Macroeconomic Dashboards

for Tactical Asset Allocation

DAVID CLEWELL, CHRIS FAULKNER-MACDONAGH,
DAVID GIROUX, SÉBASTIEN PAGE, AND CHARLES SHRIVER

A
DAVID CLEWELL wide body of academic litera- shifts take place. There’s ample evidence that
is a research analyst ture suggests that macro factors macro factors also matter.
at T. Rowe Price
can be significant drivers of asset
in Baltimore, MD.
[email protected] returns. And among practitio- PRIOR RESEARCH SHOWS THE
ners, statements such as “stocks make money IMPORTANCE OF MACRO FACTORS
CHRIS FAULKNER- in expansions and tend to lose money in
M ACDONAGH recessions” are often held as self-evident. Most of the academic literature focuses
is a global portfolio However, very little has been published on on whether macro factors get priced into
strategist at T. Rowe Price
in Baltimore, MD.
how to use these factors to inform invest- markets. Chen et al. [1986] show that the
chris_faulkner-macdonagh@ ment decisions. We show how to build sensitivities (“macro betas”) of size-sorted
troweprice.com dashboards to help integrate macro factors stock portfolios to rates, industrial produc-
into a broader, discretionary tactical asset tion, inf lation, credit spreads, and consump-
DAVID GIROUX allocation process. We view our dashboards tion explain a significant portion of their
is a portfolio manager
as trade idea generation tools that scour the relative performance over time. Fama and
and co-chair of the Asset
Allocation Committee entire set of data and highlight possible areas French [1989] use a different methodology
at T. Rowe Price of excess returns. that focuses on the broad stock and bond
in Baltimore, MD. Our goal is not to design stand-alone markets. They show that business conditions,
[email protected] systematic trading strategies based on as approximated by dividend yields, rates, and
macro factors. Rather, we submit that investors credit spreads, forecast broad market returns.
SÉBASTIEN PAGE
is the head of the multi-
should use our dashboards to introduce dis- Several other studies have confirmed the
asset division and member cipline into their asset allocation process, in importance of macro factors in explaining
of the management combination with other inputs. For example, a wide range of asset class and style premia
committee at T. Rowe for tactical asset allocation, relative valuations returns. Factors covered in the literature
Price in Baltimore, MD. matter. Even simple strategies that mechani- include consumption, unemployment, inf la-
[email protected]
cally follow the adage “buy low and sell high” tion, GDP growth, and oil prices. Examples
CHARLES SHRIVER based on valuation signals—such as the price- are highlighted in Exhibit 1.
is a portfolio manager to-earnings ratio—have outperformed static
and co-chair of the Asset benchmarks over time.1 However, valuation- ISSUES WITH PRACTICAL
Allocation Committee based investment strategies tend to be more APPLICATIONS
at T. Rowe Price effective when valuations are at extreme
in Baltimore, MD.
[email protected]
levels. Importantly, strategies that focus solely While these studies provide credible
on relative valuations can lead to disappointing evidence of the importance of macro
outcomes when important macroeconomic factors, many practitioners still struggle to

50 M ACROECONOMIC DASHBOARDS FOR TACTICAL A SSET A LLOCATION MULTI-A SSET SPECIAL ISSUE 2018

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

EXHIBIT 1
Macro Factor Examples in Literature

use these factors for tactical investment decisions at the from the econometric methods used in academic
6- to 18-month horizon. Economists and investment studies—it is meant to be simpler and more intuitive.
teams often operate independently, and the question Also, unlike historical regression analyses based on static
of what macro expectations are priced into markets is data samples, our dashboards are meant to be dynami-
often left unanswered. Moreover, the sheer amount of cally updated so that practitioners can rely on them as
macro data makes it difficult to separate noise from signal a research tool or to inform investment decisions on an
and anticipate which variables will drive returns. ongoing basis. We focus on the relative returns between
Another challenge in the practical application of pairs of asset classes. We highlight which factors may
existing studies is that macro factors may inf luence asset have a significant impact on which pair trades, under
class returns differently based on initial conditions. Boyd various scenarios. Importantly, we take into account
et al. [2005], for example, show that a rise in unemploy- current conditions, as ref lected in the macro factors’
ment during an expansion affects stock returns differently current percentile levels.
than a rise during a recession. Similarly, the effect of a In Exhibit 2, we show the macro factors included in
decline in industrial production may depend on whether our dashboards. This list broadly corresponds to the key
starting business conditions are good or bad. In fact, we factors documented in previous studies. In Exhibit 3, we
suggest that any macro factor’s impact on asset returns show the list of asset-class-level pair trades that we model
depends on the prevailing regime. Yet with the exception as a function of the macro factors. For each pair trade,
of Boyd et al. [2005], previous research does not account we partition historical asset returns to match a given sce-
for the relationship between current conditions and the nario and current conditions. Our entire framework is
subsequent impact of macro factors on asset returns. out of sample. Starting from each macro factor’s current
level, our dashboards answer the following question:
PROVIDING DISCIPLINE TO TACTICAL If an investor has a one-year view on the direction of
ASSET ALLOCATION: DATA AND the macro factor, what is the corresponding forward
METHODOLOGY one-year return?
We illustrate our methodology in Exhibit 4.
To map macro factors to expected asset returns, we To create large enough data samples, we use ranges for
propose the use of dashboards. Our approach is different

MULTI-A SSET SPECIAL ISSUE 2018 THE JOURNAL OF PORTFOLIO M ANAGEMENT 51

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

EXHIBIT 2 x (75%∗f ) is the cutoff for the third quartile. Next we define
List of Macro Factors scenarios (St+1) similarly:

S( f )t +1 = f t +1 − f t (2)

where the scenarios are predefined ranges that are

meaningful to the practitioner (such as a 25 basis point
(bp) rise in 10-year Treasury yields), and the subscript
t + 1 denotes one-year forward returns. Then we calcu-
late the conditional pair trade return (RtC+1) as

RtC+1 = E( Rt +1 IC t , St +1 ), (3)

which is the average historical return of the pair trade

when initial conditions were in the same range (low,
medium, or high), and the factor subsequently moved
according to the scenario. We also include the 10th to
90th percentile range and identify when the sign was the
same as the average (“hit rate”) at least 80% of the time.
For example, suppose we want to evaluate the
impact of the dollar on the relative performance
between small and large cap stocks. Over the entire
sample, from January 1990 to December 2016, U.S.
small caps (Russell 2000) outperformed U.S. large caps
(Russell 1000) in 51% of rolling 12-month periods. For
monthly data available through April 10, 2017, the U.S.
Notes: All series are retrieved from Haver, except for high-yield spreads dollar index stands at 100.6, which is in the top quar-
( J.P. Morgan Global High Yield Spread-to-Worst), J.P. Morgan tile of its history since January 1990. Further, suppose a
Emerging Markets Currency Index, DXY (Factset), and U.S. tactical asset allocator expects the U.S. dollar to rise fur-
investment-grade spreads (Bloomberg Barclays U.S. Aggregate Index
OAS). Historical analysis data end in December 2016. All data are
ther. Given history, when the U.S. dollar was in the top
sourced at the monthly frequency. Levels are based on data reported quartile and subsequently rose by 5% (or more) over the
April 10, 2017. We estimate real yields as the nominal Treasury yield next year, U.S. small caps outperformed U.S. large caps
less year-over-year Core CPI. 88% of the time. The average outperformance was 8.2%,
with the 10th to 90th percentile range between −2.3%
the macro factors. We define initial conditions of each and 15.9%. In this case, the outperformance of U.S. small
factor, (IC( ft )) as follows: caps in periods of rising USD may be attributed to their
lower reliance on exports, compared to U.S. large caps.
⎧“low ” x L ≤ f t < x( 25%∗ f ) Transaction costs are difficult to estimate because
⎪ they depend on the amount traded, the method of
IC ( f )t = ⎨“medium” x( 25%∗ f ) ≤ f t ≤ x(75%∗ f ) execution (physicals versus futures, for example), and
⎪ how market impact is parsed from opportunity cost.
⎩“high” x(75%∗ f ) < f t ≤ xU (1)
Nonetheless, for illustration purposes, we report returns
where ft is the level of the macro factor at time t, and net of a rough estimate of transaction costs, which are
xL and xU are lower and upper bounds that determine detailed in the appendix.
initial conditions, based on long-term percentile ranks: In Exhibit 5, we show how to read the dashboard,
x (25%∗f ) is the lowest quartile value of the factor, while and in Exhibits 6–9, we show the dashboards under uncon-
ditional, stable, rising, and declining macro factors.

52 M ACROECONOMIC DASHBOARDS FOR TACTICAL A SSET A LLOCATION MULTI-A SSET SPECIAL ISSUE 2018

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

EXHIBIT 3
Asset Class Returns

Notes: Historical analysis data end in December 2016. Unless specifically identified, asset class returns are computed using total return indices. All data
are sourced monthly. The T. Rowe Price Real Assets Blended Benchmark is the following: As of December 1, 2013, the Real Assets Combined Index
Portfolio comprises 25% MSCI ACWI Metals & Mining, 20% Wilshire RESI, 20% FTSE EPRA/NAREIT Dev Real Estate Index, 19.5%
MSCI ACWI Energy, 10.5% MSCI ACWI Materials, 4% MSCI ACWI IMI Gold, and 1.00% MSCI ACWI IMI Precious Metals and Minerals.
Prior to this date, the Real Assets Combined Index Portfolio was composed of 25% MSCI ACWI Metals & Mining, 20% Wilshire RESI, 20% FTSE
EPRA/NAREIT Dev Real Estate Index, 16.25% MSCI ACWI Energy, 8.75% MSCI ACWI Materials, 5% UBS World Infrastructure and
Utilities Index, 4% MSCI ACWI IMI Gold, and 1.00% MSCI ACWI IMI Precious Metals and Minerals.
Sources: Bloomberg Barclays, Russell, Credit Suisse, FactSet, J.P. Morgan, and T. Rowe Price.

The average return reported under “Conditional scenario-specific dashboards (Exhibits 7–9), the direc-
Returns” of the first dashboard (Exhibit 5) is weighted tion of the macro factors over the next year could matter.
based on bucket size (percentile range) for each macro Stable or improving macro conditions correlate
factor. The P-value indicates whether this average with strong returns for “risk-on” trades, such as long
return—from current starting conditions (as of April 10, stocks, small caps, high-yield bonds, and emerging
2017)—is statistically different from the full history of market bonds. On the other hand, a rising U.S. unem-
returns; it should be below 0.05 (or 95% confidence ployment rate, especially from its currently low level, is
level) to indicate a meaningful difference. likely to be connected with a selloff in stocks.
Emerging market currencies may be an important
Interpreting the Dashboards factor to watch. Stable or rising emerging market
currencies are supportive of emerging market equities,
As shown in Exhibit 6, all else being equal, based real asset equities, and emerging market bonds.
on current conditions with no forward view, macro Further, emerging market currencies have depreciated
factors are unlikely to drive relative asset class returns significantly—the index currently sits at the bottom 5%
(over the next year) to be different from long-term of its historical range. If they move significantly up or
“unconditional” returns. However, as shown in the down from their currently low level, they could corre-
late with meaningful directional volatility across assets.

MULTI-A SSET SPECIAL ISSUE 2018 THE JOURNAL OF PORTFOLIO M ANAGEMENT 53

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

EXHIBIT 4
Stylized Illustration of our Out-of-Sample Methodology

And the price of oil, if it remains stable or appreci- dashboards lies not in their academic merit, but rather in
ates from its currently medium level (63rd percentile), their value to practitioners. The confidence intervals and
could be a significant positive driver of emerging market hit rates help filter the continuous f lood of macro data.
stocks, real asset equities, and emerging market bonds. (Note that we do not report volatilities by pair, but they
Regarding style rotation, growth stocks have longer are directly proportional to the confidence intervals.)
duration than value stocks. Therefore, even though value Importantly, although the relationships among the
stocks have a higher dividend yield than growth stocks, macro factors and asset classes are reasonably persistent,
when rates decline, growth outperforms; and when rates the dashboards should be updated frequently, because as
rise, value outperforms. This effect occurs both in the initial conditions change, some of the investment con-
United States and EAFE markets. The large weight of clusions may change as well.
negative-duration financials in the value index partly
explains this effect. CAVEATS
There are several other useful interpretations of
the data, but in general, our results are in line with We don’t claim to identify causation, which is
economic intuition, as well as the findings published almost always impossible to determine given the com-
in the literature we reviewed. The contribution of our plexity and dynamic nature of which factors drive asset

54 M ACROECONOMIC DASHBOARDS FOR TACTICAL A SSET A LLOCATION MULTI-A SSET SPECIAL ISSUE 2018

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

EXHIBIT 5
How to Read the Dashboards

returns; rather, we merely identify correlations that appear to U.S. equities based on valuation metrics (price-to-
meaningful and leave it to the investor to assess causation. earnings ratios and other such metrics) and macro fac-
Also, although the academic literature suggests tors indicate ACWI ex-U.S. should outperform, then
that our selected macro factors can be significant drivers a tactical asset allocator may take a larger position in
of asset returns, the confidence intervals for one-year ACWI ex-U.S. equities than if valuation and macro
returns in our dashboards are wide, and statistical data don’t agree.
confidence is low across the board. Hence, we don’t Another caveat is that we don’t model expecta-
recommend building systematic tactical asset alloca- tions directly. In theory, we should run our scenarios
tion strategies based solely on these macro data and in against the expectations that are priced into the market.
this manner. The problem is that expectations are often difficult and
Instead, macro data should be used in combi- in many cases impossible to measure. Survey data may
nation with relative valuations and other factors such be useful, but they rarely reveal what markets are truly
as fundamentals and technicals to determine both pricing in, nor are survey results available on a timely—or
whether to invest, and in what size. Macro factors are broad enough—fashion going far enough back in time.
often used to confirm relative valuation signals. For Regarding market-implied views, forward curves incor-
example, if ACWI ex-U.S. equities are cheap relative porate a risk premium, which makes it hard to disentangle

MULTI-A SSET SPECIAL ISSUE 2018 THE JOURNAL OF PORTFOLIO M ANAGEMENT 55

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

 56
EXHIBIT 6
Dashboard of One-Year Forward Returns Based on Current Macro Conditions, April 10, 2017

M ACROECONOMIC DASHBOARDS FOR TACTICAL A SSET A LLOCATION

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

The P-value associated with (C) ≠ (A) is generally between 0.9 and 1.0 given current economic factors. This means the current average and distribution is not statistically different fom the
long-term averages.

MULTI-A SSET SPECIAL ISSUE 2018

 Dashboard of One-Year Forward Returns Based on Current Macro Conditions, Assuming Stable Factors, April 10, 2017
EXHIBIT 7

MULTI-A SSET SPECIAL ISSUE 2018 THE JOURNAL OF PORTFOLIO M ANAGEMENT 57

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

 Dashboard of One-Year Forward Returns Based on Current Macro Conditions, Assuming Rising Factors, April 10, 2017
EXHIBIT 8

58 M ACROECONOMIC DASHBOARDS FOR TACTICAL A SSET A LLOCATION MULTI-A SSET SPECIAL ISSUE 2018

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

 Dashboard of One-Year Forward Returns Based on Current Macro Conditions, Assuming Declining Factors, April 10, 2017
EXHIBIT 9

MULTI-A SSET SPECIAL ISSUE 2018 THE JOURNAL OF PORTFOLIO M ANAGEMENT 59

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

an expectations component. Ultimately, Chen, Roll, and APPENDIX
Ross [1986] mention that spreads and interest rates series
are noisy enough to be treated as unanticipated. Also, Transaction cost assumptions are for illustrative
they find that econometric methods to extract the unan- purposes. Actual transaction costs may vary.
ticipated component of industrial production do not offer
any advantage over the unadjusted series.2
Lastly, we’ve selected easily investable asset class
pairs. This list represents asset classes commonly used
in practice by asset allocators. But in theory, it would
be more elegant to isolate market factors and scale posi-
tions based on volatility. For example, we could hedge
the equity risk factor common on both sides of the
small- versus large-caps pair, or at least make sure the
trade is equity-beta neutral. Although statistical sig-
nificance would likely increase (see Naik et al. [2016]),
we would move away from implementable trades. Ulti-
mately, our goal is to add discipline to the analysis of
macro factors, and our framework is one piece of the
puzzle, focused on idea generation. Portfolio construc-
tion then involves combining macro with other fac-
tors, adjusting broad market factor exposures, as well
as risk-scaling positions between the long and the short
leg and across trades.

CONCLUSIONS

Too often, quantitative models ignore the cur-

rent state of the world. Historical data analysis can be
useful (after all, we don’t have future data), but only to
the extent it helps formulate a view about the future. ENDNOTES
Our dashboards help practitioners filter historical data
The authors would like to thank Stefan Hubrich, Ph.D.,
to try to predict the impact of macro factors on asset CFA, Sean McWilliams, the T. Rowe Price Multi-Asset
returns. Based on a wide body of academic literature, Research and Development team, the T. Rowe Price Asset
we have developed a framework that incorporates cur- Allocation Committee, Josh Yocum, CFA, and Natalie Reed
rent conditions and that investors can easily replicate. for their support and feedback to improve the framework and
Instead of empirical statistical tests on the pricing of sharpen the usefulness of the research for investors across the
macro factors—which have already been covered in firm.
1
prior academic research—we focus on how to use data See, for example, Chapter 5 in Naik et al. [2016], as
in the investment decision-making process. Our dash- well as the performance of the stand-alone value strategies
boards filter one-year forward returns for a wide range in Blitz and P. Van Vliet [2008], Asness et al. [2013], and
of asset-class-level pair trades, based on current macro Haghani and Dewey [2016].
2
However, they lead industrial production by one year.
conditions and expected movements in macro factors.
For tactical asset allocation, this obviously would be like
Our results reveal that for tactical asset allocation, macro
“cheating” because it would assume perfect foresight.
factors matter.

60 M ACROECONOMIC DASHBOARDS FOR TACTICAL A SSET A LLOCATION MULTI-A SSET SPECIAL ISSUE 2018

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018

REFERENCES Haghani, V., and R. Dewey. “A Case Study for Using Value
and Momentum at the Asset Class Level.” The Journal of
Asness, C.S., T. Moskowitz, and L. Pedersen. “Value and Portfolio Management, Vol. 42, No. 3 (2016), pp. 101-113.
Momentum Everywhere.” The Journal of Finance, Vol. 68,
No. 3 ( June 2013), pp. 929-985. Jensen, G.R., R. Johnson, and J. Mercer. “New Evidence on
Size and Price-to-Book Effects in Stock Returns.” Financial
Bernanke, B. “The Relationship Between Stocks and Oil Analysts Journal, Vol. 53, No. 6 (November/December 1997),
Prices.” Brookings Institute, February 2016. www.brookings pp. 34-42.
.edu.
Kritzman, M., D. Turkington, and S. Page. “Regime Shifts:
Blair, H., and X. Qiao, “A Practitioner’s Defense of Return Implications for Dynamic Strategies.” Financial Analysts
Predictability.” The Journal of Portfolio Management, Vol. 43, Journal, Vol. 68, No. 3 (May/June 2012), pp. 22-39.
No. 3 (2017), pp. 60-76.
Lettau, M., and S. Ludvigson. “Consumption, Aggregate
Blitz, D., and P. Van Vliet. “Global Tactical Cross-Asset Wealth, and Expected Stock Returns.” The Journal of Finance,
Allocation: Applying Value and Momentum Across Asset Vol. 56, No. 3 ( June 2001), pp. 815-849.
Classes.” The Journal of Portfolio Management, Vol. 35, No. 1
(Fall 2008), pp. 23-28. Ludvigson, S.C., and S. Ng. “Macro Factors and bond Risk
Premia.” The The Review of Financial Studies, Vol. 22, No. 12
Booth, J.R., and L.C. Booth. “Economic Factors, Monetary (2009), pp. 5027-5067.
Policy, and Expected Returns on Stocks and Bonds.” FRBSF
Economic Review, Vol. 2 (1997), pp. 32-42. Naik, V., M. Devarajan, A. Nowobilski, S. Page, and
N. Pedersen. “Factor Investing and Asset Allocation:
Boyd, J.H., J. Hu, and R. Jagannathan. “The Stock Market’s A Business Cycle Perspective.” CFA Institute Research founda-
Reaction to Unemployment news.” The Journal of Finance, tion Publications, Vol. 2016, No. 4 (December 2016).
Vol. 60, No. 2 (2005), pp. 649-672.
Page, S. “R isk Management beyond Asset Class
Chen, N., R. Roll, and S. Ross. “Economic Forces and Diversification.” CFA Institute Global Investment Risk
the Stock Market.” The Journal of Business, Vol. 59, No. 3 Symposium, cfapubs.org, September 2013.
( July 1986), pp. 383-403.
Zhang, Q.J., P. Hopkins, S. Satchell, and R. Schwob.
Dahlquist, M., and C.R. Harvey. “Global Tactical Asset “The Link Between Macro-Economic Factors and Style
Allocation.” Emerging Markets Quarterly, Vol. 5, No. 1 Returns.” Journal of Asset Management, Vol. 10, No. 5
(Spring 2001), pp. 6-14. (December 1, 2009), pp. 338-355.

Fama, E.F., and KR. French. “Business Conditions and

Expected Returns on Stocks and Bonds.” Journal of Financial To order reprints of this article, please contact David Rowe at
Economics, Vol. 25 (1989), pp. 23-49. drowe@ iijournals.com or 212-224-3045.

Franz, R. “Macro-Based Parametric Asset Allocation.”

Working paper, December 2013, SSRN 2260179.

MULTI-A SSET SPECIAL ISSUE 2018 THE JOURNAL OF PORTFOLIO M ANAGEMENT 61

Downloaded from http://jpm.iijournals.com/ by guest on August 5, 2018