Does Unemployment Increase the Risk of Suicide? : Empirical Evidence from the OECD Countries

Sammanfattning: Suicide is one of the leading causes of death around the world, where one person ends its own life every 40 seconds. Many different factors can explain the suicide decision, and one such factor is unemployment, which has negative health consequences and could increase the risk of suicide. Unemployment affects the individual in monetary terms, by reducing the income, and in non-monetary terms, as through anxiety and depression. To empirically estimate the effect of unemployment on suicide rates, a panel dataset covering all the 38 OECD countries during the period 2000–2019 is analysed with two different models, one two-way fixed effects model and one random effects model with time fixed effects. Regardless of which model that is selected, the results are similar and suggest a positive relationship between unemployment and suicide rates. An increase in the unemployment rate by one percentage point is associated with an annual increase in the total suicide rates by 0.8 people per 1,000,000 inhabitants, and similarly it would lead to an increase per year by 1.4 male and 0.2–0.3 female suicide victims. Although, the effect on female suicide rates is insignificant. The estimated effect in absolute numbers might appear small, but a one percentage point increase in unemployment rate would result in a total of 27 more suicide victims per year for the average OECD country in 2019. In relative terms, the estimated effects indicate that a 1 % relative increase in the unemployment rate is associated with a 0.08–0.09 % increase in the suicide rates, which is significant and valid for the rates in total as well as for men and women. Thus, about a tenth of the relative change in unemployment rate does successfully pass-through to the suicide rates. The results point in similar direction as some previous research. However, the great variability in modelling and analysing the effects on suicide rates complicates the comparisons across studies. In contrast to previous literature, some limitations of the models are clearly pointed out to a greater extent in this study, to highlight the sensitivity of the results, depending on the control variables included in the models and the panel dataset.