Escape analysis of object-oriented languages determines, for every program point, the escape property ɛ i.e., the set of the creation points of the objects reachable from some variables. An approximation of ɛ is useful to stack allocate dynamically created objects and to reduce the overhead of synchronisation in Java-like languages. ɛ can itself be used for escape analysis, but it is very imprecise. We define here a refinement ɛR of ɛ, in the sense that ɛR is more concrete than ɛ and, hence, leads to a more precise escape analysis than ɛ.