This Great Graphic was posted on the Sober Look blog. It took it from research from JP Morgan. It compares the bank's data surprise models for the US and Europe. It documents what we have experienced.
US data has generally surprised economists on the upside is early summer. European data was improving relative to expectations July and August, but then has disappointed consistently.
Data surprise models are common, but often misused. There are two components to the "surprise" and neither is stationary.
The first, of course, is the data itself. Much of the high frequency economic data that is tracked is volatile by the very nature and often double-count, such as durable goods orders and factory orders, or retail sales, which account for roughly 40% of personal consumption expenditures and PCE itself.
In many models, all the time series used are equally weighted, yet some data is more important than others and some data is more volatile than others. A rise of 1.0% increase in new home sales may not be as significant as a 0.5% rise in core retail sales (excluding autos, gasoline and building materials, which is used for GDP calculations).
The second component are expectations, which are also constantly being adjusted by the army of economists, policy makers, and investors. Expectations in such models are often seen as fixed points rather than a range. Some data, for a number of reasons, including the way the it is collected, is easier to forecast, like consumer prices. Other time series are notoriously difficult to forecast, like the monthly non-farm payroll number.
Some study of market expectations suggest that they are not typically normally distributed. In some ways, like market returns themselves, the tails are fatter and the is more gathered around the mean. Perhaps it is how economists and analysts are compensated, broadly understood. There might be some incentives to stick out from the crowd and make bold forecasts. At other times or in different places, there may be incentives not to stick one's neck out, so to speak.
Yet, no one likes to be wrong long. After the period where the data has generally surprised to the upside in the US, economists "chase" the market and adjust their forecasts. It is possible, like investor they get the timing wrong. Just as economists revise up their forecasts, the US economy softens a bit after a relatively respectable Q3 (which may be revised up, and such revisions are often not taken into account by the surprise models).
The surprise models are most useful when they prod us to ask the next question; not when we are told that pictures speak for themselves.