Early intervention does pay long-term dividends, but there are a couple of problems with this specific claim.
The figure of $3.2 million apparently comes from the work of Michael Ganz of the Harvard School of Public Health: "the total annual societal per capita cost of caring for and treating a person with autism in the United States was estimated to be $3.2 million and about $35 billion for an entire birth cohort of people with autism." But when Ganz writes of "societal per capita cost," he is not just talking about "social service costs." He explains that it includes much, much more:
The total costs of autism equal the sum of its direct and indirect costs. Direct costs measure the value of goods and services used and indirect costs measure the value of lost productivity due to autism. These direct and indirect costs represent the value of other activities that these resources could have purchased (i.e., opportunity costs). Physician and other professional services, hospital and emergency department services, drugs, equipment and other supplies, and medically related travel and time costs are typical components of direct medical costs. Direct medical costs were obtained either from the literature or from an analysis of the Medical Expenditure Panel Survey (MEPS) and the National Health Interview Survey (NHIS). Special education, transportation, child care and babysitting, respite care, out-of-home placement, home and vehicle modifications, and supported employment services are typical components of direct nonmedical costs. Nonmedical costs were obtained from the literature. Multiple cost estimates within categories were averaged to obtain a single cost estimate for each category. Indirect costs are the value of lost or impaired work time (income), benefits, and household services of individuals with autism and their caregivers because of missed time at work, reduced work hours, switching to a lower-paying but more flexible job, or leaving the workforce. Indirect costs were computed using a human capital approach that combines average earnings, benefits, and household services with information on average work-life expectancies and labor force participation rates for men and women at different ages.
Second, saving the $3.2 million is possible only when the treatment is so totally effective that the autistic person needs no further intervention, ever. That assumption is highly optimistic.
Again, advocates are wise to point out the cost-effectiveness of early intervention. But they should take care not to overstate their case or distort the data.