Voya’s 2022 Capital Market Assumptions are a key input into our strategic asset allocation process for multi-asset portfolios.
Executive Summary
Our 2022 Capital Market Assumptions details our research on asset class returns, standard deviations of returns and correlations over the 2022–2031 timeframe. These estimates are a key input into our strategic asset allocation process for our multi-asset portfolios; they also provide a context for evaluation of the macroeconomic inputs that support the return estimates.
We expect that the next ten-year period will be characterized by returns below historical averages, to varying degrees across all asset classes. Our current forecast is for U.S. and international developed market equities to produce mid-single-digit returns, which are similar to what was forecasted last year.
The macro inputs of a historically low potential GDP growth trajectory, reduced labor supply and more inflation inform our forecasts. To combat utilizing a single point estimate forecast, we incorporate an alternate scenario into our methodology. This step delivers a holistic approach to our macro inputs, which produces blended estimates. Each year the asset allocation team determines through its research process if the alternative scenario is to have slightly better or worse macro inputs. This year we again used marginally higher productivity and lower terminal fed funds rate inputs for the alternative case scenario.
In most cases, expected returns for developed international market assets are lower than those for comparable U.S. assets. Expected returns for emerging market equities are higher than U.S. equity returns. With a more uncertain path to growth than that of developed markets, emerging markets continue to have higher expected equity return volatility.
Our bond return assumptions imply that returns generally will be in the low single digits. We note that our projections assume that moves in both bond term premiums and real interest rate premiums will cap upside returns available to fixed income assets.
Forecast Environment
Our long-term capital market assumptions provide our estimates of expected returns, volatilities and correlations among major U.S. and global asset classes over a 10-year horizon. These estimates guide strategic asset allocations for our multi-asset portfolios and provide a context for shorter-term economic and financial forecasting.
As has been the case in the past, our forecast models an explicit process of convergence to a steady- state equilibrium for global economies and financial markets through 2031. In our modeling process, we worked with the economic consulting group Macroeconomic Advisors to provide additional analytical insight for our macro inputs.1
We believe that cyclical fluctuations are an inevitable aspect of market economies and therefore recognize that the steady-state equilibrium incorporated as the terminal point of our forecast is unlikely to be fully attained over any point-to-point 10-year period under real world conditions. Nonetheless, we believe that this is a useful theoretical construct for anchoring the forecast. As a result, the forecast does not assume a recession or contraction over the 2022–2031 horizon.
Still a Low Growth Environment
Over the 2022–2031 period covered by our forecasts, we believe the United States has the ability to move to a somewhat higher, sustained growth path than was the case in the previous business cycle. The key is for the U.S. to exit the current low productivity regime that has constrained the economy.
Productivity growth essentially comes from capital deepening and total factor productivity (TFP). The latter is an unobservable measure taken from the decomposition of real GDP growth, the remainder after accounting for the contributions of capital and labor, called Solow’s residual. This residual could reflect improvements in technology, growth in the “effectiveness” of labor, strength in property rights, quality of labor and cultural attitudes. Cultural attitudes can include risk and high levels of confidence in the outlook, which can contribute to a productivity revival through the TFP channel. Increased confidence in the outlook is necessary to raise trend growth along with the potential for deregulation to encourage the capital formation essential for moving to a higher productivity growth rate.
Our research shows that labor-force productivity growth alternates between high and low productivity regimes over time. To determine which regime we are in we fit productivity data through a Markov model (Figure 1). The latest productivity data as of third quarter 2021 show U.S. productivity growth has moved from 4.0% on a year-over-year basis in 3Q20 to -0.5% by 3Q21. U.S. labor-force productivity is now estimated to be in the “low productivity” regime. The system had been in “high productivity” equilibrium for four quarters following the Covid recession. The estimated means of the two productivity regimes are 3.7% (high) and 1.1% (low). A Hodrick-Prescott filter based decomposition of year-over-year productivity growth into trend and cycle components also shows that the current trend of U.S. productivity growth is 2.2%.
Figure 1. Productivity Growth Has Begun to Accelerate
U.S. Productivity Regimes
Source: Voya Investment Management. Note: non-shaded areas in the chart denote low productivity regimes. Data as of September 30, 2021.
As in the past, our 2022 CMA forecast is predicated on a “base” and an “alternate” scenario. The alternate scenario assumes that the U.S. does see modest improvement in output per hour; largely the result of gains in total factor productivity as the labor share shifts away from brick-and-mortar to more productive firms. We generate our forecast based on 60%/40% weights given to the base/alternate scenarios. Our forecast is for U.S. GDP growth over the 10-year period to attain 2.7%. Figure 2 shows the 2031 values from this forecast, which is consistent with our estimates of longer-term, steady-state values for key U.S. economic variables.
Figure 2. U.S. Economic and Financial Variables
Source: Voya Investment Management and Macroeconomic Advisers. Forecasts are subject to change.
Long-Run Assumptions
While 10-year forecasts guide our strategic asset allocations, our long-run equilibrium return assumptions refer to horizons much longer than our 10-year forecasts, and are the underlying inputs into our glide path estimations. Typically, we assume the horizon for these forecasts is 40 years. At that point, we think of the economy as being in a steady state where GDP grows at its trend rate, inflation is at target, unemployment equals the non-accelerating inflation rate of unemployment, the real interest rate equals the “natural” rate of interest — neither contractionary nor inflation inducing — and all capital and goods markets are in equilibrium.2
These forecasts use a building block methodology. Starting with our expectations of real short-term yield and inflation, we generate a risk-free rate forecast and, from that, derive all equity and fixed income assets by adding relevant risk premiums. We derive the risk premium for U.S. equities from the Gordon growth model as the sum of the dividend yield and the nominal earnings growth rate in excess of the risk-free rate. International equities include the addition of an international equity risk premium. Government-bond return forecasts are the sum of the risk-free rate and an appropriate term premium. Corporate bond return forecasts further include the addition of a credit-risk premium.
From a theoretical perspective, all risk premiums mean revert towards a long-run equilibrium as the economy is in a steady state. The reason for mean reversion is that investment opportunities are time varying. Since the rate of arrival of new information is time varying, return volatility and covariance are time varying as well in the short run. Our econometric work and that of academic researchers confirms the stationarity of a number of risk premiums, which in turn justifies our assumption of constant average risk premiums, term premiums and credit spreads in the long-run equilibrium. Our return forecasts are shown in Figure 3.
Figure 3. Long-Run Equilibrium Return Assumptions
Source: Voya Investment Management, as of November 2021. Assumptions are subject to change.
How We Forecast Returns
We derive asset-class return forecasts from the blend of the base and alternative case scenarios. Together, these capture the most important upside and downside risks which the global economy and markets will face over the forecast horizon. The base case forecasts growth to average 2.6% through 2031 driven by strong consumer spending, below-trend productivity growth and subdued labor force growth. The alternative scenario assumes slightly faster productivity growth, a higher dividend payout ratio, more inflation and an assumption that the Federal Reserve lets the economy run a little hotter than in the base case. Given these assumptions, returns to risky assets are modestly higher in the alternative scenario than in the base case.
For U.S. equities, we estimate earnings and dividends for the S&P 500 index using our blended macroeconomic assumptions. Earnings growth is constrained by the neoclassical assumption that profits as a share of GDP cannot increase without limit, but converge to a long-run equilibrium. We then use a dividend discount model to determine fair value for the index each year during the forecast period. We construct returns for other U.S. equity indices, including REITs, using a single index factor model in which beta sensitivities of each asset class with respect to the market portfolio are derived from our forward-looking covariance matrix estimation. For a detailed discussion of this, please see the Covariance and Correlation Matrices Methodology, in Methodological Considerations below. Each equity asset class return is the sum of the risk-free interest rate and a specific market-risk premium determined from our estimate of beta sensitivity and market-risk premium forecasts.
For U.S. bonds, we use the blended-scenario interest rate expectations to calculate expected returns for various durations. We model bond expected returns as the sum of current yield and a capital gain (or loss) based on duration and expected change in yields. For non-U.S. bonds, the process is similar and includes an adjustment for expected currency movements. Return expectations for credit-related fixed income reflect yield spreads and expected default-and-recovery rates.
Figure 4. Voya Investment Management 10-Year Returns Forecast, 2022
Source: Voya Investment Management. Returns shown are in U.S. dollar terms. Forecasts are subject to change.
Figure 5. Voya Expected Correlations Matrix
Source: Voya Investment Management, data as of November 2021. Projections are subject to change.
Methodological Considerations
Covariance and Correlation Matrices Methodology
Estimating asset-class covariance and correlation matrices are the underlying pillars of our asset-class standard deviation forecasts. This is a different process than forecasting returns, as evidence tells us that correlations wander through time. If we were to use a historical average or exponentially weighted methodology, which takes a long-run history and puts a heavier weight on recent observations, it could lead to risk forecasts that may be representative of the past but bear little resemblance to the future. Therefore, the forecast multiple asset-class risk summarized by the return covariance matrix is crucial to the capital market assumptions process.
A simple example using equities and bonds can illustrate this point (Figure 6). Over the past 20 years, the correlation of returns between the S&P 500 index and Barclays U.S. Aggregate Bond index was -0.02; however, this tells us nothing about the relationship of returns between these two asset classes during unusual periods or when financial markets are in very euphoric or pessimistic states. For example, during normal periods of returns over the same 20-year interval, the correlation between equities and bonds was -0.10, whereas during the unusual periods it was +0.07. Incorporating these periods of unusual correlation patterns can lead to a truer estimate of the durability of diversification between asset classes. We capture these unusual periods in our standard deviation and correlation forecasts in an academic framework called turbulence.
Figure 6. Normal and Turbulent Periods of Stock and Bond Correlations, 20 Years Ended 09/30/21
Source: Voya Investment Management. Data as of September 30, 2021.
Turbulence: Evolution of a Concept
The turbulence framework we use to estimate correlations and standard deviations of returns among asset classes is derived from the academic work of the applied statistician Prasanta Chandra Mahalanobis. In the early twentieth century, Mahalanobis analyzed human skull resemblances among castes and tribes in India. He created a formula to capture differences in skull size, which incorporated the standard deviation of measures of various skull parts. He then squared and summed the normalized differences, generating a single composite distance measure.3
This formula evolved into a statistical measure called the “Mahalanobis distance.” The measure was ground-breaking in that it helped analyze data across standard deviations but also incorporated the correlations among data sets. More than 60 years later, the Mahalanobis distance was used by Kritzman and Li to formulate a concept called financial turbulence.4 They postulated financial turbulence as a condition in which asset prices, given their historical patterns of returns, behave in an uncharacteristic way including extreme price moves. They further noted that financial turbulence often coincides with excessive risk aversion, illiquidity and price declines for risky assets. It is this turbulence framework (or unusualness of returns and correlations of returns) that we have used to forecast risk measures in our capital market assumptions.
Observing Turbulence
Turbulence can be calculated for any given set of asset classes. Back to our example of U.S. equities and bonds, the two dimensions can be visualized as the equation of an ellipse using the returns of the S&P 500 index and the U.S. Barclays Aggregate index (Figure 6). The center of the ellipse represents the average of the joint returns of the two assets. The boundary is a level of tolerance that separates normal from turbulent observations. The boundary takes the form of an ellipse rather than a circle because it takes into account the covariance of the asset classes. The idea captured by this measure is that certain periods are considered turbulent not only because returns are unusually high or low but also because they moved in the opposite direction of what would have been expected given average correlations.
Using Turbulence to Create Portfolios
Note that the threshold of normal and turbulence in Figure 6 is not static but rather is dynamic through time. Our process identifies turbulent market regimes by estimating a covariance matrix covering those periods of market stress alone and is the outcome of a Markov model. The model classifies regimes rather than arbitrary thresholds because thresholds would fail to capture the persistence of shifts in volatility. The Markov model output in Figure 7 illustrates turbulent and normal regimes.
Figure 7. Markov Normal and Turbulent Regimes over Time
Source: Voya Investment Management. Data as of September 30, 2021.
Turbulent market regimes make use of the concept of multivariate outliers in a return distribution. That is, we take into account not only the deviation of a particular asset class’ return from the average, but also its volatility and correlation with other asset classes. We subsequently estimate a covariance matrix based on periods of normal and turbulent market performance. Finally, we use a procedure to blend these two covariance matrices using weights that allow us to express both views about the likelihood of each normal or turbulent regime and to capture the differential risk attitudes toward each. The weights we use are 60% normal and 40% turbulent to create our strategic asset allocation portfolios.
We overweight the turbulent regime at 40% — higher than its observed frequency of 30% — to account for structural issues such as globalization, demographics and world-wide central bank intervention, which are prevalent today. From this blended covariance matrix, we then extract the implied correlation matrix and standard deviations for each asset class. In our view, this process helps create a strategic asset allocation portfolio that can account for the empirical evidence that correlations will deviate through time.
Time Dependency of Asset Returns and Its Impact on Risk Estimation
Recent research suggests that expected asset returns change over time in somewhat predictable ways and that these changes tend to persist over long periods. Thus, changes among investment opportunities — all possible combinations of risk and return — are found to be persistent. This Appendix will set out the economic reasons for return predictability, its consequences for strategic asset allocation and the adjustments we have made to control for it in our estimation process.
In our view, the common source of predictability in financial asset returns is the business cycle. The business cycle itself is persistent, and this makes real economic growth to some extent predictable. The fundamental reason for the business cycle’s persistence is that its components share the same quality. Consumers, for example, have a tendency to smooth consumption since they dislike abrupt changes in their lifestyles. Research on permanent income and lifecycle consumption provides the theoretical basis for consumers’ desire for a stable consumption path. When income is affected by transitory shocks, consumption should not change since consumers can use savings or borrowing to adjust consumption in well-functioning capital markets.
Robert Hall has formalized these ideas by showing that consumers will optimally choose to keep a stable path of consumption equal to a fraction of their present discounted value of human and financial wealth.5 Investment, the second component of GDP, is sticky, as corporate investment in projects is usually long term in nature. Finally, government expenditures also have a low level of variability. Over a medium-term horizon, negative serial correlation sets in as the growth phase of the cycle is followed by a contraction and then that contraction is followed by renewed growth.6
How does this predictability of economic variables affect the predictability of asset returns? Consider equities as an example. The value of equities is determined as the present discounted value of future cash flows and depends on four factors: expected cash flows, expected market risk premium, expected market risk exposure and the term structure of interest rates. Cash flows and corporate earnings tend to move with the business cycle. The market risk premium is high at business cycle troughs, when consumers are trying to smooth consumption and are less willing to take risks with their income; and low at business cycle peaks, when people are more willing to take risks. The market risk premium is a component of the discount rate in the present value calculation of the dividend discount model. A firm’s risk exposure (beta), another component of the discount rate, changes through time and is a function of its capital structure. Thus, a firm’s risk increases with leverage, which is related to the business cycle. The last component of the discount rate is the risk-free rate, which is determined by the term structure of interest rates. The term structure reflects expectations of real interest rates, real economic activity and inflation, which are connected to the business cycle. Thus, equity returns, and financial asset returns in general, are to a certain extent predictable. Expected returns of many assets tend to be high in bad macroeconomic times and low in good times.
This predictability of returns manifests itself statistically through autocorrelation. Autocorrelation in time series of returns describes the correlation between values of a return process at different points in time. Autocorrelation can be positive when high returns tend to be followed by high returns, implying momentum in the market. Conversely, negative autocorrelation occurs when high returns tend to be followed by low returns, implying mean reversion. In either case autocorrelation induces dependence in returns over time.
Traditional mean-variance analysis focused on short-term expected return and risk assumes returns do not exhibit time dependence and prices follow a random walk. Expected returns in a random walk are constant, exhibiting zero autocorrelation; realized short-term returns are not predictable. Volatilities and cross-correlations among assets are independent of the investment horizon. Thus, the annualized volatility estimated from monthly return data scaled by the square root of 12 should be equal to the volatility estimated from quarterly return data scaled by the square root of four. In the presence of autocorrelation, the square root of time scaling rule described above is not valid, since the sample standard deviation estimator is biased and the sign of autocorrelation matters for its impact on volatility and correlations. Positive autocorrelation leads to an underestimation of true volatility. A similar result holds for the cross-correlation matrix bias when returns exhibit autocorrelation. So for long investment horizons, the risk/return tradeoff can be very different than that for short investment horizons.
In a multi-asset portfolio, in which different asset classes display varying degrees of autocorrelation, failure to correct for the bias of volatilities and correlations will lead to suboptimal mean variance optimized portfolios in which asset classes that appear to have low volatilities receive excessive allocations. Such asset classes include hedge funds, emerging market equities and non-public market assets such as private equity or private real estate, among others.
There are at least two ways to correct for autocorrelation:
- A direct method that adjusts the sample estimators of volatility, correlation and all higher moments
- An indirect method that cleans the data first, allowing us to subsequently estimate the moments of the distribution using standard estimators
Given that the direct methods become quite complex beyond the first two moments, our choice is to follow the second method and clean the return data of autocorrelation. Before we do that, we estimate and test the statistical significance of autocorrelation in our data series.
We estimate first-order autocorrelation correlation as the regression slope of a first-order autoregressive process. We use monthly return data for the period 1979–2014. We subsequently test the statistical significance of the estimated parameter using the Ljung-Box Q-statistic.7 The Q-statistic is a statistical test for serial correlation at any number of lags. It is distributed as a chi-square with k degrees of freedom, where k is the number of lags. Here we test for first-order serial correlation, thus k = 1. About 80% of our return series exhibit positive and statistically significant first-order serial correlation based on associated p-values at the 10% level of significance.8 Khandani and Lo provide empirical evidence that positive return autocorrelation is a measure of illiquidity exhibited among a broad set of financial assets including small-cap stocks, corporate bonds, mortgage-backed securities and emerging market investments.9 The theoretical basis is that in a frictionless market, any predictability in asset returns can be immediately exploited, thus eliminating such predictability. While other measures of illiquidity exist, autocorrelation is the only measure that applies to both publicly and privately traded securities and requires only returns to compute.
Since the vast majority of the return series we estimate exhibits autocorrelation, we apply the Geltner unsmoothing process to all series. This process corrects the return series for first-order serial correlation by subtracting the product of the autocorrelation coefficient ρ and the previous period’s return from the current period’s return and dividing by 1-ρ. This transformation has no impact on the arithmetic return, but the geometric mean is impacted since it depends on volatility. This correction is thus important to make for long-horizon asset allocation portfolios.
Multi-Asset Strategies and Solutions Team
Voya Investment Management’s Multi-Asset Strategies and Solutions (MASS) team, led by Chief Investment Officer Paul Zemsky, manages the firm’s suite of multi-asset solutions designed to help investors achieve their long term objectives. The team consists of 25 investment professionals who have deep expertise in asset allocation, manager selection and research, quantitative research, portfolio implementation and actuarial sciences. Within MASS, the asset allocation team, led by Barbara Reinhard, is responsible for constructing strategic asset allocations based on their long-term views. The team also employs a tactical asset allocation approach, driven by market fundamentals, valuation and sentiment, which is designed to capture market anomalies and reduce portfolio risk.