September 07, 2003

Three Criticisms of Productivity Statistics

Morgan Stanley's Stephen Roach has two criticisms of the amazing productivity statistics, both of which are, I think, half right. The first criticism is that we are overestimating the growth in the value of the white-collar work done. The second criticism is that we are underestimating the number of hours that white-collar workers are putting in:

Morgan Stanley: This is where the productivity miracle falls apart, in my view. I honestly don't think we have a clue as to how to measure productivity in the white-collar services sector. The problems lie both in the numerator (output) and the denominator (labor input) of the productivity equation. The production of the proverbial "widget" makes measurement of tangible output in the manufacturing sector relatively easy by comparison. The intangible output of services is a different matter altogether. Measuring quality-adjusted value-added in knowledge-based activities is tough in theory and virtually impossible in practice. Yet that's exactly what the productivity metric requires us to do.

Is it correct to measure the output of a software programmer, for instance, by the lines of code that he or she writes? Or the number of words that an analyst produces? Or is less more? To me, the efficient software program and the insightful piece of analysis wins, hands down. Try measuring managerial output -- hardly a trivial consideration given that managers account for fully 25% of America's total white-collar workforce.

As tough as it is to measure the numerator in the white-collar productivity calculus, I have long been equally critical of efforts to capture the denominator -- labor input. The official data on labor input comes from the establishment surveys of the US Bureau of Labor Statistics; at the crux of this gauge is an estimate of the length of average work schedules. For white-collar knowledge workers, these numbers simply don't make any sense to me. Take financial services -- an industry in which I have spent my entire career. According to the BLS, the average workweek in the financial activities sector was 35.4 hours in July 2003 -- essentially unchanged from the level a decade earlier (35.6 hours in July 1993). I find that most difficult to fathom. Over the past decade, the IT-enabled knowledge worker has seen a radical transformation of work schedules. Courtesy of miniaturized and portable information appliances, together with near-ubiquitous connectivity, workdays have been extended as never before. Yet in this increasingly 24 x 7 mindset, the official data speak of unbelievably short and unchanged work weeks. What a disconnect!

To me, this smacks of a classic measurement problem. The official data seem to underestimate woefully actual hours worked in America's increasingly knowledge-based, white-collar economy. We are guilty of confusing extended work schedules with productivity growth. I've said it before: Productivity is not about working longer. It's all about generating more value added per unit of labor input. To the extent that government statisticians are undercounting work time, it follows they are guilty of overstating productivity. With America's newfound productivity gains skewed increasingly toward the white-collar services sector, this statistical conundrum takes on even greater meaning for the economy as a whole...

The first criticism is only half-right because the bulk of white-collar service-sector work--including virtually all of managerial work--are themselves inputs into further stages of the production process. The management of Daimler-Chrysler helps the rest of Daimler-Chrysler make cars. The management of Nike helps the rest of Nike make shoes. We know what a car is. We know what a shoe is. To the extent that we overestimate white-collar productivity in Daimler-Chrysler's and Nike's value chains, we automatically underestimate blue-collar productivity because the combined output of both--quality-adjusted cars or shoes--is something we know about. It is very possible that we are overestimating white-collar and underestimating blue-collar productivity, but such errors should cancel each other out for the economy as a whole. And yet the statistics for the economy as a whole are very impressive.

The second criticism is also only half-right. Because people are easier to reach, they are spending less time hanging around the office twiddling their thumbs waiting just in case somebody needs to reach them and learn something they know. Because people are easier to reach, they are being bugged more and made to work more outside formal normal working hours. Which effect dominates? I don't know. I do know that people seem to prefer the wired to the hanging-around-the-office lifestyle, and prefer it by quite a bit. But I live near the center of where the most action is. It's not clear to me whether Stephen Roach's point is quantitatively important, indeed, it's not clear to me that it cuts in the direction he thinks it does.

Moreover, there is a third potential criticism of the productivity numbers that Stephen Roach doesn't make, and that I wish he would: the speed-up criticism. More and more, blue-collar and lower-level white-collar workers can be watched, monitored, and assessed. The pace at which they are expected to work can be ratcheted up with much more ease than in past ages. Is this factor--either the reduction in paid on-the-job leisure, or the breaking of the wage-effort social contract in the interest of extracting more value for each wage and salary dollar, depending on your viewpoint--important? I do not know. I wish I did.

Posted by DeLong at September 7, 2003 05:23 PM | TrackBack


Thereís another effect, but I donít know if itís significant or widespread. Consider pure service sector work, where the output is the end product, like a law firm or a consulting firm. Managers want to show a low rate of overhead because it looks good on a proposal to the government, and so here is what they have done in many cases. They offload a lot administrative tasks onto the professional staff. Now the professional types his own report and prepares the graphs and figures using ďdesktopĒ publishing. He types in data to his desktop workstation if necessary. These kinds of things used to be done by secretaries, keypunch technicians and tech writers. The total costs might go up because you have a high-paid professionals doing low-level work. But the overhead rate looks smaller because those support services have been curtailed or donít exist at all. Then again the professional might just work longer and total cost goes down. A lot of people are putting in 60-hour workweeks these days, particularly at Morgan Stanley.

Posted by: A. Zarkov on September 7, 2003 06:50 PM

But for exempt employees what numbers can they collect? Nothing but their managers estimates. But managers have only a loose grasp on what technical employees are doing hour by hour - weekend work when the managers aren't there, night calls to stem emergencies. I just can't beleive the hours are even in the ball park.

Posted by: Dick Thompson on September 7, 2003 07:21 PM

Have any "quality of service" statistics been available during this productivity increase?

It occurs to me that one way to increase productivity in the service sector is simply to give lower quality service. Longer lines, fewer cashiers, longer hold times, longer waits for an appointment, dirtier stores, etc.

It would seem to me that, in the service sector, the main limit to productivity is how much discomfort the customer can tolerate. If your customers will pay the same prices, but tolerate worse service, you're golden. Just keep cutting until the customers scream.

Manufacturing and agriculture wouldn't seem to be quite so forgiving. Cutting staff is more likely to quickly result in reduced production and reduced revenue, or unsalable items.

Since so much of our economy is service-oriented now, that'd help explain part of the productivity increase.

I am not an economist, though, so I'm probably either completely wrong, or stating the glaringly obvious.

Posted by: Jon H on September 7, 2003 09:07 PM

Managers at Nike and Chrysler are NOT all in the services sector. Roach's main point was talking about services work, not just management types.
Software and software service are hard to gauge.
The same goes for all other eBusiness work units.
On the other hand, technology has obviously led to productivity gains. If a part time person can do a payroll a full time person used to do - thanks to Quickbooks - it's a productivity jump.

Posted by: Mike on September 7, 2003 11:53 PM

Thanks a lot, this is interesting, but Brad, you're missing the point about what's new in today's economy i.e. the scale of the accumulated productivity increases in white-collar vs blue:

"The management of Nike helps the rest of Nike make shoes." - No, Nike doesn't make shoes, they advertise and distribute shoes.

"We know what a shoe is." - To some extent, yes, but most of Nike's business ie about to teach and learn us what a Nike shoe is, and how wearing it makes you a better person!

"To the extent that we overestimate white-collar productivity in ... Nike's value chains, we automatically underestimate blue-collar productivity" - Only insignificantly so, making the shoe is what, a $10 business? Obviously advertising and distributing is has to be on a $100/item scale.

White and blue collar are on separate scales (as they say in physics), and hence in different countries, at least when it comes to sport shoes!

Posted by: Mats on September 8, 2003 12:01 AM

Brad writes -"And yet the statistics for the economy as a whole are very impressive."- and yes they are, aren't they? What's new here, in Brad's Nike-example, is the extent to what the improvements in GDP relies on skilled advertisers and vendors making people think that $10 shoes are worth $100.

This subject were earlier touched upon in Brad's critique of Kay's article in FT on US/Europe retail sector comparisons here:

and here:

Posted by: Mats on September 8, 2003 12:19 AM

To emphasize one point - white-collar workers aren't on the timeclock system. This means that they can be made to work a 60-hour week under certain circumstances (like, say, a lingering recessionary job market). Their nominal schedule could still be 9-5, but
they could work 8-6, with occasional nights and weekends. That's just got to distort the productivity figures. Now, some of that is slack time, but that applies to blue-collar workers as well.

A good indicator would be to look at the average workweek figures from the financial services industry (which Roach mentions has a 60-hour week as normal). If the survey figures give a lower figure, then they're biased.

Posted by: Barry on September 8, 2003 04:12 AM

Roach's argument doesn't go far enough to provide a basis for questioning the recent productivity growth estimates. A measurement problem that is static and continuous would require a once and for all revision in productivity (output per hour)estimates, but would not necessarily require a change in our estimate of productivity growth (except in transition). For that we need evidence that the measurement error is increasing as time passes (thus falsely leading us to judge that output per hour is growing more rapidly than it is).

I can't see the basis of such a claim. Can anyone?

Posted by: Jim Harris on September 8, 2003 04:39 AM
Post a comment