bank of

England‘s constitution]. ... You might as well, or better, try to alter the English monarchy and

substitute a republic.?

Some would add the New York Fed‘s rescue of the Bank of New York following its November 1985

computer glitch. We instead classify this as overnight ?adjustment?lending, reserving the term ?last

During the Great Depression, for example, the Fed departed from Bagehot‘s

doctrine first by failing to lend to many solvent but illiquid banks, and later (in

1936-7) by deliberately reducing solvent banks‘supply of liquid free reserves (ibid.,

p. 23). Since then, it has tended to err in the opposite direction, by extending credit

to insolvent institutions. The Fed made large discount window loans to both

Franklin National and Continental Illinois before their spectacular failures in 1974

and 1984, respectively; and between January 1985 and May 1991 it routinely

offered extended credit to banks that supervisory agencies considered in imminent

danger of failing. Ninety percent of these borrowing banks failed soon afterwards

(United States House of Representatives 1991; Schwartz 1992).

During the subprime crisis, Humphrey observes, the Fed ?deviated from the

classical model in so many ways as to make a mockery of the notion that it is a

LOLR?(Humphrey 2010, p. 1). It did so by knowingly accepting ?toxic?assets, most

notably mortgage-backed securities, as loan collateral, or by purchasing them

outright without subjecting them to ?haircuts?proportionate to the risk involved,

and by supplying funds directly to firms understood to be insolvent (ibid, pp. 24-28;