Risk is one of the most talked about topics in investment. Organizations spend considerable resources to measure it, understand it and, most importantly, take it. Yet, often when it matters most the tools of risk management do not provide the answers they need.
Maybe these failing are due to risk being treated as a generic property of assets and markets. Instead, what if we view risk as a manifestation of our personal (or organisational) ability to forecast accurately the future. Instead of investment risk being the variation in price it is our inability to forecast.
A forecast is a belief about a future outcome: that belief may be simple (that a security will increase to a certain price, for example, or that one security will outperform another over some specific time horizon) or complex and specific (that the value of a security will be drawn from a certain distribution, with specified parameters).
This perspective defines risk as being our prediction error. If we can predict an asset’s future price accurately then it is not risky, no matter how volatile its price may be.
Of course, precise prediction of the future state of the markets (or any complex adaptive system) remains extremely hard, but viewing risk as prediction error changes what risk is. Riskiness is no longer only a characteristic of the assets we own, it is linked to our prediction capabilities.
Importantly, it suggests that individuals and organisations should establish where their prediction edge is and is not, and then build portfolios that exploit this prediction edge (ie, be paid to own assets that others perceive as risky but which they can forecast) and are robust to the asset price movements they cannot forecast.
Relating riskiness to prediction ability seems intuitive but it is not how most investors think about investment risk.
Many criticisms, few solutions
Many investors (and traditional finance theory) use volatility, or similar metrics, as a measure of risk. The ‘risk’ that is being measured is the variability of price movements over a period. This approach says an asset with a high volatility (say an individual stock) is necessarily more risky that an asset with low volatility (say a government bond).
Other investors criticise the use of volatility as a measure of risk[1]. Risk, some investors say, is not price decline but permanent impairment of capital. For these investors, an asset with a lower potential for permanent capital impairment is less risky than an asset with a higher potential for capital impairment[2].
These views of what risk is have strengths and weaknesses but they also seem to be in conflict with each other. Relating risk to prediction ability unifies these competing perspectives around a single concept.
Risk is linked to our predictive ability
As an alternative perspective on risk, what if we simplify the idea of risk to that of being wrong in our prediction about the future.
This connects risk and our ability to forecast. When our predictions are perfect, there is no risk, and when something is unpredictable, it is risky.
Let us explore this idea using a stock and a bond. First, suppose we have a prediction error of zero (perfect prediction) for the daily return of each asset. In this case, is the stock or the bond riskier? Clearly, in this case, they are both riskless investments because we can predict their daily price changes perfectly.
Changing our abilities, let us say that our ability to predict bond prices is no longer perfect; we have a small prediction error. Are both assets still riskless? No, in this this scenario the bond is risky (we occasionally get our prediction wrong and lose money) but the equity remains riskless as our perfect predictions mean we never lose money.
The conventional view is that stocks are riskier than bonds, but when we adopt this alternative definition of risk, we see that bonds are not necessarily inherently safer than stocks: it all depends on our prediction abilities.
As our forecasting ability declines, our prediction error is increasingly related to volatility . If we have no prediction ability for either the stock or the bond then the stock is riskier due to the likely larger prediction error arising from the higher volatility of the stock. Essentially, if we are wrong about the stock’s future price it is likely to be a bigger error than if we are wrong about the bond’s future price.
Some investors might think that because they try to own the market portfolio they are not making any forecasts and therefore this notion of risk as prediction error does not apply. However, it is worth considering two things: Firstly, the portfolio still contains forecasts, the forecasts it contains are the money-weighted average forecasts of all market participants. Secondly, it is presumably the case, in general, that the overall portfolio is expected to appreciate in value – so the approach is based on at least one forecast about the future.
Separating the predictions by time horizon
The above examples are based on daily predictions: ie, assuming we have forecasts of all future periods. However, we should not expect prediction ability to be uniform across time horizons.
For an unleveraged investor this lack of uniformity may not present a challenge if they have the ability to withstand any adverse price movements within their prediction horizon, and their prediction is correct.
Where the accuracy of a prediction is not certain, or the investor is leveraged, then the journey matters. This naturally supports the notions that sizing investments and diversifying the predictions in a portfolio is important.
However, it subtly changes the point of diversification. It suggests that diversifying the sources of prediction error is what matters. This may not be the same as diversifying across traditional asset classes.
A different perspective on risk
Viewing riskiness as our prediction ability gives a different perspective compared to a traditional volatility-based approach. It also helps to reconcile the conflicting views of risk as volatility or capital impairment by explicitly incorporating the idea of our confidence in our forecast and prediction ability at different time horizons. Rather than conflicting notions of risk, we see these ideas are related by differing confidence in prediction ability between investors with different perspectives.
When an investment’s riskiness is linked to the forecasting ability of an investor, risk is no longer absolute but different for each individual and organisation. It naturally follows that different organisations should have different portfolios that reflect their differing prediction abilities.
[1] We will leave discussion of volatility being an inappropriate measure of potential variation for power-law-like distributions or those without a finite second moment for another day. [2] The potential for a realised capital impairment will be affected by the use of leverage, liquidity requirement and the time horizon of the investor.