The paper describes a method of estimating variable returns to scale in pro
duction that adaptively fits spline functions, using model selection criter
ia, to determine the appropriate number and location of break points for a
fixed factor of production. Unlike other approaches, the method obtains non
parametric estimates of variable returns to scale for small samples while e
nsuring global curvature and flexibility properties are maintained. An appl
ication of the method is presented using data from the British Columbia sab
lefish fishery.