Economics, Data and Conspiracy Theories
Today’s column is about stupidity. Perhaps that's overstating it; to be more precise, it is about the conspiracy-theorist combination of bias, innumeracy and laziness, with a pinch of arrogance thrown in for good measure.
I am talking about the manifold ways various economic reports get misinterpreted, sometimes in a willful and ignorant manner.
That’s right, I am calling some of you ignorant (not you, but the guy next to you). There are things we all should understand about how specific data series work in the real world. However, I want to make a distinction between legitimate criticisms of the details of any economic report and wild-eyed accusations of falsification.
Let’s start with the employment report. Long-time readers know I am not a fan of the monthly data releases, preferring instead to focus on the trend. This report is an attempt to assess the changes in a labor pool of about 150 million workers in real time, and is highly subject to revision. Like all data series based on a model, it is often wrong but useful.
Over the years, I have criticized many aspects of the models that underlie economic news releases: Business birth and death adjustments can yield a serious misreading of the labor market’s health. The tendency to ignore the broader unemployment landscape in the U6 numbers in favor of the U3, or the official unemployment rate, also misses significant trend changes. I have also written about reconciling the household and establishment series. The new home sales data series is also very noisy and unreliable. And don’t get me started on the many problems with the various inflation-modeling issues, such as hedonics and substitution. However, the MIT billion prices project shows that while the Bureau of Labor Statistics inflation data are quite imprecise, they are consistent, and therefor helpful.
Back to the foolishness that pervades the world of the conspiracy theorists.
Let’s begin with perhaps the most infamous charge, made by former General Electric Chief Executive Officer Jack Welch, the month before the 2012 election. His Twitter post was shocking, not just for its partisan rancor, but because it came from the former CEO of one of the U.S.’s largest and most-respected companies.
However, what makes the Welch accusation so entertaining was that it came from a CEO who allegedly engaged in earnings manipulation and accounting shenanigans. Under Welch, the GE Capital financial division could always be counted on to help parent GE beat quarterly earnings estimates by a penny a share, quarter after quarter. Welch’s successor had to settle accounting fraud accusations with the Securities and Exchange Commission for what was done during Welch’s tenure -- all of which made his assertion that the government cooked its books sound cynical and devoid of self-awareness.
But this issue is about much more than one addled ex-CEO. It comes up with regularity from people who ought to know better. Some of it is sheer laziness -- hey, it's hard learning the intricacies of all of the economic reports, and how they are assembled. Some people find it much easier to simply throw around unsupported indictments.
Let’s examine a few of these.
What it says is that sales of new homes in the United States printed at an annual rate of 458,000 for the month of October, up 0.7% from September. But wait, I remember reporting just last month that September’s print for new home sales was 467,000.
Let's be clear: I’m no mathematician but I play one on the Internet. So, even I know that 458,000 is less than 467,000. When you read the fine print of Census estimates on new home sales, you notice they give a margin of error of a whopping ±15-19%.
Yes, that you are not a mathematician is painfully obvious. If you were minimally numerate, you would understand how annualized data is modified and revised. You would also recognize what a margin of error means. It means that the economists at the Commerce Department who prepare this report are aware that this is a noisy data series, one that changes significantly from month to month. In my experience, it's best to assess these kinds of reports by using a three- or six-month moving average.
Note the difference between the phrases “noisy data series” and “cooking the books.” In my business, that’s the difference between understanding data and being a partisan hack.
Next up, let's look at the employment situation, better known as nonfarm payrolls. The biggest criticisms I see are the issue of U6 versus U3, the business birth-death model and people leaving the labor force.
Ignorance of how these models are assembled mixed with political bias can lead to bad conclusions. This Forbes article, for example, misapprehends the nature of financial crises versus ordinary recessions, and seems unaware of how economists reconcile household and establishment surveys.
As for the U3 versus U6 unemployment measures: There are multiple measures of unemployment and the BLS reports all of them. It’s a click away for anyone who wants to see both the data and the specific definitions of each. They tend to move in parallel, but when people choose one series over the other, it is because they are cherry-picking the one that supports their politics or their portfolios (or both).
The birth-death model is another issue for some. It is flawed, but it serves the purpose of capturing employment created by new company formations that the establishment survey -- which typically looks at larger and older businesses -- sometimes misses, especially early in any recovery cycle. My beef with it comes at the end of the business cycle, when it tends to over-reports job creation. We saw that in 2007, when 75 percent of the newly created jobs reported by BLS were due to a birth-death adjustment.
Last, let’s look at the issue of the labor force participation rate. The website Zero Hedge accurately observed that it is falling and is at multiyear lows. But I thought the site Calculated Risk did a much better job of providing context and balance, noting that the decline is tied to demographics -- retiring baby boomers -- globalization, increased productivity and automation.
Incidentally, you can speak to the data geeks at BLS and Commerce, who will happily walk you through any of their models. They will tell you at great length about how the sausage is made. They are completely transparent about it.
There is also a huge difference between upgrading and changing models and surreptitiously changing the resulting data. When the Bush administration altered the business birth-death methodology in 2001, the BLS made a public announcement. There was some debate about whether this would be good or bad for the accuracy of the models. The revision managed to capture more of the newly created jobs early in the business cycle, which was an improvement; but it over-counted jobs toward the end of the cycle, which was a setback.
But that’s a publicly announced change in the model, not a conspiracy. Let me remind you that when George W. Bush tried to fire a few U.S. attorneys for political reasons, that secret lasted approximately a nanosecond, eventually leading the U.S. attorney general to resign.
Now imagine trying to keep the purported book-cooking a secret at the BLS, where thousands of economists, statisticians and clerks help assemble all of the regular economic reports.
There is no second gunman in the data.
This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
To continue reading this article you must be a Bloomberg Professional Service Subscriber.
If you believe that you may have received this message in error please let us know.