article

Yields, uncertainty and expected returns

As a completely agnostic investor in a risk-free government bond, a sensible return expectation might be the yield on that bond.

The yield is indeed what you receive if you hold a bond to maturity. Any shorter term gains in excess of that starting yield are only bringing forward future returns; it simply means that subsequent returns will be lower in order to arrive at the same point.

However, over the last thirty years, any shorter term investor (i.e. not holding the bond to maturity) expecting yield to be an agnostic indicator of likely returns would have been consistently and materially surprised.

The chart below shows the starting yield on ten-year US Treasuries against the annual return implied by price gains and coupons over the next three years. For example, as shown on the chart, a ten-year Treasury in September 1990 was priced to deliver an annual yield of 8.8% over its life. However, the price appreciation of ten year Treasuries over the next three years actually resulted in a total return equivalent to 14.8% per annum.

Rarely have the returns over three years borne much resemblance to the starting point of yield. Such excess short-term returns have been the norm over the last thirty years. Figure 2 below shows how much more than the starting yield has been delivered by total returns on ten year Treasuries on a rolling three year basis (the difference between the two lines in figure 1).

It’s worth stressing that you’ve only experienced returns like these over the whole period if you’d constantly rebalanced your holdings back to a ten-year maturity.

Nevertheless, this environment has been a sensational outcome for investors; rarely have the three year returns fallen short of (or even equalled) the yield at which you bought them. This is not simply confined to treasuries. It’s true of gilts, German bunds and Japanese government bonds as well.

Yields and ‘Expected Returns’

The above is just a new way of illustrating what we all know: the trend decline in global bond yields has delivered significant capital gains. However, it also illustrates some important considerations about how we think about market efficiency.

Even for an asset such as US Treasuries, which can reasonably be assumed to have very low default risk and for which coupon payments are set out for all to see, the sense of an ‘expected return,’ even over multi-year periods can be a very loose one.

The relationship is even looser in equity markets:

Compare this to the Treasury chart again, but this time using the same scale:

In both cases, the starting yield line, can be seen to act as a valuation ‘anchor,’ however equities can depart far more meaningfully from a sense of ‘expected return’ than bonds simply because we have less of a sense what a genuine expected return is. To quote from a great note by Howard Marks on the concept of an equity risk premium:

“…a simple mathematical calculation will tell us exactly what the promised return on a bond is (albeit not the probability that it will be received), while coming up with the future return for a stock requires making some massive guesses about the far-off future.”

This challenge is even greater for currencies or commodities, where the lack of promised cash flows and/or a sense of a possible earnings stream make assessing expected return even harder.

Pricing Model Uncertainty and ‘Mean Reversion’

Being aware of these analytical challenges in forming a view of expected returns can help frame the assessment of behavioural episodes across different assets. ‘Pricing Model Uncertainty’  is the theory that assets that are harder to value will tend to depart further from any sense of ‘fair’ value, and that there can be a greater tendency for trending behaviour.

For investors who believe that ‘emotional’ forces can periodically drive markets and provide investment opportunities, this has important implications. For example, herding driven by fear or euphoria might be more likely in assets that are harder to value, such as equity, currency or commodities, than in those like bonds which have a far clearer valuation anchor. If we have little sense of what the true value of something is, the capacity to get caught up with the crowd is arguably greater.

This can give rise to potential opportunities. However, it can also mean that those assets with higher pricing uncertainty (i.e. harder to value) are less likely to gravitate toward a value anchor, and may certainly take longer to do so. Moreover, the degree of Pricing Model Uncertainty for the same asset can change over time. As a couple of possible examples:

  • It may be harder to assess earnings dynamics (and therefore expected returns) of companies if the benefits from network effects and ‘winner takes all’ dynamics prove to have a greater impact than they have been in the past. This may in part explain the dynamics behind the tech bubble of the late 1990s and or (perhaps) in the US today.
  • Developed market currencies it may be argued have less propensity to display ‘excess volatility’ when global economic policy is more co-ordinated than was the case in the 1970s and 80s.
  • The expected return anchors for peripheral European government bonds used to exert a relatively strong gravitational force but are now far looser:

These are merely possibilities, but they illustrate why investors not only have to form their own sense of an appropriate expected return (or ‘fair value’), but should also consider the very process by which they arrive at that assessment. Importantly, this may involve an acknowledgement that the difficulties associated with establishing an expected return can change the risk properties of an asset.

Several questions are worth asking:

  • Can we place more or less faith in our sense of expected return than in the past?
  • How differently should we treat price volatility in different asset classes?
  • Should we expect value to act as an anchor rapidly or only over the very long term?

The answers to all these questions will change on a case by case basis and should never be left to simple rules of thumb. Instead, the concept of Pricing Model Uncertainty could be a useful tool for framing these considerations.