December 19, 2017
Festivus 2017 Airing
of Grievances: I Gotta a Lot of Problems with You, Taylor Rule
December 23rd is almost upon us. You know what
that means. It’s time for me to work up my annual airing of grievances for
Festivus 2017. Although I have myriad political-economic grievances for 2017, I
am going to concentrate on only one in this annual Festivus epistle – the
Taylor Rule. For decades, there has been a debate as to whether central bank
monetary policies should be guided by some clearly-defined rule or should
central banks be free to operate with discretion, i.e., by the seat of their
collective pants. I come down on the side of a rule vs. discretion. But not on the side of the rule most often
mentioned, the Taylor Rule.
In 1993, John Taylor, a Stanford University economics
professor, published a research paper
in which he purported to describe how
the Federal Reserve had conducted monetary policy in terms of its movement of the
federal funds interest rate from 1987 through 1992. Essentially what Taylor did
was estimate a Fed reaction function
to consumer goods/service price inflation above or below a perceived Fed inflation target and to real GDP growth above or
below a perceived Fed real GDP
target. In his 1993 research paper, Taylor suggested that his description of past Fed monetary policy
decisions in terms of movement of the federal funds rate could be useful as a
guideline as to how the Fed should
operate in the future. Although Taylor did not
suggest in his 1993 research paper that the Fed should adhere rigidly to his
estimated reaction function, subsequently he has implicitly criticized Fed
policy for not hueing to his rule (see “A
Monetary Policy for the Future”, April 16, 2015)
In its basic form, the Taylor reaction function is:
i = r*+ p + 0.5 (p-p*) + 0.5 (y as a % of y*)
where:
i = nominal fed funds rate
r = real equilibrium federal funds rate (usually 2%)
p = actual inflation rate (yr/yr % chg)
p*= target inflation rate (usually 2%)
y = actual real output
y* = potential real output
So, this formula states that for every 1% rise in
inflation above its target, the Fed would raise the federal funds rate by 0.5% and
similarly for every 1% increase in the output gap (actual real GDP as a percent
of potential real GDP, the Fed would raise the federal funds rate by 0.5%. If
actual inflation rate were equal to the Fed’s target inflation rate and if
actual real GDP were equal to potential real GDP (i.e., the output gap were
zero) then the nominal, or observed,
federal funds rate would equal the unobservable real equilibrium federal funds
rate, assumed by Taylor to be 2%, plus the inflation rate.
My grievances with the Taylor Rule arise out of the old
saw, it’s not what you don’t know that will hurt you, but what you think you know but don’t. There are two elements in the Taylor Rule that are assumed
to be known but are not, in fact
known. The first of these elements is “r”, the equilibrium real federal funds
rate. Just as the price of wheat varies
over time because of changes in the supply and demand for wheat, so does the
real equilibrium federal funds rate vary
over time. Changes in fiscal policy, changes in business “animal spirits”,
changes in the age distribution of the population, to name just a few factors,
can change the equilibrium level of the real federal funds rate. If the actual equilibrium real federal funds
rate has risen to 3% from the assumed
level of 2% in the Taylor Rule, then the Fed would persistently be keeping the
nominal federal funds rate too low for an extended period of time, which would
result in persistently accelerating inflation rate. This was why Milton
Friedman, may he rest in peace, argued against the Fed using an interest rate
as its policy instrument. No one knows what is the equilibrium level of the
nominal interest rate, much less the real interest rate, which would also require
knowledge of what inflation expectations are. In some instances, a 3% federal funds rate
might represent an accommodative monetary policy. In other instances it might
represent a restrictive monetary policy. If the Fed persistently keeps the
level of the federal funds rate low compared to its equilibrium level, an
inflationary spiral will result. If the Fed persistently keeps the federal funds
rate high compared to its equilibrium, a deflationary spiral will result.
The second unknown element in the Taylor Rule is the
level of potential real GDP.
Potential real GDP is a function of the size of the potential labor force, the productivity of that potential labor as
well as the productivity of other production factors and the rate of
technological advances. The easiest of these variables to estimate with a high
degree of accuracy is the potential labor force. Demography actually is a science.
Alright, even if we did know with a high degree of
certainty what was the equilibrium level of the real federal funds rate and the
level of potential real GDP, how do we know that 0.5 is the correct value of
the reaction coefficients to the inflation and output “gaps”? Why should the
coefficients be of the same value for each gap? Is the lag between a change in
the nominal federal funds rate and the inflation rate the same as it is for
real GDP?
Lastly, even if we did know with a high degree of
certainty what was the equilibrium level of the real federal funds rate, the
level of potential real GDP and the correct reaction coefficients of the
“gaps”, there are time lags with respect to the availability of inflation and
real GDP data. Currently, the Commerce Department releases the Personal
Consumption Expenditures (PCE) chain price index monthly with about a one-month
lag. For example, the November 2017 PCE price index will be released on
December 22, 2017. And of course, it is subject to revisions. Perhaps the Fed could use the consumer price
index that is updated daily from MIT’s The Billion Prices Project.
The Commerce Department’s first estimate of Q3:2017 GDP was released on October
27, 2017, the second estimate on November 29, and its third estimate on
December 21. Then in 2018, 2017 GDP data will be revised still more. Talk about
navigating in the fog without a GPS!
Now, let’s look at how the Taylor Rule would have guided
the level of the federal funds rate vs. the actual level of the funds rate and
comparing that with reported nominal GDP growth with implicit Taylor Rule
“targeted” nominal GDP growth. The blue line plotted in Chart 1 is the
four-quarter moving average of percentage-point differences between the
prescribed Taylor Rule level of the federal funds rate (calculated by Haver
Analytics using the PCE price index and the Congressional Budget Office
estimate of the real GDP output gap) and the actual level of the federal funds
rate. Implicit in the Taylor Rule is that nominal GDP growth is equal to the
growth in potential real GDP plus 2% inflation. The red bars in Chart 1 are
percentage point differences between year-over-year percent changes in reported
nominal GDP and the implicit Taylor Rule “targeted” nominal GDP. The GDP data
and the inflation data incorporate the latest revisions, which, of course would
not have been available to the Fed in real time. The data start in Q1:1987,
when Taylor started estimating his reaction function, and end in Q3:2017.
Chart 1
From Q1:1987 through Q3:1992, the Taylor Rule fed funds
rate was below the actual fed funds
rate (i.e., the blue line in Chart 1 is below zero). Yet, over most of this
period nominal GDP growth was above the implicit Taylor Rule targeted nominal
GDP growth (i.e., the red bars are predominantly above zero). This means that the Taylor Rule would have
performed worse than the actual
discretionary Fed policy in terms of achieving the Taylor Rule targeted nominal
GDP growth. The Taylor Rule’s finest hour, so to speak, was from Q4:2002
through Q4:2006 when the Taylor Rule fed funds rate was above the actual fed
funds rate and nominal GDP growth was above Taylor Rule targeted nominal GDP
growth. Yes, the Taylor Rule was superior to Greenspan’s discretion – a pretty
low bar in my opinion. All of the economic masochists out there would have
loved it if Taylor had “ruled” during the Great Recession. During most of the
last recession, the Taylor Rule fed funds rate was above the actual fed funds
rate. Had the Fed followed the Taylor Rule then, the Great Recession might have
turned into the second Great Depression. From Q3:2010 through Q3:2017, the
Taylor Rule fed funds rate has been above the actual fed funds rate. All else
the same, then, this current economic recovery/expansion would have been even
more anemic than it has been if the Taylor Rule had been followed. Call me
mean-spirited and self-centered, but I was disappointed when John Taylor was
not nominated for Fed chairman by President Trump. I wanted to see how much
damage the Taylor Rule would inflict on the U.S. economy.
At the outset of the commentary, I said that I was in
favor of the Fed monetary policy being guided by a rule rather than by
discretion – just not the Taylor Rule. The rule I would prefer is what I’ll
call the modified Milton Friedman monetary quantity
rule. Friedman advocated that the Fed target a steady rate of growth in some
definition of the money supply. Friedman was against discretion when it came to
conducting monetary policy because he recognized that there is a lot we really
don’t know but think we know when it comes to the macro behavior of the
economy. You can conceive of a Friedmanesque monetary quantity rule as kind of
Hippocratic Oath for central bankers – first, do no harm. Friedman did not
expect his rule to enable monetary policy to prevent the occurrence of business
cycles. Rather, he believed that his rule could reduce the amplitudes of
business cycles. As important, Friedman realized that if his monetary-quantity
growth target were too high or too low, it would not result in inflationary or
deflationary spirals. In other words, inflation might rise if the target were
set too high, but it would stabilize at some higher level. In contrast, if an
interest rate target were set too low, there was no mechanism to keep the rate
of inflation from continuing to rise except an increase in the interest rate
target to some unknown higher level.
Okay, let’s get to the modified Friedman monetary
quantity rule. I know that by now you have guessed it involves the Fed
targeting a steady rate of growth in thin-air credit, thin-air credit being the
sum of the credit created by depository institutions (commercial banks,
S&Ls and credit unions) plus Fed-created cash reserves of these depository
institutions and currency in circulation, or the monetary base. Plotted in
Chart 2 are the quarterly observations of the year-over-year percent changes in
nominal GDP along with the quarterly observations of the year-over-year percent
changes in the sum of depository institution credit plus the monetary base
(i.e., thin-air credit).
Chart 2
The two series appear to move in close tandem. When
thin-air credit growth rises, so does nominal GDP growth and vice versa. One glaring exception to
this “rule” occurs in 2008, when thin-air credit growth went up but nominal GDP
contracted. During this period of financial crisis, households, nonfinancial
institutions as well as financial institutions dramatically increased their
demands for liquidity. The Fed extended massive amounts of credit to financial
institutions to satisfy their increased demand for liquidity. Businesses tapped
their lines of credit at financial institutions to enhance their liquidity
positions, not to purchase new equipment or build inventories of goods. Cross -correlation
tests demonstrate that growth in thin-air credit tends to affect growth in nominal GDP, not the other
way around. When the subperiod from Q1:2008 through Q4:2009 is omitted, the
contemporaneous correlation coefficient between the two series from Q1:1953
through Q3:2017 is 0.61. When thin-air credit growth is lagged by one quarter
to test whether thin-air credit growth last
quarter affects nominal GDP growth this current
quarter, the correlation coefficient rises
to 0.63.
However, when nominal GDP growth is lagged by one quarter
to test whether nominal GDP growth last
quarter affects thin-air credit growth this current quarter, the correlation coefficient falls to 0.56. Oh, and by the way, the Fed publishes weekly the level of “narrow” thin-air
credit – commercial bank credit, the bulk of depository institution credit, and
the monetary base.
What might a modified Friedman monetary quantity growth
rule look like? Both the median and average year-over-year percent change in
quarterly observations of real GDP was 3.1% from Q1:1953 through Q3:2017. Let’s
round it off to 3%. I don’t quite know why 2% is a magic number for an
annualized inflation target. Why not zero or even minus 2%? Whatever the
argument, let’s accept a 2% inflation target. That gives us a 5% annualized
nominal GDP growth target to shoot for. I ran an ordinary least-squares
regression with the year-over-year percent change in nominal GDP as the
dependent variable, year-over-year growth in thin-air credit lagged one quarter
as the explanatory variable along with a constant term. Based on the results of
this regression, in order to achieve 5% annualized growth in nominal GDP,
thin-air credit would have to grow at an annualized rate of 4.6%. (There is
evidence of a lot of serial correlation in nominal GDP growth. That is, this
quarter’s growth in nominal GDP is highly correlated with the previous quarter’s
growth. The estimate of target thin-air growth was made without correcting for
the serial correlation. When I did correct for it, the coefficient on thin-air
credit growth decreased, but remained statistically significant with a 99%
probability.) As an aside, in Q3:2017, thin-air credit was up 4.0% vs. Q3:2016.
So, the Fed has a read on the growth in narrow thin-air
credit on a weekly basis. If narrow thin-air credit is growing above its target growth rate, the Fed
would drain a sufficient amount of
reserves from the financial system so as to get thin-air credit growth back to
target. If narrow thin-air credit were growing below target, the Fed would
add a sufficient amount of reserves to get thin-air credit growth back to
target. The guys and gals in the Chicago interest rate futures pits would love
this because short-maturity interest rates would be much more variable, similar
to the price of wheat.
Now that I have aired my 2017 grievances, it is time to
gather around the aluminum Festivus pole (aluminum is best, according to Frank
Costanza, because of its high strength-to-weight ratio) and join me in singing the
only Festivus carol I know:
A Festivus Carol
(Lyrics by
Katy Kasriel to the melody of O’ Tannenbaum)
O’
Festivus, O’ Festivus,
This one’s
for all the rest of us.
The worst
of us, the best of us,
The shabby
and well-dressed of us.
We gather
‘round the ‘luminum pole,
Air
grievances that bare the soul.
No slights
too small to be expressed,
It’s good
to get things off our chests.
It’s time
now for the wrestling tests,
Feel free
to pin both kin and guests,
Festivus,
O’ Festivus,
The holiday for the
rest of us.
Paul L. Kasriel
Founder, Econtrarian LLC
Senior Economic and Investment Advisor
1-920-818-0236
“For most of human
history, it made good adaptive sense to be fearful and emphasize the negative;
any mistake could be fatal”, Joost Swarte
∆ + 6 = A Good Life
