Regions with high wind energy potential require a comprehensive understanding of possible climate change impacts on wind resources. We investigate the performance of two reanalyses (ERA5 and MERRA2) and a regional climate model (RegCM4) on simulating atmospheric circulations over Central America, Mexico, and the Caribbean (CORDEX-CAM) and projected low-level wind changes in the region. RegCM4 historical wind patterns in Mexico are evaluated using seven tall inland masts and the two reanalyses. The evaluation indicates that MERRA2 is more accurate than ERA5 in representing observed winds, and RegCM4 captures present-day wind climate patterns with overestimations (> 2 m/s) over complex terrain. The mean recurrence of weather patterns based on self-organising maps in the historical period identifies high near-surface winds close to Mexico's major wind farm areas. Large pressure gradients between semi-permanent pressure systems and orographic features drive most of these patterns. We use the high temporal and spatial resolution RegCM4's data to obtain near-future (2021--2050) wind changes under the highest emission scenario. These changes are assessed using mean differences and recurrence pattern analysis from the RegCM4 downscaled simulations of three global climate models relative to the historical period (1981--2010). The mean RegCM4 ensemble of wind projections across most of the CORDEX-CAM domain over land shows no substantial changes in regional circulation patterns nor in mean wind speeds (< 2%) at 100 m in the near future, except for Colombia and Venezuela in summer and autumn (< 10% wind speed increases). In the southern Caribbean, summer offshore winds may increase (< 10%).