Risk aggregation is a major challenge in assessing the risk of investment portfolios, trading books, or enterprises as a whole. Although there appears to be ample empirical evidence that asset returns correlate more strongly in periods of high market stress and despite the growing tendency of investors and regulators to allow for asymmetry and fat-tailedness by adopting downside-risk measures, such as Value-at-Risk and Expected Shortfall, conventional Pearson correlation still prevails in risk aggregation. Here, we propose methods  for deriving correlation matrices that are associated with specific regions of the joint return distribution. Specifically, we focus on tail areas to derive correlations for aggregating tail risks. An empirical application to U.S. blue chip stocks illustrates the concept.