When the Numbers Don't Add Up

One of the side effects of becoming so enamored of big data is that we may forget that this information is no more infallible than the people who created it.

We often accept the veracity of the numbers we receive without critically examining how it was sourced, prepared, analyzed and presented. We tend not to challenge methodologies or metrics unless we work in an academic setting where scholarly discourse in pursuit of truth is a profession - or in politics where discrediting one's opposition is also a profession, truth be damned. 

In fact, because of our ostensible dependence on numbers to bolster whatever case we may be making, in whatever venue we may be making it, the compilation and presentation of data has become a highly personal act, fraught with risk and uncertainty. Disagreements about strategy and policy have become increasingly emotional as the stakes grow and the information becomes less reliable. This is not necessarily because data quality has deteriorated, but because we are being asked to make decisions about issues on which little real data are available. Globalization, technological advances and speed have all reduced the certainty with which we work. Not a bad thing, because without taking risks there is no growth. But all too often decision and policy makers are being asked to verify a degree of confidence, even assurance, that defies logic.

It is precisely because of this pressure that the producers and consumers of data must become both more critical and forgiving. More critical about the ways and means that data are generated, and more forgiving of the fact that the degree to which our world is changing means that there are many things we simply can not know with anything approaching authority. JL

Caroline Baum reports in Bloomberg:

Just because economics relies on numbers doesn't make it a mathematical science.
For example, gross domestic income -- the costs incurred and income earned in the production of the nation's output -- should equal gross domestic product. It doesn't. Ever. The Bureau of Economic Analysis adds up the two columns, draws a line and reconciles them with the notation, "statistical discrepancy."
Sometimes there are anomalies within GDP. BEA's third guess at first-quarter GDP was a lot weaker than growth implied by labor inputs (employment and hours worked). Real GDP growth was revised from 2.4 percent to 1.8 percent -- one-fourth of last quarter's output gone in a flash! The major source of the adjustment was to real consumer spending on services, which was slashed to 1.7 percent from 3.1 percent, based on new data from the Census Bureau's Quarterly Services Survey. The QSS, which gathers revenue from the sales of a wide range of services, is a relatively new addition to the BEA's statistical library.
Joe Carson, head of global economic research at AllianceBernstein LP, was quick to point out (to BEA, too) the inconsistency between reported GDP and output implied by an alternative method of calculation: using the sum of aggregate hours worked (the number of employees times the number of hours) and productivity. "Hours worked in the private service sector is growing faster than output, which would imply a decline in productivity," Carson said. "If that were true, firms would be shedding workers rather than hiring."
Neil Dutta, Head of Economics at Renaissance Macro Research, had trouble with the math as well. Private hours worked for the overall economy rose 3.6 percent and productivity increased 0.5 percent in the first quarter, implying a 4.1 percent increase in GDP.
Could the Labor Department have overestimated employment and hours? "It's hard to see why," given solid growth in individual withholding and corporate taxes reported by the Treasury, he says. Tax data tend to be reliable because people don't withhold taxes, and corporations don't pay taxes on income they didn't earn. (Sometimes they don't pay it on earned income either.)
That leaves productivity growth, which is a derived number: output divided by hours worked. Mathematically, it has to be revised down with GDP. "Whether that's an accurate reflection, given strong tax receipts and hours, is the question," Carson says.
And one that may not be resolved anytime soon. This is a problem for Federal Reserve policy makers, who seem to be relying on high-frequency data to make decisions about the size of their monthly asset purchases, and financial markets, which are focused -- for good reason -- on the Fed's intentions.
Given the erratic nature of the data, taking a longer-term view would probably produce better results and might even allow markets to do the same. Heck, next month the BEA will release a comprehensive revision of the National Income and Product Accounts dating back to 1929. It's too soon to expect a reconciliation of the first-quarter's anomalies, but the Great Depression might undergo a tweak.