With the publication of the Atlanta Fed’s GDPNow estimates we get a better sense as to how not only the GDP calculations work internally (something largely hidden until now) but how GDP figures evolve with new pieces of information and settings. Along with their webpage snaphshot, the Fed branch has opened up its Microsoft Excel Spreadsheet for the whole world to peer inside the GDP envelope. There is an enormous amount of information contained within the various spreadsheets, but I think, as a start, it is very interesting to see how GDP has been formed over time; especially with regard to its subcomponents.

The prominence gained by the Atlanta Fed’s model more recently is certainly due to the fact that it was almost spot on for Q1 2015. While economists all over Wall Street were still expecting around 1.5% to 2% just before the BEA release, GDPNow was at +0.1%, only microscopically off of the initial BEA estimate of +0.2%. The model was not, however, always so accurate, which is something to keep in mind as it posits just +0.8% for Q2 (some turnaround).

ABOOK May 2015 GDPNow BEA Comp

The model has undergone several calculation modifications, which does make earlier estimates less reliable in some respects. As you can see above, the first runs were quite optimistic, which isn’t that unexpected given the origination here (Atlanta Fed after all). More recently, the model has been much better and much closer to the range for much of the time.


Again, the intent is not to match the revised (and revised again) BEA figures but rather the initial run from more incomplete (so they say) data sources. All of this is to that the model has been rather good at matching rawer GDP figures in and out for most of its timely calculations. That provides, then, a reasonable assertion that its subcomponents are, even if not exact then close, a decent representation of these economic pieces over time.

I think that is an important point that is often missing in the quarterly accumulations. GDP by itself is lumpy and is not well attuned to some of the more frequent economic changes, ebbs and flows that are missed by straddling nothing but three-month intervals. We have other economic accounts, of course, that provide more frequent updates, but until now we had very little specific grasp as to how those affect individually the GDP figure itself. Since mainstream convention still uses GDP as a primary means for measuring economic health, this is not a trivial exchange.


The economic narrative as set by the employment figures provided by the BLS has not been matched by the individual monthly accounts for spending like retail sales. In terms of GDP, the quarterly numbers have been more kind to the unemployment version. But as you can see above, what becomes clear by the various timely movements, for the most part PCE (GDP’s overly exhaustive proxy for consumer spending) has been stuck in a limited range. But for the end of 2013 and then 2014, PCE spending is right around 2-2.5%, which is historically paltry. Even the massive rebound predicted out of last year’s polar vortex misdirection never materialized; and further, the jump in Q4 toward the end of the calculation period was almost entirely healthcare spending from the updated Quarterly Services Survey accounts (which might also account for Q4 2013 as well).

In other words, for PCE, the jump in PCE estimates right at the end of the Q4 ’14 period is not a sudden surge in consumer spending as it appears in the final quarterly figures but rather the specific timing (December 10) of the QSS and its healthcare bump. Apart from that and before it entered the data stream, the whole of the quarter’s evolution was the same sad state of consumer exhaustion.

What that suggests is that even GDP doesn’t really have any inkling that there is an enormous payroll expansion purportedly under way, and under way for some time now. As I described yesterday, along with productivity these spending divergences show that if there is an aberration in GDP it is when it is high. The lower and even negative readings (as there is a good chance Q1 2015 in the next revision will be negative after imports and construction figures for March) have been widely proclaimed as anomalies, but the GDPNow trajectory is much clearer that the anomalies are those on the upside.

That extends to interpretations over capex, the other major source of optimism about some impending full recovery. Businesses are supposedly sitting upon loads of idle cash (but also debt), so it is very tempting for economists to extrapolate that into productive investment; they do so at the outset of every year. And for a time in 2014 that seemed to be the case, as there was a “bump” in the fixed investment calculations – but it did not long last the “rising dollar.”

ABOOK May 2015 GDPNow Fixed Inv

For a brief moment (or two) the future of capex appeared to be brightening, but only Q2 really remained that way as there was just, in the end, the same over-optimism that was never followed through. Even though Q4 fixed investment estimates were around 8% much of the time, by the time of the initial BEA release it was back down to 4% once more, and now negative. As with PCE, it is those upward estimates and figures that are the anomalies, not the negative numbers (now) well within the established downslope (going back to the 2012 slowdown).

There is, at this moment, too much data to parse in order to more accurately gauge why there are these only temporary upward irregularities. I tend to think that would be related to the fact that the economy itself is so highly unstable under so many artificial pressures, and thus the various economic accounts are themselves undergoing similar instability. I think that is the most vibrant interpretation of the GDPNow data series, as they show far more clearly this volatility. Whatever the case, it is that instability that leads to such mistaken impressions about the “recovery” idea and the economy itself. We now have more granularity with which to observe a more “native” state of the data itself, free from the quarter-end conclusions that lack some information value in and of the timeline developments.

This may not be, in the end, a totally accurate view of GDP, but GDP isn’t a totally accurate view of the economy anyway. It is constructed to produce the most charitable view on the economy, which is one reason that this downside preponderance (and only the intermittent upside) is so damning. And now we have a greater sense of that.