SUMMARY
The following are the main points covered in the reading.
The choice among forecasting techniques is effectively a choice of the information on which forecasts will be conditioned and how that information will be incorporated into the forecasts.
The formal forecasting tools most commonly used in forecasting capital market returns fall into three broad categories: statistical methods, discounted cash flow models, and risk premium models.
Sample statistics, especially the sample mean, are subject to substantial estimation error.
Shrinkage estimation combines two estimates (or sets of estimates) into a more precise estimate.
Time-series estimators, which explicitly incorporate dynamics, may summarize historical data well without providing insight into the underlying drivers of forecasts.
Discounted cash flow models are used to estimate the required return implied by an assetâs current price.
The risk premium approach expresses expected return as the sum of the risk-free rate of interest and one or more risk premiums.
There are three methods for modeling risk premiums: equilibrium models, such as the CAPM; factor models; and building blocks.
The DCF method is the only one that is precise enough to use in support of trades involving individual fixed-income securities.
There are three main methods for developing expected returns for fixed-income asset classes: DCF, building blocks, and inclusion in an equilibrium model.
As a forecast of bond return, YTM, the most commonly quoted metric, can be improved by incorporating the impact of yield changes on reinvestment of cash flows and valuation at the investment horizon.
The building blocks for fixed-income expected returns are the short-term default-free rate, the term premium, the credit premium, and the liquidity premium.
Term premiums are roughly proportional to duration, whereas credit premiums tend to be larger at the short end of the curve.
Both term premiums and credit premiums are positively related to the slope of the yield curve.
Credit spreads reflect both the credit premium (i.e., additional expected return) and expected losses due to default.
A baseline estimate of the liquidity premium can be based on the yield spread between the highest-quality issuer in a market (usually the sovereign) and the next highest-quality large issuer (often a government agency).
Emerging market debt exposes investors to heightened risk with respect to both ability to pay and willingness to pay, which can be associated with the economy and political/legal weaknesses, respectively.
The GrinoldâKroner model decomposes the expected return on equities into three components: (1) expected cash flow return, composed of the dividend yield minus the rate of change in shares outstanding, (2) expected return due to nominal earnings growth, and (3) expected repricing return, reflecting the rate of change in the P/E.
Forecasting the equity premium directly is just as difficult as projecting the absolute level of equity returns, so the building block approach provides little, if any, specific insight with which to improve equity return forecasts.
The SingerâTerhaar version of the international capital asset pricing model combines a global CAPM equilibrium that assumes full market integration with expected returns for each asset class based on complete segmentation.
Emerging market equities expose investors to the same underlying risks as emerging market debt does: more fragile economies, less stable political and policy frameworks, and weaker legal protections.
Emerging market investors need to pay particular attention to the ways in which the value of their ownership claims might be expropriated. Among the areas of concern are standards of corporate governance, accounting and disclosure standards, property rights laws, and checks and balances on governmental actions.
Historical return data for real estate is subject to substantial smoothing, which biases standard volatility estimates downward and distorts correlations with other asset classes. Meaningful analysis of real estate as an asset class requires explicit handling of this data issue.
Real estate is subject to boomâbust cycles that both drive and are driven by the business cycle.
The cap rate, defined as net operating income in the current period divided by the property value, is the standard valuation metric for commercial real estate.
A model similar to the GrinoldâKroner model can be applied to estimate the expected return on real estate: E(Rre) = Cap rate + NOI growth rate â %ÎCap rate.
There is a clear pattern of higher cap rates for riskier property types, lower-quality properties, and less attractive locations.
Real estate expected returns contain all the standard building block risk premiums:
Term premium: As a very long-lived asset with relatively stable cash flows, income-producing real estate has a high duration.
Credit premium: A fixed-term lease is like a corporate bond issued by the leaseholder and secured by the property.
Equity premium: Owners bear the risk of property value fluctuations, as well as risk associated with rent growth, lease renewal, and vacancies.
Liquidity premium: Real estate trades infrequently and is costly to transact.
Currency exchange rates are especially difficult to forecast because they are tied to governments, financial systems, legal systems, and geographies. Forecasting exchange rates requires identification and assessment of the forces that are likely to exert the most influence.
Provided they can be financed, trade flows do not usually exert a significant impact on exchange rates. International capital flows are typically larger and more volatile than trade-financing flows.
PPP is a poor predictor of exchange rate movements over short to intermediate horizons but is a better guide to currency movements over progressively longer multi-year horizons.
The extent to which the current account balance influences the exchange rate depends primarily on whether it is likely to be persistent and, if so, whether it can be sustained.
Capital seeks the highest risk-adjusted expected return. In a world of perfect capital mobility, in the long run, the exchange rate will be driven to the point at which the expected percentage change equals the âexcessâ risk-adjusted expected return on the portfolio of assets denominated in the domestic currency over that of the portfolio of assets denominated in the foreign currency. However, in the short run, there can be an exchange rate overshoot in the opposite direction as hot money chases higher returns.
Carry trades are profitable on average, which is contrary to the predictions of uncovered interest rate parity.
Each country/currency has a unique portfolio of assets that makes up part of the global âmarket portfolio.â Exchange rates provide an across-the-board mechanism for adjusting the relative sizes of these portfolios to match investorsâ desire to hold them.
The portfolio balance perspective implies that exchange rates adjust in response to changes in the relative sizes and compositions of the aggregate portfolios denominated in each currency.
The sample varianceâcovariance matrix is an unbiased estimate of the true VCV structure; that is, it will be correct on average.
There are two main problems with using the sample VCV matrix as an estimate/forecast of the true VCV matrix: It cannot be used for large numbers of asset classes, and it is subject to substantial sampling error.
Linear factor models impose structure on the VCV matrix that allows them to handle very large numbers of asset classes. The drawback is that the VCV matrix is biased and inconsistent unless the assumed structure is true.
Shrinkage estimation of the VCV matrix is a weighted average of the sample VCV matrix and a target VCV matrix that reflects assumed âpriorâ knowledge of the true VCV structure.
Failure to adjust for the impact of smoothing in observed return data for real estate and other private assets will almost certainly lead to distorted portfolio analysis and hence poor asset allocation decisions.
Financial asset returns exhibit volatility clustering, evidenced by periods of high and low volatilities. ARCH models were developed to address these time-varying volatilities.
One of the simplest and most used ARCH models represents todayâs variance as a linear combination of yesterdayâs variance and a new âshockâ to volatility. With appropriate parameter values, the model exhibits the volatility clustering characteristic of financial asset returns.
Last updated