Randolph Roth, Ohio State University History Professor, Seminar "American Homicide"
I just witnessed a really amazing seminar by Randolph Roth entitled "American Homicide." He has tried to put together homicide data for the United States over the last couple hundred years from newspaper reports and coroner reports. From this he makes claims that murder rates were basically flat from 1914 to 1933 and that prohibition had no impact on murder rates. I really wish that I could post a copy of the figure that Roth presented. The fact that murder rates seem to have risen in individual states after they adopted prohibition doesn't seem to matter, that murder rates fell dramatically as soon as prohibition ended to the month in 1933 doesn't matter. The more amazing thing is how he got his data together. When someone asked him about the subjectiveness of determining what is murder, Roth responded that it is extremely subjective: "Tell me what murder rate you want and I can get you that murder rate." In most fields you want to have some separation between those who put the data together and those who use it. Ideally it is best if those who put the data together have no idea what the data is going to be used for. But Roth who seems to have extremely strong political views has not ensured a separation in data gathering and use. Such separations are expected in most empirical work that I am familiar with. No actual bias necessarily occurs and even unconscious effects might be avoided, but the data has more credibility with others if precautions are taken. For example, those gathering the data should not even know what it is going to be used for.
It was also interesting that he had no desire to try to reconcile the data that he gets with state level and other patterns, such as those just discussed with prohibition.
I would have liked to have seen him use newspaper reports from today to construct the homicide rates that we see. Could he use newspaper reports to accurately construct the changes in crime rates? I doubt it.
I have also rarely seen an academic seminar where someone crops the ends of his figure (e.g., cutting off the crime data in 1992) to exaggerate the differences that he is trying to claim exist.
UPDATE: Here are number for the period that Roth claims that murder rates were essentially flat. Note that individual states were adopting prohibition rules over this period. Bureau of Justice Statistics, DOJ I don't disagree with the claim made that murder rates fell dramatically after prohibition ended (though his claim that it was just due to FDR wasn't explained.
1900 - 1.2
1901 - 1.2
1902 - 1.2
1903 - 1.1
1904 - 1.3
1905 - 2.1
1906 - 3.9
1907 - 4.9
1908 - 4.8
1909 - 4.2
1910 - 4.6
1911 - 5.5
1912 - 5.4
1913 - 6.1
1914 - 6.2
1915 - 5.9
1916 - 6.3
1917 - 6.9
1918 - 6.5
1919 - 7.2
1920 - 6.8
1921 - 8.1
. . .
1933 - 9.7 (last year of prohibition
posted by John Lott at 4:57 PM
9 Comments:
Perhaps he studied his statistical methods by reading the classic-pamphlet "How to Lie With Statistics" by D.Huff
I was also at that talk. I don't share your impression that Roth had strong political views that colored his findings. His central claim was that homicides of strangers tend to increase with the general breakdown of law and order, the political legitimacy of the state, and the like. This is not exactly an ideologically charged argument. Moreover, the actual murder rate in the early 20th century is not crucial to his case, so cooking the data in this respect would not help him. Your argument that he's wrong about the murder rate during Prohibition presumes that you have accurate data, something he disputed during his talk. Where do your data come from, and why should we believe they are better than his?
Dear Anonymous:
The talk was litered with comments that any woman who separates from a man should immediately get a restraining order to exaggerations on the rate of crime across races to claims about there not being a significant black/white crime differential in the US. The emphasis on the breakdown of law and order was a very small part of his talk and was brought up when his other explanations couldn't explain anything. It is disappointing that you didn't notice how he went out of his way to exaggerate the differences in crime between the US and Europe.
Unfortunately, Roth never explained why his created numbers from newspaper stories and coroner reports were better than the Bureau of Justice Statistic numbers. I have absolutely no problem arguing that the BJS numbers are not perfect, but they are not off in 1900 by a factor of seven or so.
I was also at the talk. Roth pointed out that the ultimate source for the DOJ’s estimate of the annual homicide rate--the National Center for Health Statistics, Vital Statistics--was incomplete until 1933, when Texas joined the Death Registration Area (DRA). When it was first implemented the 1900, the DRA included only 10 states--Maine, Vermont, New Hampshire, Massachusetts, Rhode Island, Connecticut, New Jersey, Indiana, Michigan, and the District of Columbia--and the District of Columbia, together containing only 26 percent of the nation’s population. As you can see by the list, most of the earlier DRA states were in the Northeast, where the homicide rate was lowest. The last two states to join, LA & TX, had much higher homicide rates. Thus, much of the apparent increase in the homicide rate between 1900 and 1933 is not real. Roth, by the way, did not explain why his "created numbers from newspaper stories and coroner reports were better than the Bureau of Justice Statistic numbers" because there were no Bureau statistics before 1900. Roth's talk focused on the eighteenth and nineteenth centuries.
I also agree with Anonymous about not seeing a link between Roth’s political views and his findings. Indeed, I couldn’t quite tell what Roth’s political inclinations even were. His earlier research on early America refuted Michael Bellesiles’ argument that a lack of guns in early America explained lower homicide rates and his emphasis on the important role of law, stability, and the authority of the state suggests that Roth is no simple “gun basher.” Nor did I see Roth's emphasis on the breakdown of law and order as a small part of his argument--it seemed absolutely central to it.
Finally, while I would agree with your assertion that “it is best if those who put the data together have no idea what the data is going to be used for,” I don’t know of any historical projects that meet that standard. With little or no funding, historians must their own construct datasets. Roth, I thought, takes the only defensible route. He (1) acknowledges the substantial subjectivity in estimating a homicide rate in the pre-1933 period, (2) makes his best judgment about which deaths were homicides, probable homicides, possible homicides, or ordinary deaths, (3) evaluates the impact of different definitions on his argument, and (4) places his data on the web for others to critique, re-define, or use in any way they see fit. This, in my estimation, is model social science history.
Thanks very much for the note David.
1) Even if I agreed with the claims, Texas and Louisiana only made up 5.7 percent of the US population in 1900 (most of that is for Texas), and even then I do not believe that Texas had a particularly high murder rate. (It wasn't until after WWI that their share of the US population seems to have really picked up.)
I don't understand why your reference to the rest of what he talked about in the seminar is relevant because Roth also made an adjustment to the crime rate during the pre-1933 period covered by the BJS numbers. That is what I was referring to.
Since you raise the issue, however, I am dubious that looking at newspapers and coroner reports (with him subjectively deciding whether the coroner and newspapers got these things correct) is very useful.
The increases in crime at the state level correspond to when those state passed prohibition. He never responded to that point. If he did, I apologize and I would be greatly appreciative if it were explained to me.
The examples that he gave on what types of crime occurred is a perfect example. My guess is that extrapolating out what types of crime exist from reading newspapers (which seemed to be where he got information on the nature of the crime) is a very risky undertaking.
His claims about crime rates also differ from other historians. Just take the case of Joyce Lee Malcolm's book "Guns and Violence" (Harvard University Press) regarding England. She does not have the sudden drops in murder at two points that Roth claimed occurred in the 1800s. There are other historians who have gone through newspaper articles on the American West and they claim that it was nowhere near as violent as others have claimed. Why not some discussion of why he disagrees with those historians?
2) I guess that I have little sympathy with the notion that it is OK to not put in even the most basic controls over quality for something that is the center piece of the research. Here is a suggestion: limit oneself to data that one can do a good job on.
3) Possibly by the New York Time standards this comes across as balanced. I gave a list of comments that I didn't think were very objective. Possibly you can respond to them. Could you explain his advice on all women getting a restraining order when they separate from men? Do you know that the murder rate by whites in the US compares similarly to the white murder rate in much of western Europe? Could you also explain why he cut off the crime data in 1992?
Thanks again very much.
John,
I’m way out of my field, here, so I can’t really respond to some of your questions and the literature you referenced. I’ll take your word on it regarding the frontier & England. I do know a little about the Death Registration Area, however. It goes from 26% coverage in 1900 to 100% in 1933. Although I used only the examples of TX & LA to illustrate the effect, the important point is that the DRA was initially dominated by New England, the region (and here I take Roth’s word) with the lowest homicide rate.
You’re absolutely correct that Roth did not respond to your comment about state-level data, however (although to be fair to Roth, we were out of time and I was pressing him to wrap it up). Although my comment above about the DRA suggests an increasing downward bias in the national homicide rate as you move back in time, state-level data would be unaffected. The positive correlation with the passing of prohibition laws and state homicide rates seems to me to be strongly suggestive, intuitive, and quite interesting. If it’s not published already, it should be.
Perhaps I am a very bad judge of political bias. As you are no doubt aware, the social sciences tilts strongly to the left. I’ve seen enough far left presentations that (perhaps) a mere left-leaning presentation strikes me as moderate and objective. I thought that Roth’s comments about restraining orders were given in a jocular way—an attempt (unsuccessful perhaps) to be humorous. Perhaps I am being too generous. I should add that I’ve met Roth on several occasions at the Social Science History conference and invited him to speak. He’s always struck me as a person with unimpeachableintegrity who is willing to let the data fall where they may. Thus his complete rejection of the Bellesile thesis from day 1 when most historians were singing his praises. We all wrestle with our biases, however.
Being a 19th-century historian, I found my eyes drawn to the first part of the plots and didn’t even notice that the data were cut at 1992. I have no idea why.
One last comment: As you note, historical data are less than ideal (to put it mildly). As historians, however, we are committed to studying the past as best we can. To conduct research on US fertilty decline (my field), I'm forced to rely in indirect methods with some (perhaps great) potential for error. I persist, however, because I find the topic fascinating and worthy of study. I do the best I can. I believe that Roth does as well. We're honest, however, with its limitations.
Thanks for your comments and thanks also for participating in the seminar.
There is another variable that you may (or may not) want to consider. The medical field has experienced incredible gains during the past century. People who died 100 years ago from result of gunshot wounds, stabbings, beatings, etc. now would live.
Our young people today wear the scars of multiple GSW like badges of honor.
If anything at all, this would skew the homicide numbers even farther. The homicide rates of today would be much greater if not for advances in the medical field: ambulances, air lift capabiltity, EMT training, antibiotics, and on and on.
As a retired police officer (1974-98) I see an astounding change in attitudes towards violence amoung the young.
As someone who has used newspapers extensively to create a database of homicides in San Francisco from 1850-2003, I too have had the source questioned. I think I've answered that ojection adequately in my books but here I would like to respond to Dr. Lott's comment that "I would have liked to have seen him use newspaper reports from today to construct the homicide rates that we see. Could he use newspaper reports to accurately construct the changes in crime rates? I doubt it." In January 2007 I started clipping San Franciso homicide stories from the San Francisco Chronicle and my count as of June10 matches that of the Police Department. The newspaper stories are a bit thin on detail but they get the count right.
Kevin J. Mullen
Dear Kevin:
Thanks very much for your useful post. I would say though that I don't think that SF Chronicle coverage is the same for all newspapers or even most of them from the work that I did for my book, The Bias Against Guns. The reason is that from interviewing crime reporters I know that they don't cover all murder victims, particularly if you are talking about cases where gang members or criminals have been killed. Among the top ten largest newspapers in the US in 2001, I found only two that had it as their policy to serve as the newspaper of record and to record all murders.
Post a Comment
<< Home