What do you think? Leave a respectful comment.

What’s the most reliable indicator of monthly job gains?

The economy added an unexpectedly high 321,000 jobs in November, the greatest monthly increase in almost three years, while the unemployment rate held steady at 5.8 percent for the second month in a row. Job gains for the previous two months were revised up by 44,000.

Inspiring even more good cheer, however, is news that average hourly earnings for private nonfarm payrolls actually rose 9 cents last month.

The Good News in November’s Report

For months, stagnant wage growth has muted excitement about strong job growth. Wages rose an average of 2.1 percent over the year, barely above inflation. The strain that’s had on workers’ pocketbooks helps explain some of the disconnect between impressive economic figures and many Americans’ economic malaise.

As with all month-to-month government figures, however, one month’s spike is typically suspect and it will take a longer-term trend to make the optimism stick. Left-leaning economist Dean Baker, director of the Center for Economic and Policy Research, argues that this month’s 9 cent/hour growth (and 4 cents for private-sector production and nonsupervisory employees) simply makes up for poor wage growth in the previous two months and that the three-month average is still too low.

But there was other good news. The average workweek for private nonfarm payrolls increased by 0.1 hour to 34.6 hours and by 0.2 hour in manufacturing, potentially increasing workers’ earning potential. Factory overtime hours also increased. Job growth was broadly based across all industries, too, although the highest gains remain in professional and business services.

Making Sen$e’s “Solman Scale U7” metric of unemployment, which adds to the officially unemployed part-timers looking for full-time work and “discouraged” workers, decreased to a record low since we started calculating it in 2011 — 13.73 percent.

Involuntary part-time workers decreased — by about 200,000; these are workers who are working part-time jobs because they can’t find full-time work, and their elevated numbers have been a consistent dark spot in employment reports during the recovery. There are still some 6.7 million involuntary part-timers.

We explored the underreporting of part-time workers last month when we discovered that people who work 35 hours or more at multiple part-time gigs are actually counted as full-time workers because of their total hours clocked. Furthermore, the headline number of jobs added does not distinguish between full and part-timers. “So if you’re on for an hour,” the Peterson Institute’s Justin Wolfers told us in November, “you’re counted as having a job.” Most job gains, however, have been full-time.

The trend of voluntary part-timers is a different story. Their numbers have risen — by almost 300,000 from October to November alone. To Baker, that’s good news, too, because it signifies that people who were working full-time just for the sake of having health insurance through their employer are now able to work the hours they want to work while obtaining coverage through exchanges. A victory for the Affordable Care Act, he thinks.

It’s Not Enough

The number that hasn’t budged, though, is the labor force participation rate, holding at about 62.8 percent all year. In October, some 92 million people were not in the labor force, which also means that they weren’t counted as officially unemployed because they weren’t actively looking for a job. Even when adjusting labor force departures for demographic trends (baby boomer retirement, for example), there aren’t enough jobs, says Baker, for all the people who still need one.

Conservative economist Michael Strain of the American Enterprise Institute agreed with some of Baker’s concerns about persistent weak spots in the labor market.

How Did We Get to 321,000?

But the number that shocked and awed this morning was 321,000 jobs. Given the volatility of all monthly data, however, and the fact that the Bureau of Labor Statistics conducts two separate surveys that paint often contrasting pictures of the labor market, a little more explication of the employment situation is in order.

Consider this: The survey of 60,000 individual households (on which the unemployment rate figure is based) paints a much weaker portrait of November’s job gains. It says only 4,000 more people were employed in November. But what about those 321,000 jobs?

The larger figure comes from a different and, many economists argue, more reliable survey of employers. Baker puts almost all of his stock in the employer survey, as does Wolfers, who weighs the employer survey about 80 percent and the household survey 20 percent. The BLS also encourages looking at the payroll survey for the most accurate estimate of jobs.

But if the household survey reflects conversations with actual workers, why is it less reliable? One reason is that it often shows big swings from month to month. In October, the same survey said 683,000 more people were employed — that doesn’t even compare to the 4,000 it showed for November. There’s no way either of those are true, Baker says. “You find often very crazy numbers that clearly didn’t happen in the world” in that survey, he adds.

The employer survey is less volatile in part because it’s a bigger sample. (The much smaller household survey has a month-to-month sampling error four times that of the payroll survey.) And not only is the employer survey sample bigger (144,000 business and government agencies representing 554,000 establishments), but when the BLS makes contact with one employer, they’re also finding out data about many employees; its reach is far wider.

Comparing Apples to Oranges

There are also minor differences in how the data for each survey are collected. For starters, the household survey is more inclusive of different kinds of employment situations, counting self-employed workers, unpaid family workers, agricultural workers and private household workers.

When the BLS asks households about their employment status, they want to know whether individuals were employed during the calendar week that includes the 12th of the month, called the reference week. Surveying employers, the BLS is interested in how many people were employed during the pay period that includes the 12th of the month, and that pay period could be weekly, biweekly or monthly. Theoretically, then, Baker explains, someone could have been counted as employed in that payroll pay period, but tell a surveyor they’d been laid off during the reference week in the household survey.

Likewise, the household survey counts employees who are absent from work without pay — perhaps on extended disability or maternity leave — whereas the payroll survey does not.

The two surveys really are apples to oranges; one looks at people, the other looks at jobs, and that will always be a source of discrepancy. For example, as the BLS points out, it’s possible that a person who changed jobs could be on two different employers’ payrolls during the same reference week. That would count as two jobs, whereas in the household survey, this person would, of course, only identify as having one job.

Even when Americans do work multiple jobs, they’re only counted as having one job in the household survey. They either have a job or they don’t. In the employer survey though, each of those nonfarm payroll positions would be a unique job.

What both surveys have in common, though, is sampling error, and although one survey’s is bigger than the others, it’s a reminder that taking any one month’s figures too seriously is an economics faux pas. To be statistically significant, the household survey has to show changes of plus or minus 436,000 employed people, and the employer survey has to show movement of plus or minus 94,000 jobs. That introduces a lot of variance. Add or subtract those figures from Friday’s reported numbers, and you’ll see how easily one month’s good news can become bad news or vice versa.

Keep Your Eyes on Payrolls

The bottom line here is that data between months is noisy. And the takeaway is that when you want to know what happened to employment, look at the payroll survey.

Support for Making Sen$e Provided By:

Support for Making Sen$e Provided By: