View the most recent version.

Capitalization of Software in the National Accounts

Printable version (PDF) of this article

All spending by businesses and governments on software will be treated as a capital expenditure. Previously, only a small portion, less than 20% of these expenditures, was treated as such. This will be a net addition to GDP, both in level and in growth, in the years in which software expenditures grow faster than other components of final expenditure. Treating software as a capital expenditure has been introduced by most countries, including the United States, within the last couple of years.

Table of contents
1.0 Introduction 2.0 The effects of capitalizing software on GDP 3.0 Trends of software investment in Canada 4.0 Software investment in Canada and the U.S. compared 5.0 The estimation of software investment benchmarks for Canada 6.0 The estimation of annual and quarterly, industry and provincial time series for software investment 7.0 Tables

1.0 Introduction

A new accounting treatment of software as investment was implemented in the Canadian System of National Accounts during 2001. Preliminary estimates of software capital stocks were included for the first time in the National Balance Sheet Accounts (NBSA) released in March 2001. Software investment was then included in GDP with the first quarter 2001 release (31 May 2001) of the National Economic and Financial Accounts (NEFA). Later in the year, it was included in the Input-output (I/O) Accounts, Provincial Economic Accounts (PEA) and the Industry Measures Accounts (IMA) with the release of 30 October 2001.

This mini historical revision brings Canada in line with a number of countries, including the U.S. and other G-7 member nations, who introduced software into their GDP over the last few years. It also brings Canada in line with the 1993 SNA recommendation that business and government acquisition of software be treated in national accounts as an investment as opposed to a current expense.¹ Software is now treated like any other capital input that is used repeatedly in production over a year or more whereas, formerly, it was treated as if it were fully used up during the production period like any other intermediate input. This new accounting for software has raised the level of GDP although the effects on GDP growth turn out to be relatively small.

Software is characterized by three types: pre-packaged, custom-design and own-account.² Pre-packaged software is of the sort that can be purchased ‘off-the-shelf’ and is typically mass-produced and sold or licensed in standardized form. It is intended for generalized uses common to the every-day operations of businesses and governments. Custom-design software, by contrast, is intended for specialized uses. It is typically developed for and tailored to a specific organization’s needs by some third party software developer under contract. Customized software has limited application beyond the particular ‘business problem’ it is designed to solve. Like custom-design, own-account software is specialized to a specific organization’s needs, and distinguished only insofar as its development is undertaken ‘in-house’ by employees within the organization rather than being contracted out.³

Because expenditures on software are not always tracked separately or treated uniformly in organizations’ accounting records and because Statistics Canada’s surveys haven’t always asked for the details of these specific expenditures, the estimation of software investment relies on indirect methods.⁴ In the case of purchased software, pre-packaged and custom-design, the method involves estimation of the components of the software market in Canada, with investment determined residually as the amount that equates demand with supply. In the case of own-account, the method rests on an estimation of the wage bill for computer programmers and systems analysts as a starting point for assessing the costs of software developed in-house for own-use.

Not all purchases or costs related to software acquisition are included as investment. In particular, expenditures on repair and maintenance (e.g., Y2K and other emergency fixes, routine de-bugging and re-coding to accommodate changes to input data) are not included, nor is spending on employees’ training on software (unless part of a package deal).⁵ Purchased software that gets embedded in hardware and then re-sold continues to be treated as an intermediate use to avoid double-counting (i.e., the initial purchase is deemed to be intermediate, while the subsequent purchase of hardware (with embedded software) is treated as investment). As well, the costs of developing software that is to be sold are excluded from the own-account estimates in order to avoid double-counting with purchased software.

Software investment here covers organizations’ capital plus non-capital spending, essentially treating all of their software purchases and own-account development costs as capital outlays, irrespective of how these outlays are treated on their books.⁶ Some direct estimates for the actual capital spending component (as reported in accounting records) are now available, as a result of the introduction of ‘software’ as an explicit asset category on Statistics Canada’s 1998 Survey of Capital and Repair Expenditures (CAPEX). The Survey has been capturing a significant portion of software investment in the past, but under computer hardware.⁷ An adjustment to the hardware investment series is made to remove these amounts in order, again, to avoid a double-count. The resulting downward revision to hardware goes a long way to explaining the minimal effects of the new treatment of software on GDP. Moreover, because the prices for hardware have fallen even faster than those for software, the reduction to hardware reduces real GDP growth more than the equivalent amount of software adds to it.

With the changes made in 2001, the estimates of software investment are now available in the NEFA on a quarterly seasonally adjusted and unadjusted basis at current, constant 1997 and chained 1997 prices, by sector, from the first quarter 1981. They are available also by industry, at current and constant prices, in the national I/O Accounts, annually for 1981-1998, and by industry and province/territory in the provincial I/O Accounts for 1997 and 1998. No details on software are available in the NBSA, PEA or IMA because it gets subsumed under broader aggregates.

As with most series in the national accounts, the correspondence between definitions and underlying concepts, the data sources and the estimation methods are revisited, refined and further developed over time. The software investment series are no exception. Indeed, these are likely to undergo refinements if only because the source surveys undergo modification in order to keep up with the rapidly evolving market for software. In addition, with each quarterly and annual update of the estimates, experience is gained in terms of how they fit with other aggregates in the national accounts. This on-going process can indicate refinements to both the sources and methods in order to improve the software estimates.

The estimates for 1998-2001 will be open for revision at the time of the next annual revision of the national accounts, scheduled for May 2002. This will allow the incorporation of more recent data, including results from the 1999 Annual Survey of Software Development and Computer Services, the 1999 Survey of Capital and Repair Expenditures, the 2000 surveys on International Transactions in Commercial Services, updated merchandise trade statistics, and preliminary I/O Accounts for 1999 and the final ones for 1998. Some parameters underlying the estimates will be updated as well, including business/government split factors, the share of intermediate costs in the own-account estimates, and the margin rates on software sales to the domestic market.

There is room for more substantive improvements in the future. Results from the 2001 Census will provide new benchmark data next year for the own-account estimates. A new made-in-Canada price index for pre-packaged software is under development and will eventually replace the U.S. pre-packaged software index currently in use. The feasibility of developing a provincial dimension to the various software price indexes and the quality-adjustment of software prices warrant further investigation. As well, the adjustment made to avoid double-counting the work of programmers and systems analysts on software that gets sold and the adjustments for software embedded in hardware, which all rest largely on assumptions, stand to be revisited. Last, following a recent survey and review of national practices, the OECD/Eurostat joint Task Force on Software Measurement in the National Accounts will make recommendations later this year on best practices in this area, and these will be considered in due course.

The following section looks at the effects of the new accounting treatment of software on GDP, its components and on GDP growth. Next comes a summary of the software investment results for Canada, followed by a comparison with those for the U.S. A brief outline of the approach to estimating the software benchmarks for 1997 and 1998 is then provided. A more detailed description follows of the data, sources and methods for the full annual time series, 1981-2000, the quarterly series, and the industry and provincial distribution of software investment. Appendix Tables A.1-E.3. summarize the data, sources and methods.