They're Selling Us a Recovery That Never Was

Story Stream
recent articles

The United States Constitution declares in Article I, Section 2 that "representation and direct Taxes shall be apportioned among the several States which may be included within this Union, according to their respective Numbers." To accomplish this foundation for proportional government, Article 1 tasked federal authorities with "actual enumeration" within three years of the first meeting of Congress and then every ten years thereafter. The first Census was carried out in 1790 with Thomas Jefferson the nominal supervisor as Secretary of State. In fact, the first Census-takers were actually Federal Marshals deputized by each of the US Judicial Districts.

That first Congress established a special committee to prepare the framework for the endeavor. The Committee recommended six questions be answered by every legal respondent, with the Census Bureau today ascribing that no less than five of those six were proposed by none other than James Madison as a Virginia Representative. The focus was on identifying sex, race, relationship to the head of the household, the name of the head of the household, and number of slaves.

By the third decennial Census in 1810, Congress tasked those Marshals with accounting of additional economic information. With industry starting to grow in the early 19th century, the government wanted to, "take account ...of the several manufactures within their several districts, territories, and divisions." In addition to the six original questions relating to demographics there were added many, many more across 25 "broad categories" of economic production spanning more than 200 goods. Thus was born the Economic Census.

In 1902, Congress finally established a permanent Census Bureau and formalized the economic version to be taken every five years instead of ten. It was greatly expanded in 1930 to include retail and wholesale trade, and in 1963 added travel and transportation sector activities. The last major upgrade was made in 1992, when the Bureau added FIRE as well as telecom and utilities to its catalog. The only areas still left out of it are agriculture, rail, and household employment.

With the Census Bureau taking full economic account only every half decade, it raises the question as to what all this monthly economic data published regularly here and there actually is. If it takes the dedicated branch of the federal government five years to tally the actual results, what is the basis for all the rest?

The monthly, quarterly, and annual figures for various economic statistics are exactly that - statistics. They are almost all measurements not of actual changes in the economy but rather stochastically estimated variations around a benchmark. Those benchmarks are compiled from larger, higher population surveys than the individual "high frequency" data, with the Economic Census being the benchmark of benchmarks, so to speak.

With economic accounts reporting only variation, that leaves open any number of ways in which error might be introduced, as well as a great deal of time during which those errors might propagate. The last Economic Census was taken in 2012, but its results didn't start to find their way into the benchmark data flow until the middle of 2014 and really throughout 2015. Unlike prior editions, the 2012 update is proving to be absolutely stunning.

The distance between the prior two Economic Census contained, obviously, the Great Recession. Thus, the 2007 effort was the last of the pre-crisis era while the one in 2012 is the first of the "recovery" cycle that truly deserves its quotation marks. What that means for individual data accounts that again measure only modeled variation is that until the 2012 Census was fully digested the major economic data was still being benchmarked under the old, pre-crisis assumptions.

When the government agencies tasked with producing the various economic data transitioned in the 1960's and 1970's to stochastic processes, they did so assuming that they had more than enough technical skill to be successful. That included not just knowledge of statistics in general, but in how to apply them to an economic setting. There are all manner of statistical processes that are undertaken to produce a single monthly estimate for an individual data series. For starters, the US, as anywhere, exhibits sharp seasonal differences; retail spending, for example, is much, much greater around the Christmas holiday than immediately after it. So measuring variation means accounting for the "normal" seasonal pattern of just how much retail spending "should" decline every year from December to January.

Part of that "should", however, must take into account more purely economic conditions. If in years before the drop-off from December to January in retail spending was (for the sake of simple illustration) $100 but in the current year January spending is only $90 less, how do we account for that variation? If we expect -$100 as "normal" and instead find -$90, is that statistical variance (error) or is that economic growth? This is where subjectivity becomes most paramount, especially under conditions where the Census Bureau might find not of -$90, and thus indicative of growth, but maybe -$130 or even more and potentially indicative of recession.

To account for this change in cycle, seasonal adjustments are augmented by the incorporation of other components and concepts such as trend-cycle. In 1967, researchers Julius Shiskin, Allan H. Young, and John C. Musgrave working in the Census Bureau's Department of Economic Research and Analysis Division published in Technical Paper #15 their assimilation of these concepts into the 11th version of the AutoRegressive Integrated Moving Average (ARIMA X-11) template for variable adjustments of economic data.

"There are various and sundry methods for seasonally adjusting economic time series, all of which are based on the premise that seasonal fluctuations can be measured in an original series (O) and separated from the trend, cyclical, trading-day, and irregular functions. The seasonal component (S) is defined as the intrayear pattern of variation which is repeated constantly or in an evolving fashion from year to year. The trend-cycle component (C) includes the long-term trend and business cycle. The trading-day component (TD) consists of variations which are attributed to the composition of the calendar. The irregular component (I) is composed of residual variation, such as the sudden impact of some political events, the effect of strikes, unseasonable weather conditions, reporting and sampling errors, etc."

As Shiskin, et al, point out in their introduction, all those components "are related in multiplicative fashion (O = S * C * TD * I) for most national economic time series." The importance of benchmarking, then, is to check as to whether stochastically modeled variation is more or less accurate extrapolation from a narrower population subset. In other words, if the statistical world of the more highly estimated high frequency data matches the more comprehensively derived baselines of the Economic Census.

In 1978, Nobel Laureate Clive Granger pointed out one difficulty especially of "C", the trend-cycle component.

"It can certainly be stated that, when considering the level of an economic variable, the low frequency components, often incorrectly labeled ‘trend-cycle components,' are usually both statistically and economically important. They are statistically important, because they contribute the major part of the total variance, as the typical spectral results and the usefulness of integrated (ARIMA) models indicates. The economic importance arises from the difficulty found in predicting at least the turning points in the low-frequency components and the continual attempts by central governments to control this component, at least for GNP, employment, price, and similar series."

The implication for any usual business cycle is easily apparent, as until the turn from growth to recession (or from recession back to growth) becomes widely accepted the statistical high frequency accounts will be figuring too much "upward" variation. That is why revisions around cycle inflections are usually quite drastic, as it takes some time for the cyclical turn to be found and then declared in either direction - and only then added to the economic data. What that means is, to oversimplify and generalize for clarity, data accounts are made to more or less "expect" that once a part of the cycle is declared that economic data will act like that part of the cycle until declared otherwise.

Enter the 2012 Economic Census. Again, the high frequency data has been until really last year predicated on a pre-crisis benchmark scheme. Straight away, that sounds alarms as this "cycle" is even by those benchmarks nothing at all like all prior cycles. It is the inherent and intrinsic flaw of all statistical processes, that among all these other subjectivities at its very core lies the biggest assumption of them all - that the future will look like the past. If that isn't the case, then we are off into the "tails" of any probability distribution, meaning "error" that "needs" to be actively corrected.

What are being corrected now are all those assumptions of data error; in short, the stunted recovery is becoming even more stunted as the 2012 Census increasingly amends cyclical tendencies. Furthermore, it is in the consumer accounts where these revisions are taking their greatest toll.

These sharp benchmark revisions started last year, but it is the more complete updates to the core of each data series taking place this year that is truly astounding. The latest is for durable goods, particularly new production orders for consumer goods. The two subsequent benchmark revisions, the first occurred in May 2015, have completely altered the accounting of especially the time period surrounding this "rising dollar" phase.

In late 2014, it was widely believed and telecast broadly that the US economy was heading into not just full recovery but "full employment" reaching the dangers of "overheating" in the parlance of orthodox monetary policy. That was based on several factors, but primarily the BLS's labor statistics showing "the best jobs market in decades." That view was given greater plausibility by GDP that suddenly rose from the reductions in 2012 to as much as 5% (original estimates) for Q3 2014. In addition, while figures for consumer spending on goods and estimates for the production of goods were subdued compared to prior cycles they were at least indicating that a fruitful economic pickup might actually be stirring.

In the case of durable goods, the 2014 benchmark series (the last before the 2012 Census inclusion) estimated that in October 2014 the total value for new orders for the production of consumer goods (excluding aircraft) amounted to $169.99 billion. That was 6.4% more than the $159.80 billion figured for October 2013. That 6.4% growth rate was not just an improvement from the 3.7% average in 2013, but it also seemed to confirm the economic momentum among US consumers consistent with a strong labor environment. From that, Janet Yellen could plausibly claim that the sudden appearance of weakness toward the end of that year and especially into 2015 would only be "transitory."

The just-released 2016 benchmark version of October 2014 durable goods is much, much different, however. Last year's major revision already knocked it back to $164.5 billion, but this year's correction to subjectivity finds only $156.67 billion. In short, the actual level, or as close as the comprehensive Census benchmarks might get, of durable goods orders in October 2014 is $13.3 billion, almost 8%, less than previously thought. This reduction in "recovery" is cumulative, amounting to $257 billion in the 27 months from the start of 2013 to March 2015 when the first of the non-cyclical benchmarks was spliced. Janet Yellen thought there had been $4.33 trillion in durable goods orders for consumer goods in those two plus years when now it is believed there was only $4.07 trillion.

A miss of 6% like that is enormous, especially in cumulative fashion and even more so spread across that big a length of time. It completely erases the 2014 narrative, drastically altering the complexion of economic interpretation such that weakness in 2015 would not be expected as temporary but instead consistent. The latest revisions find that durable goods orders in October 2014 grew only 4.2% from the prior year, reflecting that $13 billion reduction across the two benchmarks, thus less than the 6.4% originally believed, but also that October 2013's total was revised down by $9.5 billion such that what was supposed to have been 4.4% gain is now almost completely gone. So instead of 6.4% in October 2014 on top of 4.4% in October 2013, we now find just 4.2% on top of 0.3%.

This benchmark reprocessing has been widespread throughout the spending and production accounts. From retail and wholesale sales to factory orders, what little recovery that was believed to have existed up to 2015 is disappearing as the 2012 Economic Census reveals that we are in nothing like a traditional business cycle. That point is most apparent in the drastic revisions (April 2016) to industrial production, especially again consumer goods.

Industrial Production

The pre-2015 benchmark series showed the typical pattern of this "recovery"; a steady but slow advance that took far longer than "normal" to match the prior cycle peak. By that count, though, the US economy appeared to be at least moving solidly if unspectacularly in that direction and doing so at an increasing rate so as to be also plausibly consistent with the mainstream, FOMC narrative. Two benchmark revisions later, however, and there is no recovery in the production of consumer goods to speak of - at all. This is much, much worse than recession not just in what the economy might have been already, but what it might portend for what the economy will be.

A recession is a temporary deviation from trend, which is why the Census Bureau spent a lot of time and effort in determining trend-cycle components as the largest potential source of variation. The trend is supposed to remain dominant even for each given cycle. What we see here after the Great Recession is nothing like a cycle. Instead of a temporary drop below trend, even if that "temporary" was somewhat longer than usual, we increasingly find that the overall trend itself has been suppressed. Given that the trough of the Great Recession occurred in the middle of 2009, seven years is more than enough time to make such a momentous declaration: the economy didn't fall into a recession great or otherwise, it shrunk.

Far from being a surprise, this result is actually quite consistent with other major economic data points including those produced by the BLS. The major problem all along has been in labor, the 14 or 15 million workers that just disappeared from the official rolls. The BLS does not include them in the unemployment rate, but the real economy cannot be so choosy. For a while, under the countenance of the 2007 Economic Survey benchmarks, there was at least some evidence that the participation problem might not matter except to account for a slower cycle. The 2012 update, however, shows that common sense should have prevailed:

"The Fed started with a view of recovery consistent with the unemployment rate only to be forced in later years to actually agree instead with the civilian employment-population ratio. In other words, the real state of labor and therefore the whole economy does not lay in the whole unemployment rate only the denominator just as we suspected all along. Economists have tried for years to suggest that the 14-15 million that "dropped" out of the official labor force count just didn't matter; a ridiculous idea without the slightest logic to it except the prejudices about "stimulus" always working. This idea was so prevalent and religiously bonded that it was incorporated into the subjectivity that these kinds of economic statistics do allow."

The implications of all this are enormous. What happens if the US and by extension global economy isn't actually within a business cycle? And hasn't been for almost a decade? It would certainly start to explain why "stimulus" never works; economists have been operating at the start under false pretenses. But it is much more than that, as only something very radical must have happened in the most basic economic functions to possibly alter the entire global economic trajectory (and furthermore do it in unison). Fortunately, there are a number of clues:

"It is not coincidence that orthodox monetarists were taken for geniuses all the while the eurodollar standard was expanding, qualitatively as well as quantitatively, and are now increasingly (and rightfully) viewed as floundering idiots who can't do anything right as it falls apart."

A badly malfunctioning financial system that has been the basis for global "money" for decades would be right at the top of any list of suspects. The eurodollar system hasn't been right since August 9, 2007, and neither has the economy. Whatever traces of doubt there might have been in the remnants of cyclicality are being erased by more comprehensive data, revised into an unmistakable monetary strangulation.

But this is all ignored in every way possible. When monthly variation in whatever economic flavor-of-the-month is high, the media reports it as indisputable proof that Yellen was right. When that "proof" disappears into the trend-cycle statistics of time, there is only silence. You would think that someone would be interested in describing how so much of the economy just vanished. It is the greatest story of our time and nobody wants to tell it. That can only mean politics.

This indictable incuriosity led to what was perhaps the most frustrating and devious aspect of the further weakness in 2015; economists actually had the audacity (due to desperation) to try to claim that the "rising dollar" was actually the strong dollar of old. The two could not be more different; the former explains everything about why economists were not only wrong then but had it all wrong all along. In other words, the "rising dollar" did not need the benchmark revisions to suggest weakness distinctly beyond the business cycle - it was expecting them to confirm the interpretation just as it has. The "rising dollar" is a symptom of an unstable "dollar" or eurodollar system becoming even more so.

The strong dollar, by contrast, is actually the answer. So economists and even policymakers that really should have known better (but it isn't surprising that they don't living in the impenetrable orthodox bubble that forces observation to conform to theory rather than theory to observation) were trying to claim that the necessary global solution to our deep and deepening economic problems was somehow evidence that no solution at all was ever necessary; all dependent upon a view of economic data that was growing more and more dubious by the month and really by the benchmark. And because economists have been in denial this whole time we are left years behind trying to figure out what to do about circumstances that don't follow any recognizable prior patterns. It's a perfect summation for just this sort of bureaucratically centralized political incompetence. They tried to sell us a recovery that never really was one, and are left only with more evidence of what the people have known throughout this cycle that wasn't.

Jeffrey Snider is the Chief Investment Strategist of Alhambra Investment Partners, a registered investment advisor. 

Show commentsHide Comments

Related Articles