| The Standard Industrial Classification (SIC) system has been used by academic and professional researchers to classify companies into homogenous groups, with varying degrees of success. This study examines another source of historical SIC code assignments to companies that has become recently available to ascertain if this classification yields any improvements in classifying companies into homogenous groups. The historical SIC code from Compustat is contrasted with the traditionally used current SIC code, the historical CRSP SIC code, and the current CRSP SIC code, and the degree of homogeneity within industry groupings and across industry groupings is examined to see if better results can be obtained under any one system of industry classification. It is shown that the historical SIC Code from Compustat exhibits lower composite variance for a set of twenty financial ratios, compared to the historical CRSP SIC code classification, or the current Compustat or CRSP SIC code classification. Moreover, it is shown that the composite variance increases when aggregating the SIC codes to less than four digits, uniformly under the historical and current SIC codes assigned by Compustat and CRSP. The different SIC coding systems are also compared by replicating Lang and Lundholm (1996), to compare the SIC coding systems on the basis of the coefficient on the variable representing an industry information transfer effect. It is shown that the Historical Compustat 4-digit classification system performs better than the Historical CRSP four-digit SIC classification system, in capturing the industry effect, as well as in terms of a comparison of the industry effect for regressions performed on an industry basis. |
