Allostery describes the functional coupling between sites in biomolecules. Recently, the role of changes in protein dynamics for allosteric communication has been highlighted. A quantitative and predictive description of allostery is fundamental for understanding biological processes. Here, we integrate an ensemble-based perturbation approach with the analysis of biomolecular rigidity and flexibility to construct a model of dynamic allostery. Our model by definition excludes the possibility of conformational changes, evaluates static, not dynamic, properties of molecular systems, and describes allosteric effects due to ligand binding in terms of a novel free energy measure. We validated our model on three distinct biomolecular systems, eglin c, protein tyrosine phosphatase 1B, and the lymphocyte function-associated antigen 1 domain. In all cases, it successfully identified key residues for signal transmission in very good agreement with experiment. It correctly and quantitatively discriminated between positively or negatively cooperative effects for one of the systems. Our model should be a promising tool for the rational discovery of novel allosteric drugs.