AGGREGATION, UNIT ROOTS AND THE TIME-SERIES STRUCTURE OF MANUFACTURING REAL WAGES

The effects of both temporal and cross-sectional aggregation on the estimated time series characteristics of manufacturing real wage data are examined. The effects, especially of temporal aggregation, are found to be quite large and in accord with statistical theory. The well-known results of Altonji and Ashenfelter, that real wages are a random walk, and of Nelson and Plosser, that real wages are an IMA(1, 1) process, both seem to be entirely artifacts of temporal aggregation, with the true models following processes that are much more complex and that display substantial cyclical behavior.