New York Post Claims Census Falsifies Unemployment Figures

The New York Post is reporting an absolute bombshell story if true.  They claim the September 2012 unemployment report was manipulated and survey data was faked, just in time for the election.  The story quotes anonymous sources, insiders from the Census Bureau who claim to have falsified survey data for the unemployment report.

The decline — from 8.1 percent in August to 7.8 percent in September — might not have been all it seemed. The numbers, according to a reliable source, were manipulated.

And the Census Bureau, which does the unemployment survey, knew it.

Not only is the Post claiming the September 2012 unemployment surveys were manipulated but this is still going on today.  The Census actually caught one employee fabricating the unemployment statistics by falsifying survey results which should be answered by respondents.

Just two years before the presidential election, the Census Bureau had caught an employee fabricating data that went into the unemployment report, which is one of the most closely watched measures of the economy.

And a knowledgeable source says the deception went beyond that one employee — that it escalated at the time President Obama was seeking reelection in 2012 and continues today.

The story might very well be true.  The unemployment rate comes from surveys, sent out to 60,000 households spread out over 2,025 geographical areas of the country..  The Bureau of Labor Statistics gives great detail into the methodology used for these surveys.  The fraud reported by the New York Post comes from the 2200 Census employees who conduct the phone interviews each month for the Household survey.  Instead of doing their job and recording the answers from the interview questions of these households, a few individuals are falsifying the survey results.  If employees were falsifying interviews and this fudging of surveys is done with enough households, such fraud could skew the unemployment rate.

“He’s not the only one,” said the source, who asked to remain anonymous for now but is willing to talk with the Labor Department and Congress if asked.

The Census employee caught faking the results is Julius Buckmon, according to confidential Census documents obtained by The Post. Buckmon told me in an interview this past weekend that he was told to make up information by higher-ups at Census.

The Post claims the Census Bureau demands a 90% interview response rate and certain regions of the country that is hard to get.  So personnel are fudging the results and filling in the survey.  There might certainly be truth to this as employees are under pressure to produce results, no one answers their phones these days and who wants to be bothered by some Census Bureau employee asking a lot of obnoxious questions?

What is more odious is the BLS is not publishing actual response rates for the household survey.  If the Post article is true and the Census expects a 90% or greater response rate, that seems a little ridiculous demand by itself.  Below is a short explanation on how the unemployment report statistics are collected.  Read it and then imagine being part of this survey.  How accurate would you be or even responsive over time?

The survey is conducted by the U.S. Census Bureau for the Bureau of Labor Statistics. Each month, during the calendar week including the 19th day, Census interviewers contact households by telephone and in person and ask questions regarding the labor market activity of household members during the previous calendar week which included the 12th day of the month—the reference week. Personal visits are preferred in the first month in which the household is in the sample. At the first visit, interviewers prepare a roster of the household members including their demographic characteristics and their relationship to the person maintaining the household, and enter the information via laptop computers, along with responses to all survey questions.

In the months following the first interview, the interview is generally conducted by telephone. The household roster is checked for accuracy and brought up to date in each interview. About 10 percent of households are interviewed via computer-assisted telephone interviewing (CATI) by staff in three centralized calling centers. Other telephone interviews are collected by field representatives. A personal visit is generally attempted for the fifth interview. At the end of each day’s interviewing, the data are transmitted over secure telecommunications lines to the Census Bureau’s central computer in Washington, DC.

From the number of households and individual Census employees it is clear the Census would have to falsify many interviews to actually skew the survey results enough to manipulate the unemployment rate.  The Post story quotes the employee already busted for falsifying survey interview responses that he did so on the demand from management at the Census Bureau.

The question of intentional hoodwinking for political agendas versus low paid workers having to interview people all day on a quota is really unanswered.  Yet none of that matters.  If any portion of this story is true, it means data integrity is compromised.  All sorts of policies, funds, legislation and even Wall Street are tied to the unemployment rate.  Suspecting fictional statistics, just the concept, could result in no confidence of government statistics.  That is an unmitigated disaster on a host of fronts.

Census appears to have looked into only a handful of instances of falsification by Buckmon, although more than a dozen instances were reported, according to internal documents.

In one document from the probe, Program Coordinator Joal Crosby was ask in 2010, “Why was the suspected … possible data falsification on all (underscored) other survey work for which data falsification was suspected not investigated by the region?”

On one document seen by The Post, Crosby hand-wrote the answer: “Unable to determine why an investigation was not done for CPS,” or the Current Population Survey — the official name for the unemployment report.

With regard to the Consumer Expenditure survey, only four instances of falsification were looked into, while 14 were reported.

We look at the unemployment statistics in great detail every month and the way to prove the Census is falsifying data is by a probability and error margin analysis check.  Such an analysis is quite involved.  To detect Census employee fraud, such an analysis should be performed on individual employee interview results.  This still leaves the question of how honest actual survey respondents are as well.  In terms of statistical accuracy, we've never liked the Household survey.  The error margin is too great and we believe the Census should cross correlate survey results with other labor market metrics for a reality check.  While not a poll, a survey is just that, asking people to volunteer information.  To make matters worse, recently some of the revisions to GDP methodology are truly questionable and the last thing America needs is even more manipulated or skewed economic statistics

At the time the September 2012 unemployment report was released, the results were so extreme we compared it to falling through a worm hole in statistical space.  Yet that month is not the only one which seems to be skewed.  Every month we analyze these reports digging out the figures to explain the unbelievable, such as the unemployment rate dropping dramatically while the net gain of those employed is basically static.

We question the conspiracy element of the Post story that the Census would actually falsify the unemployment rate to skew an election.  The reason that seems absurd is the press pays no attention to the pathetic labor market and neither Presidential candidate in 2012 was offering a damn thing to actually increase jobs.  It is also doubtful a sudden drop in the unemployment rate would sway the election results, even though the official unemployment rate sweeps millions of Americans needing a job under the political and statistical rug.

What could very well be true is an increasing nonresponse rate to Census survey unemployment questions.  Generally speaking people do not have the time, feel obligated to answer such surveys or even pick up a land line, unlike 50 years ago.  We also question how honest people are who do answer surveys as well as the type of person to respond versus those who do not.  Just by asking individuals would seem to put bias into the sampling group.  One would think America could obtain more accurate data collection methods than a survey in order to find out what people are doing each month for work.  After all, the NSA knows every single thing we say or do these days.

Update: The Census Bureau responded to the Post's story and said they are unaware of any such fabrication of data by employees.  Now the Post is on a rampage and frankly we did not think too many would believe the New York Post in the first place.  Now the Post is doubling down while backing down from some of previous accusations.  At the same time the Post is trying to imply Congress will hold an investigation and hearings based on their news reports.  The latest New York Post blast is over the Employment Establishment survey birth/death business model.  This is the payrolls survey, also part of the employment report.  The birth/death model is an estimate of how many new businesses start and hire and how many go under.  We have repeatedly debunked conspiracy theories on the birth/death model here, here and here.

The birth/death model is a statistical model, based on different surveys as well as tax records of new businesses.  Now, come on, there is simply no way everybody in the Census is just some Machiavellian mastermind, out to manipulate large statistics and that includes the birth/death model.  Frankly the theory of the great Census conspiracy would imply everyone is not only in on the fraud, but there is some great mastermind at the top, dictating how employees cheat and lie in order to obtain the unemployment numbers that they want.  To state the obvious, the employment statistics are so bad, if this were true, that the Census was cheating, uh, don't we all believe they would cheat much better?  After all the reports still tell us in no uncertain terms the jobs market really sucks.

To the New York Post:   At this point your anonymous sources had better come forward.  While we believe some employees could be burnt out, fudging results and the squeezing of the Census budget could have put much more strain on existing personnel, it is becoming obvious the New York Post has real issues with the Census.  It is also becoming obvious those issues are not based in statistical reality.  The Census are bean counters for Christ sake.  Do you seriously believe a group of geeky federal workers are these great and powerful political spin masters, plotting and planning to throw major elections and hoodwink the American people?  Wrong organization!



large margin of error anyway

if 2200 census employees interview 60,000 households representing 110,000 individuals per month, that means each employee is responsible for 50 responses to the survey...since most of those are employed already, the opportuntiy to switch survey answers from unemployed is minimal...with the labor force participation rate less than 65%, the average number of unemployed each employee would interview would be around 4...with each survey answer representing around 22K in the final totals, it's unlikely one rouge employee could reduce the number of uemployed by more than 100K...with a margin of error of +/- 300,000 in the monthly change in the number unemployed and +/- 0.2% in the employment rate, it would take at least 4 employees falsifying every unemployed person they interview to swing a change larger than the margin of error...


nice back of the napkin

The article is implying more than 4 of the 2200 are falsifying survey results. I only covered this because over and over again we have CPS results that are so screwy I cannot track down the variance, within the normal error margins.

That said, I frankly don't believe the New York Post. Prove it, let's have an investigation, testimony, evidence.

But guess what I discovered, the BEA used unpublished inventory data in their manufacturing inventories and God knows what else. *That* is not the first time I've found not publicly available data being used in their government statistical reports and to me that is extremely uncool.

That's why I covered the Post article. Clearly there was one person and 4 out of 2200 doesn't seem out of bounds when we tally up the questions per interviewee and the number of surveys each employee would have to obtain. Sick kid, want to take off, oops, just a little fudgy.

Conspiracy and swaying elections, don't think so.

easier to manipulate unemployment up

most census employees are career number crunchers; it wouldnt surprise me that some started work there under their politics is probably in a similar proportion to that of the population...if it was known in the agency that some were fudging reports to make Obama look better, those opposed to him could much more easily switch employed to unemployed than vice-versa, cause there are more of them...


Anyone surprised?

Banks are facing civil and criminal penalties for manipulating LIBOR, any chance that anyone in the Census Bureau will face similar penalties? didn't think so.

First it has to be proven

I have issues with some of these statistics but it's most often unpublished data, methods and so on, never intentional malfeasance.

The New York Post is big on claiming the "government is lying to us" and frankly understandably so, especially Congress and the White House.

So, I think all of the government statistical branches should be audited, but not by politicians, by statisticians. ;)