Monday, July 28, 2014

Let's predict the GDP report

We can skip the consensus of 3% as BS from the business and retail economists, notorious liars. Let us extrapolate from this graph, the graph below and add in the Obamacare costs. The graph below is percent change from the previous quarter, so it includes the 3% reduction. Otherwise, 2013 shows a .5% increase, a 1% increase and a .5% decrease.

So, regarding the GDP chart, we have, in the series first differential , .5 1.0,.5,-3. From the chart above we can make out a similar series for YoY growth by grouping regions. Roughly we get:
 3.5,2.5, 1.9,1.2,-1; in 2013.

So, approximately, we can say the regions are 3 twos digits in a counter, which count from -3 to 1.0; the last count was -3, and they seldom change counts by more than 1.5, and a negative jump of 3.0 was an extreme, due to Obamacare. So, at the top end, we might say, the best we can get is a -3 jump to -.5.  Negative .5, my final answer; after the first revision. Given the need for a revision, I will say the initial print might be .5, positive.

Seriously, go back to 2010, when was the last time the regions made any combination that changed the delta by more than 1.5 points in delta GDP? Given three counters, that is about 12 combinations, and we have some 19 examples.  The maximum delta jump, in any direction being 1.5 (except for the Obamacare adjustment of -3). We are simply constrained to small changes of 1.5 relative to the previous number.. Obamacare has, if anything, made us more constrained and implies a negative bias. So a negative .5 is actually optimistic, and an initial print of .5 simply implies the usual counting errors of the BEA.

What was my method again?

I said the regions are the principal components that adjust the first differential of the economy, the change from quarter to quarter, that is what they count, quarter to quarter change.  These regions are constrained in the integral of the rate at which they count quarter to quarter change. Obamacare was the exception. The speed by which they count change is limited to about half the previous change, at best. The integral is 1/2 the first differential. Unless they over reacted to Obamacare, they are not likely to report a positive growth, quarter over quarter.

OK, Shannon again. The principal components are counters and whatever they are counting will have square noise, which I call the second differential.  The thing counted will have at least 1/2 as much power as the noise, mostly larger. Great, simple. maybe it works. Is this always the case?  No, consider President Brown and Neito in their trade meetings.  They keep signal large relative to square noise to avoid boo boo, they  undersample by 2 plus a buffer, that is square noise.  But they also keep an eye on what Texas President Perry is not doing, he is not having trade discussions. So, Brown and Neito are leaving space for a surprise Perry move, one way or the other, they are  assuming cubic noise; they triangulate. Rather than approach from the log base 2 to the natural log, they go from log base 2 to log base 3, they deliberately undersample so as to make room to maneuver whenever Rick Perry maneuvers.

No comments: