Measuring Social Media Network Effects

TEXT ANALYSIS: Measuring Social Media Network Effects

THE DISSECTION

This is an empirical validation of the infrastructure value of social graphs—the paper measures how much of the $78-$101/month consumer surplus each platform generates is attributable to specific network connections (local network effects) versus platform features. Using incentive-compatible choice experiments with ~20,000 U.S. users across Facebook, Instagram, LinkedIn, and X, the authors decompose platform value into its social components. Key findings: 8.1-23.7% of platform value is explained by local network effects; platforms function as complements rather than substitutes; value distribution varies by tie strength, gender, race, and age.

The paper performs a precise forensic accounting of who monetizes the social graph and how.

THE CORE FALLACY

The paper treats this value as evidence of platform health and consumer benefit. It is, in fact, evidence of the extraction mechanism.

The social graph generates $53B-$215B in annual U.S. consumer surplus—surplus created by users, captured by platforms. The paper's methodology measures WTP (willingness to pay) for maintaining specific connections, which tells us the labor value of social capital that users built without compensation. The paper's framing—"social media generates significant value"—is precisely backwards: users generate the value; platforms collect the rent. This is textbook platform enclosure of commons.

The 8.1-23.7% attributable to "local network effects" is not a finding that network effects matter. It is a finding that the social graph is the asset, and that the platform's ownership of that graph is what makes it valuable. The paper proves the existence of the rent.

HIDDEN ASSUMPTIONS

1. Platform sovereignty is natural. The paper assumes platforms own the social graph as a legitimate property right. It never questions why the connections users create—through labor, relationship maintenance, content creation—belong to the platform. This is not analysis. This is accounting for the landlord.

2. Consumer surplus = welfare. The paper treats "$53B in consumer surplus" as evidence of net positive social value. It ignores the counterfactual: what would happen if this labor (social maintenance, content creation, attention provision) were compensated? The surplus is real. It is also unpaid.

3. The labor being measured is voluntary leisure. The choice experiments measure what people would pay to maintain connections—treating social graph maintenance as consumption. Under DT logic, this is the human activity most structurally threatened by AI: network maintenance, coordination, signaling—precisely the cognitive-social work the post-WWII economy is being unmade from.

4. Platforms are complements. The finding that platforms are complements (connectedness on one increases value on another) is presented as a consumer benefit. It is also evidence of lock-in architecture—the more platforms you use, the higher the switching cost of leaving any of them.

SOCIAL FUNCTION

Transition management theater and prestige signaling.

This paper is careful, rigorous work within a liberal empirical frame. It will be cited by economists as evidence that platforms "create value." It will be cited by regulators as justification for antitrust scrutiny of network effects. It will be cited by platforms in advertising as evidence of consumer benefit. It serves all these masters simultaneously because it is structurally agnostic about who owns what and why.

It performs the institutional function of making platform capitalism legible and measurable without questioning its basic architecture. It is the economic equivalent of documenting the temperature gradient in a burning building.

THE VERDICT

The paper accurately measures the rent. It misidentifies who earns it.

Under DT mechanics, this paper is evidence of precisely the wrong thing from a social welfare standpoint: it documents that the social graph—the product of massive unpaid cognitive and coordination labor by billions of users—is being enclosed by capital at massive scale, and that this enclosure generates extraction value currently misclassified as "consumer surplus."

The platforms are not creating value from local network effects. Users are. The platforms are collecting it.

This paper is a forensic accounting of the extraction. It does not question the extraction. It does not ask who built the infrastructure, who pays to maintain it, or what the AI-automated economy does to the value of human social graph maintenance.

Answer: it makes it redundant at scale, then worthless, then gone.

The $53B-$215B in consumer surplus is the last measurement before the instrument breaks.