Separation of concerns is an important aspect of component-based development (CBD), and managing data is a primary concern in enterprise systems. In CBD methods, such as Catalysis and UML components, this concern is addressed by business components. Although a business component is self-contained, having no direct dependency on any of the other components, the data propagation between components may lead to indirect data dependencies across the business components, and grasping such dependencies at design-time is crucial to maintaining data consistency. In this paper we propose a method for data-flow analysis (DFA) of the business component model, in which the operational behavior is described using the Object Constraint Language (OCL) pre-postconditions. Traditional DFA techniques are aimed at procedural descriptions, while OCL is a declarative language whose essential properties include nondeterminism and incompleteness. In order to extract a data-flow from the OCL descriptions taking account of their semantics, our proposed method applies the idea of abstract interpretation. We also analyze the safety of our abstract interpretation technique, and discuss the usefulness and scalability of the method from a practical viewpoint. The proposed method, when used in conjunction with the inter-procedural DFA techniques, would allow us to extract the propagation and dependency of data across the business components automatically.