Capitalization of Software in the National Accounts

View the most recent version.

5.0 The estimation of software investment benchmarks for Canada

The following describes the sources and methods for software investment and its implementation in the Canadian System of National Accounts.

• 5.1 Purchased software
• 5.2 Own-account software
• 5.3 Software prices
• 5.4 Software already capitalized
• 5.5 Software capital stocks and capital consumption

5.1 Purchased software

As mentioned above investment in prepackaged and custom-design software is derived indirectly via estimation of both the supply and demand for software. The supply consists of domestic production, plus imports, plus the wholesale, retail and tax margins on software sales in Canada. Investment in software is then determined as the amount that equates demand to supply or, more precisely, as the total supply of software net of the purchases by resident households and by non-residents.¹⁷ The table below shows how this approach works in practice, using the 1998 benchmarks (reconciled with the I/O Accounts) as an example and omitting some of the details.

SOFTWARE COMMODITY-FLOWS, 1998
$ millions
Domestic production	6,389
+ Margins on domestic sales (incl. imports)	1,728
+ Imports	2,002
= Total supply of software	10,117
- Exports	2,151
- Personal expenditure	410
= Intermediate use of software (former accounting)	7,557
- Software embedded in hardware	373
= Investment in software (new accounting)	7,185
Note: Figures may not add due to rounding.

Domestic production of software (valued at producer prices, i.e., prices at the ‘factory gate’) includes receipts from sales as well as licensing of pre-packaged software and, in the case of custom software, it includes receipts from sales as well as royalties paid from abroad.¹⁸ The main sources are Statistics Canada’s annual surveys of Computer Services and of International Transactions in Commercial Services. Software Publishing (NAICS 511210) accounts for 85% of domestic production of pre-packaged software, while Computer Systems Design and Related Services (NAICS 541510) accounts for 92% of domestic production of custom software. As a whole, the computer service industries account for virtually all software production in Canada.¹⁹

Margins are added in order to balance supply with demand at prices paid by purchasers (which reflect additional distribution costs and taxes, on top of producer prices). The margin amounts come from the I/O Accounts, which in turn draw on Statistics Canada’s Wholesale and Retail Trade Surveys and detailed information on commodity taxes. In the case of custom-design, wholesale and retail channels are largely avoided, so the relatively low margins here reflect only taxes.

Imports cover merchandise imports of software as well as royalties paid to non-residents, in the case of custom software (from the surveys of International Transactions in Commercial Services).²⁰ Some 25% of software goods imports (1998) are actually customized software coming into Canada on physical media (e.g., tapes, diskettes). Since custom-software is viewed as a service, it is regularly removed from merchandise import data via a Balance of Payments (BOP) adjustment to avoid double-counting with BOP payments for computer services. The amount removed is included above as an import of custom software, while the amount left over is included as a pre-packaged software import.

Exports cover both merchandise exports and re-exports (which are treated here entirely as pre-packaged software), foreign receipts, excluding royalties, from custom software (from the Survey of Computer Services), as well as receipts of license fees and other software royalties paid from abroad (from the surveys of International Transactions in Commercial Services). Merchandise exports include a BOP adjustment, supplied by the U.S. BEA, which raises the Customs export values by over 50%.²¹ The adjustment is to correct an underestimation stemming from media-valuation in the U.S. Customs data on imports of software (there’s no similar concern with Canadian import data, where valuation is on content).²² Receipts from license fees are wholly attributed to pre-packaged software, while receipts of other software royalties are attributed to exports of custom-design software.

Personal expenditure covers consumer spending on pre-packaged software, obtained from Statistics Canada’s annual Survey of Household Spending. These figures exclude spending on electronic games (as do software imports and exports, where there’s a separate set of codes to capture games, electronic encyclopedias, and the like).²³

Regarding software that gets embedded in hardware (and counted as hardware investment), a deduction is made to avoid counting it a second time under software investment. There’s no direct source of information for the amounts involved, so this deduction rests on assumptions that stand to be revisited. The deduction is arbitrarily set at 50% of the purchases of pre-packaged software by the computer manufacturing industry, on the assumption that this is the only industry engaged in purchasing and installing pre-packaged software in hardware and that some of its purchases are for this purpose (and others for investment).²⁴ It is arbitrarily set at 40% of the purchases of custom-design software in printing, publishing & allied, machinery & transportation equipment, electrical and electronics and other manufacturing, as well as business services (which includes computer services), on the assumption that these industries embed some of their purchased custom software in their products.

Last, the investment in pre-packaged and custom-design software is split between business and government. Government investment is estimated from administrative data on software purchases, survey data on industry sales to government (Survey of Computer Services), and capital spending on software (CAPEX). Business investment is residually derived.²⁵ It might be noted that, starting in April 2001, the federal government has included capitalized software on its own books, thus providing a direct source of information for use in the GDP accounts.

5.2 Own-account software

There’s no source of direct information on the value of in-house software developed for own-use. Thus, as is the practice elsewhere, own-account software investment is built up from estimates of the labour cost for persons engaged in software development. The estimates (which are done at a detailed industry and provincial/territorial level) are benchmarked on quinquennial Census of Population data on the annual aggregate wage and salary income of computer programmers and systems analysts.^26,27,28 These figures are raised for additional costs faced by employers, such as employment insurance, public and private pension plan premiums, etc. to arrive at a more comprehensive labour cost.²⁹ The table below shows the various steps for own-account for the most recent benchmark year, omitting some of the details.

OWN-ACCOUNT SOFTWARE BENCHMARKS, 1995
$ millions
Labour cost for computer programmers and systems analysts	7,117
- Deduction for work on software to be embedded or sold	3,032
- Deduction for time spent on non-investment related work	2,042
= Labour cost of own-account software development	2,043
+ Cost of other inputs	939
= Investment in own-account software	2,982
Note: Figures may not add due to rounding.

A deduction is made (along the lines of the U.S. method) to avoid double-counting software to be embedded or sold, which is already accounted for under hardware investment or purchased software. In those industries not engaged either in producing software or embedding it in hardware, the labour cost for programmers and systems analysts is about 1% of all wages, salaries and supplementary labour income. This percentage is used to cap the labour cost of programmers and systems analysts in software producing and embedding industries, on the assumption that costs over and above this threshold are related to software production and/or embedding, not the everyday running, maintenance and development of software systems that is nowadays integral to operations in most industries.³⁰

The cap results in an overall deduction of $3 billion (about 43% of the total labour cost attributable to programmers and systems analysts). Almost two-thirds of this deduction is centered in the computer service industries, which employ about one-third of all programmers and systems analysts. Labour income for the industry as a whole is about $4 billion (1995), 50% of which is attributable to programmers and systems analysts, well above the threshold value of 1%. Applying the cap to total labour income for the industry gives $40 million allowed as own account software development, while the remaining $1.96 billion in programmer and systems analysts labour income is removed to avoid double-counting with purchased software.

A second deduction is made to avoid counting the time spent on, and labour cost associated with, the routine running and maintenance of computer systems. It is assumed that only half of the work of programmers and systems analysts is on own-account software development and the rest on routine operations, leading to a further 50% reduction. This step simply repeats the U.S. BEA methodology which, in this case, is based on a study of how programmers spend their time.

Last, an amount is added for other, non-labour, costs of own-account software development. This is arrived at by examining the cost structure of a sub-sample of firms in the Survey of Computer Services who derive the majority of their revenues either from custom software development or contract programming. These ‘custom software developers’ are more labour intensive than the industry as a whole, with labour costs averaging just over two-thirds of total operating expenses (i.e., non-labour costs are about 50% of the labour cost).³¹ The estimate from the previous step is then raised by 50% to arrive at own-account software investment. This assumes that the cost structure facing custom software developers adequately represents the cost structure (i.e., the technology) for own-account software development across all industries (business and government); an assumption that merits further investigation.

5.3 Software prices

At present there’s no made-in-Canada price index for pre-packaged software, although Statistics Canada is currently developing one aimed for release in 2002. In the meantime, an adjusted version of the U.S. BEA price index for pre-packaged software is used to fill the gap.³² The price used to deflate pre-packaged software investment here is just an average of the U.S. BEA price index, weighted by the domestic share of supply to the domestic market, and an exchange-rate adjusted version of the same index, weighted by the import share of supply to the domestic market. It is assumed, somewhat controversially, that exchange rate fluctuations are fully passed through to the domestic price of imported software.

The price index used to deflate own-account software investment is a fixed-weighted average of an index of the average hourly earnings for programmers and systems analysts (established separately for business and government) and an index of the costs of non-labour inputs to the computer services industry, with weights of about two-thirds and one-third, respectively. The hourly earnings index is derived from and benchmarked on Census of Population data on the derived hourly earnings of programmers and systems analysts.³³ Fixed-weighted average hourly earnings indexes from the Survey of Employment, Payrolls and Hours are used to interpolate and extrapolate the benchmarks over 1981-97.³⁴ Average hourly earnings indexes from the Labour Force Survey are used to bring the indexes forward to the current year. It is assumed that there’s no change in the productivity of programmers and systems analysts, an assumption that stands to be revisited.

The price index for non-labour inputs to own-account software development is the implicit price index for intermediate inputs to the computer services industry, obtained from the I/O Accounts up to 1997. From then on, it is updated with a Laspeyres (1997 fixed-weighted) price index for those inputs accounting for 1% or more of intermediate input costs.

Following the U.S. BEA methodology, movements in the price index for custom-design software are derived as a weighted average of the changes in the price indexes for pre-packaged software and own-account software development, with arbitrary weights of 25% and 75%, respectively.

5.4 Software already capitalized

Prior to 1998, and going back to 1988, Statistics Canada’s Survey of Capital and Repair Expenditures (CAPEX) had listed ‘software’ only as an item to be included under a broader asset category which up until now has been identified as ‘computer hardware’. Insofar as organizations have been capitalizing software (purchased separately from hardware) and reporting this capital spending under the broader ‘computer hardware’ asset category, there’s a double-count of software between the estimates discussed above and previously published estimates of investment in office machinery and equipment. This is quite apart and different from the issue of double-counting software that is physically embedded in hardware.

To avoid this double-count, a deduction is made to hardware investment, arrived at as follows. Results from a follow-up survey with CAPEX respondents who reported software, seeking a breakdown of their capital outlays by type (pre-packaged, custom-design, own-account), are used to breakdown the 1998 CAPEX software total to estimate capital spending by type of software.³⁵ These estimates are compared to capital plus non-capital spending to establish the fraction of spending on software (by type) that is capitalized.³⁶ These ‘capitalization ratios’ are assumed to have been constant over 1988-1997, and applied to the software investment series to arrive at capitalized software deemed to be previously reported to CAPEX but under hardware investment. Starting with reference year 2000, CAPEX will gather details of capital spending on software by type, allowing for continuing updates of these ratios.

Prior to 1988, CAPEX made no mention of software in the description for the hardware asset category, and it is unclear how respondents with capital spending on software would have reported it, if at all. Rather than drop the adjustment altogether, which would create a significant break in the hardware series, the capitalization ratios are gradually reduced going back to 1981 to half their initial values.

This deduction is allocated to the business and government sectors on their respective shares of hardware investment, in the case of purchased software, and 100% to the business sector, in the case of own-account.

5.5 Software capital stocks and capital consumption

Software capital stocks are estimated using the Perpetual Inventory Method (PIM), with straight-line depreciation, assuming service lives of 3 years (pre-packaged) and 5 years (own-account and custom-design). The service lives used here are the same as in the U.S., and are consistent with results on lives found in CAPEX and the follow-up.³⁷ The stocks are built up from the software investment series, net of the reduction to the hardware investment series.³⁸ Initial stocks for 1981 are obtained via PIM by carrying the software investment series back (on their growth over 1981-82) to 1978 or 1976, depending on the service life. Estimates for capital consumption, coming out of PIM, are on an historic cost basis for the business sector and a replacement cost basis for the government sector. This is in line with the estimates of business and government CCA currently in the accounts.

 

Table of contents