Forecasting Bond Returns in the New Normal

• PIMCO has a detailed framework for deriving a forecast for secular bond returns based on our most current expectations of policy rates and the inflation-adjusted (or real) bond risk premium.
• We start by defining the expected secular real policy rate as the expected average rate of the fed funds rate after adjusting for inflation over the next 10 years.
• Under our assumptions, we expect the secular real policy rate to average -1% per annum for the next decade, and if there are no major changes to our assumptions beyond that, to actually fall gradually towards -1.25% by 2030.
• If policy makers follow the optimal path of secular real policy rates from this point forward, the 10-year U.S. Treasury note (our proxy for the bond market) can be expected to deliver an average total return of about 2.0% to 3.0% per annum over the next five to 10 years.

Article Main Body

PIMCO’s asset allocation process focuses on deriving valuations for numerous risk premia based on both top-down macroeconomic and bottom-up microeconomic considerations. In last quarter’s Asset Allocation Focus. “Forecasting Equity Returns in the New Normal,” we detailed a framework for deriving a secular forecast for equity returns based on our current expectations of the equity risk premium value from forecasts of the quantity and quality of future economic growth. We aim here to detail a framework for deriving a forecast for secular bond returns based on our most current expectations of policy rates and the inflation-adjusted (or real) bond risk premium.

Decomposing bond returns

Accurately forecasting the future path of policy rates is the holy grail of bond investing. At the most basic level, current bond yields represent the market’s expected path of future policy rates, as well as a risk premium for bearing uncertainty around the expected path (known as the term premium).

Historically, bond returns i (using the 10-year U.S. Treasury as our proxy for the bond market) can be broken down into three fundamentally derivable components:

• returns related to the ups and downs of real economic activity
• returns related to the ups and downs of inflation
• returns related to the fluctuation of the term premium, or the market’s collective price of uncertainty

The first and second components can be neatly folded into a forecast for the policy rate – e.g. the federal funds rate in the U.S. The third component, like the equity risk premium, can be derived as a residual once growth and inflation related return expectations are determined.

Since 1900, the 10-year maturity bucket of the U.S. Treasury market has returned 4.94% compounded per annum ii. This return can be decomposed into 0.53% per annum for spot real economic activity (represented by the real return on three-month T-Bills), 3.14% per annum for spot inflation (represented by realized CPI inflation), and 1.27% per annum for the term premium (represented as the excess return on 10-year U.S. Treasuries after accounting for the real returns on T-Bills and inflation).

We focus on deriving future return expectations from the “real components” of bond returns, namely the compensation for future real economic activity and the compensation for collective uncertainty. We will return to explore PIMCO’s secular inflation expectations, and therefore forecasted returns from inflation in the New Normal, in next quarter’s Asset Allocation Focus.

Framing the path of real policy rates

The compensation received for future real economic activity as defined above is analytically best represented by the future path of real policy rates. The economic and bond market history of the United States provides an excellent petri dish for developing a framework for forecasting the future path of real policy rates, assuming one begins with trying to understand the past experience of real policy rates in their proper context.

At PIMCO, our secular forecast for the real policy rate is driven by a few key beliefs. First, we view today’s real policy rate and its future path as the sum of a “secular” real policy rate and a “cyclical adjustment” employed by monetary policymakers. Second, the secular real policy rate is dependent upon not only the quantity of future real economic growth, but the quality and stability of said growth. And third, the secular real policy rate is largely time-varying, therefore rendering naïve mean reversion based forecasts of real bond returns meaningless for the purpose of both fixed income and asset allocation portfolio strategies.

Since 1900, the U.S. economy has experienced a nearly complete set of economic conditions within which the realized path of policy rates can be studied with an eye to forecasting future paths. We have experienced both severe deflation and high inflation. We have experienced negative real growth as well as positive real growth. We have experienced de-leveraging as well as re-leveraging. And, we have experienced periods of highly stable economic activity with low default rates as well as periods of highly unstable or volatile economic activity with high default rates. This nearly complete set of economic experiences once again serves as the basis for our secular forecast for the real policy rate described below.

Understanding the historical secular real policy rate

We start by defining the expected secular real policy rate as the expected average rate of the fed funds rate after adjusting for inflation over the next 10 years. Since 1900, by this definition, the average secular real policy rate has been 0.39% per annum – rising as high as 5.38% (in 1933) and falling as low as -6.12% (in 1951).

Currently, the secular real policy rate stands at -1.33%, and is expected by bond markets to fall to -1.98% by 2017 before rising again to -0.58% by 2023. The key question for forecasting bond returns in the New Normal is whether this market’s expected path of the secular real policy rate is correct or not.

To begin answering this all-important question, we start by understanding and calibrating the historical path of the secular real policy rate based on the key principles described above.

We assume that the historical path of the secular real policy rate has often been suboptimal iii. To wit, there have been times when the rate has been too low relative to fundamental fair value (resulting in one type of suboptimal policy error) and other times when the rate has been too high relative to fundamental fair value (resulting in a different type of suboptimal policy error).

To help decipher fundamental fair value, we will generalize suboptimal policy errors into two main categories. When real policy rates are too low relative to their fundamental fair value, they result in inflation acceleration to levels above expectations. And, when real policy rates are too high relative to their fundamental fair value, they result in real growth deceleration to levels below potential, which when combined with existing debt/GDP ratios, result in accelerating private debt defaults to levels above expectations.

For the purpose of calibrating the historical path of the secular real policy rate to our understanding of the current monetary policy regime, we will assume that monetary policy today is trying to keep inflation rates stable at their current levels, while also trying to make sure that economic growth does not fall below potential such that the currently very high levels of debt/GDP do not result in accelerating levels of private debt defaults. This key assumption for our forecast is based on our current understanding of the Federal Reserve’s secular policy objectives.

Forecasting the secular real policy rate

The basic forecast for the secular real policy rate assumes there is a linear relationship between the optimal secular real policy rate and the potential real growth rate of the economy iv. Any deviation from the optimal path results in one or more of the suboptimal policy errors discussed above.

Clearly, as shown in the chart above, the historical path of the secular real policy rate has not been optimal at all times. In the 1930s, it appears that the secular real policy rate should have been falling to a negative rate instead of actually rising to a positive rate. And in the 1950s, it appears that the secular policy rate should have been rising to a positive rate instead of actually falling to a negative rate. Barring those two periods, it appears that the generally expected relationship between potential GDP growth and the secular real policy rate has held, and continues to do so today.

To better quantify the optimal path historically and project it into the future, we next introduce the sub-optimal policy error indicators of inflation and defaults in the chart below. The inflation error is calculated by measuring the realized acceleration of inflation above or below the prevailing rate measured using a 10-year trailing average. The default error is calculated by normalizing the realized issuer-weighted default rate by the prevailing debt/GDP ratio versus its own long-term average.

The definition of these policy error terms implies that the historical secular real policy rate is too low when inflation is accelerating versus its 10-year trailing average and/or when default rates adjusted for the prevailing debt/GDP ratio are below their historical average and importantly, vice versa.

Once these optimizing policy error terms are introduced into a multivariate model of the secular real policy rate, the result illustrates clearly historical instances when real policy rates were either too low or too high versus their optimal conditions, and allows us to forecast a path for the secular real policy rate based on inputs of (1) potential GDP growth, (2) total debt to GDP ratio, and (3) inflation expectations.

The results of our forecast are shown in the chart above, both in historical terms versus the actual realized secular real policy rate and in forecast terms versus the market’s expectation of the secular real policy rate.

What is very clearly visible from the historical optimization is that actual real policy rates in the aftermath of the Great Depression were much too high, a notion that has been explored and broadly accepted by policy makers and markets alike v. As a follow, it is also suggested by the historical optimization that real policy rates should have been lifted much sooner after the end of World War II, as a response to falling debt/GDP, falling default rates and rising inflation.

Similarly, in the second half of the 1970s, the historical optimization suggests real policy rates should have been tightened to around 3.0% – a target that was eventually reached and exceeded, but only a half-decade later in the mid-1980s.

Forecasting bond returns in the New Normal

While illustrating historical deviations of actual secular real policy rates from optimal secular real policy rates is useful, we are most interested in what our forecast for optimal secular policy rates suggests for the future in terms of bond market pricing and expected returns.

To forecast the next decade of secular real policy rates, we have assumed that (1) potential GDP growth in the U.S. gradually declines from a 2% rate today, toward a 1.5% rate by 2023 vi. We have also assumed that (2) U.S. debt/GDP will broadly stabilize at today’s levels (the risk to this forecast is squarely to the upside given low savings and declining demographics), and that (3) the path of realized inflation will broadly follow the market’s expectations (we will revisit and stress this assumption in detail next quarter).

Under these assumptions, we expect the secular real policy rate to average -1% per annum for the next decade, and if there are no major changes to our assumptions beyond that, to actually fall gradually towards -1.25% by 2030.

Against this secular forecast, the market expects secular real policy rates to rise from -1.3% today to about -0.6% by 2023, and to positive 0.6% by 2030 vii.

Assuming our forecast for the secular “optimal” real policy rate is correct, and assuming inflation follows the path of market expectations, the bond market is priced to deliver positive expected returns for the decade ahead. If policy makers follow the optimal path of secular real policy rates from this point forward, the 10-year U.S. Treasury note (our proxy for the bond market) can be expected to deliver an average total return of about 2.0% to 3.0% per annum over the next five to 10 years.

Further, an expectation of -1% secular real policy rates, combined with 2.5% expected inflation, produces a risk-neutral fair value yield of 1.5% for 10-year U.S. Treasuries. Given the current yield of 2.0%, the ex-ante term premium is determined to be 0.5%, about 0.75% below its long-term average viii.

Words of caution

If history has taught us one lesson, it is that policy makers often make large mistakes either due to policy regime changes or due to observational errors. Even though we have more confidence in the current Federal Open Market Committee to follow the historically derived optimal path of secular real policy rates, we remain concerned that either external currency pressures and/or domestic political pressures might cause a deviation from the prescribed lower-for-longer path.

As has been the case historically, the errors of suboptimal policy are likely to be repeated in generally the same way.

The result of a too high real policy rate path relative to the one prescribed by historical optimization will undoubtedly be a sharp rise in private defaults given the historically high debt/GDP ratios we are confronted with today.

Conversely, the result of a too low real policy rate path relative to the one prescribed by historical optimization will likely be a more rapid erosion of confidence in the U.S. dollar as a sustainable global reserve currency, sparking financial and then real de-globalization, leading to economic stagflation.

One thing is clear: The margin of error for monetary policy in the United States has never been smaller. High and rising debt stocks combined with low and falling potential GDP growth makes for a very uncomfortable mix of fundamental realities within which the secular real policy rate must be set.

The good news for investors is that real policy rates are unlikely to rise anytime soon, and lower-for-longer will give investors the best chance of realizing positive real returns from a risk-factor-diversified asset allocation portfolio going forward. The bad news for investors is that an unwarranted rise in real policy rates above their secular optimal path will likely result in much higher than expected private default rates, and therefore, negative real returns across multiple asset classes in the years ahead.

Saumil H. Parikh

Managing Director

i In this paper we focus on the interest rate component of bond returns.

ii Bloomberg data, PIMCO calculations as of 31 December 2012.

iii See Athanasios Orphanides and John C. Williams, 2011

iv See Thomas Laubach and John C. Williams, 2001.

v See Ben Bernanke, Essays on the Great Depression. 2004

vi See Chris Brightman, 2012.

vii Bloomberg data, PIMCO calculations, as of 31 December 2012.

viii See Joseph Gagnon, Matthew Raskin, Julie Remache, and Brian Sack, 2010

References

Bernanke, Ben (2004). Essays on the Great Depression, Princeton, N.J. Princeton University Press.

Brightman, Chris (2012). “1%. The New Normal Growth Rate?” Research Affiliates, November.

Gagnon, Joseph, Raskin, Matthew, Remache, Julie and Sack, Brian (2010). “Large-Scale Asset Purchases by the Federal Reserve: Did They Work?” Federal Reserve Bank of New York, Staff Report no. 441, March.

Laubach, Thomas and Williams, John C. (2001). “Measuring the Natural Rate of Interest,” Federal Reserve, November.

Orphanides, Athanasios and Williams, John C. (2011). “Monetary Policy Mistakes and the Evolution of Inflation Expectations” NBER Working Paper No. 17080, May.