Knowledge that Transforms

Thomas P. Murphy, Political Economy and Geographic Distribution of Federal Funds for Research and Development: The Midwest Case, Administrative Science Quarterly, Vol. 14, No. 3 (Sep., 1969), pp. 426-441

will show how unimportant the indeterminate transactions were. They account for only 1.6 percent of all the messages originally logged in Alpha (32 of 2043), and for only 1.4 percent of the premises contained in the Beta sampling blocks (17 of 1245). How did the omitted premises differ from those which were tabulated? Overall, 61 percent of the incomplete transactions dealt with programmed decisions, a proportion which does not diverge alarmingly from the figure of 66 percent calculated for the complete transactions. A little more than half (52 percent) of the completed transactions originated with superiors. The corresponding figure for the incomplete transactions was about 7 percent larger, mainly because many of the disappearing premises were aimed at officials physically remote from City Hall: the director of the water purification plant, the supervisor of a work crew laboring on the outskirts of town, and so on. Because the observers' sampling schedule called for them to rotate their observations among officials headquartered in City Hall, it became extraordinarily cumbersome to ascertain the reactions of men who worked some distance away. What saddens me most of all is my critics' gratuitous allegation that my failure to say more about the incomplete transactions reflects adversely on my sophistication. That I said next to nothing about incomplete transactions in my article does not mean that I scanted them in the analysis of the data. In point of fact, I had not ignored the data at all; but I did eliminate an explicit discussion of them in successive drafts of the article in order to conform with ASQ's suggested limits on the maximum length of submitted articles. As Mayhew and Gray indicated in one of their footnotes, I did keep records of the indeterminate exchanges. I also analyzed them to see if they differed from the completed transactions. It would have been more courteous to ask me directly about them -some others did-than to leap into print with freewheeling allegations about my blunders. Out of critics' sight is not necessarily out of author's mind, and lawyers know that it is always wise to get the facts straight before taking a case to court. It saves embarrassment. I am amused over being censured for analyzing only 10 percent of my data. This criticism genuinely puzzles me. As my critics must know, samples rather than entire populations are used simply because it is less costly to examine elements selected at random than it is to compile information about an entire population. This procedure is hardly heterodox. Social scientists and the Bureau of the Census use it all the time. I may have used an inappropriate sample design, but I do not think it inappropriate to sample.