‘Twas the land bubble that killed productivity

Even though more words have been written about Australia’s productivity performance than most other economic issues, I have learnt very little about what our productivity trends really mean.

Recently, the RBA tried to unravel the mystery. Here at MacroBusiness my wise colleagues have often penned their interpretation of events.

To throw a little more confusion into the mix,  the RBA’s D’Arcy and Gustafsson note that:

…there is considerable measurement error in the estimates of productivity growth making it difficult to be precise about the timing of changes in the underlying trend; and productivity growth is the result of the interaction of many fundamental and proximate factors.

Technological, structural and regulatory changes, as well as cyclical variation in factor utilisation, can all affect measured productivity, making it very difficult to identify and disentangle the various effects.

But we are given a hint at the important conceptual basis of the thing we refer to as productivity:

Conceptually, economists often view technology as determining the productivity ‘frontier’; that is, the maximum amount that could be produced with given inputs.

Factors affecting how production is organised, including policies affecting how efficiently labour, capital and fixed resources are allocated and employed within the economy, determine how close the economy is to the frontier.

Trend productivity growth is then determined by the rate at which new technologies become available – how fast the frontier is expanding – and the rate of improvement in efficiency – how fast the economy is approaching the frontier.

It all sounds very theoretical, but the reality actually very simple.  The chart below shows the two key measures of productivity since the 1970s, and the declining multifactor productivity (MFP) that has attracted so much attention.  Labour productivity has remained positive, if a little lower than the historical average:

There are two questions I will answer in this article:

  1. Why has is labour productivity growth historically low?
  2. Why has MFP growth been negative for the past decade?

To answer the first question we need some perspective about whether Australia’s performance is abnormal compared to other nations.  If not, then I suggest there is little that can, or should, be done.

The Productivity Commission has some good figures on our performance against other comparable nations.  It seems that our productivity performance is… wait for it… actually pretty good, and fairly stable in relation to the US and EU.   Comparing GDP per hour worked, the fundamental measure of labour productivity, Australia has made gains on the EU15 during the 2000s, and has lost just a little ground against the US up to 2007.  The chart from the RBA below clearly shows that we were middle of the road of productivity growth in global terms:

So why then would labour productivity be historically low across the world? Mostly, it has to do with significant structural declines in unemployment.  Typically the least productive people, those with few skills to utilise capital effectively – to ‘leverage’ their work with the help of machines, computers, tools and so on – are the last to be employed during periods of strong growth, and the first to lose work during economic contractions.  Thus the expected outcome is that during economic boom periods of declining unemployment, labour productivity will be biased down by these new workers, compared to if unemployment was flat.  We should also expect that during periods of increasing unemployment, that labour productivity surges again.  When the least useful one percent of the workforce is laid off, production usually declines just a fraction of that one percent.

In addition, much of the mainstream productivity discussion is dominated by the influence of mining and infrastructure, the two industries with the largest declines in productivity.  The basic arguments are as follows:

  1. As widely noted, we have a ‘wall of wire’ problem in much of our basic infrastructure.  This simply means that the honeymoon period of relatively new electrical, phone, water, and waste infrastructure is over, and major maintenance and capital expenditure is becoming more frequently required to deliver the same service.
  2. In mining, a sector showing substantial productivity declines in recent years, we have the situation where “rising minerals prices meant low-productive mines were profitable, and thus the extraction of minerals from those mines actually assisted in lowering the sector’s and the economy’s productivity”

These arguments are both true and apply to those sectors in terms of both labour productivity and MFP.   Another major factor is unpredictable seasonal changes in the agricultural sector output.

So what of our MFP performance?

If we have been tracking fine in terms of labour productivity, the actual only meaningful and useful productivity measure that reflects the benefits from economic growth, why the dismal pattern of MFP, the measure most economists prefer to fuss over? And why do they prefer it anyway, when labour productivity is the only one that matters?

As noted in the RBA report, economists believe that Total Factor Productivity (TFP), or Multi-factor Productivity (MFP), measures changes in technology and market structure that enables the ‘production frontier’ to shift outwards.  But when the idea of MFP was originally put forward, it was known as the Solow Residual, because it is “the part of growth that cannot be explained through capital accumulation or the accumulation of other traditional factors, such as land or labor”.  Essentially, it is the bit left over after we measure all the inputs and outputs of the economy.  Economists thought they might call it ‘technology’ or ‘productivity’, because it appears to measure our ability to get something for nothing.

But in reality, it is capital accumulation that almost exclusively improves labour productivity, and the scale of our per capita productive capacity.  Having more, and better, machinery, buildings, infrastructure networks and other capital equipment, is what enables each person to be more productive.  Using better machines, for example, can improve how many meters of road can be laid by a small team of workers in one day, and the quality and durability of the resulting surface.  As the economy accumulates capital, all parts of production require less labour per output.  It is one of the main reasons the agricultural sector requires such a small workforce.   If I haven’t repeated myself enough already, it is capital accumulation that explains almost all the improvement in labour productivity (for example, see here).

To recap, labour productivity is simply a measure of output, usually GDP, divided by labour input, either in per employed person, per working hour, or per capita.  MFP is a measure of output divided by the sum of inputs of labour and capital, including land. I use the term productivity  to mean MFP, or will explicitly state labour productivity when referring to it.

To answer the question of why we have experience declining MFP, we have to think about what can cause a divergence between the two productivity measures. MFP is the result of dividing output, measured by GDP, by the sum of labour and capital inputs.  So either we are using our capital less efficiently, requiring more new capital for each improvement in output (diminishing returns to capital), or we have some kind of measurement anomaly in the estimation of the balance of capital assets.  Indeed there may be some diminishing returns to capital effect, but after investigating this anomaly I found that falling MFP is substantially the result of estimates of land prices in the measure of the capital stock. 

The culprit is hidden deep in the ABS release 5204.0 System of National Accounts.  Back in 1999 the methodology for estimating MFP changed. One critical change was the inclusion of non-agricultural land in the capital stock.

“the scope of capital inputs has been changed to include the capital services of livestock, intangibles and non-agricultural land and to exclude ownership transfer costs”

The ABS believes that the exclusion of non-agricultural land biased the measure of MFP downwards in the past.  But this only applies to the situation where the value of land assets grows with inflation.  When land values significantly exceed inflation, which has especially been the case since 2001, the capital stock component in the denominator of the MFP calculation increases, for no particular reason.  Theoretically, the inclusion of land is very odd, since it is always fixed in any case.

The ABS explains that they take the balance sheet value of land from the national accounts to include as the land component of capital stock. We can observe in the chart below the rise in the value of the land balance sheet value against the estimate of MFP, and indeed against an estimate of the land balance if land values simply tracked inflation.  Quite clearly, from about 2002 onwards the abnormal increase in the value of land lead to a flattening and falling estimate of MFP.  More telling is that fall in all land asset values in 2009 lead immediately to an increase in the MFP measure, only for the next wave of land price escalation, especially FHOB stimulated residential land, to cause a deterioration in MFP during 2011.

We can dig a little deeper into the ‘land balance sheet’ in the system of national accounts, and look closely at the type of increases in land value estimated.  The chart below shows in blue the neutral holding gains – that is, the change in the value of land expected if prices tracked inflation. In red we see the real holding gains, which are market-based increases in land values.  As the ABS notesHolding gains and losses accrue to the owners of assets and liabilities purely as a result of holding the assets or liabilities over time, without transforming them in any way”. In economic terms, they are pure rents.

When red is greater than blue, we find a significant downward bias in the MFP estimate.  It is really that simple.   And we are not alone in this either.  Spain’s land price boom resulted similar pattern of declining MFP during their land price boom in the early 2000s.

Let us wrap up by summarising the key points from this analysis.

  1. Australia is not performing abnormally low by international standards in productivity growth.
  2. Labour productivity is the most important productivity measure, and improves almost entirely through capital accumulation.
  3. Labour productivity is usually biased by changes in unemployment.  Reduction in unemployment results in a downwards bias as new labour is employed before capital can be produced to help the expanded workforce produce more effectively.
  4. Multifactor productivity is the bit left over after adding up all the economy’s outputs and subtracting all the inputs.  It typically captures compositional changes in goods produced.
  5. Multifactor productivity has fallen mostly because the denominator of the productivity equation has been so heavily influenced by inflated land prices across all sectors since 2002.  I expect if the slow melt in land prices continues we will see a ‘surprising’ recovery of the multifactor productivity measure in the coming years.

Tips, comments and suggestions to [email protected] or follow me on Twitter @rumplestatskin


  1. Great article Rumplestatskin. You should get the ‘freakonomics’ award for this one.

    I am left rather perturbed by the increase in the estimated land value since the start of the century. We are told we have a land bubble, but when you see this, you are just blown away by it’s size; particularly when compared a neutral scenario when land prices track inflation.

  2. I wonder why those who rail against falling productivity growth fail to see the connection between the productivity and boom in non-productive assets. And that it takes an academic to point this out. Perhaps I am just naive…

  3. “It is capital accumulation that explains almost all the improvement in labour productivity”.

    As I have said before, many people on this forum take a glass half empty view of the high Aussie Dollar while ignoring that it also represents a terrific opportunity to upgrade equipment.

    Airline industry is one example. The business case for newer, higher capacity, more efficient aircraft can live or die on the exchange rate calculation.

  4. I have a question related to the labour productivity measure:

    “We should also expect that during periods of increasing unemployment, that labour productivity surges again. When the least useful one percent of the workforce is laid off, production usually declines just a fraction of that one percent.”

    So, according to that measure at least, it’s more productive to have people sitting idle than actually making stuff. Seems to be more micro than macro. Is there a measure which includes the lost productivity due to unemployment?

    • Rumplestatskin

      “according to that measure at least, it’s more productive to have people sitting idle than actually making stuff. ”

      If you are only counting inputs and outputs, yes.

      GDP per capita is a rough measure to take into account unemployment, but you need to make some adjustment for the changing size of the workforce. So perhaps GDP per person in the labour force is the best productivity measure to guide policy.

  5. invest-magicMEMBER

    It is possible to see the Land Balance sheet graph in log scale? We might get a better appreciation of the size of the bubble. We might also see the land prices falling.

    Awesome analysis – another reason why MSM publications are such a waste of time.

  6. Bryan Kavanagh

    Very good article, Rumple! Only part I have a bit of issue with is “Reduction in unemployment results in a downwards bias as new labour is employed before capital can be produced to help the expanded workforce produce more effectively.”

    That sounds suspiciously like ‘Wage Fund Theory’ to me, Cam.

    As a director of a real estate valuation company, it pretty soon became clear to me that our staff created their own wages (and our company’s profits). We didn’t have no fund.

    • Rumplestatskin

      Perhaps I could be more precise with my phrasing. What I mean is simply that it takes time when the labour force expands rapidly (on the expectation of future economic expansion) there is an adjustment period where new people are being trained, new equipment being ordered/produced, adjustments to business structures are taking place. In addition, of course, to the ‘regular’ amount of such adjustments.

      Lags associated with capital investment during these adjustments are not the sole cause of the bias in the productivity measure, but it was simply a rough way of saying that organising the expanding workforce to produce the same or greater quantities in per capita terms takes time.

  7. Wages. Australian minimum wages are too high compared to our competition. Our shortage of labour has allowed our lowest skilled workers to earn far in excess of their productive capacity. Nothing to do with land prices.
    Adjusted for a high AUD our wages in international terms are ridiculous seen through the productivity lens.

    • Rumplestatskin

      I’d be keen to hear how you expect lower wages to increase productivity. Especially since my current view is that wages, especially minimum wage laws and the like, have very little impact on labour productivity, or even a positive impact, in the long run (under most moderate scenarios).

    • Alex Heyworth

      If you lower wages, ceteris paribus, more people will be employed. Because they were not employed before, they are presumably less productive workers. Hence, overall productivity must fall slightly because the more productive workers have been diluted with a less productive group. I don’t see how it is possible to argue otherwise.

      • If you lower wages, ceteris paribus, more people will be employed

        Erhh no. Employment is driven by aggregate demand. Just because the minimum wage is lowered, you will not find previously inert labour now being offered the opportunity to bring product to market.

        In fact, you are more likely to cost jobs in the long run because you have lowered aggregate discretionary income. Many above minimum wage structures are informally pegged in relation to the minimum wage.

        Less productive workers do not dilute those they are already productive, they reduce the marginal increase in productivity.

        You have phrased it as “productivity must fall”, inferring aggregate output. You can’t a reduction in output because you add another worker.

        • Alex Heyworth

          RP, I must disagree. There is a well established relationship between minimum wages and the level of unemployment. See for example Leigh, 2003 (that’s Andrew Leigh, now Member for Fraser) and the 2006 NBER review by Neumark and Wascher.

          And I am not confused between aggregate output and productivity. I think that was razorack’s problem. Think of it this way. Your factory employs 10 people, each producing 6 widgets per hour. The government lowers the minimum wage you have to pay your workers, so you employ another 5 workers. However, they are less productive and only produce 4 widgets per hour. The output of your factory has gone up, but the productivity has gone down. Your workers (now 15 of them) produce 80 widgets per hour between them, an average of 5.3 each – less than the average before.

          • Rumplestatskin

            Employment impacts of minimum wage laws are definitely a debate for another time.

            However I will raise a few points to ponder in the meantime.

            1. How do countries without minimum wage laws perform regarding employment statistics?
            2. Surely for minimum wage workers the take-home wage itself is just one of many of the costs of employment. Perhaps we reduce the gap between the employment cost the the employer, and the wage benefit to the employee.
            3. Remember, statistically significant doesn’t mean actually significant, or true. It about the weight of evidence.

            Given all I have read on the topic, my current view is that under most moderate scenarios the employment impacts are negligible, if they exist at all.

          • There is a well established relationship between minimum wages and the level of unemployment.

            I personally would not consider it well established at all. I would contest it to be still a battleground fought on ideological grounds.

            See for example Leigh, 2003 (that’s Andrew Leigh, now Member for Fraser) and the 2006 NBER review by Neumark and Wascher.

            And I am not confused between aggregate output and productivity. I think that was razorack’s problem. Think of it this way. Your factory employs 10 people, each producing 6 widgets per hour. The government lowers the minimum wage you have to pay your workers, so you employ another 5 workers. However, they are less productive and only produce 4 widgets per hour. The output of your factory has gone up, but the productivity has gone down. Your workers (now 15 of them) produce 80 widgets per hour between them, an average of 5.3 each – less than the average before.

            I would say we’re differing on semantics.

            I would say before and after the hiring of the additional 5 people, the original ten are still going to produce 6 widgets per hour.

            They are still as productive.

            I would reaffirm my statement that the additional have a marginally lower rate of productivity, thus meaning the margin of profit (if any) diminshes.

            My defintion of productivty is to each of those ‘per unit of input’ in isloation thus meaning they are not impacted by additional, lesss productive units.

            I suggest that’s where we differ.

            As long as even the least productive are offering value in excess of the minumu wage, and that must be taken in light in regards to aggregate demand, then a ‘too high minimum wage’ is of no value.

            I view it more of a grab for a greater share from capital, that side has never said anything but “the minimum wage is too high”.

          • Alex Heyworth

            Cameron, OK, happy to leave this discussion for another time. It’s your thread.

            RP, I’ll just say that productivity (in the discipline of economics) already has a meaning. You don’t get to make up your own.

  8. Alex Heyworth

    Cameron, one other factor that I think must be occurring, although I don’t know how significant it is, is that as our economy has become more capital dependent, the amount of capital required to replace and maintain existing capital equipment has grown. You mention this in relation to infrastructure, but it affects many other areas of industry as well – even to the need to replace the espresso machine in the coffee shop now and then. As the capital stock ages the proportion of capital, as well as the amount, required for this purpose will rise.

    Consumer expectations affect this turnover of capital stock as well. How many times have you seen a perfectly good shop interior ripped out to be replaced with a new one? Not to mention the frequency with which consumers turn over their own capital stock – mobile phones, TVs, hi fi gear, cars – perfectly good stuff is ditched all the time, usually for very little improvement in functionality.

    • Yes, Leith, but this is just about faulty statistical analysis. Good for Cameron for pinging the number crunchers on this, but is he still agnostic on the ACTUAL effects on prices and productivity, of urban land supply constraints? See my comments below. This stuff should all be obvious; of course productivity IS reduced through these effects, it is not entirely that the statistics are distorted because of the way they are calculated.

      Before anyone says that there was no actual reduction in productivity, let me just point out that no-one knows how much better it MIGHT have been. It is all very well to wait 50 years and let a McKinsey Institute tell you you are 20% to 40% behind where you might have been, or let a “Cheshire and Sheppard” tell you that you are imposing an economic burden equivalent to a 4% tax on all incomes.

  9. anonysubscribe

    if one measure distorts conclusions, we should remove it from conventional wisdom until its flaws are not glossed over by economists who should know better than bamboozle us with invalid models

  10. Cameron, well done for this triumph of forensic statistical investigation, but do you agree that ACTUAL productivity does indeed fall (rather than just a statistical result of the way productivity is calculated) as a result of growth-constraint urban planning and inflated urban land prices? This is something I have been pointing out on this forum for some time.

    My exhibits are: The McKinsey Institute’s 1998 paper, “Driving Productivity and Growth in the UK Economy”
    Alan W. Evans’ 2004 book, “Economics and Land Use Planning”
    A series of analytical papers from the London School of Economics, starting from Cheshire and Sheppard, (2002) “The Welfare Economics of Land Use Planning”.
    A recent paper summarising the LSE research (but not Evans’, possibly because of academic rivalry), is
    Max Nathan and Henry Overman: “What We Know (and Don’t Know) About the Links between Planning and Economic Performance”


    Productivity is undermined by restrictive urban planning and zoning, because of anti-competitive effects and reduced formation of economic agglomerations; because optimum amounts of land for efficient processes are too expensive in most industries; because capital is required to be sunk into dead excessive land “value” instead of productivity-raising equipment; because workforce cost pressures are higher due to increased “dead” housing cost increases; and because of increased congestion in a forced “lower urban footprint” for the urban economy.

    All this stuff is not in the least counter-intuitive. Cheshire and Sheppard estimate the effects to be like an additional tax of 4% per annum on incomes in the UK economy; the McKinsey paper estimated the UK’s productivity to be 20% to 40% lower than its major trading partners, the main point of difference that could be blamed, being the UK’s planning system and inflated urban land costs.

    Of course this is a long-term process and may not be so immediately evident from one single unprecedented urban land price bubble phase. But I am picking that the erosion of productivity is one reason that the inflated land prices cannot be maintained indefinitely; eventually the economy is squeezed from both ends, so to speak; its costs are increased while its income growth is constrained.

    • Another factor or two that I forgot to list above, that reduces productivity when urban land is over-regulated and prices are too high: reduced mobility of workforces and businesses. Reduced “churn” of land to more efficient uses. (Property vendors holding out for inflated prices). Higher central bank interest rates targeted at property price inflation, squeezing productive investment. Investment capital diverted into chasing property capital gains rather than business investment. Reduced construction industry productivity, increased business uncertainty due to volatility of market conditions. Greater substitution of JIT inventory and low-volume transport, for bulk transport and local holding of product and materials. Infrastructure maintenance, repairs and capacity increases rendered more difficult and disruptive to the local economy.

      • Rumplestatskin

        Productivity is a measure of the quantity of inputs and outputs. Since land is a fixed quantity, regardless of the improvements we make, a change in value shouldn’t flow through to the productivity figures.

        If you are talking about inefficiencies in production because the physical arrangement of capital (buildings, infrastructure, fixed plant and equipment) is limited by planning rules, then you may be correct. But the counterfactuals are pretty difficult to even guess at. Would our cities be more or less dense? Are long commutes and transport costs in high value adding sectors reducing productivity? What about congestion costs, and infrastructure requirements for more sprawling urban areas? Which alternate world would be better for productivity in average?

        • Cameron, the EVIDENCE from the UK, is that urban growth containment has numerous effects that REDUCES productivity. As I said, this is not even remotely counter-intuitive.

          The intuitions that hold that urban growth containment INCREASES efficiency, are WRONG. Increased density compelled via planning, correlates to increased congestion, not decreased congestion. It correlates to vastly reduced traffic speeds and LONGER trip times for all purposes. It correlates to increased local emissions and pollution.

          Urban growth containment policy basically intends to cause more people to use public transport, and hopefully also, to live closer to work, and walk (but inflated housing costs overall, have the opposite effect there). But the correlation between density and public transport use is something like; for every doubling of density, there is 7% higher use of public transport. That means there is 100% more people, and 7% of the total 200% use public transport, that would not have without the doubling of density. But that also means 86% more car users in the same urban area, usually without any extra road capacity having been engineered by the (anti-car) urban planners. So, surprise, surprise, we have far worse traffic congestion, delays, commute times, and local air pollution; and these things are “exponential”. Traffic can go from “flowing” to basically “stop-start”, so that not only are there 86% more vehicles, the road network THROUGHPUT of vehicles is much LOWER, leading to travel speeds dropping exponentially and local emissions increasing exponentially.

          The low density, low land cost urban economies of the USA use more petrol per person BECAUSE their trip speeds are higher and they travel a lot FURTHER in less time, AND they have far higher discretionary spending due to lower housing costs.

          The connection between restrictive urban and transport policy, and reduced consumption of petrol and energy, is via reduced household discretionary income due to inflated housing costs (and in some countries, higher taxes on energy); NOT due to “more efficient urban form” at all.

          The NZ Commission of Inquiry Into Housing Affordability Report contained an appendix summary of research into whether infrastructure costs are lower via policies that increase density in existing areas, or on greenfields. The research was unanimously in favour of greenfields development involving LOWER costs. (The exception is the low-hanging fruit situation of increasing density in VERY LOW density areas).

          The cost of upgrading infrastructure in established areas is very much higher than doing it on greenfields, and the greater the disruption to existing daily life and economic activity, the higher the cost of the “existing area” upgrade.

          It is frequently assumed that there is “spare infrastructure capacity” in established urban areas. This is such a rare exception that to assume it is false.