Approach and methodology

This paper applies a cost-of-illness (COI) methodology, the same analytical architecture used in the Mental Health Commission of Canada's landmark estimate of the economic burden of mental illness,1 the MADD Canada cost-of-impaired-driving series,2 and CIHI's addiction cost estimates.3 It estimates what online harms are currently costing existing systems across health care, education, justice, and the broader labour market and presents these in the form of annualized costs in today's dollars. These are based on documented Canadian incidence and unit costs, supplemented by international evidence scaled to Canadian conditions where domestic data are unavailable.

For this study, we were provided with a long list of online harms by Reset Tech on behalf of the Safer Online Spaces Coalition. Based on our initial review of the literature the paper has chosen to focus on the following core areas.

• Cyberbullying and online harassment, including downstream mental health, educational, and productivity impacts.
• Exposure to content that induces self-harm and related outcomes, with particular attention to health system costs and long-run consequences.
• Hate speech and online hate-related victimization, including justice system, victim services, and mental health burdens based on reported administrative data and public health research.
• Online gambling harms among youth and young adults.
• Platform-driven screen addiction and associated mental health effects to the extent there are established clinical data on incidence of mental disorders and the lost time associated with general knowledge acquisition in classroom.
• Online child sexual exploitation (OCSE).

1. Identify a Canadian incidence base. Drawn from Statistics Canada administrative surveys, such as CIHI national databases of hospital admissions, Statistics Canada's research on youth and health, the Uniform Crime Reporting (UCR) Survey, the Canadian Centre on Substance Use and Addiction and, where Canadian data is unavailable, from internationally validated estimates scaled to Canadian conditions.

2. Apply attribution factors to online harm drawn from peer-reviewed epidemiological literature. This is then adjusted for demographic factors specific to Canadian youth to establish a baseline level of incidence. In a number of cases there are conflicting estimates in the literature; we have noted this and attempted to provide a high and low level estimate of the range of impacts that are possible.

3. Multiply incidence by unit costs derived from Canadian published sources. The unit cost data is of varying dates and timelines and so we have taken steps to forecast what we believe the best estimate for these activities in 2026 dollars based on recent trends in activity demand, population growth, trends in activity level spending by the public sector, and or macro factors like nominal GDP or price levels where appropriate to do so.

4. Sum across the various channels of impact to arrive at an estimated impact, in both low and high impact scenarios. Given the potential for overlap and duplication between certain channels of incidence we have applied a discount factor to attempt to control for these effects in aggregate. Where sub-unit incidence or activity costs were noted from international sources we have attempted to back out relevant components that are already covered elsewhere in this work to assist with comparability.

This analysis is explicitly scoped to platform-facilitated harms and does not estimate the economic value of benefits associated with online activity — including educational access, social connection, information access, and skills development — which would partially offset the costs documented here. A comprehensive social welfare analysis would need to weigh both sides of the ledger; this paper addresses only the harm side. It is anchored to the approximately 11.3 million Canadians between the ages of 0 and 244 — approximately 27 percent of the population. Where data require age bands that overlap with young adults (such as the 18-to-29 cohort in gambling research), we document the narrowing applied and flag any resulting conservatism.