Work Sampling in Financial Management--Cost Determination in Post Office Department

Work Sampling has been described by some authorities as the best and most objective method of measuring the time and cost of performing the various functions of complex organizations for use in management, budgeting and setting performance standards. Yet its use for these purposes has been quite limited. This paper is designed to encourage further use of Work Sampling. It describes what is probably the largest continuing application of Work Sampling in the world, namely, that made by the Post Office Department. About 4.5 million observations are made during a year at randomly selected instants of time to estimate the cost of the in-office activities and over 200,000 employees participate each year. The paper describes the basic objectives served by the application, the size and composition of the universe of interest, the methods of sampling, the sample size used, the estimation procedure, the methods of measuring sampling variation and controlling the quality of the information. In addition, a few areas of research to improve the sample design are described. The Post Office Department needs to know the cost it incurs to handle and deliver the mail in each class (e.g., first class and air mail) in order to compare it with the corresponding postage or other revenue earned by class. This information is needed primarily to recommend postal rate changes to Congress. While the total cost of postal activities currently amounts to about $7 billion a year, a first step taken in estimating the cost of each class by probability sampling deals with the in-office costs. This cost currently amounts to about $4 billion a year. Later probability sampling applications will be extended to cover out-of-office and other postal costs. A sophisticated three-stage sample design was adopted. First, a stratified random sample of 500 post offices was selected. Stratification was based on size of office. Then, a digital one percent sample of employees in these offices was selected to participate each week. Finally, the work which the selected employees are performing is observed at four randomly chosen instants of time during their 8-hour work day. Thus, each employee is generally work-sampled at 20 instances a week and a total of about 4.5 million readings are made annually. A quality control type of edit and control system has been established to detect and correct field reporting errors, thus helping to obtain high quality of results.