When I report that 80 percent of federal and state prisoners are drawn from the working class, a natural reaction is to wonder whether that figure is exaggerated by statistical artifacts, and yet I say in the text that 80 percent is more likely to be an underestimate than an overestimate.
Let’s start with the artifacts that might exaggerate that percentage. One possibility is that we’re looking at an artifact of education. Prison inmates have disproportionately high dropout rates from secondary school. In 2004, the year of the last inmate survey, 62 percent of white male inmates ages 20–49 had fewer than twelve years of education, quadruple the 15 percent of white males ages 20–49 in the general population. Because becoming a high school dropout is likely to consign a person to Fishtown, perhaps the prison population includes many young men who grew up in middle-class or affluent neighborhoods, got in trouble, dropped out of school, worked at blue-collar jobs, and look as if they come from Fishtown.
But this cannot possibly be a large artifact, because parents outside Fishtown have so few children who do not complete high school. Consider the 1979 cohort of the National Longitudinal Survey of Youth, born in 1957–64. Among white males who were the children of Belmont parents, only 2.5 percent dropped out of high school. Of all the white males who dropped out of high school, 85 percent had Fishtown parents. White high school dropouts in prison who were raised outside Fishtown can account for only a small proportion of the prison population.
Another possibility is that the prison data underrepresent white-collar crime. Let’s assume that when people in Belmont commit crimes, they are mostly crimes involving embezzlement or fraud—the only two offenses in the Uniform Crime Reports that might be characterized as white-collar crime—and such crimes result in prison sentences less frequently than crimes such as robbery or burglary.
This cannot represent a large artifact, because embezzlement and fraud constitute such a small proportion of serious crime. In 2008, the FBI reported 117,217 arrests of whites for fraud and 10,517 for embezzlement. The percentage of arrests for serious crime this represents depends on how you define serious. The offenses in the crime index are murder, forcible rape, robbery, aggravated assault, burglary, arson, larceny-theft, and motor vehicle theft. If you count just index crimes plus fraud and embezzlement as serious, then they represent 10 percent. If you add in some other crimes that are not index crimes but seem as serious as fraud and embezzlement—assault, forgery and counterfeiting, and dealing in stolen property—then the percentage is 6 percent. If you add in drug offenses, the percentage drops to 4 percent.
Furthermore, fraud, for which whites were arrested ten times more frequently than for embezzlement, often consists of traditional con-man frauds. People doing bait-and-switch cons on the street are not what we have in mind when we think of white-collar crime. Consider also that the numbers I just gave are based on 2008, when about 18 percent of the people classified as white by the FBI were Latino whites. Even if we assume that the Latino crime rate is no higher than the non-Latino crime rate (an incorrect assumption), the estimated total arrests for non-Latino whites must be adjusted downward accordingly.
Taken together, it is impossible to postulate a rate for white-collar crime committed by non-Fishtown residents that would materially affect their responsibility for all serious crime, unless you assume, with no corroborating evidence, that there is serious undetected white-collar crime of mammoth proportions.
A third possibility is that people outside Fishtown have better lawyers and so go to prison less often than people from Fishtown who are accused of the same type of crime. There is no exact way to estimate the size of that effect, but a few observations are possible. Some offenders from Belmont probably avoid prison time because they (or Mom and Dad, in the case of young offenders) hire a good lawyer—but even the best lawyers have a hard time getting probation for their clients if the court is looking at the second or third arrest for a class 1 felony. By the same token, the prevalence of sentencing guidelines means that first-time offenders often avoid prison sentences even without good lawyers. One may accept that Belmont offenders come to the criminal justice system with better representation than many Fishtown offenders without having a basis for thinking that the discrepancies in sentencing will produce statistically important changes in the proportions of offenders who appear to come from Fishtown.
Now let’s turn to the other side of the ledger, and the opposite hypothesis: The estimated percentage of white criminal activity coming out of Fishtown is underestimated, perhaps grossly underestimated, because I am counting prisoners instead of crimes.
Ever since criminologist Marvin Wolfgang’s pioneering longitudinal study of all of the males born in Philadelphia in 1945, scholars have found that a small proportion of those who are ever arrested account for about half of all offenses.1 The exact size of that proportion has varied by study, but it has usually been in the neighborhood of 7 percent, leading to a term of art in the criminological literature, “the dirty seven percent.” Since people are incarcerated partly because of their past criminal history as well as their current offense, people in prison have a much higher mean number of arrests than do members of an entire birth cohort, but the pattern is the same. Figure E.1 shows the story graphically.
FIGURE E.1. CONCENTRATION OF ARRESTS AMONG A MINORITY OF PRISONERS
Source: 2004 survey of state and federal inmates. Sample limited to white males ages 20–49.
That seemingly perfect mathematical function represented in figure E.1 is not a fitted line. It was drawn from the raw data. In the 2004 inmate survey, more than half of all prior arrests of white male prisoners ages 20–49 were accumulated by just 13 percent of them. More than three-quarters of all their prior arrests were accumulated by just 31 percent of them.
Prisoners from the different neighborhoods had different arrest histories. Those from Belmont averaged 4.0 arrests prior to the one that landed them in prison and those from Fishtown had 6.3. The differences are actually even greater than that, because age is strongly related to number of prior arrests (as one would expect), and the average ages of prisoners from Belmont and Fishtown were 38.0 and 33.6, respectively. After controlling for those differences in age, the number of prior arrests for a typical prisoner age 30 from Fishtown was 2.4 times that of a typical prisoner age 30 from Belmont.2 These differences alone would make the proportion of crimes coming out of Fishtown much larger than it appears from counting prisoners.
Consider next that offenders are arrested for only a fraction of the crimes they commit. The typical prisoner is believed to commit somewhere in the neighborhood of twelve to fifteen non-drug-related crimes in the year prior to his imprisonment.3 That distribution is highly skewed. In a study of Wisconsin prisoners by John Dilulio and Anne Piehl, the median of non-drug-related crimes is 12, but the mean is 141.4 In an earlier Rand study of self-reported crimes, 50 percent of convicted robbers reported fewer than 5 robberies in the year prior to incarceration, but 10 percent said they had committed more than 87 that year. Among active burglars, 50 percent had committed fewer than 6 in the year prior to incarceration, while 10 percent said they had committed more than 230.5 Even if we discount for braggadocio, members of the top quartile of prisoners probably committed dozens of crimes in the year before they were locked up. If you have to predict which prisoners fall into that top quartile based on their prior arrests, the logical expectation is that the arrests and self-reports are correlated. Since Fishtown prisoners have substantially more prior arrests than prisoners from Belmont, an estimate of the proportion of total crimes coming out of Fishtown would once again rise.
All in all, it is a lot easier to make the case that 80 percent is too low, not too high, as an estimate of the degree to which white male crime is produced by men with very low education and working (when they work at all) in blue-collar jobs.
There are two main technical questions that arise about Figure 10.3: Can we legitimately use arrest rates as a proxy measure for criminal activity? Can we use the profile of the prison population to draw inferences about the profile of the arrestee population? I discuss each in turn.
Using the white arrest rate as a proxy for the white crime rate. In the FBI’s Uniform Crime Reports (UCR), statistics on reported offenses do not include the race of the offender (which is often unknown, especially for property crimes), let alone the educational and occupational background of the offender. But we do have arrests reported by race, and we do have the neighborhood breakdown of prison inmates; using those resources, we can establish a plausible estimate of the white crime rate by neighborhood.
We begin with the fact that reported offenses and arrests of whites are highly correlated. From 1960 to 2008, the correlation between the overall violent crime rate and white arrests for violent crime was +0.92. For property crime, the correlation was +0.91.6 Figure E.2 shows how the proportional changes look for violent crime, using 1960 as the baseline equal to 1.
FIGURE E.2. WHITE ARREST RATE FOR VIOLENT CRIME AND THE OVERALL VIOLENT CRIME RATE
Source: UCR crime data.
There is good reason to think that changes in white arrest rates tell us a lot about changes in white criminal activity.
Using the socioeconomic profile of prisoners as the basis for estimating the neighborhood distribution of whites arrested for index crimes. The next question is what percentage of whites who are arrested for index crimes comes from Belmont and what percentage comes from Fishtown. We know what the profile of white prisoners looks like and that the profile did not change appreciably from 1974 to 2004. Is it reasonable to assume that the educational and occupational profile of persons arrested for index crimes is similar to the educational and occupational profile of prisoners? The answer is not only yes, it is once again quite possible that the educational and occupational profiles of persons arrested for index crimes are more heavily skewed toward Fishtown than the prison population is.
The key to that conclusion is the specification of arrests for index crimes. If we were talking about a minor offense such as drivers who are stopped for speeding, the socioeconomic profile of offenders would probably be not that much different from the profile of the general population of the same age and sex. If we were talking about a somewhat more unusual offense such as arrest for possession of marijuana, the population of offenders would deviate further from the general population of the same age and sex, but it wouldn’t be like the profile of the prison population. But an arrest for an index offense means an arrest for murder, forcible rape, robbery, aggravated assault, burglary, arson, larceny-theft, and motor vehicle theft. In 2008, arrests for violent index offenses amounted to only 4 percent of all arrests and arrests for the property index offenses amounted to only 12 percent of all arrests. This is a highly selective group of arrestees.
Meanwhile, the educational and occupational profile of prisoners is based on everybody who is in prison, many of whom did not commit index offenses. In the 2004 survey, for example, the offenses for which the prisoners were serving time were split almost evenly between index and nonindex crimes (53 percent were incarcerated for an index offense), and the educational and occupational levels were both higher for prisoners found guilty of nonindex offenses. The differences were small (for example, 18 percent imprisoned for a nonindex offense had education beyond high school, compared to 13 percent of those imprisoned for an index offense), but the data do indicate that persons imprisoned for index crimes are more intensely concentrated in Fishtown than persons imprisoned for nonindex crimes.
The text of chapter 10 references my exploration of data from the Securities and Exchange Commission and the Internal Revenue Service and states that the evidence from them cannot be used to demonstrate systematic changes in business integrity. The following summarizes the results of those efforts.
The SEC is responsible for policing securities markets. They identify and prosecute wrongdoers, which would seem to make the SEC an excellent source of data. But the SEC has not published anything comparable to the FBI’s Uniform Crime Reports data for assessing changes in corporate malfeasance over time.
The one partial exception to this is the Accounting and Auditing Enforcement Release (AAER) that the SEC issues at the completion of an investigation of alleged wrongdoing. But the annual number of AAERs did not pass 100 until 1994, and has never been higher than 232. The number of publicly traded companies is about 15,000.7 With a numerator in the low hundreds and a denominator in five figures, trends are uninterpretable—they could as easily reflect changes in staffing or administrative policy as real changes in corporate malfeasance, or simply represent random noise.8
The most serious violation of the tax code is tax fraud. As of 2005, tax returns were filed for almost 6 million corporations and 25 million proprietorships and partnerships. That same year the IRS assessed 217 civil penalties for corporate income tax fraud. It is impossible to interpret trends with such data, for the same reason that trends in AAERs are uninterpretable.9
The data on lesser tax infractions are a little more interpretable, but not much. Two categories of offense have consistent definitions over the years: delinquency, which refers to the failure to file tax returns on their due date, and failure to pay, which can include any amount short of the total that the IRS eventually decides you really owed the federal government. Figure E.3 shows the trends for corporations.
FIGURE E.3. TRENDS IN DELINQUENCY AND FAILURE TO PAY FEDERAL INCOME TAXES: CORPORATIONS
Source: 2009 IRS Data Book, table 17, and comparable tables from earlier editions.
For corporations, the picture since the mid-1980s has been one of steady decline in both the rate of delinquencies and failures to pay.
Scholars have tested other indicators of corporate malfeasance with varying degrees of success, but all of the reliable ones use financial measures that must be extracted from the detailed financial statements filed by corporations, not ones that are reported in aggregate data about U.S. corporations. See Cecchini et al., 2010, for a description of one of the most recent ideas and a literature review of other attempts. The simplest measure, applied in Prechel and Morris, 2010, is to use a restatement of corporate finances given to the SEC within a given year (for reasons other than a change in accounting standards) as suggestive of malfeasance. Apart from the merits and shortcomings of that measure, I was unable to find a way of assembling a longitudinal database using that measure with anything short of a major research project.
In addition to the evidence on bankruptcy presented in chapter 10, I looked at IRS data for individuals, but ran into the same problems of interpretability that IRS data have for corporations. For individuals, unlike corporations, the annual cases of tax fraud run into the thousands, so it is at least worth looking at the trendline. It is shown in figure E.4.
FIGURE E.4. TAX FRAUD: INDIVIDUALS
Source: 2009 IRS Data Book, table 17, and comparable tables from earlier editions.
From 1984 to 2000, the rate of tax fraud dropped steeply, leveling off thereafter. Perhaps this reflects reductions in IRS investigative resources or other administrative artifacts but, at the least, there is certainly no evidence of increased dishonesty.
For the much lesser but also more common offenses of delinquency and failure to pay, the trends go in the opposite direction. The trends since 1978 are shown in figure E.5.
FIGURE E.5. TRENDS IN DELINQUENCY AND FAILURE TO PAY FEDERAL INCOME TAXES: INDIVIDUALS
Source: 2009 IRS Data Book, table 17, and comparable tables from earlier editions.
In contrast to the corporate trends, both delinquency and failure to pay rose over the period. The problem is that delinquencies and failure to pay on the part of individuals coping with a notoriously complex American tax code can reflect carelessness, procrastination, or an honest mistake, with no implications for integrity. They can also reflect a decline in integrity. There is no way to untangle which causes play what role.