[Previous] [Main Index] [Next]

Wednesday, August 13, 2003


This article is actually a composite of four sets of comments that the buggy prof left at another web-site, run by Professor Brad DeLong --- an economics professor at UC Berkeley, who brings to his impressive academic work an unusual historical perspective that helps illuminate contemporary economic trends, both within the US and globally. That perspective shows up in his daily commentaries left on his site, most of them stimulating even when you don't agree with him . . . and a source of extensive and vigorous comments by others. Three of the four sets of comments that I left on his site, yesterday and today, were prompted by DeLong's commentary entitled "Some Optimism in the Forecasts" for the US economy, and in particular by some speculation about job- and productivity trends afoot for three years now. The fourth set --- actually the earliest --- was left in response to a DeLong commentary called "Productivity Growth Trends" (August 9, 2003), which has a nifty chart of the growth of productivity since the early 1970s and how unusual it is for productivity to grow so rapidly in a recession and for two years after.

Revised and extended, it's these comments, which deal with productivity trends in the US economy --- and some comparisons with other countries --- that you'll find here.

First, Some Lead-In Remarks To Those Comments

In effect, what we're experiencing in the US economy --- as three earlier buggy prof articles noted for July 21st, 23rd, and 26th --- is a curious short-term conflict between the surging growth in labor productivity and the poor performance of the economy in employment matters since the end of the short-lived shallow recession of 2001 almost two years ago. Unemployment continued to grow for a good 21 or 22 months afterwards, something we've never experienced before; and as those earlier buggy prof articles have argued, the long-term benefits of rising labor productivity in our economy --- which bode well for ever higher living standards in the future --- have created a short-term problem for the job market. If productivity is rising, say, around 3.3 - 3.5% annually, then short-term economic growth has to rise at least into that range to get firms to start hiring back laid-off workers and create new jobs. With GDP growth averaging about 2.6% in the seven quarters since the end of September 2003, that hasn't created enough sales and profits for the owners and managers of firms to create those jobs. Instead, they have managed to keep profit margins from falling too much by squeezing out ever more productivity from their smaller work forces.

Will this jobless recovery continue? Not likely.

For one thing, if any one trend stands out in the US economy's long-term performance over the decades, it's a matchless ability to create new jobs . . . most of them good jobs, despite recurring flurries of worry that manufacturing has been declining steadily as a percentage of GDP (down from 40% in the early 1950s to less than 20% today) and manufacturing jobs declining even faster (down to less than 14% of the total work force), or worry that we'll all end up flipping hamburgers. These new jobs require ever greater levels of skill. They also pay, usually, higher than average wages. Even in manufacturing industries, high-school diplomas are required for new job-seekers. No other economy can match the US on these scores. For a good, readable study of this put out by the Clinton Council of Economic Advisers in 2000, see "20 Million Jobs, January 1993 - November 1999."

For another thing, as the buggy prof article for July 26th noted, our prospects for a far better short-term performance of GDP --- whether for the rest of this year or next year --- have noticeably improved of late. Profits are up; business investment is up; the service sector expanded last month at its fastest pace in six years; long-term interest rates --- which started rising for complex reasons in July --- have fallen again; and consumer confidence remains generally robust. And even with the summer doldrums, the stock markets are doing well too. As a result, GDP growth should be way above 3.0% for the remainder of the year, and we are likely to see unemployment --- which peaked in May at around 6.4% and was down in June --- continue to fall steadily in the quarters to come.

What to do about the huge inroad of Chinese imports --- putting tremendous pressure on certain key US industries (above all, textiles and wood furniture) --- remains a short-term problem, aggravated by the Chinese government's fixing the Yuan at an artificially low rate to the dollar, then spending $300 million of Yuan to buy dollars daily in order to keep its currency as a prime-pump of exports to the US. For the buggy prof's advice, see the article for August 2nd here on that topic. Remember, the Chinese import surge is a short-term problem: in the long-term, China's growing wealth is likely to be a big boon for the high-tech, knowledge-based US economy --- and simultaneously, as a second article on China's economy will show, it's unlikely to be anywhere near the levels of US technological prowess in two or three decades . . . whether in the civilian or military realms. It will be lucky if it's much more advanced than Taiwan's economy today


First Set of Buggy Comments

1] Mr. Bakho asked a probing question, which a couple of other contributors here took up . . . without answering it. Let me take a stab, adding that as a political scientist with a shared Ph.D. in economics, I would rather have a specialist in productivity --- not just an economist, but someone who specializes in this difficult area and knows the measurement problems --- be the one to answer Bathko.

Here's Bakho's comment and then his query:

"However, I was reading an interesting but disturbing article about computer programmers in the US training their replacements in India and elsewhere before being let go. Companies are replacing $60/h US workers with $6/h foreigners in high tech jobs. Ouch. Business Week is highlighting this.

How much is this contributing to productivity gains?"

Posted by: bakho on August 11, 2003 09:09 AM


The Buggy Response:

[1] As far as I can tell, the question as it stands confuses two things: measurement of overall US business-sector productivity --- something the Bureau of Labor does quarterly --- with its efforts, far more controversial, to measure the productivity of specific industries. For the more general measure of business-sector productivity --- which excludes general government, non-profit organizations, , and one or two other things --- the BL stats (BLS) take quarterly and annual GDP, deflate it for inflation, subtract the excluded factors just mentioned, and get the Output numerator.

For the denominator, the BLS runs surveys to calculate worker hours, and that together with the numerator gives the productivity of worker-hours in the business sector for the quarter or year. More specifically, of course, the resulting figure reflects the growth of the output side (real GDP contributed to by businesses) and the growth of labor hours . . . if any.

There has been something of a flap about whether the BLS surveys reflect accurately the hours worked by Americans in the denominator, but an intelligent study carried out at Princeton by Alan Krueger and a group in 2000 or so found that there wasn't much discrepancy from what they found and the BLS reports. See Krueger


[2] Against this background, we should be able to answer Bakho's query.

Remember first, real GDP can be measured in two different ways in the National Income and Product accounts: on one side, by the production of consumption goods C, business investment I, government goods G, and X-M (exports - imports). For business sector productivity, exclude government goods. On the income side, you have personal income (wages, interest, profits), which can be divided in the accounts into C, S (savings), and T (taxes). Again, to answer Bakho, subtract government employee incomes and non-profit organizations and their incomes in order to pin down the business sector. (The Bureau of Labor also divides the business sector further into non-agricultural productivity trends, manufacturing, durables and so on.)


[3] What follows?

If a computer firm substitutes foreign workers for high-paid American workers, productivity in the firm remains the same on one assumption: the Indian programmers are just as efficient as the American programmers. The same number of software programs and their quality that the firm sells to other business firms and to end-users haven't changed at all. The firm, though --- faced with lower wage costs --- reaps higher profits, assuming further that the price of its software programs to its customers remains the same. Alternatively, the firm could reduce the price, but it then sells more software programs over the year, and its profits still rise.

3a) In that case, nothing has changed as far as the Bureau of Labor's productivity analysis goes, either for the firm or for general US business sector productivity. The firm's ouput --- software programs --- doesn't change, only its labor costs. And though lower labor costs in the aggregate for the US business sector are down, they are offset by an equal amount of reported profits.

If, alternatively, the Indian programmers are more efficient --- they write software programs faster for the firm to sell --- the productivity will go up for the firm, and GDP (the numerator in the productivity equation) goes up too because the firm now reports higher profits. Oppositely, though, the inverse could happen. If the Indian programmers are less efficient --- fewer programs are written and hence sold --- the firm's productivity is reduced, and for the US in the aggregate business sector productivity would fall off because fewer profits might be reported. The latter, to happen, would require profits to fall faster than the reduction in the wage-costs that occurs when the high-wage US programmers are replaced by the low-wage Indian programmers.


4. There are other scenarios too, all depending on what happens to the displaced US programmers themselves. Specifically:

4a. On one assumption, assume that the laid-off US programmers are unemployed for a whole year. Assume further these laid-off programmers get unemployment insurance for half the year, and live the remaining half on their savings and loans or help from others. Well, nothing would change in overall productivity even if they got unemployment benefits for the whole year. Why? Well, GDP might go up for the year --- with unemployment compensation now paid to the laid-off workers (financed, let's say, by growing national debt for the year, not taxes) --- but not in the business sector. The laid-off workers aren't in the business sector at all. In that case, business sector productivity isn't affected at all.

4b. On another assumption, assume that the laid-off programmers get jobs flipping hamburgers at McDonald's and Wendy's. How would aggregate business sector productivity then be affected? The answer: well, for the numerator, there would now be an increase in reported GDP as calculated by national income (profits, wages, and interest). But, on the denominator side of the productivity equation, the hours worked would rise too; and what with the lower level of overall productivity in such hamburger joints compared to that of software firms, the hours worked in the aggregate for the US economy would likely rise even faster. The result? Lower aggregate national productivity in the business sector.

4c. On yet a third assumption, conceivably business sector productivity could rise even if all the laid-off programmers are hired by McDonald's and Wendy's as burger-flippers. How so? Much brighter than the average employee at those firms, the former programmers might see various ways to improve the number of hamburgers --- and possibly the quality too --- and McDonald's and Wendy's would report higher profits. Alternatively, the number of hamburgers and quality could stay the same, but the mentally agile programmers think of ways to cut costs. Either way, the corporate firms reap higher profits.

Alternatively, the heads of these chains (locally and nationally) might share the profits by awarding higher wages to the astute, efficiency-minded. And in the aggregate --- given this third assumption --- the US economy's overall productivity would rise too: there'd be more profits and more income reported for the numerator (GDP in the business sector).

Michael Gordon http://www.thebuggyprofessor.org

P.S. Just promised to take my wife and dog for a walk, so I don't have time to review the reasoning or prose here.

Posted by: michael gordon on August 11, 2003 06:40 PM


PART TWO: RESPONSE TO SNS, As a Follow-Up to This

SNS, many thanks for your stimulating reply, which is set out here again:

"Buggy -- aren't relative wage levels derived from the productivity? So if a computer programmer is well paid it is because of some unmet need, and then if it turns out there are tens of millions of workers overseas qualified to do this same job there is no longer an unmet need and the high value added job is now a low value added one. Presumably the out of work programmer believes work is available at a wage above what the overseas programmer makes, or else would take a pay cut to that level. So if the 100K job becomes a 20K job overseas and new work as a housecleaner is found at 30K, shouldn't that be an easy case to be made that productivity has risen? Maybe we are just used to programmer being higher value added and housecleaner being lower."

[1]You seem to be confusing productivity --- either of a firm or the economy as a whole --- with labor costs . . . and in the case of Indian programmers compared to American ones, relative wage costs.

[2] Aggregate labor productivity is defined as Output / hours worked by the Bureau of Labor, where output is GDP minus the government sector and non-profit organizations to get business sector output, deflated for any price increases. Since the mid-1980s, I believe, the Bureau also tries to weigh the contributions of specific industries to the resulting Business-Sector output, so that the more important an industry is in accounting for GDP, the more heavily it's output is weighted against the hours worked by the employees in it.

[Keep in mind that GDP output in this sense doesn't just reflect labor contributions: it also reflects anything that causes GDP, including the growth of capital inputs, the growth of capital productivity (hard to measure, but not impossible even across countries), ever better management and marketing, and technological progress of a general sense. For that matter, the productivity in the use of raw materials and fuels.

An example: international comparisons of manufacturing undertaken by the McKinsey Global Institute in the late 1990s showed that British labor productivity in mfg. was about 10-15% below that of German and French mfg, but by contrast capital productivity --- the growing efficiency of output / unit of capital investment measured in dollar terms --- was about 10% higher, and the result was that British manufacturing was just about as efficient.

Despite that, German wage costs in manufacturing were extraordinarily high --- way above any productivity difference compared to the rest of the EU or the US --- because of skyrocketing social security taxes. In 1995, if I recall --- given the DM/$ ratio --- German mfg. labor's basic wage was about $20 hour, slightly higher than the US but not by much. But after taking into account social security taxes and medical payments etc, US wages in mfg. rose only by about 10% to $21 or so, whereas German wages rose to around $37 an hour! (The DM went down about 35- 40% against the dollar in the late 1990s, then as the euro another 25% between 1999 and the start of 2002, only to rise about 25% since then again.)

All this still on the numerator side, and GDP can be calculated either on the product side or the income side (wages, profits, interests), again for the business sector. The two sides of national income should, of course, be equal, and they usually are, though there was a strange wedge driven between them in the early and mid-1990s.


[3] The Denominator.

Surveys are taken to check on the hours worked, and there are efforts by the Bureau, it seems, to subtract paid vacations and paid leave from the output side where no hours are worked by the employees. Also, efforts to consider benefits not included in gross income per se. (There are also further refinements about supervisory and non-supervisory employees, with different survey and reporting techniques).


[4]Wage costs.

Whether a firm pays on an average higher wages or not --- ditto the firms in a specific industry --- isn't part of this definition of either in the numerator or denominator in the equation (where wages are wages, in a gross sense). True, higher productive industries --- which may be higher value ones (not necessarily) --- will pay, on an average higher wages; and within a specific firm or industry as a whole, certain workers will also get higher wages depending on their roles (managers vs. non-supervisory) or education and talents and the scarcity of not of such labor. But the causal influence runs from higher productivity to higher wages, not vice versa.


[5] Generally, it's true in the abstract, firms won't make a profit in competitive markets if they pay wages higher than productivity increases justify in the long-run . . . something that would be reflected in rising wage costs that would eat up profit margins and eventually all profits.

But note:

Some firms or industries may also pay higher than average wages --- or wages not justified by their levels of productivity --- and still stay in business for long periods of time, making a profit in the process. How? That will normally be because they're protected from foreign competition by tariffs or quotas or subsidies, and possibly from domestic competition (new start-ups as potential competitors) by various regulations. Either way, the costs are passed on to other businesses (think of high-priced steel in the US economy when new protection was added last year) or to end-consumers.

This happens all the time, and especially in the EU, where on the Continent lots of small retail and department stores are protected by laws that regulate how many discount stores can be started in a region or municipality, and where they locate and so on.


[6] Oppositely, a firm or an industry can have higher levels of productivity and still not be competitive internationally.

Take general steel-making in the US (as opposed to specialized steel-makers). At one time, in the 1970s and 1980s, their output (a ton of steel, say) / per man-hour lagged noticeably behind EU, Japanese, and Korean steelmakers. Since then, having shed 50% or more of the labor force and modernized, US general steelmakers have just about the highest productivity in the world, somewhere around 4 man-hours per ton compared to the 10 hours it took in the 1970s and 1980s.

And generally, as opposed to the 1970s and early 1980s, wages (including benefits) don't exceed the level of productivity any more.

All the same, in part because of subsidized steel in Europe and Japan --- less so in the rest of Asia or Brazil --- but also because of very high pension costs to retired workers, US steelmakers face declining market-share in the US and protest they can't meet the competition head-on.


[7] In all these cases, then, we can say this: wages in a firm or industry will ordinarily reflect levels of labor productivity (if this can be measured directly, as in steel output per man hour), but only if the markets are competitive and there aren't inherited fixed costs, etc.

But there are several other considerations that might work contrary to competition or cause inescapable high costs (like high pension costs that will take decades to run down).


[8] One more complication here --- and a significant one.

As a general thing, the wages of workers in any firm or industry compared across countries will also reflect not just whether the firm or industry is a high or low-productive one, but also average wages across the country itself.

That's why Indian programmers, who might be as talented as American programmers --- some of whom are Indian immigrants anyway --- can get far lower wages. What's at stake here is average levels of productivity in the US and India as calculated by GDP/hours worked in the business sector, and not just the contributions of Indian programmers to final output of the US multinational.

The multinational firm might, in the process, reap higher profits, but productivity within the firm across its US and foreign branches could stay the same. In the case just described, to repeat, assume that the Indian programmer in Bombay is just as productive as his Indian immigrant counterpart here.

Productivity within the firm hasn't changed, only wage costs and profit margins.


[9] Finally, a last observation here. It deals with alternative definitions of labor productivity.

As we noted, labor productivity is ordinarily calculated in aggregate terms by using hours worked in the denominator. Nonetheless, alternative definitions may be more useful for cross-country comparisons: for instance, GDP/worker or even GDP/population. (The latter, of course, gives you per capita income).

The reason for preferring at times alternative definitions for comparative purposes? There may be big differences in the hours worked across countries, as well as the participation ratio of the work force: the percentage of adults, say 18-65, working. There may also be jobs being created, but part-time or temporary, rather than full-time. And much of these differences may not reflect preferences for leisure vs. work.

To clarify briefly, assume what some specialists say --- specifically, assume that in the EU, there is more preference among the population for leisure as opposed to work compared with the US. That preference shows up in hours worked per year on an average: the US worker recording around 1900 hours, and say at the opposite end the German worker around 1550. A huge difference. As Robert Gordon showed in a study earlier this year, taking into account the preference for leisure here reduces the per capita income gap --- a proxy for labor productivity (with of course all other contributions like capital efficiency and multifactor productivity reflected in the GDP output that labor hours have nothing to do with) --- by about half. In particular, the gap with the US rises from roughly 70% of the US average in per capita income to about 85% for the EU.

But note: what if lots of German and EU workers want to work more hours, but can't? Or, more plausibly, what if they're in part-time jobs only --- most of the EU jobs created since the early 1990s seem to be this --- and can't get full-time jobs because of rigid labor markets, high minimum wages, union rules, problems with lay-offs etc. Or, more to the point still, since about a third of EU youth (16-late 20's in age)aren't employed but almost certainly would like to be --- or are in endless undergrad or grad programs with no payoff (with the average age of a German student's getting his first degree now 29 years!) --- then using GDP/hours worked might be very misleading as an accurate gauge of productivity.

From that viewpoint, per capita income (GDP/population) or GDP/worker might be better estimates across countries.


For the Robert Gordon study and further reflections on this, see the buggy prof article in an exchange with a lawyer in this country who worked a long stint in the EU that appeared this last February:


(I add that owing to a hacker attack, the archives are still scrambled, with my web manager promising to fix them. This, though, is the accurate IP for the article, found thanks to Google.)



Just wanted to add a postscript remark or two to the lengthy set of comments I made here earlier today. These remarks are prompted, I add, by my chasing down the original Robert Gordon paper comparing EU and US differences in productivity levels and living standards (per capita income, in purchasing power parity terms), and finding some curious data about both --- especially the EU living standards compared to the US's.

1) Robert Gordon's paper can be found at http://faculty-web.at.northwestern.edu/economics/gordon/355.pdf In it, he uses data from two sources --- the IMF and OECD --- that lead to two findings as a basis of his later explanatory efforts: 1. in 2000, EU living standard (per capita income) is 77% of the US's, whereas 2] its productivity ratio is 93%. (see p. 10 in his article, based on data explained and used in charts),

[2] On that basis --- with a very useful long-term historical perspective, I add (it's a good paper except for this starting data, as I'll show in a moment) --- he seeks in his paper at two points to find explanatory influences that would narrow the gap between living standards (the EU ratio is 77% of the US's) and labor productivity levels (93%). His conclusion? That the actual figure for both should be about 85% of the US's levels.

[3] In short, the EU living standard is higher by about 15%, and the productivity level lower by about the same amount (slightly more).


[4] The problem is, the EU Commission --- which puts out careful studies each year of EU economic prospects and especially the EU's commitment, languishing, to make the EU the dynamo of the world economy's information and communications technology --- finds far different living standards and productivity levels in the EU compared to the US.

. Go to chapter one of the EU Competitiveness Report 2002, p. 20, and look at the chart and table. http://europa.eu.int/comm/enterprise/enterprise_policy/competitiveness/doc/competitiveness_report_2002/cr_2002.pdf

The EU average per capita income --- in PPP terms --- is 69% of the US's in 2001 (the ratio will have fallen to about 65% given differential growth in 2002 and likely 2003 for the entire year). That is far below the 77% that Robert Gordon starts with.

Now jump to table 1.4 on p. 22 of the report, and you get the EU figures for labor productivity in the EU and Japan as a ratio of the US's in 2001. As you can see there, the EU ratio is 78% (and Japan's 67%). Again, this is far lower than the 93% that Robert Gordon starts with for the year 2000.


[5] I cannot account for the big discrepancy in Robert Gordon's use of figures for the ratio of either per capita income in the EU compared to the US or its ratio of labor productivity. An excellent scholar, there's no reason to think he was being careless with the figures. (And to repeat, he not only uses good sources like the stats put out by the IMF and the OECD, he also has written a very stimulating article on the sources of US productivity advantage -- originally in the 19th century and throughout the 20th despite in his view the EU's narrowing this gap. In fact, if the EU Commission's statistics are right --- and they appear to be (who knows the EU per capita income and productivity levels better than the EU Commission and its economists?) --- then the gap in living standards and productivity between the EU and the US has not narrowed at all since the mid-1960s.

On the other hand, while estimates of labor productivity across countries can vary depending on whether you use Output/labor hours vs. Output/ worker --- with the former narrowing the productivity ratio for the EU whatever the actual starting ratio might be for the latter --- it is confounding to figure out why there should be such discrepancy in GDP (output)/ population figures and the resulting ratio of per capita income.

Remember, GDP has to be real (essentially, the deflators on both sides of the Atlantic aren't that different), and then translated into Purchasing Power Parity terms . . . itself a pretty reliable way to bring prices in line with one another for comparative purposes. At most, depending perhaps on different index-years for the dollar or Euro in different PPP surveys, you might get a different of 3-5% for countries or regions like the US and the EU. At least, I've never seen any that diverge beyond that range. (When it comes to comparing, on the other hand, the US and China in per capita terms, PPP comparisons are much more dubious . . . what with the large percentage of non-market transactions, including sheer barter, that comprise the living standards of developing countries. Large numbers of people, especially in the countryside and small towns, build their own shelters, raise their own food, engage in barter, etc.)


6) I would add one more remark.

The underground economy --- outright criminal activities as well as non-criminal economic activities that aren't reported in order to escape taxation --- is much higher in the EU than in the US. Even in Scandinavia, traditionally know for its civic discipline, decades of high taxation and a sense of grievance --- why the hell is that organized group getting more from the government than my group? --- have led to large and inefficient use of barter and fraud to avoid taxing.

Essentially, the underground economy is about 15-20% of GDP in the EU and about 7% in the US, at any rate it is according to the best study of this that uses a variety of ingenious cross-checking techniques: see the link to the Austrian team headed by Schneider and Frey, and to their findings as cited in a gordon-newspost article, the now defunct listserver subscription daily commentary that I put out for a few years:


That listserver, which is indexed by google (enter gordon-newspost and try an economic or political topic: there were a couple of thousand articles or so it catalogued), has been superseded by http://wwws.thebuggyprofessor.org

Note that though including the underground economy would raise GDP in the EU compared to the US --- and hence per capita income --- it would, oppositely, lower productivity in any meaningful sense. After all, the techniques used for tax evasion such as a dentist fixing the teeth of a Mercedes car mechanic free in return for his fixing the dentist's car --- or slipping capital abroad --- are notoriously inefficient. And that leads to big problems, mentioned earlier by me, of making sense of productivity calculations . . . especially on a cross-country basis.

--- Michael Gordon