OECD, Growth, and Measurement: Philpott Revisited1

by

Robin Johnson2

This paper derives from Brian Easton's observation that the “knowledge wave” conference in Auckland in 2001 lacked any consideration of systematic quantitative estimates and forecasts with regard to economic growth. The paper looks at what is available in this regard, some of its shortcomings, and the broad lessons it demonstrates. The approach is essentially based on estimates of productivity growth using GDP factor productivity models. It is clear that long term productivity growth in this sense of around 1% per annum, as measured by Bryan Philpott in a series of path breaking papers, represents some kind of achievable growth rate given New Zealand's position with regard to international trade, the terms of trade and the exchange rate. This translates to an economy growth rate of around 2.7% per annum. The question is then asked whether greater attention to economic fundamentals is likely to be accepted by the government and the electorate?

Introduction

This conference has as its theme Back into the Top Half of the OECD: New Zealand's Long-run Economic Performance. A major conference was held in Auckland in 2001 to examine how the knowledge wave could contribute to future economic growth in this country. The conference was characterised by a wide ranging set of contributions which looked at everything except economic growth itself. As Brian Easton observed (The Listener Sept 1 2001):

As one Herald reporter wrote, the conference was ‘an act of faith’...the anecdotes and random data that littered the conference-inspired articles certainly maintained that tradition of a lack of systematic analysis. Reading the deluge, one can only conclude that, if New Zealand businesses operated on the same loose thinking, it would be no surprise if the economy was doing worse than it is.

There is a good literature of growth in New Zealand, the most important of which is the preparatory work done by Bryan Philpott first at Lincoln University and later at Victoria University. Most of the productivity ratios he estimated were based on simple average factor share weighted index numbers using the national income data base and his own estimates of labour equivalents and gross capital stock. Most of the other studies, referenced by Diewert and Lawrence (D & L), draw on this same stock of data. More refined methods of weighting were later introduced by D & L based on a new data set of their own and using chain-linked Fisher indexes. D & L also run an eye over the national income sectoral data base so we can make some direct comparisons with Philpott's results. At a national level, the result are very similar but at the industry level, there are discrepancies which require further investigation. In this paper I concentrate on the Philpott and Diewert results. Diewert and Lawrence review most of the NZ literature in their Ch 6.

The paper proceeds to make a number of observations and propositions about growth drawn from the literature and my own investigations with a background of selected statistical material which illustrate the points made and provide for the readers' scrutiny in their own time.

The Philpott Data Base

The data base contains national income information on a sectoral basis back to 1960. Gross output, intermediate inputs and value added are given in both nominal and real terms. In addition, sectoral employment on a full-time equivalent basis (from the Quarterly Employment Surveys) and levels of sectoral capital stocks are estimated. In two different publications, capital stocks are estimated on a gross and net basis. Gross stocks are based on estimated life times of the assets involved; net stocks are estimated with standard depreciation rates (Philpott 1994, 1995a, 1999a). The latter tend to be around two thirds of the former.

There are certain problems with the compilation of these statistics which are discussed below. Some of the individual sectoral series are subject to serious measurement errors and hence any results drawn therefrom must be seriously qualified.

Philpott advanced most of his productivity analysis on a simple version of the total factor productivity ratio (TFP). The ratio is realGDP/real(L+K) and the two indexes of labour and capital inputs are combined in a weighted average input according to their average shares of product in the period under analysis. (That the shares of product may vary considerably over a period of time is thus ignored). TFP thus measures the excess of product or value added that accumulates to society over and above the average rewards that go to the labour and capital factors of production if they were held constant.

The relationship of TFP to other growth rates is shown in Table 1. The measure of GDP is all production sectors (i.e. excluding government and owner occupied dwellings). Over the longer period 1960-98, real gdp increased by 2.7% per year, in terms of labour productivity it increased by 1.5% per year and in terms of factor productivity by 0.9% per year. In the sub periods 1960-84 and 1985-98 gdp growth was greater in the earlier period, but gdp growth per head was greater in the latter period. Total factor productivity was greater also in the latter period as capital input growth was less. Capital productivity has consistently declined over the whole period. The often quoted figure of 2.4% growth in real national income appears to relate to the 1980-98 period. In terms of the all groups totals, including government and dwellings, growth has been slightly lower.

Table 1: Philpott's Growth Parameters (% per year)
Real GDPGDP/FTE GDP/unitKGDP/L+K*
1960-84(market sector)3.0 1.5 -0.8 0.7
1985-98 2.2 1.8 -0.4 1.3
1960-98 2.8 1.6 -0.6 0.9
1960-98(all totals) 2.7 1.4 -0.6 0.7
Average factor shares weighting

On a sectoral basis, Philpott (1995b) found that TFP growth was greatest over the period 1960-94 in the agriculture, fishing and hunting, food and beverages, textiles, energy and communications sectors (Table 2). Low growth or negative growth was exhibited by the basic metal industries, chemicals, non-metallic mineral products, building and construction, trade and restaurants, and financial services sectors. Some reasons for the variations between sectors and time periods are discussed further below. More sophisticated weighting methods also appear to change the rankings of the different sectors as well.

 

Table 2: Philpott Sectoral Growth Rates (%per year)
Sector 1960-651965-70 1975-851985-94 1960-94
Agriculture 5.5 1.8 1.8 5.9 3.4
Fishing&Hunting-0.4 -2.5 4.9 9.1 3.1
Forestry and Logging -2.4 1.4 2.5 4.31.9
Mining&Quarrying 5.32.8 -6.3 7.61.8
Food&Beverages-0.81.27.6 1.83.0
Wood&Wood Products5.54.30.4-1.81.6
Paper&Printing4.73.01.9-0.91.9
Base Metals-1.2-1.3 -7.13.8-1.7
Textiles&Apparel 4.3 0.4 4.9 5.83.8
Chemicals etc-8.11.6-3.2 0.5-0.8
Non-Metallics-2.9-1.1-0.51.3-0.5
Machinery&Other1.63.0-2.5 0.30.5
Elect,Gas&Water4.73.83.52.63.4
Building&Construction2.3 0.7 0.2-0.90.3
Trade&Restaurants4.00.1-2.10.1 0.1
Transport&Storage0.71.00.94.61.8
Communications1.82.43.88.84.5
Finance,Insurance etc0.4-4.81.2-4.2-2.1
Pvt Services1.6-3.30.30.5-0.5
Total Market Economy 2.50.80.11.41.0
Source: Philpott (1995b)

 

The Role of Investment

There are some changes within sectors in Table 2 that defy economic logic. In some sectors there is a rapid transition from negative growth to positive growth (e.g. Chemicals in 1965-70). Given that capital per unit of net output has been expanding throughout the period, it is plausible that a short term focus (such as 5 year periods in Table 2) obscures industries where capital investment has been increasing relatively rapidly yet output was not responding in the short run (say, in mining or aluminium processing). In some other cases it is likely the data is inconsistent (e.g. forestry labour in 1987-88 and the subsequent spurt in productivity).

It is plausible that some sectors in the economy are capital hungry and do not pay a full return on the investment made even in the long run. In some sectors, capital investment was dominated by Government spending on social capital and infrastructure and hence returned low dividends in GDP terms. In other sectors (e.g. communications) there has been large investment and large returns in national income.

It would be normal to expect that private investors would not expand their business without reasonable prospects of getting their money back. There are risks in this, of course, and some businesses fail. However, in the case of public enterprises and services, social goals may be different to private enterprise, and the discipline of a payback period may not be enforced. Further, in some industries like forestry the payback period is over 30 years and even in farming it is generally considered to be about 4-5 years.

Examination of investment patterns across sectors reveals some industries always do better than others in terms of capital employed (Table 3). On the other hand, some industry groups are expensive in their use of capital. Is there a connection between the level of investment and the economic surplus generated? An economic surplus is defined as a level of GDP that is surplus to the normal costs of labour and capital employed (average factor shares).

The table is based on the accumulated capital assets "in use" in each industry sector (Philpott's gross capital stock series). New assets are added each year and assets deducted when they are no longer in use. When the total asset base in a sector declines, "wastage" exceeds replacements (Normal rates of depreciation are not used in the tables). There is then less total capital to enhance growth unless the results of previous investments have a delayed effect.

There are distinct differences in the two periods identified in Table 3. Before 1984, all sectors were investing in growth and expansion of output. Most rapid expansion was taking place in fishing, mining, forestry, paper, chemicals, basic metals, non-metallics, machinery, energy, transport, communications, finance and community services. They averaged a 400% increase in assets deployed over 21 years. The remaining six sectors only averaged a 66% increase.

But in the group of 13 high-investing sectors, six sectors were not generating an economic surplus over the 21 year period: mining, chemicals, basic metals, non-metallics, finance and community services. The trade and finance sectors also returned a negative surplus. These seven industries represent 42.3 % of market production GDP, hence they weighed down the other sectors of the economy in this period to a considerable degree.

The corollary of this is that the remaining 12 sectors were the only sectors generating economic growth as conventionally measured between 1962 and 1984.

In the second period, the annual rate of growth of investment in assets has slowed. At least six sectors (farming, forestry, textiles, non-metallics, building and transport) did not add to their capital stock. Some sectors were still expanding capital rapidly (fishing, chemicals, basic metals, communications, finance and community services). At least three of these sectors also had a negative economic surplus. Overall the extent of the negative surplusses was less pronounced though 8 sectors in all were still drawing on the economy rather contributing to it.

Most significant is the turnaround of some sectors from the pre-reform period. Mining, chemicals, basic metals and non-metallics all turned in an economic surplus; the aluminium smelter fits in here no doubt. Past investment is paying off in national income terms.

In the second period, the farm sector stopped full replacement of its capital stock in 1986; forestry in 1983; primary processing in 1991; building and construction in 1979; textiles in 1988, and transport in 1989. These industries have adjusted to the changing situation. In transport the results will reflect in part the high labour productivity gains obtained by rationalisation and de-manning in Tranz Rail.

Some of these structural changes can be attributed to Government policy: the withdrawal of support for farming, less support for forest plantings, roll-on effects on primary processing, the slow-down in the public building and construction programme (including public transport), the removal of import licensing, and the setting up of state enterprises. Most of the changes reflect the then Government's mission to get investment decisions made on a commercial basis.

The role of government policy is harder to discern in the investment-expanding sectors. Export incentives were withdrawn early in the reforms but tariff reforms have come much later. Textile investment has declined only since 1988/89 for example. The finance sector has grown considerably after the lifting of controls on interest rates, foreign exchange and foreign investment.

In the case of the continued negative surplusses in trade and community services, Diewert and Lawrence drew attention to the difficulty of measuring the outputs these sectors produce. The trade sector includes hotels and restaurants and the measures fail to adequately account for quality changes and the introduction of new services. In the community services sector the output measure is dependent on hours worked indicators rather than quality of service. These authors also express concerns about a possible over-statement of the capital stocks used in financial services.

The incentives for continuing high capital replacement come from expectations of higher returns in the future. If expectations are not so high then firms will cut their cloth accordingly, as farming, forestry, fishing and textiles have done since 1984. Investment in mining, chemicals, basic metals and communications has continued and paid a productive dividend. The service sectors all show high investment levels and negative surplusses. It is these sectors which pull the national average labour productivity measures downwards.

This data suggests that higher labour productivity is dependent upon the right investment at the right time in the right sector. This raises the question whether output per worker is improved by capital alone or other factors are at work? These might be a better trained work force (investment in education), the application of new scientific knowledge (investment in research), or better organisational techniques (as in Telecom). Some of capital investment in a sector may also incorporate facilities for better training, use of scientific advances, and organisational changes.

In summary,

Table 3: Sector investment rates and surplus generated
1962-831984-98
Sector
Share
of
GDP
%
increase
in
investment
%
increase
in
surplus
%
increase
in
investment
%
increase
in
surplus
Farming.080+30+101-4+106
Fishing.004+280+31+89-30
Forestry.011+50+21+7+98
Mining.012+395-39+59+142
Food.102+215+60+40+9
Wood.022+77+294+43-5
Paper.034+176+76+33+9
Textiles.034+84+371-1-1
Chemicals.037+262-60+141+25
Basic Metals.014+2477-75+104+52
Non-metallics.014+190-90+2+30
Machinery.088+164+13+104-7
Energy.031+132+114+20 +62
Building&Const..097+78+15-11-3
Transport.058+179+13-1+60
Trade&Restaurants.215+79-19+47-6
Communications.017+186+122+146+139
Finance and Insur..085+350-72+150-31
Community Serv..046+108-48+130-1
All production sectors1.00+152+12+44 +19
All sectors - +111+6+41+15
Sources: Economic surpluses are derived from Philpott (1994) and Ch 5 in Diewert and Lawrence (1999); Capital Stocks are derived from Philpott (1994,1999a)

Weighted Index Numbers

The Diewert and Lawrence Report, commissioned by Treasury, the Reserve Bank and the Dept of Labour, employs more sophisticated index number methodology than Bryan Philpott but does not modify his basic conclusions about growth in the national economy. D&L make several changes to the analysis. First they define TFP in a different way; they estimate their index numbers by more sopisticated methods; they define capital stocks in a different way, and use data on hours worked (instead of labour equivalents) for their "labour" index. Their modification of the capital stock series tends to overstate the size of the aggregate capital stock according to Philpott (1999b).

D&L approach the measurement of TFP in two different ways. When D&L are using their own data base they actually estimate a productivity ratio based on the expenditure approach to national income. When they use the official data base, their productivity ratios are based on production national income measures. The advantage of the latter is that the national accounts can be divided into the individual product sectors.

D&L use chain linked index numbers based on the Fisher ideal index; for each period the previous period's observation is defined as the base (Report, pp.10-11). They state that the Tornqvist index can also be used as it closely approximates the Fisher index (ibid. p.10). These methodologies clearly overcome the weighting problem created by using factor weights that get out of date.

Table 4: Major Diewert and Lawrence TFP Results (%growth per year)
1972-98
Expenditure Total Output1.6
Expenditure Total Input 1.3
Expenditure TFP0.4
Labour partial productivity 1.0
Capital partial productivity -0.9
1972-97
Australian Total Output 2.2
Australian Total Input 1.2
Australian TFP 0.9
1978-98
Production Total Output 2.3
Production Total Input1.3
Production TFP 0.8
1975-94
Philpott average shares TFP 0.8
Sources: D&L 1999; Philpott 1995b

Table 4 shows a selection of results from the expenditure and production methodology compared with Australian and Philpott's main result. In the aggregate, the expenditure method gives a lower estimate of TFP than the production method; the production method gives a similar result to that estimated by Philpott. New Zealand's rate of productivity growth appears to be lower than that of Australia.

On a sectoral basis exact comparisons are not possible because of different time periods but Table 5 shows a comparison of Philpott's estimates on a sectoral basis for the period 1975-94 with the D&L estimates for 1978-98.

Table 5: Sectoral total factor productivity(%growth per year)
SectorPhilpott 1975-94 D&L 1978-98
Agriculture3.9 3.9
Fishing and hunting7.0 0.3
Forestry and logging 3.4 6.3
Mining and quarrying0.74.9
Food and beverages4.70.7
Wood and wood products -0.7 0.3
Paper and printing 0.51.3
Basic Metals-1.6 1.0
Textiles and Apparel 5.4 0.2
Chemicals-1.4 0.3
Non-Metallics0.42.4
Machinery( -1.1 0.1
Other(2.4
Energy 3.13.5
Building and Construction-0.40.6
Trade and Restaurants-1.0-0.7
Transport and Storage2.8 3.8
Communications6.36.7
Finance etc-1.5-2.1
Private Services0.40.1
Aggregate0.80.8
Source: Philpott 1995b: weighted average of 1975-85 and 1985-94; D&L 1999 p. 76-77.

While the weighted aggregate estimate of TFP is the same from both studies, there are wide discrepancies in the results for the majority of individual sectors. Only agriculture, forestry, paper, energy, transport and communications and finance appear to be drawn from the same population. Without sight of the original calculations it is impossible to make a reconciliation here. Box 1 shows some of the reasons given for the variation in results. Diewert and Lawrence (1999 p.132) set out the case for giving Statistics NZ more resources to remedy these and other matters. Philpott (1999b) gives a cogent criticism of the perpetual inventory methodology used by Diewert and Lawrence. Easton (1998) discusses the same data from the point of view of labour productivity.

Easton (1999) critiques the data in D&L and draws attention to differences in the deflators used for capital expenditure This has effects back on their expenditure definition of GDP.

Box 1: Identified measurement differences in official data base between Philpott and Diewert and Lawrence (as identified by the author and others)
(list not complete but indicative)

The Role of Research and Development

The author has presented two papers to this association on the relationships between investment in Research and Development and real national income (Johnson 1999, 2000). This exercise involved developing a time series of provider expenditure (universities, private and government) on R&D back to the 1960s extrapolating from the Morst surveys which started in 1989, and then relating sectoral R&D expenditure to sectoral factor productivity changes. Intuitively, some of the Solow residual productivity increase is explained by better production methods, new products and better organisation of resources. Behind such production changes sit opportunities opened up by current and past scientific effort. Can any direct line of causation be traced from the expenditure on R&D effort and consequent increases in factor productivity? Two hypotheses were developed: one using R&D expenditure to develop a "stock" of accumulated knowledge at cost; the other using econometric methods to analyse past patterns in annual R&D expenditure. In the light of policy positions at the time, "private" sector expenditure and stocks were examined separately from "public" expenditure (universities and government).

In the first paper, Research and Development was seen as a stock variable in a capital sense and annual expenditure was converted to an accumulated stock allowing for natural wastage of ideas. In the statistical model, changes in sectoral factor productivity were related to private and public stocks of R&D of a year earlier, educational attainment and a variable representing overseas availability of R&D a year earlier. The results showed that:

Thus only about half the sectors showed some relationship between Research and Development as defined and factor productivity. The other half showed negative or non-significant coefficients. This seems to tell us that only sectors with a good buildup of background research effort can demonstrate a statistical linkage between R&D effort and productivity increase. I believe this is a measurement problem as only the major sectoral categories (agriculture, processing , manufacturing, building, and energy) have a clear historical record. Put another way there is a mismatch between the industrial classification used by Morst and the industrial classification in the gdp data. I cannot explain why changes in Australian stocks of R&D can have any influence on sectoral productivity in NZ.

In the second paper, a statistical model was developed to explore lagged responses to investment in Research and Development. The model posits that current sectoral TFP is more influenced by recent annual expenditure on R&D that that incurred in earlier years. The test takes the form of a polynomial distributed lag (PDL). The results obtained showed no clear relationships between patterns of past annual expenditure on public and private R&D and TFP. In three sectors (forestry, manufacturing and transport) the estimated coefficients were negative thus indicating a perverse movement between the two variables. There was suggestion that in some sectors there was response to a lag of 5-6 years earlier and then a gradual decline out to 11-12 years. It is a possibility that there is a cycle of R&D investment in NZ with higher rates of gdp growth in a sector encouraging higher immediate investment which is followed by quite a delay in obtaining the next lot of benefits. Further research would be worthwhile.

Intuitively, we can appreciate that technology does have a relationship to the annual productivity surplus in TFP terms, but the kind of aggregate data available in NZ does not allow us to to be specific in measurement terms. Individual examples like per cow milk production are easy to document, but the total social return to national investment in Research and Development is more difficult to demonstrate. More research is needed at a lower level of aggregation to identify practices and managerial decisions that affect productivity. Earlier work in this paper showed that most sectoral productivity increases were generated by heavy investment in land and building and equipment. R&D investment will be hidden in such investment programmes and would need to be teased out by micro studies if their separate effects were to be measured.

Discussion

1. Quality of statistics: Some indication has been given of measurement errors in the national income statistics. Some of these may have been rectified in recent years. Although the production GDP data allows a disaggregation to production sectors, the expenditure measure of GDP probably gives a more accurate measure of changes in net output. [But Easton 1999 points out that the investment series is overstated] As D&L point out (p.94) the final demand components of GDP are more easily measured than the production components and also have a better set of deflators. In the production data, there is a lack of good deflators for gross output and intermediate inputs for many but not all sectors. It is only in the farming, electricity distribution, gas distribution and parts of the transportation sector that have real GDP measures estimated by the double deflation method. Incidentily, those sectors with double-deflation accounts also lend themselves to weighted index number measures of sectoral productivity.

2. Measurement errors: Since sectoral productivity estimates have to be based on the production accounts and other measures of labour and capital, they are subject to considerable measurement error. There is therefore not a lot to gain by using sophisticated weighting methods like those used by D&L as compared with the simple methods used by Bryan Philpott. The two methods give some quite different results for different production sectors which need to be resolved by constructing a standard data set. Easton (1998) also doubts the compatibility of the labour and capital estimates.

3. The role of investment: D&L/Jannsen (p.91) note that growth accounting is not a theory of economic growth; it simply tells us what has happened, not why it happened. It does elucidate whether growth occurs on the capital or labour side and how much capital has been needed in the past. This study shows that a lack of investment means no growth at all. Hence further analysis must start with an examination of the incentives to invest in productive activities. This then leads on to whether private or public investment is more effective (from a growth point of view) and what are the conditions for optimum investment?

4. The terms of trade: There is some debate over the role of the terms of trade. D&L (pp.32-3) decompose changes in nominal GDP into effects due to changes in the terms of trade, growth in labour and capital endowments. productivity change and changes in non-traded goods prices. Their analysis indicates that changes in the terms of trade had a relatively minor minor impact on real GDP; productivity change was the largest contributor, all else unchanged, followed by increases in the economy's capital stock, and increases in labour endowments. This appears to overlook the investment effect of changes in the terms of trade on the tradable sector. Nominal incomes fall when the terms of trade declines, producers seek to maintain nominal incomes, and ration capital investment including maintenance of assets. Growth almost ceases until the cycle starts again.

5. Piggy backing: There remains the possibility that some technology is piggybacked in on high levels of capital investment while some may be purely innovative and less demanding of new capital. Solow type productivity is a grab bag of everything left over after measured labour and capital and includes organisational changes, new products, new inputs, better educated and trained workers, as well as more effective investment. The task is to unravel these.

6. Necessary conditions: Investment thrives in an economy where stability is protected, compliance costs are under control, profitability can be foreseen, and public intervention is at a minimum.

These are the kind of answers that the likes of Scobie (2002) have recently presented.

He identifies the basics as:

7. Summary: It seems to me that the government and the electorate prefer stability and mediocrity to growth and displacement. It is surprising that the Treasury website can still emphasise the economic solution rather than the social preference solution. I think we are likely to continue much as before and that New Zealand's particular mix of commodity production, exchange rates and government intervention will dictate the outcome. This also appears to be the view of NZIER (2002) who recently forecast economic growth across all industries to average 2.6% per annum between March years 2002 and 2006.


References

Diewert E. and Lawrence D. (1999), Measuring New Zealand's Productivity, Treasury Working Paper 99/5.

Easton, B. (1998), Microeconomic Reform: the New Zealand Experience, Macroeconomic Reform and Productivity Growth Workshop Proceedings, Productivity Commission and ANU, pp.155-181.

Easton, B. (1999), Measuring New Zealand's Economic Activity, A Report to the Dept of Labour based on Measuring New Zealand's Productivity (D&L Report).

Johnson, R.W.M. (1999), The Rate of Return to New Zealand Research and Development Investment, New Zealand Association of Economists Annual Conference, Rotorua (www.geocities.com/rwmj2001).

Johnson R.W.M. (2000), Methodologies for Measuring the Accumulated Knowledge Base in Research and Development, New Zealand Association of Economists Annual Conference, Wellington (www.geocities.com/rwmj2001).

MoRST (Var), NZ Research and Experimental Development Statistics: All Sectors, Ministry of Research, Science and Technology.

NZIER (2002), New Zealand Industries and regions: outlook and issues to 2006, Press release on 24.04.02 (www.nzier.co.nz).

Philpott B.P. (1994), Data Base of Nominal and Real Output, Labour, and Capital Employed by SNA Industry group 1960-1990, RPEP Paper 265, Victoria University.

Philpott B.P. (1995a), Real Net Capital Stock by SNA Production Groups New Zealand 1950-1991, RPEP Paper 270.

Philpott B.P. (1995b), New Zealand's Aggregate and Sectoral Productivity Growth 1960-95, RPEP Paper 274.

Philpott B.P. (1999a), Provisional Estimates for 1990-1998 of Output, Labour and Capital Employed by SNA Industry Group, RPEP Paper 293.

Philpott B.P. (1999b), Deficiencies in Diewert-Lawrence Capital Stock Estimates, RPEP Paper 294.

Scobie G. (2002), We'd like better growth? Then how about some back to basics? (www.treasury.govt.nz/et/#8).

Footnotes

1) Paper prepared for Annual Conference of the NZ Association of Economists, 26-28 June 2002, James Cook Hotel, Wellington.

2) Private Consultant, Wellington (johnsonr@clear.net.nz).

Print this page