The Innovation That Never Reaches the Books

Services accounted for 66.2% of world value added in 2024 according to the World Bank’s development indicators, against 65.3% the year before and levels near 62% at the close of the 1990s. The figure is familiar and usually functions as a closing argument: the economy has tertiarised, therefore services deserve study. We are interested in a less comfortable reading. If two thirds of output is generated in activities whose innovative effort is booked as current expense rather than investment, then the instruments we use to measure innovation - built around research and development spending - are observing a shrinking fraction of the phenomenon. Our thesis is that innovation in services is, before it is a theoretical problem, an accounting boundary problem, and that correcting that boundary changes two things at once which are rarely discussed together: the technology policy diagnosis and the valuation of firms.

Where the mismatch comes from

Since the early 2000s the literature on service innovation has organised itself around three positions. The assimilation approach holds that service innovation should be studied with the tools developed for manufacturing, treating the new or improved service as the analogue of the new or improved product. Rubalcaba (2013) notes that this view assigns services a subordinate role in the innovation process, consistent with Pavitt’s (1984) taxonomy, which classifies them as supplier-dominated sectors: recipients of innovation generated elsewhere. The demarcation approach answers that non-technological dimensions are constitutive of the service and require a dedicated apparatus, given its intangible nature, the difficulty of storing it and protecting it through conventional intellectual property mechanisms, and the high degree of customer interaction its production demands (Coombs and Miles, 2000). The synthesis approach, finally, observes that within-sector differences are often sharper than differences between manufacturing and services taken as aggregates (Gallouj, 1998), and that the activity dimension of any sector always contains service components.

We side with the third position, but we think it stopped too early. If the product-service boundary blurs - the servitisation the literature describes, visible in services on the good, services complementary to the good, and services that substitute for owning the good - then the problem ceases to be one of classification and becomes one of accounting recognition. And there the answer is uncomfortable: under prevailing accounting frameworks, internally developed processes, platforms and organisational capabilities qualify for capitalisation only under restrictive conditions, so most of the innovative effort of a service firm lands in selling and administrative expense or directly in cost of service. This is not an error by accountants. It is the consistent application of a recognition criterion designed for assets with identifiable and separable future benefits, a condition that organisational capital satisfies poorly by construction.

Knowledge, learning and the appropriability problem

The conceptual substrate has been available for a long time. Johnson and Lundvall (1994) argue that the production structure works as the frame for the routine learning that takes place within the system, and that these learning processes tend to reinforce the existing structure: innovation systems tend to specialise rather than diversify. The second axis of their argument is the institutional frame, understood as the structure of routines, norms, rules and laws that governs behaviour and determines personal relations, and which bears on how interactive learning occurs and therefore on the pace and direction of innovation. Translated into our problem: if the learning that sustains service innovation is embedded in routines, client relationships and people’s capabilities rather than in technical artefacts, we should expect it to be hard to measure and hard to appropriate, and those two difficulties share a root.

Schumpeter (1934) defined the competition that matters as the one that introduces new combinations. The observation keeps its edge once we accept that the combination can be organisational. A firm that redesigns how it delivers a service and captures extraordinary rent during the window in which competitors fail to imitate it is doing precisely what Schumpeter described, with the difference that its innovative spending will leave no trace in any R&D statistic and its intangible asset will not appear on the balance sheet. This is where the argument becomes operational rather than merely descriptive.

The case that is usually told wrong

The canonical service innovation narrative is Netflix against Blockbuster: designing a rental service without late fees is said to have rendered the traditional model obsolete. The story is accurate as description and misleading as mechanism, and it is worth saying so because it appears in nearly all the popular literature, including an earlier version of this text that I wrote in 2022.

Three corrections. First, the decisive displacement was not the removal of late fees but the shift to streaming, and that transition was capital intensive: content delivery infrastructure, licensing agreements and, later, original production, which is in fact capitalised as a content asset. The service innovation worked as an entry wedge, not as a sustained source of rent. Second, Blockbuster’s collapse carried a capital structure component that is hard to separate from the competitive one: a leveraged company has less room to cannibalise its own cash flow than one financed with growth equity. Third, and this matters most to us, the asset that actually sustains Netflix’s margin today - the recommendation system, the ability to forecast content demand, the production processes - is precisely the kind of asset that accounting does not recognise and innovation statistics do not capture. The case does not illustrate the power of service innovation. It illustrates that we do not know how to measure it.

This is also the fact that sits awkwardly with our own thesis. If organisational capital were the principal source of rent, we should observe margin persistence in service firms with neither physical capital nor proprietary content, and that does not hold generally: a large share of service innovations is imitated quickly, precisely because the entry barrier is low. Extraordinary rent from service innovation appears to be, on average, more intense and shorter lived than rent from protectable technological innovation. That asymmetry is poorly documented, and it is a genuine evidence gap: no harmonised series of innovative effort in services exists that is comparable to R&D spending, and measurement manuals capture non-technological innovation through self-reported survey rather than through booked expenditure.

The conjecture: capitalise what is currently expensed

Here we move from review to a proposal of our own, untested in this text and best read as a research programme rather than a result.

The asset pricing literature solved, for its own purposes, a problem formally identical to the one the service innovation literature has been posing for twenty years. Eisfeldt and Papanikolaou (2013) build a measure of organisation capital by capitalising selling and administrative expense through perpetual inventory, and document that firms holding a high stock of that capital display differential expected returns. Peters and Taylor (2017) formalise a measure of intangible capital that adds knowledge capital, accumulated from R&D spending, to organisation capital, estimated as a fraction of selling and administrative expense, and show that the investment-q relation behaves better once intangibles enter the denominator. Our conjecture is that the same adjustment, applied to a different purpose, resolves the measurement problem in service innovation.

The specification is straightforward. We define adjusted innovative intensity as R&D spending plus a fraction of selling and administrative expense, divided by sales, and compare it against the conventional R&D-to-sales metric across a panel of listed firms classified by sector. The first falsifiable prediction is that the sector ranking changes materially: sectors that look weakly innovative under the conventional metric rise significantly under the adjusted one, and the magnitude of the reordering grows with the service share of the sector. The second is that the gap between the two metrics, taken as a proxy for measurement error, correlates negatively with the coverage that official innovation statistics report for that sector. The third, more ambitious, is that a factor built on adjusted innovative intensity within the services universe captures return variation that the conventional R&D factor does not.

The identification problems are serious. The capitalisable fraction of selling and administrative expense is an assumption rather than a measurement, and the literature converges on a value by convention rather than by structural estimation. Reporting practice for that line differs by sector and by jurisdiction, which introduces bias into exactly the comparison we care about. The depreciation rate of organisation capital is a free parameter with first-order impact on the estimated stock. And applying the exercise to Latin American markets, where it would carry the most policy value, runs into small panels, heterogeneous reporting and short series. None of this invalidates the programme. It does require presenting it as a conjecture.

Scope

We deliberately leave outside this work the question of which institutional framework best promotes service innovation, a matter of political economy rather than measurement, and the analysis of service innovation within global value chains, where the relevant unit of analysis ceases to be the firm. Nor do we address whether the knowledge economy is confined to the most advanced fringes of world capitalism or can extend to developing regions, a question we regard as open and more consequential than it is usually taken to be.

---

References

Barletta, F., Suárez, D. and Yoguel, G. (2013). Innovación en servicios: un aporte a la discusión conceptual y metodológica, 61-74.

Coombs, R. and Miles, I. (2000). Innovation, measurement and services. In J. S. Metcalfe and I. Miles (eds.), Innovation systems in the service economy. Kluwer.

Eisfeldt, A. L. and Papanikolaou, D. (2013). Organization capital and the cross-section of expected returns. The Journal of Finance, 68(4), 1365-1406.

Gallouj, F. (1998). Innovating in reverse: services and the reverse product cycle. European Journal of Innovation Management, 1, 123-138.

Johnson, B. and Lundvall, B. (1994). Sistemas nacionales de innovación y aprendizaje institucional. Comercio Exterior.

Pavitt, K. (1984). Sectoral patterns of technical change: towards a taxonomy and a theory. Research Policy, 13, 353-369.

Peters, R. H. and Taylor, L. A. (2017). Intangible capital and the investment-q relation. Journal of Financial Economics, 123(2), 251-272.

Rubalcaba, L. (2013). Innovation and the new service economy in Latin America and the Caribbean. Discussion Paper IDB-DP-291, Inter-American Development Bank.

Schumpeter, J. A. (1934). The theory of economic development. Harvard University Press.

World Bank. Services, value added (% of GDP), indicator NV.SRV.TOTL.ZS, World Development Indicators.