STD/NA(2001)26
15
STD/NA(2001)26
Table of contents
1. Issues relating to definitions 3
2. French valuation of software expenditure 4
3. Valuation of GFCF in software using the demand approach 7
Split between GFCF and IC (GFCF/(GFCF+IC) ratio) 8
4. Valuation of own-account software production 9
Models and results 9
5. Valuation of GFCF in software using the supply approach 11
Annex: Sources 14
Annual business surveys (EAE) 14
Employment survey 14
Structural Employment Survey (ESE) 14
Annual Declarations of Social Data (DADS) 15
ESTIMATES OF GFCF IN SOFTWARE IN FRANCE:SOURCES, RESULTS AND OUTLOOK
1. Issues relating to definitions
The latest System of National Accounts (SNA 93) provided an opportunity to extend the concept of GFCF to a number of intangible fixed assets. This extension was incorporated into the 1995 European System of Accounts (ESA 95). Software (assigned to code AN.1122 in the classification of assets) and databases are now treated as assets subject to GFCF. According to the 1993 SNA, these products are defined as:
"Computer programs, program descriptions and supporting materials for both systems and applications software. Included are purchased software and software developed on own account, if the expenditure is large. Large expenditures on the purchase, development or extension of computer databases that are expected to be used for more than one year, whether marketed or not, are also included."
It should be noted that the scope of this definition extends to:
• customised software;
• originals for publication;
• published products.
The same is true for databases.
No explicit estimates are currently made of GFCF in databases in France in that no mention is made of databases under code 7220 "software development" in the Classification of Products by Activity. Databases could be included under code 7240 "Databank services", but for the time being there is no GFCF associated with this code.
The borderline between gross fixed capital formation and intermediate consumption is not always easy to establish in the case of software and databases. The first criterion to take into account, since it is explicitly stated in the 1995 SNA, is the amount of expenditure, which must be "substantial" if the acquisition or own-account production of software is to be classified under the gross formation of fixed capital. While admittedly it is relatively easy to establish a more or less conventional threshold for determination of what should be deemed to constitute "substantial expenditure", simply setting a threshold does not solve all the problems. This is particularly true in the case of office software, which individually is of relatively little value but which firms purchase in large quantities. It would seem logical here to consider, not the value of each software product, but the amount of total expenditure, which is precisely what is recommended at the European level. However, this solution does present certain disadvantages in that the size of the software order placed by a firm largely depends upon its IT policy. For example, two identical firms may have identical IT expenditures in the medium term, but one may renew its software resources once every four years while the other renews a quarter of its resources every year. The expenditure by the former can be classified as GFCF, while the latter's can be reported as intermediate consumption. The use of a threshold therefore makes the contribution to GDP of a firm dependent upon its IT policy.
The lack of a distinct boundary poses many other problems. For example, it is unclear how software developed as part of a research and development project, such as the design of a new model of car or a plane, should be classified. In principle, if the software developed as part of such a project has a highly specific purpose and cannot be reused, if should be classified as intermediate consumption. It can solely be classified as gross fixed capital formation if it is likely to be reused for other work. In practice, however, the situation is rarely quite so straightforward in that large parts of a specific software program can be reutilised in another, equally specific software program. It is therefore rarely possible to determine precisely whether or not software developed under a research and development programme should be classified under GFCF.
Similar problems arise in the case of software embedded in either computer products or robots.
The mode of capitalisation of software expenditure also merits further attention, particularly in the case of own-account production of an original software program. Like any form of own-account production, software products can be estimated at either the basic price or, where precluded, the cost of production. Unfortunately, these two values can differ substantially in the case of a software program designed for publication.
2. French valuation of software expenditure
In accordance with the recommendations of SNA 95, an estimate was made of the level of GFCF in software in 1992 in order to provide a benchmark for the 1995 base. This level was established in accordance with three sources of information:
• The results of French annual business surveys (EAE[1]) of the industry carried out by the government statistical agency;
• A study carried out by a consultancy, which broke software expenditure in 1992 down by sector of the industry into purchases and own-account production;
• A study, also by a consultancy, which provides a breakdown of the turnover of computer service companies by type of customer (banks, administration and non-financial enterprises).
The annual business surveys were used to determine the volume of software purchases by industrial enterprises (including enterprises in the energy and agro-food sectors) with 20 or more employees. These purchases have been fully reported as GFCF.
No information was available in the EAE for other enterprises, either because the question had not been asked or because the replies were not sufficiently reliable. An estimate was therefore made by considering software purchases to be equivalent to purchases of intangible assets by enterprises. The percentages used in this estimate were based on expert opinions.
The estimate of own-account production of software for all sectors of activity was similarly obtained by considering such production to be equivalent to immobilised production.
The GFCF in software of financial enterprises and public administrations was estimated on the basis of the structure of software expenditure by type of customer established in one of the above-mentioned studies.
The final annual accounts are established by updating the data for the previous year on the basis of the EAE results and, in the case of own-account production, the trend in immobilised production.
The following table presents the ratios calculated on the basis of the data currently published in French national accounts. These ratios are based on the European CPA (Classification of Products by Activity). It would be interesting for individual countries to recalculate these figures, since it would allow them to consider the differences that exist between countries.
The ratios are calculated at two levels of the classification: division (two positions) and class (four positions). It is possible that some splits will occur solely at an aggregated level. The products presented are:
• Software development assigned to code 7220[2];
• computer services assigned to code 72;
• computers and computer equipment 3002;
• office machines and computer hardware 30.
The parameters used for these products are:
• total GFCF, which in the case of software can be broken down into own-account software and purchased software;
• total IC + GFCF since the split between IC and GFCF poses a problem with regard to software.
The ratios were calculated on the basis of data relating to value and volume (i.e. at the previous year's price related to 1995 as the reference year). The price indices for GFCF and IC of computer services and of computers and computer equipment are "total" indices, i.e. indices obtained after deflating the other items in the employment resource balances (ERE) and after imposing equality of volume.
The production price index for computer services was calculated by the federation of professional trade unions (Syntec). It is based on wage costs and therefore takes no account of any productivity gains in the sector.
The index for production of computers and computer equipment was calculated on the basis of three price indices:
• producer price index[3] for microcomputers in the French market (hedonic index[4]);
• producer price index for microcomputer printers in the French market (hedonic index);
• export price index.
The price index used to deflate household final consumption is the consumer price index . The price indices for exports and imports are based on the unit value indices used by Customs.
The ratios thereby calculated revealed:
• the split between IC and GFCF at several levels of the classification;
• the share of own-account production in GFCF;
• the share of GFCF in purchased software in domestic software sales;
• the relationship between software expenditure (IC + GFCF) and GFCF in computer equipment;
• the share of software expenditure (GFCF + IC) in the GFCF for all products combined (excluding household accommodation).
Table 1: characteristic ratios of GFCF in software in France in 1998
Ratio / As a percentage of value / As a percentage of volumeGFCF (7220)
/
[CI + GFCF] (7220) / 64 / 64
GFCF (72)
/
[CI + GFCF] (72) / 25 / 25
GFCF in own-account software (7220)
/
Total GFCF (7220) / 41 / 41
GFCF in own-account software (72)
/
Total GFCF (72) / 41 / 41
GFCF in purchased software (7220)
/
[GFCF + IC] in purchased software (7220) / 51 / 51
GFCF in purchased software (72)
/
[GFCF + IC] in purchased software (72) / 17 / 17
[GFCF + IC] (7220)
/
GFCF (7220 + 3002) / 103 / 82
[GFCF + IC] (7220)
/
GFCF (3002) / 302 / 173
[GFCF + IC] (7220)
/
GFCF (all products excluding household accommodation) / 7.9 / 7.6
[GFCF + IC] (72)
/
GFCF (all products excluding household accommodation) / 20 / 19
[GFCF + IC] (7220 + 3002)
/
GFCF (all products excluding household accommodation) / 15 / 18
[GFCF + IC] (72 + 30)
/
GFCF (all products excluding household accommodation) / 21 / 23
GFCF: Gross Fixed Capital Formation
IC: Intermediate Consumption
The figures in brackets are the CPA codes (Classification of Products by Activity).
The estimates currently used by national accountants are the outcome of the first attempt to capitalise GFCF in software. On the next change of base, a new estimate will be made in the light of the new statistical sources now available. The questions relating to software purchases have thus been extended to all the major enterprises covered by the EAE. The areas covered are now as follows:
• Industry
• Business
• Services
• Transport
• Construction
The EAE now also allow software purchases by major enterprises to be broken down into immobilised and non-immobilised expenditure.
These new data sources make it possible to use two separate approaches (supply and demand) to the estimation of GCFC in software as well as an estimation of own-account production.
3. Valuation of GFCF in software using the demand approach
Current estimates of GFCF in software in France are based on demand. It is now possible to make use of more accurate data from the annual business surveys (EAE) which explicitly ask questions about software expenditure. The interest to using this method is that it dispenses with the need to use fixed percentages for intangible investment, which fluctuate too widely.
The EAE each comprise a common framework based largely on the French general accounting plan; however, this framework is usually sent solely to enterprises above a given size. The criteria regarding size differ from one EAE to another. They are based on the number of employees or size of turnover (over 20 employees in the manufacturing sector, over 30 employees or turnover of 30 million francs for the business sector, etc.). The data field is therefore partial. It is for this reason that data are presented in the form of ratios: software expenditure to tangible investment net of sales corresponding to the same field.
It is thereby possible by roughly applying ratios to tangible investment net of sales in each sector to estimate the expenditure on software in the sectors studied. Furthermore, the questionnaire distinguishes between immobilised and non-immobilised expenditure. In the tables this distinction has been taken as that between GFCF and IC[5].
Table2: Ratios: software expenditure to tangible investment net of sales
by activity group in 1998 according to the EAE
Transport
Services
Industry
Business
Construction / (I)
ND
2.92
3.85
1.26
ND / ND
0.27
0.74
0.31
ND / (II)
ND
3.19
4.59
1.57
ND / (I)/(II)
92
84
80
Applying these ratios to tangible investment net of total sales[6] in each sector makes it possible to obtain an initial estimate of GFCF in software through the demand approach. In this case, the estimated GFCF is 2.5 billion euros.
According to current accounts, the services, industrial and business sectors consume 91% of the GFCF in software of non-financial companies and individual entrepreneurs (SNFEI). It is markedly lower than the figure currently published.
This low level is attributable to the relative degree of difficulty enterprises experience in providing a specific figure for expenditure on software in their replies to EAE questionnaires. Questions on software are a recent development both for enterprises and in surveys. In addition, software embedded in products poses a problem. Moreover, since the EAE field is incomplete, it may prove necessary to review the estimates that have been made to complete it.
Split between GFCF and IC (GFCF/(GFCF+IC) ratio)
It is possible to calculate the first ratio presented in paragraph 2 on the basis of EAE data for the market services, industry and commerce sectors. It needs to be borne in mind, however, that the field only represents part of the SNFEI.
Table 3: GFCF/IC split according to information given by firms in response to EAEs in 1998