John Huss­man..... Anatomy of Bubbles

John Huss­man is is once again explor­ing the topic of val­u­a­tions and expected future returns. Please con­sider this snip from Anatomy of a Bubble

When val­u­a­tions are rea­son­able, investors can expect sat­is­fac­tory long-term returns sim­ply on the basis of the stream of cash flows they receive over time. But once val­u­a­tions are ele­vated, investors become increas­ingly reliant on pure increases in prices and val­u­a­tions in order to achieve sat­is­fac­tory returns. This is eas­ily seen in his­tor­i­cal data for the S&P 500.

The chart below is based on post-war U.S. data, and illus­trates the inter­ac­tion between val­u­a­tion lev­els and val­u­a­tion changes in pro­duc­ing long-term total returns for the S&P 500, and expands on some of Mike Shedlock’s recent obser­va­tions on val­u­a­tions and prospec­tive returns.

As I’ve fre­quently noted, Depression-era data is far more hos­tile than post-war data, as is data sur­round­ing other his­tor­i­cal and inter­na­tional credit crises, so investors would have needed more strin­gent val­u­a­tion cri­te­ria in order to accept mar­ket risk dur­ing these peri­ods. Still, post-war data is suf­fi­cient to con­vey some impor­tant ideas.

The first two columns below reflect the Shiller P/E (also known as the “cycli­cally adjusted” price/earnings ratio), and the fre­quency of var­i­ous P/E ranges in his­tor­i­cal data. The next two columns show the aver­age annual total return of the S&P 500 Index, based on whether the Shiller P/E rose to a higher level dur­ing the 10– or 5-year hori­zon, or whether it fell to a lower level by the end of that hori­zon. The next col­umn is the per­cent­age of obser­va­tions where the P/E was higher at the end of the hori­zon, and the last col­umn is the weighted aver­age return for each level of Shiller P/Es.

Notice that regard­less of whether P/E ratios rose or fell dur­ing these invest­ment peri­ods, sub­se­quent returns were sub­stan­tially higher from low val­u­a­tions than from high ones. Of course, sub­se­quent returns were higher for hori­zons where the P/E increased than for those where the P/E fell.

Since data is avail­able for 5-year returns through 2006, but only through 2001 for 10-year returns, the two tables cover slightly dif­fer­ent hori­zons since 1940. Notably, the fre­quency of Shiller P/Es greater than 24 is 15.3% in 5-year data but only 9.3% in 10-year data. This is because extreme val­u­a­tions have been the norm in recent years (where the P/E has exceeded 24 over half the time, inter­rupted only briefly by the recent plunge and rebound).

Presently, the Shiller P/E stands at 24. Be care­ful how you inter­pret the data in the table for Shiller P/E’s above 24, since these lev­els were almost never observed in data prior to the late-1990’s mar­ket bub­ble. You can see the odd effect of the bub­ble on the P/E cat­e­gories above 20. The recent ten­dency for high val­u­a­tions to move even higher over the short-term, cou­pled with the rapid recov­ery of much of the 2008–2009 loss, cre­ates a “hump” in the 5-year pro­file — aver­age returns first decline as val­u­a­tions increase, and then actu­ally improve for the 20–24 bracket. This is an arti­fact of recent years, and appears nei­ther in pre-1995 data nor in 10-year return data.

The impli­ca­tion of this data for long-term returns is clear. With the Shiller P/E presently at 24, we observe about the same impli­ca­tions for 10-year S&P 500 total returns as we obtain from our broader val­u­a­tion meth­ods (expected total returns aver­ag­ing about 3.5% annu­ally). Still, the actual course of total returns will depend on whether val­u­a­tions become even more extended over the next 5–10 years, or if they con­tract instead. Even if one includes data from the late-1990’s bub­ble, the prob­a­bil­ity of ris­ing P/E mul­ti­ples from these lev­els is less than 1-in-5.

The post of mine John Huss­man referred to is Neg­a­tive Annu­al­ized Stock Mar­ket Returns for the Next 10 Years or Longer? It’s Far More Likely Than You Think

Thanks John, I appre­ci­ate the mention.

Here is another chart I would like every­one to consider.

click on chart for sharper image

The above chart was pro­duced by an asso­ciate of mine, JJ Abodeely.

Please see his post Expen­sive Mar­kets Mean Low or Neg­a­tive Prospec­tive Return for addi­tional charts and com­men­tary includ­ing a chart show­ing “real” inflation-adjusted returns.

Data for the charts is from Robert Shiller.

The chart above shows the fre­quency dis­tri­b­u­tion of 10-year annu­al­ized gains (div­i­dends rein­vested) when the start­ing PE was in the range 22–24.

Fre­quency rep­re­sents the num­ber of occur­rences, total­ing 66 as opposed to a per­cent­age that would total 100%.

There have been 66 monthly occur­rences of PEs in the range 22–24 since 1881. The 10-Year annu­al­ized returns were only hugely pos­i­tive a grand total of 6 of those 66 times, and all of those occur­rences hap­pened between 1995 and 2005.

This is what Huss­man referred to when he said “Be care­ful how you inter­pret the data in the table for Shiller P/E’s above 24, since these lev­els were almost never observed in data prior to the late-1990’s mar­ket bubble.”

There are 33 occur­rences 1% and below, and 33 occur­rences 2% and above. The aver­age nom­i­nal annu­al­ized return is 1.2%

On that basis I think Huss­man is likely a bit high with his esti­mate of 3.5% annu­al­ized returns for the next 10 years. How­ever, rather than pick­ing a pre­cise num­ber, I pro­pose a range of –2% to +3%, with a skew to the down­side for the next 5 years or so.

Mean­while, pen­sion plan assump­tions are 7.5–8.5% every­where you look. Odds of that are minus­cule in my opinion.

Com­mod­ity Bubble

Let’s return to Huss­man for a look at the bubbles.

In the stock mar­ket, I believe that there is indeed a “bub­ble” com­po­nent in cur­rent prices, but it is not nearly as large as we observed in the approach to the 2000 peak, nor as extreme as we observed on the approach to the 2007 peak. My hope is that investors have learned some­thing. That’s not entirely clear, but we’ll be as flex­i­ble as we can while also being mind­ful of the risks.

While my view is that bub­ble com­po­nents can come and go in the mar­kets, they some­times become so large and well-defined that they take on a very dis­tinct pro­file. Such bub­bles included the advance to the 2000 stock mar­ket peak, the hous­ing bub­ble, the advance in oil prices to their peak in 2008, the advance in the Nikkei in the late 1980’s, and other clearly par­a­bolic advances.

On that note, it’s clear to me that we’re see­ing clas­sic bub­bles in a vari­ety of com­modi­ties. It is very unlikely that this is sim­ply due to global demand growth. Even with an exhaustible resource, it is a well-known eco­nomic result (Hotelling’s rule) that the opti­mal extrac­tion rule is one where the price rises at a rate not much dif­fer­ent from the inter­est rate. What we’ve seen lately is com­mod­ity hoard­ing, pre­dictably result­ing from neg­a­tive real inter­est rates pro­voked by the Fed’s pol­icy of quan­ti­ta­tive easing.

For­tu­nately for the world’s poor, the spec­u­la­tive dynamic that has cre­ated a mas­sive surge in com­mod­ity prices appears very close to run­ning its course, as we see very sim­i­lar “micro­dy­nam­ics” in agri­cul­tural com­modi­ties as we saw with oil in 2008. That’s not to say that we have a good idea of pre­cisely how high prices will move over the short term. The blowoff phase of a bub­ble tends to be steep, but so short-lived that it affords lit­tle oppor­tu­nity to exit. As prices advance in an uncor­rected parabola, the one-sided nature of the spec­u­la­tion typ­i­cally gives way to a fran­tic effort of spec­u­la­tors to exit simul­ta­ne­ously. Crashes are always a reflec­tion of illiq­uid­ity in two-sided trad­ing — the inabil­ity of sell­ers to find eager buy­ers at nearby prices.

On the sub­ject of com­modi­ties, it’s a nat­ural ques­tion whether gold falls into the same cat­e­gory as agri­cul­tural com­modi­ties. After all, gold and other hard assets have an impor­tant role as an alter­na­tive to money to store value, and it appears clear that the world is mon­e­tiz­ing in a way that is unlikely to be fully reversed even if pol­icy mak­ers wish to do so down the road.

In my view, it’s not clear that gold is in a bub­ble here, but it will be impor­tant to watch for the ear­marks of a clas­sic bub­ble. Below, I’ve plot­ted the price of gold against a “canon­i­cal” log-periodic bub­ble. Already, we’re see­ing some behav­ior that is char­ac­ter­is­tic of a bubble-type advance. A Sornette-type analy­sis gen­er­ates a finite-time sin­gu­lar­ity as early as April, but there are other fits that are con­sis­tent with a more sus­tained advance. If we observe a vir­tu­ally uncor­rected advance toward about 1500 in the next sev­eral weeks, the steep and uncor­rected advance would imply an increas­ing haz­ard probability.