Productivity and the Productivity Commission (updated)

For well over a decade, I’ve been debating the claim made by the Productivity Commission that Australia experienced a productivity surge in the 1990s. My claim has been that the apparent high rate of productivity growth in the mid-1990s was the result of measurement error, most importantly the failure to take account of the increase in the pace and intensity of work that was apparent to everyone (except PC economists) at that time. This view led me to conclude that the supposed productivity gains would dissipate as more normal labor market conditions returned, which was exactly what happened.

In most of these debates, one of my chief antagonists was Dean Parham, who worked for the PC at the time, and is now a Guest Researcher there. Today I heard that Parham had written a new paper on the weak productivity growth of the 2000s. So, I was keen to see what response he would have to my latest work and to my arguments about work intensity. The answer, quite literally is “Nothing”. I have, it appears become an un-person at the PC. Parham doesn’t cite any of my work and, more importantly, fails to mention work intensity at all.

Update The original version of the post contained a somewhat snarky suggestion that Parham had been negligent in ignoring my work. He has written to me to say that this is incorrect. The reason he doesn’t mention it is because, in his view, nothing I have written on this topic, at least since 2004, merits a response.

Further update Dean Parham writes that

the reason I did not mention your work or the work intensity thesis in my paper is that I did not consider it central to the focus of the paper (industry contributions) or even to the contextual motivation of the paper.

Since the contextual motivation of the paper is (as the title suggests) the slump in productivity, I can’t see that this differs from my summary. If Parham thinks my work merits a response, he’s welcome to provide that response here or in any other venue that suits him.

I’ve got some urgent commitments over the next few days, so I won’t be able to return to this topic until later. But in the meantime, here are some of the things I’ve written about this in the last few years. Agree or disagree, I think I’ve put forward a serious case that deserves an answer.

72 thoughts on “Productivity and the Productivity Commission (updated)

  1. the below from the links is an excellent point by John that invites an answer:

    “In this paper, it will be argued that, given its relatively short duration and high year-to-year variability, the MFP data set does not contain enough information to allow clear statistical discrimination between competing hypotheses.

    As a result of this lack of information, combined with the human predilection for observing patterns, a range of alternative stories, each of which may be supported by an appropriate interpretation of the data, has been produced.”

  2. Interpreting the squiggles created by a few years of MFP point estimates, especially given the magnitude of uncertainty surrounding each point estimate, is a bit like reading tea leaves.

    Of course, some do swear by tea leaf readings…

    From a Bayesian perspective, ‘correct’ interpretation can be a whiz if one starts with a tight enough prior. Thus, tea leaf reading can be provided with a sound theoretical foundation.

  3. JQ, I read your 2004 post, most of Dean Parham’s paper, a bit of the ABS material referenced in the Parham paper.

    Please let me know where I am wrong in the following.

    1. Your 2004 material introduces the theoretical concept of ‘productivity’, makes clear what ‘multifactor productivity’ is supposed to mean, and then provides sufficicient hints as to the difficulties in going from the theoretical concept to measurements, particularly when aggregate GDP data is used.

    2. You introduced ‘labour’ or ‘work intensity’ and you commented on the lack of interest in this notion among some or possibly many economists. I do not know of a conceptual definition but I found it relatively easy to come up with one. Here it is.

    Consider a production technology for 1 type of output, denoted by q(i) (physical quantity). The technology requires labour inputs (typically more than 1). Consider onlt 1 type of labour input, measured in time units (ie physical quantities) and denote it by L(j), where j is an index, say j = 1, …, n. Now introduce a time index, such that L(j)[t] denotes the intensity j of (the one type of) labour applied at time t to earn wage W[t] (fixed for the type of labour at t).

    An increase in labour intensity (for the one type of labour) from t=1 to t=2 is representable by two n-dim. vectors such that for t=1 the vector is [L(1), 0, …,0] and for t=2 the vector is [0, L(2), 0, …, 0].

    Now we can relate quite a lot of empirical phenomena to the notion of ‘labour intensity’ and productivity.

    a) Those who expect labour productivity to increase continuously seem to imagine that n tends to infinity. This is contradicted by empirical observations such as physical and mental resources of any type of labour being finite (scientific knowledge).

    b) Data on accidents rates during ‘productivity cycles’, workers compensation rates, psychological injury cases can be correlated with survey data on ‘L’s’ subjective evaluation of what is demanded of them to earn W, as well as objective data.

    c) Increases in labour productivity over the biologically feasible range may not be linear.The notion of a technological optimum, as understood in micro-economics, isn’t all that silly. As work intensity pressure is increased on L, the quality of the work changes (cutting corners) such that in the limit only ‘errors’ are produced.

    d) The theoretical time dependent technological optimum isn’t directly observable. Hence it is quite conceivable that ‘managers’ try out to push more work onto L and it may or may not lead to an increase in L-productivity. Sensible people stop doing something which is counterproductive. Not every manager of L is an idiot.

    e) The time index, t, assumes discrete but constant time intervals. In reality the time interval for W(t) is linked to the micro-economic reform invention of enterprise agreements. These ‘instruments’ have an enterprise specific time horizon. This wage time horizon may or may not correspond to the managerially (or ‘voluntary’ market pressure – want to keep my job idea) determined time horizon of work intensification.

    If I were to keep at it a little longer, other examples of complexities even for a single type of labour and a single output would come to mind.

    Now, Parham says his main work relates to multi-factor productivity [MFP](a residual, in your terminology, after ‘capital’ and ‘labour productivity’ adjustments have been made).

    So Parham is concerned with the explanation of an error term, so to speak.

    I seem to recall the ABS talks about MFP statistics being experimental.

    Your argument is immediately plausible. But this does not constitute a critique of Parham’s work.

    Parham’s arguments (analysis) is not plausible on the grounds that it assumes time runs backward, from ‘output’ to ‘inputs’. You don’t believe me? I quote: “Typically, output growth provides the additional income needed to finance additional growth in inputs.”

    The most generous interpretation I can find for Parham’s ‘argument’ is that he implicitly assumes complete Arrow-Debreu future markets. In this case there is no need for the ‘finance industry’, because the value of future state contingent output is realised at the time when inputs are acquired. While there are some futures markets, these markets have different properties to the theoretical future markets.

    Parham aim to disaggregate national accounts data. Setting aside the problem of measurement errors and the problem or revisions, the theoretical conditions for going from the aggregate to its components are severe. Parham has not even touched on this problem. On the contrary, he excludes data.

    Parham seems to be puzzled by the growth in profitability despite the alleged decline in MFP growth. I am not. All that is required to ‘explain’ this, using my simple model above is considering the possibility of a period where, thanks to the micro-economic reform invention of enterprise agreements, the wage rate is fixed over 2 or more episodes of work intensification (ie an implicit reduction in the wage rate).

  4. EG, John has done a good job since the early 1990s challenging the loose talk about the net gains of economic reform.

    There is a literature dating back to Harberger (1954) puzzling over why the social losses from public policies and monopoly are small. If the welfare loss triangle is small, so are the gains from closing it.

  5. Jim Rose, I am not convinced you read my post because I am not in disagreement with Professor Q but I am asking him to review my critique of Parham’s article.

  6. EG, my post was not intended to criticise your post.

    The prevailing model of deregulation is essentially a nirvana model in that the gains from deregulation can essentially be had without cost.

    The transitional gains trap are just a subset of a more general phenomenon indicating that deregulation can never replicate the status quo ante. The rent-seeking costs of the original privileges were capitalised and lost forever. They are not regained by economic reform.

    Further rent-seeking costs are incurred in lobbying for and against proposed reforms. Reform is not a free lunch.

    google tollison and romance, realism and economic reform

  7. Talking of ‘just so’ stories, fact free, but with a strong narrative and policy message, here is an interesting paper which suggests which side of politics has a predilection for the minimal intellectual effort outlook on life: “Low-Effort Thought Promotes Political Conservatism”

    Ain’t that remarkable?

  8. Ernestine – the model you sketch in point (2) looks like a good way of representing the work intensity problem.

    On your critique of Parham, I think you may be reading a theoretical analysis into a passing remark. I doubt that Parham has a fully-developed growth model in mind. Rather, he is making some ad hoc references to Australia’s dependence on imported capital and its implications for medium-term measures of productivity.

  9. In relation to productivity measurement there is also the issue of capacity utilization. An increase in productivity may be an artifact of higher utilization, which could also be thought of as an increase in intensity and may involve mismeasurement of depreciation if captial stock ‘wears out’ at a faster rate as a consequence. Likewise, more intensive use of humans can wear them out at a faster rate (as slave owners were probably aware). (I suppose that might be an efficiency argument for transferable property rights in labour – on efficiency grounds. )

  10. Erneestine says;

    “Parham’s arguments (analysis) is not plausible on the grounds that it assumes time runs backward, from ‘output’ to ‘inputs’. You don’t believe me? I quote: “Typically, output growth provides the additional income needed to finance additional growth in inputs.” ”

    I am not defending Parham overall. However, does Ernestine not consider that positive feedback loops exist? Output growth (from some cause, any cause) creates income growth, some of which can be used to finance growth by re-investment. Referring to positive feedback does not allow us to infer that time runs backwards, surely?

    But back to the main game. I was in the Federal Public service in the Howard era, supposedly the golden age of productivity growth. It was very clear where I was there that computerisation drove most productivity growth plus the removal of some public service “evils” like gold-bricking (bludging, goofing off, going to the pub). However, computerisation was not invented by the Liberals and Hawke / Keating had already removed most of the gold-bricking.

    Howard’s “reforms” were negative and partly offset the computerisation gains by subjecting us to endless and pointless re-structures like turning DSS into Centrelink. Under-staffing and increased pressure did create a decline in work quality and computerisation was (mis)-used to make welfare calculations ever more ridiculously complicated.* In that way, many of the efficiency gains were frittered away to pander to the ideologies of the neo-cons.

    * I always argued that the attempt to measure welfare need too finely (by over-complication of legislation, policy and computer systems) was futile as it was subject to measurement error. Over-complication of data collection confused recipients and led to much data being of very dubious quality or just dwonright wrong.

  11. I will make one exception to my rule of not participating in blog conservations to address a factual error. John states in his update about the fact that I did not cite him or the work intensity argument:

    ‘The reason he doesn’t mention it is because, in his view, nothing I have written on this topic, at least since 2004, merits a response.’

    That is not correct. What I did say about the reason for not citing him is as John reports in his ‘further update’. I did also say – making it clear that this was not a reason for not citing him – that he has not responded to the criticisms of the work intensity argument that I made in my 2004 Economic Record review article on Australia’s productivity surge in the 1990s

  12. To respond briefly, Parham (2004) asserts

    Increases in work intensity through
    reductions in slack on work time or increases in pace of work would be genuine sources of productivity improvement.

    I have frequently responded to this claim, but will do so at greater length in a subsequent post. For the moment, observe that, on Parham’s definition, increases in productivity, from an initial starting point of labor market equilibrium, will generally reduce welfare.

  13. Adam Smith thought that if labour was used more intensively, then, in a competitive market, an employer would have to pay extra in compensation. The so called theory of equalizing differences. And there is some evidence to support this. Hence, there is “no free lunch “” and no real productivity improvement when labor is used more intensively.

  14. when is a reduction in transaction costs not a productivity improvement? When is less principal-agent slack within a firm (or a bank) not a productivty gain?

    Armen Alchian’s insight was that the set of rules (the distribution of property rights) determined the level of output of the firm (and the economy) because they determined the incentives of each individual.
    – When property rights are perfect, by definition no theft (including shirking on work effort) can take place and as a result, no effort is made to protect these property rights

    – When property rights are incomplete, individuals attempt to increase their ownership in an effort to increase their wealth. This attempt to capture property rights may be dissipating (as in the case of theft), or wealth generating (as in the case of assets brought out of the public domain or common property, all of which can be equilibriums).

    – When property rights are incomplete, individuals are always in the process of maintaining their existing property rights and attempting to establish new ones.

    This leads to the property right definition of transaction costs: the costs of establishing and maintaining property rights.

    Microeconomic reform affects the costs of establishing and maintaining property rights.
    – There is a problem of monitoring of employee and team effort within the firm – collaboration is subject to shirking.
    – The net gain from collaboration depends on the contract governing it.

    The formation and organisation of the firm can be attributed to ways of monitoring team effort. Replacing solo artisans with assembly lines in a large firm is an example.

    Improved property rights definition and enforcement and harder budget constraints, for example, because of microeconomic reforms redirects efforts away from rent dissipation (such as through shirking by employees) towards wealth creation.

    As an example, the abolition of academic tenure is justified in part because the reduction in shirking outweighs the benefits of tenure. Academic tenure is defended on the grounds that any academic shirking that there might be is worth the price of giving strong property rights to academic employees over their jobs. This balance changes with the times.


  15. Neoconservative “economic reform” including micro-economic reform and privatisation of publicly owned enterprises has been wholly and solely about the transfer of accumulated wealth from public to private ownership and the transfer of income from labour to capital. The theory and justifictions are the typical right-wing pseudo-evidence; propaganda paid for by the monied classes.

  16. @Ikonoclast

    Is private ownership the real problem? Presumably, a workers cooperative can have private ownership, and in the right legal structure, would not generate problems. If a single operator business, merely makes their own wage, is this private ownership a problem?

    On the other hand a public owned enterprise which engages in capitalist activity, creates huge problems.

  17. @Ikonoclast

    “does Ernestine not consider that positive feedback loops exist? Output growth (from some cause, any cause) creates income growth, some of which can be used to finance growth by re-investment. Referring to positive feedback does not allow us to infer that time runs backwards, surely?”

    My preferred economic theories are those which involve robust concepts that can be related to daily life. ‘Production’ and ‘productivity’ are such concepts. And, as I shall try to illustrate, production involves time, my criticism of Parham’s paper stands, and positive feedback loops are irrelevant.

    Fried eggs don’t ‘grow’ but weed grows.

    Fried eggs is the ‘output’ of a production process requiring the following inputs: raw eggs, a heat source, a surface suitable for heat transfer. We call the knowledge of how to produce fried eggs ‘production technology or ‘technological knowledge. Fried eggs sunny side up and fried eggs sunny side down involve 2 production processes. (The number of production techniques can be expanded by considering cooking surfaces that require some fetty substance versus those that don’t, etc). The technological knowledge may be emboddied in machines or it may be in the head of a single person at a particular time and place. Notions such as ‘economies of scale’ crucially depend on a the notion of a production technology. A ‘production’ consists of inputs and at least 1 type of output. Even the fastest production technology for fried eggs, you will agree (because I know you are not averse to reality) requires time. Moreover, it is not possible to retrieve the inputs, in their original state from the output. We call this the irreversibility of production. (‘you can’t unscramble an egg’).

    Now weeds grow in many gardens all by themselves. These weeds by themselves do not require inputs, from the perspective of the concept of ‘a production’ (although they require inputs in a biological sense, namely soil and some water and no weed eating animals). It is only when the owner (public or private or communal) of a garden wants to get rid of the weed that the inputs are required. So the output is a weeded garden. The inputs consist of spontaneously grown weed, labour, compost or other disposal facility. The weeding takes time (even with a weed killing agent, another input).

    No matter how often the ‘fried egg production’ and the ‘weeded garden’ productions are repeated and no matter how large the quantities (as long as they are finite) of these productions grow, inputs preceed outputs in real time.

    Both, the production ‘fried eggs’ and ‘weeded gardens’ may be produced and consumed by the same person (ie within a household) or sold ‘in the market’. Parham uses national accounts data; only market transactions matter.

  18. @Ernestine Gross

    I am sorry Ernestine. Your answer baffles me. I understand it in its own terms but I don’t understand its application to the point of contention. However, you may be arguing a technical point that has eluded me. The key could be your last statement:- “Parham uses national accounts data; only market transactions matter.” The form of Parham’s argument may be susceptible (can’t think of a better word right now) to your argument because of this.

    I was thinking in a more prosaic way. In a diner, outputs (eggs sunny side up or over easy) can be sold for a profit. Profits can be used to buy another stove and more frying pans if customer numbers warrent it. This was the sense in which I was (loosely) using terms like inputs, outputs and feedback loops.

    I still feel reasonably sure this is the sense in which Parham meant “Typically, output growth provides the additional income needed to finance additional growth in inputs.”

    Be that as it may, you will be aware that I am a strong ideological opponent of the pseudo-science of the Productivity Commission. The Commission is largely a tool of neoconservative propaganda.

  19. So, in summary, the Productivity Commission is a taxpayer funded pack of capitalist running dogs (?).

  20. Facts, reasoned arguments and anticipated consequences win arguments. Insinuations of malevolence and moral turpitude do not.

  21. But might, might be right? And who could argue with those anticipated consequences?

  22. “Increases in work intensity through reductions in slack on work time or increases in pace of work would be genuine sources of productivity improvement.”

    What Parham asserts in 2004 is just playing around with words; especially with the word “genuine”.

    Lets put it this way, hypothetically assuming labour does slack off some time per work hour and assume if labour market work intensity is 0.65/hour using a range between 0 and 1. This implies that the employees work hard for about 39 minute per hour and the rest is slacked off. Even if the work intensity increases by reducing time slacked off, say work intensity of 0.75/hour after productivity gain; this would imply the employees now work hard for 45 minute per hour and the rest is slacked off. Furthermore assume the work hour per business day is 8 hours paid. This would imply that the employee is working 5.2 hour of actual intense work prior to the productivity gain and 6 hour after the gain.

    Then you look at the fact that employers actually do measure how much their labour produces and how much to pay them. This is only simply measurement that every employer have to do to determine how much to pay their employee, how long is the work hour per day and how much should their product cost etc. During this measurement process prior to productivity gain, the employers would assume that they are hiring their employees for 5.2 actual intense work hour per day instead of 8 as offical work hour. From this perspective if the actual intense work hour per day increased to 6, the offical work hour don’t actually have to increase, but technically the work hour for the employees had increased already.

    If productivity gain is purely from this increase in work intensity, this is nothing ‘genuine’ compare to improvements in technology and skills of the employees (or increase in pace of work as suggested by Parham, which any reasonable person knows there is a limit to this increase) that actually increases the output per intense work minute or hour. Also, this increase in work intensity is very finite as well as putting stress in employees.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s