November 09, 2004

Department of "Huh?": Capital Deepening and Economic Growth

The London Economist writes:

Economist.com | Economics focus: Economic commentators toss around the term “productivity growth” as if there were one widely agreed definition. There isn't. America's favourite measure is output per man-hour in the non-farm business sector. Since 1996, this has increased at an annual average rate of 3%, double the pace of the first half of the 1990s.... [G]rowth in GDP per man-hour quickened in America after 1996, while it slowed in most European economies. Yet all this reflects only labour productivity. A better measure is multi-factor productivity (also called total factor productivity) which tries to capture the efficiency with which inputs of capital as well as labour are used. If workers are given better machines and equipment, this will automatically boost output per man-hour, even if there is no gain in overall economic efficiency once the extra capital spending is taken into account...

Too which the only response is, "Huh?"

Due to technological progress, a dollar spent on computers and peripherals today buys twelve times the information-processing power of a dollar spent on computers at the start of the 1990s. The extra computer power at the disposal of the average worker boosts labor productivity and incomes. The total factor productivity measures don't count the effects of this capital deepening driven by technological improvements in the manufacture of information processing equipment. Yet this is the very definition of an increase in economic efficiency: no increase in the share of resources diverted from consumption to investment, yet a hell of a lot more output.

Capital deepening driven by the extraordinary fall in the real price of information-technology capital is *the* big fact about economic growth in the industrial core over the past decade. To choose a particular measure of productivity growth that deliberately excludes this factor from consideration is perverse. For workers to be given better machines and equipment because we have learned to make better and cheaper machines--it is simply not true to say that there is "no gain in overall economic efficiency once the extra capital spending is taken into account." First of all, there is a great gain in economic efficiency. Second, there is little increase in capital spending: there is a twelve-fold increase in information-technology capital accumulation, but that's because equivalent machines can be made one-twelfth as cheaply, not because we are spending twelve times as much on computers.

Posted by DeLong at November 9, 2004 06:57 PM | TrackBack
Comments

Hmm. My reading of the excerpt from the Economist was that it was talking about better machines that did require more capital spending.

In that case, it would seem to make sense that increases in labour productivity would not necessarily indicate improved economic efficiency, depending on the relative value of capital and labour inputs.

The capital deepening I understand you to be talking about arises from the improved bang-for-buck of faster computers at the same price, raising labor productivity without increasing nominal capital expenditure.

Is that a separate factor altogether (technological progress, better education?)

Not an economist...

Posted by: ToonArmy at November 9, 2004 07:27 PM

Hmm. My reading of the excerpt from the Economist was that it was talking about better machines that did require more capital spending.

In that case, it would seem to make sense that increases in labour productivity would not necessarily indicate improved economic efficiency, depending on the relative value of capital and labour inputs.

The capital deepening I understand you to be talking about arises from the improved bang-for-buck of faster computers at the same price, raising labor productivity without increasing nominal capital expenditure.

Is that a separate factor altogether (technological progress, better education?)

Not an economist...

Posted by: ToonArmy at November 9, 2004 07:32 PM

Assuming that increases in processing power can be directly counted as an increase in assets based on 1980's computer pricing seems unfounded. Not only is this practice not done for any other type of equipment (hence introducing a bias for IT heavy economies), but it's fairly obvious that most uses of computers are not limited by the speed of the machine, but by the operator. A data entry clerk doesn't work 20 times as fast as he/she did in 1980 because he/she has a faster computer.

Posted by: ExLurker at November 9, 2004 07:50 PM

The software actually drives the quality/ROI, not the hardware.
It's like the difference between a hayburner, a stick shift, and an automatic transmission. The first substitution dramatically drops the horseshit ratio to zero, but the second one just means you can drink your coffee on the way to work. That's how the hardware works.
The software is different. You can deskill a job by putting pictures on a cash register to the point that you can prefer immigrants to anglos because the immigrant can explain in a customer's language exactly what a hamburger is, and whether it is dietary pollution to eat one...

Posted by: wkwillis at November 9, 2004 08:07 PM

I would suspect that some of the productivity gains could be a result of salaried employees working longer hours, but the standard 40 hour workweek still being used in calculating their productivity -- or is an adjustment made for this?

I agree with wkwillis that software is probably more important than hardware -- though contemporary hardware may be necessary for hosting certain kinds of software that would have been unfeasible 10 years ago (complicated internet applications, such as online order fulfillment apps, for example).

Posted by: hamstak at November 9, 2004 08:34 PM

I would suspect that some of the productivity gains could be a result of salaried employees working longer hours, but the standard 40 hour workweek still being used in calculating their productivity -- or is an adjustment made for this?

I agree with wkwillis that software is probably more important than hardware -- though contemporary hardware may be necessary for hosting certain kinds of software that would have been unfeasible 10 years ago (complicated internet applications, such as online order fulfillment apps, for example).

Posted by: hamstak at November 9, 2004 08:37 PM

I would suspect that some of the productivity gains could be a result of salaried employees working longer hours, but the standard 40 hour workweek still being used in calculating their productivity -- or is an adjustment made for this?

I agree with wkwillis that software is probably more important than hardware -- though contemporary hardware may be necessary for hosting certain kinds of software that would have been unfeasible 10 years ago (complicated internet applications, such as online order fulfillment apps, for example).

Posted by: hamstak at November 9, 2004 08:38 PM

Hamstak: yes, an adjustment is made. Hours worked are determined by asking the people who work them, not by counting employees and multiplying by something.

Posted by: Jonathan Goldberg at November 9, 2004 09:04 PM

I find basing productivity increases on increased processing power to be a highly dubious proposition. Maybe for some fields it can be directly related to increased productivity, but I doubt it holds for, say, word processing. I can boot up my old original PC and it isn't really any slower than my current WinDozer for typing purposes. Sure, there are huge gains from internet connectivity, but how does one measure those? Is there a reasonable base to assume that processing power speaks to that?

Posted by: praktike at November 9, 2004 09:10 PM

ToonArmy: To put more on top of your argument, not only does the machine cost money, but somebody has to manufacture it, and it burns fuel and/or electric power.

ExLurker: And it is debatable whether the use of the computer has made the office more paperless, and has reduced the number of memos, printed or not (I doubt that), or whether the number of meetings has been cut.

hamstak: I'm sure you're right about salaried workers. I don't know how you would adjust for something that you don't measure. Another aspect is the attribution of labor in outsourcing or offshoring situations where the product is sold domestically or under the name of the domestic headquarters. I'm sure adjustments are made for material or service imports, but surely at deflated prices. Any idea anybody?

And otherwise my standard counter to the supposed efficiency increase through computers & internet outside of offices (although it applies only to one aspect of business) is that customer "service" and support is increasingly pushed to web FAQs and "self-help" searchable support case collections, while the response time and quality of telephone support has gone down. Well, sure I can give you more product with fewer people any day, at inferior quality.

Posted by: cm at November 9, 2004 09:21 PM

Jonathan Goldberg: How would this be done? I don't write a time sheet, so how will my boss, or rather the HR or payroll department, know how many hours I work? At best my boss will know what work has been assigned to me, but not how many hours I take to complete it. And a link would be welcome.

Posted by: cm at November 9, 2004 09:25 PM

*Only* a 12-fold improvement?

Lessee...

In the early '90s, my computer was a 50 MHz 68030 with 8 MB of RAM and an 80 MB hard drive.

It cost $4000.

Today, it is replaced by a 1GHz G4 with 640 MB of RAM and an 80 GB hard drive.

It cost me under $1000.

That's a 1000-fold increase in disk space, an 80-fold increase in RAM and a 20-fold increase in clock speed (not a useful comparison: the G4 does many times as much per clock-cycle as the 68030; we're talking easily a 100-fold increase in processing power). All for a quarter the price (even less, in constant dollars).

Raw power, of course, isn't everything. There are so many classes of things that I can *do* with the machine than I could in the early 1990s.

The image of data-entry clerks slaving away at much faster, but ultimately not much more useful machines is largely illusory. Nowadays, most of the information that might previously have been hand-entered by such clerks is entered automatically, via bar-code scans, or is received via "paperless" electronic form submission. Today's workers using computers are not simply using them to do the same task faster. They are, frequently, using those computers to do entirely different tasks than they were in the early 1990s.

Posted by: Jacques Distler at November 9, 2004 09:38 PM

Jaques Distler: That's right, but my (seriously meant) question to you is how has your life quality improved? I'm not saying it has not. But I suspect when you go to the supermarket you are waiting approximately as long at the computerized checkout as on the manually operated one prior (because while the checkout is quicker, what the store management tries to optimize is throughput of merchandise per unit of operating cost, not individual wait time). With 25% higher checkout speed, 4 instead of 5 checkouts will be manned. When ordering an airline ticket on the internet, I'm sure you pay more in real terms than previously at the travel agent (although in fairness 9/11 had a part in this). What drives automation is not primarily expansion of production, but staff reduction, to the extent that market saturation is approched.

And your 100-fold increase in processing power has been partially matched by a 30-fold increase in processing power requirements on the part of "evolving" applications (just to say some number), with much less than a 30-fold increase in functionality or effectiveness.

Of course that's not the whole picture. Computers have facilitated new qualities of research, and e.g. improved medical procedures are available that did not exist before. Cars have become safer and more comfortable. Computers have created new ways of communicating with others, and of entertainment, etc.

But they have also facilitated the squeezing of more work out of employees through cell phones, pagers, wireless email devices, and laptops that you can take home so you "don't have to stay in the office".

Posted by: cm at November 9, 2004 11:09 PM

Increasing processing power is only going to increase the productivity of a subset of all people using computers in their daily work. People who only use word processors and spreadsheets are not going to get miraculously faster at typing as new machines become available.

However, industries which rely upon large database updates and queries are going to see an increase in speed per employee. The more computationally intensive the task, the more the processor will help.

Who thinks word processors are more important to the current economy than databases? Raise your hands...

Posted by: John Bates at November 10, 2004 12:22 AM

Doesn't Brad's argument fail on two counts?

- It is based on increased productivity in the manufacturing of computers. Machines are now made much more cheaply than before, but, once they exist, they should be treated like any other piece of capital equipment. If there are more machines (in value terms) per worker, this is capital deepening, regardless of how much cheaper it is to build machines.

- Expenditure items are expressed in real terms in the national accounts. If, as it seems, "machines" are not only cheaper but also of much higher quality than before, this should show in the statistics (large deflator, etc).

Posted by: Luis at November 10, 2004 01:38 AM


Brad is correct, but the usual figures need to be qualified in various ways. Rapidly declining prices for computers implies rapid depreciation of the existing capital stock, which means that the gap between net product (the economically relevant measure) and gross product (the usual one) is increasing.

Posted by: John at November 10, 2004 02:12 AM


In addition, for international comparisons (what was done in The Economist) changes in the price of capital goods shouldn't matter.

Posted by: John at November 10, 2004 02:15 AM

Andrew Sullivan has an absolute disaster of a post on Bushian tax reform today. It's here:

http://www.andrewsullivan.com/index.php?dish_inc=archives/2004_11_07_dish_archive.html#110006191390426317

My rather long-winded takedown here:

http://reconomist.typepad.com/reconomist/2004/11/bad_tax_ideas.html

Posted by: Marshall at November 10, 2004 02:52 AM

I think I detect a tendency to think of computers in the work-place as generalized desk top units. We should probably think instead of the proliferation of imbedded chips in equipment that doesn't look like a computer. The cash register at the Gap is not just better because it reads the tag for the clerk, it's better because it does the inventory, and in a limited way, writes orders for fresh stock. Delivery vehicles help drivers find their way. Quality control takes place with much less human intervention.

Thinking in terms of the desk top computer is natural for those involved in intellectual work. Seeing limits to the benefits of increased computing power is pretty natural, too. We can caluculate, store, retrieve, draw, test - same as a decade ago, only better. Much of the massive power of the chips we use lays dormant while we type on a word-processor that is only better by the number of new bells and whistles. We can glance over at the data entry chappy who is doing things much the same was as a decade ago, and wonder how much the new chip really does to improve his efficiency. That is not the case for workers in many specialized fields (think about design of complex materials), nor is it the case for workers who are are using super-tools away from the office.

Posted by: kharris at November 10, 2004 04:47 AM

How about network externalities as "the" productivity story of the last decade? Better computers and software certainly has been a help, but tying everything together has really made the difference. I'm trying to imagine my current company without the network, and I'm not sure we'd even be in business.

Posted by: Jim Harris at November 10, 2004 05:59 AM

Deployment of assets/capital!

Surely the productivity gain is in how the extra technology power is deployed. The thread of the argument is that just because you have 20x processing speeds does it mean that you actually do your work 20x faster. You may invest in 20x faster technology, but it only means you actually do things 2x faster for example.

The issue is that its often easy to measure investment in capital, and the production possibility frontier of the capital, but its not rational nor feasible to achieve this. The question for economists is how to relate IT expenditure back to productivity gain.

This relates back to skills - e.g. a fundamental flaw in much UK skills policy is that a supply side approach - i.e. pumping out academically qualified people - will solve the productivity gap. But it is not just in the supply of skills, but in how they are used and deployed in an organisation that counts (demand).

As ever, economists are left to try and make informed estimates of the relative impact of factors of productivity on productivity performance. I would be interested to see more studies on how IT has impacted on business performance. There are interesting examples of zero impact, and highly positive impacts.

Posted by: GlennA at November 10, 2004 06:06 AM

Deployment of assets/capital!

Surely the productivity gain is in how the extra technology power is deployed. The thread of the argument is that just because you have 20x processing speeds does it mean that you actually do your work 20x faster. You may invest in 20x faster technology, but it only means you actually do things 2x faster for example.

The issue is that its often easy to measure investment in capital, and the production possibility frontier of the capital, but its not rational nor feasible to achieve this. The question for economists is how to relate IT expenditure back to productivity gain.

This relates back to skills - e.g. a fundamental flaw in much UK skills policy is that a supply side approach - i.e. pumping out academically qualified people - will solve the productivity gap. But it is not just in the supply of skills, but in how they are used and deployed in an organisation that counts (demand).

As ever, economists are left to try and make informed estimates of the relative impact of factors of productivity on productivity performance. I would be interested to see more studies on how IT has impacted on business performance. There are interesting examples of zero impact, and highly positive impacts.

Posted by: GlennA at November 10, 2004 06:07 AM

On Hamstaks (rather OT) question: In my job downsizing was done by going from 3.6 FTE to 1.0 (me), assigning some tasks to other areas, and leaving me with the rest, requiring me to decide which of the remaining tasks really needed to be done. My pay increase barely took me as high as the lowest-ranking person I replaced.

By having about 1000 lbs. of unnecessary and unusable files shredded, I did in fact improve efficiency. The previous operation was highly traditionalistic. But inevitably there came to be disagreements about what needed to be done, and I was the fall guy. Ultimately, after I had the situation almost fixed, I was replaced by a younger, more cheerful, less burned-out person.

I never worked more than 40 hrs. a week, but I was worried about something all the time, day and night, especially because some of the companies we relied on were also downsizing and merging and thus could not be relied upon to do anything right. I spent about 5-10 hours a month for about 10 months dealing with problems with CMS-Fisher-Coulter, and Baxter-Scientific Products-Fenwall-Travenol was also a constant problem.

The quality of life cost was enormous. In a lot of ways I'm still paying, since I never did get my job back on track.

So anyway, I never read stories about increased productivity with the joy that others do. It's not a good thing for everyone. For a lot of people, the jobless recovery was an especially wonderful thing, and it's hard for me to see why we shouldn't ask whether a jobless recovery wasn't the goal.

Some things are win-win, and some aren't. One poster here (Surowiecki) claimed that the labor victims over the last 10-20 years are only the 2% in manufacturing, but I wasn't in manufacturing, and I wasn't laid off and in fact got a series of smallish raises. Not at all worth it.

Posted by: Zizka at November 10, 2004 06:19 AM

I have always assumed that productivity is measured simply by dividing output by labor costs. How else could you do it?
Which leads to conclusion that productivity can be increased either by increasing output or decreasing labor costs. I suspect the latter is what's driving the productivity increases.
As for the increased productivity of IT I think Brad is simply arguing from common sense, an untrustworthy guide in the realm of economics.

Posted by: Bil at November 10, 2004 06:23 AM

Just to add my wrinkle to what kharris says above, the physical increase in IT processing power, combined with ever cheapening costs, serves as a quasi-metaphorical proxy for a very wide range of distributed effects. Most of all, increased physical processing power goes hand in hand with increased connectivity, and it is especially the synergy of IT with telecommunications that is novel and potent. This diffusion of increasing information-processing capacity results not just in the emrgence of more efficient or new markets, (Ebay!), but the development of diverse applications across different sectors of production and distribution. Perhaps most of all, it impacts upon design, resulting, together with an increased rate of obsolescence that accompanies decreasing costs, in an increased flexibility in real capital goods. And that, in turn, allows for a reorganization of business structures, (i.e. corporations). Perhaps one day in retrospect this would all be looked back upon as the equivalent of a second industrial revolution. (It's too soon to tell, but then so was the original industrial revolution a lengthy process.) But like the first industrial revolution, this one carries with its productive progress, its own forms of immiseration. One of the spurs of development and investment in applications of IT is precisely the increased control over production and distribution processes that it allows management/capital vis-a-vis labor/employees, together with the increased substitutability of capital for labor that comes with ever diminishing costs. This, in turn, would effect the capture and distribution of increased productive surpluses between, broadly speaking, capital and labor.

Posted by: john c. halasz at November 10, 2004 06:30 AM

There are also issues of valuation. The distortion of prices -- the basic signaling mechanism of a market economy -- by Asian government exchange rate manipulations must corrupt all calculations.

Posted by: jm at November 10, 2004 06:49 AM

Brad DeLong wrote, "Due to technological progress, a dollar spent on computers and peripherals today buys twelve times the information-processing power of a dollar spent on computers at the start of the 1990s."

praktike wrote, "I find basing productivity increases on increased processing power to be a highly dubious proposition. Maybe for some fields it can be directly related to increased productivity, but I doubt it holds for, say, word processing."

I agree with praktike. I'm typing this on a cruddy 550 MHz Compaq. We in my particular office (part of a larger organization, with more office space nearby) "downgraded" to these cruds because a thief stole all the good computers and tore out the building's alarm system, and the alarm hasn't been fixed yet.

And yet, would having a 2.8 GHz machine make any difference? Nope. All I do is e-mail (connecting to a server), connect to a UNIX server, etc. I *did* stick in some extra RAM, without which the machine was very slow. But that's largely necessitated because the cheapening of RAM led to MS and other software companies writing bloatware that adds very little extra value.

So if you want to claim that all these wonderful semiconducting devices are leading to vast increases in productivity, you should also factor in (into all macro calculations---not referring to productivity here in particular) the fact that (a) when the computer is off or just doing a mindless task like wordprocessing, all that good capital is going to waste, (b) the increased rate of capital depreciation given all the rapid improvements in tech.

And don't forget that the overall measure of productivity depends on the measure of real GDP, which depends on inflation, which depends---for the past few years, anyway---on hedonistic adjustments made by the BLS. IIRC the size of the hedonistic changes are extremely large.

BLS is a great organization, overall, but do I have faith that they actually look at issues like "what good does a 2.8 GHz CPU really do for most endusers over say a 800 MHz CPU" in an intelligent way? No.

Posted by: liberal at November 10, 2004 08:23 AM

>And yet, would having a 2.8 GHz machine make any difference?

Nope. Brad would do well to Google "Thin Client" to find out how the computing paradigm (god, I just hate the word "paradigm" but if you want to speak to the people you gotta use their language) is already changed to a point where heuristic evaluations like "12x the processing power" have been meaningless for at least 3 years.

Jaques - how much of your new hard drive is filled? Porn doesn't count!! ;>

Posted by: a different chris at November 10, 2004 09:02 AM

I’ve often said that when we talk about efficiency we should always ask “Efficient at What? Efficient for whom? Efficient by what standard? Efficient for how long (and then what?)?” Similarly for productivity, accountability, and so on.

Another nagging concern has long been this, “What do people mean when they say economic efficiency?” I have a stack of economic books (including textbooks) close by so I don’t lack for definitions. I Just don’t understand how flippantly people use the terms. Consider Brad’s use here:

…"it is simply not true to say that there is 'no gain in overall economic efficiency once the extra capital spending is taken into account.' First of all, there is a great gain in economic efficiency."

Huh? I learned many years ago that we ought to be careful to consider impacts to systems both in total and at the margin. I learned also to recognize the wisdom of Garrett Hardin’s restatement of the second law of ecology, "You can never do just one thing!"

I am therefore skeptical of Brad’s claim, while admitting that we have made great strides in the ability to substitute so-called capital for so-called labor in making things, in service, etc. We are clearly transforming society and culture in wild ways. Capitalism may or many not survive the transformation.

In the meantime, here we sit – fat, dumb, and deeply divided – in a country that has botched fiscal and monetary policy for so long that we have given up our jobs (and seem hell-bent on accelerating that exodus), and have wandered into what Kevin Phillips (WEALTH AND DEMOCRACY) calls “financialization.” We sell houses, financial paper, etc. to one another at ever-increasing bubble prices in the broad pretense that we are doing something good for society and culture.

It all looks like one of Hyman Minsky’s Ponzi Finance moments to me.. Welcome to Jesus bubbleland!

Posted by: Dabbler Dave at November 10, 2004 09:05 AM

Desktop computers are not capital equipment. They are office supplies.

But chips built into capital equipment are capital equipment.

How much of the improved information provided by IT is worth anything at all? In my limited experience the important decisions are usually based most importantly on things that still aren't known. Next year's demand. What the competition will be doing in the meantime. The price of oil in china. Getting some of the rest of the data exact instead of within 10% is not all that useful.

How much is the improved information provided by the web worth? It's easier to find new suppliers, cheaper to advertise, etc. Instead of a hard scrabble to find leads, you get swamped in misinformation and disinformation. It doesn't necessarily result in better choices at all, or even quicker, easier, or cheaper choices.

However, all this stuff isn't going away. Whether or not it improves productivity it's the sea we have to swim in. If I was part of a government study looking at whether to recommend that the US government provide incentives for businesses to avoid the net and avoid excessive computer use, or looking into another innovation and choosing whether to fund it, then I'd be interested in the degree of productivity created by current practices. Since I'm not, I'll keep a degree of skepticism and beyond that just swim as best I can.

Posted by: J Thomas at November 10, 2004 12:39 PM

Real-world example:

I can electronically match customer orders with inventory through buying a $200,000 machine. This allows me to eliminate ten of my twenty $20,000 clerks. (The other ten are needed for error checking, telephony, etc.)

So I invest $200K in capital and eliminate $200K in jobs. At the end of Year 1, my costs and production are the same.

In Year 2, the improved processing allows me to either (1) double my business and ten more clerks
(I am now back at the same staffing levels as pre-machine, but my sales and profit are doubled) or (2) reduce operating expenses for year two by $200K.

The choice of (1) depends on whether there are supply constraints (other bottlenecks in my production) or demand constraints (the market is saturated at my price point--even if I adjust the price point down for the lower operating expenses). Otherwise, I go with (2), passing on some of that gain to the customers (market) if it fits long-term strategy and market conditions.

In either case, there is not likely to be a direct relationship if I increase processing speed (which is why the 1,000x increase in processing ability has only led to about a 12x increase in processing).

It is possible to buy a machine that produces exactly the cost savings of its expense, just as it is possible to hire a person who produces only enough work to cover the cost of their salary. But the latter is charity, not business, and the former is equally non-business--even when it is incented by other factors (e.g., the Bush administration's tax benefit that keeps getting extended), it's not an optimal use of capital.

Posted by: Ken Houghton at November 10, 2004 01:57 PM

Suppose it used to take 10 dollars of labor to produce a widget. Now it only takes 8. But if 2 additional dollars of capital expenditure were needed for each widget, that would cancel out the labor gains. So I see it as reasonable to look at the total factor cost.

Posted by: Ionides at November 10, 2004 03:47 PM

Is there some confusion here between unit costs and productivity? Several comments refer only to the cost of labor, rather than hours of labor, which is the relevant unit when discussing productivity. That is not to say the decisions about the mix of labor and capital, as well as the motive for employing either, are not settle by looking at costs and revenues, but we need to distinguish between units of input with costs of input in our analysis.

Posted by: kharris at November 10, 2004 04:01 PM

kharris: Then please clear up the methodology for me the layman. Does "productivity" always necessarily imply "output per (paid) human worker hour"? What is the concept of "output per unit of worker hour and material/energy input" called? Is that not the factor productivity?

Which is not to say that any of the measures is right or wrong. But excluding inputs like energy and material that are not free (in either a financial, scarcity, or environmental sense) does not provide the whole picture and may eventually be misguided.

Posted by: cm at November 10, 2004 08:47 PM

"Due to technological progress, a dollar spent on computers and peripherals today buys twelve times the information-processing power of a dollar spent on computers at the start of the 1990s. The extra computer power at the disposal of the average worker boosts labor productivity and incomes.

Too which the only response is, "Huh?"

=========

The predominent concensus today is that, despite the IT marketplace "hurry, hurry" sales boasts of great productivity gains with cheapification of hardware, nevertheless, productivity is some 50% greater in terms of raw processing, but at an overall labor and materials cost 200% higher.

I'll give you an example. Early in my career, I graduated with a computer science degree on the IBM mainframe and punchcard system of the era.
I graduated into a public works job, where the highest technology was using a mechanical adding machine, where you'd type in your formula, and then listen as the seconds dragged into minutes, while the machine ground out one digit at a time.

Projects took some years to reach fruition, with a team of shirt-and-tie engineers, backed up by pencil-and-ink drafting technicians. Projects still take years today, with more permitting and administrative overhead than ever.

Now, calculations on computers happen faster than you can input the numbers. The achievable limit of calculation power, except for the rarified atmospheric and aerospace sciences, was reached a long time ago. Bar codes, RFID, all that check-out automation only marginally improves productivity at the retail level. GPS farming, again, marginal productivity increase at huge overhead, while the tilth continues to become depleted, and yield oscillates about its all-time highs achieved years ago due to drought.

Drafting by computer is only slightly faster than the old pen-and-ink days, again, except in CADAM direct-to-manufacturing systems. Still, even the CAD software vendors admit, processing is faster, but at a much higher IT investment and maintenance cost. The limit of productivity gains was reached late in the 1990's. Even the IT experts have looked at micro-circuitry, and determined the scaling limits have been reached.

Today "productivity increase" is downsizing and outsourcing labor, squeezing more throughput out with less massagenation but more defects and code bloat, or by sheer economic manipulation, throwing off a whole division to re-balance your quarterly P&L. That's how these corporations achieve impossible 20% "productivity" gains YOY.

Look at Fed-Ex or UPS for productivity reality.

The only real true productivity gains will be achieved in India and China, where software is pirated and free, and labor is $0.10 on the US$.

Which is why American IT and manufacturing production is headed south, literally, and why you shouldn't believe anything in the market.

Posted by: Tante Aime at November 11, 2004 09:21 AM

A question from someone who has never been much interested in economic theory: What is it that is being 'produced.' What's the constant in the equation? So we are able to work faster and faster, but at what? Do we have more leisure or less? Is there more food grown or less? Is it simply a feedback loop? Are we creating jobs that produce ideas that enable us to push more paper around that require us to create more jobs? It all seems absurd to me- it always has of course- but since we're on the subject, I'm curious what partisans of economic progress would respond.

Posted by: seth edenbaum at November 11, 2004 03:06 PM

Multifactor productivity is pretty deep water, so i hesitate to wade in ... but it seems like Dr. DeLong's objection to multifactor productivity - that it doesn't reflect technology improvements - is valid only if those improvements are assumed to be captured in the calculation of the contributions of capital services to productivity. I'm not sure whether that's supposed to be true, or if so, the extent to which it really is true in practice.

As i understand it, multifactor productivity (the way BLS calculates it, anyway) is output divided by an artifical denominator that combines labor inputs (denominated in hours, with a variety of adjustments to reflect changes in the age, experience mix, and education of the labor force) and capital inputs (denominated in dollars, and imputed based on estimates of the total value of productive assets, plus the expected real return on assets in each sector of the economy, minus the expected rate of depreciation on those assets by type of asset).

Maybe in theory they're supposed to capture technology improvements in the rates of return & depreciation when they calcualte the capital inputs. If that were the case, you'd expect that there'd be some noticable movement in the rate of output per unit of capital over time, and probably an upward trend at least during the internet boom, right? But it's not there - the long-term trend in output per unit of capital was negative from the mid-60s through the mid-80s, and since then has been flat (with a negative turn in the past few years as the economy slowed down).

To me, that says - whether by oversight or by design - their measure of capital inputs doesn't really capture technology improvements. That means, to the extent such improvements are captured in productivity data at all, they ought to show up in the multifactor productivity number - which is, after all, just captures the residual gains in output that aren't accounted for by any of the specific factors that go into the combined output denominator.

Of course the impact of technology is buried in there, along with other uncountable factors like improved organization, better morale, smarter management, more pliable customers, etc. [Several other quantifiable things are left out, too, such as capacity utilization and competitiveness of markets, which you'd think would impact productivity - but no measurement's perfect, and the sheer amount of math involved here already gives me a headache.] The main point is, while multifactor productivity misses the impact of technology improvement, it shouldn't be any worse in that respect than the regular productivity measures like output per hour.
It's really the output measure that's missing this dimension, isn't it? Since the definition of output has to assume that what people are paying for a product is in fact what it's worth - which may be true enough at a point in time, but over time it's pretty obvious that what constitutes fair value for something changes in ways that the price can't reflect very well, especially for complex purchases like cars and computers.

I think that's the point that several other posters have made, more concisely & on a more intuitive level than me ... but my curiosity was peaked by all this, and having spent the past few hours digging into the MFP literature, I had to do something with it!

Thanks for the great blog.

Posted by: Tom at November 11, 2004 10:25 PM

Tante Aime: Good presentation. And don't forget that computers run on oil/fossil fuel (electric power), whereas the human engineers run on food!

Posted by: cm at November 12, 2004 12:24 AM

To Tante Aime: You're right that there are limits to the productivity gains you can get from any technology or innovation, and that we're probably in the range of diminishing returns for many of the productivity boosters that have become commonplace in the past 10 or 20 years. You're right that we may be in a cycle right now where the only strategies that a lot of companies have available to increase their productivity in the short term (or at least the only ones they can think of) are to force down wages, reduce headcount, and/or dispense with or outsource functions that they don't do very efficiently. You're right that some of the biggest productivity gains we're likely to see in the next few years will be job-killers and/or trade-downs in quality (e.g. "customer self-service" websites to replace call centers, low-cost airlines replacing the traditional carriers as the latter go out of business).

If you're making the argument i think you are about offshoring, i agree with you that offshoring is a lazy, short-term approach to boost profitability at the expense of real gains in productivity - i.e. companies are charging ahead on the assumption (implicit, at least) that the much cheaper price of labor offshore is a market failure that they can cash in on to save a few bucks without trading down in quality or control. A lot of those companies are going to learn the hard way - as some already are learning - that offshoring carries a lot of hidden costs, and that to the extent there ever was a market failure to take advantage of, that window is very rapidly closing & in many instances is already closed. The wage gaps will close, the real tradeoffs in quality & responsiveness for call centers & other offshored services will be translated into prices, and more than a few comapnies are going to want to cancel or not renew those contracts a few years down the road.

The million-dollar question is, will anyone be left with capacity, either onshore or nearshore, to pick that business up when the day comes? Whoever does (and if they're smart, Tata & the other offshore firms will cover their bets & use some of their excess profits to buy up North American capacity in their core lines of business, while it's cheap) will have a great opportunity - and probably end up w/ market power to burn. End of the day, some of the firms that are taking the short-end savings now may someday pay through the nose for the same services when they bring them back home.

Still, for all the things we agree on, I don't share your pessimism about the productivity future, either for IT or for manufacturing in general. I think at most times in our history, we've been on the horns of this kind of dilemma: the last wave of productivity gains is always running out of steam, and nobody knows where the next one is going to come from, yet more often than not it materializes anyway. The thing that has always pulled us through before is that the openess of our economy, combined with the flexibility & capacity of our financial system, has made it possible for new ideas to sprout up.

The thing that worries me about our future isn't that we're gonna outsource & offshore ourselves into obsolesence, but rather that the galactically stupid macroeconomic policies of the current administration are going to saddle us all with so much foreign debt that - either through a current account crack-up, a significant currency devaluation, or some other calamity (see DeLong's post about Obstfelt & Rogoff's NBER paper) - we'll hose up the financial system to such an extent that the innovation "virtuous cycle" will grind to a halt.

Posted by: Tom at November 12, 2004 07:38 AM

Some of the salient points so far (IMHO) of this discussion (apart from some technical discussion on the definition of productivity):

- It's not the hardware, it's the software (and while it has gained in convenience and in multiplicity of features, it's not better or faster by orders of magnitude than it used to be a few years back); and since many people have not upgraded the productivity oriented software to the tune of their hardware, there is a degree of loss here.

- It's not the software, it's its usage (and, unless you are a PhD student in non-linear physics, you just have to look at your CPU usage% to understand that the only way to reach double digits on a sustained basis is to do some video processing or gaming, which for most people is entertainment and can be productivity loss if practiced at work); so all that high power and rich software is going for naught

- In software usage, the critical bottleneck is the end users (esp. for desktop machines, which after all, is the primary environment of about 300 million information workers worldwide); we can only type so fast, or build Excel formulas so complex

So in other terms, simply stating as Brad did that one can buy X times the car power and speed for the money that people could buy 50 years ago (I am paraphrasing) does not mean people travel that much faster. They actually only have a relatively small marginal improvement in "productivity of travel", but probably a larger improvement in "enjoyment and convenience".

Two missing elements in the conversation:
- Software and hardware are not one time cost items. The complexity of system integration, maintenance, training, and continuous adaptation of the systems defy simplistic computations like Brad's or Ken's
- One point ignored is the gain in productivity through improved efficiency in communication and collaboration. Networks, email, internet, IM, mobile phones (which are specialized computers) are tools that have revolutionized (and in the future, together with data collaboration and VoIP, with revolutionize again) the way we work, and create a substantial increase in productivity from reduce time and cost in collaboration. That increase is not predicated as much on power increases as on network externalities and availability of technology.

Posted by: ergos at November 12, 2004 09:16 AM

Some of the salient points so far (IMHO) of this discussion (apart from some technical discussion on the definition of productivity):

- It's not the hardware, it's the software (and while it has gained in convenience and in multiplicity of features, it's not better or faster by orders of magnitude than it used to be a few years back); and since many people have not upgraded the productivity oriented software to the tune of their hardware, there is a degree of loss here.

- It's not the software, it's its usage (and, unless you are a PhD student in non-linear physics, you just have to look at your CPU usage% to understand that the only way to reach double digits on a sustained basis is to do some video processing or gaming, which for most people is entertainment and can be productivity loss if practiced at work); so all that high power and rich software is going for naught

- In software usage, the critical bottleneck is the end users (esp. for desktop machines, which after all, is the primary environment of about 300 million information workers worldwide); we can only type so fast, or build Excel formulas so complex

So in other terms, simply stating as Brad did that one can buy X times the car power and speed for the money that people could buy 50 years ago (I am paraphrasing) does not mean people travel that much faster. They actually only have a relatively small marginal improvement in "productivity of travel", but probably a larger improvement in "enjoyment and convenience".

Two missing elements in the conversation:
- Software and hardware are not one time cost items. The complexity of system integration, maintenance, training, and continuous adaptation of the systems defy simplistic computations like Brad's or Ken's
- One point ignored is the gain in productivity through improved efficiency in communication and collaboration. Networks, email, internet, IM, mobile phones (which are specialized computers) are tools that have revolutionized (and in the future, together with data collaboration and VoIP, with revolutionize again) the way we work, and create a substantial increase in productivity from reduce time and cost in collaboration. That increase is not predicated as much on power increases as on network externalities and availability of technology.

Posted by: ergos at November 12, 2004 09:17 AM

Ergos, I agree on all points. I had tried to summarise some of your points with the simple claim that desktop computers are not capital equipment but office supplies. They get replaced every three years maximum. So the improvement needs to be amortised over 3 years, or 2, or 1.

And the software often gets replaced just as fast. The result is new features that must be learned, which is a new expense. The value of the new features is often dubious. And new releases create new bugs that users must learn to work around; at least they were used to the old bugs.

When the software doesn't get replaced then the result is often to do the same things subliminally faster.

Some subliminal improvements may be significant improvements in ways that are hard to measure. Graphic displays that don't flicker as bad. Displays with more resolution. Does that reduce eyestrain and maybe improve employee health? Hard to tell.

Improvements are likely to follow a logistic scale rather than exponential.

Better hardware *allows* new capability that wouldn't have been practical on older systems, but you only find out how valuable the new stuff is by using it, after it gets invented.

Efficient communication is a double-edged kitchen knife. It allows much more communication than was practical before. Parkinson's Law says the result will be a vast increase in unproductive communication. "The amount of traffic increases to fill the available bandwidth." Is the result a net improvement? I don't know. Maybe.

Apart from the increased spamming potential, we also have a more elaborate presentation process. Things that used to be given to a secretary to turn into a memo now often turn into PowerPoint presentations. Things that used to go to Graphics to be slideshows now turn into Flash. It's a better product, but are the ideas presented worth the flash? If top management can pick up much more information in the same time then it's a clear win, but if it's a zero-sum game for the people who want top management's ear, then productivity hasn't gone up much at all.

Posted by: J Thomas at November 12, 2004 12:34 PM

J. Thomas, we are in agreement, with the caveat that I am not as skeptical as you are of the benefits of improvements in communication and collaboration. I agree on the risks, but in my opinion the benefits far outweigh them.
Regards

Posted by: ergos at November 13, 2004 09:24 AM

Ergos, I lack data to claim that the increased communication reduces productivity. If the amount of communication that was necessary (or extremely useful) was more than the previous channels could manage, then the increased bandwidth *allows* a useful increase. And if the bandwidth was technicaly adequate, but useless communication often crowded out the good stuff, then increased bandwidth might allow important communication to get through when it wouldn't have before.

But there are no guarantees. Ideally you'd measure it.

Posted by: J Thomas at November 13, 2004 12:37 PM