Fooling All The Experts With Seasonal Adjustments, All Of The Time

Tyler Durden's picture

Reading the economists' comments in response to today's ISM report (which, incidentally, missed expectations) one would think that the US has practically entered a second golden age. Here is a sample:

  • Manufacturing index "has now stabilized at a level reflecting a solid pace of expansion," Thomas Simons, economist at Jefferies, writes in note
  • June data consistent with Barclays estimate of 2Q GDP growth rate of 4%, according to note from Cooper Howes, economist at firm.
  • June’s reading of 55.3 “has to be viewed as a good result, even if it was lower than expectations,” Rob Carnell, economist at ING, writes in note
  • ISM index shows factories humming along in Q2, according to UBS
  • And especially this one from TD Securities: Increase in new orders, as tracked by ISM factory report, is "especially encouraging as it augurs very well for future manufacturing sector activity"

So what exactly are all these "experts" looking at to be so convinced, once again, that the "imminent" economic surge that thay have all been predicting for so long, incorrectly, is finally here.

The answer - the all import New Orders index - the key driver of the headline ISM print and the one most important sub-headline index.  And if we were also simply looking at the reported number of 58.9, which printed at the highest level since December, we too would assume that the US economy is finally rebounding.

Alas, here lies the rub: what none of the abovementioned experts realize is that for some inexplicable reason, the ISM survey is, just like the vast majority of all other economic indicators, also seasonally adjusted.

Recall that it was ISM's seasonal adjustment SNAFU last month, when it used the wrong "adjustment factor", that caused the reported number to become a humiliating farce after the ISM had to revise it not once but twice with what ultimately ended up being a "factor" leading to a far higher, and consensus expectation-beating, headline ISM print of 55.4.

But what really happened in June? For the answer we need a refresher of just how the ISM survey results in reported numbers.

What the ISM does is ask respondents to comment on how they are seeing any given query category as performing in the current month. The response options are simple: better, same, or worse.

The ISM then takes the percentage of "better" responses and adds half the percentage of "same" (ignoring the worse answers) for any of the following categories:

  • New Orders (58.9 in June)
  • Production (60.0)
  • Employment (52.8)
  • Delivery Time (51.9)
  • Inventories (53.0)

Then it simply takes the equal-weighted average of these 5 series and gets the final number (in the case of June 55.3 down from May's adjusted 55.4).

However, before the final tabulation, the ISM also applies a little-known seasonal adjustment factor to the actual unadjusted survey reponse result before getting a seasonally adjusted number that feeds into the above calculation. Why a survey needs to be seasonally adjusted - considering it merely captures sentiment which already reflects the periodicity of the seasons when it is, well, experienced - is beyond the scope of this article, and/or logic.

What adds to the confusion is that for some unknown reason, in June of 2013 the seasonal adjustment factor (1.051) actually subtracted from the unadjusted number (converting 54.5 into 51.9), while in June of 2014 the factor (0.976) managed to add to the unadjusted number of 57.5, making it appear the abovementioned highest for 2014 print of 58.9. Was June of 2014 more "seasonal" than the June from a year earlier?

This is shown in actual practice below:

What the table above shows is the survey responses (courtesy of Stone McCarthy) for the all important New Orders data. The cell highlighted in green, namely the 58.9 June New Orders print, is what the "experts" are looking at. The cells shaded in red is what is relevant: namely what the people are really saying about the economy as it is right now, not adjusted for some arbitrarily assigned fudge factor.

According to the actual data, when it comes to "New Orders" the number of respondents who responded "Better" dropped to 30%, below the 35% in May, far below the 37% in April, and in fact, the lowest since January when the economy was crashing under the weight of "harsh weather." Furthermore, while the number of respondents saying the economy is doing the "Same" rose to 55%, or the highest sinec 2012, those responding "Worse" rose once again to 15, higher than April, and the highest since the "weather impacted" February.

Visually, here is the difference between the Unadjusted and Seasonally Adjusted New Orders survey.

In summary: the actual, and unreported, New Orders number dropped from 60.5 in May to 57.5 in June, which also was the weakest print since January... some improving trend. Compare that to the seasonally adjusted New Orders number of 58.9, the highest of 2014. That's right: thanks to seasonal adjustments what was otherwise a downward sloping trendline, and a print that was the weakest in 5 months, magically was transformed to the best print of the year!

And that is how you fool all of the experts all of the time with something as simple as a completely unnecessary seasonal adjustment factor, which leads everyone to believe that the economy is soaring. Well, it is... on a seasonally adjusted basis, the same seasons that everyone says to ignore when they lead to a -2.9% GDP print!

In reality, the US economy, as represented by actual New Orders surveys, is the weakest it has been since January. And to think- very soon everyone will be shocked, shocked, that Q2 GDP (and Q3, and Q4 and so on) was not nearly as wonderful as everyone had prayed it would be.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
PartysOver's picture

Can you say "Delusional Stupidity"?

NotApplicable's picture

'When it becomes serious, you have to lie'

Vampyroteuthis infernalis's picture

I wish my profession could just make up numbers as they go. Oh wait, I actually produce useful things as an engineer.

lordylord's picture

"Oh wait, I actually produce useful things as an engineer."

Depends on the engineer. NIST made up all kinds of shit in their 9/11 report.

NoDebt's picture

Just scanning down the seasonal vs. non-seasonal numbers..... to me, it looks like the NON-adjusted numbers are less "noisy" than the adjusted ones.  I thought seasonal adjustment was supposed to, you know, SMOOTH OUT seasonal factors so different months could be compared against eachother more easily.

isudas's picture

Yes, the engineers at GM keep producing useful products.  Even better than the ones they made last year.  So useful that year after year they ask you to bring your car back to the dealer to install the even more useful product they made this year.  Without engineers dealers would have to develop innovative ways to get you to come back time and time again.   

hedgeless_horseman's picture



It is a slippery slope.  I would bet the government already seasonally adjusts election returns. 

I consider it completely unimportant who in the party will vote, or
how; but what is extraordinarily important is this—who will count the
votes, and how.

-Josef Stalin

 Eventually, people will seasonally adjust their tax returns.  This will be followed by people seasonally adjusting bullet drop and windage.

Got range card?

Gringo Viejo's picture

Nobody's fooling nobody. Everyone's whistling through the graveyard pretending not to notice. Everyone's just holding their breath and keeping their fingers crossed.
Because deep down they know that once this fraud implodes...
they lose EVERYTHING>

lasvegaspersona's picture

Seasonally adjusted the temp here in Vegas is a cool 74 degrees. My unseasonally adjusted electric bill will be over $400 though.

NoDebt's picture

Love it.  Why NOT seasonally adjust temperatures?  I mean, that's pretty much an IDEAL set of data for "seasonal adjustment".

sodbuster's picture

The New World Order Normal.

adr's picture

A lot of women seasonally adjust their breast size. Then you get that top off and find out those fabulous D's are actually B-.

The stock market is exactly the same. That company with the fantastic balance sheet is just using a pushup bra filled with padding.

sunny's picture

"In reality, the US economy, as represented by actual New Orders surveys, is the weakest it has been since January. And to think- very soon everyone will be shocked, shocked, that Q2 GDP (and Q3, and Q4 and so on) was not nearly as wonderful as everyone had prayed it would be."

Don't be silly.  The next batch of GDP numbers will be seasonally adjusted as well.  They will be wonderful.  What is the problem?  Didn't you folks get the memo?

machineh's picture

It is easy to end-run seasonal adjustment, by comparing to the NSA value from 12 months ago:

June 2013 -- 54.5

June 2014 -- 57.5

Result: up +3.0 in 12 months, non seasonally adjusted.

So what was this lame-ass article all about? 

Mercury's picture

Shouldn't all the seasonally adjusted and non-seasonally adjusted figures net out to the same endpoints after, you know, all the seasons have come to pass?

I mean, it seems logical that a seasonal adjustment is zero-sum meaning that a boost here means a debit somewhere else in the same year - no?

Or is it more like: "But I've only had three in dog-beers" kind of thing?

rwe2late's picture

the previous monthly reports could be revised almost endlessly (especially seasonally)


month           Jan      Feb      Mar        Apr        May

1st report        8.1      8         7.9       7.8        7.7

prior mo rev               7.9       7.8       7.7        7.6

publicized change      +0.1       +0.1    +0.1      +0.1   steady improvement!

rev & forgotten           -0.1       -0.1     -0.1      -0.1      or not!

machineh's picture

'For some inexplicable reason, the ISM survey is, just like the vast majority of all other economic indicators, also seasonally adjusted.'

Seasonal adjustment is done because some data series ARE seasonal.

For example, house sales go down every winter. It doesn't mean the housing market is collapsing. You have to compare to the previous winter, or else seasonally adjust the data.

As the author could easily have confirmed, ISM modifies its seasonal adjustment factors annually, and announces them in advance. Here is the 2014 announcement:

D-minus for research.

MortimerDuke's picture

I agree with you that the ZH focus on SA is sometimes excessive.  And I think the comments on these articles make it clear that not everyone has taken a class on econometrics or forecasting (probably to their credit).  Still, I have two problems with SA as used by the agencies and private sector responsible for most of the reporting.  First, I'd like to see the models used as well as the coefficients.  I think many go overboard on their adjustments.  Look at the change in adjustment foactors  between May and June for both 2013 and 2014.  Seriously?  That much difference because, presumably, a different number of work days?  Does the ISM use economic theory, common sense or pure data mining to come up with this discrepency?  At some point common economic sense has to play a role.  Second, the whole idea that you need to be able to compare a MoM change in monthly time series leads to a fascination with economic noise by markets and analysts.  Although there are problems with a simple YoY on NSA data, I still prefer it because the trend in momentum over a longer time frame, absent someone's balck box SA algorithm, is probably more meaningful.  By the way, the whole idea of seasonally adjusting a better/worse/same response to a survey question borders on nonsensical even though I understand those are answers that correspond to changes in seasonal variables.  The whole thing seems a little too stupid to take seriously.     

TabakLover's picture

Seasonal adjusters..........can blow me.

Kaiser Sousa's picture

sorry to interrupt...but can anybody give me a filter for blocking the fucking "ALL" ads above coming from this website...

ejmoosa's picture

Better?  By how much a little or a lot?

Worse?  Same question.

30% could be better by 1% while 15% could be down by 5%.  The index would look better while things get worse.

They could hvae actual values here.   But that would prohibit their efforts to baffle us with bullshit.

NoDebt's picture

Good catch.  I had heard something about that on the radio, but hadn't come across an article yet.

You would think stuff like this would end the global-warming debate (by finally admitting it's all lies, fudged data and goal-seeked projections), but no.... we'll still be having it shoved down our throats for the next 100 years as if it's true or even matters.

undercover brother's picture

The writer is correct, except that none of it matters when the fed and other cb's are intervening in the interest rate markets and moderating the equity markets every single day.   They are and have been the overwhelming influence for many years.  Until they step out of the way and allow capital markets to function in the manner for which they were originally intended, this rise will continue unabated.  Central banks were foolish to attempt to prevent losses and bankruptcy from occurring and they compuounded their error by continuing that policy.

orangegeek's picture

MSM doing their job pumping the herd.


POMO schedule is interesting - big pomo days followed by no pomo days.


I am sensing some volatility coming back soon.

kchrisc's picture

Winston Smith lives!

optimator's picture

And I suspect the Algos can't figure that out and take the numbers as printed.

max2205's picture

Who long as the Fed is buying the moon

oklaboy's picture

wish I could seasonally adjust my pay check...