Tell Me Again About Your Models

Tell Me Again About Your Models

Despite (and often because of), my overtly pessimistic take on the plight of humanity, I find occasion to laugh more often than I probably should throughout the day.

Understatements are a perpetual source of amusement for me. People make them all the time, especially in the market context. On Tuesday, for example, while summarizing new research from Ned Davis’s Ed Clissold, one Bloomberg reporter wrote that “assessing the exact probability” of an equity correction that “result[s] in a mini bear-market” between now and the end of the year “isn’t easy.”

No, it surely isn’t. In fact, it’s basically impossible.

As I’m always keen to remind folks, if anyone had anything like a reliable model for “assessing the exact probability” of equity corrections, you wouldn’t know about it. They’d have built it into an algo, hit “go” and retired to… I don’t know, and island somewhere. (Wink, wink.)

That said, Clissold reckoned that, based on a pair of models that purport to gauge the behavior of US stocks during comparable cycles, one of the two suggests “a median S&P 500 drop of 18% over 6.8 months.”

You can go ahead and laugh. And not because it’s necessarily wrong. Rather, because without exception, attempting that kind of precision (both in terms of guessing the magnitude and timing of a pullback) is an exercise in abject futility.

What is the S&P 500 (or any benchmark, for that matter)? The S&P 500 is an index which, stripped of the niceties, just measures the trajectory and relative value of corporate profits (which capture the business cycle) plus a premium (discount) that corresponds to the current level of euphoria (despair) among market participants, adjusted for hedging flows and buying (selling) associated with a hodgepodge of systematic strategies that adjust exposure based on volatility. That’s pretty much it.

You can forecast (and discount) profits and you can model the mechanical flows, but you can’t accurately predict when, why or even how humans will experience ecstasy and depression.

That definition explains why “buy and hold” works so well and also why people like Nomura’s Charlie McElligott can nail tactical call after tactical call, but also why, despite our best efforts, we’re all generally bereft when it comes to projecting where equities will be in six months, one year, or even three years.

On a longer horizon, simply being confident that capitalism will plod along, generally unimpeded, is a good enough rationale for owning an index fund. On a very short horizon, being able to model systematic flows can help you predict (or at least make sense of) the daily zigs and zags. But nobody (and no model) can help you very much in the medium-term, because that’s where manias and fear live, and you can’t model those.

In any case, that’s a window into the kinds of things I muse about internally on days like these — days when folks are just waiting around on the oxymoronic “important non-event” or, as Paul Tudor Jones billed the June FOMC during a pointless CNBC cameo, “the most important Fed meeting in Jay Powell’s career.”

Writing in his latest, Deutsche Bank’s Aleksandar Kocic noted that “the Fed’s shift to average inflation targeting came without a firm commitment to a particular window of averaging [and] by that maneuver, the [Committee] retained an option to select the window that would ex-post exonerate its stance without affecting its credibility.” That’s some insightful analysis. “This offers the Fed considerable flexibility given the trajectory of inflation in the last decade,” Kocic added.

Parsing the data is still largely a pointless exercise, if for no other reason than the Fed is sure to ignore it, at least until another two NFP prints are on the books. Retail sales missed on Tuesday except, when you take account of revisions to April’s (no longer disappointing) report, it was a wash. PPI came in hot, but despite being a technical “surprise” (i.e., above consensus), nobody was literally surprised. Empire missed, but prices paid fell. And so on. Someone called it a “mixed bag.” I’ll call it “noise.”

Commodities are coming off the boil. Lumber is down ~40% from the highs and copper has cooled, but then there’s crude, which traded at the highest levels since October 2018 Tuesday (figure below, with breakevens, which are down from multi-year peaks).

Apparently, Trafigura CEO Jeremy Weir thinks it’s possible oil will trade in the triple-digits again thanks, in part anyway, to underinvestment. $100 is a long way from negative $35.

What does it all mean? How does it all fit together?

Who knows. There are conflicting signals everywhere you look, efforts to craft coherent narratives notwithstanding. The Fed told you on Wednesday just how confounding it all is.

Writing earlier this week, Goldman’s Peter Oppenheimer noted that “investors didn’t really expect the high levels of inflation in the early 1970s to persist, at least to begin with [and] they were willing to accept a relatively low yield compared with reported inflation as it began to spike.” That’s where we are now — with a disconnect between rates and realized inflation.

“CPI continues to rise and rates continue to ignore it,” Deutsche’s Kocic remarked, in the same noted cited above. “The new CPI print and subsequent (absence of) reaction of rates underscore the widening of the formal disconnect between the two as well as demand for conditional interpretation of the CPI index in the context of the post-COVID developments,” he added.

Commenting further on the last real inflationary episode in the US, Goldman’s Oppenheimer recalled that “it took some time for expectations to adjust to the reality that a high inflation regime was likely to last.”

Once inflation was embedded, investors were similarly reluctant to accept that it was likely to recede. The figure (below, from Goldman) gives you some additional historical context.

“By the time inflation started to come down in the early 1980s, investors were doubtful that this was the start of a new trend,” Oppenheimer went on to say. “Bond yields remained much higher than reported inflation for a long time.”

Now, tell me again about your models and predictions.

6 thoughts on “Tell Me Again About Your Models

  1. Lots of billionaires get air time but billionaires who were pioneers often don’t. How often do you hear from Jim Simons? (Of course, why would he want public attention?) If net worth means anything in regards to predictions, Jim Simons is worth three Paul Tudor Jones. It is all noise. I would listen to predictions by the philosopher Thales (on olive presses) or Simons but not anybody on CNBC (or Bloomberg).

  2. Powell has repeatedly said (reading between the lines), that he is NOT raising interest rates….almost REGARDLESS OF WHAT HAPPENS WITH INFLATION ( sorry about the All Caps, but I just finished rereading ‘A Prayer for Owen Meany’).
    The Fed does not even care about the inflation in higher education or medical costs- two areas where government intervention resulted in significantly higher, not lower, costs….and I do not believe they care about CPI, either!!!

    The USA can not afford higher rates- end of story.

    Don’t fight the Fed.

  3. The crystalline concision, with a damn-near-iambic-pentameter rhythm, of your description of the S&P 500 is classic Heisenbergian style. Another great column. Thanks, H.

  4. H

    Thanks for this great take. For those who want to say something like … harrumph, so and so can do this prediction stuff … No he can’t. Nobody can. There are only two time frames possible in the universe of security prices, the past and the future. Since forecasting the past isn’t useful, only the future is interesting. BTW, before continuing I should say that I was a participant in a meeting of the Eastern Finance Association some years ago where I was privileged to listen to the presentation of a paper reporting the results of an exhaustive study of the ability of qualified financial analysts to predict corporate earnings ,,, in the past. Picture that, folks. These were analysts who would update their estimates of corporate performance every quarter, including the year end numbers — after the year was over. Even though the year was over these analysts were shown to be only about 90% accurate in predicting the past year’s results. After I heard that paper I pretty much decided that the direction academic finance was taking was a waste of my time and I started focusing on other areas I could teach without laughing out loud.

    So, as to predicting the future, note that stock prices are found to be continuously distributed in the past and the future out to infinity. Continuous distributions are characterized by containing an infinite number of individual points. Therefore, the probability of predicting any specific point in that distribution for any time in the future may be found by the equation: Probx = 1/infinity, which any good math whiz will tell you equals zero. It can’t be done for two reasons. First, a specific point in a continuous distribution is infinitely small. Second, the values of the individual points in the distribution are, by definition, unknown so good luck guessing the point value of any one of them.

    As always, H has it dead on. Thank you, again, sir, for one more try to dispel the myths surrounding market forecasts.

Speak your mind

This site uses Akismet to reduce spam. Learn how your comment data is processed.