The Bureau of Labor Statistics (BLS) is responsible for collecting data to estimate price indices such as the Consumer Price Index (CPI). BLS accomplishes this task by sending field staff to places of business to price actual products. The field staff are given product checklists to help them determine whether the products found are comparable to products priced the previous month. Prices for noncomparable products are not included in the current month's price index calculations. A serious problem facing BLS is developing product checklists for dynamic product areas, new industries, and the service sector. It is difficult to keep checklists up to date and quite often simply to develop checklists for service industry products. Some people estimate that more than 50% of U.S. economic activity is not accounted for in the CPI. The objective it to provide the results of tests on a method for helping BLS staff build new product checklists quickly and efficiently. The domain chosen for studying the method was the telecommunications industry. The method developed by ORNL is based on behavioral science and knowledge-engineering principles. The method has ten steps, which include developing a sample of domain experts, asking experts to list products in the domain, culling the list of products to a manageable number, asking experts to group the remaining products, identifying product clusters using multidimensional scaling and cluster analysis, asking experts to compare pairs of products within clusters, and, finally, developing checklists with the comparison data. The method performed as expected. Several prototype checklists for products in the telecommunications domain were developed, including checklists for paging services, digital cell phones, web browsers, routers, and LAN modems. It was particularly difficult, however, to find experts to participate in the project. Attending a professional meeting and contacting experts from the conference's mailing list proved to be the best approach for this domain. The method has performed well in two domains: the telecommunications industry, as demonstrated in this project, and the PC software industry, as demonstrated in a previous project. It is recommended that the method be further tested in additional service industries, such as the nursing home industry. In addition, further attention needs to be devoted to developing procedures for the method to improve its cost and time efficiency. For example, if automated methods were used to collect information from the experts and if the experts could be assembled at one time, it could be possible to create prototype checklists in one day.
