<p>The widely cited estimate that there are approximately ’six others’ affected per person with problem gambling implies, when projected to the population level, a count of affected others that is highly inconsistent with prevalence data — likely due to some combination of biased reporting and network clustering that inflates relationship counts relative to unique individuals. By reconciling conflicting reports from gamblers and those they harm, and explicitly modelling network clustering—where the same people are often harmed by multiple gamblers—we estimate that, at the population level, approximately 4.7 people are affected for every person with problem gambling. Correcting this population-level overestimate does not eliminate the prevention paradox: low- and moderate-risk gamblers still account for the majority (70%) of all vectors of harm transmission in the population, despite lower reconciled fan-out than problem gamblers. We also find that nearly one in four affected others are themselves at-risk gamblers due to network effects, experiencing roughly double the exposure of non-gamblers. Our reconciliation of the available evidence confirms that the transmission of gambling harm concentrates in dense social networks (including families) rather than dispersing randomly. The edge-conservation framework we develop can be updated as new data becomes available, and applies broadly to any behavioural harm that propagates through social networks.</p>