Topic 46 of 64· Public value in digital government

Open Data Value

Open data value is the problem of estimating what government and public data is worth when it has no price: it isn't sold, so there's no revenue line, yet releasing it (weather records, transport timetables, postcode boundaries, company registers) demonstrably generates economic and social activity downstream. Valuing it well matters because "it's free to release" and "it's worthless" are both wrong, and a software engineer deciding whether to open an API or a dataset needs a better argument than either.

Why it matters

The most-cited top-down estimate comes from the McKinsey Global Institute's 2013 report "Open data: Unlocking innovation and performance with liquid information", which put the potential annual value of open data across seven domains — education, transportation, consumer products, electricity, oil and gas, healthcare, and consumer finance — at $3 trillion to $5 trillion a year globally, through mechanisms including increased transparency, matching supply to demand more efficiently, and enabling new products and services built on the data. That figure is a scenario estimate, not a measured outturn, and it is routinely misquoted as if it were revenue government could capture directly, when the value mostly accrues to third parties — businesses, researchers, citizens — using the data, which is precisely the point of opening it rather than selling it. The UK's Open Data Institute, co-founded by Sir Tim Berners-Lee and Sir Nigel Shadbolt in 2012, has since built a body of more granular, bottom-up case studies — sector by sector, dataset by dataset — that are far more useful for a real business case than the McKinsey headline number, because they show the mechanism of value creation, not just its aggregate size.

The math

Open data has no market price, so valuation methods substitute for one; three approaches recur, and none is sufficient alone:

1. Cost-avoided / replacement-cost method:
   value ≈ what users would have paid to produce or license the
   equivalent data themselves — a lower bound, ignores value created
   by uses the original producer never anticipated

2. Market-analog / downstream-activity method:
   value ≈ revenue or savings generated by businesses/services built
   on the data (e.g. satnav apps built on open mapping and traffic
   data) — captures real economic activity but is hard to attribute
   cleanly to the data release itself (see additionality-and-deadweight)

3. Contingent/stated-preference method:
   value ≈ what users say they would pay, or the time they say it
   saves them — see stated-preference-valuation for the general
   method and its biases

None of these produces a figure as clean as a market price; credible
open data business cases triangulate across two or more, and are
explicit about which mechanism is doing the work.

Worked example

Illustrative national mapping/address data release (methodology after ODI-style case studies, figures illustrative of the scale such studies typically find):

Cost-avoided estimate:
  Businesses that would otherwise license equivalent address-matching
  data commercially, at an estimated average license cost of £4,000/
  year, across an estimated 15,000 SMEs now using the free open dataset
  = 15,000 × £4,000 = £60,000,000/year in avoided licensing cost alone

Downstream-activity estimate (more speculative, needs a counterfactual):
  New delivery-routing and logistics products built on the open data
  that would not exist, or would be materially worse, without it —
  requires a comparison against the counterfactual of the data staying
  closed or commercially licensed (counterfactual-analysis), because
  some of that activity would happen anyway on paid data at a higher
  price, which is deadweight in the "value created by opening it" sense

A defensible business case reports the cost-avoided figure as the
solid lower bound, and treats the downstream-activity figure as an
upper-bound scenario, not a fact.

Software engineering connection

For engineers, the practical open-data-value question is usually narrower than the national headline figures: does opening this specific API or dataset (rather than keeping it behind a partner agreement) increase reuse enough to justify the ongoing cost of documenting, versioning and supporting it as a public interface? That maintenance cost is real and is the counterpart to the build-once-reuse-often economics of government-as-a-platform — the two topics are close cousins, one about shared code and infrastructure, the other about shared data. Any open data value claim should be checked against additionality-and-deadweight before it goes in a business case: activity that would have happened anyway, on commercially licensed data, isn't value the opening created.

Pitfalls

  • Quoting the McKinsey $3–5 trillion figure as UK-specific or as this dataset's share: it's a global, seven-sector scenario estimate from 2013 — using it as a precise multiplier for a single national dataset misrepresents what the number is.
  • No counterfactual: claiming credit for all downstream economic activity built on open data, without asking how much of it would have happened anyway on paid or licensed data at a higher price (see additionality-and-deadweight and counterfactual-analysis).
  • Confusing cost of production with value created: a dataset that was expensive to collect isn't automatically valuable to release, and a cheap one isn't automatically low-value — value tracks downstream use, not upstream cost.
  • Ignoring the ongoing maintenance cost of "open": publishing a one-off CSV extract is not the same commitment as running a documented, versioned, supported open API — underfunding the latter after the launch announcement is a common failure mode.

Sources