Capitalization of Software in the National Accounts

View the most recent version.

6.0 The estimation of annual and quarterly, industry and provincial time series for software investment

While the general approach described above applies to the full time series, sources and methods vary over time due to data gaps in earlier periods or because survey results are not yet available for more recent years. Appendix Tables A.1-E.1 and A.3-E.3 give details for the annual, national software investment series 1981-2000, while Tables A.3-E.3 include details on the quarterly, national series for 1997 onwards. The following summarizes these details and discusses briefly the industrial and provincial/territorial distribution of software investment for the I/O Accounts and the Provincial Economic Accounts.

• 6.1 Annual national estimates by sector
• 6.2 Annual national estimates by industry
• 6.3 Annual provincial/territorial estimates
• 6.4 Quarterly national estimates by sector

6.1 Annual national estimates by sector

On the supply-side, domestic production 1986-98 for pre-packaged and custom software come from the I/O Accounts and the Survey of Computer Services (SCS), respectively. Selected receipts from the SCS (1981-85) are used to carry these series back to 1981. Payroll data from the Survey of Employment, Payrolls and Hours (SEPH) for the Software Publishing and Computer Design and Related Services industries, respectively, are used to carry them through to 2000. Margins on pre-packaged software are estimated as a fixed percentage of domestic sales, with the percentage based on I/O data for 1997-98. For custom software, they are set at the same rate as margins on ‘professional and processing computer services’ with the 1997 rate held up to 2000. Merchandise imports of software (including custom software on physical media) come from Customs data from 1996 on. This series is taken back to 1988 on imports of all commodities that ‘include software’ (according to HS commodity descriptions),³⁹ and by linear extrapolation back to 1981. Custom software is deemed to make up 25% (based on the 1996-99 data) of software merchandise imports over 1981-95. Software royalties paid abroad 1990-2000 from BOP are carried back to 1981 on custom software imports.

On the demand-side, merchandise exports of software (treated as pre-packaged) come from Customs data for 1997 on. Custom software exports for 1997-98 are from the SCS, carried to 2000 on BOP computer services receipts. Both series are taken back to 1981 on foreign revenues from the SCS. The BOP adjustment for under-valuation of software merchandise exports, available for 1989-2000, is set at 57% for 1981-88, with the percentage based on 1989 data. License fees for prepackaged software and other software royalties from abroad for 1990-2000 are from BOP. The former series is carried back to 1981 assuming a fixed ratio to merchandise exports at content-value, with the ratio based on 1990 data. The latter series is set equal to 28% of custom software exports for 1981-89, with the percentage based on 1990 data. Last, personal expenditure on software comes from the I/O Accounts (1981-98), from the SHS, 1999, and is grown to 2000 on personal expenditure on ‘office machines and equipment’ from IEA.

The adjustments for software embedded in hardware are based on I/O data for 1981-97. For pre-packaged software, a negligible amount, is trended forward for 1998-2000 on a judgmental basis. The adjustment for custom-design is set at 11% of domestic sales for 1998-2000, with the percentage based on 1997 data.

Government investment is built up from two series, for both pre-packaged and custom, one for software already reported as capital spending and one for non-capital spending. As mentioned earlier, the former is arrived at by allocating estimates for total software already capitalized to government and business on their respective shares in hardware investment. The latter series is obtained as a fixed percentage of total investment, in software of each type, with the percentages based on I/O data for 1997-98. The series for business investment are residually derived.

The own-account benchmark estimates for 1981, 1985, 1990 and 1995, follow the steps outlined earlier.⁴⁰ The adjustment for other employer costs increases over time, from 10% of gross wages and salaries in 1981 to 14% in 1995. The adjustment for work on software to be embedded and/or sold also rises over time, from an overall deduction of 34% in 1981 to 43% in 1995. The deduction rises through time because employment of programmers and systems analysts is increasingly concentrated in the computer service industries, where almost all their labour cost is removed to avoid double-count. The 50% reduction to avoid counting routine, non-development work is made in each year,⁴¹ as is the 50% add-on for other non-labour costs. This add-on is fixed for 1981-1997, but will be updated for more recent years on the latest SCS results with annual revisions of the accounts.

Inter-censal estimates are obtained by straight-line interpolation, between benchmarks, of the ratio of own-account investment to labour income by industry, with the interpolated ratios then applied to labour income. For 1996 and 1997, the 1995 benchmark ratios are used. Thereafter, business own-account investment is raised from 1997 levels on annual Labour Force Survey (LFS) data on the growth of aggregate earnings of programmers and systems analysts employed outside of the public sector and, to avoid double-count, outside of the computer service industries.⁴² Own-account investment in the government sector is grown on the earnings of programmers and systems analysts employed there. This assumes a fixed distribution of own-account investment across industries within each sector, to be revised when the 2001 Census data become available.

6.2 Annual national estimates by industry

Annual estimates of supply and intermediate use have been built up from the commodity level for each industry in the I/O Accounts for 1981-1998. ‘Software products development’, a service commodity in I/O which formerly covered only pre-packaged software, is redefined to include pre-packaged software, custom-design (formerly part of the commodity ‘professional and processing computer services’) and own-account software (a new entry in the I/O system). In addition, some traded software is retained under the commodities ‘recorded media’ and ‘royalties and license fees’.⁴³ Software purchased by businesses that is embedded in hardware and re-sold continues to be treated as intermediate consumption of the commodity ‘software products development’, and is allocated entirely to the computer services industries (on software expenses reported to the SCS).

The amounts now recorded as investment in ‘software products development’ for each business sector industry start from pre-packaged and custom-design software expensed. For 1981-1997, this is re-classified to investment from intermediate use for each industry (as estimated under the former accounts’ definition).⁴⁴ For 1998, its industry distribution is based on SCS data on sales by client industry. Added to this is capitalized software. For 1981-1997, this is re-classified from hardware to software investment, and allocated across industries on the distribution for computer hardware investment. For 1998, the CAPEX estimates for capital spending on software by industry (and province/territory) are classified directly as software investment. Last, own-account software, which is determined at a detailed industry (and provincial/territorial) level, is added. Software retained under ‘recorded media’ and ‘royalties and license fees’ is also re-classified from intermediate use to investment. It is allocated across industries on the distribution of investment in pre-packaged software, in the former case, and on the industry distribution of software royalty payments from BOP, in the latter.⁴⁵

Estimates of software investment for government sub-sectors for 1997 and 1998 are based on administrative data on software purchases, SCS data on industry sales to government and capital spending on software (CAPEX). Each sub-sector’s share of government software purchases over this period is assumed to hold back to 1981. These shares are determined separately for expensed pre-packaged software, expensed custom-software, and capitalized purchases of software. They are applied to government sector capital and non-capital spending on pre-packaged and custom-design software (discussed earlier) over 1981-1996 to arrive at sub-sector details for investment in purchased software. Own-account software investment by government 1981-1998 is built up from the detailed sub-sector level, as discussed earlier.

6.3 Annual provincial/territorial estimates

Annual estimates for software investment at current prices are also built up by province and territory, as well as by industry, in the I/O Accounts starting with 1997 and presently up to 1998. These estimates are carried back to 1981 and forward to 2000 in order to incorporate software investment in the Provincial Economic Accounts. No deduction is required from hardware investment for software already capitalized in the PEA, because these accounts carry only aggregated information on investment by broad asset category. Instead, only a net addition is made to investment in machinery and equipment for the software expensed by organizations (software capitalized by organizations is already included as investment in machinery and equipment).

The national I/O industry benchmarks for purchased software that is expensed (i.e., software investment net of own-account and software already capitalized, and excluding software that continues to be treated as intermediate) are distributed on industry GDP by province and territory, for 1984-1996. The 1984 results are carried back to 1981 on labour income by industry, province and territory, and constrained to meet the national I/O industry benchmarks for 1981-1983.⁴⁶ The 1998 benchmarks are carried forward to 2000 on real GDP by industry, province and territory, and constrained to satisfy the national estimates for the business and government sectors.

In the case of own-account software, the national estimates are allocated across the provinces and territories on the provincial/territorial distribution of wages and salaries of computer programmer and systems analysts within each industry. The distribution for each industry is based on census benchmarks for 1980, 1985, 1990 and 1995, interpolated between adjacent census benchmarks, for inter-censal years, and follows the 1995 distribution in the post-censal years through to 1998. For 1999-2000, LFS data on the aggregate earnings of computer programmers and systems analysts in the business sector (excluding the computer services industries) and in the government sector, by province, are used to distribute the national estimates for business and government own-account software investment.⁴⁷ These provincial distributions will be updated in the future with the results from the 2001 census.

Estimates at constant 1997 prices are arrived at by deflating the net addition to investment in machinery and equipment for software at current prices in each province/territory by the national implicit price index for all three types of software combined. There is room here for improvement in the future with the incorporation of provincial/territorial prices for each type of software.

6.4 Quarterly national estimates by sector

For 1981 to 1996, investment in purchased software and software already capitalized are both patterned after quarterly hardware investment, by sector. This is the case with the current and constant price, seasonally adjusted and unadjusted series, where the corresponding hardware series is used as a quarterly distributor. Regarding own-account, at current and constant prices, investment is equally distributed across the quarters within the year for both business and government (resulting in no seasonal pattern here). Quarterly price indexes, seasonally adjusted and unadjusted, are derived implicitly from the corresponding current and constant dollar software investment series.

From 1997, and reflecting the greater availability of quarterly data, the methodology is quite different. Rather than distributing the ‘bottom-line’ investment series (pre-packaged and custom), the various components of the commodity flow balances are estimated quarterly, on a seasonally adjusted and unadjusted basis. The quarterly software investment series fall out of the corresponding balances. Own-account investment is distributed on the quarterly earnings of programmers and systems analysts, by sector, available from the LFS starting in 1997 (introducing a seasonal pattern to the series). Last, the quarterly software price indexes are constructed and seasonally adjusted directly rather than being derived implicitly as in the back-period 1981-1996. The quarterly constant price series, seasonally adjusted and unadjusted, are derived implicitly from the corresponding current dollar software investment series and the software price indexes.

The software trade components of the pre-packaged and custom-design commodity flow balances are available quarterly, up to the current quarter. For other components, however, the estimates are obtained by distributing the annual on some quarterly indicator variable or by assuming that annual relationships between variables hold on a quarterly basis. Domestic production of pre-packaged software, for instance, is distributed on the quarterly payroll of the software publishing industry from SEPH. Likewise, personal expenditure on pre-packaged software, domestic production and exports of custom software, are distributed on the quarterly patterns of related indicators (see Tables A.3 and B.3). The quarterly investment in pre-packaged software of government and business, on the other hand, relies on the assumption that each sector maintains its annual share during each quarter. Likewise, the quarterly series on margins, the adjustment for embedded custom software and software already capitalized are all obtained on the assumption of fixed ratios from the most recent annual benchmarks (see Tables A.3 and B.3). These ratios are updated as new benchmark data become available.

Last, the sources and methods for 2000 (as detailed in Tables A.3-E.3) carry forward to the current quarter of 2001. Some series are available for the current quarter (e.g., software trade), others are brought forward either on the indicator series used to obtain their quarterly distribution in 2000 or on the assumption that relationships between variables observed in 2000 hold into 2001.

 

Table of contents