How Data Fuels the New Economy
By Cai Hui Lien.
Introduction
Data is the new oil. In the digital economy, the abundance of data has reached astronomical proportions, with digital data forecasted to reach 181 billion trillion bytes in 2025 (IDC & Statista, 2021). In concrete terms, this amounts to a stack of reports reaching from Earth to beyond Pluto (World Bank, 2016). Giant leaps in data collection, storage and analysis capabilities have engendered the exponential growth of big data in the last decade, resulting in fundamental shifts in firms’ operations and the workings of the economy.
Big data technologies play an integral part in this process by increasing the productivity of large data sets and reducing the diminishing returns to data. The value of data hinges on extracting meaning from data and its economic value is only realised when the product of data science is applied to a commercial context. As a corollary, the growth of data is closely tracked by the overall growth of the data science field. At the same time, the provision of zero-price products (such as apps) offers an estimate of the economic value companies derive from data.
The burgeoning interest in data thus has profound implications on both the microeconomic and macroeconomic levels. In this article, I examine how the newfound role of data and the dynamics of the data economy have shaped and will continue to shape economic growth’s past and future.
Past literature
Data feeds back into economic growth in two key ways (Veldkamp and Chung, 2019). Firstly, the predictive power of data allows it to be used for forecasting. In the short run, transaction-based data optimises production by reducing information frictions, increasing productivity and output from accumulating production experience through a learning-by-doing model (Arrow, 1962). However, diminishing returns set in when this causal analysis is extended to the long run because prediction errors naturally have a zero lower bound. Moreover, even if a technology with perfect predictive power exists, the inherent randomness of future events means that its predictions cannot be wholly accurate. Second, data can be seen as equivalent to new technology. As a factor input in research and development, data accelerates innovation with increasing returns through a data feedback loop within a firm — innovations generate new data, which generates innovations. Innovation, in turn, drives economic growth (Schumpeter, 1911).
Empirical evidence has pointed to the bounded nature of productivity growth through learning-by-doing (Thompson, 2008), as in the former case where data is used for forecasting. In the latter case, however, the magnitude of the economic transformation brought about by data needs to be revised. An important caveat is that data does not value-add but requires complementary data analytics and software. This is akin to the role of new manufacturing processes in the industrial revolution, mirrored by the industrialisation of knowledge creation with big data technologies (Abis and Veldkamp, 2021). In the respective cases, new technologies drastically reduce the diminishing returns to capital and data, leading to increased productivity. More precisely, the ability of data to drive growth through research and development relates to the idea that innovation is a combinatorial process where new inventions surface from novel combinations of ideas. General purpose technologies such as machine learning have the potential to facilitate this process, with past precedence, such as the Internet, that did so through increasing connectivity (Carlsson, 2004). Hence, the change such technologies bring to the knowledge production frontier has the potential to bring about an economic singularity (Agrawal et al., 2018).
Methodology
To fit data into an existing growth model, the two fundamental attributes of an economic good — its degree of rivalry and excludability — have to be considered. Data is non-rivalrous, given that it can be used by multiple firms simultaneously. It is also excludable since firms keep the data they own private. In addition, as previously noted, data can generate increasing returns. This arises from data as a by-product of economic activity since each transaction generates information used for product innovation and optimising production. As firms with higher-quality goods invest more, produce more, and sell more, the accumulation of aggregate knowledge accelerates (Farboodi and Veldkamp, 2021). Combining these properties, there is a clear parallel to the role of ideas in endogenous growth models.
For analysis, I refer to Romer’s endogenous growth model (Romer, 1990) to understand how data contributes to long-run economic growth. Compared to exogenous growth models that take technological progress — a determinant of growth — as exogenously given, endogenous growth models attempt to account for technological advancement.
Using a simplified model — omitting capital entirely — (Jones, 2021):
where Y_t is output, A_t is the stock of knowledge, and L_{yt} is the amount of labour devoted to output.
The flow of new knowledge is Delta A_{t+}, and its production is given by
where z bar is the productivity of idea generation, and L_{at} is the amount of human capital devoted to idea generation, that is, L_{at} = \bar{lL}, where \bar l is the fraction of labour devoted to idea generation and L_{yt}=(1-\bar l)\bar L). Thus:
Since parameters \bar z, \bar l, and \bar L are taken to be constant, rewriting the long-run growth rate of the economy is given by
In Romer’s model, the stock of knowledge A generates increasing returns. The cumulative nature of idea generation — or research — arises from intertemporal knowledge spillovers that increase the productivity of future research.
Following this line of logic, the accumulation of data is analogous to ideas in the production function. An increase in data increases the productivity of idea generation \bar Z when complemented with analytics tools. Reiterating Abis and Veldkamp (2021) findings on the industrialisation[1] of knowledge production, big data technologies change the data-labour ratio in knowledge creation, raising the productivity of analysing large datasets, the parameter \bar z. In doing so, the growth rate of knowledge accumulation and hence the long-run growth permanently increases due to the accumulation of data.
While it is difficult to isolate the individual impact of data growth on overall economic growth trajectories, there are several indicators of data’s tangible impact on economic growth. The first is the increase in the share of labour devoted to knowledge production, or more relevantly, experts in the data science field. A straightforward indicator is the growth of jobs and salaries in the data science and analytics field, in line with the explosion of data. On job platforms, there has been an increasing trend in jobs posting related to data science (Figure 1).
Source: Van der Aalst (2014)
The logical result from the increased demand for data scientists is a corresponding increase in parameter l, the share of labour devoted to idea generation. More labour working on idea generation leads to an increase in the growth rate of idea generation and hence the economy’s long-run growth rate.
Another notable result of the impact of data on firms’ profits is the rapid proliferation of zero-price goods, for example, digital platforms. Such goods have economic value to users — in the United Kingdom, hedonic regression estimates of the aggregate gross value derived by households from the consumption of video-conferencing, personal email, and online news were between £6.1 billion and £22.7 billion in 2020 alone (Poquiz, 2022). Yet, they are provided for free. This points to a new form of economic transaction, the implicit barter trade of data for free digital services, as firms generate revenue from owning users’ data. The Financial Times estimates that the average person retails for less than a dollar (Steel et al., 2013), varying depending on an individual’s characteristics. Thus, because data has strong predictive power, firms see inherent economic value in possessing such data in the short term before diminishing returns dominate.
Conclusion and Limitations
Both examples paint a broad picture of the economic value data holds on both the macroeconomic and microeconomic levels. While the findings of the analysis may be broad, they are still significant and reflective of broader trends in the digital economy.
However, there are a few important caveats. Firstly, the assumption that the growth rate ratio is proportional to the amount of human capital devoted to idea creation is flawed. By that measure, an exponential increase in data accumulation and population should have led to explosive economic growth, but such a result is unobserved. It is thus more realistic to infer that other parameters, such as the availability of complements to data such as data analytics and skilled data analysts, constrain the link between data volume and idea creation. The actual ability to capitalise on data thus goes beyond data volume. Moreover, treating data like a commodity akin to gold and oil generalises across all sorts of data. How data is collected, managed, and processed leads to significant heterogeneity in its quality.
Thus the Romer model reinforces the argument that data, like ideas, charts the trajectory for sustained economic growth as it is not subject to decreasing returns like capital. Moreover, the properties of data as an essential asset may mean that firms choose to produce zero-price services for data, as in barter trade. To further extend this analysis and complement this idea, it might be useful to model data as a factor input that directly affects the output of firms.
This article was written by Cai Hui Lien, an economic research analyst within the Central Banking Society’s Economic Research Division. This article was reviewed by Yating Zhang, the present Head of Economic Research.
References
· Abis, S., & Veldkamp, L. (2020). The Changing Economics of Knowledge Production. Working paper.
· Agrawal, A., McHale, J., & Oettl, A. (2018). Finding Needles in Haystacks: Artificial Intelligence and Recombinant Growth, No 24541, NBER Working Papers, National Bureau of Economic Research, Inc.
· Arrow, K. J. (1962). The Economic Implications of Learning by Doing. The Review of Economic Studies, 29(3), 155–173. https://doi.org/10.2307/2295952
· Carlsson, B. (2004). The Digital Economy: What is new and what is not?. Structural Change and Economic Dynamics. 15(3). 245–264.
· Farboodi, M. and Veldkamp, L. (2021). A Model of the Data Economy. NBER Working Paper.
· IDC, & Statista. (2021). Volume of data/information created, captured, copied, and consumed worldwide from 2010 to 2020, with forecasts from 2021 to 2025 (in zettabytes) [Graph]. In Statista. Retrieved January 20, 2023, from https://www.statista.com/statistics/871513/worldwide-data-created/
· Jones, C. (2021). Macroeconomics. W.W. Norton & Company.
· Pouquiz, J. L. (2022). Measuring the value of free digital goods.
· Romer, P. M. (1990). Endogenous Technological Change, Journal of Political Economy 98, S71–S102
· Schumpeter, J., 1911 (English edition 1934). The Theory of Economic Development. Harvard Economic Studies, vol. XLVI. Harvard University Press, Cambridge, MA.
· Steel, E., Locke, L., Cadman, E. & Freese, B. (2013). How much is your personal data worth? The Financial Times. https://ig.ft.com/how-much-is-your-personal-data-worth/#:~:text=The%20average%20person's%20data%20often,or%20%240.50%20per%201%2C000%20people.'
· Thompson, P. (2010). Learning by doing. Hall, B., Rosenberg, N. (Eds.), Handbook of the Economics of Innovation, vol. 01. Elsevier/North-Holland, pp. 429–476 (Chapter 10)
· Van der Aalst, W.M.P. (2014). Data Scientist: The Engineer of the Future. Enterprise Interoperability, 6th ed. pp. 13–26. Springer.
· Veldkamp, L., & Chung, C. (2019). Data and the aggregate economy. Journal of Economic Literature.
· World Bank. (2016). World Development Report 2016: Digital Dividends. Washington, DC: World Bank. doi:10.1596/978–1–4648–0671–1
[1] Industrialisation can be interpreted as using new production technologies with less human input and less diminishing returns to capital.