This paper performs an empirical analysis of the international cross-sectional distribution of gross domestic product (GDP) growth rates and business cycles. We consider a balanced panel of 91 countries in the period 1960-2010 and two different measures of GDP fluctuations: the logarithmic growth rates and the Hodrick-Prescott cycles. Both measures are characterized by fat-tailed distributions and strong heteroscedasticity. The latter is the result of a scale relation between the variance of the fluctuations and the size of the country. The analysis of the time evolution of these properties shows that distribution tails become asymmetrically fatter during the period of study, suggesting an increased probability of finding high amplitude fluctuations in more recent years. Moreover, we observe significant changes in the scale parameter characterizing the relation between volatility and country size. These findings enrich the discussion about robust properties of business cycles and reveal more evidence about scaling-law relations in economic systems.