When introduced in 1976, personal computers (PC’s) had only several kilobytes of randomaccess memory (RAM) and no hard disk, processed commands at speeds of less than 1 megahertz (MHz), yet typically cost several thousand dollars. Today’s PC’s have megabytes (MB) of RAM and gigabytes of hard-disk memory, process commands at speeds exceeding 1,000 MHz, and often cost less than $1,000. Ever more powerful PC boxes have been transformed into increasingly smaller and lighter notebooks. In the context of such rapid technological progress, government statistical agencies face the formidable tasks of reliably tracking real output growth in the PC industry, and adjusting PC prices for quality changes. In 1986, following collaboration with IBM researchers, the Bureau of Economic Analysis (BEA) began adjusting mainframe computer prices for quality change, using hedonic price methods (Rosanne Cole et al., 1986). In 1990, the Bureau of Labor Statistics (BLS) first published an experimental price index for mainframe computers, also based on hedonic methods. Today the BLS publishes producer and consumer price indexes for a number of electronic goods, including informationprocessing equipment, using hedonic price methods (Dennis Fixler et al., 1999).

The hedonic regression methods employed by researchers and government statisticians to adjust PC prices for quality changes are usually based on pooled cross-section and time-series data, and they typically assume some parameter equality across models, as well as some parameter stability over time. Here we report on results from an examination of the stability of annual PC hedonic price parameters over almost the first quarter-century of their history (1976–1999), and on the equality of estimated hedonic price parameters between desktop and mobile models. We compute price indexes based on alternative assumptions concerning parameter equality and stability, and we assess at what point in time, when constructing price indexes, mobile and desktop models were appropriately treated as distinct product groups.