«Jonathan Haskel Imperial College Business School, CEPR, Ceriba, IZA and UK-IRC Keywords: copyright, IP, innovation, knowledge, investment, ...»
Note to figure: Data for the dashed line taken from Figure 17.2 of the UK Film Council Statistical Yearbook, 2009, and represents all film production spending that took place in the UK, but not constrained to UK-certified films. It therefore excludes the expenditure that took place in say, a coproducing country. However, the reason that the values implied by the dashed line in are on occasion higher than those of the dotted line in Figure 11, is that it includes expenditure on the production of films that either did not apply for, or meet the requirements, for UK certification (see Box 1). The series is compared with the time-distributed series for total production costs of UK-certified films in our datasets.
Source: own calculations and thenumbers.com.
Figure 13 presents Film GFCF as included in the National Accounts alongside our three final estimates: central, upper-bound and lower-bound. Our preferred estimate is the central estimate, which suggests that official estimates understate GFCF by a factor of eight in 2009.
The upper-bound suggests the understatement could be as a high as a factor of ten in 2009.
Note however that even our lower-bound, produced using quite conservative assumptions, implies that official GFCF is underestimated by a factor of around seven in 2009. Note that these data are based on production costs, ΣP(X)X, that is there has been no estimation of the monopolists mark-up so it is implicitly assumed that μ=1. The estimates are also consistent with our speculative discussion on UK production data earlier in the report.
Figure 7: Final Estimates (ΣPxX), CP £mns
Note to figure: Higher, lower and central estimates are our own estimates based on our custom dataset built on data from the-numbers.com, UKFC and BFI. For details on the ONS series, see section 1.1.
Source: own calculations and ONS 8.1.1. Alternative method of estimating ΣP(X)X, based on ASHE Although the dataset described above is used as our preferred estimate, an alternative source for calculating upstream input costs is available to us. A much used method in measuring GFCF for other intangible assets is to use the wages for occupations involved in asset creation (Haskel et al, 2009). So for instance, in the case of own-account organisational investment, this involves collecting data on the wages of managers and using an estimate of the proportion of their time that, on average, is spent on improving the organisation of business processes. A similar method is used for software in the National Accounts, although that method also involves the allocation of overheads to software-creation. We explore here whether, at least for certain asset types, a similar method can be used to estimate GFCF in artistic originals, using data for relevant creative occupations.
The main data source is the Annual Survey of Hours and Earnings (ASHE). The ASHE is a business survey sent out to employers, based on a random sample of the National Insurance numbers of employees. As well as limited companies, it is sent to partnerships and sole proprietors, so in theory should also provide information on freelance and self-employed workers that have set up their own small companies. A possible alternative is the Labour Force Survey (LFS), a household survey that asks respondents for information on occupation, industry and pay. In theory this could provide a cross-check on the data from ASHE.
However, there are limitations to the use of the LFS in this way. Firstly there is no stratification of respondents by either industry or occupation. Secondly the LFS does not record responses on the income of the self-employed, 30 and so its use would require the imputation of data on gross pay using the responses of employees. If for any reason the characteristics of employees and the self-employed in that field differed, the final estimate would be biased. In the context of this project it turned out that these issues were insurmountable, and so we use ASHE as our sole source.
Before we identify occupations involved in the creation of motion picture originals, by inspecting the Standard Occupational Classification (SOC), it is worth remembering exactly what we are trying to measure. ASHE gives us data on income by industry and occupation, not capital income that flows to the owners of originals. Therefore we are not seeking to generate an estimate of P(R)R or P(N)N, instead we are looking to measure upstream labour input costs, ΣP(L)L, which we can use to inform an estimate on total upstream input costs, ΣP(X)X. Table 6 provides a summary of the occupations that we consider to be involved in the upstream production of film.
The reason is conceptual as well as practical. The income of the self-employed includes a return for both labour and capital. In the National Accounts this is referred to as Mixed Income. At an aggregate level Mixed Income can be split using data on operating surplus and compensation of employees.
However, splitting at the person-level would be much more complex and it is likely that many selfemployed workers would not even recognise such a distinction within their income. In any case, LFS data on the incomes of the self-employed is not available, and Mixed Income in the National Accounts is estimated using administrative data from HMRC. Imputing pay for the self-employed based on that of employees of the same occupation would not be appropriate if it is likely that they have significantly different characteristics. In the case of occupations involved in artistic creation this seems likely, and employment status likely reflects the market and negotiating power of the artist and their relationship with funders.
Note to table: We ensure that no occupations already used in the calculation of investment for other intangible assets are used, including managers. Workers recorded in industries dominated by the public sector (defined as Public Admin & Defence (L), Education (M) and Health (N)) are also excluded so our final estimates are reflective of our definition of the market sector (A-K & OP).
To ensure no double counting with any other types of artistic original or knowledge asset, we only include those workers recorded in the film industry. As can be inferred from Table 6,
the main data issues are:
- a lack of fine detail in both the industrial and occupational classification, resulting in the inability to distinguish between assets e.g. freelance script-writers that work in either film, or television, or possibly both; or between and scriptwriters and other types of writers recorded in ‘Artistic & Literary Creation’
- a lack of time-use information associated with labour input in creative occupations e.g. how much of labour input goes toward asset creation (investment) and how goes towards the production of intermediate goods
- a limited dataset that only extends back to 1997 Data on wages do not provide us with estimates of upstream input costs since labour is not the only cost in artistic creation. We scale up the data by allocating overheads, based on the published breakdown of intermediate purchases at detailed industry-level as recorded by the
ABI. Following a similar methodology to that used by the ONS for the capitalisation of ownaccount software (Chamberlin et al, 2007), we calculate non-labour costs as follows:
Non-labour costs of asset creation 31 = Total Purchases of Goods and Services
- Purchases of goods for resale without further processing
- Purchases of energy and water products for own consumption
- Purchases of Road Transport Services
- Purchases of Computer Services
- Purchases of Advertising and Marketing Services
- Purchases of Telecommunications services
- Commercial insurance premiums paid + Total Taxes, Duties and Levies Paid
- Total stocks and work in progress at end of year Purchases of goods for resale without further processing are deducted as by definition they are not used as an input to the asset creation. Purchases of Road Transport Services are deducted because we also assume that these are, in the main, not an input to asset production.
Purchases of Computer Services are likely an input to creation, but software is an already capitalised asset, therefore we deduct these purchases to avoid double-counting. We deduct Advertising and Marketing Services, firstly because they are not an input to the production itself, more to the distribution of the asset, and secondly because branding is capitalised as a separate asset in the intangibles framework, and so again we wish to avoid double-counting.
We also deduct Purchases of Telecommunications Services and the payment of Commercial Insurance premiums for conservatism, since they are likely a considerable proportion of any other production that is not UK film asset creation. In addition we add on paid taxes and we deduct the value of stocks and works-in-progress at the end of the year. The reasoning for the deduction of stocks is that we assume that these are largely inventories of unused intermediate The ONS methodology is based on a view of the inputs to production of own-account software. The composition of deductions is therefore slightly different since we are considering the production of film originals. Additionally, the ONS adjust for consumption of fixed capital or depreciation. Data on depreciation at a detailed industry-level are no longer published as part of the ABI release and therefore we are unable to use them to adjust our overhead factors. However, any impact on the final estimate would have been limited.
goods. To account for remaining input costs, we use these data described to derive a ratio of other costs to employment costs, which we use as a factor, γ.
ΣP(X)X = γ(wN) For the industry ‘Motion Picture and Video Production’ (SIC 92.11) our average estimate (2001-07) of the ratio of non-employment costs to employment costs, derived from published ABI data, is 2.28. Adding one to account for labour costs gives us an estimate for γ of 3.28.
Unfortunately we do not have any information on time-use in this industry in order to adjust our resulting estimate of ΣP(X)X, presented in Figure 14. It is for this reason that we expect this series to be an overestimate of UK GFCF in this asset type, and more reflective of UKlocated production, rather than UK-owned asset creation.
Figure 14 shows the resulting estimates. Data for 2008 is around £500m. This exceeds the estimate for the ONS but is closer to our estimate based on the production budgets of UKowned films. Note that a time-use assumption of 50% would result in estimates close to those from our preferred source.
Figure 8: Film Originals (PxX using data on labour input from ASHE and the ABI), CP (£mn) Source: Our calculations, based on ASHE
8.2. PrR, Downstream Rental Payments. Film Royalty payments and licence fees flow to the owners of asset rights (IPRs) from downstream users. A relatively complete dataset on these rentals, along with information on whether they are one-off, intermediate or long-term payments would allow an estimation of investment as the NPV of the stream of royalties. At present we have no access to data on the rental payments made by users and distributed to the owners of UK Film Originals. In the case of Film, relevant payments would include those for cinema projection, DVD production, television broadcasting, and merchandise, among others.
8.3. P(N)N, Upstream Output. Film As explained and shown in Figure 5, data structured to the SIC does not match the upstream/downstream distinction used in our model. Therefore we feel that estimates based on industry aggregates are inferior to those from our data on the universe of UK films, but industry data are readily available from the ABI. The following text provides further insight into the structure of industries involved in asset creation, a description of what industry data relate to, and discussion of some issues presented by the use of industry aggregates.
Additional data is also provided in Appendix 2.
In practice, data for the film industry are broken down into data on production, distribution (studios) and projection (cinemas). In a sense, the upstream is actually made up of both Production (SIC 92.11) and Distribution (SIC 92.12). The former carries out production in return for a payment from the owner of the final asset, usually the studio located in the Distribution industry. At the end of production, the production company hands over the final asset to the studio, along with the associated rights to commercialise the asset and receive revenues from the downstream industries. In effect the studio is outsourcing production activity. Therefore the innovator is the studio, since it is they who fund and own the final original. In practice the production company will sometimes be a subsidiary of the studio.
Thus on the face of it, it might seem possible to treat the output of the production industry as an estimate of ΣP(X)X, and output of the distribution industry as an estimate of P(N)N (plus some costs for advertising and distribution). The output of cinemas and the rest of the downstream, including DVD production, TV broadcasting etc., is P(Y)Y.
Figure 15 charts gross output for both the Distribution and Production industry. As can be seen they do have a similar profile, but revenues are significantly higher in distribution, as we would expect, since they should reflect their market power as owners of IPRs. The chart also includes an approximation of μ, implied by the data on the revenues of production and distribution.
Figure 9: Output of the Film Industry, ABI, (CP £mns) 2,500 3.5 2,000
0.5 Source: Published aggregates from the ABI and Services Trades Sector Review (pre-94) The implied series for μ is volatile, which is understandable since we would expect each singular film original to differ greatly in commercial value. On average, the value for μ between 1992 and 2007 is 2.26, supporting our point that whether or not one feels an estimate for μ should be implicit in the final estimates, its impact is important and it is currently treated inconsistently in both the OECD recommendations and the practices of National Statistical Institutes (NSIs). We should note that we consider this to be a poor approximation of μ, since the ABI data includes UK-located production of foreign assets (exports) and data on revenues are inconsistent in the sense that they are a mix of royalties, asset purchases and fees.