I’ve been working on a piece about productivity, targeted as an Economics Briefing for the Fin Review. This used to be the Schools Brief, but they’re trying to broaden the appeal a bit. The target audience, I think falls into two groups

* Students with a year or two of economics under their belts

* The subset of the Fin audience with an interest in economic policy issues, as opposed to business news

Any suggestions or criticisms will be much appreciated.

By the way, I hope no-one minds being used as unpaid editorial advisers. The payoff is that you get to read stuff well before the print audience.

Productivity may not be the only thing contributing to a difference between high and low living standards, but it is easily the most important. By comparison, factors such as mineral resources and other natural advantages are almost insignificant. Australia’s mineral resources, for example, contribute about 4 per cent of our national income. This is less than the amount added through productivity growth every three or four years.

Ultimately, all productivity growth comes from new knowledge about science, technology and economic organisation. The standard of living enjoyed by the average Australian today could not have been achieved by any society 100 years ago, no matter how hard people worked, how much they saved and invested or what government policies were adopted.

However, the links between technical progress and productivity growth are many and varied. Some are clear and direct, but others are complex and poorly-understood. Partly because of this, there are many different concepts of productivity, and many complex disputes about its measurement.

The simplest form of productivity improvement is called ‘embodied technological change’. This is what happens when Intel or IBM comes out with a new computer chip that is faster, smaller and cheaper than the one it replaces. From the point of view of a country like Australia, which mostly imports its computers, this kind of embodied technical change is like a free gift. When we replace our old computers with new faster ones, we can do more of whatever we do with no additional effort on our part.

Improvements in physical capital are not, however, the most important way in which improvements in knowledge contribute to productivity. What matters even more is the skills and knowledge of workers, or ‘human capital’. In part, this depends on the progress of knowledge in the world as a whole – doctors in 1900 could not possibly know about antibiotics, for example. But more significantly, human capital depends on the availability and effectiveness of education (of course, education is costly, and this cost needs to be taken into account when we think about improving productivity through education).

Technical innovations and improvements in human capital create the potential for profitable new investments, thereby giving rise to a third potential source of productivity growth, known as capital deepening. Capital deepening is usually said to arise when the value of the stock of capital per worker increases. In this sense, there has been almost continuous capital deepening ever since the Industrial Revolution. However, the evidence suggests that the ratio of physical to human capital has been more or less constant (if you assume that workers are paid roughly in line with their contribution to output this observation follows from the fact that the wage share of national income has been stable over time).

Finally, there may be improvements in economic organisation, including improvements in economic policy. An important example of an improvement in economic organisation is the invention, in the mid-19th century, of the limited liability corporation. In Australia, it is commonly claimed that changes in policy during the 1980s and 1990s, collectively referred to as microeconomic reform, were a source of productivity growth.

Assessing the contribution of different factors in promoting productivity growth is a complex exercise, fraught with error. Consider the simplest possible concept of productivity, which is output per unit of labour. The output of an economy is typically measured by Gross Domestic Product. This aggregate measure has well-known problems, but there’s not enough space to discuss them all here. Even more problems arise when we try to measure the value added in some particular sector of the economy.

But the real issue is with the measurement of labour input. Until relative recently, it was usual to look at output per worker. This made sense when most workers were full-time employees, with standard hours of 35 to 40 hours a week common to most of the workforce. But when some workers have one or more part-time jobs, and others are working 60 hours a week, this no longer makes sense, and it’s more usual to look at output per hour of work.

But even output per hour is not really the right measure. Over the last fifteen years or so, there’s been a lot of pressure to speed up the pace and intensity of work. This squeezes more labour input into a given time, but there’s no real change in productivity. Since there are no good measures of work intensity, there is no easy way to tell how much this change has affected measured productivity, but it’s been argued that the bias could be 5 or even 10 percentage points, which would account for much of the observed increase in productivity between, say, 1989 and 1999 . On the other hand, many economists dismiss claims about increases in work intensity during the 1990s as being based on little more than anecdotal evidence.

While labour productivity is interesting, the discussion above suggests that it’s necessary to take account of capital as well. The standard approach, known as growth accounting involves estimating how much of any observed increase in output can be explained by increases in inputs of labour and capital. The unexplained residual is given various names, of which the most common is ‘multi-factor productivity’.

The size of the residual attributed to multi-factor productivity varies depending on how much the capital input is adjusted to take account of embodied technological progress and how much the labour input is adjusted to take account of human capital. The more quality adjustment that goes on, the less is left for the residual to explain. But, almost invariably, there is a positive residual.

In Australia, estimates of multifactor productivity have been produced by the Australian Bureau of Statistics since the mid-1990s. On statistical grounds, the ABS divides the data for the period since 1964 into subperiods called ‘productivity cycles’. In general, productivity cycles are about five years in length and a typical macroeconomic cycle of ten years contains two productivity cycles, one corresponding to the upswing and the other to the downturn.

This series has excited a lot of interest because, after declining, with occasional upticks, during the 1970s and 1980s, the estimated rate of multi-factor productivity growth rose substantially in the productivity cycle from 1994 to 1999. Initial estimates suggested a growth rate of 2.4 per cent, an all-time record, leading a number of commentators referring to an economic ‘miracle’, or the emergence of a ‘New Economy’. Subsequent revisions yielded an estimated growth rate of 1.8 per cent, comparable to the ‘Golden Age’ of the 1960s, but not unprecedented.

Sceptics argued that the increase in measured productivity was, at best, a once-off blip in the data, and at worst, a statistical illusion. This implied that improvements in measured productivity growth would prove transitory.

The case that the productivity ‘miracle’ was illusory had three main elements. The first, and most important, was that of increased work intensity. It was argued that the apparent improvement in productivity simply reflected unmeasured increases in labour input arising from a faster pace of weak, the loss of informal tea breaks and so on.

The second element of the argument related to the ABS concept of productivity cycles. It was argued that productivity moves in line with the economic cycle, being weak in recessions and strong in recoveries. Thus the upsurge in productivity growth in the mid-1990s was partly due to things like improved utilisation of capital due to economic recovery.

Finally, there were technical disputes about the composition of the index. The MFP index is constructed for the market sector, but excludes a range of business services, which expanded rapidly in the mid-90s as a result of contracting out. On the other hand, the index includes agriculture, and was boosted somewhat by generally favourable seasons in the mid-1990s.

Sceptics claimed vindication when the measured rate of productivity growth dropped sharply after 1999. The incomplete productivity cycle beginning in 1999 has shown MFP growth of less than 1 per cent per year, below the long-term average.

By contrast, supporters of the view that microeconomic reform has boosted productivity have argued that productivity growth is merely ‘taking a breather’ as a result of random shocks such as the drought conditions that have been widespread over the past few years. They point to the long economic expansion since 1992 as evidence that the economy has become more resilient and adaptable. However, some have argued that reform has slowed down under the Howard government and that this is already being reflected in the productivity statistics.

Economic disputes are rarely resolved quickly, and the dispute over productivity growth seems likely to be no exception. A slide into recession, with an accompanying decline in productivity might settle the argument in favour of the sceptics. Conversely, a renewed acceleration in growth would strengthen the case for an Australian economic ‘miracle’. It is quite possible, however, that the actual outcome will be somewhere in between, a slowdown in growth but not an outright recession, leaving the issue unresolved.

19 thoughts on “Productivity

  1. I don’t get this work intensity argument. If I work harder, but not more hours, and I produce more, why isn’t that an improvement in my productivity?

  2. The simple economic answer is that the relevant cost is the disutility of effort. This will generally depend on both hours and intensity.

    But we can look at it another way. I assume we’re agreed that, if you increase your weekly hours from 50 to 55 and your output rises by 10 per cent, there’s no change in productivity.

    One way of doing that would be to cut official lunch and tea breaks by an hour a day. No change in actual or measured productivity.

    Now suppose that there’s no change in official hours, but everything is tightened up – you have to be at your workstation and ready when work starts, private phone calls are banned, toilet breaks are timed etc. In this case, measured hours probably won’t change much, and measured productivity will rise, but the increase is illusory. Output per hour of actual work hasn’t changed.

    Now go to the micro-level and think of a job as consisting of a series of tasks with brief breaks between them. On an assembly line, for example, each time the line moves is one task. An increase in work intensity arises when you shorten the breaks, for example, by speeding up the line. I hope its clear that, in conceptual terms, this is no different from the previous examples. The increase in productivity is illusory.

    Finally, go to the case when you just work harder and faster. It ought to be apparent that this is just a limiting case of the previous example.

  3. JQ,

    Interesting and well-written. Are you going to provide the productivity data as well, e.g. a chart? On that subject, where is the data on productivity for Australia?

  4. Pretty good, though I guess the role of the column precludes a “where do we go from here” paragraph, the lack of which makes the column feel a bit incomplete to me. Also, I would add an example of one of the microeconomic reforms Australia undertook, and the two sets of parenthesis near the top may be unnecessary.

  5. I assume we’re agreed that, if you increase your weekly hours from 50 to 55 and your output rises by 10 per cent, there’s no change in productivity.

    Well, yes, but we also know if your hours are increased it is probably due to an increase in demand and that the products being made have a higher realized incremental profit margin.

    The problem I have with productivity discussions is the question of how owns productivity increases.

    Lets take the U.S. discussion on Social Security and worker productivity – the argument being that there will be less workers per retiree in the future and therefore the system is unworkable.

    If we posit the extreme case of (almost) maximum productivity – the entire GDP is created by one worker, the 150 MM others are retired – the question isn’t – can the worker sustain the tax increases necessary to pay the retirees, but instead – how will the productivity increases be distributed througout society in some kind of equitable fashion.

    Productivity measures will always be problematic due to the political necessity of distribution and that output per hour is not coincident with profit extraction.

  6. I continue to be struck by the distinction between labor saving technological improvements and capital saving technological improvements, which I read in Samuelson forty years ago. Labor saving improvements increase output per manhour, other things being equal, which by definition is an increase in productivity (output per manhour). But, at the same time, it decreases the demand for labor, which, again, pari passu, means lower wages and a reduced average marginal productivity of labor. Since there has obviously been a great deal of labor saving technology over the years — most obviously in agriculture — why haven’t wages fallen as a result? The answer, at least in part, would seem to be because labor, through the political process, secured a lower standard workweek, enforced through overtime legislation, and the like. So the question is: could further reductions in the standard workweek likewise increase hourly wages? If so, by how much? (Like sprinters, we can work harder for faster periods of time, so there are two aspects here: reduced supply of labor, plus increased pay of work: can these be disentangled?) Also, trade with low-wage countries is tantamount to an increase in the relative supply of labor. Is this theoretically equivalent to technological unemployment as described above? If so, is it amenable to the same corrective measures? I.e., to what extent can we offset the impact of free trade of wage rates in the West with shorter work weeks, and how does it compare with measures to redistribute income between labor and capital (via a progressive consumption tax and wage subsidies, for example). I know this is an old-fashioned way of thinking, but is it truly irrelvant? Will the new global economy lead to heaven or hell for low-skill workers in the West over the next 50 years? Under lassie faire? Under artificial restriction of the supply of labor and the distribution of income? Is the latter compatible with a reasonable amount of market efficiency? I would appreciate your comments.

  7. John

    Two thoughts:

    (1) as you imply, the problem with the ABS’s productivity cycles is an empirical issue rather than a conceptual one ie the ABS’s dating of peaks is dubious. Your point might be clearer if you refer to specific points in history eg the unemployment rate in 1993-94 was much higher than at the next productivity peak, and hence the utilisation of L and K can be expected to have been much lower in 93-94 etc;

    (2) MFP calculations implicitly assume Cobb-Douglas technology. This can be a problem when factor prices move ie if the technology is eg CES, measured MFP will be high when real wages are high and low when real wages are low even if there is no change in the underlying rate of tech change. In Australia, this effect is material with productivity growth (as measured by MFP) understated in the 80s relative to other periods.

    I found when I did the exercise a few years ago that if you allow for both (1) and (2) in a labour demand equation (ie explicitly incorporate capacity utilisation and use CES technology), there is no change in the coefficient on tech change (ie a dummy for TIME) from the early 1990s. That is, all the measured increase in MFP is accounted for by these two factors.

  8. Ric,

    Thanks for these comments

    I’ll add a sentence to illustrate the first point.

    On the second, I assume you have an elasticity of substitution

  9. John

    Yep. The elasticity of substitution in most equations I have seen tends to be around 2/3.

  10. The simple economic answer is that the relevant cost [of work intensity] is the disutility of effort.

    And your evidence that workers incurred additional disutility from each working hour during the 1990s is …?

  11. What sort of evidence would you like, DD? I can offer

    1. An argument from first principles. I assume you either don’t like my demonstration that higher work intensity is equivalent to longer hours, or believe that workers gain utility from extra hours or work. If you indicate which of these you’re claiming, I’m happy to spell my argument out in more detail

    2. Survey evidence from ACIRRT and elsewhere

    3. Tons of anecdotal evidence – there have been hundreds of newspaper articles on the theme of how much people dislike the increased pace of work

    4. Evidence from the claims of microeconomic reformers themselves, for example about the “cold shower” effect of competition.

    5. International evidence of similar processes overseas

  12. Ric, my last comment got cut off, because I foolishly used a less-than sign, which got interpreted as HTML code. I repeat the relevant para

    On the second, I assume you have an elasticity of substitution less than 1, and I can see the role of capacity utilisation, but I still haven’t got the argument completely clear. Is there a paper you can send me? I’d like to add it as an additional point in this piece and, if it’s accessible, to cite it in future work.

  13. John

    I’ve tried to email you something. Let me know if it doesn’t turn up.

  14. My default position as a layman in economics is with Dave Ricardo. If I was in charge of an operation with measurable outputs my concern may well be to maximize the output of workers per elapsed hour that I was paying them. The more time-on-task within the hour the better, provided that it wasn’t producing adverse effects, such as mistakes and worker burnout.

    I can see that you are rigorously applying logic and trying to keep your concepts clean, but I’m not entirely sure how it all fits together and why. I would have thought output per hour in relation to what you are paying for the labour would be important.

    OTOH when I was a boss person in the real world (if the public service be deemed real) we had production workers (pick and packers), drivers, higly technical people who could maintain and fix things, plus provide expert advice, teachers, creators, critics, planners and much else besides.

    Literally, the strength of the bladder of a technician controlling a studio can be important. The driver who can zap around the suburbs delivering stuff is much more effective if he/she can exchange a bit of cheer in the few seconds a delivery takes.

    For some creative people, if they had one really good idea each year it would be heaps. A quick perspective when a complaint from the minister’s office hits the deck might be worth heaps also.

    And when your business is education, how do you measure output or outcomes? It may show up in terms of a correlation between levels of education and dollars in GDP, but what about personal fulfilment and contributions to social wellbeing?

    Just wondering.

  15. No, your theoretical approach is OK, though perhaps we ought not take it for granted that if people are working harder per hour than they used to that they are liking it less (that’s the sort of thing we might expect to be true but we should still try and confirm).

    I have a Gradgrindian liking for facts. I’d like evidence that workers are in fact working harder in each hour than they were in the eighties – if ACIRRT, whose work I generally respect, has solid data that shows workers are working harder in each hour than they were in the eighties then point me to it (I haven’t seen it). But ‘anecdotal evidence’ is not evidence – the plural of anecdote is not data, especially where people are using it to buttress an idea they are already disposed to think true.

  16. The above discussion ignores or gives little weight to the vital importance of fixed-capital investment to productivity. The trilogy – triumvirate – trio of fixed-capital investment, productivity, production are linked in practice, in history, in undeniable fact, by man, by economists and even – dare it be said? – by the Almighty Himself. If you whack fixed-capital investment, you take the economy off the highway and into the bush tracks of stagnation. See my comment elsewhere –
    Like the US, Australia’s problems go back more than thirty years and fundamentally derive from the attempt to “fight inflation” by hiking interest rates and “damping down” the economy. All that did was produce, first, stagflation, and then shift domestic inflation to the balance of trade/payments. In the process, we – both the US and Australia – gutted our industry, crippled our rate of fixed-capital investment, abandoned full employment, burdened ourselves with massive debt, etc etc. A major adjustment – a mild term for collapse – is inevitable in the US and, sadly, Australia too, probably quite soon.

  17. DD, here’s a fairly recent report. The money quote is

    — work intensification. All occupations and industries report that workloads have increased and work has become more intense. The proportion of employees working 50 or more hours per weeks has risen to 21%. Half the people who work overtime don’t get paid for it and about the same number wish they didn’t have to do it. Understaffing and the resulting increase in workloads and responsibilities have become fixed features of the contemporary workplace.

  18. I think this covers most of the issues very lucidly, John. But I suspect that an analytically minded lay reader would benefit from a more sequential of issues of (1) definition, (2) measurement, and (3) explanation. For example, the early paragraphs talk a lot about the growth of productivity and its causes, without venturing a definition of the variable whose growth is in question. The difference between labour and multi-factor productivity is introduced very late, when it would have been helpful earlier. You list capital deepening as one of several causes of higher labour productivity, but then tell readers that productivity growth measures the increase in output that isn’t caused by capital deepening. This is quite right, but possibly confusing for our target audience.

    Would it not be better to start something like this:

    Productivity measures output per unit of some input. If we want to know why some countries have higher GDP per capita than others, we might focus on output per worker. Until a few decades ago economists tried to explain the differences in terms of capital accumulation: country X had higher output per worker because it had more machinery per worker. Attempts to explain this in turn focused on how saving and investment pattern in country X brought about a higher degree of capital accumulation. But more recently conomists have found that differences in capital per worker explain only a small part of differences in output per worker. Most of the differences seem to be due to less tangible causes. Some of these intangible causes might affect labour specifically and make it more productive (such as knowledge and training) while others might make machines specifically more productive (such as faster micro processors). In principle one can try to separate these, but it’s not so easy in practice, so economists look for an overall measure of the intangible contribution. If we adjust labour productivity for the quantity of capital per worker we get a measure of output per unit of combined capital and labour, which the statisticians call multi-factor productivity (MFP). Thus the growth of MFP can thought of as the part of GDP growth that isn’t caused by increases in the number of workers and machines…
    But whatever you do, get rid of the hyphen from ‘poorly-understood’.

  19. Of course, these articles always represent the bias of the author, and you do give the sceptics more time in the sun than most would give. It’s a pitty that I presume this type of article doesn’t lend itself to admissions of bias, like a quick comment explaining that you are one of the sceptics?

    Also, in some parts I get the impression that the productivity debate in Australia is between two even and united teams. I don’t think this is the case.

    More constructively, I think you could say more about the debate as to whether the 1990s productivity growth was a short-term or long-term change in productivity. That is, was it a change in the level (with associated short-term catch-up growth) or was it a permanent change in the growth rate. To some degree, I think you emphasis on the growth-sceptics steals space from this more interesting debate.

    Finally, I don’t think that future growth or recession will give an answer to the debate. Microeconomic reform that increases long-run growth rates do not preclude the possibility of future recessions. And if the micro-reform lead only to a change in levels (not growth rates) then the slow down in growth rates is to be expected, and is not a sign of failure.

    Hope that was helpful.

Comments are closed.