Wednesday, 19 March 2014

How productive are we?

Yesterday Stats came out with the latest productivity statistics. They're very important numbers: in the long run, it's productivity that largely determines our standard of living, and if you've ever engaged, or want to engage, with why incomes here seem to have slipped compared to the OECD as a whole, or compared to Australia, then you need to start out armed with the basic info on what's actually been happening here.

Here are some of the key findings, taken from the full Stats release. And if you'd like someone else's coverage as well, Patrick Smellie at the NBR did a piece (outside the paywall) which quotes the ANZ's chief economist Cameron Bagrie.

Here's a table of the overall picture.

Two explanatory comments. You'll see that these are numbers for "the measured sector", which is where it is easier for Stats to measure productivity. The missing bits are central government, education, and health, where measuring productivity is iffier (though not impossible), but even so the "measured sector" covers quite a big slab of the economy (about 80%).

And those time periods, which look a bit arbitrary at first, are based on the idea that you ideally need to measure productivity from one point in one business cycle to the same point in the next business cycle. Productivity numbers can get badly thrown off if, for example, you went to measure from peak of boom to bottom of bust (or vice versa). The latest cycle (2008-13), incidentally, is still underway, so we're far from the last word on what's happened this time round.

So, what's been happening?

First, output growth per annum (0.8%) in this latest cycle is low, but that's because we've got the GFC in there (output -3.0% in the year to March '09, -0.4% in the year to March '10), and anyway it's accelerating since. The year to March '13 was good, the year to March '14 will have been better, and it's highly likely that the year to March '15 will be even stronger again.

Second, the GFC and its aftermath were not kind to employment and hours worked, so the quantum of labour input has actually fallen over the period (largely due to a contraction by -2.4% in the year to March '10). This will obviously change as we add on more recent data.

Third, growing output (however slowly) but falling labour input means that labour productivity went up. Obviously we'd all prefer if we got higher labour productivity by labour input going up quite a bit and output growth growing even faster, but there we are. These were difficult times, and the good news is that labour productivity increased by 1.5% a year, about the same as our longer-term average (1.6% a year over 1996-2013). Not a bad outcome on the back of one of the most difficult business cycles in living memory.

If you rearrange a formula a bit, you can get an insight into where this increased labour productivity came from. It can come from each person working with more capital equipment ("capital deepening"), or from something else ("multifactor productivity", MFP, which is shorthand for everything that isn't more labour input or more capital input, and includes everything from technological and process innovation to better management and learning by doing). Here's the outcome.

You'll see that this time round (or at least up to March '13) people have been able to produce more largely because they have had more gear and equipment to work with. The good news is that the capital spend has kept up. The bad news is that usually we get more of a productivity kicker from all that "everything else" MFP stuff, and that hasn't kicked in to date in this cycle. But as you can see, 2013 showed more of a MFP boost, and hopefully we'll see more again this year and next.

We don't need to spend all our time comparing ourselves with the Aussies, but it's an interesting exercise all the same. Here are the numbers.

You'd be tempted to conclude that the reason for the gaps in output growth and labour productivity growth, in Australia's favour, are all down to the benefit that the Australian workforce gets from bigger increases in the capital equipment they've got to work with.

That's actually a bit of a statistical illusion. It's true that the capital deepening component of the gap between our output and productivity performance is growing, yes. But it's still a smaller component than the MFP contribution. Geoff Mason of the UK's NIESR has done the comparative exercise in detail (I covered it here, a short form of his results is here and the whole thing is here), and broadly speaking 60% of the productivity difference is down to better and more inventive ways of organising their affairs in Australia (MFP) and 40% is down to the Aussies having more gear to deploy (there's also a small difference in aggregate skill levels, which I've ignored).

A lot of this will be of greatest interest to hard core macroeconomics types, but I think there's also an important lesson for management everywhere lurking in these numbers, and it's this.

On average, the measured sector has been able to generate MFP growth of 0.7% a year over 1996-2013: in other words, without hiring a single extra employee or installing a single extra machine, organisations have been able to produce 0.7% more each year with the same resources. I'd like to see that number being used. It ought to be built into people's performance targets and business units' business plans - and public service budgets, if it comes to that.


  1. Nice note. But it might also be worth noting that, once SNZ allow for changes in the composition/quality of the labour force, average MFP growth drops back to 0.5% pa.

    1. You're quite right. I dithered for a while about whether to go into the composition issue or not and explain the 0.2% contribution to 0.7% MFP from higher skills and decided to run with the 0.7% as it was being used more as a headline number (as in the Stats comparison with Australia) but conceptually the 0.5% number would have been better to use in my last para

  2. NZ growth in the end is driven by the pace of technological diffusion to NZ.

    these large variations in productivity growth rates and the increase in labour inputs as a main driver of recent economic growth implies large variations in the pace of technological diffusion to NZ. is that possible or is there just measurement error?

  3. Thanks - that's a good observation. Can I think about that, rather than dash off something lightweight? My first reaction was that sometimes cyclical and underlying movements don't always get disentangled neatly, which would be a kind of measurement error, but let me chew on it a bit

    1. thanks, measurement of prosperity has three major biases.

      The average quality of goods: new goods and improved versions of older goods can provide variety and entirely new products and services previously unavailable at any price. Measurement of the impact of new goods is ‘pretty much guesswork at present’ and ‘some very large gains in consumer welfare’ may be missed (Moulton 1996, p. 173).

      An important bias affecting the measurement of prosperity is greater longevity. The life expectancy of males at birth improved by 5.9 years between 1970-72 and 1995-97, and by another 3.6 years by 2005-2007 (Statistics New Zealand 2009a). Life expectancy of males at birth increased by a mere 1.3 years from 1950-52 to 1970-72!

      Becker et al. (2005) estimated that the six year increase in New Zealand life expectancy between 1965 and 1995 amplified the 34.3 per cent increase in New Zealand GDP per capita to the income equivalent of a 47.3 per cent rise. Becker et al. (2005) estimated that the 73.5 per cent rise in Australian GDP per capita between 1965 and 1995 was enhanced to 95.5 per cent after adjustment for the seven year increase in life expectancy.

      even short periods of time are subject to large measurment errors

  4. Hi - I've been thinking about your comment about the implied (and perhaps implausible) variation in the pace of technological diffusion to NZ, and whether this in turn implies MFP growth or other components of the MFP calculation must be mismeasured. First I'd certainly agree that MFP calculations are approximate rather than exact - Brian Easton, quoting some UK economists, likes to call MFP our "coefficient of ignorance", and in any event as a residual it will always be at the mercy of errors in other variables. But I'd also wonder if there can't be genuine and perhaps sizeable variation both in the rate of overseas innovation and in the rate of domestic adoption of that innovation, variation that isn't down to data issues. Were the '60s sizeably more innovative than the '50s? The '90s more than the '70s? Could be. And have we genuinely varied in the speed with which we approach that overseas tech frontier? In principle we could (eg if we had two three-year terms of domestic anti-innovation economic policy) and in practice I'd guess we have: I'm not too surprised that MFP growth dropped quite a lot in NZ in a period containing the GFC, for example, where management attention may well have been temporarily more focussed on survival issues than on productivity issues. On your point re possibly inaccurate measurement of new goods and versions, I agree that it's very much a work in progress. Stats NZ for example is doing good stuff eg in using hedonic regression to capture quality change in cars, but (a) that's only one product and (b) while I like the approach, I've personally been left with the impression (admittedly with no knock-out facts to support it) that it's still underestimating the scale of quality improvement. I like your blog by the way, and in a mo I'll have it added to the 'Sites I like' and 'Latest posts' sections here

  5. Thanks, a quick response
    David Andolfatto argues for the Schumpeterian view of economic development where the distinction between growth and business cycles is artificial. His macro textbook is on line and is a great read.
    Everyone agrees that long-run growth is the product of technological advancement.
    The New-Keynesian school views trend growth as being relatively stable
    In the Schumpeterian view, there is no reason to believe that the process of technological advancement proceeds in a smooth’ manner.
    It is more reasonable to suppose that new technologies appear in clusters.

    These technology shocks may cause fluctuations in the trend rate of growth through what Schumpeter called a process of ‘creative destruction.

    Technological advancements that ultimately lead to higher productivity may, in the short run, induce cyclical adjustments as the economy restructures: resources flow from declining sectors to expanding sectors, and people retrain and learn the next technologies and invest in the secondary innovations to make say computer of practical application

    The productivity slowdown is the 70s is attributed in this view to a doubling of technology adoption costs that were no measured as investment.

    Variations in technology diffusion is a big issue for Diego comin in cross-national differences in incomes.

    More later this week.