This paper explores whether the stationarity hypothesis of non-life insurance consumptions is supported during the period 1979-2005 for 31 countries. The stationarity of insurance consumption has important implications for modelling and forecasting insurance activities. On a global scale, this paper first implements the recent panel seemingly unrelated regressions augmented Dickey-Fuller unit root test, which allows us to account for possible cross-country effects and to identify how many and which countries of the panel contain a unit root. The main conclusion is that whether non-life insurance consumptions are stationary or not will be affected by different regions and their levels of development. Overall, our empirical results illustrate that non-life insurance consumptions in these countries are a mixture of stationary (integrated of order zero) and non-stationary (integrated of order one) processes. Higher risk aversion, lower income level and lower level of insurance market development may lead to non-stationarity. Finally, for the estimated half-lives of Africa, the degrees of mean reversion are greater than those for Europe and America. [PUBLICATION ABSTRACT]