Daniel Lewis et al., U.S. Economic Activity During the Early Weeks of the SARS-Cov-2 Outbreak (2020)

Summary

National Bureau of Economic Research Working Paper 26954, dated April 2020, by Daniel Lewis, Karel Mertens and James H. Stock, introducing a Weekly Economic Index to track U.S. activity during the early weeks of the COVID-19 outbreak. The index takes the first principal component of 10 weekly series from private sources such as industry groups and polling firms, using data from January 2008, and leaves out financial data. It is scaled to four-quarter GDP growth. The authors find a strong and sudden decline beginning in the week ending March 21, 2020, and a further fall to -6.19% in the week ending March 28.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Cite this document

“Daniel Lewis et al., U.S. Economic Activity During the Early Weeks of the SARS-Cov-2 Outbreak (2020).” Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/daniel-lewis-et-al-u-s-economic-activity-during-the-early-weeks-of-the-d707f60a04/, SHA-256 59abf5b6771560dfa3eecdc72750253311de1870a6014ede5e9913909fae7be9.

Full text

NBER WORKING PAPER SERIES

U.S. ECONOMIC ACTIVITY DURING THE EARLY WEEKS OF THE SARS-COV-2
OUTBREAK

Daniel Lewis
Karel Mertens
James H. Stock

Working Paper 26954
http://www.nber.org/papers/w26954

NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
April 2020

Bi-weekly updates of the Weekly Economic Index are available at https://www.jimstock.org/.
The views expressed are those of the authors and do not necessarily reflect the position of the
Federal Reserve Bank of New York, the Federal Reserve Bank of Dallas, the Federal Reserve
System, or the National Bureau of Economic Research. We are grateful to Mihir Trivedi and Eric
Qian for research assistance, to Tyler Atkinson for useful suggestions, and to Mark Booth for
sharing the tax collections data.

NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.

© 2020 by Daniel Lewis, Karel Mertens, and James H. Stock. All rights reserved. Short sections
of text, not to exceed two paragraphs, may be quoted without explicit permission provided that
full credit, including © notice, is given to the source.
U.S. Economic Activity During the Early Weeks of the SARS-Cov-2 Outbreak
Daniel Lewis, Karel Mertens, and James H. Stock
NBER Working Paper No. 26954
April 2020
JEL No. C51,E01,E66

ABSTRACT

This paper describes a weekly economic index (WEI) developed to track the rapid economic
developments associated with the response to the novel Coronavirus in the United States. The
WEI shows a strong and sudden decline in economic activity starting in the week ending March
21, 2020. In the most recent week ending March 28, the WEI indicates economic activity has
fallen further to -6.19% scaled to 4 quarter growth in GDP.

Daniel Lewis James H. Stock
Federal Reserve Bank of New York Department of Economics
33 Liberty St Harvard University
New York, NY 10045 Littauer Center M26
daniel.lewis@ny.frb.org Cambridge, MA 02138
and NBER
Karel Mertens James_Stock@harvard.edu
Research Department
Federal Reserve Bank of Dallas
2200 N Pearl Street
Dallas, TX 75201
mertens.karel@gmail.com
Economists are well-practiced at assessing real economic activity based on a range of familiar aggregate
time series, such as the unemployment rate, industrial production, or GDP growth. However, these series
represent monthly or quarterly averages of economic conditions, and are only available at a considerable
lag, after the month or quarter ends. When the economy hits sudden headwinds such as the COVID-19
pandemic, conditions can evolve rapidly. How can we monitor the high-frequency evolution of the
economy in “real time”?

To address this challenge, this paper develops a Weekly Economic Index (WEI) that measures real
economic activity at a weekly frequency and that can be updated relatively quickly. 2 Few of the
government agency data releases macroeconomists often work with are available at weekly or higher
frequency. Our weekly series instead come mostly from private sources such as industry groups that
collect data for the use of their members, or from commercial polling companies. Financial data, such as
stock market prices and interest rates, are also available at high frequency. We do not use financial data
in the construction of the WEI, as our objective is to obtain a direct measure of real activity, and not of
financial conditions.

To compute our index, we extract the first principal component from 10 weekly time series, using the
sample from January 2008 to present. We scale our baseline index to four-quarter GDP growth, such that
a reading of 2 percent in a given week means that if the week’s conditions persisted for an entire quarter,
we would expect, on average, 2 percent growth relative to a year previous. 3 The top panel in Figure 1
plots the WEI based on data through April 2, 2020. The trough of the Great Recession is clearly visible, as
well as the subsequent recovery. The WEI index also shows a modest decline during the 2015-2016 mini-
recession, during which the energy and agricultural sectors as well as certain segments of the
manufacturing economy experienced substantial slowdowns in growth.

The bottom panel in Figure 1 shows the evolution of the WEI from January 2019 to its most recent value.
As is clear from the figure, developments related to the Coronavirus pandemic have led the index to fall
to levels below those of 2008 in recent weeks. Specifically, the WEI registers a strong and sudden decline
in economic activity starting in the week ending March 21, 2020, falling to -3.22%. For reference, the WEI
stood at 1.58 for the week ending February 29. The week ending March 21 saw an unprecedented 3.28
million initial UI claims (seasonally adjusted), a sharp decline in consumer confidence and fuel sales, and
a more modest decline in steel production. There was also a countervailing surge in retail sales, as
consumers took to stores to stock up on consumer staples. In the most recent week ending March 28, the
WEI has fallen to -6.19%. This further decline is driven by another sharp increase in unemployment
insurance initial claims, which came in at 6.65 million (seasonally adjusted), far surpassing the prior week’s
record-setting release. The drop was reinforced by further major declines in fuel sales in response to stay-
at-home orders and other restrictions, as well as modest decreases in electricity output and rail traffic.

2
The WEI builds on earlier work by Stock (2013). A preliminary version of WEI was presented in a NY Fed blog post,
see Lewis, Mertens and Stock (2020).
3
Specifically, the mean and standard deviation of the Weekly Economic Index have been adjusted so that they
match the mean and standard deviation of the four-quarter growth of GDP from 2008 through the fourth quarter
of 2019.

1
To track the rapidly evolving economic fallout of the Coronavirus pandemic, the WEI is updated weekly
every Tuesday and Thursday. The weekly updates contain preliminary estimates for the prior week based
on the available data. The latter are based on estimated historical relationships between the WEI and the
series available at the time of the update. The final values of the WEI are available after two weeks.

Figure 1: Weekly Economic Indicator (WEI)

Notes: Based on data available through April 2, 2020. The units are scaled to 4-quarter GDP growth.

The rest of this paper describes the underlying weekly data series as well as the details behind the
construction of WEI. We also document the close relationship between the WEI and widely used lower
frequency indicators of aggregate economic activity in the US, such as real GDP growth and industrial
production.

I. The Weekly Data Series

Table 1 below lists the series we use to construct our baseline WEI. These include a measure of same-
store retail sales, an index of consumer sentiment, initial and continued claims for unemployment
insurance, an index of temporary and contract employment, tax collections from paycheck withholdings,
a measure of steel production, a measure of fuel sales, a measure of railroad traffic, and a measure of
electricity consumption. Unless the source already provides year-on-year growth rates, we transform all
series to represent 52-week percentage changes, which also eliminates most seasonality in the data.
Figure 2 plots all the transformed series that serve as inputs to the index.

2
Table 1: Weekly Variables

Series Native Units Notes
The index is sales-weighted, year-over-year same-store sales growth
Redbook Research: Same
NSA, Y/Y % Chg. for a sample of large US general merchandise retailers representing
Store, Retail Sales
about 9,000 stores. By dollar value, the Index represents over 80% of
Average, Y/Y % Chg.
the "official" retail sales series collected by the Department of
Commerce. http://www.redbookresearch.com/
Daily survey of 1500 American adults Sun-Thurs. Index is a 3-day
Rasmussen Consumer Index
moving average based on five questions about the current state of both
Index
the economy and personal finances, whether the economy and
personal finances are getting better or worse, and whether the
economy is in a recession. https://www.rasmussenreports.com/
Unemployment Insurance: Number of claims filed by unemployed individuals after separation
NSA, Thous.
Initial Claims from an employer. Data collected from local unemployment offices.
https://oui.doleta.gov/unemploy/
Insured Unemployment Number of continued claims filed by unemployed individuals to receive
NSA, Thous.
(Continued Claims) benefits. Data collected from local unemployment offices.
https://oui.doleta.gov/unemploy/
American Staffing
NSA, Jun-12- The ASA Staffing Index tracks temporary and contract employment
Association Staffing trends. Participants include a stratified panel of small, medium, and
Index 06=100
large staffing companies that together provide services in virtually all
sectors of the industry. They account for about one-third of industry
sales offices. https://americanstaffing.net/
Federal Withholding Tax Y/Y % Chg. Treasury receipts of income and payroll taxes withheld from
Collections paychecks. The series is filtered for daily volatility patterns and
adjusted for tax law changes. https://taxtracking.com/

Raw Steel Production NSA, Thous. Net Weekly production tonnage provided from 50% of the domestic
Tons producers combined with monthly production data for the remainder.
https://www.steel.org/industry-data

US Fuel Sales to End Users NSA, EOP, Thous. Weekly product supplied minus change in stock of finished gasoline
barrels/day and distillate fuels. This estimates wholesale gasoline + diesel sales to
retailers and large end users (e.g., UPS). Published by the U.S. Energy
Information Administration in the Weekly Petroleum Status Report.
https://www.eia.gov/petroleum/supply/weekly/
NSA, units Total carloads and intermodal units reported by railroad companies to
U.S Railroad Traffic
the Association of American Railroads https://www.aar.org/data-
center/

Electric Utility Output NSA, Gigawatt Total output for U.S. (excluding Alaska and Hawaii) investor-owned
Hours
electric companies. https://www.eei.org/

As can be seen in Figure 2, some of the weekly series exhibit considerable noise from week to week, such
that gleaning broader trends from any one series can be difficult. The series, however, also display a clear
cyclical pattern, which suggests that they might usefully be combined into a single index. The WEI is

3
computed from these ten series using the method of principal component analysis. The first principal
component of these ten series provides an estimate of a signal about the economy which is common to
all variables. By construction, the Weekly Economic Index is a weighted average of the ten series. The
mathematics of principal components analysis is summarized next.

Figure 2 Weekly Variables and WEI

4
Notes: Based on data available through April 2, 2020. For sources, see Table 1.

II. Construction of the Weekly Economic Index 4

A leading framework for the construction of an economic index from multiple time series is the so-called
dynamic factor model, developed by Geweke (1977) and Sargent and Sims (1977). The dynamic factor
model posits the existence of a small number of unobserved or latent series, called factors, which drive
the co-movements of the observed economic time series. Application of dynamic factor models to
estimating economic indexes range from the construction of state-level indexes of economic activity
(Crone and Clayton-Matthews, 2005) to large-scale indexes of economic activity (for example, the Chicago
Fed National Activity Index, or CFNAI). See Stock and Watson (2016) for a review.

The premise of a dynamic factor model is that a small number – in our application, a single – latent factor,
ft, drives the co-movements of a vector of N time-series variables, Xt. The dynamic factor model posits
that the observed series is the sum of the dynamic effect of the common factor and an idiosyncratic
disturbance, et, which arise from measurement error and from special features that are specific to an
individual series:

4
Parts of this section are adapted from the appendix in Stock (2013).

5
Xt = λ(L)ft + et (1)

where L is the lag operator. The elements of the N×1 vector of lag polynomials λ (L) are the dynamic
factor loadings, and λi(L)ft is called the common component of the ith series. The dynamic factor can be
rewritten in static form by stacking ft and its lags into single vector Ft, which has dimension up to the
number of lags in λ(L):

Xt = ΛFt + et (2)

where Λ is a matrix with rows being the coefficients in the lag polynomial λ(L).

The two primary methods for estimating the unobserved factor ft are by principal components and using
state space methods, where the factor is estimated by the Kalman filter. Broadly speaking, early low-
dimensional applications used parametric state-space methods and more recent high-dimensional
applications tend to use nonparametric principal components or variants. We used both methods in
developing the WEI, but found the results using the parametric DFM to be sensitive to specification details
(lags, sample length, etc.), so principal components estimation is used in this paper.

Table 2: PCA Results

Series Weights Baseline Weights Trimmed (ALS)

Same-Store Retail Sales 0.29 0.29
Consumer Confidence 0.23 0.21
Tax Collections 0.30 0.31
Initial Claims -0.38 -0.38
Continued Claims -0.41 -0.41
Temp Staffing 0.40 0.40
Steel Production 0.37 0.36
Fuel Sales 0.17 0.18
Electricity Output 0.12 0.13
Railroad Traffic 0.34 0.34

Total variance explained 54.4% 52.7%

Notes: Estimation sample is first week of 2008 through last week of February 2020. The first
column uses all observations. The second column is based on a trimmed sample in which outliers
were removed. In this case, the weights are estimated using alternating least squares, see for
instance Stock and Watson (2002b).

6
An alternative approach to using high-frequency data for real-time monitoring (“nowcasting”) is to focus
on forecasting a specific economic release, such as the monthly change in employment, and to construct
a model that updates those forecasts as new data comes in. The dynamic factor model and its state space
implementation is useful for this purpose because a single model automatically adapts to new data
becoming available to estimate the variable of interest. For applications of dynamic factor models to
nowcasting, see Giannone, Reichlin and Small (2008) and Aruoba, Diebold and Scotti (2009).

Table 2 provides the weights associated with the first principal component, as well as the total variance
explained based on the 10 weekly series described above. The first column provides the weights using the
full sample between the first week of January 2008 and the last week of February, 2020. The second
column shows the weights over the same sample period, but after treating outliers in the weekly series
as missing observations. Removing outliers overall has little effect on the weights, and for WEI we
therefore use the full-data weights. We find that WEI explains 54% of the overall variance of the
underlying series.

Robustness The WEI is robust to changes in the details of its construction. Subtracting or adding
individual series has little effect on the overall path; the same is true for estimating the weights on each
series using only more recent data. The left panel of Figure 3 compares our baseline index to one with a
subset of 7 variables (omitting railroad traffic, tax withholdings and continuing claims). The middle panel
plots a version in which we omit consumer sentiment. Both figures illustrate that the common signal is
not driven by the precise choice of series. The right panel of Figure 3 plots the baseline WEI against a
series computed with weights estimated using only data from 2015 onward, showing that the relationship
between these series has been fairly constant during and after the Great Recession.

Figure 3: Robustness Checks

Notes: Based on data available through April 2, 2020.

Real Time Updating The WEI is published every Tuesday and Thursday following the releases of the
underlying weekly data. Each WEI release reports a preliminary estimate for the prior week based on the
data available at that point. The Tuesday update reflects incoming data on retail sales, consumer
confidence and steel production for the prior week, as well as data on temp staffing for the week before
the prior week. The Thursday update reflects incoming data on fuel sales, tax collections, initial claims,
electricity use, and railroad traffic for the prior week, as well as data on continued claims for the week
before the prior week. Final values for the WEI are available with a two-week lag.

7
Our procedure to produce the preliminary values is based on forecasting regressions for the WEI on two
lags of WEI as well as the non-missing data. For example, in the Thursday update, where we miss the
latest data points for two series, we use the estimated value from the forecasting regression of the WEI
on its lags and the current values of the eight available series. The forecasting regressions use the same
estimation sample (Jan 2008 to Feb 2020) as the one used to estimate weights in Table 2. The Tuesday
update misses observations of two consecutive weeks of continuing claims. In that case, we first produce
a preliminary WEI for two weeks prior based on this procedure, and subsequently repeat the forecasting
step for the latest week using the preliminary WEI as input for the forecast.

III. Relationship Between the Weekly Index and Lower Frequency Measures

Figure 4 plots the WEI together with the four-quarter growth rate of real GDP. The latter series is used to
scale the baseline index. A reading of 2 percent in a given week indicates that if the week’s conditions
persisted for an entire quarter, real GDP would be 2 percent greater than the same quarter a year ago.
The panels in the first row of Figure 5 plots the index against the monthly change in nonfarm payrolls and
the twelve-month percentage change in industrial production (IP). The figure shows that the index tracks
both payroll changes and IP growth closely. The bottom panels of Figure 5 plot the index against two
additional monthly activity indicators: capacity utilization and the ISM manufacturing index. The index
also tracks these series relatively well. The close relationship with the lower frequency measures indicates
that, despite the noise inherent in the raw high-frequency data, our methodology to combine these data
into a weekly index produces an informative and timely signal of real economic activity.

Figure 4: WEI and GDP growth

Notes: Based on data available through April 2, 2020.

8
Figures 4 and 5 help to illustrate two important differences between our index and a nowcast,
like those for GDP growth produced by the Federal Banks of New York, Atlanta or St. Louis. First, a nowcast
focuses on a single important target series, and uses the information contained in intermediate data to
predict that series. In contrast, while we report the WEI in GDP growth units, this is simply an ex post
normalization; the WEI does not focus on a single outcome by targeting either a consumption variable or
a production variable—both are important to get a sense of real activity. Second, most nowcasts
(including those of the New York, Atlanta and St Louis Feds) focus on lower-frequency targets like GDP
growth, which are very informative about the economy. But, since GDP is a quarterly variable, such models
are not equipped to highlight variation from one week to the next (see also McCracken, 2020). The goal
of these nowcasts is only to predict average variation in the target series over thirteen weeks, which they
generally do well.

Figure 5: Relationship Between WEI and Other Monthly Activity Measures

Notes: Based on data available through April 2, 2020.

9
This said, it is useful to examine the predictive content of the WEI for relevant lower frequency activity
indicators. A complication in comparing weekly and monthly data is the non-alignment of the calendar.
To address this non-alignment, we introduce the concept of “pseudo-weeks”, which divide the month into
four weeks, the first starting on the first day of the month, the first three having seven days (and thus 5
weekdays and 2 weekend days), and the final pseudo-week running from 22nd through final day of the
month (so including between 7 and 10 days). Each day of the month naturally falls into a calendar week
of the original WEI, so we compute the pseudo-week WEI as an average of the WEI of the constituent
days. With these pseudo-weeks, we have an approximate measure of the signal provided by the index
after the first, second, third, and fourth weeks of the month. We also calculate a monthly WEI by
computing the average WEI for all constituent days.

Payroll Employment We first explore predictive power for changes in payroll employment. Specifically,
we begin by computing a monthly regression,
𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚ℎ𝑙𝑙𝑙𝑙
Δ𝑌𝑌𝑡𝑡 = 𝑐𝑐 + 𝛽𝛽𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡 + ∑2𝑠𝑠=1 𝛾𝛾𝑠𝑠 Δ𝑌𝑌𝑡𝑡−𝑠𝑠 + 𝑢𝑢𝑡𝑡 , (4)

where 𝑌𝑌𝑡𝑡 is monthly private payroll employment. We compute heteroskedasticity and autocorrelation
robust standard errors using the EWC estimator recommended by Lazarus et al (2018). Column (I) of Table
3 reports the results; the WEI is a highly significant predictor of employment changes, with an R2 of 0.83
(regressing on WEI alone gives 0.66).

We run an additional regression, reported in column (II), adding the change in employment from the ADP
release as a control. The ADP release, which precedes the official payroll numbers by two days, is known
to be highly informative for the eventual BLS release. We find that despite this strong relationship, the
WEI provides additional information, above and beyond the ADP release.

Next, we turn to intra-month regressions. Week by week, we run “nowcasting” regressions based on the
information flow from the WEI. These take the form
𝑤𝑤
Δ𝑌𝑌𝑡𝑡 = 𝑐𝑐 + ∑𝑤𝑤 2
𝑖𝑖=1 𝛽𝛽𝑖𝑖 𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡 + ∑𝑠𝑠=1 𝛾𝛾𝑠𝑠 Δ𝑌𝑌𝑡𝑡−𝑠𝑠 + 𝑢𝑢𝑡𝑡 , 𝑤𝑤 = 1,2,3,4 ;
𝑖𝑖
(5)
𝑤𝑤
where 𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡 𝑖𝑖 is the average WEI for the ith pseudo-week of month t. For employment, since the payroll
survey is conducted during the second week of the month, we consider the last two pseudo-weeks of the
prior month and the first two pseudo-weeks of the current month. The results are reported in columns
(III) to (VI). In regression (VI), we find that the second pseudo-week of the month, that on which the payroll
survey is focused, is a significant predictor of employment changes. Moreover, from the last week of the
prior month onwards, the weekly information provided by the WEI is jointly significant (from the F-test
that all weekly coefficients are zero).

Industrial Production The WEI also helps to nowcast industrial production (IP). While Figure 5 shows a
clear relationship between 12-month percentage changes in IP and the WEI, we now consider the more
conventional monthly percentage change. We regress this first on the monthly WEI and lagged IP growth
according to (4), where 𝑌𝑌𝑡𝑡 is monthly log IP. Column (I) of Table 4 shows that the monthly average WEI
(and lags) explains 17% of variation in IP growth, about two weeks before the official release (still 16%

10
dropping lags of IP growth). We then proceed with the weekly nowcasting regressions, following (5). We
find that, from the second week of the month onwards, the flow of information from the WEI is a
significant predictor of monthly IP growth; the explained variation rises from 15% to 28%. The most recent
week is a significant positive predictor of IP growth, while the first week is a negative predictor, since it is
closely related to production in the prior month.

Table 3: Employment Regression Results

Regressors (I) (II) (III) (IV) (V) (VI)
𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚ℎ𝑙𝑙𝑙𝑙 21.90*** 10.54*
𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡
(6.78) (6.05)
Δ𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 0.84***
(0.09)
WEI week 2, 111.51*
current month (60.87)
WEI week 1, 51.13 -46.11
current month (39.06) (62.10)
WEI week 4, past 54.02 -2.32 8.64
month (42.37) (67.83) (68.51)
WEI week 3, past 10.77 -42.09 -33.66 -56.13
month (10.30) (49.64) (50.62) (54.78)
F-test: weekly 1.09 6.41 6.55 3.61
coefficients = 0 (0.32) (0.01) (0.01) (0.05)
F-test: weekly 0.50 0.79 0.76
coefficients equal (0.62) (0.53) (0.58)
SER 96.42 73.20 97.84 97.01 96.40 94.39
Adjusted R2 0.83 0.90 0.83 0.83 0.83 0.84
Notes: HAR standard errors computed using the EWC estimator of Lazarus et al (2018). Results starred at the 1%,
5%, and 10% levels, ***,**,*.

GDP growth Finally, the WEI also aids in nowcasting GDP growth. To show this, we first regress GDP
growth on the quarterly WEI, following (4), where Δ𝑌𝑌𝑡𝑡 is 4-quarter GDP growth (percent) and we replace
the monthly average WEI with the quarterly average WEI. The results in Column (I) of Table 5 show that
the quarterly WEI is a significant predictor of GDP growth, with 89% of variation explained (85% without
lagged GDP growth), nearly a month before the advance release. We then regress the 4-quarter growth
rate on the flow of information from the WEI, starting with the WEI for just the first month of the quarter,
and so on, following
𝑚𝑚
Δ𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 = 𝑐𝑐 + ∑𝑚𝑚 2
𝑖𝑖=1 𝛽𝛽𝑖𝑖 𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡 + ∑𝑠𝑠=1 𝛾𝛾𝑠𝑠 Δ𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡−𝑠𝑠 + 𝑢𝑢𝑡𝑡 , 𝑚𝑚 = 1,2,3 ;
𝑖𝑖
(6)

where Δ𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 is 4-quarter GDP growth (percent). Columns (II) to (IV) report the results. For the first two
months of the quarter, the most recent month’s WEI is a significant (positive) predictor of growth, with
the adjusted R2 rising from 0.86 to 0.92. Data on the final month does not appear to add much additional
information, although the coefficients on monthly WEI are jointly significant for all specifications. We

11
conclude that a strong signal of GDP growth is available from the WEI from the second month of the
quarter, nearly two months before the advance release.

IV. Conclusion

In normal times, familiar macroeconomic aggregates provide accurate descriptions of economic
conditions with a modest delay. When conditions evolve rapidly from day to day and week to week, as is
the case in the current environment, less familiar sources of data can provide an informative and timely
signal of the state of the economy. The WEI provides a parsimonious summary of that signal.

Table 4: Industrial Production Regression Results

Regressors (I) (II) (III) (IV) (V)
𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚ℎ𝑙𝑙𝑙𝑙
𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡 0.17
(0.11)
WEI week 4, 0.99**
current month (0.36)
WEI week 3, 0.57* -0.11
current month (0.31) (0.26)
WEI week 2, 0.76** 0.18 -0.05
current month (0.31 (0.33) (0.27)
WEI week 1, 0.14 -0.60** -0.57** -0.63**
current month (0.10) (0.28) (0.26) (0.25)
F-test: weekly 1.94 2.99 1.92 2.78
coefficients = 0 (0.19) (0.09) (0.19) (0.09)
F-test: weekly 0.00 2.50 1.84 1.80
coefficients equal (0.13) (0.20) (0.21)
SER 0.67 0.68 0.66 0.65 0.62
Adjusted R2 0.20 0.28
0.17 0.15 0.18
Notes: HAR standard errors computed using the EWC estimator of Lazarus et al (2018). Results starred at the 1%,
5%, and 10% levels, ***,**,*.

12
Table 5: GDP Regression Results

Regressors (I) (II) (III) (IV)
𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞
𝑊𝑊𝑊𝑊𝑊𝑊𝑡𝑡 0.70**
(0.22)
WEI month 3 0.25
(0.57)
WEI month 2 1.66*** 1.25
(0.45) (0.86)
WEI month 1 0.64* -1.12** -0.97**
(0.28) (0.42) (0.37)
F-test: weekly 5.42 12.81 8.14
coefficients = 0 (0.06) (0.01) (0.04)
F-test: weekly 4.28 2.87
coefficients equal (0.08) (0.17)
SER 0.55 0.63 0.48 0.48
Adjusted R2 0.89 0.86 0.92 0.91
Notes: HAR standard errors computed using the EWC estimator of Lazarus et al (2018). Results starred at the 1%,
5%, and 10% levels, ***,**,*.

References

Aruoba, S.B., F.X. Diebold, and C. Scotti, (2009), “Real-Time Measurement of Business Conditions,” Journal
of Business & Economic Statistics 27, 417-427.

Bai, J., (2003), “Inferential Theory for Factor Models of Large Dimensions,” Econometrica, 71, 135-172.

Bai, J., and S. Ng, (2006), “Confidence Intervals for Diffusion Index Forecasts and Inference for Factor-
Augmented Regressions,” Econometrica, 74,1133-1150.

Chamberlain, G., and M. Rothschild, (1983), “Arbitrage Factor Structure, and Mean-Variance Analysis of
Large Asset Markets,” Econometrica, 51,1281-1304.

Crone, T.S. and A Clayton-Matthews, (2005), “Consistent Economic Estimates for the 50 States,” The
Review of Economics and Statistics, 87, 593-603.

Engle, R.F., and M.W. Watson, (1981), “A One-Factor Multivariate Time Series Model of Metropolitan
Wage Rates,” Journal of the American Statistical Association, 76, 774-781.

Geweke, J., (1977), “The Dynamic Factor Analysis of Economic Time Series,” in Latent Variables in Socio-
Economic Models, ed. by D.J. Aigner and A.S. Goldberger, Amsterdam: North-Holland.

Giannone, D., L. Reichlin, and D. Small, (2008), “Nowcasting: The Real-Time Informational Content of
Macroeconomic Data,” Journal of Monetary Economics, 55, 665-676.

Lazarus, E., D.J. Lewis, J.H. Stock, and M.W. Watson, (2018), “HAR Inference: Recommendations for
Practice,” Journal of Business and Economic Statistics, 36, 541-559.

13
Lewis, D.J., K. Mertens, and J.H. Stock, (2020), “Monitoring Economic Activity in Real Time,” Liberty Street
Economics, March 28, 2020.

https://libertystreeteconomics.newyorkfed.org/2020/03/monitoring-real-activity-in-real-time-
the-weekly-economic-index.html

McCracken, M., (2020), COVID-19: Forecasting with Slow and Fast Data, On the Economy Blog, Federal
Reserve Bank of St Louis, https://www.stlouisfed.org/on-the-economy/2020/april/covid-19-
forecasting-slow-fast-data

Sargent, T.J., (1989), “Two Models of Measurements and the Investment Accelerator,” Journal of Political
Economy 97:251–287.

Sargent, T.J., Sims, C.A., (1977). Business cycle modeling without pretending to have too much a-priori
economic theory. In: Sims, C. et al., (Ed.), New Methods in Business Cycle Research. Federal
Reserve Bank of Minneapolis.

Stock, J.H., (2013) “Economic Activity During the Government Shutdown and Debt Limit Brinkmanship,”
Council of Economic Advisors, Report October 2013.

Stock, J.H., and M.W. Watson, (1989), “New Indexes of Coincident and Leading Economic Indicators,”
NBER Macroeconomics Annual 1989, 351-393.

Stock, J.H., and M.W. Watson, (1999), “Forecasting Inflation,” Journal of Monetary Economics, 44, 293-
335.

Stock, J.H., and M.W. Watson, (2002a), “Forecasting Using Principal Components from a Large Number of
Predictors,” Journal of the American Statistical Association, 97, 1167-1179.

Stock, J.H., and M.W. Watson, (2002b), “Macroeconomic Forecasting Using Diffusion Indices,” Journal of
Business and Economic Statistics, 20, 147-162.

Stock, J.H., and M.W. Watson, (2016), “Factor Models and Structural Vector Autoregressions in
Macroeconomics,” Handbook of Macroeconomics: Volume 2 (2016), 415-525.

File and source

File: lewis-2020-u-s-economic-activity-during-the-early-weeks-of-the-sars-cov-2-outbreak.pdf
Size: 645,783 bytes
SHA-256: 59abf5b6771560dfa3eecdc72750253311de1870a6014ede5e9913909fae7be9
Our copy: lewis-2020-u-s-economic-activity-during-the-early-weeks-of-the-sars-cov-2-outbreak.pdf
Original: www.nber.org