The increasing frequency of very high temperatures driven by global warming has motivated growing interest in how the probability distribution of summertime temperatures will evolve in the future. Climate models predict increasing temperature variance in global warming simulations, but given their biased representations of historical temperature variability, it is important to use simple models to evaluate and understand these predictions. In this study we show that the projections of increasing temperature variance are indeed credible and are driven primarily by the magnitude of local warming. A simple analytic theory based on the surface energy and water budgets reproduces the increased midlatitude summertime temperature variance shown by state of the art climate models using only the local change in summertime mean temperature and relative humidity. The relative contributions of local warming and relative humidity changes to the increases in summertime temperature variance are roughly equal.