Some Notes on New Zealand Economic Growth Research

Revised Paper for a Seminar, 15 May 2008

Keywords: Growth & Innovation; Statistics;

I have been asked to make a few remarks on research into economic growth in New Zealand. Let me begin with some preliminaries.


I am going to avoid suggesting any policy conclusions which might follow from the research, because policy oriented research is almost invariably bad research, when the concern with policy obscures the insights which research is able to offer. My research strategy has always been to pursue an understanding of the phenomenon, without concern to what policy conclusions it may generate. When I have that understanding I begin to think about the research’s policy implications. .

I shall, however, draw attention to places where I think the policy conclusions do not, or cannot follow, from the research.


It is striking how often in economic research in New Zealand, data is taken at face value, with no consideration of its quality. Data preparation and assessment is a time consuming but necessary part of empirical research. Moser’s Law is one of the most fundamental laws of social research: if a statistic looks interesting it is probably wrong. The law’s research conclusion is that one should check to see whether the interesting data is correct (if it is it, the statistic is really interesting). Sadly, I regularly see research reports which do not discuss the quality of the data and often do not even bother to properly define it.

As an example, consider the OECD’s measure of what it calls ‘GDP at purchasing power parity prices’. GDP is a measure of output, but the statistic is actually measured on the expenditure side not the production side (aggregating total expenditure outlays and making an adjustment for the balance of payments). In fact the measure is of GDI.

Students are taught that they GDP equals GDI, but that only applies if the same prices are used. New Zealand economists are aware of this distinction, because once upon a time the Government Statistician had to go before the Court of Arbitration and explain that the growth in GNP was not the same as the growth of GNI (he called it ‘effective GNP’) if there was a change in the terms of trade, and so the production side prices no longer corresponded to the expenditure side prices.

Aside that purchasing power parity prices are not necessarily consistent between expenditure and output prices, the balance of payments adjustment presents another complication. Internationally traded products are valued at international prices in the balance of payments and domestic prices in the expenditure. The two prices can vary greatly where there is substantial domestic protection. Consider butter, which is valued in New Zealand’s balance of payments at its internationally traded price. But in the expenditure estimates it is valued at the average domestic price, which is higher because of protection. The export country (e.g. New Zealand) has a relatively lower GDP at PPP measured the OECD than if all its produce was valued at the expenditure price. In contrast the country which protects has a higher relative GDP. (This is a measurement phenomenon. It says nothing about the allocative gains and losses of protection.)

What is happening is the OECD is not measuring output but income (or as the Government Statistician would have once put it, ‘effective output). Thus the research which purports to be looking at production (and productivity) is actually looking at income. One of the explanation as to why countries such as New Zealand have lower income (that is appears to have lower production) is because their exports are discriminated against.

Surely this should be taken into account when GDP rankings are interpreted as measures of production (or productivity). Instead researchers have been using income as if it were production without being aware of it, which gives one little confidence in the research.

Another problem of the OECD GDI at PPP statistics are they dont track properly over time. Suppose you have them for two different years. Then the change ought to be equal to the growth in GDP. Right? Wrong! Even when there is an adjustment for terms of trade (that is they are treated as GDI), the measured year-to-year changes dont track the endpoint years very well.

In other words the per capita GDP statistics are open to some wide and perplexing errors. Use them with caution.


Research needs to go beyond arithmetic. If someone said that in order to arrive earlier at one’s destination, all you had to do was to go faster, we would not be impressed. Speed equals distance divided by time, so the proposition is a tautology. And so is the proposition that to get more output with the same resources, we need higher productivity. Yet the proposition is repeatedly said as if it a serious contribution to growth analysis.

Much of the growth research suffers from the failure to recognise the tautology. Its argument tends to be circular, which is why much of the research does not progress, but goes round and round. So, perhaps in frustration , the researcher leaps off into a hypothesis which has the most tenuous connection with the tautology. I am often struck that one can clip off the first third or so of many research papers on growth, and not make an iota of difference to the analytic argument.

What the research should be doing is finding an explanation for the speed of the car, or the level of productivity. It is not sufficient to find a variable which correlates with productivity. Correlation does not prove causation. What we need is a comprehensive account which explains all (or most) of any productivity differences. Most of the difference has to be explained: low r-square bivariate cross-national comparisons does not count as useful explanations. At best they are suggestive.

What is needed instead of tautologies is the addition of theories which may be wrong and can be empirically tested.

That is why multi-factor productivity is not a tautology in the way that, say, average labour productivity is. To calculate it, one has to add in a neoclassical pricing assumption. Whether the assumption is correct is a matter for reflection. Experience shows that the choice of weightings for the contribution of capital and labour does not matter much. I worry though that the weightings may change over time and that the MFP with fixed weights may be misleading.

In any case, as Bob Solow understood right at the beginning, MFP is a residual; it is what has not been explained, after growth of capital and labour has been allowed for. Tommy Balogh and Paul Streeten provocatively suggested that it be called the ‘coefficient of ignorance’. When someone argues that MFP should be increased, they are usually advocating we should increase the coefficient of ignorance. Sometimes I think New Zealand is well placed to do that.

The obvious research strategy is to reduce the coefficient of ignorance. For instance rather than numbers of workers we might use hours worked and allow for the increasing quality of the workers. That makes some progress, but it still leaves a very large residual.

What I think is happening, is that the approach is crudely measuring technological change. I’ve written in a number of places about this and so I wont go through that here. But if one really thinks technology is important, one cant simply assume that knowing MFP is enough, or that because technology is a good thing, we dont have to think about it carefully.

Aggregate technology is not a very productive notion, and we have to think about it in more microeconomic terms. Practically that means sectors, although we may have to go below them to processes and products. Otherwise, the aggregate accounts of the growth process are going to be almost entirely tautologous.

The Mason-Osborne Research
I want to illustrate some of these issues by considering the paper by Geoff Mason and Matthew Osborne on Productivity, Capital Intensity and Labour Quality at Sector Level in New Zealand and the UK. (Treasury Working Paper 07/01). They compared the British and New Zealand productivity performances for 19 market sectors. (The have little to say about the non-market sector.) This requires pricing each sector’s outputs and inputs in common prices.

Table 1: Productivity Comparisons for Market Sectors

<>% share of NZ value added
(UK = 100)
(UK = 100)
(UK = 100)
<>Finance and insurance
<>Cultural & recreational services
<>Agricultural, forestry & fishing
<>Accommodation, restaurants & bars
<>Food beverage & tobacco mfg
<>Business services
<>Metal product manufacturing
<>Wholesale trade
<>Transport & storage
<>Average: all market sectors
<>Electricity , gas & water supply
<>Machinery & equipment mfg
<>Communications Services
<>Retail trade
<> 7.8
<> 5.8
<>Non-metallic mineral product mfg
<>Wood & paper product mfg
<>Furniture & other mfg
<>Textile & apparel mfg
<>Petroleum, chemical, plastic & rubber products mfg                                               
<>Printing, publishing & recorded media

From Mason & Osborne (2007) Tables 5, 7, 10 (KLP constructed from Tables 5 and 7)

Table 1 shows some of their results. Before commenting on its implications, – and as I counselled earlier – we need to consider the accuracy of the data.

International data comparisons are always difficult, but this study also depends upon the assumption that the ratio of value-added to output is the same in New Zealand and Britain for each sector. To be certain, that requires that composition of the sectors are the same, whereas often they are not. There is not a lot of armaments production in New Zealand’s machinery and equipment manufacturing industry; New Zealand’s electricity power generation is more hydro and less thermal than Britain’s.

Moreover the prices of the inputs relative to the outputs have to be the same. Because of different real exchange rates, this assumption seems unlikely for import inputs.

Some might argue that these problems are so large that the data is of little value, but that would be equally true for estimates of aggregate national productivity. Instead proceed with caution.


Focusing on the ‘multi-factor-productivity’ measure, which is adjusted for capital, labour and the quality of the labour force, one is struck by the enormous variation in the relative productivities. It is almost five times between the 38% relativity of the printing, publishing and recorded media sector, and the 180% of the mining sector. Even if the ratio were half that, there remains a salutary implication for aggregate productivity studies. It appears that different sectors perform at different rates of efficiency. Presumably our explanation of overall performance – and, indeed, any policy conclusions that might be drawn – needs to take these differences into account.

We have to be careful when we focus on particular sectors. The relative productivity differences may reflect measurement problems. One of the lessons – research questions – is that three sectors, construction, total distribution (transport, communication and distribution1 ) and manufacturing seem to be sufficiently problematic and large enough to be important in New Zealand’s overall productivity performance.

Subject to these caveats, it would appear that 73 percent of the productivity deficit can be explained by the weaker performance of the manufacturing sector, 38 percent by the total distribution sector, and 27 percent by the construction sector. (They sum to more than 100 percent because there are some New Zealand sectors which are productivity stronger than their British equivalent.)

Composition of Output

Suppose we used New Zealand production methods to produce Britain’s output. It turns out they would require 20 percent more resources. This should not surprise us, because we believe Britain to use its resources more productively that New Zealand.

What about the other way around? Suppose we use British production methods to produce the New Zealand output. As it happens, New Zealand would use only 1 percent less resources if it used British processes and technologies, so the British productivity seems to be much the same as the New Zealand productivity, when it comes to producing New Zealand output.

What seems to be happening is that there is a sense that while there are variations in the productivities between sectors, the Brits are better at deploying their resources in sectors where they are highly productive, rather than in sectors where there relative productivity is lower. At the very least this exercise shows sectoral composition matters.

Factor Intensities

We observe that average labour productivity is generally lower in New Zealand than in Britain, whereas average capital intensity is somewhat higher. We need to be careful with such conclusions, because particularly capital is difficult to measure (including the relative price assumptions may be wrong because of differences in real exchange rates).

Average productivity is not the same as marginal productivity, which is the relevant variable for investors. Even so, the estimates raise the question as to why there is not more investment in New Zealand.

Comparing the Mason-Osborne Productivity Estimate with the OECD Estimate

I’ve left the checking of the data to last. The OECD reports per capita GDP at purchasing power parity is 113% of the OECD average for Britain and 86% for New Zealand. So the gap between the two economies on a per capita basis is 100 to 76. How does this compare to the Mason/Osborne estimates of Britain being ahead 100 to 77 on a per hourly labour productivity basis in market sectors?

To answer this we first need to adjust the Mason-Osborne estimate for differences in average annual hours worked per person. Then we need to add the non-market sectors to the market sectors considered by Mason and Osborne.

The Brits work fewer hours per person than New Zealanders – I calculate about 10 percent less.2 Adjusting the Mason-Osborne figures for the differences in hours worked relative to the population we get the ratio changes to 100 versus 69 per hour worked in the non-market sector.

How to include the non-market economy? There are a couple of simple assumptions that could be made. The first is that the relative productivities of the non-market sector are the same as the market ones, in which case the Mason-Osborne ratio remains at 100 to 69.

An alternative assumption would be that the relative productivity of the non-market sectors in the two countries are the same. This arises from the artificial way that the OECD measures their output at PPP. Non-market production is measured as an input at PPP, so that all OECD economies have the same unit productivity in the non-market sector.

At the time the non-market sectors contributed 22 percent to GDP. Including them the Mason-Osborne relativity becomes 100 to 77, after adjusting for hourly differences and including the non-market sector. That ratio is much the same as the OECD figure based on comparable definitions.

So the Mason-Osborne estimates are not inconsistent with the OECD figures. From one perspective this is not entirely surprising, since they are based on the same data base. From another, the final match is much closer than one might have expected, since one is based on the production side, the other on the expenditure side.

Existing Research

It will be clear that the general points I have been talking about apply to much of the existing research program to some degree. It therefore unnecessary to evaluate every paper. But I want to make a few over-view points.

The empirical research I have drawn upon suggests that major factors causing New Zealand’s lower productivity performance are international protection against its exports, problems in three key sectors: logistics-distribution and manufacturing and the balance of our sectors. Too much of New Zealand’s research is about aggregate productivity, and draws policy conclusions based on that indicator with little – if any – attention to the sectoral differences. It seems likely that unless they address these issues, any conclusions drawn from the research must be at best limited and inefficient, if not darned wrong. Or if the conclusions are right, they are not supported by the research that purports to underpin them.

Second, suppose someone proved that the New Zealand per capita was in the top half of the OECD. Would that change any policy recommendation that is in the research. My impression is that it would not; that one would pursue exactly the same policies. So what has the research taught policy makers that they would not have known anyway?
(Although it is not this paper’s purpose to pursue policy conclusions, I might point out that this short paper raises a policy issue which has not been addressed for two decades: to what extent should we be trying to get out of low (relative) productivity sectors. I mention also that it has alerted me much more than in the past to the capital/investment problems that New Zealand faces.)

Third, there is a disconnect in many of the papers between the diagnosis of the poor economic performance, and the proposed policy treatment. The treatment is assumed to be effective with the most causal evidence that it is related to the diagnosis. Too often it amounts to that on some vague indicator purporting to have some relevance to some fuzzy concept, New Zealand is ranked in the bottom half. It is not enough to claim that the (selected) evidence is not inconsistent with the hypothesis. That is an extremely weak test, and contributes little to understanding. What we really need is a quantitative estimate of how much of any productivity gap is due to a particular phenomenon.

My final point here is that there is over-emphasis on aggregation. I would have thought that the aggregate analysis was exhausted a decade ago – longer if you take Bryan Philpott’s work into consideration. The view that you can discuss the New Zealand economy on the basis that it produces a single commodity is – as I argued in my In Stormy Seas – first year economics. Any familiarity with the New Zealand economy would suggest a single commodity approach is a nonsense. For instance a single commodity economy needs either no exports or no imports. (It needs one if it is borrowing or lending overseas.) No wonder we cock up tradeable sector policy.

There are hints in the above about how the research program can be improved. However, had I the time to do more proper research I would look at the following issues.

1. I would try to improve the data. We need to pay particular attention to measurement in the more problematic sectors (including some of the high relative productivity ones which might also have surprised one). My understanding is that the international research frontier is also moving, and that it may be possible to adjust for the different relative prices of inputs. I worry greatly about some of the capital measures, because their prices do not allow for the importance of imports and how differences in real exchange rates can change magnitude. On the issue of labour measurement I note that the Szeto-McLoughlin, the Máre-Hyslop and the Mason-Osborne papers are progressing on an important front because they are looking at a more disaggregated level.

2. We need more international comparisons. Comparing productivity levels between countries and trying to identify reasons differences is a standard scientific method. The Mason-Osborne paper is a start. Since Australia is our main comparator, we need a direct comparison with Australia.

3. I have indicated that there is a need for more sectoral analysis. I identified three sectoral groups which appear problematic and need further attention. It may be there are problems of measurement error. Sometimes there may be little we can do, but it would be good to know that more precisely: there may be in the case of the distribution sector given the population density and physical configuration of the country.

4. We also need to pay more attention to the international interface. I would like to get a precise estimate of the effects of international protection on the relativities (which is different from the allocative effect). It also ought to be possible to say more about the costs of distance on the direct return to exporters and hence of their productivity.

5. I am supportive of research on regions, with its current focus on Auckland. That would lead to more attention to infrastructure and the costs of distance.

6. At the more theoretical level, it would be nice to be able to say something more systematic about the relationship between MFP and production technologies, than the current hand-waving does And we need to know more about how the tradeable sector functions relative to the non-tradeable sector, especially in terms of price and profitability signals.

This is a research program which would keep me going for a while. It would certainly increase our understanding of New Zealand’s economic performance, although it no doubt it would suggest some new research lines. And no doubt, after we had improved our understanding it, some original policy insights might occur to us.

I note, however, this research program does not address the growth of productivity. I accept that is a different one, but it needs to be founded on differences in productivity levels. Indeed we may find that New Zealand productivity levels generally grow at about the same rate as the rest of the world except when they experience major external shocks: that was the broad conclusion of In Stormy Seas. In so far as that is true, it does not invalidate existing policy conclusions. Rather policy is about ensuring we keep up with the rest of the world, which in fact looks very much the way I see today’s policy advocacy.

Go to top