Wednesday, September 09, 2009

NY Fed Treasury Spread Model: No Chance of Recession in 2010, Economic Recovery Has Started

The New York Fed just released its latest "Probability of U.S. Recession Predicted by Treasury Spread," with data through August 2009, and the Fed's recession probability forecast through August 2010 (see chart above, click to enlarge). The NY Fed's model uses the spread between 10-year and 3-month Treasury rates (3.42% spread in August, the second highest spread since May 2004, just slighly below the 3.54% spread in June) to calculate the probability of a recession in the United States twelve months ahead.

The Fed's data show that the recession probability peaked during the October 2007 to April 2008 period at around 35-40%, and has been declining since then in almost every month (see chart above and chart below). For August 2009, the recession probability is only 1.45% and by August next year the recession probability is only .08%, the second lowest level since May 2005.

Further, the Treasury spread has been above 2% for the last 18 months, a pattern consistent with the economic recoveries following the last six recessions (see chart above). The pattern of the recession probability index so far this year (going below double-digits and declining monthly) is very similar to the pattern starting in March 2002 that signalled the end of the 2001 recession (see chart below).


At 9/09/2009 2:36 PM, Blogger David Foster said...

What mechanism would explain why this model is a good predictor?

It's dangerous to make decisions based purely on historical correlations without any cause-and-effect analysis as to why these correlations might be valid.

At 9/09/2009 2:59 PM, Anonymous morganovich said...

this is a classic example of a model that backtests but has little predictive value. short term rates and especially forward short term rates are fickle and volatile. by the time next august rolls around, the numbers that get displayed in the chart to make it look like it works may be totally different from those now. thus, it looks like it works. but it doesn't.

if you made a graph of forward one year predictions from the past, it would show you nothing useful. if it were this easy, it would have already been priced in and wouldn't work anyway.

At 9/09/2009 3:36 PM, Anonymous Gabriel said...

Morganovitch said : "if it were this easy, it would have already been priced in and wouldn't work anyway."

Another extreme fundamentalist of the efficient market theory... I hope that you're an investor or trader so that I can take your money ^_^
And here I thought that people would have learned something from this crisis !
Hihi it looks there always be someone ready to let his money slip away from his fingers. And that's fantastic.

At 9/09/2009 4:05 PM, Anonymous morganovich said...


i am, in fact, a hedge fund manager. if you'd like to invest according to this fed model, i'd be happy to take your money. it's a classic rookie trap. your model will back test and you will hemmorage money, just like trading with MACD or stochastic bands or any number of others.

i have specialized in small and micro caps for a decade, so believe me, i have no particular faith in efficient markets. i know just large they can be and for how long they can persist.

that said, before that, i built program trading systems, so i've seen both sides of the argument.

if there was a simple, straightforward, linear 2 element model that predicted something as broad and important as the overall economy with even rudimentary accuracy, trust me, it'd be in the models and get arbed out faster than you can say "flash trade".

the whole tricky bit of efficient markets is that any investment inefficiency can only work for so long. the market is very good at scraping out any obvious patterns. the best data miners on earth are working on the problem.

the fact that they still miss a great deal is not a testament to their stupidity or ineffectiveness, it's a sign of just how difficult the problem is.

back chained non linear cross linked massively multidimensional models are a bitch, especially id you have to stress test them for non-discrete movements. (like the one that killed LTCM)

At 9/09/2009 4:24 PM, Blogger Matt Young said...

They simply estimate the likelihood that the yield curve will invert based on volatility of the steepness.

Still this could very well be an L shaped recovery.

At 9/10/2009 12:50 AM, Anonymous Six Ounces said...

The Beige Book came out today. Let's just say there was something for everyone in there.

Just about every sentence was filled with weasel words, vague adjectives and double talk.

Twelve Fed regions and every line of every report sounded like:

X is showing signs of improvement, but Y is weak or unchanged.

It was a masterpiece of many words, saying nothing, and meaning whatever they want it to mean ex-post and leaving escape hatches.

The best information were the statements which were unambiguously negative, like commercial real estate.

At 9/10/2009 1:07 AM, Anonymous Six Ounces said...

Morganovich, it's refreshing to finally read something from someone who knows what he's talking about here. It's amazing how many people are stuck in their two-variable world without a clue about endogeneity, omitted variable bias, simultaneous equations, and the generally nagging problems of time series analysis.

I'm a CFA charterholder but my knowledge of finance is limited. I do know econometrics and there's not an econometrician among them.

They also vascillate between unperturbed faith in perfect, free markets yet think themselves expert market timers. Any good economist understands there are market imperfections. Your comment demonstrates the extreme difficulty, expense, and risk of trying to identify and exploit those inefficiencies net of transaction costs and taxes.

But let gabriel read his candlesticks or value his companies pretending they have a crystal ball.

At 9/10/2009 12:54 PM, Blogger Jeremy said...

I think the Treasury's current $5 trillion/year rate of issuance should pretty much invalidate any useful information that might otherwise be gleaned from bond spreads.

At 9/10/2009 9:49 PM, Blogger Unknown said...

I don't know what manipulation was done to create the second graph in the pdf, but the first graph seems to show that a sharp flattening and/or inversion of the yield curve precedes a recession, by six months to a year.

I didn't think this was controversial. Hasn't it has been the conventional wisdom for a long time?


Post a Comment

<< Home