CONCEPT

GDP-B

Brynjolfsson's proposed supplement to traditional GDP that captures the consumer benefit from free digital goods — demonstrating through choice experiments that services the standard metric values at zero are worth thousands of dollars per person annually.

GDP-B is a measurement framework Brynjolfsson developed with collaborators including Avinash Collis, Erwin Diewert, Felix Eggers, and Kevin Fox to capture the welfare value generated by free digital goods that conventional GDP records as contributing nothing to the economy. Through incentive-compatible choice experiments — offering participants real money to give up access to digital services — Brynjolfsson's team estimated that the median American would require over $17,000 per year to give up search engines alone. Facebook, email, maps, Wikipedia, and other free services carried similar consumer valuations. Because GDP measures output at market prices and these services have market prices of zero, their enormous welfare contribution is entirely invisible in national accounts. GDP-B proposed adjusting the framework to capture this consumer benefit — a reform whose urgency increased as AI tools expanded the category of valuable services provided outside traditional market pricing.

In the AI Story

The methodology of choice experiments addresses a classic measurement problem: how do you value goods that are exchanged outside market transactions? Standard approaches like survey valuation are notoriously unreliable — people systematically overstate their willingness to pay for things they value. Brynjolfsson's team used incentive-compatible designs where participants made real binding choices with real monetary stakes. The resulting valuations were conservative by construction and therefore particularly striking — revealing that conservative estimates still produced numbers in the thousands of dollars per person per year for single digital services.

The implications for AI measurement are direct. If free digital services are massively undervalued by GDP, then AI tools that are either freely available or priced far below their use value will similarly escape capture. The individual who uses Claude or ChatGPT at subscription costs of $20-$200 per month may be capturing use value of thousands of dollars per month — the exact magnitude of valuation gap that GDP-B was designed to reveal. National productivity statistics will show modest effects. The actual welfare generated will be far larger.

The policy consequence aligns with Brynjolfsson's broader argument about intangible capital: what the economy cannot see, it cannot manage. Without GDP-B or something like it, policymakers making decisions about AI investment, regulation, and distribution are operating with systematically distorted information. The economy's dashboard is broken in a specific, correctable way — and fixing it is not a technical nicety but an economic imperative.

The framework connects to broader debates about economic measurement reform. Diane Coyle's work on reimagining economic statistics for the digital age, the Stiglitz-Sen-Fitoussi Commission's recommendations on well-being measurement, and the ONS experimental statistics program in the UK all point in similar directions: the metrics designed for an industrial economy measuring physical goods are systematically inadequate to an economy increasingly organized around information, services, and digital value creation.

Origin

The GDP-B framework was developed in a series of papers beginning with Brynjolfsson's 2019 NBER working paper GDP-B: Accounting for the Value of New and Free Goods in the Digital Economy, co-authored with Avinash Collis, Erwin Diewert, Felix Eggers, and Kevin Fox. The paper proposed specific methodological innovations for integrating welfare effects of free goods into national accounts.

The choice experiment methodology itself draws on decades of work in experimental and environmental economics, particularly in valuing non-market goods like environmental quality. Brynjolfsson's contribution was to apply these methods systematically to the digital economy and to demonstrate the magnitude of what standard accounting was missing.

Key Ideas

GDP values free goods at zero. The market-price framework is structurally blind to welfare from services that are not priced.

Choice experiments reveal hidden value. Incentive-compatible methods produce conservative estimates of willingness-to-accept that still show thousands of dollars per person annually for single services.

The gap is macroeconomically significant. Integrated across the economy, the unmeasured welfare from free digital goods runs to trillions of dollars annually.

AI will amplify the gap. As more valuable services are delivered outside traditional market pricing, the share of real economic welfare that GDP misses grows.

Better measurement enables better policy. Without seeing the gains, policymakers cannot make calibrated decisions about distribution, investment, or regulation.

Debates & Critiques

Methodological debates focus on the reliability of choice experiments at scale and whether willingness-to-accept estimates properly capture welfare. A more fundamental debate concerns whether GDP should be modified at all or whether a parallel set of metrics (well-being measures, welfare accounts, digital economy statistics) should complement rather than replace it. Brynjolfsson has generally argued for supplementation rather than replacement — keeping GDP for the purposes it serves while adding GDP-B and similar measures to capture what GDP misses.

Appears in the Orange Pill Cycle

Erik Brynjolfsson — On AI