THE MEASURE OF GDP PER CAPITA IN PURCHASING POWER ...



[pic]

The measure of GDP per capita in purchasing power standards (PPS):

A statistical indicator tricky to interpret

Francois Magnien, INSEE, France[1]

Note: This paper uses solely European data. The author unfortunately had neither the time nor the means to extend the analysis to all the OECD countries, which it would be interesting to do.

The comparison, for a given period, of the real GDPs of OECD countries, i.e. in purchasing power standards (PPS), is a very important indicator among the wide range of statistical indicators available. It is an important analytical tool in the OECD, and is watched closely by policy-makers seeking elements of comparison of the level of economic development of Member countries, and even of economic policy in the European Union with regard to the allocation of structural funds. It is thus essential that estimates be reliable. It is however very difficult to gauge their reliability given the complexity of purchasing power parity (PPP) calculations (see Box 1). PPPs make it possible to deflate the value of the aggregates being compared. Very roughly, the total value of the consumption of a product that was twice as expensive in France as in Germany would be halved in the comparison of the real GDPs of the two countries. Fundamental to the calculation of PPPs is the collection of the prices, in the countries concerned, of all the products that constitute the various uses of GDP (essentially consumer and investment goods).

When the comparison of GDPs in PPS is repeated for at least two periods, it is possible to infer the relative growth rates for different countries between the two periods. However, an alternative reliable estimate of real GDP per capita growth exists -- that provided by national accounts. By using the growth figures published annually by the national statistical institutes of the OECD countries, it is possible to calculate real GDP per capita growth in each of the countries concerned, over the same period.

The comparison of the estimates of relative growth rates provided by national accounts and the International Comparison Programme (ICP) gives information on the limits of the comparison of real GDP per capita in PPS in Europe (Section 1 of this paper). This conclusion ties in with that presented by Danish statisticians at the last IARIW Conference. Comparison with non-European countries is not covered for lack of time, but the problems posed are probably the same. Examination of construction prices sheds light on the nature of the difficulties (Section 2). It is essential that the method of calculating PPPs address these problems if the credibility of the ICP is to be fully restored (Section 3). It’s essential that statisticians of Member States and of international organizations should work together in order to improve those indicators.

1. PPP calculations are not coherent with national accounts

1.1 A fundamental inconsistency in the measurement of real growth…

The European results are presented in Table 1 below. The first part of the table (column 1) shows population growth over the period 1992-1999 compared with the European Union average. The second part (columns 2 and 3) show the trend of GDP and real GDP per capita compared with the EU average according to national accounts sources. The figures are total growth rates over the period 1992-1999 at constant domestic prices. The third part of the table (columns 4 and 5) shows the same calculations, but for real GDP per capita in PPS terms. The fourth part of the table (the last column) shows, for each country, the difference between its relative performance as measured by the usual national accounts figures, and its relative performance measured in terms of PPS.

Table 1: real GDP per capita growth relative to the European average 1992-1999

(% over the seven-year period)

|  |Population |National accounts ( 1995 prices) |Source Eurostat (in PPP) |Difference|

|  |Population |GDP growth |GDP growth per capita (1) |GDP growth |GDP growth per |(2) - (1) |

| | | | | |capita (2) | |

|Belgium |-0.4 |-0.4 |-0.1 |-3.5 |-3.2 |-3.1 |

|Denmark |0.8 |4.1 |3.3 |13.1 |12.2 |8.6 |

|Germany |-0.3 |-5.1 |-4.9 |-2.7 |-2.4 |2.6 |

|Greece |-0.1 |0.9 |1.0 |8.8 |9.0 |7.9 |

|Spain |-0.5 |4.7 |5.2 |3.7 |4.2 |-0.9 |

|France |0.5 |-2.1 |-2.6 |-8.0 |-8.4 |-6.0 |

|Ireland |3.3 |49.1 |44.3 |44.4 |39.7 |-3.2 |

|Italy |-0.7 |-3.4 |-2.7 |-2.4 |-1.6 |1.1 |

|Luxembourg |8.7 |NA |NA |23.9 |14.0 |NA |

|Netherlands |2.0 |7.1 |5.1 |12.0 |9.8 |4.5 |

|Austria |0.1 |0.3 |0.1 |2.4 |2.3 |2.2 |

|Portugal |-0.8 |4.4 |5.3 |10.0 |10.9 |5.3 |

|Finland |0.3 |12.2 |11.9 |16.2 |15.8 |3.5 |

|Sweden |0.0 |NA |NA |0.5 |0.4 |NA |

|United Kingdom |0.3 |7.0 |6.7 |3.3 |3.0 |-3.5 |

| | | |  |  | |  |

| European Union |0.0 |0.0 |0.0 |0.0 |0.0 |0.0 |

|Source : Eurostat | | | | | |

Both estimates of the wealth produced are in volume, and unaffected by inflation. If there were no structural effects, they should be equal (see Box 2). It may be observed that, for many countries, there are large differences: + 8.6 for Denmark, + 7.9 of Greece, - 6.0 for France, + 4.5 for the Netherlands, - 3.5 for the United Kingdom. The sign of the difference varies: for example, for countries like Denmark, Germany and the Netherlands, the relative growth shown by the national accounts is well below that measured implicitly by the comparison of GDP in PPS. In contrast, other countries -- France, the United Kingdom, Belgium and Ireland -- would rank higher in the ranking of countries by GDP per capita if it were based on the 1992 ranking extrapolated by the measure of real GDP growth provided by national accounts. It may also be observed (Magnien 2002) that the difference is usually the same from one year to the next: structurally, depending on countries, growth is implicitly over- or understated in PPP calculations compared with national accounts.

Without a detailed investigation of its nature[2], the difference between the two measures of real GDP per capita growth is too large for it to be explained solely by the structural effect. Furthermore, the well-known non-robustness (Boxes 3 and 4) of the measure of levels (the case of the ICP) compared with that of trends (national accounts) lessens the credibility of the ICP.

1.2 stemming from differing estimates of price levels in national accounts and PPP calculations

The difference observed between the two measures of annual real GDP per capita growth reflects a profound divergence between estimates of the temporal trend of prices in national accounts and PPP calculations. There is no divergence on the measure of GDP in value terms: both measures agree on this starting point. If the measure in PPS suggests a different level of real growth, this is because it is based on a different estimate of relative price levels during the period. The price trends in each country relative to the European average are shown in Table 2.

Table 2: National price trends relative to the European average between 1992 and 1999: comparison of Eurostat and national accounts measures (in %)

|  |National |PPP Estimate | |Difference |

| |accounts |Eurostat | | |

|Belgium |1.2 |4.5 | |-3.1 |

|Denmark |3.6 |-4.6 | |8.6 |

|Germany |0.2 |-2.4 | |2.6 |

|Greece |14.0 |5.7 | |7.9 |

|Spain |NA |-11.7 | |NA |

|France |-0.4 |6.0 | |-6.0 |

|Ireland |8.3 |11.9 | |-3.2 |

|Italy |-9.4 |-10.3 | |1.1 |

|Luxembourg |NA |NA | |NA |

|Netherlands |1.4 |-3.0 | |4.5 |

|Austria |0.3 |-1.8 | |2.2 |

|Portugal |3.2 |-2.0 | |5.3 |

|Finland |-3.8 |-7.0 | |3.5 |

|Sweden |NA |NA | |NA |

|United Kingdom |16.1 |20.2 | |-3.5 |

| | | | | |

| European Union |0.0 |0.0 | |0.0 |

|Source : Eurostat | | | |

The first column shows the difference between GDP inflation measured by national statistical institutes, and the European average, adjusted, obviously, for exchange rate changes[3]. The second column shows the same estimate based on PPPs. The last column corresponds to the difference between the two estimates. By construction, it matches the last column in the previous table. By virtue of the method, the differences in the estimates of relative growth mirror the differences in the estimate of relative price trends. In the case of France, inflation measured on a PPS basis over the period was 6 points higher than the European Union average, whereas GDP inflation measured by the INSEE was 0.4 point lower. Denmark was in the opposite situation (+8.6 points).

To this should be added the margin of error affecting the estimate of nominal GDP (number of billion euros produced on national territory), as Box 4 shows. This estimate is subject to numerous conventions such as, for example, making allowance for the underground economy[4].

2. Measuring construction prices: an example that illustrates the difficulties that the ICP must overcome

The difficulties that the ICP must overcome to compare price levels are very different depending on which GDP aggregates are involved. Household consumption expenditure, excluding rents, is probably the aggregate that poses the fewest problems. It is also, most fortunately, the biggest use of GDP. This is, moreover, one of the reasons for limiting the ICP to household consumption (see Box 5). But other aggregates -- together representing half of GDP -- pose major difficulties: rents -- not only actual but imputed[5] --, government consumption expenditure (roughly the wage level of civil servants -- teachers, hospital staff, police, justice, defence personnel ), capital goods, software, construction. The last aggregate (accounting for nearly 10 per cent of GDP in most countries) is a particularly significant example of the difficulty of comparing price levels across countries.

2.1 Comparison of construction prices: a complex method

How do Eurostat and the OECD go about comparing constructions prices? In the same way as for consumption and capital goods, from very detailed lists of product prices (in theory). The classification first distinguishes between three main items (or analytical categories): residential buildings, non-residential buildings and civil engineering works. These categories are in turn broken down into elementary positions, the most detailed level at which Eurostat and the OECD calculate price level indices by the EKS method (see Box 2) from the estimated prices of various construction projects. For residential buildings, the elementary positions are apartments, houses and renovation. For non-residential buildings, they are agricultural buildings, factories, office blocks, administrative buildings and renovation. Lastly, civil engineering works cover transport infrastructure, energy transport and communications infrastructure, and other civil engineering works.

Thirty construction projects (but the number will be reduced from 2001) are priced in each country by an expert[6] who has a detailed bill of quantities for each project. Projects are broken down into main operations (internal fittings, roofing, etc.) which in turn are broken down into elementary components (for example, “enamelled steel bath (1.70m x 0.70m ), including mixer tap, shower attachment and hose”). The expert calculates the “unit” prices, then adds them up to obtain the total price of the project.

2.2 Which leaves too large a margin of uncertainty surrounding the price level

As described, the procedure for comparing prices seems reliable. But in practice, a large margin of uncertainty surrounds relative price levels across countries because many other elements enter into the formation of the prices of a construction project. These elements -- works organisation and management, ancillary costs and regional differences in prices -- have to be taken into account. But for this, the experts have only general guidelines, and these elements are not costed separately in the bills of quantities at their disposal. They therefore have to increase unit costs, which would otherwise reflect only the cost of materials. Analysis of these additional construction costs shows that the estimates of the costs of projects are subject to a large margin of uncertainty.

The costing of works management (“delegation of works”) is thus left to the appraisal of national experts. However, a wide variety of practices may be observed. The client, whether individual, corporate or government, can either deal directly with individual trades [7](masonry, etc, ) or entrust the whole project to a general contractor who will handle the rough work and sub-contract the remaining work to specialised firms. The general contractor will take 8 to 10 per cent for this, which will push up prices by as much. But other arrangements are possible: groupings of firms, separate packages or macro-packages with or without scheduling, supervision and co-ordination, which would give different overall costs.

Ancillary costs also have to be taken into account. They are varied and complicated to measure:

Site installation costs (at least 5 per cent of the total cost): levelling, installation of cranes and scaffolding, fencing, installation of site offices, etc;

Insurance;

Design costs, as the architect can never do all the specifications for a project.

The prices transmitted to Eurostat and the OECD must be domestic prices. But while environmental considerations make this question particularly tricky, there are no spatial comparison surveys, unlike in the case of consumer prices, that make it possible to confirm that the prices supplied to Eurostat and the OECD are indeed domestic prices.

Differences in the way countries estimate and take these elements into account, compound the appreciable differences in the prices of materials. The prices of elementary components vary widely depending whether basic or more elaborate goods are involved, which bills of quantities do not specify.

2.3 The prices supplied by national experts are often amazing

It follow that two essential conditions have to be met for prices to be really comparable: firstly , the quality of the national experts must be ensured; secondly, Eurostat and the OECD must check the prices they receive, and ensure that they are consistent. In practice, these conditions are far from being met.

In some countries, construction projects are priced by the National Statistical Institutes themselves, in which case it is to be feared that the prices correspond more or less to the prices of materials. Only firms specialised in financial engineering and construction management are able to evaluate all the costs of a project. That said, it has been observed that when the experts have changed, the price estimates for the same project have varied widely.

Up to now, prices have not been checked for want of resources. And yet relative prices are sometimes surprising. Two examples among others may be cited: the prices of agricultural buildings vary by a factor of one to four between countries like France, Italy and Austria; the price of an asphalted road varies by a factor of one to three between France on the one hand, and Germany, Italy and Belgium on the other. Still more surprising, for some projects the price of a more complex variant is higher than the initial project in most countries, which is to be expected, but well below it in one or two countries. Other anomalies have unfortunately been observed: a project which is not considered to be characteristic of a country is given a small weighting in the computation of that country’s relative price level. But the choice of what is considered a characteristic project is sometimes surprising: for example, up to now concrete roads have characterised French road infrastructure[8], not asphalted roads[9].

2.4 Insufficiently representative sampling

Lastly, the representativeness of the project sample is open to criticism: agricultural buildings are over-represented compared with industrial buildings or major infrastructure such as airports, rail networks, ports, hospitals, thermal power stations, which are missing from the sample. The importance of such infrastructure, which to a large extent is publicly financed, varies quite a lot from one country to another, and has its own specific mode of price formation since prices are governed by public procurement rules. Its impact on relative construction prices is not taken into account as it should be in the current procedure for computing PPPs.

3. the revision of PPPs: an opportunity to restore credibility to the ICP which should not be missed

The persistence of these problems prompted criticisms that found a wide echo in the French press. The adoption by all European countries of SNA 93 (ESA 95) from 2000 convinced Eurostat and the OECD of the need to revise the calculations for the period 1995-2000.

The European Task Force on the revision of PPPs 1995 to 1999 comprises Eurostat, the OECD and a few Member States: France, Belgium, Portugal, Italy and Austria. It has two objectives: first, to recalculate PPPs for the period 1995-1999 in ESA 95 (the European system of national accounts now used by the EU Member States); second, to rectify the biggest errors in the calculation of PPPs over the period 1995-2000.

3.1 Two objectives -- but one priority: to correct price levels

It is now possible to recalculate PPPs in ESA 95. For the year 2000. for the first time all European countries provided national accounting data, essentially weightings, in the ESA 95 format. The impact of this could be quite significant, especially for the earliest years (the switch-over from ESA 79 to ESA 95 was gradual), as the change in nomenclature probably distorted these weightings.

In contrast, the change-over to ESA 95 did not affects prices at the elementary level of the price lists. But as we saw, for some GDP aggregates -- rents, construction, capital goods -- those prices are highly questionable. The second -- and most important objective of the Task Force -- is thus to correct prices. The first thing to do is to detect errors in prices. For this, two complementary approaches can be used.

3.2 It is essential to reconcile PPP-based figures with national account figures

The first approach consists in comparing systematically PPP- based temporal trends in prices with national accounts data (the approach adopted by this study). A big difference indicates that there is a serious problem (at least for countries whose GDP structure does not differ too much from the European average). However, this method has two drawbacks.

It will show if there is error in prices but not the date of the error: if national accounts data and PPPs diverge between two periods, one does not know whether it is the price levels of the first or second period that should be adjusted;

It does not reveal all the errors: a country whose prices were systematically too high could have temporal trends that were identical to those shown by national accounts.

3.3 “Cross-sectional” comparison of basic price data

A complementary approach is thus essential. This involves comparing, for each year and at the most detailed level, product prices across countries. To be done properly, the comparison has to be “cross-sectional”: the expert (or the same team) checks the prices of a small number of products for all countries.

Once the errors in prices have been detected, the second stage consists in correcting them, which poses two types of difficulties.

3.4 Correcting prices -- a politically sensitive exercise

The first difficulty is a technical one: how does one correct price deemed to be incorrect when they are not regulated prices (rail fares, civil service salaries, etc.) since it is not possible to observe them retrospectively? As regards rents or construction prices, prices are obtained by adding various items (insurance, taxes, charges, etc,) which countries do not take into account in the PPP exercise. One solution is to recalculate the prices using a common “perimeter”. However, this is tricky to do; it requires a detailed analysis of prices, the possibility of which depends on complex information being available. Otherwise, the best solution seems to be to make an imputation from price trends in national accounts.

In Europe, there is a second -- political -- difficulty. Can one impose a price correction on a country that would alter its GDP per capita ranking, and thus the structural funds disbursed by the European Union?

The revision of PPPs is thus an important but difficult task. It constitutes an unprecedented opportunity to restore credibility to a key essential statistical indicator. It would be useful to extend this type of analysis to all the OECD countries.

Once the revision has been completed, it would be essential to take the following measures to overcome the fundamental difficulties besetting the ICP:

Ensure temporal consistency between PPP-based GDP with national accounts

Carry out “cross-sectional” comparisons of price levels

Systematically revise the figures in the event of manifest errors detected belatedly

Increase the role of comparisons in terms of final consumption per capita.

.

Bibliography

Balk, B.M. (1996), A comparison of ten methods for multilateral international price and volume comparison, Journal of Official Statistics, Vol 12. N°2. pp. 199-222.

Dalgaard, E and Sorensen H.PPS, Consistency Between PPP Benchmarks and National Price and Volume Indices, paper presented for the 27th General Conference of the International Association for Research in Income and Wealth, Stockholm, Sweden, August 18-24. 2002

Eurostat (2002), GDP per capita in PPS, website, Collection : Key Indicators, Theme : General Statistics.

Eurostat (2002), Minutes of the first task force meeting on the “Revision of the PPP 1995 to 1999“, Luxembourg, 8 July 2002

Magnien, F. (2002), Analyse de la position de la France en termes de PIB par habitant, note interne Insee N° 2/G420. 10 January 2002

Varjonnen, PPS. (2001), Consistency Between GDP Based on PPPs and National Accounts Time Series , Document presented at the OECD meeting of national accounts experts.

Box 1. Computing PPPs

Annual comparisons of the economic activity of OECD countries are based on GDP per capita. But GDPs cannot be expressed in national currencies -- the comparison would be meaningless -- nor even simply converted into a single currency, for example the euro, using exchange rates, because differences in price levels between countries would not be taken into account. The correction is done by calculating with national currencies the price of a basket of goods and services that is representative of overall economic activity (final consumption, investment, foreign trade, changes in inventories) in all the countries being compared. The basket plays the role of a unit of account, worth a PPS (purchasing power standard). It thus possible to express the currencies of the different countries in PPS. GDPs in PPS are then directly comparable in volume terms, and in principle make it possible to rank countries by wealth produced per capita.

Box 2. The structural effect

This effect results from cross-country differences in the volume structure of uses of GDP. The mechanism can be described simply. Let us consider a product whose price is set internationally and whose share in the GDP of country P is bigger than in the other countries. If the price of the product increases, then, other things being equal, the GDP in PPS of country P will increase more than that of the other countries[10], even if the quantities change in the same proportions. “Extrapolated” PPP-based GDPs will therefore be distinct, independently of any problems of statistical measurement. This effect, which is known but difficult to measure, may have operated favourably in the case of the United Kingdom (and even more so Norway) for petroleum products.

This structural difference raises a methodological problem for the calculation of PPPs. How does one construct, for example, a basket of goods and services representative of consumption in two countries as different as Portugal and Denmark? This is done by “graduality”: for example, one compares Denmark with Germany, then Germany with France, France with Spain, and lastly Spain with Portugal. But a difficulty arises: unlike for a temporal index (CPI), there are several possible “paths’ for comparing Denmark with Portugal. “Chaining”, which is the best solution for temporal indices, is unsuited. “Transitive’ indices, i.e. which do not depend on the path followed, require other methods, two of which are used by Eurostat and the OECD: the EKS and GK methods. Very roughly, the EKS method consists in calculating the price in each country of a representative basket of goods and services common to all the countries being compared. The GK method involves calculating, symmetrically, the price of baskets specific to each country using a system of median prices. The difficult lies in constructing the common basket and the system of median prices.

Box 3: The difficulty of measuring nominal GDP

Besides the difficulty of comparing price levels across countries, calculations of wealth per capita are complicated by the non-robustness of the measure of nominal GDP. To a large extent this measure is normative, and based on conceptual choices that can change. Thus, in the previous version of national accounts (SNA 68) software was treated as intermediate consumption, whereas in SNA 93 it is classified under investment, the effect of which is to increase GDP but in proportions that vary markedly from one country to the next. The non-robustness of the statistical measure itself (on the basis of the given concepts) renders comparisons of nominal GDP even more uncertain. The split between software intermediate consumption and investment is especially illuminating: the ratio “investment/intermediate consumption + investment” varies widely between the United Kingdom, France, Italy Germany, and even more so with the United States. Business surveys provide scant details on the nature of firms’ fixed assets. The value of software developed by firms in-house for their own needs is even more difficult to estimate.

Measures of the underground economy -- moonlighting, tax evasion and avoidance -- are also very approximate. Factoring in the underground economy pushes up GDP considerably, by nearly 5 per cent in the case of France.

Conceptual and methodological changes give rise in each country to frequent “rebasings” of national accounts, resulting in large revisions of nominal GDP, though basic trends remain relatively unchanged.

Box 4. Spatial and temporal price comparisons: different statistical operations

In practical terms, the reliability of PPP calculations hinges on the quality of price data collection. It might be thought that the method of collection is the same for spatial comparison -- computing PPPs -- as for temporal comparison -- the consumer price index (CPI) for example. In fact, to measure the temporal trend of prices in a given area, price surveys are carried out simultaneously in all parts of the area, and then the exercise is repeated later. When price levels in the area are compared during a given period, price surveys are also carried out simultaneously in the entire area, and if the spatial comparison is repeated, prices are collected again. On the face of it, therefore, it might seem that the statistical operations involved in collecting prices for spatial and temporal comparisons are the same.

In fact, it is not the case. The quite different aims underpinning the two types of comparison mean that the price data collected are exploited in very different ways:

for a spatial comparison, prices collected during the same period are compared with one another, but no link is established with prices observed during a later period;

for temporal comparisons, in contrast, prices during the same period are not compared with one another but with a single price -- that of the same item during a later period.

Consequently, the samples of items are different:

in the case of a spatial comparison, the items whose prices are observed simultaneously in all parts of the area must be absolutely similar since the slightest difference results in an unwarranted difference in price levels;

in the case of a temporal comparison, the similarity concerns pairs of items whose prices are being compared during each period. These items, whose prices in the area are compared simultaneously, do not have to be absolutely similar for two reasons: i) they are selected so as to be representative of the prices prevailing in the area; ii) the time trend of the prices of two articles is substantially the same, even though they are slightly different. This second consideration warrants a certain amount of flexibility in the collection of prices for temporal price indices such as the CPI.

Box 5. Place more emphasis on household consumption in per capita comparisons

Consumption of goods and services being the aim of economic activity, it would be desirable to give more importance to it in per capita comparisons. Household consumption also has the advantage that it can be estimated more accurately than other uses of GDP. The consumption considered is actual consumption, i.e. that which includes the part financed by government. This is because the aggregate confined solely to household consumption expenditure is unsuited to international comparisons of living standards since, for the same volume of consumption, it depends on the social policies implemented in each country. It is all the smaller in that social cover is extensive. Publication of such an aggregate, which is irrelevant for international comparisons, is liable to contribute to misunderstanding on the part of users and to provoke misinformed public debate.

-----------------------

[1]. French delegate to the Eurostat Working Party on PPPs. The views and opinions expressed in this paper are those of the author only and do not commit the INSEE in any way.

[2]. Which has never been done in the ICP.

[3]. For example, if the Italian lira is devalued against the ecu, Italian prices will fall against the European average by as much. The relative price trend thus takes account of this effect.

[4]. Whereas the measure of changes in GDP (growth rate) is considered to be more reliable.

[5]. The share of which varies widely from one country to another, as do the methods of evaluation also.

[6]. One per country.

[7]. One speaks of “separate building trades”.

[8]. This anomaly was rectified for the 2000 comparison.

[9]. There are no concrete roads in France

[10]. Value GDPs are directly comparable since the price of the product is set internationally.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download