Measuring obsolescence using big data

Over the last few years, interest in measuring the stock of intangible capital as grown very substantially. But estimation of both the stock of any type of capital and its contribution to output requires knowledge of the rate at which capital is scrapped. This is generally calculated using an assumption of geometric decay at a known depreciation rate.

For much intangible capital, the assumptions made about the depreciation rate may be relatively arbitrary. It is also not clear that the concept of depreciation is relevant to types of capital such as software which do not decay but may become obsolete.  Estimates of rates of capital scrapping based on observed behaviour would be very valuable.

This project attempts to measure the rate of obsolescence of software and creative originals by looking at internet searches for them. The rate at which internet searches decay over time will be used as an indicator of the rate of obsolescence.

We will use Google Trends to generate estimates of the average depreciation of software and creative originals. One can view depreciation as a form of obsolescence. As software and creative originals become obsolete, their ability to generate future output/revenues diminishes. Google Trends shows the number of searches, offering a practical way of gauging the popularity of an intangible asset. If one assumes that the rate of the decline in search results is proportional to obsolescence, then it might be possible to estimate an average depreciation rate for intangible assets using this data.

The research uses Python (or RStudio) to scrape Google Trend results for all software, films, music, books, and TV series released from 2002 to 2020. A non-linear regression is then fitted to the data, and the slope parameter interpreted as the rate of obsolescence of the given asset. Ultimately, this can be interpreted as a form of depreciation.

The aim is to generate a weighted average depreciation rate for the asset types mentioned above. After generating estimates of depreciation rates for various types of intangible assets, we will analyse the impact of changes in these parameters to levels of capital stock and net domestic product. Building on 2021 work by O’Mahony and Weale, we will then develop indices of depreciation and net capital services which reflect its findings.

Future analysis could then look at specific types of software that are currently popular and try to gauge to what extent their development could be linked to earlier developments in software. This would use text analysis on web scraped data. It may also be possible to extend the analysis to look at obsolescence of R&D (Research and Development) spending by examining patent data.

This project will support development of the evidence base to understand depreciation rates and asset lives of software and artistic originals, which are intangible assets recorded within the National Accounts. It will provide us with a more comprehensive measure of capital depreciation by bringing in novel measures that have not been considered or captured in estimates to date.

It will also provide users of the National Accounts with better data, facilitating an understanding of what is happening to the economy. It may also stimulate further research based on similar methods.

EM2024 presentation: Contributed session K: Measuring output and productivity, John Lourenze Poquiz ‘Measuring the depreciation of intangibles using Search Volume Data: A proof of concept’, 16 May 2024.

Poquiz, J. L. ‘Household production’ New Measures of the Economy Workshop, The Digital Economy Lab, Stanford University, 19 March 2024

Poquiz, J.L. ‘Measuring the Value of Free Digital Goods’ ESCoE Economic Measurement Webinar Series, 2 Nov 2023 (Discussant Carmit Schwartz, UNSW).

Poquiz, J.L. (2023) ‘Measuring the Value of Free Digital Goods’ ESCoE Discussion Paper Series, ESCoE DP 2023-16

Poquiz, J.L., Mahony, M. and Balisacan, F.H.M. ‘The Value of Free Digital Goods’ National Bureau of Economic Research SI 2023 Conference on Research in Income and Wealth, 17-18 July 2023