PIMCO Quantitative Research Stock Bond Correlation Oct2013
PIMCO Quantitative Research Stock Bond Correlation Oct2013
PIMCO Quantitative Research Stock Bond Correlation Oct2013
November 2013
Analytics
In Figure 3 we report the parameter estimates alongside Lastly, bond and equity yields have opposite sensitivities to
t-statistics in both the levels regression and the error the fed funds policy gap, based on the levels regression. This
correction model.5 result is likely due to the Federal Reserve’s policy to keep rates
low (good for bonds) when unemployment is high (bad for
stocks) and vice versa.
2.0% Based on our models for stock and bond yields, we use a
simulation model to build forward-looking estimates of their
0.0% correlation as a function of various macroeconomic projections.
Mar Jun Sep Dec Mar Jun Sep Dec Feb
‘88 ‘91 ‘94 ‘97 ‘01 ‘04 ‘07 ‘10 ‘13 We focus on yields as opposed to returns for econometric
convenience, but the correlation of returns should be very close
Actual 10 year Treasury yield Fitted 10 year Treasury yield
to the correlation in yields. Even at the two-year horizon, stock
Source: PIMCO, Bloomberg. Data as of Q1 1988 – Q2 2013.
Hypothetical example for illustrative purposes only. returns have a correlation greater than 95% with changes in
cyclically adjusted earnings yields. And for Treasuries, changes
FIGURE 5: ACTUAL VERSUS FITTED CYCLICALLY ADJUSTED S&P 500
in yield are a very common approximation for returns per
EARNINGS YIELD unit of duration. Hence, for the purposes of estimating the
stock-bond correlation, using changes in valuations is roughly
8.0% equivalent to using returns (or at least well within the margin
7.0% of error for any model-based forecast).
6.0%
In Figure 6 we show the multi-step process we use to generate
5.0%
model-implied correlations. We follow three broad steps:
4.0%
1. We simulate (through a Monte Carlo simulation) 1,000
3.0%
paths for growth, unemployment, inflation, and policy rates.
2.0%
1.0%
2. Next we derive 1,000 simulated paths in stock and bond
yields, based on our econometric model (using the
0.0%
Mar Mar Mar Mar Mar Mar Mar Mar Feb coefficients on levels and changes from Figure 3).
‘88 ‘91 ‘94 ‘97 ‘00 ‘03 ‘06 ‘09 ‘13
3. From these simulated changes we calculate the stock-bond
Actual S&P earnings yield Fitted S&P 500 earnings yield correlation at various time horizons.
Source: PIMCO, Shiller, Bloomberg. Data as of Q1 1988 – Q2 2013.
Hypothetical example for illustrative purposes only. The mean outcomes for the macroeconomic variables are
based on FOMC projections, and their dynamics
(autocorrelations, volatilities and correlations) are calculated
on historical data from Q1 1988 to Q2 2013. Our framework
includes error terms in the simulated macroeconomic variables’
paths and in the implied stock and bond yield changes. (The
appendix provides details on this simulation process and the
underlying parameters.)
0.05
Simulate paths
for macroeconomic 0.00
variables
-0.05
-0.10
For each path,
calculate implied -0.15
stocks and bond
yield changes -0.20
-0.25
Rates implied
Real GDP growth Unemployment Inflation rate Real policy rate
by Taylor rule
FOMC Projections
Gaps
Sources: FOMC (data as of June 2013), Haver, Bloomberg, PIMCO. Historical data from Q1 1988 – Q2 2013.
Figure A2 shows our estimated parameters alongside the Feinman, J. N. (2005, Summer). Inflation illusion and the
correlations and volatilities of the error terms (εj ). (mis) pricing of assets and liabilities. The Journal of
Investing, 29–36.
The realizations of the driving macroeconomic variables are
jointly simulated 1,000 times from Q3 2013 through Q4 2015 Fuller, R. J., & Petry, G. H. (1981). Inflation, return on equity,
with correlation and volatility structure dictated by parameters and stock prices. The Journal of Portfolio Management,
in Figure A2 and average realizations along the path given by 7(4), 19–25.
the FOMC projections in Figure A1.
Geske, R., & Roll, R. (1983). The fiscal and monetary linkage
References between stock returns and inflation. The Journal of Finance,
38(1), 1–33.
Amenc, N., Martellini, L., & Ziemann, V. (2009, Summer).
Inflation-hedging properties of real assets and implications Gulko, L. (2002). Decoupling. The Journal of Portfolio
for asset-liability management decisions. The Journal of Management, 28(3), 59–66.
Portfolio Management, 94–110. Hughes, M. (1992). U.K. equity-gilt study 1918–1991.
Andersson, M., Krylova, E., & Vahamaa, S. (2007). Why does The Journal of Investing, 1(2), 37–43.
the correlation between stock and bond returns vary over Johnson, N., & Page, S. (2012, October). Inflation regime
time? Journal of Applied Financial Economics, 18(2), 139–151. shifts: Implications for asset allocation. PIMCO In Depth,
Asikoglu, Y., & Ercan, M. R. (1992). Inflation flow-through www.pimco.com
and stock prices. The Journal of Portfolio Management, Leibowitz, M. L. (1986). Total portfolio duration: A new
18(3), 63–68. perspective on asset allocation. Financial Analysts Journal,
Baele, L., Bekaert, G., & Inghelbrecht, K. (2009). The 42(5), 18–29, 77.
determinants of stock and bond return comovements. Leibowitz, M. L., & Kogelman, S. (1993). Resolving the
The Review of Financial Studies, (23)6, 2374-2428. equity duration paradox. Financial Analysts Journal,
Bhardwaj, G., Hamilton, D. J., & Ameriks, J. (2011, April). 49(1), 51–64.
Hedging inflation: The role of expectations. Vanguard Leibowitz, M. L., Sorensen, E. H., Arnott, R. D., & Hanson,
Research, 1–9. H. N. (1989). A total differential approach to equity
Efron, B., & Tibshirani, R. J. (1993). An introduction to the duration. Financial Analysts Journal, 45(5), 30–37.
bootstrap. Monographs on Statistics and Applied Li, L. (2002). Macroeconomic factors and the correlation of
13-1347_GBL