Several empirical studies have found that extended household units do not appear to be highly altruistically linked, thereby violating the very premise of the Ricardian Equivalence Hypothesis (REH). This finding has a very strong implication for the effectiveness of fiscal policies that change the allocation of resources between generations. We build a two-sided altruistic-linkage model in which private transfers are made in the presence of two types of shocks: an “observable” shock that is public information (for example, a public redistribution like debt or pay-as-you-go social security) and an “unobservable” shock that is private information (for example, individual wage innovations). Parents and children observe each other’s total income but not each other’s effort level. In the second-best solution, unobservable shocks are only partially shared, whereas, for any utility function satisfying a condition derived herein, observable shocks are fully shared. The model, therefore, can generate the low degree of risk sharing found in previous empirical studies, but REH still holds.