May 2008 - Evidence of Manipulation of Key Current Headline Data
Issue Number 43
June 9, 2008
__________
Section Four
(View all Sections of this Issue)
Reporting/Market Focus
As discussed previously, including in the June 3rd Flash Update, there are two types of manipulation that distort economic reporting. Manipulation of the first kind includes long-term methodological changes to the definition, gathering, analysis or reporting of key data, with the impact of building in a reporting bias that generates overly positive results. As a result of such changes, government reporting increasingly has strayed from common experience. Examples of this type of manipulation include the creation of the monthly bias-factor/birth-death model adjustment added into payroll employment, or the elimination of accounting for millions of "discouraged workers" due to redefinition.
Manipulation of the second kind involves direct adjustment of targeted, current economic reporting for perceived near-term political or financial market needs. An example would be Lyndon Johnson’s reviewing the GNP reports before they were published, and his sending them back to the Commerce Department for "correction," if he did not like the result.
The matter at hand involves manipulation of the second kind, specifically manipulation of the headline numbers tied to first reporting of current monthly payroll employment change, GDP and the CPI. If indeed such manipulation is taking place, it offers some political buffer to the Bush Administration from the inflationary recession that otherwise would help political opposition in November. It also would be an inexpensive alternative to other policy tools that might be considered by the Federal Reserve and the Administration in their efforts to support troubled financial markets and related institutions.
There have been several instances of the second kind of manipulation in earlier administrations that I have been able to document (see the Primer Series available at www.shadowstats.com). At present, though, there are no whistleblowers or other direct proof of what appears to be happening in current headline reporting, only significant circumstantial evidence in unusual features of reported results, in the presence of motivation and opportunity, as well as indications of contrary results from better-quality series.
Payroll Employment Biased in Concurrent Seasonal Adjustments. In the case of the headline Payroll Employment change — the first estimate — published monthly by the Bureau of Labor Statistics (BLS) (Department of Labor), the data that go into the monthly calculations and seasonal adjustments are massive and complex. Out of necessity, very few individuals would be involved or have direct knowledge of political massaging of the data. Indeed, in some earlier documented cases, the manipulations were orchestrated by the Fed or a given administration from outside the statistical agency that did the actual reporting.
When irregular changes are made within a complex reporting system, however, such sometimes leaves unintended evidence of the manipulation that can be found in careful examination of the available data. During the Clinton Administration, for example, an examination of monthly revisions to payroll employment reporting showed that seasonally-adjusted monthly jobs growth was being targeted for an extended period of time at exactly 250,000 jobs per month or exactly 500,000 jobs per two months (a target of 3.000,000 jobs per year). After the BLS was questioned on the matter, those patterns disappeared from further reporting.
Impossible Seasonal Adjustments. A generally unrecognized issue with current payroll reporting is highlighted in the following graph, a circumstance that has been enabled by the BLS’s "concurrent" seasonal adjustment practices, which calculate current-month and recalculate recent-month seasonal adjustments each month. Over the period of a year, seasonally-adjusted and unadjusted series should be equal to each other. Instead, unusual seasonal-adjustment patterns appear to have "created" 595,000 jobs in the headline employment numbers in the last 12 months, with nearly 370,000 of those being generated in January through May 2008.
The purpose of seasonally-adjusting payrolls is to redistribute reported employment activity throughout the year, so as to smooth out monthly activity for regular variations tied to calendar events, holiday-season employment, school year, etc. At the end of a year, both the seasonally-adjusted and unadjusted series should equal each other. Using seasonal adjustments should not end up creating the reporting of new jobs, only redistributing the numbers over the period of a year.
One way of avoiding having to use seasonal adjustments to assess current monthly trends is to look at the year-to-year change in the monthly series, as such neutralizes the bulk of seasonal variation. The exception would be where calendar variations, such as an early or late Thanksgiving, might result in some irregular (when viewed year-to-year) month-to-month shifting of jobs.
That said, under most circumstances, the year-to-year percent change in monthly payrolls should be virtually identical for both the seasonally-adjusted and unadjusted series. One of the regular cross-checks I run on the monthly employment data is to look at the adjusted and unadjusted year-to-year change in the employment levels that generate the headline jobs creation number. Therein lies a situation that cannot be happening with honest reporting.
For each headline employment number in the last year (subsequent revisions are not relevant here, as the headline number is what would be targeted and what is followed by the markets and media), the year-to-year unadjusted change was calculated and used to work out an implied seasonally-adjusted set of numbers. For example, for the May 2008 jobs report, the unadjusted employment levels for April and May 2008 were divided by the same numbers for 2007. If the year-to-year percent changes in the numbers on an unadjusted basis were the same as the adjusted, then applying the unadjusted annual rates of growth from the unadjusted series to the adjusted April and May 2007 numbers would yield the same April to May 2008 change as officially published headline number.
That, however, did not work out. As reported, May payrolls of 137,754,000 fell by 49,000 from April’s 137,803,000, while estimating the adjusted series using the unadjusted growth patterns showed May at 137,626,000, down by 134,000 jobs from April’s 137,760,000.
The problem is that this pattern has been repeated in 11 of the last 12 months, suggestive of some intelligent intervention in what otherwise should be something of a random process. While the math may be somewhat convoluted (a worksheet on the data is available on request), and comparative adjusted and unadjusted annual growth rates will vary some month-to-month, the variations should lead to irregular patterns of higher and lower implied change versus reported headline jobs change (adjusted and unadjusted series should equal each other over time).
Instead, with the exception of September 2007 (which may have involved unusual Administration versus Fed pressures on the approach to the banking crisis), every month in the last year has shown an implied upside bias in the headline reporting. The total upside bias over the last 12 months was 595,000 (just headline, not net of revisions), with monthly biases in January through May 2008 running respectively 79,000, 104,000, 63,000, 37,000 and 85,000.
As with the Clinton Administration’s apparent 250,000 per month jobs targeting, this circumstance likely will disappear as it gets increased exposure.
The circumstantial case for massaged jobs data considers the preceding, in conjunction with potential political/financial-market motivation and with other employment-related data, such as help-wanted advertising, jobless claims and purchasing managers surveys, all of which suggest recent monthly employment declines should have been six-digit.
One comment I have received is that the 0.5% surge in May’s unemployment rate surely was not manipulated. As discussed in the Opening Comments, the report may well have been the result of poor-quality seasonal adjustments. Next month’s report may show an unusual swing in the other direction. While an unusually large change in unemployment will take headlines, it is the payroll change that usually is considered the headline number from the monthly employment report and would be targeted for manipulation.
Guesstimating GDP. No special gimmicks are needed to adjust GDP reporting, since everything needed already is in place. The "advance" estimate of GDP, which usually is the primary headline number for the quarter, has a 95% reporting confidence interval around it of +/- 3%. The number is generated based largely on underlying assumptions — guesstimates — not on hard data. The Bureau of Economic Analysis (Department of Commerce) generates three estimates: high, low and best, and tries to target the economic consensus estimate, which tends to be overly optimistic going into recessions.
The government can report — and justify with its underlying assumptions — whatever growth rate it desires. Few will question it, if it comes close to the consensus outlook. Does the Administration have a political interest in the results? Of course it does. Might overly optimistic assumptions be used to generate desired results?
Key economic series, such as retail sales and industrial production, suggested quarterly contractions in both fourth-quarter 2007 and first-quarter 2008 inflation-adjusted GDP growth, but reported growth was positive for both quarters. Yet, as discussed in the GDP section, GDI (Gross Domestic Income), which is the theoretical equivalent of the GDP, contracted in the fourth quarter and was virtually flat in the first quarter.
Unusual Seasonal Adjustments Mask Oil Price Impact in CPI. Seasonal factors have been suppressing the reporting of gasoline and energy price increases in recent, seasonally-adjusted monthly CPI inflation reported by the BLS. The argument goes that where prices have been rising this year, they also were rising at the same time in 2007 and 2006, hence the need for seasonal adjustments.
Accordingly, there should be a period of catch-up, since the raw CPI numbers do not get revised, and the monthly seasonal factors are not recalculated every month as they are with the payroll data.
John Crudele of The New York Post was able to get comments from a BLS spokesman in this area, indicating that the seasonal-adjustment reduction in gasoline prices would continue in May, but begin to reverse with the June CPI. As described in Crudele’s May 20th column:
"A top government official who helps calculate the nation’s inflation rate says gasoline costs in the consumer price index will surge in a couple of months - even if prices at the pump don’t.
" ‘We are going to show huge increases,’ predicted Pat Jackman in a telephone interview with me last week. ‘If gas prices are stable from May forward, we are going to end up showing roughly a 16.3 percent increase [for the period] between May and December.’ "
That well may be, but the recent surge in oil prices goes far beyond regular seasonal variation. The following two graphs show monthly average oil price levels, and year-to-year percent change in same, for the period January 2006 to date. I can find no meaningful seasonal patterns in either series.
Instead of adjusting away these large changes, the BLS has an option known as "intervention analysis" to remove the effects of unusually large changes in prices — that are not seasonal in nature — before calculating its seasonal factors. Such was done for gasoline prices impacted by Hurricane Katrina. Had such been applied to the current energy price circumstance, recent seasonally-adjusted inflation reporting would have been much higher.
While reporting catch-up should follow in the months ahead, recent low inflation reporting certainly has been helpful to the Fed during the current financial crisis. Other issues as to why core inflation does not reflect the carry-through impact of higher energy prices will be addressed at a later date.
Next Reporting/Market Focus
Money Supply in Theory versus Available Hard Data
Given some ongoing debate as to what is the proper measure of money supply, what is happening to the velocity of money and what constitutes inflation, etc., the next Reporting/Market Focus will explore some basic monetary theory and what can and cannot be observed in related existing economic reporting, as well as how money supply measures can be used to predict CPI.
End of Section Four
(View all Sections of this Issue)
PLEASE NOTE: The next SGS Newsletter is targeted for around the end of June. Intervening Flash Updates and Alerts will be posted in response to key economic or financial-market developments.
Earlier editions of the SGS Newsletter, referenced in the text, can be found on the Archives tab at www.shadowstats.com.
OCCASIONALLY, BRIEF UPDATES ARE COMMUNICATED DIRECTLY BY E-MAIL. IF YOU ARE NOT RECEIVING E-MAIL COMMUNICATIONS FROM US, PLEASE LET US KNOW at johnwilliams@shadowstats.com or by using the "Feedback" option on www.shadowstats.com.