The Accuracy of Data Collected From Firms

Abstract

How reliable are our facts and figures on the working of British industry? This was one of the problems considered by a conference on the difficulties found in the collection of data from firms, held recently by the Acton Society Trust. The conference was held in the belief that, although people looking for different things must adopt different approaches, there is a sufficient common basis to the search for facts in industry to make an exchange of experience among research workers worthwhile. The twenty-five people invited to the conference were therefore drawn from a range of studies in the social sciences including economists, sociologists, psychologists, statisticians, and those all-in wrestlers who do operational research. Some of the topics discussed were of interest chiefly to research workers. Should the research worker, for example, have any technical knowledge? If he is studying the details of processes in the factory he clearly needs such knowledge, but if he is making a more general study of, say, prices, a display of too much knowledge may make firms unwilling to give him information. There was also some discussion on how to get an entrance into the firm in the first place, and, having got there, how to win the co-operation of everyone on whom one is dependent for information being on good terms with one department in a firm frequently makes one persona non grata with another. But when the conference discussed the sort of errors to which data collected from firms are subject, it was dealing with matters of interest to the much wider public who use these data when they have been processed and published in the form, for example, of economic statistics. To understand the type of error that is liable to occur, we need first to look at our published facts and figures in their embryonic stage, as primary data. One of the most common forms in which primary data are collected is 'the little black book' the notebook kept by the foreman on the job. If, as often happens, the foreman is only semi-literate, one source of error is obvious. A similar type of primary data is the return completed in a small branch depot and sent to the head office one investigator, inquiring into the efficiency of horse transport, discovered what appeared to be a remarkable performance by the horses at one depot. In fact the depot had had no horses for twenty years, but the man who sent in his'stastistics'