Temporal and regional variations in the seasonality of births in Aland (Finland), 1653-1950.

Monthly numbers of births are influenced by the length of the month and, thus, rates per day should be used. Comparisons of seasonality in data sets of different sizes must be based on standardized indices. Although strong seasonality exists, a poor model may incorrectly yield low seasonality measures. The standard deviation of the inter-monthly variation in the indices has been used as a seasonality measure. We introduce a new measure based on analysis of variance (ANOVA). Our methods are applied on 107,896 births in the archipelago of Aland (Finland), 1653-1950. For all subregions, the birth indices for the period 1653-1850 showed marked peaks in March to April and in September to October. The seasonality is weakest for the Main Island and strongest for the most isolated parish, Kökar, which has many local/extreme socioeconomics and genetic characteristics. For the period 1851-1950, the seasonality disappeared almost totally. The variation in the seasonality between subregions can mainly be explained by socioeconomic differences, and the temporal decrease in all regions seems to be connected to diminishing differences in living conditions. Comparisons with data from Sweden and Finland show higher seasonality in Aland.