This paper derives an exhaustive set of restrictions implied by the multistock flexible accelerator specification of the adjustment-cost model for the firm. The flexible accelerator specification of the demand for capital and labor is estimated using U.S. manufacturing annual data for the 1947-1976 period. The consistency of the data with the model is examined by testing the validity of the derived restrictions. Finally, the conditions which permit aggregation over firms are determined and tested empirically. THE FLEXIBLE ACCELERATOR SPECIFICATION has been popular in the empirical literature on (interrelated) investment and employment decisions. (See Nadiri and Rosen [38], Coen and Hickman [10], and Schramm [40]. More recent examples include Sargent [39], Berndt, Fuss, and Wavermann [4], Berndt and Morrison [6], and Meese [34].) Under suitable assumptions the specification may be rationalized by the adjustment-cost model of the firm (Eisner and Strotz [13], Lucas [30]). Mortensen [35] extends the work of Lucas and derives some empirically testable restrictions of the accelerator specification that are implied by the adjustment-cost framework. The broad objective of this paper is to provide a more rigorous and comprehensive test of the adjustment-cost model than is presently found in the literature. This objective is achieved in three stages. First, we offer an alternative derivation of the Mortensen restrictions, and more importantly we prove (Theorem 1) that his restrictions are exhaustive, i.e., we establish an integrability result for the flexible accelerator, adjustment-cost model. Second, the flexible accelerator specification of the demand for capital and labor is estimated using total U.S. manufacturing annual data for the 1947-76 period. It is assumed that firms minimize expected production costs subject to a technology which implies that capital and labor stocks are costly to adjust, while a third factor, materials, is freely adjustable. The consistency of the data with the adjustmentcost framework is examined by testing for the validity of the derived restrictions. Further, we test the validity of some separability restrictions on the adjustmentcost technology that to varying degrees are imposed a priori in all of the cited empirical studies. Finally, we provide a complete characterization of the individual firm technologies which justify the common assumption of the existence of a representative firm. This problem of aggregation over firms, which has been