Economic Surprise Indices Offer No Value

Howard Simons
  MAR 24, 2014 11:41 AM

Using surprise indices to forecast returns? You might as well be flipping a coin; at least then you would get some exercise.


Spin-doctoring and the managing of expectations are based on the premise that you can fool most of the people most of the time. Thus you have candidates campaigning furiously across the frozen cornfields of Iowa for months on end every four years while trying to convince the self-adoring commentariat they can hope for, at most, 15% of the vote in a crowded field. Alternatively, you have investor-relations managers otherwise handicapped by Regulation FD and Sarbanes-Oxley trying to manage earnings expectations lower so that the actual earnings announcement seems like a "beat."
I used to ask students to share a little bit about themselves and their careers at the start of a program. One stated he was a software manager for a trading technology firm that just beat its quarterly estimate by (the requisite) $0.01 per share and saw its stock price jump. Restated, he identified himself by his firm's managed beating of an imaginary target.
Economic Surprises
This silliness has extended to what are called "economic surprise indices," managed by the likes of Bloomberg and Citigroup (NYSE:C). The concept is simple: Define a set of market expectations as set by analysts, economists and maybe a Druid or two, and then compile how the actual report compared to the expectation. Then aggregate the various hits and misses and, voila -- you have an economic surprise index.
Forgive me for being less than impressed or, alternatively, congratulate me for testing to see whether the Citigroup economic surprise index for the US had any non-random forecasting content for US stocks, 7-10 year Treasuries, the Dow Jones-UBS index for industrial metals and for both high-yield and investment-grade corporate bonds.
I examined the correlation structure of these markets against the surprise index in one-week forward increments from the January 2003 inception of the surprise index and selected for the maximum positive correlation for the stocks and metals, and minimum negative for the bonds. This was data-mining in favor of trying to establish a leading relationship.
The pencil-pushing did not help, and indeed could not help. Let's take a look at the surprise index against forward 11-week returns for the Russell 3000 Index (INDEXRUSSELL:RUA) and for forward two-week returns for 7-10 year Treasuries. The respective r-squared or percentage of variance explained levels were 0.0022 and 0.00064. The other r-squared levels were similarly near-random as well. You might as well be flipping a coin; at least then you would get some exercise.


No Surprises
No one should be surprised at how useless this index is for forecasting returns.  Once the actual economic data -- whether higher or lower than the consensus -- arrives, it is incorporated immediately into prices, intermarket relationships, and forward curves. This sets price expectations right back to the geometric Brownian motion, or random-walk process, ensconced in the efficient market hypothesis, plus or minus a random drift term. The impact of the surprise thus is felt today and has no bearing on future returns.  Those are dependent on as-yet-unknown developments.
Finally, economists and analysts are herding animals without the social utility of dairy cattle.  While there are always headline-grabbers out there willing to say and do outrageous things just to get attention and the usual assortment of perma-bulls and perma-bears who, like the broken clock, will be right twice a day, most forecasters converge on the consensus. Markets move on actual information, not this backward-looking convergence. 

More from Minyanville:

Todd Harrison: Biotech Is a Focus on Turnaround Tuesday

Jason Haver: Charts Suggest New All-Time Highs After Correction

Vince Foster: Rate Hikes or Not, the Federal Reserve Does Not Set the Cost of Credit

No positions in stocks mentioned.