The need to deal with variability in wind power output is one of the greatest challenges connected with adopting a considerable amount of wind power into power grids. Power system operators need to acquire more information on this variability, which can be utilized in the mitigation of high ramping events, especially when these events synchronize with a large error in the prediction, ensuring flexibility and reliability in the power system besides the economic considerations. The paper analyses short-term variability in output power using actual data obtained from aggregated wind farms from 2015 to 2020, where power ramping characteristics are described using a variety of measurements. The use of the standard deviation of short-term wind power variation as a reserve measure will be investigated in detail since there is no consensus about the ideal confidence level value as a multiplier of σ, which ranges from 3 to 6 times σ. The paper addresses how large this confidence level should be, as well as developing a data-driven approach for estimating this reserve with increasing wind shares and evaluating the proper distribution of short-term wind variation. The results illustrate that the stochastic variations in wind power can retain many of their characteristics from year to year, even when the share of wind capacity is raised.