realized and anticipated macroeconomic conditions ...mbrandt/papers/working/...this nding is...
TRANSCRIPT
Realized and Anticipated Macroeconomic
Conditions Forecast Stock Returns∗
Alessandro Beber†, Michael W. Brandt‡, Maurizio Luisi§
November 2014
Abstract
We construct daily real-time indices capturing the public information on realized andanticipated economic activity. The one-month change in realized fundamentals predictsU.S. stock returns across horizons with strongest results between a month and a quarter.The information in anticipated fundamentals that is orthogonal to the realized datapredicts returns even more strongly, particularly at longer horizons of up to two quarters.Splitting the sample into times of high versus low uncertainty, as measured by the cross-sectional dispersion of economist forecasts, we show that the predictability is largelyconcentrated in high-uncertainty times. Finally, extending the analysis internationally,we find similar results that are curiously stronger when U.S. data are used as predictorsrather than global composites or local data.
Keywords: stock market predictability, state of the economy, macroeconomic uncertainty.
JEL classification: G12
∗Earlier versions of this paper were circulated under the title “Economic Cycles and Expected Stock Returns.”We thank Inquire UK for financial support. We thank Daryl Caldwell, Robert Darwin, Fabio Fornari, Amit Goyal,Ana-Maria Tenekedjieva, and seminar participants at BlackRock, City University, PanAgora, the 2012 Asset PricingRetreat at Cass Business School, the Fall 2012 Inquire UK Conference in Bath, the Imperial College Hedge FundConference, the London Quant Group Conference, the Stockholm School of Economics, and the University of York,for their comments and suggestions.†Cass Business School, City University London, and CEPR‡Fuqua School of Business, Duke University, and NBER§Bloomberg L.P.
1 Introduction
The risk premium on stocks varies, both through time and across countries. Characterizing this
variation empirically and modeling the underlying economic mechanisms theoretically preoccupies a
substantial part of the academic finance profession. Despite sharing a broad objective, however, the
empirical and theoretical literatures can appear disjointed. Empirical papers tend to use forecasting
variables whose variation is primarily driven by financial markets (e.g., valuation ratios, interest
rates, option implied volatilities), whereas theoretical models are based on economic fundamentals
(e.g., GDP growth, consumption, inflation). This disconnect arises because the fundamental data
empiricists observe is infrequent, backward-looking, and often restated after the initial release,
whereas market based predictors can be measured daily and reflect current forward-looking information.
Empiricists therefore tend to favor market based predictors as implicit proxies for economic fundamentals,
but market prices also reflect other things, such as aggregate preferences and potential misvaluation.
We construct daily measures of economic activity based on the almost continuous flow of
macroeconomic data releases. Our measures are meant to summarize in real-time the public
information about the economy available to market participants. We particularly focus on information
about economic growth (abstracting from information about inflation, housing, trade, and the
public sector) and explicitly differentiate between (i) realized ex-post measures, such as quarterly
GDP releases, and (ii) anticipating ex-ante information, namely data from surveys of consumers
and firm managers. The resulting economic indices overcome some of the concerns about using
economic data for stock return predictability: the indices are measured daily and capture the
information of the entire news flow, not just a few select data series; they are both backward- and
forward-looking, so we can potentially capture leading information as market variables do; and
they are based on carefully dated and unrestated data to alleviate concerns about look-ahead bias
through restatements. We use these real-time economic indices to reexamine whether and, if so,
to what extent stock market returns are predictable by economic fundamentals through time and
across four major economies.
We find that the one-month change in realized growth, constructed from backward-looking
and delayed data, still predicts U.S. stock market returns one to four months into the future. In
addition, the part of anticipated growth, constructed from forward-looking data, that is orthogonal
to realized growth predicts market returns at horizons of two to six months. Due to the orthogonal
construction of the two predictors, the results are roughly additive in a multivariate regression,
leading to return predictability that far exceeds what is generated by the usual suspects: valuation
ratios (D/P and E/P), interest rate spreads (term and default), and the difference between option-
1
implied and realized stock market volatility.1 Moreover, since realized and orthogonal anticipated
growth predict returns at different horizons – the realized growth results are strongest at one and
two month return horizons whereas the orthogonal anticipated growth results are strongest at
one and two quarter horizons – the joint predictions are substantially improved across horizons
compared to univariate specifications.
Our second set of empirical results shows that the predictability by economic fundamentals,
both realized and orthogonal anticipated, is state dependent. The results are much stronger during
times of greater disagreement among economists about economic growth. We measure disagreement
as the weighted average, across economic news series, of the cross-sectional, across economists,
standard deviation of forecasts, where the weights are the same as for our real-time growth index,
reflecting the relative importance and correlation structure of different economic series. We find
that this disagreement measure is highly counter-cyclical, so that return predictability is also much
stronger in economic recessions than in expansions, where we define recessions as below average
growth periods. When we combine greater disagreement and being in a recession, the conditional
predictability is further strengthened, suggesting that each conditioning contains at least some
separate information. For example, unconditionally the bivariate specification has an R2 of 2.6
percent at the monthly horizon. Conditioning on greater disagreement or on being in a recession
raises the R2 to 6.5 and 5.1 percent, respectively. Conditioning on both further raises it to 10.3
percent.
We extend the evidence to international data, specifically equity indices in Europe, Japan, and
the U.K.. We find that economic fundamentals predict equity returns as strongly if not more so
internationally. More interestingly, though, at least for Japan and the U.K., global aggregates and
even just the U.S. factors are more predictive than local versions. For Europe, both are relevant.
This finding is consistent with those of Ang and Bekaert (2007) and Bollerslev et al. (2012), who
also find global aggregates to have stronger forecasting power than local predictors. To quantify
the economic significance of these international results, we construct a global equity market timing
and country selection portfolio based on realized and anticipated global growth. Consistent with
the strong statistical evidence, the Sharpe ratio of this investment strategy is 0.7 annualized, as
compared to a Sharpe ratio close to zero for being long world equities over the same time period.
The value of forecasting with economic fundamentals is as strong during the first half of our sample
as during the second, suggesting that our results are not entirely driven by the 2008-2009 financial
crisis when global economic fundamentals fluctuated the most since the 1930s. Finally, we find that
1The following is a partial list of academic papers that document various degrees of return predictability and thevariables they use: Bollerslev et al. (2009, 2012), variance risk premium; Campbell (1987), term spread; Campbelland Shiller (1988a, 1988b), dividend yield; Cochrane (2008), dividend yield; Fama and French (1988, 1989), defaultspread, dividend yield, term spread; Fama and Schwert (1977), Treasury bill yield; Ferson and Harvey (1991), defaultspread, dividend yield, lagged returns, term spread, Treasury bill yield; Keim and Stambaugh (1986), default spread,trend; Lamont (1998), dividend-to-earnings ratio.
2
the international data corroborates our results on conditioning. In all three regions, the forecasting
ability of realized and orthogonal anticipated growth is much stronger during times of greater
economist disagreement and recessions.
Our paper blends two literatures: that on stock return predictability surveyed partially in
footnote 1 and that on measuring the state of the economy based on economic news data, commonly
referred to as “nowcasting” (see Banbura et al., 2012, for a survey). There are two general
approaches to nowcasting. The first approach is to use a balanced panel regression, along the
lines of the seminal paper of Stock and Watson (1989), now the Chicago Federal National Activity
Index (CFNAI). This first approach uses a large set of news releases but results in a relatively low
measurement frequency because the econometrician has to wait for the panel to be complete before
the index can be constructed. The second approach to nowcasting is to model macroeconomic data
using a latent state-space model (e.g., the ADS business conditions index of Arouba et al., 2009).
The advantage of this second approach is to produce an indicator at a higher frequency, since a
state-space model can effectively handle the sparse and delayed reporting of economic data, but
this technique is impractical for large cross-sections of news releases.2 Our approach to nowcasting
uses a large cross-section of news data but, by forward-filling missing data and making appropriate
correlation matrix adjustments, still produces high frequency indicators. We show in Beber et
al. (2014) that the resulting real-time indices are highly correlated with the CFNAI and ADS
business condition index, but that they appear to be more timely and informative about future
economic fundamentals.
The two closest papers to ours are Ludvigson and Ng (2009) and Bai (2010). Both papers use a
large panel of macroeconomic variables, among other things, to forecast bond returns (Ludvigson
and Ng, 2009) and stock returns (Bai, 2010). Their emphasis is on return predictor selection
and combination by searching over a very large set of not only macroeconomic but also financial
variables. Our focus, in contrast, is on documenting the link between economic fundamentals
and equity returns. We do so by taking a two-step approach. We first extract real-time indices
that best capture the dynamics of economic fundamentals and only then examine how these factors
forecast future stock returns. We explicitly exclude financial variables from the factors, although we
show that the predictability we uncover is not subsumed by the usual suspect financial predictors.
Finally, unlike the vast majority of predictability and nowcasting papers, we exclusively work with
precisely date- and time-stamped initial data releases, as opposed to restated macroeconomic data.
Ghysels et al. (2012) demonstrate the importance of using unrestated data in the case of bond
return predictability.
The paper proceeds as follows. In Section 2, we describe the data and carry out some preliminary
2For example, Evans (2005) only considers the set of different (preliminary, advance, and final) GDP releases.Arouba et al. (2009) construct their business condition index using four indicators at different frequencies, includinga continuously observable financial markets variable.
3
analyses. Section 3 explains our methodology for constructing our real-time economic factors as
well as an empirical proxy for uncertainty surrounding those factors. We present our empirical
results in Section 4, and Section 5 concludes.
2 Data
2.1 Macroeconomic news and forecasts
We obtain data on the dates, release times, and actual released figures for 43 distinct U.S. macro-
economic announcements covering the period from January 1997 through December 2011, for a total
of more than 8,000 announcements over about 3,800 business days.3 This data is obtained from
Bloomberg through the economic calendar screen (i.e., “ECO <Go>”), which provides precisely
time-stamped and unrestated announcement data.4 We also collect data on economist forecasts
for each announcement. Bloomberg surveys economists during the weeks prior to the release of
each major indicator to obtain a consensus estimate. We work with the individual economist
level forecasts, rather than the aggregated consensus forecasts, in order to construct cross-sectional
measures of disagreement for each news release.
Bloomberg contains data for many of our series prior to 1997, but those data are stored in
historical fields which (a) are not associated with clear announcement dates and times (rather they
are dated according to the period they reference) and (b) are restated over time.5 We collect this
more problematic data from January 1990 through 1996 simply to construct an initial correlation
matrix estimate, which is required by our methodology (see Section 3). In order to date the releases
prior to 1997, we compute for each news series the median time between the reference period and
the announcement. For example, the employment report is traditionally released four days after
the end of the month to which the report refers. We then apply this median reporting lag to the
reference period of the older data in order to obtain an approximate announcement date.
We complement the U.S. data with equivalent information for the Eurozone, the U.K., and
Japan. Specifically, we obtain information for 183 European macro releases, for 43 U.K. releases,
and for 45 Japanese releases, over the same sample period. The surveyed economist forecasts are
only available for a slightly shorter sample period, starting in June 1997 for Europe and the U.K,
3We emphasize the fact that we work with distinct announcements because there are a lot more than 43 statisticsif we included multiple versions of essentially the same data released in the same economic report. For example, theCFNAI uses 13 industrial production statistics, resulting in 20 percent of the index being determined by a singlerelease. In contrast, we include in our analysis only the headline month-over-month figure.
4The importance of using real-time versus final data in macroeconomic forecasting has been discussed extensivelyin the literature (e.g., Koenig et al., 2003, or Ghysels et al., 2012).
5For example, there are monthly releases of quarterly GDP labeled “advance,” “preliminary” and “final” allreferring to the same quarter. Bloomberg’s historical field for GDP is dated according to the referenced quarter,so that the advance release gets overwritten by the preliminary release, which in turn gets overwritten by the finalrelease. Historically only the final releases are stored.
4
and starting in May 2000 for Japan.
Most macroeconomic indicators are released on different days and at different frequencies,
making it difficult to process the flow of information in a systematic and consistent way. Figure 1
shows that actual news releases occur with a variety of different lags with respect to the month they
are referencing. Furthermore, news on different indicators are frequently released simultaneously.6
For example, the employment report traditionally announced on the first Friday of the month
contains four different indicators: nonfarm payrolls, nonfarm payrolls in the manufacturing sector,
the unemployment rate, and average weekly hours. Finally, the release frequency varies across
different economic aggregates. Data releases of different economic indicators are usually observed
at different frequencies; e.g., GDP data are sampled quarterly, the nonfarm payrolls are released
monthly, initial jobless claims are sampled weekly, etc. These features of our large cross-section of
macroeconomic news releases generate a sparse matrix of data that our methodology will have to
take up. The Appendix describes in detail the set of U.S., Europe, U.K., and Japan macroeconomic
news in our sample, including their frequency, source, and units of measurement.
2.2 Market returns and other predictors
We collect daily returns data on four major equity indices: the S&P 500 index for the U.S., the
EURO STOXX 50 index for the Eurozone, the FTSE 100 index for the U.K., and the Nikkei 225
index for Japan. The sample period is January 1997 through December 2011. We compute daily
excess returns using corresponding LIBOR rates from Datastream.
We also collect daily data on a number of return predictors used in the literature. Specifically,
we obtain price-to-earnings (P/E) and dividend-to-price (D/P) ratios for the four equity indices
from Datastream. We calculate a default spread variable as the difference between Moody’s BAA
and AAA corporate bond spreads for U.S. issuers and use this variable as a proxy for the default
spread of the other countries as well. We also compute a term spread variable as the difference
between the ten-year and three-month zero coupon Government bond yields, using U.S., German,
U.K., and Japanese data obtained from Datastream.
Finally, following Bollerslev et al. (2009), we construct for each market a daily variance risk
premium V RPt as the difference between the one-month ahead option implied variance and the
expectation of realized variance for the next month:
V RPt = IVt − Et [RVt,t+d] , (1)
where d denotes the number of days in a month, IVt denotes the implied variance from date t to
6For example, for the U.S., there was at least one data release on approximately 80 percent of days. Multipledata releases occurred much less frequently, on approximately 60 percent of the days in the sample. Obviously in thecase of Europe, with over four times as many data series, coincident releases occur much more frequently.
5
t + d, and RVt,t+d =∑d
s=1RVt is the one-month ahead realized variance. The notation reflects
the fact that the implied variance is known at date t whereas the corresponding realized variance
is partially unknown until date t+ d. Both variance measures are assumed to be expressed in the
same units of time – daily, monthly, or annualized.
For the implied variance IVt, we use the square of the VIX volatility index for the U.S., of the
VDAX volatility index for the Eurozone, of the VFTSE volatility index for the U.K., and of the
VXJ volatility index for Japan. All of these volatility indexes are based on highly liquid equity
index options and are constructed by the exchanges with the same model-free calculation approach
of Britten-Jones and Neuberger (2000).
We construct daily realized variances RVt as the sum of the five-minute squared returns over
normal trading hours of each market, as in Bollerslev et al. (2009). The required high frequency
data is obtained from Tickdata. We then forecast one month ahead realized volatility using the
information contained in both realized and implied volatility, as in Drechsler and Yaron (2011).
Specifically, we regress the future realized volatility from t to t + d on the current (i.e., date t)
realized and implied volatilities, the realized volatility from t− d to t− 1, and the average implied
volatility over that same period:7
RV1/2
t,t+d = α+ β1RV1/2
t + β2IV1/2
t + β3RV1/2
t−d,t−1 + β4IV1/2
t−d,t−1 + εt. (2)
We estimate the coefficients of this regression by OLS using a trailing one-year window so as to avoid
any look-ahead bias in the following regression. This forecasting model is sufficiently parsimonious
and delivers large explanatory power in the forecast of future volatility, mainly thanks to the high-
frequency measurement of realized volatility and the degree of persistence imposed by the use of
implied volatility. We also explored more involved specifications with further lags and volatility
measured over different horizons, but those specifications improve the explanatory power only
marginally.8
2.3 Categorizing the macroeconomic news flow
Our aim is to extract a set of factors describing the state of the economy. Rather than relying on a
statistical procedure to obtain orthogonalized factors that are increasingly difficult to interpret with
the order of the factor, we impose a specific economically motivated structure on the macroeconomic
news flow. Based on both empirical evidence and economic rationale, we first separate the aggregate
7Using lagged volatility terms measured on different horizons is consistent with the heterogeneous auto-regressivevolatility forecasting models proposed by Corsi (2009) and used in Corsi, Fusari, La Vecchia (2013) and Mueller,Vedolin, and Yen (2012).
8Our realized volatility forecasting model has an average (or median) R2 of 37 (33) percent. Adding one-weekrealized and average implied volatilities as predictors increases the explanatory power by only one percent.
6
economy into two broad dimensions: the nominal and the real side.9 In practice, we split the set of
announcements into nominal inflation-related announcements and news that relates to real growth.
Growth data, in turn, come in two flavors – objective realizations of past economic activity and
subjective often forward-looking views derived from surveys which we label “anticipated growth.”
Finally, realized growth can be split one last time into information relating to output versus
employment.
Through this structure, we can potentially obtain two (inflation and growth), three (inflation,
realized growth, and anticipated growth), or four (inflation, output, employment, and anticipated
growth) factors:
• Inflation
• Growth
Realized Growth
Output
Employment
Anticipated Growth
where, for example, the realized growth factor combines information relating to output and employment.
In that sense, the information is nested from right to left.
We use information from different subsets of news to construct the different macroeconomic
factors. For example, we extract the U.S. anticipated growth factor from the news flow generated
by 10 surveys: ABC consumer confidence, Chicago purchasing manager, consumer confidence,
Dallas Fed manufacturing activity, Empire manufacturing survey, leading indicators index, NAPM-
Milwaukee, Philadelphia Fed business outlook survey, Richmond Fed manufacturing index, and the
University of Michigan confidence index. For completeness, the Appendix lists the assignments of
all macroeconomic announcements for the U.S., Europe, U.K., and Japan to the four categories:
inflation, output, employment, and anticipated growth.
It is worth reiterating at this point that we do not include any market-based data (such as stock
prices, interest rates, credit spreads, or implied volatilities) in our analysis, unlike, for example,
Arouba et al. (2009) and Giannone et al. (2008). While such data are very timely and undoubtedly
informative about the state of the economy, they represent already the market’s interpretation of the
macroeconomic news flow. Our aim is to objectively summarize and describe the macroeconomic
news flow itself, so that we can relate the actual state of the economy to expected stock market
returns.
2.4 Transformation and temporal alignment
We examine the stationarity of each data series in two ways. First, we conduct a Dickey-Fuller test
on each series. Second, we read the definition and description of each statistic to determine from an
9The economy is often separated into nominal and real sides because shocks to the two should be treated differentlyfrom a policy perspective. For example, many argue, from the perspective of monetary policy, that nominal shocksshould be minimized, whereas real shocks should not be intervened upon.
7
economic perspective whether it is a non-stationary index or a stationary quarterly growth rate, for
example. In a few cases where the conclusions from the two approaches differ, usually because the
available data is too short to examining statistically, we rely more on the description to determine
whether the series is stationary. All series that are deemed non-stationary are first-differenced in
news release time. The Appendix contains more details.
The final data task is to align the data temporally by moving from announcement time to
calendar time. We do this by populating the news releases in a T×N matrix where T denotes the
total number of week days in our sample and N refers, for example, to the 43 announcement types
for the U.S.. The data at this stage looks like the top panel of Figure 2.
There are two important aspects of the data to discuss. First, there are a vast number of
missing values, as we can think of each news series as a continuously evolving statistic that is
observed only once per month or quarter. Second, not all announcements have a complete history.
Some announcements are initiated in the middle of the sample and/or are terminated before the
end of the sample. To solve the missing data problem, we simply forward fill the last observed
release until the next announcement. Forward filling can be rationalized as replacing missing
values with expected values under a simple independent random walk assumption for each news
series. Of course, both independence in the cross-section and random walk dynamics through
time are simplifying assumptions that are rejected by the data (in fact, the motivation for our
methodology described below is the cross-sectional correlation structure within news category). A
more sophisticated approach for filling in missing data would be to compute the expectation of
the missing values given the full cross-section of previous releases as well as the cross-sectional and
intertemporal correlation structure of the data. An optimal solution would also allow for sampling
error, which is the case in Kalman filter or Bayesian data augmentation algorithms. However, there
is a clear trade-off between statistical complexity and ability to process a large cross-section of news
series. Since the goal of our approach is to utilize the entire cross-section of news, we choose a very
simple statistical model for filling in missing observations. After forward filling, the data looks like
the bottom plot of Figure 2.
Note that the second data issue, the fact that some series do not span the entire sample
period, cannot be solved with missing values imputation. It is instead explicitly addressed in
our methodology below.
3 Methodology
3.1 Subset principal component analysis
Our goal is to extract from the cross-section of macroeconomic news releases a set of factors that
capture in real-time the state of inflation, output, employment, and anticipated growth, as well as
8
the two more overarching factors measuring realized growth and growth. As we already discussed,
the most obvious ways of accomplishing this, full data principal components analysis (PCA) and
forecasting regressions, do not appeal to us. First, with full data PCA we obtain factors that are
mechanically orthogonal, whereas the dimensions of the economic news flow we want to capture are
likely correlated (e.g., output and employment are both high at the peak and low at the trough of
an economic cycle). This orthogonalization makes is practically impossible to assign an economic
meaning to higher order factors. Second, trying to identify the factors through predictive regressions
on candidate variables in each category, such as final GDP for output, would require us being able
to identify a single series that represents each category. While this is a common approach in the
nowcasting literature, it relies on ex-ante knowledge of the key statistic to track and assumes that
there is only one such statistic that does not change over time (see also Stock and Watson, 1989).
Instead, we rely on our ex-ante categorization of the news and, within each category subset,
let the data speak for itself by extracting the first principal component of that subset of data.
Specifically, on each day of our sample t, we obtain for each news category i the first principal
component from the correlation matrix Ωt,i of the stationary news series in category i. We work
with the correlation matrix to abstract from arbitrary scaling of data. Moreover, in order to obtain
a real-time measure, we use a telescoping (meaning, with a common historical start date and rolling
end dates) correlation matrix starting in 1990.10 We denote the Ni×1 principal component weights
by ct,i, where Ni is the number of news series in category i. Consistent with extracting principal
components from a telescoping correlation matrix, we standardize the news series using telescoping
estimates of their means and standard deviations.
3.2 Economic new series correlation matrix
The key inputs to our methodology are the within news category correlation matrices Ωt,i. Specifically,
we need to calculate from historical data up through date t the correlation of all news series of
category i that are “active” on that date, where active means that the news series was previously
initiated and has not yet been terminated. There are two issues that need to be addressed in
computing these correlation matrices. First, the data is in the form of an unbalanced panel due to
some of the series being initiated after the start date of the estimation window (e.g., series j = 5
in Figure 2). Second, the data is naturally persistent, partly due to autocorrelation of the data in
announcement time, partly due to the cross-sectional misalignment of the news in calendar time,
and largely due to the forward filling of missing data.
We address the first unbalanced panel issue by using a correlation matrix estimator along the
lines of Stambaugh (1997), who shows how to adjust first and second moments estimates for unequal
sample lengths. The intuition of his approach is to use the observed data on the longer series, along
10We also experimented with fixed window size rolling correlation matrices for 5, 10, 15, and 20 years. The resultsare qualitatively similar, particularly for the longer data windows.
9
with a projection of the shorter series onto the longer ones estimated when both are observed, to
adjust the moments of the shorter time series.
To correct for the persistence, we could use the standard approach of Newey-West (1987), where
due to the nature of the data we need to account for up to one quarter of autocorrelation and cross-
autocorrelation. Unfortunately, the kind of persistence in our data is not ideally captured by the
non-parametric Newey-West approach for two reasons. First, we have daily data, so adjusting for
up to a quarter of autocorrelation would involve approximately 60 cross-autocorrelation matrices.
Second, the (cross-) autocorrelations are not exponentially decaying as a typical autoregressive
model might predict. Instead, the data is locally constant, due to the forward filling, and over
longer intervals only moderately (cross-) autocorrelated due to the statistical nature of the news
series.
This peculiar correlation structure of economic news forward filled onto a daily calendar is
actually identical to that found in high-frequency asset prices, where asynchronous and infrequent
trading creates a misaligned and locally constant panel of observations. In that literature, Ait-
Sahalia, Mykland, and Zhang (2005) propose a “two-scales realized volatility” estimator to handle
this specific structure of short-term constancy versus long-horizon weak dependence. Specifically,
their estimator subsamples the data at a sufficiently low frequency that overcomes the local constancy
and then averages over the set of all possible estimators that start the subsampling schemes at
different times.
We adopt exactly the same approach, except of course our application is very different. Specifically,
at date t we subsample the forward filled news series backward at a monthly frequency and then
compute a Newey-West estimate of the correlation matrix using four lags. We repeat the same for
monthly sampling starting at dates t − 1, t − 2, ..., t − d + 1 (assuming d days per month) and
then average the resulting d correlation matrix estimates.
3.3 Level versus disagreement factors
Given the vector of principal component weights ct,i obtained with our methodology, we then
construct for each news category two times-series. First, we sum at each date the product of the
weights multiplied by the most recent releases to obtain our real-time level factors. Second, we sum
the product of the same weights multiplied this time by the cross-sectional standard deviations of
the economist forecasts for the most recent releases to obtain our real-time disagreement factors.
Throughout our sample not every news series has economist level forecasts data available, particularly
for the international announcements. We therefore construct the disagreement factor using the
available data, re-normalizing first the principal component weights to account for the proportion
of missing data.
10
4 Results
We first plot the data to examine their joint behavior and as potential preliminary evidence for our
main results. We then predict U.S. stock market returns at different horizons with regressions on
realized growth and orthogonalized anticipated growth, unconditionally and conditional on periods
of high economic uncertainty proxied by large dispersion of economist forecasts. This analysis is
extended to the international data, where we examine the role of local versus global factors as well
as quantify the economic significance of our findings through a portfolio choice problem. Finally,
we confirm that our measures of realized and anticipated growth indeed forecast future realized
growth in the U.S. and internationally.
4.1 Preliminaries
We begin our graphical description of the data in Figure 3, where we plot the U.S. real-time growth
index extracted from growth related (both realized and anticipating) economic news in the upper
panel and the corresponding economic uncertainty measure obtained from economist disagreement
about growth-related news releases in the lower panel. The vertical shaded regions in this and
all of our graphs are ex-post dated NBER recessions. The growth index dips sharply throughout
both recessions, in 2001 and 2008-2009, and recovers quickly afterward. The economic uncertainty
surrounding growth exhibits peaks of disagreement in the final stages of the recession periods,
while uncertainty is well below average at the beginning of recessions. This pattern suggests that
economists tend to disagree much more about the economy exiting a recessions then entering it.11
In Figure 4, we break out the growth index to its two components, namely the realized growth
index and the anticipated growth index. Recall that both sub-factors are constructed entirely
from macroeconomic news releases without the use of financial market data. The realized growth
index relies on objective realizations of past economic activity such as employment and output
figures, whereas the anticipated growth index is based on subjective forward-looking consumer and
manufacturer surveys on the perceived state of the economy. It is clear from the graph that the
two factors are highly correlated, as one would expect, but we also notice intriguing divergences.
More specifically, anticipated growth appears to be leading realized growth around turning points
of the business cycle. This finding is consistent with anticipated growth being forward-looking,
but at the same time perhaps surprising given the absence of financial market information. The
difference between realized and anticipated economic growth will become an important feature of
our empirical analysis in the next sections.
11Note that the dispersion of growth forecasts is also relatively large and noisy at the beginning of our sample.While there might have been a higher degree of economic uncertainty at that time, it is more likely that this patternis due to the small number of news releases for which economist forecasts were available in the first year of our sample.Out of the 34 variables used to construct the growth index, only 11 had forecasts reported on Bloomberg in 1997 andfor those releases only an average of four economists were providing their forecasts.
11
We present summary statistics and correlations of U.S. equity market returns, our real-time
macroeconomic indices, and the other return predictors in Table 1. We first notice that both the
realized growth and the anticipated growth factors exhibit negative skewness, consistent with the
sharp drops during recessions exhibited in Figure 4, whereas dispersion, our proxy for macroeconomic
uncertainty, is quite naturally positively skewed. Similar positive skewness is observed in the
variance risk premium and default premium, both of which are known to spike in economic
downturns.
Turning to the correlation matrix, the aggregated growth index and its two sub-factors, realized
and anticipated growth, are all highly correlated. The correlation is higher between the aggregated
index and its sub-indices (0.97 and 0.93) than between the two sub-indices themselves (0.82)
suggesting the potential for some differences in information in realized versus anticipated growth.
All three indices are largely uncorrelated with our measure of economic uncertainty, with correlations
ranging from 0.05 to -0.23. Economist disagreement about growth-related data clearly captures
different information then the data releases themselves.
The correlations of our macroeconomic indices with the usual stock market predictors are also
interesting. The growth index exhibits strong negative correlations with the dividend yield, default
spread and term spread, all of which are countercyclical predictors, with values ranging from -0.51 to
-0.84. The growth index is equally strongly but positively correlated with the price-earnings ratio,
which in turn has a correlation of -0.85 with the dividend yield. Comparing the correlations across
the two sub-indices, the correlation is substantially stronger for anticipated growth in the case of the
price earnings ratio (0.65 versus 0.46) and for realized growth in the case of the term spread (-0.57
versus -0.38). The correlations are very similar with the dividend yield and default spread. Finally,
we observe strong negative correlations of our indices with the variance risk premium confirming
that this predictor is also countercyclical.
We complete our description of the data by plotting in Figure 5 the various real-time indices
against the S&P 500 index. The purpose of this preliminary analysis is to correlate the indices
with equities contemporaneously in levels, to detect common business cycle variation, in contrast
to our main analysis below in which we compute lead-lag correlations in first differences (i.e., do
changes in economic conditions forecast stock returns?). The top left chart is for the real-time
growth index which combines realized and anticipating information. Unconditionally, real-time
growth and the equity market are only mildly correlated, with a contemporaneous correlation of
0.36. During recessions, however, the two series are clearly much more strongly related, moving
nearly in locksteps. A very similar behavior is observed for the realized growth factors in the top
right plot, although the unconditional correlation is somewhat higher at 0.45.
Rather than working with two very highly correlated series, realized and anticipated growth,
we orthogonalize the later to the former using rolling sample regressions in order to focus on
12
the forward looking information contained in anticipated growth that is uncorrelated with past
economic activity. Specifically, for each day in our sample we regress anticipated growth on the
realized growth using the previous five years of data. We then take the regression residual for that
day as the realization of a new “orthogonal anticipated growth” index. For the remainder of this
paper, we break up the aggregate real-time growth factor into the realized growth factor and this
orthogonal anticipated growth factor.
The bottom left chart of Figure 5 shows the relation between orthogonal anticipated growth
and the equity market. These two series are contemporaneously negatively correlated at -0.41,
suggesting a contrarian aspect to the forward looking information. This negative correlation also
explains why the aggregate growth index that incorporates all growth related data is less correlated
with equities than the realized growth index that excludes the information of the orthogonal
anticipated growth index. Finally, in the bottom right plot of the figure, our measure of economic
uncertainty, namely the disagreement across economists about growth-related data, is even more
negatively related to the equity market, with a correlation of -0.50. This result is consistent with
our earlier observation that forecast dispersion tends to be low at the peaks of economic expansions
and high at the troughs of recessions.
4.2 U.S. return predictability
4.2.1 Forecasting regressions
We now turn to the main objective of the paper, namely to examine whether the risk premium on
stocks covaries with changes in realized and anticipated economic conditions as measured in real
time by our macroeconomic news indices. For this purpose, we estimate the following regressions
for return horizons, denoted lead below, ranging from one week to six months:
Rm;t,t+lead −Rf ;t,t+lead = α+ βXt + εt, (3)
where Rm;t,t+lead −Rf,t,t+lead denotes the log return on the S&P 500 index in excess of Libor from
date t to t + lead. Xt is either the monthly first difference of the realized growth index, the level
of the orthogonalized anticipated growth index, or both. We first-difference the realized growth
index to overcome the extreme degree of persistence in levels (a monthly autocorrelation of 0.96).
In contrast, the orthogonalized anticipated growth index is far less persistent because, intuitively,
it is already a difference between two highly persistent time series.
The results in Table 2 show strong evidence of return predictability by both predictors individually
as well as jointly. We report in the table point estimates and t-statistics for the slope coefficient
β along with adjusted R2 of the regressions.12 Our inferences are based on Newey-West robust
12We include an intercept α in the regression but do not report the estimates in the table to save space.
13
standard errors that correct for the autocorrelation induced by overlapping observations.13 In
Panel A, which focuses on the univariate predictability of the realized growth index, we observe
statistically significant slope coefficients at horizons of one to four months, with the strongest
predictability for monthly returns (a t-statistic of 3.8 and an adjusted R2 of 2.2 percent). In terms
of magnitude, the coefficient at the monthly horizon implies that a one standard deviation change
in the realized growth index, which equates to a change of 0.36 in the level series, is associated with
an additional 75 basis point excess return per month.
The evidence for predictability is even stronger in Panel B, which shows the univariate results
for the orthogonalized anticipated growth index. Orthogonalized anticipated growth predicts future
stock market returns significantly at all horizons starting at a month. Not only are the t-statistics
and adjusted R2 substantially larger than in Panel A, but the predictability also peaks at a much
longer horizon of five to six months suggesting, consistent with the orthogonalization, that the
results in panels A and B are different phenomena. For example, at the quarterly horizon the
slope coefficient has a t-statistic of 4.1 with an adjusted R2 of 6.2 percent. The point estimate
implies that if orthogonalized anticipated growth is one standard deviation (which is 0.68) above
average, the one-quarter ahead expected excess return is 2.2 percent higher. Both statistically and
in magnitude, the divergence between ex-ante consumer and manufacturer expectations of economic
conditions and ex-post realized macroeconomic metrics is an unusually strong predictor of future
stock market returns.
Given the orthogonalization of anticipated growth, the first differencing of realized growth, and
the strikingly different horizon patters of the univariate results, we would expect the two predictors
to be additive in a multivariate specification. That is precisely what we observe in Panel C of
Table 2, which shows the results for bivariate regressions. Neither set of coefficients is changed
substantially, both in terms of magnitude and statistical significance, relative to the univariate
regressions. Moreover, the adjusted R2 is increased across horizons. For example, the bivariate
regression at the one-month horizon has a 20 percent higher adjusted R2 than the univariate
regression on realized growth alone in Panel A. The improvement is even greater (one-third) for
the quarterly horizon compared to the univariate results in Panel B.
There are a number of variables that have been shown to predict stock returns, some of which
are, according to Table 1, fairly highly correlated with our real-time macroeconomic indices. This
begs the question of whether the predictability documented in Table 2 is truly a new finding. To
address this question, we simply include other commonly used predictors in our forecasting model.
13As a robustness check we also estimated the forecasting model with quarterly non-overlapping observations.Even with such severely reduced sample size and the related loss in information, we obtain a statistically significantrelation between the real-time macroeconomic news factors and future stock returns.
14
Specifically, we estimate the following regression:
Rm;t,t+lead −Rf ;t,t+lead = α+ βXt + γZt + εt, (4)
where, as above, Rm;t,t+lead − Rf,t,t+lead denotes the log return on the S&P 500 index in excess
of Libor from date t to t + lead. Xt contains the monthly first difference of the realized growth
index and the level of the orthogonalized anticipated growth index. Added to this specification is
Zt, a set of other predictors commonly used in the return forecasting literature. Specifically, we
include the variance risk premium V RP defined as the difference between option implied variance
and expected variance obtained through the time-series model in equation (1), the log price-earning
ratio, the log dividend yield, the default spread defined as the difference between Moody’s BAA
and AAA corporate bond yields, and the term spread defined as the difference between ten-year
and three-month Treasury yields.
Table 3 presents the results for monthly and quarterly forecast horizons.14 First notice that in
the univariate regressions with the other predictors, specifications (2) through (6), only the price-
earnings ratio and the dividend yield pass the usual statistical significance hurdle with Newey-West
standard errors and daily observation frequency. Both the default spread and the term spread are
not only insignificant but the coefficients also have the wrong sign relative to the typical results
in the literature (e.g., Fama and French, 1989). Most likely this is due to the changing dynamics
of interest rates during and after the financial crisis. The default spread series is overwhelmed
by an unusually large spike in high-yield rates that peaked in the last quarter of 2008 and lasted
through the first quarter of 2009. The term spread, in turn, was inverted in 2007 as it tends to
be before recessions but then did not increase as much as it has historically toward the end of the
recession because quantitative easing by the Federal Reserve depressed long-term yields through
the end of the sample. Finally, the coefficient on the variance risk premium has the right sign but
is insignificant, in contrast to the results of Bollerslev et al. (2009) who used a monthly observation
frequency, a shorter pre-financial crisis sample, and a different approach to obtain the expectation
of realized variance. It remains an open question whether the financial crisis has permanently
changed the forecasting power of the default spread, term spread, and variance risk premium.
In specifications (7) through (9) in Table 3, we check whether any of these other predictors drive
out the explanatory power of the first difference of realized growth or the orthogonalized anticipated
economic growth. We find that this is never the case. If anything, including some of the other
predictors increases the explanatory power and improves the statistical significance of our real-time
macroeconomic indices. For example, including the price-earnings ratio as an additional regressor
increases the explanatory power of the model by more than 40 percent for both horizons. The
same is true when we include instead the dividend yield, which we do not report in the table given
14The results for the other horizons are qualitatively the same.
15
how highly (negatively) correlated the price-earnings ratio and dividend yield are. In summary,
the evidence in Table 3 shows that the U.S. macroeconomic predictors constructed from the flow
of scheduled macroeconomic announcements have genuine forecasting power that is not subsumed
by any of the predictors traditionally used in the literature.15
4.2.2 Conditional predictability
Both realized and anticipated macroeconomic conditions, as measured by our real-time indices,
forecast the equity premium unconditionally. We now try to identify environments in which
economic fundamentals are particularly important. Specifically, we focus on two ways of conditioning
the predictability results. First, we sort observations into periods of high and low disagreement
among economists about the growth index. Intuitively, when disagreement is high, the news flow
captured by our indices is more informative. Second, we separately analyze periods of low and
high economic growth. When growth is low, marginal utility is likely higher, and a given change in
economic fundamentals has a greater impact on investors. In both cases, we sort observations into
subsamples associated either with high/low disagreement or with low/high growth and re-estimate
the forecasting regressions for these subsamples.
In panels A and B of Table 4 we present results conditioning on the dispersion of economist
forecasts about the news series that comprise our growth index being above or below the full-sample
median, respectively. Recall from footnote 11 that the time-series of growth disagreement is very
noisy at the beginning of our sample because detailed economist forecasts are available only for a
small subset of releases. We therefore use January 2000 instead of January 1997 as the sample start
date for these regressions (although including the more noisy data does not substantially change
our results). We notice a striking pattern. Predictability is substantially stronger, both statistically
and economically, when economists disagree more about growth prospects. For example, at the
quarterly horizon, the explanatory power is about 15 times larger when macroeconomic uncertainty
is high, and the coefficients on the real-time indices are at least four times larger. In contrast, when
growth dispersion is below its median, only orthogonalized anticipated growth is significant and
only at short horizons.
In panels C and D we instead condition the analysis on the overall growth index being negative
or positive, respectively. Since the individual news series that comprise our real-time indices are
standardized, this roughly translates to periods of below or above average economic fundamentals.
Contrasting the results in the two panels shows clearly that during times of weak growth the equity
risk premium covaries considerably more with macroeconomic information than during times of
15We also considered specifications in which we control for stock market momentum (or reversals) by includinglagged returns over the same horizon as we are predicting forward. Our findings are virtually unchanged and thereforenot reported. If anything, controlling for lagged returns at the shortest 5-day horizon helps to better identify the roleof macroeconomic fundamentals, resulting in more significant coefficients.
16
strong growth. For example, at the quarterly forecast horizon the R2 is about 17 percent when the
growth index is negative (in Panel C), compared to an unconditional eight percent in Table 2 and
less than one percent when the growth index is positive (in Panel D). These results are consistent
with our conjecture that economic fundamentals are more important to investors during recessions
when marginal utility is more likely to be high.
The two sets of findings (panels A and B versus C and D) are consistent with economist
forecast dispersion peaking in recessions, as Figure 3 indicates graphically. This raises the question
of whether the two sets of results are perhaps the same phenomenon. Double sorting the data
based on dispersion and growth suggests that this is not the case. For example, unconditionally
the bivariate specification has an R2 of 2.6 percent at the monthly horizon. Conditioning on
greater disagreement or on being in a recession raises the R2 to 6.5 and 5.1 percent, respectively.
Conditioning on both, further raises it to 10.3 percent. Unfortunately, the number of observations
in the counter-factual high dispersion and expansion or low dispersion and recession states is too
small (about 10 percent of the data) to draw reliable conclusions. We therefore do not include the
double-sorted results in Table 4.
4.3 International return predictability
We now turn to the cross-section of global equity markets to see whether the set of results we
uncovered in U.S. data is a broader phenomenon. We first plot the data in Figure 6. The three
graphs show the total growth indices, realized growth indices, and anticipated growth indices for
the U.S., Eurozone, United Kingdom, and Japan, respectively. For all three indices, we observe
a strong degree of co-movement across countries. This is particularly the case for realized growth
and least so for anticipated growth.16 Because of this commonality, we consider below regression
specifications with local macroeconomic news indices, global indices, and the difference between
local and global indices. We defined the global indices as the first principal component of their
local counterparts based on a full-sample covariance matrix. We also consider as robustness check
the U.S. macroeconomic news indices as a proxy for global economic conditions.
4.3.1 Forecasting regressions
Table 5 shows the results from running the forecasting regression (3) for the U.S. (S&P 500) in
Panel A, Euro zone (EURO STOXX 50) in Panel B, U.K. (FTSE 100) in Panel C, and Japan
(Nikkei 225) in Panel D with the predictors Xt being the one-month change in the realized growth
index and the level of the orthogonalized anticipated growth index (as in Panel C of Table 2). Each
panel has three sub-panels corresponding to the use of local indices extracted from country-specific
16However, even with this strong cross-sectional correlation, we still observe country-level growth variation thatcorresponds to well-known idiosyncratic events. For example, notice the drop in Japanese realized growth in 2011triggered by the Tsunami that hit the country in March of that year.
17
news, global indices obtained through principal component analysis of their local counterparts, and
U.S. indices as proxies for global ones. In each sub-panel we then report results for return horizons
of 20, 40, 60 and 80 days.17
Panel A shows the results for the U.S., where obviously the local indices coincide with the global
ones proxied for by U.S. data. In this case, the local predictors tend to perform overall better than
the global ones, especially for the monthly change in realized growth, suggesting that for U.S. equity
markets the domestic macroeconomic environment is more relevant than global conditions. This
finding stands in stark contrast to the results for Europe, the U.K., and Japan for which the global
and particularly the U.S. data appears to be more important.
Consider first the results for the U.K. in Panel C. With the local predictors, only the orthogonalized
anticipated growth index is marginally significant, and the R2 are relatively small. Using global
predictors increases the significance and explanatory power of orthogonalized anticipated growth
considerably. Using instead the U.S. indices as proxies for global ones produces results that are
remarkably close to the U.S. findings in Panel A. The results for Japan in Panel D are very similar.
When local indices are used, only orthogonalized anticipated growth is significant (the change
is realized growth has insignificant and negative coefficients) and the R2 are low. Switching to
global predictors increases the significance and degree of predictability. The change in realized
growth is only significant when we use U.S. data. Finally, the results for Europe in Panel B are
a little different in that in this case the strongest predictability is found for the local predictors
and by the orthogonalized anticipated growth index alone. The change in realized growth is again
only significant when we use U.S. predictors, but the R2 are lower than with the local predictors.
We conclude from these results that the predictability we uncovered in U.S. data appears robust
globally, but that particularly for smaller economies the predictability is stronger using global and
even just U.S. predictors than local ones.18
In order to make sure that our international findings are not subsumed by the traditional
predictors used in the literature, we also estimate regression specifications that include either
local or U.S. versions of the alternative forecasting variables considered in Table 3. While in
most cases the explanatory power of the regression is enhanced by including other predictors, the
statistical significance and magnitude of the coefficients on our real-time macroeconomic indices is
never substantially changed, suggesting, consistent with our findings for the U.S., that the global
phenomenon we uncovered is in fact a new one. Given how similar the results are to the univariate
specifications in Table 5, we do not tabulate them here to save space.
17We obtain comparable results for the three horizons we omit, relative to Table 2, to save space.18In unreported follow-up work, we estimate all four sets of regressions in a fixed-effect panel framework in order
to understand whether the predictability obtained in the time-series for each country in isolation carries over to ajoint cross-section in which the coefficients are constrained to be the same across countries. The results are stronglysupportive. We find that particularly the U.S. real-time macroeconomic news indices forecast future equity marketreturns across all horizons and countries, with economic magnitudes that are statistically indistinguishable from thecountry-specific regressions.
18
Given the differences in the results for local versus global predictors across countries, we estimate
a richer specification in which we include both sets of predictors. Specifically, we estimate a model
with two set of predictors, like equation (4). Xt is the one-month difference in the global realized
growth index and the orthogonalized global anticipated growth index. Zt includes the differences
between the corresponding local predictors and their global counterparts. We specify the regression
in terms of global predictors and differences between local and global predictors, as opposed to just
global and local predictors outright, to overcome the high correlation between the latter. We
present results using U.S. information to proxy for global economic conditions, but the results are
very similar when instead we use the first principal component of the local indices as measure of
global economic activity.
Table 6 presents the results. The first four columns of each panel correspond to single-country
regressions, and the final column is for a pooled specification estimated with country fixed-effects.
In panels A and B the forecast horizon is one month and one quarter, respectively. Focusing
first on the monthly horizon in Panel A, the change in global growth is generally significant across
countries with similar magnitude positive coefficients, and it is particularly significant in the pooled
specification. The difference between local and global growth, in contrast, is insignificant across the
board. In three of the four specifications the coefficient has a negative sign, which means that if a
country grows faster than the rest of the countries, its equity market is likely to underperform the
following month, perhaps because the growth gap reverts. The results are somewhat different for
orthogonalized anticipated growth. The global predictor has again very similar magnitude positive
coefficients across countries and is significant for the U.S., Japan, and pooled regressions. However,
in this case, the difference between the local and global indices also has positive coefficients that
are significant for Europe as well as in the pooled sample. A positive coefficient means that if
a country has stronger anticipated growth than the rest of the countries, its equity market is
expected to outperform the following month, consistent with the anticipation coming true. The
results for the quarterly horizon in Panel B are qualitatively the same. The statistical significance
is generally weaker for the change in global realized growth and stronger for the orthogonalized
global anticipated growth, consistent with the horizon pattern in the results in Tables 2 and 5.
4.3.2 Conditional predictability
Recall that our empirical analysis for the U.S. uncovered a strong conditional pattern in the results,
where return predictability is considerably stronger during periods of above median macroeconomic
uncertainty or during periods of negative growth. Table 7 considers the same conditioning for the
international analysis. As above, we use U.S. data to proxy for the world. In panels A and B we
condition on high or low macroeconomic uncertainty as measured by the dispersion of U.S. growth
forecasts being above or below the median, respectively. We can readily observe much stronger
predictability when economists disagree more. For example, with a quarterly return horizon, the
19
explanatory power of the regressions in Panel A is about six times larger than in Panel B for
the Eurozone and the U.K., and almost 20 times larger for Japan. Both regressors are statistical
significant at all horizons for all countries in Panel A, but only orthogonal U.S. anticipated growth
is sometimes marginally significant in Panel B.
In panels C and D we condition instead on negative or positive U.S. growth, respectively. In this
case, we observe much stronger predictability during recessions, just as we did in the U.S. analysis.
For example, focusing again on the quarterly return horizon, the explanatory power in Panel C is
about 16 times larger than in Panel D for the Eurozone, about ten times larger for the U.K., and
almost three times larger for Japan. Both regressors are statistical significant at all horizons for all
countries in Panel C, with only a few exceptions. Interestingly, during U.S. expansions in Panel D,
U.S. orthogonal anticipated growth predicts significantly negative excess returns for the U.K. and
the Japanese stock market – that is, the sign of the coefficients is flipped during expansions relative
to recessions and the unconditional case.
4.4 Economic significance
We conclude from the previous subsections that the predictability we uncovered in the U.S. data is
corroborated internationally and hence is less likely to be the outcome of data mining. Having thus
far focused on statistical significance, we now measure the economic significance of our results. We
do so by embedding three key findings in a global tactical asset allocation problem. The three key
findings we focus on are: (1) both the change in realized growth and orthogonalized anticipated
growth predict future stock market returns at different horizons, with the most significant results for
quarterly returns, (2) international predictability is stronger when we use global or U.S. predictors,
as opposed to local factors, as some countries appear more sensitive to global macroeconomic
developments than domestic ones, and (3) there is some evidence that cross-country deviations of
changes in realized growth from the global average mean-revert while orthogonalized anticipated
growth divergences persist. Combining these findings, our portfolio choice problem features both
a directional choice of whether to be long or short global equities and a cross-sectional choice of
which countries to buy or sell.
For the global market timing part of the problem, we construct a balanced global equity index
as the volatility weighted average of the four international equity indices. We normalize this global
equity index to have a constant unit risk through time, where both the return weighting and the
subsequent normalization are done with trailing three-month standard deviations. We construct
global real-time macroeconomic news factors as averages of the corresponding local factors.19 Each
day in our sample, the portfolio is long (short) one unit of the global equity index when the current
global realized growth factor is above (below) its 60-day moving average. In addition, the portfolio
19Very similar results are obtained when we use the U.S. indices as proxies for the global ones.
20
is long (short) one unit of the global equity index when global anticipated growth exceeds (falls
short of) global realized growth. We chose these two simple rules to mimic the predictors in our
return regressions in a way that is intuitive and easy to compute in real time (i.e., does not require
a separate orthogonalization regression).
For the country selection part, the portfolio is long (short) one unit of risk each in the two
equity indices with highest (lowest) pickup in local realized growth. We do the opposite for the
differences between local anticipated and realized growth, buying (selling) the two indices for which
the anticipated to realized growth gap is the largest (smallest). These two rules capture the apparent
mean-reversion of realized growth divergences and continuation of anticipated growth divergences,
respectively.
In Figure 10 we plot the cumulative returns on this global tactical asset allocation portfolio as
well as on the global market index over our sample period. We scale the passive market returns to
have the same volatility of the active portfolio returns that we construct in order to facilitate a fair
comparison between the two. The macroeconomic news based market timing and country selection
strategy delivers a stable and strong performance that clearly outperforms the global market index
for the same level of risk taken. The active portfolio has a Sharpe Ratio of 0.69 as compared to a
ratio close to zero for the global market. The outperformance is particularly strong during the two
recessions, but it is also present during the other periods. This brief portfolio application clearly
illustrates that the predictability we uncovered is economically significant.
4.5 Macroeconomic news-based predictors versus economic growth
One obvious question is to what extent our results are driven by the new predictors we propose as
opposed to the different regression specifications, relative to what has previously been considered in
the literature. To address this question, we repeat a subset of the analyses in tables 2 and 5 using
quarterly headline GDP figures instead of our realized growth indices.20 For our sample period,
the headline GDP data consists of 60 quarterly observations for each of the four countries. In line
with the absence of similar results in the literature, we find that there is no significant predictive
relationship between changes in quarterly GDP growth and stock returns. More specifically, for
the four countries and five forecasting horizons (ranging from five to 80 days ahead) considered,
20 regression coefficients are insignificant, nine coefficients are significantly positive, and 11 are
significantly negative. The explanatory power of changes in headline GDP growth varies between
zero and three percent. It appears that the results above rely critically on our use of real-time
macroeconomic news indices.
One possible explanation for why the results are so different is that the information contained
in the two predictors, the one-month change in realized growth and the level of orthogonalized
20We use the first GDP release for a given quarter, which for the U.S. corresponds to the preliminary release.
21
anticipated growth, contain information about future growth, as proxied by headline GDP or more
generally by our real-time growth index. To investigate this possibility, we estimate the following
predictive regression specification:
Yt+lead = α+ ρYt + βXt + εt, (5)
where Yt denotes the real-time U.S. growth index (which aggregates realized and anticipated
information) and Xt contains the monthly first difference of the realized growth index and the
level of the orthogonalized anticipated growth index. To control for the persistence in the growth
index, we include the lagged level of growth as a predictor.
Panel A of Table 8 presents the results. Jointly the one-month change in realized growth and
the level of orthogonalized anticipated growth explain between four and 27 percent of the variation
of future growth left unexplained by the current level of growth.21 The change in realized growth is
significant at horizons up to a quarter. The level of orthogonalized anticipated growth is significant
at all horizons but, as with the change in realized growth, the explanatory power is considerably
stronger at shorter horizons with t-statistics above five. We conclude from these results that our two
predictors contain information about future economic growth that is not contained in the current
level of growth. This potentially explains why our predictors are more strongly related to future
stock returns.
In panels B and C we repeat this analysis but with samples that are conditioned on the
dispersion of economist forecasts about the growth index being above or below the full-sample
median, respectively. The purpose is to see whether the same conditional predictability pattern
exists for growth as we observed for stock returns. Indeed, growth predictability is much stronger
in periods of high macroeconomic uncertainty. For example, at the quarterly horizon, the marginal
explanatory power of the two regressors is twice as large in Panel B than it is in Panel C. Both
regressors are always significant when growth forecast dispersion is high, but only orthogonalized
anticipated growth is significant in periods of low forecast dispersion. The fact that the growth
predictability results exhibit the same conditional pattern as the stock return predictability results
further strengthens the case that the two may be related.
5 Conclusions
We constructed daily measures of economic activity based on the almost continuous flow of macroeconomic
data releases. Our measures summarize in real-time the public information about the economy.
We focused on information about growth and explicitly differentiated between (i) realized ex-post
21Table 8 presents regular adjusted R2s as well as ones computed net of the AR(1) dynamics of the growth index.The latter measure the fractions of the AR(1) residual variance explained by the exogenous predictors Xt.
22
measures, such as quarterly GDP releases, and (ii) anticipating ex-ante information, namely data
from surveys of consumers and firm managers. The resulting economic indices overcome some of the
concerns about using economic data for stock return predictability: the indices are measured daily
and capture the information of the entire news flow, they are both backward- and forward-looking,
and they are based on carefully dated and unrestated data. We used these real-time economic
indices to reexamine whether and, if so, to what extent stock market returns are predictable by
economic fundamentals.
We found that the one-month change in realized growth predicts U.S. stock market returns one
to four months into the future. In addition, the part of anticipated growth that is orthogonal to
realized growth predicts market returns at horizons of two to six months. Due to the orthogonal
construction of the two predictors, the results are roughly additive in a multivariate regression,
leading to return predictability that far exceeds what is generated by the usual suspect predictors.
We then showed that the predictability we uncovered is state dependent. The results are much
stronger during times of greater disagreement among economic forecasters about future growth as
well as during recessions.
We extended the analysis in a number of ways. First, we confirmed it internationally. We found
that economic fundamentals also predict equity returns in Europe, Japan and the U.K.. More
interestingly, at least for Japan and the U.K., global aggregates and even just the U.S. factors
are more predictive than local versions. For Europe, both are relevant. We also found that the
international data corroborates our results on conditioning. In all three regions, the forecasting
ability of realized and orthogonal anticipated growth is much stronger during times of greater
economist disagreement and recessions. To quantify the economic significance of these international
results, we then constructed a global equity market timing and country selection portfolio based
on realized and anticipated global growth. The Sharpe ratio of this investment strategy is 0.7
annualized, as compared to a Sharpe ratio close to zero for being long world equities over the same
time period. Finally, we showed that the return predictors are informative about future economic
growth and that this more fundamental predictability shares the same conditional pattern as the
stock return predictability we uncovered.
23
References
Ait-Sahalia, Yacine, Per A. Mykland, Lan Zhang, 2005, How often to sample a continuous-timeprocess in the presence of market microstructure noise, Review of Financial Studies 18, 351–416.
Ang, Andrew, and Geert Bekaert, 2007, Stock return predictability: Is it there?, Review of FinancialStudies 20, 651–707.
Aruoba, S. Boragan, Francis X. Diebold, Chiara Scotti, 2009, Real-time measurement of businessconditions, Journal of Business and Economic Statistics 27, 417–427.
Bai, Jennie, 2010, Equity premium predictions with adaptive macro indexes, Working Paper,Federal Reserve Bank of New York.
Banbura, Marta, Giannone, Domenico, Modugno, Michele and, Lucrezia Reichlin, 2012, Now-casting and the real-time data flow, CEPR Discussion Papers 9112.
Beber, Alessandro, Michael W. Brandt, and Maurizio Luisi, 2014, Distilling the macroeconomicnews flow, Journal of Financial Economics, forthcoming.
Bollerslev, Tim, George Tauchen, and Hao Zhou, 2009, Expected stock returns and variance riskpremia, Review of Financial Studies 22, 4463–4492.
Bollerslev, Tim, James Marrone, Lai Xu, and Hao Zhou, 2012, Stock return predictability andvariance risk premia: Statistical inference and international evidence, Working Paper, DukeUniversity.
Britten-Jones, Mark, and Anthony, Neuberger, 2000, Option Prices, Implied Price Processes, andStochastic Volatility, Journal of Finance 55, 839–866.
Campbell, John Y., 1987, Stock returns and the term structure, Journal of Financial Economics18, 373–399.
Campbell, John Y., and Robert J. Shiller, 1988a, The dividend price ratio and expectations offuture dividends and discount factors, Review of Financial Studies 1, 195–228.
Campbell, John Y. and Robery J. Shiller, 1988b, Stock prices, earnings, and expected dividends,Journal of Finance 43, 661–676.
Cochrane, John H., 2008, The dog that did not bark: A defense of return predictability, Review ofFinancialStudies 21, 1533–1575.
Corsi, Fulvio, 2009, A Simple Approximate Long-Memory Model of Realized Volatility, Journal ofFinancial Econometrics 7, 174–196.
Corsi Fulvio, Nicola Fusari, and Davide La Vecchia, 2013, Realizing smiles: Options pricing withrealized volatility, Journal of Financial Economics 107, 284–304.
Drechsler, Itamar, and Amir Yaron, 2011, What’s vol got to do with it, Review of Financial Studies24, 1–45.
24
Evans, Martin, 2005, Where are we now?: Real-time estimates of the macro cconomy, TheInternational Journal of Central Banking.
Fama, Eugene F., and Kenneth R. French, 1988, Dividend yields and expected stock returns,Journal of Financial Economics 22, 3–25.
Fama, Eugene F., and Kenneth R. French, 1989, Business conditions and expected returns on stocksand bonds, Journal of Financial Economics 25, 23–49.
Fama, Eugene F., and G. William Schwert, 1977, Asset returns and inflation, Journal of FinancialEconomics 5, 115–146.
Ferson, Wayne E., and Campbell R. Harvey, (1991), The variation of economic risk premiums,Journal of Political Economy 99, 385–415.
Ghysels, Eric, Casidhe Horan, and Emanuel Moench, 2012, Forecasting through the rear-viewmirror: Data revisions and bond return predictability, Working Paper, University of North-Carolina.
Giannone, Domenico, Lucrezia Reichlin, and David H. Small, 2008, Nowcasting: The real timeinformational content of macroeconomic data releases, Journal of Monetary Economics 55, 665–676.
Keim, Donald B., and Robert F. Stambaugh, 1986, Predicting returns in the stock and bondmarkets, Journal of Financial Economics 17, 357–390.
Koenig, E., S. Dolmas, and J. Piger, 2003, The use and abuse of real-time data in economicforecasting, Review of Economics and Statistics 85, 618–628.
Lamont, Owen, 1998, Earnings and expected returns, Journal of Finance 53, 1563–1587.
Lettau, Martin and Sydney Ludvigson, 2001, Consumption, aggregate wealth, and expected stockreturns, Journal of Finance 56, 815–850.
Ludvigson Sydney, and Serena Ng, 2009, Macro factors in bond risk premia, Review of Financialstudies 22, 5027–5067.
Mueller, Philippe, Andrea Vedolin, and Yu-Min Yen, 2012, Bond Variance Risk Premia, WorkingPaper, London School of Economics.
Newey, Whitney K., and Kenneth D. West, 1987, A simple, positive semi-definite, heteroskedasticityand autocorrelation consistent covariance matrix, Econometrica 55, 703–708.
Schwert, G. William, 1989, Why does stock market volatility change over time?, Journal of Finance44, 1115–1153.
Stambaugh, Robert F., 1997, Analyzing investments whose histories differ in length, Journal ofFinancial Economics 45, 285–331.
Stock, John H., and Mark W. Watson, 1989, New indexes of coincident and leading economicindicators, in O.J. Blanchard and S. Fischer (eds.), NBER Macroeconomics Annual, 352–394.
25
6 Appendix: Macroeconomic announcements
This appendix summarizes the categorization and transformation of the macroeconomic news
releases for each country. Category is either inflation (INF), employment (EMP), output (OUT),
or anticipated growth (ANT). If the series is stationary in our sample, we make no adjustment
(Adj.=0), otherwise we use first differences with respect to the previous period (Adj.=1) or previous
year (Adj.=12). We also indicate Units, Frequency (M for monthly, W for weekly, or Q for
quarterly), and the source of the release.
U.S. Releases
Cat. Release Name Adj. Units Freq Source
INF U.S. Import Price Index by End Use All MoM 0 Rate M Bureau Labor Statistics
INF U.S. PPI Finished Goods Total MoM 0 Rate M Bureau Labor Statistics
INF U.S. PPI Finished Goods Except Foods Energy 0 Rate M Bureau Labor Statistics
INF US CPI Urban Consumers MoM 0 Rate M Bureau Labor Statistics
INF US CPI Urban Consumers Less Food Energy 0 Rate M Bureau Labor Statistics
INF BLS Employment Cost Civilian Workers QoQ 0 Rate Q Bureau Labor Statistics
INF US GDP Price Index QoQ SAAR 0 Rate Q Bureau Economic Analysis
INF US Personal Cons. Expenditure Core Price Index MoM 0 Rate M Bureau Economic Analysis
INF US Output Per Hour Nonfarm Business Sector QoQ 0 Rate Q Bureau Labor Statistics
EMP ADP National Employment Report Private Nonfarm Change 0 Volume M Automatic Data Processing
EMP US Initial Jobless Claims 1 Volume W Department of Labor
EMP US Continuing Jobless Claims 1 Volume W Department of Labor
EMP US Employees on Nonfarm Payrolls Total Net Change 0 Value M Bureau Labor Statistics
EMP US Employees on Nonfarm Payrolls Manufact Net Change 0 Value M Bureau Labor Statistics
EMP US Unemployment Rate Total in Labor Force 1 Rate M Bureau Labor Statistics
EMP US Average Weekly Hours All Total Private 1 Volume M Bureau Labor Statistics
OUT ISM Manufacturing PMI 0 Value M Institute Supply Management
OUT US Manufacturers New Orders Total MoM 0 Rate M U.S. Census Bureau
OUT US Auto Sales Domestic Vehicles 1 Volume M BLOOMBERG
OUT ISM Non-Manufacturing NMI NSA 0 Value M Institute Supply Management
OUT Federal Reserve Consumer Credit Net Change 1 Value M Federal Reserve
OUT Merchant Wholesalers Inventories Change 0 Rate M U.S. Census Bureau
OUT Adjusted Retail Food Services Sales Change 0 Rate M U.S. Census Bureau
OUT Adjusted Retail Sales Less Autos Change 0 Rate M U.S. Census Bureau
OUT US Industrial Production MoM 2007=100 SA 0 Rate M Federal Reserve
OUT US Capacity Utilization of Total Capacity 0 Rate M Federal Reserve
OUT US Manufacturing Trade Inventories Total 0 Rate M U.S. Census Bureau
OUT US Durable Goods New Orders Industries 0 Rate M U.S. Census Bureau
OUT US Durable Goods New Orders Ex Transp. 0 Rate M U.S. Census Bureau
OUT GDP US Chained 2005 Dollars QoQ SAAR 0 Rate Q Bureau Economic Analysis
OUT GDP US Personal Consumption Chained Change 0 Rate Q Bureau Economic Analysis
OUT US Personal Income MoM 0 Rate M Bureau Economic Analysis
OUT US Personal Consumption Expend. Nominal Dollars 0 Rate M Bureau Economic Analysis
ANT Bloomberg US Weekly Consumer Comfort Index 1 Price W BLOOMBERG
ANT University Michigan Survey Consumer Confidence 1 Price M U. of Michigan Survey Research
ANT Empire State Manufact. Survey Business Conditions 1 Value M Federal Reserve
ANT Conference Board US Leading Index MoM 0 Rate M Conference Board
ANT Philadelphia Fed Business Outlook General Conditions 1 Price M Philadelphia Fed
ANT Conference Board Consumer Confidence SA 1985=100 1 Rate M Conference Board
ANT Richmond Fed Reserve Manufacturing Survey 0 Rate M Richmond Fed
ANT US Chicago Purchasing Managers Index SA 1 Price M Kingsbury Intern.
ANT ISM Milwaukee Purchasers Manufacturing Index 1 Rate M NAPM - Milwaukee
ANT Dallas Fed Manufact. Outlook Business Activity 1 Rate M Dallas Fed
26
Eurozone Releases
27
Cat. Release Name Adj. Units Freq Source
INF Austria PPI 2010 MoM 0 Rate M Austrian Institute of Economic
INF Belgium CPI MoM NSA 100=2004 0 Rate M Belg. Inst Nat’l Stat.
INF Estonia CPI MoM 0 Rate M Statistical Office of Estonia
INF Estonia PPI MoM 0 Rate M Statistical Office of Estonia
INF Finland CPI 2010=100 MoM 0 Rate M Finnish Statistics Office
INF Finland PPI 2005=100 MoM 0 Rate M Finnish Statistics Office
INF France CPI MoM 1998=100 0 Rate M INSEE National Statistics Office
INF France European Harmonised CPI MoM 2005=100 0 Rate M INSEE National Statistics Office
INF France PPI MoM 2005=100 0 Rate M INSEE National Statistics Office
INF Germany CPI MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany HICP MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany CPI Saxony MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany CPI Baden Wuerttemberg MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany CPI Bavaria MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany CPI Hesse MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany CPI North Rhine Westphalia MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany CPI Brandenburg MoM 2005=100 0 Rate M German Fed Statistical Office
INF Germany Producer Prices MoM 0 Rate M German Fed Statistical Office
INF Germany Wholesale Prices MoM 2005=100 0 Rate M German Fed Statistical Office
INF Greece Harmonized CPI YoY 2005=100 0 Rate M National Statistical Service
INF Greece CPI YoY 2009=100 0 Rate M National Statistical Service
INF Ireland CPI All Items MoM 0 Rate M Central Statistics Office
INF Ireland HICP MoM 2005=100 0 Rate M Central Statistics Office
INF Ireland WPI MoM 2005=100 0 Rate M Central Statistics Office
INF Italy HICP MoM NSA 0 Percent M ISTAT
INF Italy CPI NIC Incl Tobacco MoM NSA 0 Percent M ISTAT
INF Italy PPI Manufacturing MoM 2005=100 0 Rate M ISTAT
INF Netherlands CPI MoM 2006=100 0 Rate M Dutch Statistics Office
INF Netherlands HICP MoM 0 Rate M Dutch Statistics Office
INF Portugal HICP MoM Base year 2005 0 Rate M Instituto Nacional de Estatist
INF Portugal CPI 2008=100 MoM 0 Percent M Instituto Nacional de Estatist
INF Portugal Producer Prices Total 2005=100 MoM 0 Rate M Instituto Nacional de Estatist
INF Slovakia CPI MoM 0 Rate M Statistical Office of the Slovakia
INF Slovakia CPI Harmonized MoM 0 Rate M Statistical Office of the Slovakia
INF Slovakia PPI MoM 0 Rate M Statistical Office of the Slovakia
INF Spain CPI MoM 2006=100 0 Rate M Instituto Nacional de Estadist
INF Spain CPI Core MoM 2006=100 0 Rate M Instituto Nacional de Estadist
INF Spain Harmonized CPI 2005=100 MoM 0 Rate M INE
INF Spain PPI MoM 2005=100 0 Rate M INE
INF Slovenia CPI MoM 0 Rate M Statistical Office of the Republic
INF Slovenia PPI MoM 0 Rate M Statistical Office of the Republic
INF Eurostat Eurozone MUICP All Items MoM NSA 0 Rate M Copyright European Communities
INF Eurostat Eurozone MUICP All Items YoY Flash Estimate NSA 0 Rate M Copyright European Communities
INF Eurostat Eurozone Core MUICP YoY NSA 0 Rate M Copyright European Communities
INF Eurostat PPI Eurozone Industry Ex Construction MoM 0 Rate M Copyright European Communities
EMP Belgium Unemployment Rate SA 1 Rate M National Bank of Belgium
EMP Estonia Unemployment Rate 1 Rate M Estonian Labour Market Board
EMP Estonia Average Gross Monthly Wages (Quarterly figures) YoY 0 Percent Q Statistical Office of Estonia
EMP Finland Unemployment Rate 1 Rate M Finnish Statistics Office
EMP France Non-Farm Non-Government Payrolls Total Quarterly 0 Rate Q INSEE National Statistics Office
EMP France Monthly Wage Index QoQ 0 Rate Q French Labor Office
EMP France Unemployment Rate ILO Method - Mainland France 1 Rate Q INSEE National Statistics Office
EMP France Unemployment Rate ILO Method Net Change (000s) 0 Volume Q INSEE National Statistics Office
EMP France Unemployment Rate ILO Method - Mainland & Overseas 1 Rate Q INSEE National Statistics Office
EMP France Jobseekers Total SA net change 1 Volume M French Labor Office
EMP Germany Unemployment Change SA 1 Rate M Deutsche Bundesbank
EMP Greece Unemployment Rate Monthly 1 Rate M National Statistical Service
EMP Ireland Unemployment Rate SA 1 Rate M Central Statistics Office Ireland
EMP Ireland Total Persons on Live Register SA 1 Volume M Central Statistics Office Ireland
EMP Ireland Total Persons on Live Register SA MoM 0 Rate M Central Statistics Office Ireland
EMP Italy New Hourly Wages MoM SA 2005=100 0 Rate M ISTAT
EMP Italy Unemployment Rate SA 1 Rate Q ISTAT
EMP Netherlands Unemployment Registered SA Per 1 Rate M Dutch Statistics Office
EMP Portugal Unemployment Rate NSA 1 Rate Q Instituto Nacional de Estatist
EMP Portuguese Labor Cost Index: Year over Year Percentage Change 0 Percent Q Instituto Nacional de Estatist
EMP Slovakia Unemployment Available to Work Rate 1 Rate M The Center for Labor
EMP Slovakia Avg Monthly Real Wages Industry YoY 0 Rate M Statistical Office of Slovakia
EMP Spain Unemployment Level MoM Net Change Latest Rev 0 Volume M Spanish Labour Ministry
EMP Spain Labor Costs Avg Monthly Labor Cost Worker YoY 0 Rate Q INE
EMP Spain Unemployment Rate 1 Rate Q INE
28
Eurozone Releases (cont.)
29
Cat. Release Name Adj. Units Freq Source
EMP Slovenia Unemployment Rate Unemployed of Active Population 1 Rate M Rep Statistical Office
EMP Slovenia Avg Gross Real Wages YoY 0 Rate M Rep Statistical Office
EMP Eurostat Unemployment Eurozone SA 1 Rate M Copyright Euro Communities
EMP Eurostat Eurozone Employment SA WDA QoQ 0 Rate Q Copyright Euro Communities
EMP Eurostat Labor Costs Nominal Values Eurozone YoY WDA 0 Rate Q Copyright Euro Communities
EMP Eurostat Eurozone Employment NSA YoY 0 Rate Q Copyright Euro Communities
OUT Austria GDP Constant Prices QoQ 0 Rate Q Austrian Institute of Economic
OUT Austria Industrial Production MoM SA 0 Rate M Statistik Austria
OUT Belgium GDP Constant 2008 Prices SA QoQ 0 Rate Q National Bank of Belgium
OUT Estonia Chain Linked GDP Seas Working Day Adj QoQ 0 Rate Q Statistical Office Estonia
OUT Estonia Retail Sale Enterprises Constant YoY 0 Rate M Statistical Office Estonia
OUT Finland GDP Constant Prices SA QoQ 0 Rate Q Finnish Statistics Office
OUT Finland GDP Working Day Adjusted 0 Rate Q Finnish Statistics Office
OUT Finland Industrial Production Volume MoM SA 2005=100 0 Rate M Finnish Statistics Office
OUT Finland Retail Sales Volume Index YoY Per 0 Rate M Finnish Statistics Office
OUT France GDP QoQ 0 Rate Q INSEE National Statistics Office
OUT France Industrial Production MoM SA 2005=100 0 Rate M INSEE National Statistics Office
OUT France Manufacturing Production MoM SA 2005=100 0 Rate M INSEE National Statistics Office
OUT Germany GDP Chain Linked Pan German QoQ 0 Rate Q German Fed Statistical Office
OUT Germany GDP Chain Linked Investment in Construction QoQ 0 Rate Q German Fed Statistical Office
OUT Germany GDP Chain Linked Exports QoQ 0 Percent Q German Fed Statistical Office
OUT Germany GDP Chain Linked Imports QoQ 0 Percent Q German Fed Statistical Office
OUT Germany GDP Chain Linked Private Consumption QoQ 0 Percent Q German Fed Statistical Office
OUT Germany GDP Chain Linked Government Consumption QoQ 0 Percent Q German Fed Statistical Office
OUT Germany GDP Chain Linked Domestic Demand QoQ 0 Rate Q German Fed Statistical Office
OUT Germany GDP Chain Linked Gross fixed capital investment QoQ 0 Rate Q German Fed Statistical Office
OUT Germany Industrial Production MoM SA 0 Rate M Deutsche Bundesbank
OUT Germany Manufacturing Orders MoM SA 0 Rate M Deutsche Bundesbank
OUT Germany Retail Sales Constant 2005 Prices MoM SA 0 Rate M German Fed Statistical Office
OUT Greece Real GDP QoQ SA 0 Rate Q National Statistical Service
OUT Greece Industrial Production YoY 0 Rate M National Statistical Service
OUT Greece Retail Sales YoY 2005=100 WDA 0 Rate M National Statistical Service
OUT Ireland GDP Constant 2005 Prices QoQ SA 0 Rate Q Central Statistics Office Ireland
OUT Ireland Industrial Production SA MoM 2000=100 0 Rate M Central Statistics Office Ireland
OUT Ireland All New Vehicle Registrations 1 Volume M Central Statistics Office Ireland
OUT Ireland Retail Sales Volume All Businesses MoM SA 0 Rate M Central Statistics Office Ireland
OUT Italy Real GDP QoQ SA WDA 0 Rate Q ISTAT
OUT Italy Industrial Production MoM SA 0 Rate M ISTAT
OUT Italy Industrial Orders MoM SA 2005=100 0 Rate M ISTAT
OUT Italy Industrial Sales MoM SA 2005=100 0 Rate M ISTAT
OUT Italy New Car Registrations YoY NSA 0 Rate M ANFIA
OUT Italy Retail Sales MoM SA 2005=100 0 Rate M ISTAT
OUT GDP at Real 2000 Prices Seasonally Adjusted in Euros QoQ 0 Rate Q Dutch Statistics Office
OUT Netherlands Industrial Production MoM 2005=100 SA 0 Rate M Dutch Statistics Office
OUT Netherlands Industrial Sales YoY 2005=100 0 Rate M Dutch Statistics Office
OUT Netherlands Retail Sales Turnover Index 2000=100 YoY 0 Rate M Dutch Statistics Office
OUT Portugal GDP Constant 2006 Prices QoQ 0 Rate Q Instituto Nacional de Estatist
OUT Portugal Industrial Production Index MoM 0 Rate M Instituto Nacional de Estatist
OUT Portugal Industrial Sales Index 2005=100 MoM 0 Rate M Instituto Nacional de Estatist
OUT Portugal Retail Sales Index MoM 0 Rate M Instituto Nacional de Estatist
OUT Slovakia GDP Constant Prices YoY 0 Percent Q Stat. Office of the Slovakia
OUT Slovakia Industrial Production Index Adjusted 0 Percent M Stat. Office of the Slovakia
OUT Slovakia Industrial Sales Constant Prices YoY 0 Rate M Stat. Office of the Slovakia
OUT Slovakia Industrial Orders MoM 0 Rate M Stat. Office of the Slovakia
OUT Slovakia Retail Sales Ex Motor Vehicles Constant YoY 0 Percent M Stat. Office of the Slovakia
OUT Spain GDP SA Chained Linked at Constant 2008 Prices QoQ 0 Rate Q INE
OUT Spain Industrial Production YoY 2005=100 0 Rate M Instituto Nacional de Estadist
OUT Spain Industrial Production Workday Adjusted YoY 0 Rate M Instituto Nacional de Estadist
OUT Spain Retail Sales Constant Prices 2005=100 YoY 0 Rate M INE
OUT Spain Retail Sales Constant Prices WDA YoY 0 Rate M INE
OUT Slovenia GDP Constant Prices YoY 0 Rate Q Statistical Office of the Republic
OUT Slovenia Industrial Production MoM 0 Rate M Statistical Office of the Republic
OUT Slovenia Retail Trade MoM 0 Rate M Statistical Office of the Republic
OUT Eurostat GDP cons prices Euro QoQ 0 Rate Q Copyright Euro Communities
OUT Eurostat GDP cons prices Euro Household CExp 0 Rate Q Copyright Euro Communities
OUT Eurozone Government Expenditure cons prices 0 Rate Q Copyright Euro Communities
OUT Eurostat GDP cons prices Euro Gross Fixed Cap Form 0 Rate Q Copyright Euro Communities
OUT Eurostat Ind Production Euro Ex Constr MoM SA 0 Rate M Copyright Euro Communities
OUT Eurostat New Orders Euro Manufact Ind Orders MoM SA 0 Rate M Copyright Euro Communities
OUT Eurostat Euro Monthly Prod Construction SA MoM 0 Rate M Copyright Euro Communities
OUT Eurostat Retail Sales Volume Eurozone MoM SA 0 Rate M Copyright Euro Communities
30
Eurozone Releases (cont.)
Cat. Release Name Adj. Units Freq Source
ANT Belgium General Index Business Confidence 0 Value M National Bank of Belgium
ANT Belgium Sen Indicator 0 Value M National Bank of Belgium
ANT Finland Industrial Confidence Indicator 0 Value M Conf. of Finnish Industries
ANT Finland Sen Indicator 0 Value M Finnish Statistics Office
ANT France Manufacturing PMI Markit Survey Ticker 0 Value M Markit
ANT France Services PMI Markit Survey Ticker 0 Value M Markit
ANT France Business Confidence Manuf Sent Index 0 Value M INSEE National Statistics Office
ANT Bank of France Business Sentiment Indicator 0 Value M Banque De France
ANT France Business Confidence General Prod Expect 0 Value M INSEE National Statistics Office
ANT France Business Confidence Personal Prod Expect 0 Value M INSEE National Statistics Office
ANT France Bus Conf Mfg Industry Demand Past 3 Month 0 Value M INSEE National Statistics Office
ANT Ifo Pan Germany Business Climate 0 Value M IFO Institute
ANT IFO Pan Germany Business Expectations 0 Value M IFO Institute
ANT ZEW Germany Assessment of Current Situation 0 Value M ZEW Zentrum
ANT ZEW Germany Expectation of Economic Growth 0 Value M ZEW Zentrum
ANT IFO Pan Germany Current Assessment 0 Value M IFO Institute
ANT Germany Manufacturing PMI Markit Survey Ticker 0 Value M Markit
ANT Germany Services PMI Markit Survey Ticker 0 Value M Markit
ANT GfK Sen 0 Value M GfK AG
ANT Ireland Consumer Sentiment Index 0 Value M IIB Bank
ANT Italy Business Confidence 0 Value M ISTAT
ANT Italy Services PMI Markit Survey Ticker 0 Value M Markit
ANT Italy Manufacturing PMI Markit Survey Ticker 0 Value M Markit
ANT Italy Sen Indicator SA 0 Value M ISTAT
ANT Netherlands Producer Confidence 0 Price M Dutch Statistics Office
ANT Netherlands Sen Seasonally Adjusted 0 Value M Dutch Statistics Office
ANT Portugal Sen Indicator 3Mth Moving Average 0 Value M Instituto Nacional de Estatist
ANT Portugal Economic Climate Indicator 0 Value M Instituto Nacional de Estatist
ANT Slovakia Industrial Confidence Indicator 0 Yield M Stat. Office of the Slovakia
ANT Slovakia Sen Indicator SA 0 Yield M Stat. Office of the Slovakia
ANT Spain Business Confidence Indicator 1 Value Q Spanish Chamber of Commerce
ANT Slovenia Sentiment Indicator SA 0 Value M Statistical Office of the Repuic
ANT EC Manufacturing Confidence Euro Ind Confidence 0 Value M European Commission
ANT EC Composite PMI Output 0 Value M Markit
ANT Eurozone Manufacturing PMI Markit Survey Ticker 0 Value M Markit
ANT Eurozone Services PMI Markit Survey Ticker 0 Value M Markit
ANT EC Economic Sentiment Indicator Eurozone 0 Value M European Commission
ANT EC Euro Area Business Climate Indicator 0 Value M European Commission
ANT EC Services Confidence Indicator Eurozone 0 Value M European Commission
ANT ZEW Eurozone Expectation of Economic Growth 0 Value M ZEW Zentrum
ANT Sentix Economic Indices Euro Aggregate Index 0 Value M Sentix Behavioral Indices
ANT EC ANT Indicator Eurozone 0 Value M European Commission
31
U.K. Releases
Cat. Release Name Adj. Units Freq Source
INF BRC Nielsen Shop Price Index All Items YoY 0 Percent M The British Retail Consortium
INF UK CPI EU Harmonized MoM NSA 0 Rate M UK Office National Statistics
INF UK CPI Ex Energy Food Alcohol and Tobacco YoY 0 Rate M UK Office National Statistics
INF UK RPI MoM NSA 0 Rate M UK Office National Statistics
INF UK RPI Less Mortgage Interest Payments YoY NSA 0 Rate M UK Office National Statistics
INF UK PPI Input Prices Mat Fuels Purchased Manufact MoM NSA 0 Rate M UK Office National Statistics
INF UK PPI Manufact Products MoM NSA 0 Rate M UK Office National Statistics
INF UK PPI Output Prices Ex Food Beverages Tobacco Petro MoM NSA 0 Rate M UK Office National Statistics
INF Bank of England GfK Inflation Expect Survey Inflation 0 Rate M Bank of England
EMP Lloyds TSB Consumer Barometer Job Prospects 1 Value M Lloyds TSB Corp Markets
EMP UK Claimant Count Rate SA 1 Rate M UK Office National Statistics
EMP UK Unemployment Claimant Count Monthly Change SA 0 Rate M UK Office National Statistics
EMP Avg Weekly Earnings 3 Month Avg Growth Whole Economy YoY 0 Rate M UK Office National Statistics
EMP UK AWE Regular Pay Whole Economy 3M Avg YoY SA 0 Rate M UK Office National Statistics
EMP UK Unemployment ILO Unemployment Rate SA 1 Rate M UK Office National Statistics
OUT UK IOS Index Total Service Industries MoM 0 Percent M UK Office National Statistics
OUT UK IOS Index Total Service Industries 3 Mth/3 Mth 0 Rate M UK Office National Statistics
OUT UK Net Lending to Individuals Consumer Credit GBP/Billion SA 12 Value M Bank of England
OUT UK Net Lending to Individuals Secured On Dwellings in Billions 12 Value M Bank of England
OUT UK New Car Registrations YoY 0 Rate M Society of Motor Manufacturers
OUT UK Industrial Production MoM SA 0 Rate M UK Office National Statistics
OUT UK Manufacturing Production MoM SA 0 Rate M UK Office National Statistics
OUT UK Business Investment Chained QoQ SA 0 Rate Q UK Office National Statistics
OUT UK GDP Chained GDP at Market Prices QoQ 0 Rate Q UK Office National Statistics
OUT UK Nat Inst of Econ and Social Research GDP Estimate QoQ 0 Rate M National Institute of Economic
OUT UK GDP Chain Mkt Prices Gross Fixed Capital Formation QoQ 0 Rate Q UK Office National Statistics
OUT UK GDP Chain Mkt Prices General Govt Consumption QoQ 0 Rate Q UK Office National Statistics
OUT UK GDP Chain Volume Measures Household SA QoQ 0 Value Q UK Office National Statistics
OUT UK GDP Chain Mkt Prices All Exports QoQ 0 Rate Q UK Office National Statistics
OUT UK GDP Chain Mkt Prices Gross Final Expend Less Imports QoQ 0 Rate Q UK Office National Statistics
OUT BRC KPMG Retail Sales Monitor Like For Like YoY 0 Percent M The British Retail Consortium
OUT UK Retail Sales All Sales Ex Auto Fuel Volume MoM SA 0 Rate M UK Office National Statistics
OUT UK Retail Sales Volume Including Auto Fuel MoM 0 Rate M UK Office National Statistics
ANT CBI MTE Full Volume of Total Order Book Balance 0 Value M Conf. of British Industries
ANT CBI MTE Full Average Sell Prices Next 3Mo Balance 0 Value M Conf. of British Industries
ANT CBI ITS Q1 Quarterly Optimism Balance 0 Value M Conf. of British Industries
ANT CBI Retail Q1Monthly Volume Sales to Y Earlier 0 Value M Conf. of British Industries
ANT UK Manufacturing PMI Markit Survey Ticker 0 Value M Markit
ANT UK Purchasing Managers Index Construction 12 Value M Markit
ANT UK Services PMI Markit Survey Ticker 0 Value M Markit
ANT GFK UK Consumer Confidence Indicator 0 Value M GfK NOP
ANT Lloyds TSB Business Barometer Current Per Balance 0 Value M Lloyds TSB Corporate Markets
ANT UK Nationwide Consumer Confidence Index SA 12 Rate M Nationwide Building Society
32
Japan Releases
Cat. Release Name Adj. Units Freq Source
INF Japan CPI Tokyo YoY 0 Rate M Ministry Internal Affairs
INF Japan CPI Tokyo Ex Fresh Food YoY 0 Rate M Ministry Internal Affairs
INF Japan CPI Tokyo Ex Food and Energy YoY Per 0 Rate M Ministry Internal Affairs
INF Japan CPI Nationwide YoY 0 Rate M Ministry Internal Affairs
INF Japan CPI Nationwide Ex Fresh Food YoY 0 Rate M Ministry Internal Affairs
INF Japan CPI Nationwide Ex Food and Energy YoY Per 0 Rate M Ministry Internal Affairs
INF Japan GDP Chained Deflators YoY 0 Rate Q Economic and Social Research
INF Japan Domestic Corporate Goods Price MoM 0 Rate M Bank of Japan
INF Japan Corporate Services Price YoY 0 Rate M Bank of Japan
EMP Japan Unemployment Rate SA 12 Rate M Ministry Internal Affairs
EMP Japan Jobs to Applicants Ratio SA 12 Rate M Ministry Health Labour
EMP Japan Labour Statistics Avg Monthly Cash Earnings YoY 0 Rate M Ministry Health Labour
EMP Japan Manpower Survey 0 Percent Q Manpower Inc.
OUT Japan Vehicle Sales YOY 0 Rate M Japan Auto Manufacturers
OUT Japan Machinery Orders: Private Sector (exc. volatile orders) MoM 0 Rate M Economic and Social Research
OUT Japan Bankruptcies Cases with Total Debt of 10 Million Y or More 0 Rate M Tokyo Shoko Research
OUT Japan Machine Tool Orders YoY 0 Rate M Japan Machine Tool Builders
OUT Japan Indices of Industrial Production MoM SA 2005=100 0 Rate M Ministry of Economy Trade
OUT Japan Capacity Utilization Operating Ratio Manufacturing MoM SA 0 Rate M Ministry of Economy Trade
OUT Japan Tertiary Industry Activity MoM SA 0 Rate /Mid M Ministry of Economy Trade
OUT Japan All Industrial Activity MoM SA 0 Rate /Mid M Ministry of Economy Trade
OUT Nomura/JMMA Seasonal PMI 12 Value M Markit/Nomura Securities
OUT Japan Vehicle Production YOY 0 Rate M Japan Auto Manufacturers
OUT Japan Big 50 Constructors Orders by Contract Value YoY 0 Rate M Ministry of Land, Infrastructu
OUT Japan Capital Investment Excl Software YoY 0 Rate Q Ministry of Finance
OUT Japan Capital Investment YoY 0 Rate Q Ministry of Finance
OUT Japan All Households Living Expend Real YoY 0 Rate M Ministry of Internal Affairs
OUT Japan GDP at Current Prices QoQ SA 0 Rate Q Economic and Social Research
OUT Japan Real GDP Annualized QoQ SA GDP expenditure approach 0 Rate Q Economic and Social Research
OUT Japan GDP Real Chained QoQ Per SA 0 Rate Q Economic and Social Research
OUT Japan Nationwide Department Store Sales YoY 0 Rate M Japan Dept Store Ass.
OUT Tokyo Department Store Sales YoY 0 Rate M Japan Department Store Ass.
OUT Japan Convenience Same Store Sales YoY 0 Rate M Japan Franchise Ass.
OUT Japan Chain Store Sales YoY 0 Rate M Japan Chain Stores Ass.
OUT Japan Retail Trade MoM SA 2005=100 0 Rate M Ministry of Economy Trade
OUT Japan Large-Scale Retail Store Sales YoY 0 Rate M Ministry of Economy Trade
ANT Japan Composite Index of Bus. Cycle Indicators Coincident Index 12 Value M Economic and Social Research
ANT Japan Composite Index of Bus, Cycle Indicators Leading Index 12 Value M Economic and Social Research
ANT Japan Economy Watchers Survey Current Conditions 12 Value M Economic and Social Research
ANT Japan Economy Watchers Survey Expectations 12 Value M Economic and Social Research
ANT Japan Small Business Confidence All Industries 0 Value M Shoko Chukin Bank
ANT Japan BSI Business Condition Large Co All Industry QoQ 0 Rate Q Ministry of Finance
ANT Japan BSI Business Condition Large Co Manufacturing QoQ 0 Rate Q Ministry of Finance
ANT Japan Consumer Confidence Overall Nationwide NSA 12 Value M Economic and Social Research
ANT Japan Consumer Confidence Households Confidence NSA 12 Value M Economic and Social Research
33
Table 1: Summary Statistics
This table presents summary statistics of the data. The data is sampled daily from January 1997 throughDecember 2011. Rm −Rf denotes the log return on the S&P 500 index in excess of Libor. The “all” indexis constructed from all growth related news. The realized growth index is based on objective realizations ofpast economic activity. The anticipated growth index is obtained from subjective forward looking surveys.Growth dispersion measures the across economists disagreement about the real-time growth factor. V RP isthe variance risk premium constructed as the difference between the squared VIX index and the expectationof realized variance based on a time-series model with lagged VIX and realized volatility as predictors.ln(P/E) and ln(D/P ) are the log price-earnings ratio and dividend yield. DEF is the default spread definedas the difference between Moody’s BAA and AAA corporate bond yields, TERM is the term spread definedas the difference between the ten-year and three-month Treasury yields.
Growth Indices Other Predictors
Rm −Rf All Realized Anticipated Dispersion V RP lnPE lnD
P DEF TERM
Summary Statistics
Mean 0.63 -0.04 -0.15 0.10 -0.00 0.04 2.98 0.55 1.03 1.68Std Dev. 21.42 1.15 1.18 1.20 0.99 0.05 0.23 0.25 0.48 1.30Skew -0.20 -1.30 -1.63 -0.83 1.40 4.96 0.05 0.48 2.82 -0.06Kurtosis 9.77 4.97 5.95 3.72 4.19 36.57 2.19 3.56 12.25 1.66
Correlation Matrix
Rm −Rf 1.00 0.01 0.00 0.01 0.01 -0.13 0.04 -0.03 -0.01 0.01
Growth Indices:
All 1.00 0.97 0.93 -0.14 -0.46 0.55 -0.71 -0.84 -0.51Realized 1.00 0.82 -0.23 -0.44 0.46 -0.66 -0.81 -0.57Anticipated 1.00 0.05 -0.44 0.65 -0.71 -0.80 -0.38Dispersion 1.00 0.15 0.19 0.09 0.14 0.02
Other Predictors:
V RP 1.00 -0.30 0.40 0.65 0.19ln(P/E) 1.00 -0.85 -0.52 -0.28ln(D/P ) 1.00 0.69 0.42DEF 1.00 0.40TERM 1.00
34
Table 2: U.S. Stock Market Return Forecasting Regressions
We estimate the following return forecasting regression:
Rm;t,t+lead −Rf ;t,t+lead = α+ βXt + εt,
where Rm;t,t+lead−Rf ;t,t+lead denotes the log return on the S&P 500 index in excess of Libor from date t tot+ lead with the return horizon lead ranging from five to 120 days. Xt is the monthly change in the realizedgrowth index (in Panel A), the level of the orthogonalized anticipated growth index (in Panel B), or both(in Panel C). The realized growth index is based on objective realizations of past economic activity. Theanticipated growth index is obtained from subjective forward looking surveys. All of the regression are basedon daily observations from January 1997 through December 2011. A constant is included in the regressionsbut not reported to save space. Robust Newey-West t-statistics are reported in parentheses.
Panel A: Realized growth
Horizon 5 20 40 60 80 100 120
∆t−22,t Realized 0.0036 0.0211 0.0316 0.0367 0.0447 0.0323 0.0296(1.70) (3.76) (3.25) (2.90) (2.45) (1.49) (1.24)
Adj. R2(%) 0.2 2.2 2.5 2.3 2.5 1.0 0.6
Panel B: Orthogonal anticipated growth
Horizon 5 20 40 60 80 100 120
Anticipated 0.0016 0.0055 0.0171 0.0323 0.0484 0.0647 0.0782(1.77) (1.96) (3.24) (4.07) (4.46) (4.67) (4.59)
Adj. R2(%) 0.1 0.5 2.6 6.2 10.2 13.7 15.8
Panel C: Realized growth and orthogonal anticipated growth
Horizon 5 20 40 60 80 100 120
∆t−22,t Realized 0.0035 0.0208 0.0306 0.0350 0.0423 0.0288 0.0252(1.67) (3.66) (3.15) (2.80) (2.41) (1.39) (1.0)
Anticipated 0.0015 0.0052 0.0166 0.0318 0.0478 0.0642 0.0778(1.72) (1.90) (3.20) (4.03) (4.40) (4.64) (4.56)
Adj. R2(%) 0.3 2.6 4.9 8.3 12.4 14.4 16.2
35
Table 3: Realized and Anticipated Growth versus Other Predictors
We estimate the following return forecasting regression:
Rm;t,t+lead −Rf ;t,t+lead = α+ βXt + γZt + εt,
where Rm;t,t+lead − Rf ;t,t+lead denotes the log return on the S&P 500 index in excess of Libor from datet to t + lead with the return horizon lead being either 20 days in Panel A or 60 days in Panel B. Xt
includes the monthly change in the realized growth index and the level of the orthogonalized anticipatedgrowth index. The realized growth index is based on objective realizations of past economic activity. Theanticipated growth index is obtained from subjective forward looking surveys. The set of other predictors Zt
contains the variance risk premium V RP , log price-earning ratio, log dividend yield, default spread DEFand term spread TERM . All of the regression are based on daily observations from January 1997 throughDecember 2011. A constant is included in the regressions but not reported to save space. Robust Newey-Westt-statistics are reported in parentheses.
Panel A: 20-day horizon
(1) (2) (3) (4) (5) (6) (7) (8) (9)
∆t−22,t Realized 0.0208 0.0242 0.0225 0.0216(3.66) (4.25) (3.96) (3.79)
Anticipated 0.0052 0.0047 0.0071 0.0067(1.90) (1.71) (2.37) (2.22)
V RP 0.0336 0.0834(0.64) (1.76)
lnPE -0.0152 -0.0241
(-1.75) (-2.71)lnD
P 0.0167(1.76)
DEF -0.0050 -0.0008(-0.87) (-0.14)
TERM -0.0006 -0.0023(-0.44) (-1.57)
Adj. R2(%) 2.6 0.1 0.4 0.6 0.2 0.0 3.1 3.7 2.9
Panel B: 60-day horizon
36
(1) (2) (3) (4) (5) (6) (7) (8) (9)
∆t−22,t Realized 0.00350 0.0401 0.0406 0.0381(2.80) (3.02) (3.25) (2.86)
Anticipated 0.0318 0.0309 0.0373 0.0371(4.03) (3.94) (4.91) (4.19)
V RP 0.0725 0.1260(0.51) (0.99)
lnPE -0.0464 -0.0781
(-2.05) (-3.94)lnD
P 0.0465(1.86)
DEF -0.0101 -0.0025(-0.68) (-0.19)
TERM -0.0012 -0.0084(-0.34) (-2.03)
Adj. R2(%) 8.3 0.1 1.4 1.8 0.3 0.0 8.7 12.1 9.8
37
Table 4: Conditional U.S. Stock Market Return Predictability
We estimate the following return forecasting regression:
Rm;t,t+lead −Rf ;t,t+lead = α+ βXt + εt,
where Rm;t,t+lead − Rf ;t,t+lead denotes the log return on the S&P 500 index in excess of Libor from date tto t + lead with the return horizon lead ranging from five to 120 days. Xt includes the monthly change inthe realized growth index and the level of the orthogonalized anticipated growth index. The realized growthindex is based on objective realizations of past economic activity. The anticipated growth index is obtainedfrom subjective forward looking surveys. Estimates are conditional on high growth dispersion (Panel A), lowgrowth dispersion (Panel B), recession periods (Panel C), or expansion periods (Panel D). Growth dispersionmeasures the across economists disagreement about the real-time growth factor. Expansions and recessionsare periods of positive or negative growth. All of the regression are based on daily observations starting ineither January 2000 for panels A and B or January 1997 for panels C and D. The sample ends in December2011. A constant is included in the regressions but not reported to save space. Robust Newey-West t-statistics are reported in parentheses.
Panel A: Growth dispersion above median
Horizon 5 20 40 60 80 100 120
∆t−22,t Realized 0.0106 0.0355 0.0421 0.0456 0.0386 0.0223 -0.0071(2.93) (3.97) (3.05) (2.83) (1.58) (0.72) (-0.22)
Anticipated -0.0000 0.0048 0.0247 0.0472 0.0701 0.0880 0.1057(-0.02) (1.14) (3.18) (3.98) (4.41) (4.75) (4.76)
Adj. R2(%) 1.8 6.5 10.3 17.0 20.9 22.11 23.9
Panel B: Growth dispersion below median
Horizon 5 20 40 60 80 100 120
∆t−22,t Realized -0.0018 0.0011 -0.0036 -0.0022 0.0132 -0.0001 0.0143(-0.87) (0.18) (-0.32) (-0.15) (0.79) (-0.01) (0.56)
Anticipated 0.0038 0.0078 0.0105 0.0146 0.0274 0.0435 0.0507(2.72) (1.84) (1.25) (1.18) (1.45) (1.59) (1.40)
Adj. R2(%) 0.6 0.8 0.9 1.2 3.7 6.3 6.2
Panel C: Recession (growth level < 0)
Horizon 5 20 40 60 80 100 120
∆t−22,t Realized 0.0062 0.0264 0.0305 0.0312 0.0299 0.0084 -0.0179(1.83) (2.92) (2.08) (1.80) (1.27) (0.28) (-0.57)
Anticipated 0.0022 0.0087 0.0258 0.0455 0.0688 0.0925 0.1148(1.93) (2.36) (3.92) (4.70) (5.28) (5.83) (6.12)
Adj. R2(%) 1.0 5.1 9.6 16.8 25.7 30.0 33.4
Panel D: Expansion (growth level > 0)
38
Horizon 5 20 40 60 80 100 120
∆t−22,t Realized -0.0019 0.0052 0.0132 0.0161 0.0241 0.0086 0.0219(-0.88) (0.85) (1.18) (0.96) (1.00) (0.34) (0.72)
Anticipated -0.0019 -0.0079 -0.0100 -0.0051 -0.0039 -0.0022 -0.0002(-1.41) (-2.35) (-1.67) (-0.55) (-0.31) (-0.15) (-0.01)
Adj. R2(%) 0.1 1.2 1.6 0.7 1.0 0.1 0.5
39
Tab
le5:
Inte
rnati
on
al
Evid
en
ce.
Local
an
dG
lob
al
Facto
rs
We
esti
mat
eth
efo
llow
ing
retu
rnfo
reca
stin
gre
gre
ssio
n:
Rm
;t,t+lead−R
f;t,t+lead
=α
+βX
t+ε t,
wh
ereR
m;t,t+lead−R
f;t,t+lead
den
otes
the
log
retu
rnon
the
S&
P500
ind
ex(i
np
an
elA
),E
UR
OS
TO
XX
50
ind
ex(i
nP
an
elB
),F
TS
E100
ind
ex(i
nP
anel
C)
orN
ikke
i22
5in
dex
(in
Pan
elD
)in
exce
ssof
corr
esp
ond
ing
Lib
or
from
datet
tot
+lead
wit
hth
ere
turn
hori
zonlead
ran
gin
gfr
om20
to80
day
s.X
tis
the
month
lych
an
ge
inth
ere
ali
zed
gro
wth
ind
exan
dth
ele
vel
of
the
ort
hog
on
ali
zed
anti
cip
ate
dgro
wth
ind
ex.
Th
ere
aliz
edgr
owth
ind
exis
bas
edon
ob
ject
ive
reali
zati
on
sof
past
econ
om
icact
ivit
y.T
he
anti
cip
ate
dgr
owth
ind
exis
obta
ined
from
sub
ject
ive
forw
ard
look
ing
surv
eys.
Th
eth
ree
sub
-pan
els
inea
chp
an
elco
rres
pon
dto
usi
ng
loca
l,glo
bal,
or
U.S
.d
eriv
edm
acr
oec
on
om
icn
ews
ind
ices
,w
her
egl
obal
corr
esp
ond
sto
the
firs
tp
rin
cip
al
com
pon
ent
of
the
fou
rlo
cal
ind
ices
.A
llof
the
regre
ssio
ns
are
base
don
dail
yob
serv
atio
ns
from
Jan
uar
y19
97th
rou
ghD
ecem
ber
2011.
Aco
nst
ant
isin
clu
ded
inth
ere
gre
ssio
ns
bu
tn
ot
rep
ort
edto
save
space
.R
ob
ust
New
ey-W
estt-
stat
isti
csar
ere
por
ted
inp
are
nth
eses
.
Pan
elA
:U
.S.
Loca
lG
lob
al
U.S
.as
Glo
bal
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d0.
0208
0.03
060.0
350
0.0
423
0.0
122
0.0
132
0.0
177
0.0
106
0.0
208
0.0
306
0.0
350
0.0
423
(3.6
6)(3
.15)
(2.8
0)
(2.4
1)
(2.0
4)
(1.3
2)
(1.3
4)
(0.6
5)
(3.6
6)
(3.1
5)
(2.8
0)
(2.4
1)
Anti
cip
ated
0.00
520.
0166
0.0
318
0.0
478
0.0
029
0.0
085
0.0
146
0.0
218
0.0
052
0.0
166
0.0
318
0.0
478
(1.9
0)(3
.20)
(4.0
3)
(4.4
0)
(2.0
3)
(3.5
8)
(4.1
5)
(4.4
8)
(1.9
0)
(3.2
0)
(4.0
3)
(4.4
0)
Ad
j.R
2(%
)2.
64.
98.3
12.4
2.3
4.9
8.8
11.8
2.6
4.9
8.3
12.4
Pan
elB
:E
uro
pe
Loca
lG
lob
al
U.S
.as
Glo
bal
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d-0
.002
5-0
.023
2-0
.0190
-0.0
294
0.0
130
0.0
102
0.0
183
0.0
092
0.0
240
0.0
386
0.0
457
0.0
513
(-0.
31)
(-1.
72)
(-1.0
1)
(-1.5
0)
(1.8
8)
(0.8
4)
(1.1
3)
(0.4
6)
(3.6
7)
(3.3
5)
(2.9
1)
(2.4
2)
Anti
cip
ated
0.01
040.
0234
0.0
329
0.0
421
0.0
036
0.0
099
0.0
152
0.0
221
0.0
050
0.0
154
0.0
296
0.0
432
(4.3
9)(5
.72)
(5.8
9)
(5.5
4)
(2.1
1)
(3.4
6)
(3.7
6)
(4.0
4)
(1.5
0)
(2.5
3)
(3.4
7)
(3.7
8)
Ad
j.R
2(%
)3.
17.
49.5
10.9
2.0
3.6
5.7
7.1
2.1
3.7
5.4
6.8
40
Pan
elC
:U
.K.
Loca
lG
lob
al
U.S
.as
Glo
bal
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d0.
0063
0.01
620.0
228
0.0
182
0.0
105
0.0
109
0.0
119
0.0
020
0.0
165
0.0
276
0.0
291
0.0
328
(1.2
6)(1
.58)
(1.6
2)
(1.0
6)
(2.1
0)
(1.2
4)
(1.1
1)
(0.1
6)
(3.3
3)
(3.3
7)
(2.8
1)
(2.3
4)
Anti
cip
ated
0.00
810.
0149
0.0
239
0.0
294
0.0
026
0.0
078
0.0
138
0.0
199
0.0
038
0.0
127
0.0
271
0.0
392
(2.0
3)(1
.91)
(2.1
8)
(2.0
1)
(1.9
6)
(3.4
7)
(4.5
3)
(4.9
2)
(1.4
1)
(2.4
8)
(3.8
4)
(4.2
8)
Ad
j.R
2(%
)0.
71.
62.7
2.4
2.0
4.5
8.7
11.3
1.7
4.0
7.2
10.1
Pan
elD
:Jap
an
Loca
lG
lob
al
U.S
.as
Glo
bal
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d-0
.001
6-0
.014
2-0
.0130
-0.0
133
0.0
061
-0.0
011
-0.0
062
-0.0
199
0.0
247
0.0
331
0.0
264
0.0
190
(-0.
32)
(-1.
71)
(-1.0
9)
(-0.7
8)
(0.8
9)
(-0.1
0)
(-0.3
6)
(-0.9
2)
(3.5
7)
(2.9
3)
(1.7
8)
(0.8
9)
Anti
cip
ated
0.01
240.
0246
0.0
304
0.0
387
0.0
058
0.0
136
0.0
200
0.0
255
0.0
065
0.0
174
0.0
318
0.0
449
(3.0
1)(3
.43)
(2.8
9)
(2.6
3)
(3.3
6)
(4.7
6)
(4.5
4)
(3.9
3)
(1.9
5)
(2.5
3)
(2.8
9)
(2.8
9)
Ad
j.R
2(%
)1.
63.
73.4
3.9
2.8
5.9
8.0
8.6
2.7
3.9
5.0
6.0
41
Table 6: Global versus Local Factors
We estimate the following return forecasting regression:
Rm;t,t+lead −Rf ;t,t+lead = α+ βXt + γZt + εt,
where the Rm;t,t+lead − Rf ;t,t+lead denotes the log return on the S&P 500 index (U.S.), EURO STOXX 50index (Europe), FTSE 100 index (U.K.), or Nikkei 225 index (Japan) in excess of corresponding Libor fromdate t to t + lead. In Panel A the return horizon lead is 20 days and in Panel B it is 60 days. Xt is themonthly change in the global realized growth index and the level of the orthogonalized global anticipatedgrowth index. The realized growth index is based on objective realizations of past economic activity. Theanticipated growth index is obtained from subjective forward looking surveys. The global indices are proxiedfor by U.S. data. Zt is the difference between the local predictors and their global counterparts, or Xt. Allof the regressions are based on daily observations from January 1997 through December 2011. A constantis included in the regressions but not reported to save space. Robust Newey-West t-statistics are reportedin parentheses.
Panel A: 20-day horizon
U.S. Europe U.K. Japan Pooled
∆t−22,t Global Realized 0.0208 0.0110 0.0179 0.0183 0.0182(3.66) (1.00) (2.47) (2.52) (5.08)
Global Anticipated 0.0052 0.0065 0.0069 0.0102 0.0076(1.90) (1.82) (1.66) (2.50) (4.50)
∆t−22,t Local Realized – ∆t−22,t Global Realized -0.0029 0.0028 -0.0041 -0.0006(-0.38) (0.56) (-0.81) (-0.21)
Local Anticipated – Global Anticipated 0.0099 0.0046 0.0074 0.0069(3.36) (1.01) (1.45) (3.71)
Adj. R2(%) 2.6 3.7 1.8 3.1 2.8
Panel B: 60-day horizon
U.S. Europe U.K. Japan Pooled
∆t−22,t Global Realized 0.0350 -0.0018 0.0338 0.0059 0.0227(2.80) (-0.08) (2.24) (0.34) (2.83)
Global Anticipated 0.0318 0.0348 0.0302 0.0373 0.0365(4.03) (4.11) (2.91) (3.05) (7.87)
∆t−22,t Local Realized – ∆t−22,t Global Realized -0.0194 0.0068 -0.0209 -0.0074(-1.05) (0.53) (-1.80) (-0.96)
Local Anticipated – Global Anticipated 0.0288 0.0053 0.0075 0.0169(4.27) (0.44) (0.71) (3.63)
Adj. R2(%) 8.3 9.7 7.3 5.9 7.4
42
Tab
le7:
Inte
rnati
on
al
Evid
en
ce
Con
dit
ion
al
An
aly
sis
We
esti
mat
eth
efo
llow
ing
retu
rnfo
reca
stin
gre
gre
ssio
n:
Rm
;t,t+lead−R
f;t,t+lead
=α
+βX
t+ε t,
wh
ere
theR
m;t,t+lead−R
f;t,t+lead
den
ote
sth
elo
gre
turn
on
the
S&
P500
ind
ex(U
.S.)
,E
UR
OS
TO
XX
50in
dex
(Eu
rop
e),
FT
SE
100
ind
ex(U
.K.)
,or
Nik
kei
225
index
(Jap
an
)in
exce
ssof
corr
esp
on
din
gL
ibor
from
datet
tot
+lead.X
tin
clu
des
the
month
lych
an
ge
inth
eU
.S.
real
ized
grow
thin
dex
and
the
level
of
the
U.S
.ort
hogon
ali
zed
anti
cip
ate
dgro
wth
ind
ex.
Th
ere
ali
zed
gro
wth
ind
exis
base
don
ob
ject
ive
real
izat
ion
sof
pas
tec
onom
icac
tivit
y.T
he
anti
cipate
dgro
wth
ind
exis
ob
tain
edfr
om
sub
ject
ive
forw
ard
lookin
gsu
rvey
s.E
stim
ate
sare
con
dit
ion
alon
hig
hgr
owth
dis
per
sion
(Pan
elA
),lo
wgro
wth
dis
per
sion
(Pan
elB
),re
cess
ion
per
iod
s(P
an
elC
),or
exp
an
sion
per
iod
s(P
an
elD
).G
row
thd
isp
ersi
onm
easu
res
the
acro
ssec
on
om
ists
dis
agre
emen
tab
ou
tth
eU
.S.
real-
tim
egro
wth
fact
or.
Exp
an
sion
san
dre
cess
ion
sare
per
iods
ofp
osit
ive
orneg
ativ
eU
.S.
grow
th.
All
of
the
regre
ssio
ns
are
base
don
dail
yob
serv
ati
on
sfr
om
Janu
ary
1997
thro
ugh
Dec
emb
er2011.
Aco
nst
ant
isin
clu
ded
inth
ere
gres
sion
sbu
tn
ot
rep
ort
edto
save
space
.R
ob
ust
New
ey-W
estt-
stati
stic
sare
rep
ort
edin
pare
nth
eses
.
Pan
elA
:G
row
thd
isp
ersi
onab
ove
med
ian
EU
UK
JP
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d0.
0324
0.06
020.0
760
0.0
780
0.0
236
0.0
328
0.0
339
0.0
247
0.0
435
0.0
469
0.0
432
0.0
299
(3.1
1)(2
.98)
(3.3
3)
(2.9
6)
(3.0
1)
(2.7
9)
(2.3
5)
(1.2
4)
(4.2
9)
(2.8
9)
(2.4
6)
(1.1
5)
Anti
cip
ated
0.01
110.
0167
0.0
318
0.0
596
0.0
057
0.0
222
0.0
439
0.0
614
0.0
048
0.0
233
0.0
469
0.0
659
(2.8
0)(2
.20)
(3.2
3)
(4.0
5)
(1.4
1)
(3.0
2)
(4.3
0)
(5.1
0)
(1.0
4)
(2.2
9)
(2.7
8)
(2.9
7)
Ad
j.R
2(%
)7.
611
.216.6
22.6
4.1
9.3
16.4
19.6
6.5
7.7
11.3
12.6
Pan
elB
:G
row
thd
isp
ersi
onb
elow
med
ian
EU
UK
JP
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d0.
0026
0.00
580.0
010
0.0
089
-0.0
027
-0.0
062
-0.0
075
0.0
017
0.0
011
0.0
028
-0.0
185
-0.0
247
(0.4
3)(0
.57)
(0.0
6)
(0.5
1)
(-0.4
6)
(-0.7
2)
(-0.7
5)
(0.1
3)
(0.1
4)
(0.2
1)
(-0.8
7)
(-0.8
2)
Anti
cip
ated
0.00
920.
0167
0.0
205
0.0
253
0.0
047
0.0
088
0.0
164
0.0
283
0.0
031
-0.0
039
-0.0
102
-0.0
036
(2.3
4)(2
.21)
(1.8
5)
(1.7
5)
(1.0
1)
(1.0
5)
(1.5
6)
(1.7
2)
(0.5
0)
(-0.3
1)
(-0.5
6)
(-0.1
3)
Ad
j.R
2(%
)1.
32.
92.6
3.2
0.3
0.8
2.5
5.1
-0.1
-0.1
0.6
0.4
43
Pan
elC
:U
.S.
rece
ssio
n(g
row
thle
vel<
0)
EU
UK
JP
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d0.
0246
0.05
070.0
613
0.0
717
0.0
147
0.0
218
0.0
224
0.0
189
0.0
303
0.0
244
0.0
222
-0.0
026
(2.4
7)(2
.59)
(2.2
9)
(2.3
6)
(1.8
7)
(1.7
7)
(1.5
9)
(1.0
7)
(3.1
1)
(1.5
8)
(1.2
0)
(-0.1
0)
Anti
cip
ated
0.01
280.
0219
0.0
337
0.0
517
0.0
101
0.0
256
0.0
443
0.0
641
0.0
110
0.0
299
0.0
484
0.0
717
(3.6
2)(3
.39)
(3.8
8)
(4.7
5)
(2.8
3)
(4.0
0)
(5.4
7)
(6.6
5)
(2.9
0)
(3.7
7)
(3.8
4)
(4.1
6)
Ad
j.R
2(%
)6.
612
.115.8
21.7
3.8
10.5
20.3
30.9
5.3
8.0
12.2
17.1
Pan
elD
:U
.S.
exp
ansi
on(g
row
thle
vel>
0)
EU
UK
JP
Hor
izon
2040
60
80
20
40
60
80
20
40
60
80
∆t−
22,t
Rea
lize
d0.
0066
0.01
940.0
167
0.0
277
0.0
063
0.0
106
0.0
054
0.0
060
0.0
033
0.0
217
-0.0
018
0.0
026
(1.1
8)(1
.80)
(1.0
2)
(1.2
6)
(1.1
8)
(1.1
1)
(0.3
9)
(0.3
1)
(0.4
2)
(1.6
5)
(-0.0
8)
(0.0
8)
Anti
cip
ated
-0.0
040
-0.0
104
-0.0
074
-0.0
040
-0.0
141
-0.0
222
-0.0
193
-0.0
240
-0.0
194
-0.0
367
-0.0
454
-0.0
655
(-1.
15)
(-1.
68)
(-0.8
6)
(-0.3
4)
(-4.0
4)
(-3.5
7)
(-1.9
9)
(-1.9
7)
(-2.8
6)
(-2.8
8)
(-2.3
5)
(-2.5
7)
Ad
j.R
2(%
)0.
62.
51.0
1.3
3.3
4.6
2.0
2.2
2.6
6.6
4.4
6.4
44
Table 8: Forecasting future growth
We estimate:
Growtht+lead = α+ ρGrowtht + β1 (Realizedt −Realizedt−22) + β2Anticipatedt + εt
where Growtht is the real-time U.S. growth factor. Realizedt is the U.S. real-time realized growth factor,Anticipatedt is the U.S. real-time anticipated growth factor that is orthogonal to realized growth. Thesample period extends from January 1997 (January 2000) to December 2011 in Panel A (in Panel B,C).ARResidualR2(%) represents the proportion of explained variance of future growth that has been leftunexplained by the current level of growth. All of the regression are based on daily observations and includea (non-reported) constant. Robust Newey-West t-statistics are reported in parentheses.
Panel A: Unconditional
5 20 40 60 80 100 120
Growth 0.9956 0.9739 0.9357 0.8929 0.8499 0.8084 0.7655(331.16) (88.96) (40.77) (25.31) (17.84) (13.97) (11.65)
∆t−22,t Realized 0.0355 0.1075 0.1997 0.2922 0.2716 0.2229 0.1564(2.44) (2.16) (2.09) (2.04) (1.64) (1.21) (0.86)
Anticipated 0.0320 0.1087 0.1755 0.2569 0.3728 0.4963 0.6068(6.71) (6.31) (5.22) (4.76) (4.79) (4.62) (4.53)
Adj. R2(%) 98.9 94.6 88.1 81.6 74.8 69.4 64.9AR Residual Adj. R2(%) 4.7 9.5 11.8 15.6 19.1 23.3 27.1
Panel B: Conditional on growth dispersion above median
5 20 40 60 80 100 120
Growth 0.9984 0.9826 0.9502 0.9171 0.8817 0.8368 0.7866(292.76) (87.34) (38.81) (23.26) (16.80) (13.01) (10.74)
∆t−22,t Realized 0.0694 0.2978 0.4883 0.6431 0.6696 0.6207 0.5129(3.95) (4.90) (3.93) (3.60) (3.39) (2.82) (2.32)
Anticipated 0.0380 0.0975 0.1773 0.2715 0.4109 0.5335 0.6265(6.1670) (4.29) (3.50) (3.02) (3.27) (3.29) (3.21)
Adj. R2(%) 99.3 96.4 90.8 85.1 80.2 75.5 71.1AR Residual Adj. R2(%) 9.6 19.6 21.7 25.2 29.3 31.7 32.8
Panel C: Conditional on growth dispersion below median
5 20 40 60 80 100 120
Growth 0.9909 0.9511 0.8999 0.8202 0.7569 0.7335 0.7237(108.19) (31.17) (17.82) (12.53) (8.31) (6.52) (5.50)
∆t−22,t Realized 0.0017 -0.0743 -0.0784 -0.0189 -0.0796 -0.1473 -0.2009(0.09) (-1.28) (-0.86) (-0.14) (-0.48) (-0.74) (-0.86)
Anticipated 0.0290 0.1313 0.2026 0.3023 0.4030 0.5084 0.6052(2.43) (3.56) (3.45) (4.05) (4.26) (3.81) (3.45)
Adj. R2(%) 97.8 89.9 82.1 74.8 65.7 60.0 56.2AR Residual Adj. R2(%) 1.5 7.0 8.6 12.6 15.2 18.0 20.3
45
Conference Board Consumer ConfidenceChicago Purchasing Managers Index Inflation
University Michigan Consumer Survey EmploymentADP National Employment Report Output
ISM Manufacturing PMI Anticipated GrowthNonfarm Payrolls Total,Manufacturing + Unemployment Rate + Average Weekly HoursISM Non-Manufacturing PMI
Retail Sales + Retail Sales Less AutoImport Price Index
PPI + PPI CoreIndustrial Production + Capacity UtilizationEmpire State Manufacturing Survey
Manufacturing Trade InventoriesCPI + CPI Core
Durable Goods OrdersConference Board Leading Index
GDP + GDP Price IndexPersonal Income + Pers. Consum. Exp. + PCE Price Index
Manufacturers New Orders
24 26 28 30 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 1 3 5 7 9 11 13 15 17 19 21 23referencemonth M month M+1 month M+2
Figure 1: This figure illustrates the typical flow of U.S. macroeconomic news. On the horizontalaxis are days of the reference month M and the subsequent two months. Above that, we listin order of reporting the macroeconomic releases for the reference month, highlighting as shadedregions the typical reporting time-frame. The news releases are color-coded corresponding to thenews category: inflation, employment, output, and anticipated growth news.
46
j=1 j=2 … j=5 j=6 … j=N1 ... ... ... missing ... ... ...
… ... ... ... missing ... ... ...
… ... ... ... missing ... ... ...
t-22 At-22,1 not released ... missing not released ... ...
t-21 not released At-21,2 ... missing At-21,6 ... ...
… not released not released ... missing not released ... ...
t At,1 not released ... At,5 not released ... ...
t+1 not released At+1,2 ... not released At+1,6 ... ...
… not released not released ... not released discontinued ... ...
… ... ... ... ... discontinued ... ...
T ... ... ... ... discontinued ... ...
j=1 j=2 … j=5 j=6 … j=N1 ... ... ... missing ... ... ...
… ... ... ... missing ... ... ...
… ... ... ... missing ... ... ...
t-22 At-22,1 E[At-22,2]=At-43,2 ... missing E[At-22,6]=At-43,6 ... ...
t-21 E[At-21,1]=At-22,1 At-21,2 ... missing At-21,6 ... ...
… E[A ...,1]=At-22,1 E[A...,2]=At-21,2 ... missing E[A...,2]=At-21,6 ... ...
t At,1 E[A t,2]=At-21,2 ... At,5 E[A t,2]=At-21,6 ... ...
t+1 E[At+1,1]=At,1 At+1,2 ... E[At+1,5]=At,5 At+1,6 ... ...
… E[A ...,1]=At,1 E[A ...,2]=At+1,2 ... E[A ...,5]=At,5 discontinued ... ...
… ... ... ... ... discontinued ... ...
T ... ... ... ... discontinued ... ...
Figure 2: This figure shows a stylized example of the macroeconomic announcement data, for Nannouncement types over a daily sample period between 1 and T . The releases j = 1 and j = 2are monthly indicators released on two different days of the month. The macroeconomic indicatorj = 5 is a news release that did not exist at the beginning of the sample, but was included inthe sample from day t onwards. The macroeconomic indicator j = 6 did exist at the beginning ofthe sample, but was subsequently discontinued. The top panel represents the matrix of the actualmacroeconomic releases in real-time as it is constructed from the data. The bottom panel showshow our simple forward filling algorithm is used to fill in the expectation of the indicator, based ona local random walk assumption, when it is not released.
47
1997 1999 2001 2003 2005 2007 2009 2011
-4
-3
-2
-1
0
1
2Growth Index
1997 1999 2001 2003 2005 2007 2009 2011-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5Dispersion
Figure 3: The upper plot shows the real-time growth index extracted from growth related economicnews. The lower plot shows the corresponding economic uncertainty measure obtained fromeconomist disagreement about growth-related news releases. Grey shaded regions represent ex-post dated NBER recessions.
48
1997 1999 2001 2003 2005 2007 2009 2011
-4
-3
-2
-1
0
1
2Realized and Anticipated Growth
Figure 4: This figure plots the realized and anticipated growth indices over time. The realizedgrowth index in blue is based on objective realizations of past economic activity. The anticipatedgrowth indiex in red is obtained from subjective forward looking surveys. Grey shaded regionsrepresent ex-post dated NBER recessions.
49
1997 1999 2001 2003 2005 2007 2009 2011
-4
-2
0
2
Inde
x
1997 1999 2001 2003 2005 2007 2009 2011S
&P
500
Growth
1997 1999 2001 2003 2005 2007 2009 2011
-4
-2
0
2
Inde
x
1997 1999 2001 2003 2005 2007 2009 2011
S&
P 5
00
Realized Growth
1997 1999 2001 2003 2005 2007 2009 2011
-2
-1
0
1
2
Inde
x
1997 1999 2001 2003 2005 2007 2009 2011
S&
P 5
00
Orthogonalized Anticipated Growth
1997 1999 2001 2003 2005 2007 2009 2011-2
0
2
4
Inde
x
1997 1999 2001 2003 2005 2007 2009 2011
S&
P 5
00
Growth Dispersion
Figure 5: This figure plots the S&P 500 index in green against various real-time macroeconomicnews indices in blue – the aggregate growth index and realized growth index in the first row andthe orthogonalized anticipated growth index and growth forecast dispersion in the second row.The growth index is extracted from all growth related economic news. The realized growth indexis based on objective realizations of past economic activity, and the anticipated growth index isobtained from subjective forward looking surveys. Orthogonalized anticipated growth is the residualfrom regressing anticipated growth on realized growth in telescoping samples. Grey shaded regionsrepresent ex-post dated NBER recessions.
50
1997
1999
2001
2003
2005
2007
2009
2011
-5-4-3-2-1012
Gro
wth
US
EU UK
JP
1997
1999
2001
2003
2005
2007
2009
2011
-5-4-3-2-1012
Rea
lized
Gro
wth
US
EU UK
JP
1997
1999
2001
2003
2005
2007
2009
2011
-5-4-3-2-1012
Ant
icip
ated
Gro
wth
US
EU UK
JP
Fig
ure
6:T
his
figu
rep
lots
the
aggr
egat
egr
owth
,re
aliz
edgr
owth
,an
dan
tici
pat
edgr
owth
ind
ices
for
the
U.S
.in
blu
e,E
uro
zon
ein
red
,U
nit
edK
ingd
omin
gree
n,
and
Jap
anin
bla
ck.
Th
egr
owth
ind
exis
extr
acte
dfr
omal
lgr
owth
rela
ted
econ
om
icn
ews
for
agi
ven
cou
ntr
yor
regi
on.
Th
ere
aliz
edgr
owth
ind
exis
bas
edon
obje
ctiv
ere
aliz
atio
ns
ofp
ast
econ
omic
act
ivit
y,and
the
anti
cip
ate
dgr
owth
ind
exis
obta
ined
from
sub
ject
ive
forw
ard
look
ing
surv
eys.
Gre
ysh
aded
regi
ons
rep
rese
nt
ex-p
ost
date
dN
BE
Rre
cess
ion
sfo
rth
eU
.S.
econ
omy.
51
1997 1999 2001 2003 2005 2007 2009 2011-50
0
50
100
150
200Macro Timing Portfolio
Figure 7: This figure plots the cumulative return on the global macroeconomic market timing andcountry selection portfolio in blue against the world market index comprised of the U.S., Eurozone,United Kingdom and Japanese stock market indexes in green. Grey shaded regions representex-post dated NBER recessions for the U.S. economy.
52