How can you cyclically adjust when there is no cycle? The Shiller P/E and the US market

11 min read 16 Jun 17

Summary: The Shiller price-to-earnings ratio (or ‘cyclically adjusted price-to-earnings ratio’) has become the poster child for bubble-hunters everywhere. Ever since Robert Shiller himself became associated with the boom and bust, investors have looked to the metric as an indicator of irrational exuberance in the markets.

So why is Shiller himself apparently so sanguine about a metric which has reached levels not previously seen since the start of the tech boom?

Shiller makes the simple observation that the last time stocks were at similar levels they went on to perform extremely strongly for a number of years.

This may make many investors uncomfortable. Does this mean we are now entering a game of ‘greater fool,’ riding the rally with the simple hope that we can get out of the market before the inevitable crash? Or is Shiller making a more nuanced point?

Naïve mean reversion

Shiller is saying that the excess returns on US stocks over inflation look likely to be low but positive over the next ten years:

“Stocks have generally outperformed other investments through history. They are highly priced now, which means I don’t expect them to outperform so much. But for a long term investor… there should be a place for stocks in a portfolio – and they could go up a lot…

…they could also go down a lot”

This is very different to the argument that ‘stocks must go down simply because value metrics are high.’ As we have noted before the history of equity valuation is not a steady undulating wave of mean reversion. Instead it looks like a series of regimes.

Meanwhile, anyone who had spent the last few decades backing mean reversion in fixed income would likely be out of business now:

Irrespective of the cause of the trends shown above, it seems clear that investors need to be wary of placing too much faith in the mean as a useful measure, particularly over short time horizons. Critics will argue that QE is distorting the bond market (and the equity market), but that does not explain the twenty-five year trend prior to the financial crisis or the continued decline in rates after purchases were halted in October 2014.

It seems that there is a structural dynamic in global interest rates and this will clearly have an impact on the level of compensation that investors require for equity risk. This has been discussed by many, so I won’t rehash the arguments here; macro man’s blog provides as good a summary of the issues as any.

Cyclically adjusted?

When the idea of cyclically adjusting stocks was proposed by Benjamin Graham, the intention was to assess how a particular company’s earnings would be able to stand up in the face of different market conditions; it was not intended as a short term timing tool for the entire stock market.

Even when Shiller applied the approach to the entire market, the objective was the same: to get as good a sense as possible of the true earnings power of the corporate sector. Why should ten years be the right amount of time to do this? The nature of the economic cycle today is very different to that which prevailed when Graham was writing, and so is the nature of the companies that exist.

The US economy used to look cyclical when it was more domestically focused and manufacturing dominated. At that time, rolling a ten year period seemed as likely as any other to pick up both expansions and recessions in a way that was relevant to those companies.

Today ‘boom and bust’ has been replaced by ‘crawl and crash.’ Technical advances and the changing structure of the economy have dampened the extent to which manufacturing over- and underproduction drive the cycle, while the economy is also more diversified. In a manufacturing economy impacts on the ‘real economy’ are more correlated (for example when employment is dominated by workers in large factories which all struggle at the same time) than in a diversified service economy.

Similarly the natures of the companies that exist today are very different. Most obvious is the fact that earnings are global in nature and less subject to the cyclical dynamics of a single country. Global economic cycles are somewhat diversified meaning that ten years won’t necessarily capture a full cycle (Australia hasn’t had a recession for twenty-five years for example). Instead synchronised hits to global earnings appear to be the result of periodic crises (in fact as Noah Smith has noted, thinking of economic cycles as being like waves has always been a misleading abstraction).

Of course, the US market is expensive on other measures, not just the Shiller cyclically-adjusted p/e. The above merely highlights that, when looking at the history of any financial series, one has to consider whether you are really comparing like with like. Which brings me to my final point.

Comparing apples with apples

As the above highlights, the nature of the stocks that make up the US equity market, and how they operate, is very different to twenty years ago. On a very simplistic level, IT, which typically trades on higher multiples, was 12.9% of the index in 1997 and is now 22.3% of the index. At the same time two of the traditionally ‘more expensive’ sectors, Consumer Staples and Industrials, have fallen from 14.1% to 9.2% and 12.2% to 9.7% respectively. Even the business models of the companies within the major sectors have changed. The leaders of the tech sector are very different to those who led the sector at the end of the 1990s.

Lastly, the way that the earnings are measured throughout the history of the US stock market is not consistent over time. Considering changes in accounting standards, or whether a company chooses to issue dividends or buyback shares, is far less interesting than forecasting the next crash but they are significant elements behind why today’s Shiller p/e is not a fair comparison with the same measure thirty years ago.

This time it’s different

All of this may sound like classic ‘this time it’s different’ reasoning. The reality is that this time is always different. What does not change, however, is the fact that we can never delegate thought to an equation. Last month I noted a recent paper on value investing which highlighted just this point.

No purely quantitative approach can answer the ultimate questions investors need to ask themselves: What returns do you think you will receive, and how comfortable are you with a) the chance that you are wrong, and b) how bumpy the journey is going to be?

If you believe in mean reversion in valuations you ultimately need to consider why this will happen: prices may fall because investors lose faith that earnings can be delivered , or simply decide they want a higher compensation for access to that earnings stream (perhaps due to a rising discount rate in other assets). Today the threats of these forces being relevant in the US seem higher than they have for some time, but this need not necessarily be associated with an imminent crash (the risks of which are always present). Perhaps tellingly, no-one is giving much credence to the possibility that valuations become cheaper because earnings grow faster than prices.

There are always a range of possible outcomes, this is the nature of investment risk and why investors can be rewarded for bearing it. In the case of the US, the compensation on offer for  taking on risk is undoubtedly lower than it has been. Fortunately global markets generally give you an opportunity to diversify this risk, and today is no exception.

By Stuart Canning

The value of investments will fluctuate, which will cause prices to fall as well as rise and you may not get back the original amount you invested. Past performance is not a guide to future performance.

Related insights