View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.



APRIL 2010
	 NUMBER 273

Chicag­ Fed Letter
Chicago Fed National Activity Index turns ten—Analyzing its
first decade of performance
by Scott Brave, business economist, and R. Andrew Butters, associate economist

This article discusses how the Chicago Fed National Activity Index has performed as a
“real-time” indicator of economic activity and related inflationary pressure.

The Chicago Fed National Activity Index

Entering its tenth year, the
CFNAI has proven to be a
useful real-time indicator of
economic activity and related
inflationary pressure.

(CFNAI) is a monthly index constructed
to summarize variation in 85 data series
on U.S. economic activity.1 It is also an
example of a “Goldilocks” index, reflecting deviations around a trend rate of
economic growth represented by a
zero value of the index. Zero is “just
right” and suggests that the U.S. economy is proceeding along its historical
growth path; a negative value is “cold”
and suggests that growth is below average; and a positive value is “hot” and
suggests that growth is above average.
The ability of the CFNAI to capture sustained deviations from trend in economic
activity has led to its frequent use as an
indicator of business cycles.2 However,
the CFNAI was originally suggested as
an indicator for forecasting inflation,
based on the relationship between deviations in economic activity from trend
and the level of economic slack in the
U.S. economy.3 The more slack (or the
more negative the index value), generally
the less upward pressure there is on prices;
conversely, the less slack (or the more
positive the index value), the more upward pressure there is on prices.
In this Chicago Fed Letter, we examine how
the CFNAI has performed as an indicator of both economic activity and related
inflationary pressure since its initial release in March 2001. With the index
reaching its tenth year of publication,
we now have a reasonable sample with

which to judge its ability to identify recessions and periods of sustained increasing inflation as they are happening.
Now we can also more reliably assess its
ability to forecast common measures of
economic growth, such as gross domestic
product (GDP), and measures of inflation, like the Personal Consumption
Expenditures (PCE) Price Index. We
find that, overall, the CFNAI has proven
to be a useful indicator for both purposes
over the past decade.
Identifying business and inflation cycles

The CFNAI can be very volatile, since
many of the monthly series that make up
the index vary significantly from month
to month. For this reason, the release
of the monthly index is accompanied
by a three-month moving average index,
i.e., the CFNAI-MA3 , which smoothes
these month-to-month variations over
time and provides a more consistent
picture of variations in economic growth
around trend. When the CFNAI-MA3
reaches certain levels that have been
identified in previous research as “too
hot,” the likelihood of a period of sustained increasing inflation rises; when
it gets “too cold,” the likelihood of a
recession rises.
For instance, a CFNAI-MA3 value below
–0.7 after a period of economic expansion indicates an increasing likelihood of
a recession, as substantial resource slack
begins to build up in the U.S. economy.
Conversely, a value above –0.7 after a

1. CFNAI-MA3 and business cycles









NOTES: CFNAI-MA3 is the Chicago Fed National Activity Index’s three-month moving
average. Shading indicates official periods of recession as identified by the
National Bureau of Economic Research; the dashed vertical line indicates the
most recent business cycle peak. A CFNAI-MA3 value below –0.70 following a
period of economic expansion indicates an increasing likelihood that a recession
has begun. Conversely, a CFNAI-MA3 value above –0.70 following a period of
economic contraction indicates an increasing likelihood that a recession has ended.

data may not yet be
available, and the
data that have been
released are often
subject to revision.
As a result, the NBER
typically determines
the beginning and
end dates of business
cycles several quarters
after the event. With
its monthly release
schedule, the CFNAI
serves as a “real-time”
measure of the
business cycle.

In fact, for the 2001
recession, the
CFNAI-MA3 identified the start of the
recession as December
2000 in the March 5,
2. CFNAI-MA3 and inflation cycles
2001, release. Ten
months later, the
NBER identified the
start date of the recession as March 2001.
then identified the
end of the 2001 recession as February
2002 in the March 27,
2002, release. Sixteen
months later, the
NBER put the end
date at November
2001. The index’s real1970 ’75
2000 ’05
time performance
NOTES: CFNAI-MA3 is the Chicago Fed National Activity Index’s three-month moving
during the recent
average. Shading represents periods of substantial inflation increases. A CFNAI-MA3
value above +0.70 more than two years into an economic expansion indicates an
recession has been
increasing likelihood that a period of sustained increasing inflation has begun.
even better. In the
March 24, 2008, release, the CFNAI-MA3 correctly identiperiod of economic contraction indified December 2007 as the recession’s
cates an increasing likelihood that a
start date, eight months before the
recession has ended, as idle resources
NBER announcement doing the same.
begin to be put back to use.
While the NBER has yet to identify this
The history of the CFNAI-MA3 shown
recession’s end, the CFNAI-MA3 idenin figure 1 demonstrates that, based on
tified the likely end date as September
these thresholds, the index has been
2009 in the October 26, 2009, release.
successful in identifying the beginnings
Using the index to identify periods of
and ends of U.S. recessions since 1967
within one to three months of the dates sustained increasing inflation is more
difficult. A formal arbiter of such periods,
determined by the National Bureau of
Economic Research (NBER). Of course, like the NBER, does not exist. Instead,
we rely on an algorithm that looks for
identifying recessions as they are hapsubstantial increases in measures of core
pening is much trickier. Important

inflation (which ignore more volatile
food and energy prices) to identify these
dates.4 If we compare them against the
CFNAI-MA3 as shown in figure 2, we
see that the timing of such signals from
the index is often not precise. Unlike
business cycles where false positive signals tend to be rare, the index has shown
several false positive signals of increasing inflation in its history. Still, it has
generally been true that when the
CFNAI-MA3 has increased above +0.7
more than two years into an economic
expansion, inflation has increased substantially over the following year.
Since its initial release, the CFNAI-MA3
has exceeded +0.7 only twice in real time:
for the months of May 2004 (in the
June 28, 2004, release) and December
2005 (in the January 23, 2006, release).
In subsequent releases, the May 2004
value of the index was revised below
+0.7, while the December 2005 value
remained above +0.7. These months correspond with the month leading up to
and the middle month of this past decade’s lone period of sustained increasing
inflation that we identify. This contrasts
with the previous decade where +0.7 had
been reached on several separate occasions without a subsequent sustained
rise in inflation.
Forecasting GDP growth and core
PCE inflation

To further demonstrate the usefulness of
the index, we compared the CFNAI-MA3’s
ability to forecast current quarter real
GDP growth within each quarter from
2001:Q1 through 2009:Q4 relative to
forecasts based on other well-known
monthly indicators of economic activity.
We then did the same for the change in
core inflation, as measured by the PCE
deflator excluding food and energy
prices. All of our forecasts were made in
a real-time sense, using the actual data
on real GDP, core PCE, the CFNAI-MA3,
and the other monthly indicators available at the time the current quarter
forecasts would have been made.5
Real GDP and core PCE are both quarterly measures. Still, each of these series
has a pseudo-monthly release schedule
incorporating two rounds of revisions
after an initial release; so, if we count

3. Within-quarter forecast performance







(MSFE relative to random walk forecast)

Real GDP growth	





0.70	 0.85	


1.22	 0.83






0.54	 0.68	


0.87	 0.49

Δ Core PCE inflation	 2001–03	




0.91	 0.87	


0.94	 0.98





0.89	 0.86	


0.82	 0.90


Notes: This figure reports the MSFE (mean squared forecast error) of each monthly indicator’s forecast of current quarter GDP growth
and the change in core PCE inflation relative to a random walk forecast. The random walk forecasts specify that the current quarter
log annualized growth of real GDP and core PCE will be the same as in the previous quarter. Prior to estimation, all of the monthly
indicators were transformed in a manner similar to the CFNAI-MA3 by taking a three-month moving average before a stationary
transformation was applied. This transformation involved taking the log difference of industrial production (IP) and payroll employment
(EM); the arithmetic difference of manufacturing capacity utilization (CUMFG), unemployment rate (LR), and average weekly hours
worked in manufacturing (LRMANUA); and the log of housing starts (HST). No stationary transformation was made to the Institute for
Supply Management’s Purchasing Managers’ Index (ISM).
Sources: Authors’ calculations based on data from the Federal Reserve Bank of Chicago and the Federal Reserve Bank of Philadelphia.

the revisions, each is released within a
quarter the same number of times as
the CFNAI-MA3. The CFNAI-MA3 is a
timelier indicator of economic activity;
both real GDP and core PCE experience
a lag of one to three months between
the time period they describe and the
date they are released, whereas the
CFNAI-MA3 and all of our monthly indicators lag by only one month. This fact
makes it possible to produce current
quarter forecasts for each of the three
releases within the quarter, using the
previous quarter’s data on real GDP and
core PCE, along with current quarter
data on the monthly indicators.
For instance, in any given month when
the CFNAI is released, all of the monthly
indicators we examine incorporate data
up to one month prior to the date of this
release. In contrast, the real GDP and
core PCE release that corresponds with
this same month incorporates data only
through the previous quarter. To create
our current quarter forecasts, we aligned
within the quarter the last available value
of real GDP growth and the change in
core PCE inflation to the monthly data in
accordance with this release schedule.
In this way, we used monthly indicators
from the first month of a quarter to predict the first release of real GDP and core
PCE, the same indicators from the second
month to predict the second release,
and the same indicators from the third
month to predict the third release.
To obtain our monthly forecasts, we then
ran a series of regressions of real GDP
growth and the change in core PCE inflation on one of their own lags and the

contemporaneous value and up to five
lags of each of the monthly indicators,
where the number of lags was chosen
by the Bayesian Information Criterion.6
The sample period for these rolling
regressions began in 1967 and extended
to the date of each CFNAI release
over the past decade, ending with the
November 23, 2009, release (October
2009 being the latest period for which we
have matching GDP and PCE data for
our forecasts).7 Using the coefficients
from this regression, we then projected
forward one quarter using the real-time
data to obtain a current quarter forecast.
Figure 3 reports our results.8 As a benchmark, we used a “random walk” forecast
that specified that real GDP growth or
core PCE inflation would be the same
as in the previous quarter. The evaluation
criterion we use in figure 3 is the ratio of
the mean squared forecast error (MSFE)
of each indicator’s forecast relative to the
random walk forecast. A value below 1
indicates that the indicator forecast outperformed the random walk forecast,
while a value above 1 indicates it underperformed. We consider two subsamples:
2001:Q1–2003:Q4 and 2004:Q1–2009:Q4.
We do this to account for the fact that
in November 2003 we replaced several
of the original 85 CFNAI data series.
For 2001–03, the CFNAI-MA3’s GDP forecasts are relatively weak compared with
those for production indicators, such as
industrial production and manufacturing capacity utilization (first row of figure 3). However, the relatively poor
performance of the CFNAI-MA3 during
this period was driven substantially by

large forecast errors in 2003:Q2. This
represents the lone quarter in the past
decade that the index registered a false
positive in signaling an increasing likelihood of a recession. During that time,
the forecasts from several nonproduction
indicators underperformed relative to the
index and even random walk forecasts.
In contrast, all of the monthly indicators demonstrate much improved forecasting performance over the period
2004–09, when GDP growth was particularly volatile (second row of figure 3).
The CFNAI-MA3 forecasts more accurately than all of the other indicators,
with the exception of manufacturing
capacity utilization, which performs just
slightly better. Looking at just the first
GDP release within each quarter, the
CFNAI-MA3 forecasts are roughly on
par with the Philadelphia Fed’s Survey
of Professional Forecasters (SPF) median
quarterly forecasts.9
The CFNAI-MA3’s performance as an
inflation indicator during this ten-year
period is less impressive, but it’s not unlike that of a number of other common
inflation indicators. The results in figure 3 (third and fourth rows) are not significantly different across the indicators,
but they are uniformly lower than the
CFNAI-MA3’s in the 2004–09 period. The

Charles L. Evans, President; Daniel G. Sullivan, Senior
Vice President and Director of Research; Douglas D. Evanoff,
Vice President, financial studies; Jonas D. M. Fisher,
Vice President, macroeconomic policy research; Daniel
Aaronson, Vice President, microeconomic policy research;
William A. Testa, Vice President, regional programs, and
Economics Editor; Helen O’D. Koshy and Han Y. Choi,
Editors; Rita Molloy and Julia Baker, Production
Editors; Sheila A. Mangler, Editorial Assistant.
Chicago Fed Letter is published by the Economic
Research Department of the Federal Reserve Bank
of Chicago. The views expressed are the authors’
and do not necessarily reflect the views of the
Federal Reserve Bank of Chicago or the Federal
Reserve System.
© 2010 Federal Reserve Bank of Chicago
Chicago Fed Letter articles may be reproduced in
whole or in part, provided the articles are not
reproduced or distributed for commercial gain
and provided the source is appropriately credited.
Prior written permission must be obtained for
any other reproduction, distribution, republication, or creation of derivative works of Chicago Fed
Letter articles. To request permission, please contact
Helen Koshy, senior editor, at 312-322-5830 or
email Chicago Fed
Letter and other Bank publications are available
ISSN 0895-0164

index does, however, outperform the
random walk forecast; and only the unemployment rate outperforms the index
in both subsamples. In general, the index’s performance puts it on par with
other common measures of economic

slack, such as the unemployment rate
and manufacturing capacity utilization.

1	 For details, see



Entering its tenth year, the CFNAI has
performed reasonably well as a real-time
	 We obtained the real-time GDP and PCE

inflation data from the Federal Reserve
Bank of Philadelphia’s real-time database
real-time-center. We focus on quarterly
PCE inflation instead of the monthly value
in order to make use of real-time data. We
took real-time data for the CFNAI and each
of the monthly indicators from our own
CFNAI archives.

	 Charles L. Evans, Chin Te Liu, and Genevieve


Pham-Kanter, 2002, “The 2001 recession
and the Chicago Fed National Activity
Index: Identifying business cycle turning
points,” Economic Perspectives, Federal Reserve
Bank of Chicago, Vol. 26, No. 3, Third
Quarter, pp. 26–43; and Scott Brave, 2009,
“The Chicago Fed National Activity Index
and business cycles,” Chicago Fed Letter, Federal
Reserve Bank of Chicago, No. 268, November.

	 This algorithm is described in detail at

	 We excluded forecast months that were


impacted by benchmark revisions to GDP.
We also considered excluding forecasts
impacted by annual revisions to GDP; however, doing so did not much alter the results.

	 Real GDP growth was de-meaned prior to

estimation by using estimates of the mean
shifts prior to 2001 implied in Robert J.
Gordon, 2003, “Exploding productivity
growth: Context, causes, and implications,”
Brookings Papers on Economic Activity, Vol. 34,
No. 2, pp. 207–298.

	 James H. Stock and Mark W. Watson,


manufacturing, and the Institute for Supply
Management’s Manufacturing Purchasing
Managers’ Index forecast results do not
include January and February of 2001. The
other series were obtained for these months
by using the Federal Reserve Bank of
Philadelphia’s real-time database.



1999, “Forecasting inflation,” Journal of
Monetary Economics, Vol. 44, No. 2, October,
pp. 293–335; and Jonas D. M. Fisher,
2000, “Forecasting inflation with a lot of
data,” Chicago Fed Letter, Federal Reserve
Bank of Chicago, No. 151, March.

indicator of economic activity and related
inflationary pressure. In conjunction with
the publication of this article, we are releasing the complete real-time history of the
index. We hope this will encourage additional research on its real-time properties.10

	 The CFNAI archives begin in March 2001.


As a result, the unemployment rate, housing
starts, average weekly hours worked in

	 The SPF forecasts are available only once


each quarter and most closely correspond
to the first release of real GDP and core PCE.

	The full real-time history of the CFNAI


can be found at

Federal Reserve Bank of St. Louis, One Federal Reserve Bank Plaza, St. Louis, MO 63102