Model vs. Reality: Reality Wins

The bond market ended Thursday nearly unchanged, although short TIPS did very well because energy markets continued to trend higher. Gasoline rose 0.8% to $3.1136/gallon and NYMEX Crude added 1.5% to $107.83. Precious Metals were also higher. Stocks gained 0.4%. It is hard to believe this can merely be enthusiasm over growth and a “risk on” trade associated with the purported resolution of Greece’s troubles. In fact, I will say that with the almost unanimous acceptance of the notion that “the crisis is over” among the mainstream media makes me very nervous. Apple has recovered its losses from last Thursday, although on a fraction of the volume it had on the selloff, but I am accumulating equity hedges. Implied vols are at a 7-month low, but I don’t think risk is.

That is all I am going to say about market action today, because I want to mention a research publication that crossed my desk today and discuss what it means to a trader who is also an econometrician.

Goldman Sachs Global Economics, Commodities and Strategy Research today produced a piece called “The Top-Down Logic for Our Inflation Forecast.” In it, the economics team explains why they are calling for core inflation to fall to 1.5% in 2012, and 1.3% next year. Their reasoning is the “the combination of labor market slack and anchored inflation expectations should reassert itself in lower core price inflation over time.”

Frequent readers of this column will know that I have rebut the labor market slack hypothesis a number of times, and while I haven’t explicitly rebut the ‘anchored inflation expectations’ argument (mainly because modeling this requires complicated regime-shifting models that make it hard to refute null hypotheses) I am highly critical of them since the evidence in support of the notion that inflation expectations matter is based on measures of inflation expectations that demonstrably fail to measure inflation expectations.

So, you would think that these few paragraphs would be criticizing the forecast of Goldman Sachs. But that’s not really my point. Really, I want to point out the really hysterical part of the note, and observe why sell-side economic analysis is so useless (although some economists at Goldman, to be fair, are quite good). Goldman says “Although the model failed to capture the sharp pick-up in inflation in 2011, our bottom-up analysis suggests that this deviation was chiefly driven by special factors outside of the scope of the model, including pass-through from surging commodity prices and a spike in auto prices.”

Yes, that’s right: if there’s a discrepancy between reality and the model, then obviously it is reality at fault and not the model!

Get The News You Want

Read market moving news with a personalized feed of stocks you care about.

Get The App

To make the hilarity of this point clear I reproduce below the chart from their piece:

So all the model failed to do is to pick up the most-dramatic rise in core inflation in the last 35 years or so.

To make this fair, here’s my own model (now Enduring Investments’ model) covering the same period. Note that the model forecasts are actually finalized about 12 months ahead.

As I’ve said frequently, the current surge in inflation has not been fully captured by our model (although if we distributed the lags, rather than doing a simple lag, it would probably have done better, as the strange spike in late 2011 suggests). But at least it got the direction right, and began to rise at the right time, and for the right reasons. And, unlike Goldman, I think the reason for the difference is that our model isn’t quite right and is missing or underestimating some effect – so I expect our model will catch up with reality, rather than the other way around as Goldman does.

If an economist is highly confident in the model, then it can be reasonable to expect minor deviations to be mean-reverting. But it’s the model that’s mean-reverting, not reality (it is a model, after all – it isn’t supposed to capture all the nuance of reality). And if there is a significant deviation in reality from the model, then it can make sense for an economist to hew to the model if he’s 100% confident in the model. But then, if he’s 100% confident in a model, he’s an idiot. Which reminds me of this:

Bernanke: Well, this fear of inflation, I think is way overstated. We’ve looked at it very, very carefully. We’ve analyzed it every which way…We’ve been very, very clear that we will not allow inflation to rise above 2% or less…We could raise interest rates in 15 minutes if we have to. So, there really is no problem with raising rates, tightening monetary policy, slowing the economy, reducing inflation, at the appropriate time…

Pelley: You have what degree of confidence in your ability to control this?

Bernanke: One hundred percent.

Aside from the irony that we’re already above 2% (although not in core PCE, which he will claim should be understood), the other commonality (besides utter confidence in the model) between Goldman’s economists and this Princeton economist is that they’re absolutely confident that they have the right model: labor slack and inflation expectations. So far, we have seen from neither one of these sets of economists any deep introspection about whether maybe, perhaps, labor slack and inflation expectations don’t really matter, but things like money supply do.

Incidentally, while our model’s forecast for year-end is 2.10%, if we’re right on the trend but wrong on the starting level (that is, if the effect we missed was around the bottom and isn’t something persistent) then the more-relevant figure is the 0.8% acceleration in y/y core CPI our model expects in 2012. That would put core inflation at a cool 3%. Interestingly, if we remove the dampening effect the high level of private debt has in our model – that is, if we simply hypothesize that the ratio of private/public debt has a discontinuous effect so it either dampens (at high private/public ratios) or does not dampen (at moderate or low private/public ratios), then we also get 3%. That hypothesis, obviously, is difficult to test, which is why it’s important to not be 100% confident or reliant on your model, and to always be skeptical that you may be missing a key dynamic. Life is not linear. The reason that these economists don’t know that is that they’ve never tried to trade a model! If you are an investor who relies on models, a healthy – and continuous – skepticism is your best defense against reality diverging from your model.

I suspect something in between our model’s forecast (2.1%) and our model’s forecast for acceleration (implying 0.8% acceleration) is right, and so 2.7%-2.8% is our official forecast. By the way…not that it matters, if labor slack and inflation expectations are all that matters…but M2 rose $27.9 billion in the latest week and continues to grow at a better-than-10% pace year/year. It first hit that pace exactly half a year ago.

On Friday, New Home Sales for January (Consensus: 315k from 307k) is due out at 10:00ET. More interesting is that a number of regional Fed Presidents are in NY to speak at a conference on monetary policy. The conference begins at 9:00ET, so be prepared for tape bombs all day from San Francisco Fed Williams, St. Louis Fed President Bullard, Philly Fed President Plosser, and NY Fed President Dudley – a former Goldman Sachs economist, by the way. It will be interesting to hear if any of them is interesting in examining his model, or if they all expect reality to conform to the model.

It being Friday, we are supposed to have some rumors about a Greek deal over the weekend, but I am not sure how to play that since we supposedly have a Greek deal already. Perhaps this means that there won’t be such great expectations, and we can all enjoy the weekend for a change.