«UK Economic Performance: How Far Do Intangibles Count? Rebecca Riley and Catherine Robinson March 2011 INNODRIVE Working Paper No 14. The research ...»
(1) where and are the wage and labour input of intangible workers type IC in firm i at time t. Figures 1 illustrates the share of workers in occupations that we classify as “intangible” occupations. On average, ICT workers account for around 3½% of employment, R&D workers 6% of employment, and organizational workers 14% of employment (which can be disaggregated as 10% management workers and 4% marketing workers). In other words, our occupational classification implies that almost 1 in 4 workers, in the industries we consider, spends some of their time producing intangible capital goods. Figure 2 illustrates the hourly labour costs of these workers and contrasts these to the costs of other workers. Clearly, workers in “intangible” producing occupations are expensive, costing nearly twice a worker in other occupations. Basically, workers in “intangible” occupations are highly skilled and highly paid. Of the three intangible categories, workers in organizational occupations tend to 13 be the most costly. The break in figures 1 and 2 is to remind the reader of the discontinuity in SOC classifications between 2001 and 2002.
We capitalize these intangible investments for the firm according to the perpetual inventory
(2) where denotes the end of year stock of intangible capital type IC and denotes the depreciation rate. As described in Görzig et al. (2011), we assume starting stocks (for the year
before we first observe a firm in the data) are proportional to the sample average of intangible investment (discounted appropriately), ̅, for the firm:
where T is set to 100 and g is set to 0.02. Inevitably, with a relatively short span of firm-level data, our intangible capital stocks are sensitive to these initial assumptions. Depreciation rates for intangible capital are usually set substantially higher than the equivalent for tangible capital, i.e. intangible asset lives are assumed to be shorter than tangible asset lives. Informed by the literature we use:,, and (Table 3, Görzig et al. (2011)).
4.2 Scaling parameters We use two approaches to setting the three scaling parameters discussed above. The first of these, which we call the expenditure approach, is akin to the methodology typically adopted in deriving national estimates of intangible capital items that are based on the wages and salaries of particular workers. The second of these, the performance based approach, provides a robustness check on our main approach to estimating organizational capital.
To assess labour services that go towards the production of intangibles, we distinguish three types of labour input: R&D, ICT, and ORG-related personnel. We assume that only a fraction of workers in these occupations are engaged in the production of intangible capital goods; with the remainder of these workers engaged in current production (i.e. production of goods and services with a service life less than a year). Specifically, in the expenditure approach we assume that 70% of R&D workers‟ time, 50% of ICT workers‟ time, and 20% of organizational workers‟ time is spent on the production of intangible capital goods (i.e.
,, and ).9 To account for the capital services and materials that complement this labour in the production of intangible assets we scale the relevant labour expenditures with the cross-country10 average ratio of total production to labour costs
in the ICT, R&D and Business services sectors; SIC 72, 73 and 74 respectively. These are:
,, and (Table 5, Görzig et al. (2011)). Finally we assume that the productivity-wage gap of intangible type workers is zero so that for all types of intangibles.
As in much of the literature on intangibles these scaling parameters are based on simplifying assumptions and the evidence in support of some of these is as yet relatively scarce. The performance approach in INNODRIVE goes some of the way in gauging the robustness of these assumptions. Here we outline the basic estimating equation in the performance approach. The framework is discussed in more detail in Görzig et al. (2011). We assume a simple framework where firms produce intangible capital goods (own-account) in addition to final output. In producing these goods firms use tangible and intangible capital, as well as
labour services. We can write a stylized production function for the firm as:
where ̃ is value added adjusted to include intangible investment, is quality adjusted labour, and are intangible and tangible capital respectively. The rest are production function parameters. and are as defined in the previous section 4.1.
Note that in practice we have three types of intangible capital, rather than just the one in this illustration. This stylized production function can be likened to the production function underlying the standard growth-accounting framework used in macroeconomic studies that evaluate the contribution of intangible capital to productivity growth.
Labour is adjusted for the quality of intangible workers and is given by, where denotes labour in all other occupations than “intangible” occupations, and is the ratio of the marginal product of intangible workers to the marginal product of other workers. Note that this can be rearranged so that, where is the share of intangible workers in the firm‟s labour force. This specification allows us to evaluate the relative marginal products of workers in intangible occupations (following Hellerstein et al., 1999).11 Also, note that in our basic approach to measuring intangible investment we can write:, where is the ratio of the average (across firms) wage of intangible workers to the average wage of other workers, so that we have simply replaced in the expression for intangible investment, assuming that other workers are paid their marginal product. Finally, note that we can rewrite the expression for ̃ so that
above to formulate the following estimating equation12:
11 Although we essentially assume two production functions, one for the production of intangibles and one for the production of final output, it seems reasonable to evaluate the marginal product of intangible workers within a single production function setting as we do here because it is likely that workers‟ time would be distributed between activities to equate their marginal products between the two. The single production framework can also be justified if we assume that the two outputs can be exchanged at zero cost within the firm.
( ) (5) Using the estimated coefficients from this equation we can produce estimates of some of our scaling parameters. We derive the relative productivity of intangible workers to other workers as, so that our estimate of the ratio of the marginal product of intangible
Table 4 reiterates the scaling factors used in the expenditure approach for all three types of intangible assets, alongside estimates of these from the performance approach for R&D and ORG investment. The total scaling factors derived from a simple pooled OLS regression of the equation above do not differ substantially from the scaling parameters in the expenditure approach. The performance approach does suggest that the marginal products of R&D and ORG workers exceed their wages. For ORG workers, where there are good reasons to consider such a productivity-wage gap (Görzig et al., 2011), we explore this in more detail below.
5. INTANGIBLE CAPITAL IN UK FIRMSUsing the expenditure approach above we report our main estimates of intangible capital in UK firms. Next we provide some additional analysis of organizational capital constructed using a performance based approach.
5.1 Expenditure-based estimates of intangible capital in UK firms Table 5 provides a summary of the UK intangibles data characteristics, unweighted. In all cases, we divide through by sales to give an intensity measure comparable across firms. With respect to our organisation measures we see that the average value of compensation over the period is around 8 per cent of sales. In investment terms, this figure is much smaller – around 3 per cent of sales, however organisation capital represents around 10 per cent of sales. Note also that the median values are not dissimilar to the means, suggesting a relatively even distribution. R&D measures reveal around a 2 per cent mean for both compensation and investment for the period, thus on average compensation is relatively lower than for organisation workers. R&D capital is larger at 8.5 per cent of sales, reflecting its lower depreciation rate. Exploring the median values shows that there is much more variation in these data over time and sector, as we would expect. In the case of IT workers, as a proportion of sales, compensation, investment and capital are relatively small in comparison with the other intangibles, with ICT capital accounting for only 2 per cent of sales. Median values are significantly lower.
Overall, the table shows that on average across firms intangible investments account for around 6 per cent and intangible capital is approximately 21 per cent of sales, compared with 26 per cent for tangible capital. These are unweighted results, a mean across firms in the sample. Thus, the two forms of capital are comparable in magnitude. Our findings can be considered in the light of other UK studies however it is important to note that there are some marked differences in what the measures are designed to capture and construction.
Haskel et al (2011), incorporate both purchased and own account intangibles in their measure. Recent international analysis by Jona-Lasinio et al (2011) finds that in Anglo-Saxon countries (Ireland and the UK), intangible shares of GDP are around 8.9 per cent in 2005 which is noticeably larger than our 6 per cent, however ours is an unweighted mean and relates to own account intangibles only which are a sub-component of what these other studies capture.
In Figure 3 we present the intangible intensity of value added using the expenditure based approach to constructing intangible capital. Here data are weighted to be broadly nationally representative. Intensities here are constructed as a proportion of the newly revised GVA, 18 incorporating intangible investment. We see a slight increase in IT and organisation investment over the period. R&D investment over time is relatively sable, and both R&D and organisation investments are of similar magnitude. IT on the other hand is lowest, with an intensity of around 2 per cent of the new value added. Combined, intangible investments account for around 10 per cent of new value added, increasing their share very gradually over the 9 year period.
Table 6 provides correlations across firms of the capital measures with each other, including tangible capital (K). Whilst all are statistically significant and positively correlated, the greatest correlation is between IT and organisation capital (0.43). Given the nature of production, we would expect a positive and significant correlation and overall, the level of correlation is observed, but we feel that joint inclusion in econometric estimation is perfectly possible.
Figure 4 presents intangible capital intensity by broad sector. These data are averaged over the full 1998-2006 period. We see that Business Services (K) and Office Machinery and Precision equipment (DL) have the highest levels of intangibles. Composition of intangibles varies, R&D is more sector specific than organisation capital and IT capital tends to make the smallest contribution of the three assets. In Hotels and Restaurants (H), almost all intangible investment is organisation capital.
Figure 5 compares the UK under two different industrial structure scenarios – the existing one, and the EU average. Thus, the blue bars (to the right) indicate how the UK would look in terms of capital intensities were it to have an industrial structure of employment to the EU average. If the UK industrial structure looked more like the EU average manufacturing employment would be around a third higher. In terms of organisation and IT intensity, there would be very little difference although the UK is slightly more service intensive, which leads us to conclude that for these forms of capital, industrial structure is less influential. In the case of R&D, there is a small but clear difference between the UK as it is and the EU industrial structure. Were the UK to look more like the EU average, it is likely to have a higher R&D intensity. The biggest difference is in tangible capital intensity, although it is 19 worth noting that in terms of tangible capital we are only able to capture plant, machinery and equipment, not property. This may affect our findings in relation to tangible capital.
In the constructed measures of intangibles here we focus exclusively on own account intangibles, embodied in the wages of the occupational groups identified here. We therefore do not include those that are purchased such as advertising, software or copyright/licence costs.
Macro studies, for instance Haskel et al (2011) and Jona-Lasinio et al (2011) use as far as possible official data sources that are harmonized across countries to construct national measures of intangibles. For own account production they use Structure of Earnings Surveys (SES) from Eurostat and LFS data. Estimates of purchased intangibles are constructed from a variety of sources including Eurostat‟s Structural Business Statistics (SBS) and for R&D they use BERD data. Their core national accounts data are derived from EUKLEMS.
5.2 Performance-based estimates of organizational capital in UK firms The above expenditure approach can be criticised because, in reality, wages can at best be only a proxy for marginal product of workers. The returns to successful management, for example, may well be reflected in the value of the firm over and above the amount that managers are paid in terms of salary. Under such conditions, we need to consider how good our proxy of the relative productivity of intangible workers is. Following Hellerstein, Neumark and Troske (1999), who focused on the gender pay gap, we adopt a means of testing expenditure based measures of marginal product with the construction of a performance based measure of organisation capital.