I recently finished reading Baumol's excellent "The Cost Disease", about the happy/sad story of modern societies' reliance on sectors -- medicine, education, etc. -- with substandard productivity growth. I won't reiterate his arguments here -- many of you are already familiar, and you would be better reading the book anyway -- but I did want to briefly revisit something he uses to make one flavor of his argument.
Baumol points out that having some sectors with substandard productivity is a mathematical certainty given other sectors with rapidly increasing productivity and declining costs. In particular, he points to the IT sector, and the declining cost of computing equipment. Using some data from a 1997 Dallas Federal Reserve report, he shows how the time to buy one "MIPS" (million instructions per second") of computing power at the prevailing average wage has declined over 70 years. I updated his figures, bringing them up to do date with 2012 wages and CPUs, with the following result. It's remarkable.