From the Radio Free Michigan archives ftp://141.209.3.26/pub/patriot If you have any other files you'd like to contribute, e-mail them to bj496@Cleveland.Freenet.Edu. ------------------------------------------------ Prohibitionists are never short of heart-rending anecdotes, but they are ALWAYS short of data. This explains why they are constantly attempting to resurrect studies that have already been laid to rest. There has been a recent effort to resurrect the Loftin "study" of the 1976-77 Washington DC gun freeze. It is my hope that this lengthy post will reacquaint readers with the numerous major flaws in the Loftin polemic so that they may drive a stake through the heart of the study if they encounter any effort to resurrect the Loftin piece. A garland of garlic may also help. ************************************************************************* * Edgar A. Suter, MD suter@crl.com * * Chair, DIRPP Doctors for Integrity in Research & Public Policy * ************************************************************************* The draconian Washington DC 1976-77 gun law is often inaccurately described as a gun ban. In fact, the law was primarily a _handgun freeze_ since existent _registered_ guns were grandfathered (the law also includes certain stringent storage requirements even of long arms). Since new handgun acquisitions are banned, it is understandable that some might casually, but inaccurately, describe the law as a "ban." The freeze was enacted in October 1976 and became effective in February 1977. Critics of Loftin's article noted that the Washington DC homicide rates _began_ a _gradual_ fall in 1974, two years before the DC gun freeze. After this observation was called to Loftin's attention, Loftin responded, but did not test that _actual_ "alternative hypothesis" of a _gradual_ decline _beginning _two_ years before the gun freeze. Instead Loftin tested a "straw man" hypothesis, a hypothesis proposed by no one [except Loftin], that there was an _abrupt_ decline _at_ (not "beginning") in January 1974, almost _three_ years before the ban became effective. Loftin claimed that publicity and discussion about the gun freeze _proposal_ might have caused a reduction in violence even before its passage. Since the 1976-77 law was a freeze, not a ban and confiscation, there is no reason to expect an _abrupt_ drop. It is difficult to envision any plausible reason for an abrupt drop (Did DC's predators suddenly get religion and comply with the restrictions in anticipation even before passage of the law? Were they so scared that the law _might_ be passed that they stopped killing each other? Possible, but implausible.). _If_ any effect were to be noticed, it would be expected to be _gradual_, if and as the number of guns changed either downward by attrition (grandfathered guns leave DC as the owner moves, guns become unserviceable and unable to be replaced, etc.) or upward from the introduction of new guns (because the law has and had no effect on the steady introduction of new guns through illegal gun running). Loftin claims to have shown an _abrupt_ change at November 1976, _after_ the law, but _before_ its enactment. A gradual change is expected, but Loftin claimed to have demonstrated an _abrupt_ change coinciding nearly exactly with enactment of the gun freeze. There is no plausible epidemiological or other explanation for an _abrupt_ change. Since the abrupt change violates the "rule of biological plausibility," we have _another_ reason to be skeptical of Loftin's claim. The ARIMA (Automated Regressive Integrated Moving Average) method used by Loftin generates some interesting statistical artefacts. These artefacts allowed Loftin to contrive the appearance of a drop when no drop occurred. In their grant proposal to CDC (to obtain tax money to subvert our civil rights) for the DC study, Loftin discussed the ARIMA method. To demonstrate the method, Loftin analyzed IBM stock prices during a period of fluctuation. His "results" showed that the IBM stock fell. Statistical models have their strengths and their weaknesses. If a statistical test generates a result that is at variance with the primary [observed] data collected, the statistical result is suspect. If a statistical method makes an increase appear to be a decrease, one must explain how this could be so. Either the primary data collection is at fault (e.g. someone did not correctly count or enter the primary data points) or the statistical method is at fault (e.g. computational error, the method is being misapplied). In the case of the Loftin article, the problem lies not with the primary data count (nobody miscounted the _number_ of DC deaths), but with the hypothesis, the statistical model, and/or the application of the model. As I said in an earlier post on this subject, no statistical method turns "black" into "white." The ARIMA analysis does not undercut the observation the homicide rates increased every year between 1976 and 1991 (except 1985). Numerous additional problems exist with Loftin's methods. Kleck, Cowan, and many others have used Loftin's method to "test" a variety of hypothetical intervention dates. Any number of random dates ("random" in the sense that these dates do not coincide with any chronological landmarks of the DC gun freeze) obtain the same results, an _abrupt_ "fall" in the _number_ of homicides. Any number of dates _before and after_ the law can be "shown" [by ARIMA artefact] to be points of _abrupt_ fall in homicide raw numbers. If Loftin and Mr. Van Meurs claim that _one_ of those dates of abrupt fall - specifically the November 1976 date - was due to the gun freeze, how do they explain the _other_ abrupt falls on _other_ random dates? "Other causes," is the usual answer given by Loftin and his supporters --- which is EXACTLY my point. "Other causes" explain the random date drops, "other causes" explain the apparent drop noted by Loftin, AND "other causes" explain the INCREASES as well AND the _apparent_ "drop" is a statistical artefact, a spurious claim. Q.E.D. Another problem: Loftin's method works only if you use the National Center for Health Statistics (NCHS) data. His method fails to show a "drop" if one uses FBI data. Why not? NCHS data does not cull out _justifiable_ homicides. Even though FBI data undercounts justifiable homicides (FBI Uniform Crime Reports data are based on the _preliminary_ impression of the investigating officer and suggest about 2% of homicides are justifiable. On final analysis, about _20%_ of homicides are adjudicated as justifiable) that small difference is enough to cause Loftin's entire analysis to collapse. It is a bitter irony that the protective uses of guns, justifiable homicides, provide the increment of data that allows a spurious statistical contrivance to suggest that eliminating private gun ownership is desirable. Another problem: Loftin and his supporters pretend that the predominantly white, affluent suburbs of Washington DC (that had a 25% INCREASE in population) are an appropriate control group for overwhelmingly black, impoverished, crack-infested urban Washington DC inner city (that had a 20% DECREASE in population). A more appropriate control would have been a demographically similar urban area, such as Baltimore MD. Another problem: The population shifts (DC's decrease and the suburbs' increase) exaggerate Loftin's ARIMA artefact. Another problem: If one expands the baseline period (e.g. 10 years before the law compared with 10 years after the law), Loftin's method fails to demonstrate a drop. If you wish to invoke "other causes," I will happily agree AND I will note that "other causes" also account for the drop beginning two years _before_ the law AND "other causes" explain Loftin's contrived abrupt "drop." Bottom line: No statistical legerdemain can obscure the real world observation that following Washington DC's 1976 gun freeze, DC's homicide rates ROSE from 26.9 per 100,000 in 1976 to 80.6 per 100,000 in 1991. The homicide rate rose in EVERY year between 1976 and 1991 except for 1985. Whether the homicide rate increased _despite_ the law (the "other causes" explanation) or _due to_ the law (by gradually disarming the victims), I cannot say. I can say that, regardless of what reasons one may invoke, homicide rates _inceased_ following_ the DC gun freeze to the _highest levels _of _any jurisdiction_ at _any time_ in this nation. No rational person can find any scientific support whatsoever for "gun control" in these observations. Criticism in the _medical literature_ of Loftin's article includes the length-limited letters in New England Journal of Medicine (Letters. "Effects of Restrictive Handgun Laws." NEJM. 1992; 326(17): 1157-61.) and my article in JMAG (Suter E. "Guns in the Medical Literature - A Failure of Peer Review." Journal of the Medical Association of Georgia. March 1994; 83; 133-48.). The most detailed refutation of Loftin's statistical contrivance, however, is Kleck's presentation to the American Society of Criminology (Kleck G. "Interrupted Time Series Designs: Time for a Reevaluation." a paper presented to the American Society of Criminology annual meeting. New Orleans, LA. November 5, 1992.). The paper addresses not only the basic conceptual problems of Loftin's article, but exposes Loftin's methodological and statistical contrivance in meticulous detail. "Hand waving"? "proof by assertion"? I don't think so. Let the readers decide for themselves whether or not Loftin is competent and convincing. The articles and criticism are readily available to those interested. The relevant excerpt from my JMAG article summarizes: Foretelling the future - gun prohibitionists and criminals share a crystal ball... Loftin C, McDowall D, Wiersema B, and Cottey TJ. "Effects of Restrictive Licensing of Handguns on Homicide and Suicide in the District of Columbia." N. Engl J Med 1991; 325:1615-20. methodological and conceptual errors: *** the apparent, temporary, and minuscule homicide drop occurred 2 years before the Washington DC law took effect *** the "interrupted time series" methodology as used by Loftin et al. has been invalidated *** the study used raw numbers rather than population-corrected rates - not correcting for the 20% population decrease in Washington, DC during the study period or for the 25% increase in the control population - exaggerating the authors' misinterpretations *** the study conveniently stopped as Washington, DC's overall homicide rate skyrocketed to 8 times the national average and the Black, male, teen homicide rate skyrocketed to 22 times the national average *** used a drastically dissimilar demographic group as control *** the authors virtually failed to discuss the role of complicating factors such as the crack cocaine trade and criminal justice operations during the study period Loftin et al. attempted to show that Washington, DC's 1976 ban on new gun sales decreased murder.[1] Loftin and his co-authors, using tax money, produced "research" with several negating flaws that were ignored or overlooked by "peer review" and the editorial board of the New England Journal of Medicine - perhaps a corollary of the editor's no-data-are-needed[2] policy. Not only has the "interrupted time series" methodology as used by Loftin et al. has been invalidated,[3] but the temporary and minuscule homicide drop began during 1974, 2 years before the gun law - How could the law, even before its proposal, be responsible for the drop? Since homicidal maniacs and criminals could not clairvoyantly anticipate the law, other causalities should have been considered. The authors, however, side-stepped the question and dismissed non-gun causalities without any analysis whatsoever. The study conveniently stopped as the Washington, DC homicide rate skyrocketed. If the gun freeze law, which has not changed, were responsible for the homicide drop, we would expect the "drop" to continue. If the "guns-cause-murder" theory is valid and if the gun freeze were effective, as "grandfathered" guns leave circulation (owner moves, dies, guns become unserviceable, etc.), the homicide rate should drop steadily. Quite the opposite is observed. The 1976 Washington, DC homicide rate before the law was 26.9 (derived from population[4] and homicide[5] statistics) and then tripled to 80.6 by 1991[6] despite or due to the law; Justifiable and excusable homicides, including those by police officers, were treated the same as murders and were not excluded from the study. The study used raw numbers rather than population-corrected rates. This did not correct for the 20% population decrease in Washington, DC during the study period or for the 25% increase in the control population - exaggerating the authors' misinterpretation. The study used the adjacent suburbs as a control group, an area with demographics drastically different from the study group. The authors examined and allowed only a single cause interpretation - guns are to blame. They offhandedly discarded any other possible explanation. They specifically ignored the role of the crack cocaine trade, FBI stolen property and Bureau of Alcohol, Tobacco and Firearms illegal weapon sting operations in progress during the study, and measures instituted during the study period that improved the efficiency of the Washington DC court system. They generally ignored the role of poverty and myriad other factors related to criminal violence. Homicide has declined for every segment of American society except teenage and young adult inner-city residents. The Black teenage male homicide rate in Washington, DC is 227 per 100,000,[7] yet less than 7 per 100,000 for rural, middle-aged white men,[8] the US group for whom gun ownership has the highest prevalence.[9] If the "guns-cause-violence" theory is correct why does Virginia, the alleged "easy purchase" source of all those illegal Washington, DC guns, not have a murder rate comparable to DC? [According to the most recent FBI Uniform Crime Reports 1993: Virginia's homicide rate is 8.3, Washington DC's is 78.5] The"guns-cause-violence" theory founders. Even in their responses to criticism,[10] the authors' intransigent bias is evident. Their position? If a drop in murder is discovered (or statistically contrived), gun control must receive the credit, but when attention was drawn to the failures of gun control and their study design, the skyrocketing murder rate must be credited to "other causes." Shall we examine gun control as science or religion? It appears that the faith of true believers is unshakable heedless of data and the scientific method. [1] Loftin C, McDowall D, Wiersema B, and Cottey TJ. "Effects of Restrictive Licensing of Handguns on Homicide and Suicide in the District of Columbia." N. Engl J Med 1991; 325:1615-20. [2] Kassirer JP. Correspondence. N Engl J. Med 1992; 326:1159-60. [3] Kleck G. "Interrupted Time Series Designs: Time for a Reevaluation." a paper presented to the American Society of Criminology annual meeting. New Orleans, LA. November 5, 1992. [4] US Department of Commerce. Statistical Abstract of the US. - 96th. Edition. 1976. Washington DC: US Government Printing Office. [5] FBI. Uniform Crime Reports Crime in the United States 1976. 1977. Washington DC: US Government Printing Office. [6] FBI. Uniform Crime Reports Crime in the United States 1991. 1992 Washington DC: US Government Printing Office. [7] Fingerhut LA, Ingram DD, Feldman JJ. "Firearm Homicide Among Black Teenage Males in Metropolitan Counties: Comparison of Death Rates in Two Periods, 1983 through 1985 and 1987 through 1989." JAMA. 1992; 267:3054-8. [8] Hammett M, Powell KE, O'Carroll PW, Clanton ST. "Homicide Surveillance - United States, 1987 through 1989." MMWR. 41/SS-3. May 29,1992. [9] Kleck G. Point Blank: Guns and Violence in America. New York: Aldine de Gruyter. 1991. [10] Loftin C et al. Correspondence. New England Journal of Medicine. 1992; 326:1159-60. ------------------------------------------------ (This file was found elsewhere on the Internet and uploaded to the Radio Free Michigan archives by the archive maintainer. All files are ZIP archives for fast download. E-mail bj496@Cleveland.Freenet.Edu)