Risky Business: Focusing On Monthly Employment Numbers

Some economics pundits needlessly bang their heads against the wall when it comes to the art/science of digesting the numbers. Today’s example: the Labor Department’s employment report for July. Although economists focus on the so-called establishment survey for assessing nonfarm payrolls, the government also publishes an alternative measure of the labor force via what’s labeled as the household survey. These are two different methodologies for quantifying changes in the labor force that, not surprisingly, don’t always agree on a month-to-month basis—July’s statistical conflict was unusually wide. The divergence creates a fair amount of confusion, but it shouldn’t. You can find a clearer view of the trend in the year-over-year changes, which is an antidote of sorts for the monthly noise.

As an example, let’s begin by comparing the monthly figures over the past year for the establishment and household surveys. No one will confuse the two series as dispatching similar, much less identical, results in any given month, as the chart below illustrates. Last month was particularly chaotic: the household survey reported that the number of employed persons in the U.S. fell a hefty 195,000 while the establishment survey advised that total nonfarm payrolls (including changes for government employees) jumped by 163,000 (private nonfarm payrolls increased by 172,000 last month).

There are some in the blogosphere who look at the statistical chasm in the chart above as a sign that the employment numbers, if not quite worthless, are deeply misleading overall. Agreed—if you’re looking at monthly data, which can be a bit like looking for patterns in snowflakes in a wind tunnel.
The good news is that it’s a different story if we compare the data for the two surveys on a year-over-year percentage basis, as shown in the second chart.

Note that the pair of labor market estimates generally agree when we review annual percentage changes. It’s still not a perfect match—we shouldn’t expect it to be. The numbers reflect two different methodologies. But it’s no trivial point that the pair offer comparable messages about the trend. For example, both labor market measures went negative at roughly the same time during the early stages of the last recession. Both also started turning up around the same time just after the recession ended.
Fast forward to today’s labor market update and we find that both indices are rising on a year-over-year basis at relatively high levels vs. their histories. In other words, we see some corroborating evidence from two different measures of employment activity for thinking that the economy is still expanding. (By the way, nothing really changes if we look at year-over-year changes for unadjusted numbers, i.e., before seasonal adjustments.)
The monthly numbers, by contrast, are a mess. But that’s not unusual. Seasonal distortions and variations in calculation rules bedevil the short term. Some of the misleading noise is stripped away if we look at the annual changes. Annual profiling is still less than perfect, but there’s no reason to make macro analysis harder than it has to be. Then again, some analysts are gluttons for statistical punishment. Go figure!