Higher Earnings That Aren't

[Michael Eisenscher <meisenscher@xxxxxxxxxxx>]

 Barron's Online -- December 8, 1997
 Day Labor

 A surprising factor in hourly earnings

 By GENE EPSTEIN

 Sudden Impact

 Call it the Perfect Report, especially since its apparent
 imperfection is just a mirage. Friday, the Bureau of Labor
 Statistics released its data on November's employment
 picture, confounding Wall Street's expectations and
 gladdening the hearts of those of us in the bleeding-heart
 capitalist camp. In the words of phrase-maker Ed
 Yardeni, the report revealed a hiring panic nationwide.

 According to the BLS, the number of nonfarm jobs grew
 by a whopping 404,000 in November, twice the
 consensus estimate of 200,000. At the same time, the
 unemployment rate ticked down to 4.6% from October's
 4.7% rate, putting it at the lowest level since October '73.
 And even the reported surge in average hourly earnings
 of an annualized 7.2%, a supposed harbinger of
 wage-push inflation that helped roil bond prices, turns out
 to be the one figure that just ain't so. If not for distortions
 in the way the BLS calculates the data, the real increase
 would be roughly half that, or about 3.6%.

 My authority on this is Brian Garvey, a 28-year-old
 graduate student at Boston University. Garvey has
 completed a study showing a strong bias in the monthly
 estimates of average annual earnings that correlates with
 a certain quirk in the calendar. A call to BLS division chief
 Patricia Goetz turned up the interesting news that the
 agency acknowledges the problem and is in the "early
 stages" of looking into it. But until this work is completed
 (and may that day come soon), it's important to
 understand just how this closely-watched indicator is
 being buffeted by the number of weekdays there happen
 to be in any survey month. In fact, listen closely and you'll
 see how I'm able to predict that next month's estimate for
 average hourly earnings should show a zero increase or
 even a decline, simply because December will have 23
 weekdays and November had 20.

 Uncanny? For sure, but as any schoolboy knows, there
 can be 20 to 23 weekdays in a month, leading to
 differences in days between one month and the month
 before of -3, -2, -1, 0, 1, 2, and 3 days. For instance,
 since November had 20 weekdays and October had 23,
 November was a "-3" month. And since, as noted,
 December is a 23-weekday month, it gets a 3 label.

 Now look at the top chart on this page, which shows a
 striking picture: The change in average hourly earnings
 as reported by the BLS has been closely correlated with
 this weekday differential. In fact, based on 87
 observations from January 1990 to April '97, Garvey
 found that the average annual increase in months labeled
 -3 has run about 6%, while in 3 months it's averaged
 about zero. And notice that a similar yawning gap exists
 between -2 and 2 months.

 Now, unless we're to believe that this New Age economy
 has brought a peculiar numerological method to the way
 the nation's bosses bestow raises, something else must
 be going on. BLS chief Goetz believes it's happening in
 her own backyard, but she's not quite sure what it is. In
 her opinion, it probably has something to do with the fact
 that a number of the reporting firms pay their workers
 salaries on a semi-monthly or even monthly basis. The
 workers involved are professionals, like journalists and
 economists. The agency must then convert those data
 into an hourly rate, and it's apparently forced to use the
 calendar in order to do this.

 As noted, I wish the BLS luck in working through this
 particular morass. Meanwhile, the second chart on this
 page, which shows exactly how the next 13 months fall,
 should prove a useful guide to the monthly ups and
 downs of average hourly earnings. Since December '97 is
 a +3 month, look for a near-zero change.

 But let's return to the good news the agency can be
 trusted with. November's employment gains in the private
 sector were widespread, with especially large increases
 in manufacturing, services and retail trade. Manufacturing
 added 44,000 jobs, marking the third month out of the last
 four in which this sector has had a large increase. More
 than 200,000 manufacturing jobs have been added since
 the industry's last employment trough in September '96.
 Pacing the strength in manufacturing, wholesale trade
 added 24,000 jobs, mostly in durable goods. Despite
 November's unusually cold weather, construction
 employment rose by 29,000. And to cap it all off the good
 news, government employment was unchanged.

 At this point, good news for jobs is usually interpreted to
 mean bad news for inflation, but on that score more sober
 thinking should prevail. To begin with, it's worth noting
 that Fed Chairman Greenspan's favorite indicator of labor
 militancy, the "quit rate," fell in November, from 11.2% of
 the unemployed to 10.4%. For the uninitiated, this
 involves the take-this-job-and-shove-it types who were
 confident enough to abandon their situations without
 having another one lined up-what the BLS calls "job
 leavers." When the chairman originally sounded the alarm
 about the rising quit rate, it stood at 11.3%. So despite
 the tight labor market, workers are maintaining that
 humility to which the 'Nineties have been accustoming
 them.

 Moreover, the Asian contagion will not only continue to
 bring increased disinflation to our shores when it comes
 to prices, but when it comes to wages as well. That's
 because currency devaluations make the Asian worker
 cheaper in relation to his American counterpart. And as
 noted above, the recent increase in average hourly
 earnings is somewhat exaggerated and will likely show a
 downside correction come next month.

 But even so, the laws of supply and demand have not
 been repealed. With labor scarcities fairly widespread,
 wages have been rising and will continue to do so as
 long as the unemployment rate remains as low as it is.
 But whether this will lead to a ratcheting up of inflation is
 another matter. A recent study from the New York Fed
 has shown that in the goods sector of the economy,
 increases in the rate of compensation growth simply
 doesn't lead to higher prices, and that in the service
 sector, the relationship is fairly weak. That's because
 goods-producing firms no longer have pricing power,
 bedeviled as they are by global competition, and many
 service-producing firms always did have to exist in a high
 competitive environment. So as much as they may want
 to hike prices in the face of rising wages, they simply
 can't.

 This lack of pricing power means that, in order to absorb
 higher wages, companies will either find new ways to
 boost productivity or will simply have to eat it in terms of
 lower profits. But as Wall Street knows even better than
 Main Street, earnings are more than healthy, and if some
 erosion starts to set in, the companies affected won't be
 declaring bankruptcy anytime soon.

 Greenspan should keep his eyes on the ball. The Great
 American Job Machine is vital for addressing the main
 priorities of our time: to reduce crime, to push people off
 the welfare rolls and onto the employment rolls, and to
 help narrow the widening gap between rich and poor.
 Seeing it do its thing is cause for cheers.

 Postscript: Wrightson & Associates chief economist, Lou
 Crandall, inspired by Brian Garvey's finding about the
 calendar-quirk-driven average hourly earnings, has found
 another such quirk. For some reason, changes in the
 average workweek as reported by the BLS are correlated
 with the number of days in the survey month. In months
 with a relatively greater number of weekdays, the
 workweek is biased upward. Since that's true of
 December, look for an upward bias in the next report.

 But just to confirm that the gods had not gone completely
 crazy, Crandall did another kind of test and has
 reassuring news: The number of weekdays in a month
 shows no correlation at all with the change in nonfarm
 payrolls.


 E-Mail: gene.epsteinnews.barrons.com
Copyright © 1997 Dow Jones & Company, Inc. All Rights Reserved.

===============================