A brief discussion period followed, during which Dr. Bresnahan suggested that the Department of Commerce’s framework might offer two challenges to the federal statistical system. One is a wide discrepancy between the department’s greatly improved measures of output in the information technology sector itself and the less effective way it measures the improvement of output when information technology is used in other sectors. The second is that the network model constituted a serious challenge to traditional ways of measuring firm establishment and might require new concepts for capturing the locus of economic activity in the network world.

Dr. Shapiro responded to the first question by saying that the difficulty in measuring the effects of information technology in the major service sector reflects a problem in statistical technique. The Commerce Department has made significant progress in financial services and banking, in which the productivity numbers have been revised. Such progress, however, is difficult to sustain without resources: The budget for the Bureau of Economic Analysis had been cut for every one of the last seven years, so that by 2000 the economy was 25 percent larger and the bureau’s real budget to measure the economy was 12 percent smaller.

Dr. Cerf commented that gathering good economic data is important to making good economic decisions and added an anecdote about productivity in the telephone system. Years ago when a person wanted to make a phone call, an operator had to place it. With direct-distance dialing, everyone became their own operator. Given that self-service is similarly becoming an important component in what happens on the Internet, bringing an important impact to the cost of operations, he inquired as to whether economic statistics can take self-service into account.

Dr. Shapiro answered that the department is currently trying to do this, beginning with the first “e-tail” statistics and proceeding next year with the first annual business-to-business measures.

He mentioned a larger conceptual problem, which is to adapt the structural logic of the statistics to large changes in the economy. That is, how can they develop measures that capture new activities and still be able to compare what is happening today with what has happened in previous years. This is particularly difficult when changes in a firm or in the nature of work alter the way the Bureau of Labor Statistics (BLS) needs to measure unemployment and employment. The