Abstract: Wind integration studies are an important tool for understanding the effects of increasing wind power deployment on grid reliability and system costs. This paper provides a detailed review of the statistical methods and results from 12 large-scale regional wind integration studies. In particular, we focus our review on the modeling methods and conclusions associated with estimating short-term balancing reserves (regulation and load-following). Several important observations proceed from this review. First, we found that many of the studies either explicitly or implicitly assume that wind power step-change data follow exponential probability distributions, such as the Gaussian distribution. To understand the importance of this issue we compared empirical wind power data to Gaussian data. The results illustrate that the Gaussian assumption significantly underestimates the frequency of very large changes in wind power, and thus may lead to an underestimation of undesirable reliability effects and of operating costs. Secondly, most of these studies make extensive use of wind speed data generated from mesoscale numerical weather prediction (NWP) models. We compared the wind speed data from NWP models with empirical data and found that the NWP data have substantially less power spectral energy, a measure of variability, at higher frequencies relative to the empirical wind data. To the extent that this difference results in reduced high-frequency variability in the simulated wind power plants, studies using this approach could underestimate the need for fast ramping balancing resources. On the other hand, the magnitude of this potential underestimation is uncertain, largely because the methods used for estimating balancing reserve requirements depend on a number of heuristics, several of which are discussed in this review. Finally, we compared the power systems modeling methods used in the studies and suggest potential areas where research and development can reduce uncertainty in future wind integration studies.