When do firms want to share data?
How can we encourage more companies to share data? Of course, data intermediaries will likely play a pivotal role. But public-interest intermediaries must take a nuanced understanding of both commercial incentives and potential competition issues when designing their sharing policies and business models.
As readers of the ODI’s Medium publication, Canvas, we most likely agree that the troves of non-Personally Identifiable Information (PII) data collected and held by private firms have the potential to yield huge public benefits¹. Those benefits might be direct — for example, anonymised loyalty card data could help public health agencies target local obesity services. Or they might be indirect — data and insights from digital platforms can inspire commercial or public policy innovation.
And yet we are only too aware of how little data is actually shared. Public benefits are held back precisely because they are diverse and innovative, while the actual and perceived private cost to the firm of sharing is immediate and specific. In research for DCMS, London Economics estimated that the direct, out-of-pocket cost of sharing data was £760k for large firms². But this was dwarfed by the perceived loss of competitiveness — valued at £1.6m. Whether we feel these numbers are accurate or not, it is clear that strategic competitive reasons are a significant, if not the most significant, reason that firms do not share more data.
If solving this problem were not hard enough, there would be an increasing (albeit smaller) competition risk if firms shared too much data. Take tourism, for example. Firms and public bodies might agree that sharing sales data by nationality helps predict local demand for accommodation, events and restaurants. Demand is volatile, and each culture looks for something different. But, of course, firms could just as easily use the same data to coordinate prices and maintain margins. This risk gets more pronounced as firms become more sophisticated and data-driven. Price-setting algorithms, with all the data that they can muster, have the potential to coordinate pricing and, even without firms knowing, induce machine-to-machine collusion³. Higher prices hurt consumers.
Without a doubt, the relationship between data sharing and competition is complex. More data can be good or bad for competition. It depends⁴. Firms might use consumer data to create better products. Or they might bug people with too much advertising. Similarly, data can increase efficiency today, but more efficient firms can become too dominant and decrease competition tomorrow. Which effect wins out depends on the context. Greater consumer control over PII data sharing greatly helps⁵. So with Open Banking, for example, consumers can share their financial data in exchange for better services, deals or accountancy help. But not all data is PII and directly controlled by consumers.
In this context, the EU has a clear policy direction, tipping the scales toward public benefit by creating institutions and frameworks to encourage data accessibility. And data intermediaries are at the heart of that policy (see Data Governance Act). But, of course, this doesn’t dissolve the tension between data sharing and competition in specific markets and situations — particularly those not covered by the Digital Markets Act. So, will this tension be de facto internalised by data intermediaries? Will they need to balance the broad public good from making commercial data accessible with harm to consumers or competitors in specific markets?
For public-interest intermediaries focused on sharing non-PII industry data, it is hard to see how this tension won’t become part of their calculations to create sustainable data ecosystems. In the tourist example above, a data intermediary might decide it is in the public interest to discourage some data sharing, no sharing of real-time prices, for instance. So, in addition to all their other responsibilities, many data intermediaries must also understand the specific ways in which ‘it depends’ — that is, the balance between firm incentives to share and the competitive benefits of sharing. To some extent, this trade-off will — or should — shape their data strategy and policies.
In this blog, we zoom in on the commercial incentives to share data, looking at how market characteristics and the nature of data flip incentives. Importantly, firms can be encouraged to share more data through selective disclosure — what gets shared with whom. In a companion blog, we turn to recent economic research on ‘information design’ to explore the nuance of ‘how to share data’.
Firms’ incentive to share data
There are many business models for public-interest data intermediaries, including Cooperatives, Trusts, and Unions⁶. One characteristic they all share is the need to manage the incentives to source and distribute data (“data-in” and “data-out” policies).
The regulatory environment certainly plays its part. Through its incentives, sanctions and ability to establish ‘norms’, it should be able to encourage greater sharing from traditional firms (i.e. non-gatekeepers). It is entirely plausible, for example, that firms have overstated the strategic reasons not to share data simply because it is not the norm. Maybe they over-value the data they hold simply because they think it is ‘theirs’. But, of course, regulation is only part of the story. Firms must also be incentivised to share high-quality data. There are too many ways to achieve perfunctory compliance, from paying lip service to sharing obligations even when standards are explicit.
So, what are a firm’s incentives to share data?⁷ Generally, there are two effects: shared data allows a firm to make better predictions of its demand and costs, but also allows its competitors to be smarter. The incentive calculus is how much the firm cares about having smarter competitors. Of course, ‘it depends’ — but, to be specific, it mostly depends on two fundamental characteristics — the nature of competition and the nature of the data.
The key market characteristic is that firms are strategically interdependent — the value they extract from data depends on what their competitors know and do. Above all else, it is how coordinated they want to be with their competitors that matters. For example, faced with data predicting a recession, would they like their prices, product development, and hiring policies to move together or apart? If there is a sustained boom in American tourists visiting London, then London hoteliers might all want higher prices. But the West End will want a diversity of shows. In the jargon of economics, hotel prices are ‘strategic complements` and shows are ‘strategic substitutes’.
The second characteristic is the nature of commercial uncertainty. Ultimately, firms want to make more accurate predictions of consumer demand and their costs to make better decisions. They can use (and share) the data they control — for example, on sales, prices, costs, salaries, consumer segments, market research and so forth. But there are plenty of areas where they don’t have data. Some of those data gaps will be of specific interest to them more than their competitors — growth segments for their products or supply chain costs for their suppliers. This information has a private value. It is only useful to them. Some information will impact all firms in the market — overall consumer demand, tightness of the labour market, supply cost inflation, etc. This information has a common value.
To summarise, the two characteristics that influence sharing decisions are
1. Nature of competition. How ‘coordinated’ the firm would like to be with its competitors.
- Strategic Substitutes — the profitability of an action decreases the more the competitor does the same. For example, if a theatre company invests in an adaptation of a Jane Austin novel, it doesn’t want a rival company to do the same. It wants decisions to be uncorrelated in response to the data.
- Strategic Complements — the profitability of an action increases the more the competitor does the same. For example, hoteliers and Airbnb hosts all benefit from higher prices. They want pricing decisions to be coordinated.
2. Type of uncertainty. What information is being revealed
- Firm-specific (private value). Information that is specific to a firm or its market positioning. For example, data could be provided through a question-and-answer algorithm that predicts consumer trends based on a firm’s input data and query. It gives one firm-specific information, which might differ from another.
- Aggregate uncertainty (common value). For example, the overall nature of demand. Hoteliers want to know if there is a shift in tourist demand, or if Brexit has increased their employment costs. For example, data might be fairly granular to enable ‘surge pricing’.
The table below summarises how these two characteristics combine to impact a firm’s incentive to share its data. For example, when firms’ strategies are substitutes, they won’t want to share if the data results in excessive coordination. However, if data can be disseminated in a more personalised, tailored form, then firms will share. A familiar example is Google Maps. If the same route recommendation is given to everyone, it results in congestion. Recommendations should be personalised or randomised.
Source: (Raith 1996) & (Vives 2016)
A framework to understand a firm’s incentive to share data is an important first step. But, of course, what public-interest data intermediaries would really like to know is how to design their sharing policies to encourage more firms to share more data. This is the topic of the second blog in this series, where we look at specific ways to design data-sharing policies that encourage greater firm participation and mitigate potential harm to consumers.
A quick spoiler. It turns out that the ODI Data Spectrum is central. Shared data is not open data. However, choices about what to share are — at least in part — constrained by the economic fundamentals described in this blog.
Conclusion
The regulatory environment is critical to encourage firms to share data, but it will never be sufficient. Sustainable, public-interest data intermediaries must incentivise firms to share through their data strategy and policies.
So, the first observation is that in order to perform their roles well, intermediaries will need deep insight into firms’ business models and the markets they operate in. The nature of their strategic interaction is central to how shared data must be designed.
The second observation is that some intermediaries will need to make data strategy choices that trade off the welfare of consumers with wider public benefits. The stronger the degree of consumer control over data about them, the easier those choices will be.
Notes
¹See (Jones and Tonetti 2020) for a discussion and high-level quantification of the potential public benefits of data.
² Large firms are defined as having revenue greater than £250m(London Economics 2022)
³ For a review of competition issues, see (CMA 2021)
⁴ ‘It depends’ on the nature of competition in the main market where the data is used. Some of the issues are discussed later in this blog, but (de Corniere and Taylor 2023) provide a more comprehensive theoretical framework.
⁵ (Bergemann 2023) describes a market for personal data with a pivotal role for data intermediaries. Their vision for intermediaries combines delegated choice with collective bargaining. Intermediaries are regulated and licensed to create comparability. Consumers choose one intermediary exclusively.
⁶ For a comprehensive review of public-interest data intermediaries and their business models, see (European Commission. Joint Research Centre. 2023)
⁷ Despite a substantial body of economic research on firm data sharing, it is relatively lightly covered in more general reviews of commercial data sharing. Foundational research of why competing firms might share data, for example, through a trade body, is summarised by (Raith 1996) and (Vives 2016). More contemporary approaches emphasise a data intermediary's active role in designing information policies that encourage sharing (Bergemann and Morris, Stephen 2013; Bergemann and Morris 2019).
⁸ Sharing cost data when competition is price-driven is a complicated case. On the one hand, better information on costs can help firms coordinate prices which they like. On the other hand, higher-cost firms are vulnerable to being priced out of the market.
References
Bergemann, Dirk, Jacques Cremer, David Dinielli, Carl-Christian Groh, Maximilan Schafer, Monika Schnitzer, Fiona Scott Morton, Katja Seim, and Michael Sullivan. 2023. ‘Market Design for Personal Data’. Yale Journal on Regulation 40.
Bergemann, Dirk, and Morris, Stephen. 2013. ‘Robust Predictions in Games With Incomplete Information’. Econometrica 81 (4): 1251–1308.
Bergemann, Dirk, and Stephen Morris. 2019. ‘Information Design: A Unified Perspective’. Journal of Economic Literature 57 (1): 44–95.
CMA. 2021. ‘Algorithms: How They Can Reduce Competition and Harm Consumers’.
Corniere, Alexandre de, and Greg Taylor. 2023. ‘Data and Competition: A Simple Framework’. Working Paper.
European Commission. Joint Research Centre. 2023. ‘Mapping the Landscape of Data Intermediaries: Emerging Models for More Inclusive Data Governance.’ LU: Publications Office.
Jones, Charles I., and Christopher Tonetti. 2020. ‘Nonrivalry and the Economics of Data’. American Economic Review 110 (9): 2819–58. https://doi.org/10/ghbvzj.
London Economics. 2022. ‘Research into the Cost Considerations of Data Sharing’. DCMS. Raith, Michael. 1996. ‘A General Model of Information Sharing in Oligopoly’. Journal of Economic Theory 71 (1): 260–88.
Vives, Xavier. 2016. ‘Information Sharing Among Firms’. In The New Palgrave Dictionary of Economics, 4. Palgrave Macmillan UK.
