A Critique of "Reinventing Government in the American States: Measuring and Explaining Administrative Reform". (A Difference of Opinion)
本文批评了Brudney等人关于美国各州政府再造的获奖文章，指出其数据、方法论和设计存在根本缺陷，包括回应率不一致和样本代表性不足。
One of least welcome tasks in scholarly business is to publicly criticize work of colleagues. The work in question is Jeffrey L. Brudney, F. Ted Hebert, and Deil S. Wright's (to be referred to as the authors) article, Reinventing Government in States: Measuring and Explaining Administrative Reform (Brudney, Hebert, and Wright 1999, 19-30). It is an even more unpleasant task when article in question is not only multiauthored, but it also receives an award from field's premier journal (PAR 2000, 292). It does seem surprising, however, that not one of authors, editors, reviewers, or awarders noticed that article is fundamentally flawed--in its data, methodology, and design. Commentary will be made in that order, followed by a proposed exploratory hypothesis that sheds different light on subject. Reinvention of Government as Antagonist But, first, it is all too easy to see why article was so well received, despite it being too good to be true. At last, here was living proof of triumphant victory over reinvention. Osborne and Gaebler in 1992 authored most widely read work on reinvention. Ever since, but especially during past few years, Public Administration Review has published many articles about reinvention--nearly all unfavorable (Mae 1994; Terry 1998; deLeon and Denhardt 2000). Some of articles are replete with intemperate language that characterizes them as editorials, but, then again, they were simply preaching to already converted. The principal reached by authors merely reconfirmed this overt antagonism, in that ... a concerted reinvention movement does not appear to be underway across (19). (1) The authors identify survey as American State Administrators Project, but they say little about its objectives. In 1994-95, authors mailed questionnaires to 3,365 agency heads or directors of 93 agencies in 50 states. The stated--but soon to be challenged--overall return rate was 37 percent, or 1,229 respondents. Questionable Respondent Return Data Rates The authors make egregious errors in data reporting and analysis, ranging from inconsistency in usage to unevenness in distribution of respondents. First to inconsistencies: A major disagreement exists between what authors assert is number of respondents compared with figures they actually report. The authors claim to have 1,229 respondents, yet subtotals of states add to only 1,135 (24). The latter is 34 percent, not 37 percent. In displaying these data, it is strange that authors neglect to give a total number of respondents in their accompanying table, as is customary, that is, N=s. The authors fail to offer any reason why they are remiss in reporting total number of respondents or, apparently, why they discarded 94 respondents. Assuming authors can explain away this disparity over return rate, their main conclusion becomes tainted if lower figure is accurate. In particular, authors employ two statistics of central tendency: one is correlation, and other is percentage, which would be influenced through a change in denominators. This commentary utilizes (N=1,135) figure in its calculations. Large Populous States Data Sorely Under-Represented The second data problem is unevenness in reported respondents--a disproportionate number of them came from smaller, less populous states. Indeed, a scant number of replies turned up from most populated states. None of five largest states had 30 or more agency chiefs return their surveys, yet, of 11 states that did, only one is populous (see table 1). Combining two largest states, California and Texas, yielded only 27 respondents, 12 and 15, respectively. The third-largest state, New York, had least, with eight returned surveys. …