| Dinesh D'Souza |
| On page 408 of his book The End of Racism are false numbers about US government estimates for how often the crime of rape is committed when police have recorded the race of perpetrator and the race of the victim. On that page D'Souza claims that in a recent year (1992) the ratio of the estimated numbers of black-against-white rapes to white-against-black rapes is 200 to 1. As I describe below, this ratio is 16 times larger than the real ratio one gets when one investigates the books D'Souza cites in the bibliography of The End of Racism. I phoned D'Souza and found his explanation for the false claim totally unbelievable. The reference in the bibliography for this ratio oddly cites three different books as sources for two numbers used to create the ratio (as if one source wouldn't be sufficient to show readers two numbers needed for such a ratio!). The first source D'Souza gives is an FBI yearly crime almanac (a volume of their Uniform Crime Reports series of books). I found a copy of that book and examined it carefully. The two pages D'Souza cited in that book have data about murder only, and about no other crime. The only crime in issues of this publication for which race of perpetrator and race of victim is given is for the crime of murder, and only on those two pages. A second source he gives is another yearly crime-data summary book: one of the yearly BJS (Bureau of Justice Statistics) publications called Crime in the United States. Each of the yearly books in this series has a page that shows a table of data about race of perpetrator and race of victim for the crimes of rape, murder, robbery, and a few other crimes. However, for many cells there wasn't enough data to permit calculating a sufficiently-reliable estimate (an asterisk is put in such cells instead of a number), and I saw that the two cells needing data for D'Souza's page-408 claim were all empty, containing only asteriskes. The explanation for the empty cells (given at the bottom of the page) reads "estimate based upon about 10 or fewer sample cases". Introductory pages in issues of this yearly explain the empty cells this way: they obtain such data by randomly selecting, then phoning police departments around the USA, and asking for files about crimes to be pulled out and inspected. Often the race of perpetrator or victim wasn't noted by the policeman who originally made the report. However, I searched other years of the same series of books, and I was however able to find years that have data in the cells for black-on-white and white-on-black estimates of occurances of rape. The ratios of estimates from those cells was 12.5 to 1, which 16 times less than the 200 to1 claim by D'Souza's claim in on page 408 of his book. The third source given is a page of the book The Myth of a Racist Criminal Justice System, by William Willbanks. Like the page of the FBI book, that page has no data about rape either. Note: I phoned D'Souza and spoke with him at length about these "empty references" that totally fail to support the claim on page 408 of his book. He blamed it on a research assistant. I find that totally unbelievable, because four strange things about the wording in the main text (written by himself, presumably) seem clearly intended to cause confusion to anyone considering investigating the three source-books cited in the bibliographic entry. Firstly, the wording does not explain why he cited only one year's worth of data to support the generality about whites and blacks. Secondly, he merely implied the generality the data was being used to defend. The generality could be worded this way: "whites are less likely to commit racially motivated crimes against blacks than blacks are to commit racially motivated crimes against whites". Why not state this claim explicitly? Why merely imply it? Thirdly, he oddly selected a specific year (1992) without giving any explanation for why he selected this year rather than some other recent year. The text implies a generality about all years, but doesn't defend his selection of that specific year as being more representative than other years. His failures to explain why he selected only one year, and why he selected the year 1992, combine to help him hide a strange thing about the reported data for the one year that he mentioned: the year 1992 has no data on the relevant cells. If I had stopped there (and not bothered to look at other issues covering other years) I wouldn't have discovered the huge discrepancy between D'Souza's 200 to 1 ratio and the 12.5 to 1 ratio that I derived from issues containing data for the relevant cells. Another bizarre thing about D'Souza's wording on page 408 of this book is the very odd assumption that behavior of criminals of a particular race of human beings can be reasonably used to support generalizations about all of the persons of that race. This assumption is outrageous to me. For one thing, if it really is valid to do this, D'Souza needs to explain why estimates from the same books reveal that white criminals also show a strong preference for white victims. E.g., something like %95 of robberies with a white perpetrator have a white victim. Would D'Souza be consistent, and claim that this also is evidence of anti-white racism: one felt by white criminals against victims of their own race? What could explain such "racism"? D'Souza also needs to explain why these numbers aren't readily explained in other ways. E.g., he needs to account for what could be called "victim-availability bias". This factor alone would readily explain why both white and black criminals in our society exhibit a tendency to select whites as victims. There are far more whites than blacks in our society. Whites outnumber blacks by around 7 to 1. If criminals of each race selected victims in a totally random manner, the ratio D'Souza cites would be equal to the ratio of white and black potential victims in the general population (in the case of America, 7 to 1). |
| A lie about government crime data (and it surely was intentional) |
| I, Jeff Opal, am responsible for everything on this web page. I last modified the page on 11-JUL-2003. |