Background: Electroconvulsive therapy (ECT) has a long history of use in treating depression. Repetitive transcranial magnetic stimulation (rTMS) has been introduced more recently to the treatment spectrum. Its cost-effectiveness has not been explored. Method: Forty-six right-handed people with severe depressive episodes referred for ECT were randomised to receive either ECT twice weekly or rTMS on consecutive weekdays. Health and other service use were recorded for retrospective periods of 3 months prior to initiation of treatment and during the 6 months following the end of allocated treatment. Costs were calculated for the treatment period and the subsequent 6 months, and comparisons made between groups after adjustment for any baseline differences. Cost-effectiveness analysis was conducted with incremental change on the 17-item Hamilton Rating Scale for Depression (HRSD) as the primary outcome measure, and quality-adjusted life years (based on SF6D-generated utility scores with societal weights) as secondary outcome, cost-effectiveness acceptability curves plotted. Results: Based on the HRSD scores and other outcome measures, rTMS was not as effective as ECT. The cost of a single session of rTMS was lower than the cost of a session of ECT, but overall there were no treatment cost differences. In the treatment and 6-month follow-up periods combined, health and other service costs were not significantly different between the two groups. Informal care costs were higher for the rTMS group. Total treatment, service and informal care costs were also higher for the rTMS group. The cost-effectiveness acceptability curves indicated a very small probability that decision-makers would view rTMS as more cost-effective than ECT. Limitations: Small sample size, some sample attrition and a relatively short follow-up period of 6 months for a chronic illness. Productivity losses could not be calculated. Conclusions: ECT is more cost-effective than rTMS in the treatment of severe depression.