We ask a simple question over and over again

  • Our investment process starts with a deceptively simple question: “What is the probability that equity returns will beat bond returns on a risk-adjusted basis?”
  • To answer it, we calculate something called the PRATER – the Probability of a Risk-Adjusted Total Excess Return.
  • We can ask exactly the same question about the returns of emerging and developed equity markets, or any two assets we want to choose between.
  • We need to be clear at the outset that we are only interested in total returns, not yield or capital appreciation or performance relative to an index.
  • There are literally dozens of ways in which we could calculate the PRATER, but nearly all of them require forecasts, but we don’t believe in forecasts (see investment philosophy).
  • We use the most recent price data, to construct a cumulative frequency distribution of total returns based on the last 52 weeks’ return and standard deviation.

We always adjust for risk

  • Using two distributions, one for equities and one for bonds, we can calculate the simple probability of equities beating bonds.
  • But we also need to be paid for the extra risk of investing in equities – often called the equity risk premium.
  • This acts as a hurdle rate and automatically reduces the probability of equities beating bonds.
  • There are many way of estimating the equity risk premium. Some are based the long run forecast returns of equities and bonds. Others use the historic excess returns of equities relative to bonds.
  • We use a very simple identity. Extra risk means extra risk, in other words excess volatility – the volatility of equities minus the volatility of bonds.
  •  This number will vary depending on market conditions, but that’s the whole point. Change in risk appetite is one of the main drivers of equity and bond performance.

We set simple rules

  • The purpose of calculating any probability is to improve decision-making, so we set a very simple rule.
  • The PRATER is the weight of equities in a two asset portfolio, the rest is bonds.
  • This number is not stable (non-stationary), so the model has to be rebalanced once a week, but this is not a disadvantage.
  • Using the latest data ensures that the model is in tune with current market conditions, not some five year rolling average, or conditions as we would like them to be.
  • We reduce the sensitivity to base effects and rogue data by using multiple sample periods.
  • Convention would typically start with an annual or semi-annual period, but at the time of investment we cannot know which sample period will turn out to be the best predictor of the future, so we need to take the average of several periods.

We aim for higher returns with lower volatility

  • Building a portfolio using the weights recommended by a PRATER model typically produces higher long run returns than either of the underlying assets, and better risk-adjusted returns.
  • We have tested this process using total return indices for the equity markets of over forty different countries.
  • We typically achieve higher absolute returns than either the high or the low risk asset or any fixed combination of the two, and…
  • … higher risk-adjusted returns than either the high or the low risk asset or any fixed combination of the two, and…
  • .. smaller and shorter peak to trough drawdowns than the high-risk asset and most fixed combination of the two, and…
  • ..volatility which is  in line with, or lower than, a fixed 50/50 combination of the two assets.
  • Full details of all these back-tests are available on request.

We use simple tools to build multi-asset portfolios

  • Institutional investors are not restricted to a choice between equities and bonds in just one country.
  • To build a multi-asset portfolio, we need to create a decision tree and a chain of conditional probabilities.
  • We can use the PRATER process on any pair of assets. We can calculate the probability that emerging market equities will beat US equities, or the probability that US corporate bonds will beat US government bonds.
  • To create a conditional probability chain, we just link them by calculating the probability that emerging market equities will beat US equities, given the probability that US equities beat US bonds, and so on.
  • The more assets classes we want to incorporate, the longer the chain of conditional probabilities and the more complex the decision tree, but the basic principles don’t change.
  •  We just need to make sure that we calculate the conditional probabilities of all the different orders in which the asset classes can be ranked.

We have a transparent process which delivers clear decisions

  • All our models are subject to a 15 year back-test before we publish them.
  • At the end of the process we get a clear numerical solution to the question we asked at the beginning.
  • The process is transparent and rules-based, and produces a clear set of actionable recommendations.
  • Maximising returns and minimising volatility are given equal importance at all times.
  • We do not need to put a risk overlay on top of the investment recommendations, because they have already been selected on the basis of their return per unit of risk.
  • We can calculate a number for value at risk, but we don’t need it to control risk.
  • If we want to know why the asset mix is moving in a particular direction, we can check any part of the chain, look at its individual return per unit of risk, and see how it links into the wider portfolio.