In response to:
Genes & Crime from the February 12, 1987 issue
To the Editors:
I’m pleased that Christopher Jencks thinks my book has the “enormous virtue” of focusing attention on the societal determinants of crime [NYR, February 12]. But I’m a little bewildered by the substance of his review; at times, I get the odd feeling that the book Jencks reviewed isn’t quite the one I wrote.
Jencks says I argue that crime is “a product of economic inequality.” Well, that’s partly true, but it greatly oversimplifies what I really said. The central argument in Confronting Crime is that our unusually high crime rates can be traced—in part—to a pattern of development distinguished by the “relatively unbuffered play of market forces.” One result was an unusually wide spread of income inequality, but there were many others: greater economic insecurity, less provision for the casualties of technological and economic change, greater disruption of families and communities, and a culture that—even more than in other developed capitalist societies—underemphasized communal and cooperative values and overemphasized individual acquisition. Other factors, I said, were also important, especially our uniquely tragic racial history.
Now, I’m not sure what Jencks thinks about this larger argument, since he doesn’t really engage it. Instead, he too often reduces the debate to a battle between straw men; a narrow focus on the effects of income distribution on the one hand and an equally narrow focus on cultural factors on the other. Some examples:
Early on, Jencks chides liberals—presumably including me—for thinking that inequality is simply a matter of money, and not understanding that reducing it would require us to “reorganize work as well as redistribute income.” Sure it would; but I said so myself, at some length. Jencks acknowledges that I’d “doubtless” agree about the importance of work—but doesn’t tell his readers that inadequate work is central to the book’s diagnosis and a decent employment policy central to its strategy.
Later, Jencks uses the example of India, a poor country with a relatively low crime rate, to argue that income inequality alone can’t explain international differences in crime rates, and that cultural factors must be involved too. Fine; but again, I’m not sure who Jencks is arguing with. I spent some time explaining why some poor countries managed to maintain low crime rates, and India was my main example. I argued—following the work of Clayton Hartjen at Rutgers—that India’s poor were often integrated into a supportive communal life, involving useful work for the young, strong ties of family, religion, and subcaste, and a cultural ethos encouraging a sense of membership in a larger, cohesive community. I suggested that the disruption of these bonds by a process of market-driven development was partly responsible for the high crime rates in many Third World countries. Jencks may or may not agree with this view, but it’s hardly one that rests on the effects of income differences alone.
Jencks similarly takes me to task for suggesting that rising unemployment—and hence rising crime—among black youth in the Sixties resulted from “economic conditions” alone. But I didn’t say that. Early in the book, I said I thought that the weakening of cultural supports for racial subordination played a part in the rising rates of inner-city crime in the Sixties. But so, I thought, did a lot of other things: the relative shrinking of opportunities for low-skilled blue-collar work, just when agriculture, traditionally a crucial employer of black youth, was employing fewer and fewer of them; the corrosive long-term effects of this on black family life and community cohesion; the impact of crowding increasing numbers of marginal poor cheek to jowl with increasingly affluent strata in the cities; the growth of the hard-drug trade; and more. (There is also the sheer size of the youth cohort reaching adolescence in the late Sixties, which I noted briefly but should have said more about.) A mix, in short, of economic, cultural, social, and demographic factors. But Jencks apparently wants to argue that the roots of this volatile situation were cultural rather than economic; indeed, he pins most of the blame on the civil rights movement, for leading black kids to want jobs they weren’t qualified to get. But does Jencks really believe that it was the baleful influence of the civil rights movement alone, and nothing else about the structural situation of black youth, that led to their rising crime rates? And if, as I suspect, he doesn’t really believe that, what’s the point of exaggerating?
A final example. Jencks believes that the argument for a close connection between inequality and crime is “discredited” because, though income inequality has recently increased in the United States, crime has gone down. This argument isn’t exactly wrong, but, again, it’s misleadingly simplistic. Inequality has indeed risen, but to say, without further elaboration, that crime has gone down obscures the complexity of what’s been going on. Crime dipped, sometimes strongly, in the early Eighties. But the homicide rate—which both Jencks and I think is our most reliable indicator—has been rising again since the end of 1984, and in 1986 rose sharply, even terrifyingly, in some cities; equaling or nearly equaling the peaks they suffered in the late Seventies.
Moreover, this has taken place in the face of the doubling of the national prison population and its near-tripling in many states since the early Seventies. Now, I’d be the last to exaggerate the effectiveness of imprisonment in reducing crime; but I know few criminologists who’d argue that this incarceration “binge” has had no effect on the crime rate. Hence I think it’s more appropriate to ask what has kept the rates so high even though we’ve locked up hundreds of thousands of serious offenders: and the growing deprivation and marginalization at the bottom of the social ladder (and its complex relationship to the illicit drug trade) is a prime suspect.
Even more importantly, Jencks’s argument, once again, depends on an overly narrow view of the dynamics of inequality. We wouldn’t really expect crime in shoot upward dramatically whenever the index of income distribution moved upward a few points. A deeper view of the effects of inequality looks to the long-term more than the short—to the impact of growing deprivation (and declining social services) on family functioning, health care, and nutrition, long-range prospects for decent work, levels of child abuse and neglect, and much more. The most troubling consequences are unlikely to appear overnight; but I think we’d be naive not to expect them to haunt us down the road.
Jencks concludes with a singularly bleak view of the state of our knowledge about what we might do to reduce crime—telling us, indeed, that criminologists don’t have “any reliable advice” on how to bring down the crime rate. But if Jencks is really going to accuse the entire criminological profession of having nothing useful to say about reducing crime, I think he’s under some obligation to engage the arguments they’ve in fact made more seriously than he has done, and show why he thinks they’re wrong.
In Confronting Crime, I suggested a range of strategies I thought could help bring down the crime rate. They included, among others, a tougher response to men who beat their wives, intensive probation for less dangerous criminals, better rehabilitation programs for young offenders, more resources for family support programs, family planning services, assistance to teen-aged parents, and quality preschool education; and, above all, comprehensive, intensive job training and the direct creation of permanent, useful jobs. I didn’t suggest that these were the only strategies we should consider, or that they would be easy, or that they could reduce crime overnight even if we mustered the political power to tackle them seriously. I did think they showed promise. Now Jencks may really believe that none of them would make a difference; we don’t know, since he doesn’t mention any of them. If he does believe that, he ought to be willing to defend that rather extreme position more directly.
To the Editors:
My first purpose in this letter is to express appreciation for Christopher Jencks’s fairminded and thoughtful review of Crime and Human Nature. But, as long as I’m writing, it may not be out of place to comment further.
First, Jencks observes that our book has chapters on “genes, gender, age, IQ, personality, schools, television, drugs, even race,” then criticizes it for having none on “the effects of either social class or poverty.” But Chapter 12, on “labor markets,” which his review fails to mention, presents what evidence there is on crime in relation to economic need, to objective inequality between the rich and poor as well as the envy of the poor for the rich, to the temptations of being poor among the affluent, and to unemployment. These economic variables per se (as distinguished from the individual characteristics associated with them) account for little of the crime in America. In the chapter’s first sentence, we point out that economic variables would probably be more important if the extremes of wealth and poverty or the sheer needs of the poor were greater than they are here.
Second, Jencks dismisses studies showing identical twins to be more correlated than fraternal twins for criminal behavior. This is the crucial datum for the twin method of quantitative genetics, inasmuch as it suggests that genes, which identical twins share more of than fraternal, make a contribution to the trait in question. Jencks dismisses the evidence from twins because he believes that identical twins are likely “to have more influence on one another than fraternal twins.” That’s a plausible hypothesis, but, as our book said, the evidence, gathered by David Rowe of the University of Oklahoma and others, fails to support it as far as criminal behavior is concerned. Since the book came out, additional evidence has all but refuted it. The main reason identical twins are more similar in criminality than fraternal is, it seems even surer now than when the book was written, their greater genetic overlap. Jencks further says that the large Danish adoption study, one of the major studies implicating genes, implies only a 2 percent genetic contribution to the variance in criminality. However Jencks derived this estimate, it is at odds by a factor of about twenty or more with any general conclusion that is supported by the totality of available data.
Third, Jencks interprets the much larger crime rate in American cities and suburbs, as compared to the countryside, as evidence of a major nongenetic influence on crime. The reasoning tacitly presupposes, most improbably, that the genetic makeup of rural and city-suburban populations are the same. The racial composition of cities and countryside, for example, differs, and there are probably other genetic differences. Robert Gordon of Johns Hopkins University has concluded that the lifetime prevalence of crime for individuals is independent of community size above communities of 10,000, after correcting for the differing crime rates of blacks and whites. Jencks’s lapse of reasoning here is a prime example of a common weakness in sociological theory, which is to infer a purely sociological explanation for data after having tacitly assumed it.
Fourth, Jencks discusses some ideas he has about how genes affect IQ, which he thinks we were remiss in neglecting. It is not necessarily that genes endow some people with more learning ability than others, he suggests, but that they affect how people treat one and also how one behaves. For example, perhaps a “high IQ” gene is really a gene that gives one a certain sort of face which people have arbitrarily decided is “intelligent.” Then, such people are treated according to expectations, with the result that they become intelligent. Or, perhaps a high IQ gene gives one an appetite for reading and other cerebral activities, the satisfaction of which produces a high IQ. These ideas (some of which have, in fact, failed to be supported in studies evaluating them and others of which have not been, and perhaps cannot be, evaluated) would not have changed our conclusions about the correlation between IQ and crime, and that is why they were omitted. Rather than restate all those conclusions here, it should be enough to say that we did not conclude that the correlation between crime and IQ is inevitable or necessary. It is merely a fact, which most criminologists have steadfastly ignored for several decades. Jencks, in other words, agrees with us that a society can do better in diverting academically disadvantaged people from crime than ours is doing now. The more interesting issue is whether the liberal outlook is a help or a hindrance in this regard, which brings me to the final point.
Jencks has written his review primarily to convince those on the political left not to assume that genetic influences on socially significant behavior are so unwelcome that the only acceptable response is to deny them. That, I believe, is a worthy purpose, but my reasons differ. What is wrong with denial is not that the news about genetic influences is not so bad (it may or may not be), but that the news is probably true. If it is true, however unwelcome or not it is, then denial is both futile and destructive. Jencks believes that the leftist outlook can be maintained intact in the face of evidence that people differ significantly in ways that society cannot readily change. I doubt it. Here, as often before, science is challenging belief. In my way of thinking, belief should yield, and I think that that may be how Jencks and I differ most.
Richard J. Herrnstein
Christopher Jencks replies:
My discussion of both Currie’s Confronting Crime and Wilson and Herrnstein’s Crime and Human Nature deliberately focused on what I took to be their most politically important arguments. I warned readers that Wilson and Herrnstein’s book dealt with a much wider range of issues than my review did. Unfortunately, I did not issue a comparable warning about Currie’s book. I therefore have considerable sympathy with Currie’s complaints. I did not mean to convey the impression that Confronting Crime treated income inequality as the sole determinant of crime rates. It does not.
While I regret not having conveyed the full range of Currie’s argument, I am as skeptical about this larger argument as about the portion I discussed in my review. Currie tells us that a “decent employment policy” is central to his strategy for reducing crime. At a minimum, this presumably means ensuring that everyone who wants a job can get one. All the available evidence suggests that this would reduce crime. But none of the evidence suggests that the reduction would be large.
Richard Freeman recently reviewed ten different studies that had looked at trends in unemployment and crime. Nine found that crime fell when unemployment fell. Nonetheless, Freeman concluded that cutting unemployment from 10 to 5 percent would only have reduced the crime rate by about 5 percent in the early 1980s.1 These studies look only at the short-term effects of unemployment, but comparisons among communities with high and low unemployment rates suggest that the long-term effects of unemployment are also fairly modest.
Currie’s conception of a “decent” employment policy involves more than full employment, of course. Confronting Crime also proposes to “upgrade the quality of low-paying jobs.” At a minimum this means raising the minimum wage. Like cutting unemployment, this might reduce crime a little. But the evidence discussed in my review makes me doubt that the effect would be large, either relative to the increase in American crime over the past generation or relative to the difference between America and other industrial nations.
Currie also wants to improve working conditions, provide workers with more job security, make sure that their jobs offer reasonable chances of promotion, and so forth. We have no systematic evidence on the effects of such changes. Based on what we know about unemployment and income inequality, I would expect such changes to reduce crime, but only a little.
Confronting Crime also suggests “permanent public or public-private job creation in local communities, at wages sufficient to support a family breadwinner, especially in areas of clear and pressing social need like public safety, rehabilitation, child care, and family support.” In plain English this means more jobs for policemen, social workers, and daycare workers. I favor this kind of job creation, because I think we need such services more than we need additional private consumption. But jobs of this kind won’t go to people who would otherwise commit significant numbers of crimes. They will go to relatively well-educated, reliable workers who can pass examinations. This will not do much to reduce crime. Nor do extra policemen appear to deter crime.
My skepticism about the connection between Currie’s economic agenda and crime does not mean that I think economic arrangements have no effect on crime. But I think the effects are largely indirect. I see cultural values and behavioral norms as the proximate determinants of crime rates. Like Currie, I assume economic arrangements affect values and norms. But I believe that these effects are both slower and less certain than Currie suggests.
We know, for example, that crime declined among urban whites during the second half of the nineteenth century. This happened not just in America but in Britain and Australia as well. This decline appears to have been a byproduct of industrialization, which brought millions of men into relatively large hierarchical organizations for the first time. These organizations demanded a lot of self-control and regimentation. Their existence also led to the creation of a public school system that regimented the lives of the young in new ways and tried to make them internalize “Victorian” habits and values. Apparently as a result, these habits and values spread throughout the urban labor force. Declining crime rates appear to have been one byproduct of this cultural transformation.
Among America’s urban blacks, in contrast, crime rates increased steadily during the second half of the nineteenth century. Roger Lane argues that black and white crime rates diverged because urban blacks were excluded from the industrial and white-collar jobs that were transforming white immigrants into law-abiding citizens. Black culture was linked to a different sort of labor market and evolved in a fundamentally different way.2 This argument is hard to test rigorously. Nonetheless, it links economic institutions to cultural values in a way that appeals to both my economic liberalism and my cultural conservatism. Both economic conservatives and cultural liberals will probably find it less appealing.
My cultural conservatism also makes me cautious about Currie’s analysis of the cultural determinants of crime. Currie invokes Clayton Hartjen’s vision of communal solidarity to explain India’s low crime rate, for example. But Hartjen advances this argument to explain the low crime rate in rural India. The puzzle that mainly concerned me in my review was why Bombay had such a low crime rate. India’s rural poor may feel themselves part of a “larger, cohesive community,” as Hartjen claims, but the law-abiding behavior of the hungry and homeless in Bombay cannot be readily explained in these terms. India’s urban poor appear to have been almost totally abandoned by their fellows. They are, moreover, almost completely exposed to the “play of market forces” that Currie sees as the prime culprit in American crime. Nonetheless, India’s urban poor commit far fewer crimes than their American counterparts. (It should be added that America’s urban poor were also far more exposed to the play of market forces in the late nineteenth century than they are today. Yet they commit far more crimes today than they did then.)
Currie concludes his letter with a list of proposals that he believes would help reduce crime. I favor most of these proposals, and I believe that some of them would help reduce crime. But I do not think any dispassionate observer who reviewed the available evidence would bet heavily on their having a large effect on crime, at least in the near term (say ten years). Currie’s proposals might have more effect in the long run, just as industrialization did, because they might transform our ideas about morality, justice, order, and rules. But since neither Currie nor I nor anyone else can describe the links between economic institutions, culture, and crime with any precision, there is no way to be sure of this.
This does not mean I think we should reject Currie’s proposals. Few good proposals are free of risk, and even fewer come with impeccable academic credentials. There was no good evidence in 1776 that England’s North American colonies would be better off under a republican government than under King George. Nonetheless, the experiment worked. Had the colonists waited for a definitive quantitative estimate of the likely benefits of decolonization, America would still be governed from London.
Social science is restricted to the study of the past. Worse yet, rigorous social science is restricted to studying those aspects of the past that have been recorded in fairly consistent categories, like births, marriages, votes, tax rates, and murders. It can never hope to achieve much rigor in studying the subjective meanings that people attach to events, even though these meanings are critical to understanding why things happened as they did and what is likely to happen in the future. As a result, those who want to remake society can never hope to get reliable advice from social scientists. Incremental reformers may get some hints, but even they must look elsewhere most of the time.
Unlike Currie’s letter, Herrnstein’s deals largely with technical issues. First, in response to my passing remark that Crime and Human Nature said nothing about the effects of poverty or social class, Herrnstein rightly notes that the book includes a chapter on labor markets. But while this chapter includes a brief discussion of why affluence might increase crime, it does not discuss the likely effects of poverty. Nor does it discuss social class. It contains three pages on “economy and community” that echo sociologists’ ideas about the ways in which class differentiation and class conflict can shape people’s feelings about crime, but it says nothing about the extensive ethnographic literature on how those at the bottom of the social pyramid really think about such matters.
Second, Herrnstein claims that David Rowe and others have gathered evidence refuting my suggestion that identical twins have more influence on one another’s criminal behavior than fraternal twins do. In fact, Rowe’s research does not deal with criminal behavior. Rowe asked high school students to report on what he describes as “antisocial behavior” (ASB). Even “antisocial” seems too strong for what his “ASB scale” really measures. The most common forms of antisocial behavior on his scale are having “gone onto someone’s property when you knew he didn’t want you there,” having “damaged or messed up something not belonging to you,” and having “got into a fist fight to defend yourself.” Behavior of this kind has nothing to do with crime in the conventional sense of the term. Perhaps it is why Rowe’s middle-class respondents score as high on his scale as his lower-class respondents.
Nonetheless, Rowe’s findings are consistent with my argument that identical twins exert more influence on one another than fraternal twins do. First, Rowe found that most twins engage in antisocial acts together. This surely means that twins influence one another’s anti-social behavior. Second, he found that identical twins spend more time together than fraternal twins. This almost certainly means that identical twins influence one another more than fraternal twins do. Third, he found that twins who spent a lot of time together tended to have more similar scores on his index of antisocial behavior than twins who spent relatively little time together.3 Rowe’s findings do not, in my judgment, prove that comparing identical to fraternal twins yields an upwardly biased estimate of genes’ effects on criminal behavior. But they are at least as consistent with that view as with Herrnstein’s.
While I agree with Herrnstein’s claim that the “main” reason identical twins’ criminal records are more alike than fraternal twins’ criminal records is probably that identical twins have more genes in common, that is not sufficient to tell us how much of the variation in criminal behavior is traceable to genetic factors. For twin comparisons to yield an unbiased answer to this question, genetic resemblance must be the only reason identical twins are more alike than fraternal twins. Neither Rowe’s work nor anything else I have seen in print convinces me that this is the case.
Herrnstein also rejects my alleged conclusion that genes explain only 2 percent of the variation in criminal behavior. But that was not what I said. Rather, I said that “natural parents’ criminal records account for only about 2 percent of the variation in their sons’ criminal records.” As Herrnstein surely knows, the phenotypic association between natural parents and their children is not a heritability estimate.
Third, Herrnstein faults me for ignoring the possibility that genetic differences help explain urban-rural differences in crime. To support this argument he notes that blacks are now overrepresented in urban areas. Since blacks are genetically different from whites, and since blacks also commit more crimes than whites, genes contribute to urban-rural differences.
This argument is perfect illustration of the “either-or” thinking that my review deplored. There is no question that blacks commit more crimes than whites. In a sense, therefore, the genes that cause dark skin can be said to cause crime as well. But this “genetic” explanation of crime does not rule out an “environmental” explanation. Most liberals assume that having a black skin leads to crime because it alters the environment to which individuals are exposed. If that is so, Herrnstein’s “genetic” explanation of urban crime and my “environmental” explanation are perfectly compatible.
I take it, however, that Herrnstein really believes that people with a genetic predisposition to commit crimes are overrepresented in urban areas. This alleged genetic difference is presumably a product of selective migration: law-abiding folks stay on the farm, while troublemakers move to the city.
This is a new twist on an old argument. Fifty years ago, when it became apparent that urban schoolchildren had higher IQ scores than rural schoolchildren, hereditarians claimed that the brightest farm children were moving to the cities, making the urban gene pool superior to the rural one. Since high IQ youngsters commit relatively few crimes, this kind of selective migration should lead to less crime in urban than rural areas. It obviously didn’t.
Now Herrnstein seems to be advancing the opposite argument: the scum of rural society moves to the city, while the decent citizens stay behind. Does this mean that he rejects the older theory that the cleverest farm youngsters moved to the city? Or does he believe that the city attracts smart troublemakers while rural areas retain the dull but law-abiding? This whole line of argument seems to me fanciful. No one doubts that selective migration occurs, but I know no evidence that it creates important differences between urban and rural gene pools.
Nor can I see the relevance of Robert Gordon’s argument that city size has no effect on crime rates. Gordon has never denied that rural areas (i.e., communities of 2,500 or less) have less crime than urban ones.
Fourth, Herrnstein takes the position that it does not matter how genes affect IQ scores. For statistical purposes this is true. But if you want to reduce genes’ effects on IQ scores or crime, as most people do, understanding the mechanisms that link them is critical. Are we then to conclude that he is not interested in changing this?
Herrnstein’s conclusion takes my breath away. Having said that he and I agree that “society can do better in diverting academically disadvantaged people from crime,” he adds the astonishing observation that “the more interesting issue is whether the liberal outlook is a help or a hindrance in this regard.” To me, this seems one of the least interesting questions around.
The interesting question is what works, not what political label best describes it. Programs that work are sometimes best described as “liberal,” sometimes as “conservative,” sometimes as “radical,” sometimes as “reactionary,” sometimes as “none of the above,” and often as “several of the above.” Why Herrnstein needs to believe that other people’s “liberal” prejudices are all based on wishful thinking, while his “conservative” prejudices are all based on scientific truth, is beyond me. The world is not that tidy.
Richard Freeman, "Crime and Unemployment," in James Q. Wilson, ed., Crime and Public Policy (Institute for Contemporary Studies, 1983), pp. 89–106.↩
Roger Lane, The Roots of Violence in Black Philadelphia: 1860–1900 (Harvard University Press, 1985).↩
Rowe's original paper ("Biometrical Genetic Models of Self-Reported Delinquent Behavior: A Twin Study," Behavioral Genetics, Vol. 13, 1983, pp. 473–489) may have misled Herrnstein on this point. Rowe reports correlating the amount of time twins spent together with their similarity on his index of antisocial behavior in four subsamples: identical male twins, identical female twins, fraternal male twins, and fraternal female twins. Only one of these correlations was "statistically significant." But this does not mean the three remaining correlations were zero, or even that they were very small. It just means that the three remaining correlations are based on so few cases that they could have arisen as a result of sampling error, even if the true correlations for all such twins were zero. Yet as Herrnstein surely knows, the fact that a relationship could be traceable to sampling error does not necessarily mean that it is traceable to this kind of error. It just means that the sample in question is too small to justify any strong conclusion either way.
In this case, however, the problem of small sample size can be cured by pooling the data for all four kinds of twins. Professor Rowe was kind enough to make the original correlations available to me, and when one takes them all together rather than one at a time they are too large to be attributable to chance. Professor Rowe has also calculated the correlation in question for his full sample, and it is much too large to attribute to chance. Here as elsewhere, the mischief done by social scientists' use of the term "statistically insignificant" is incalculable. Even Harvard professors seem unable to remember what it really means.↩
Richard Freeman, “Crime and Unemployment,” in James Q. Wilson, ed., Crime and Public Policy (Institute for Contemporary Studies, 1983), pp. 89–106.↩
Roger Lane, The Roots of Violence in Black Philadelphia: 1860–1900 (Harvard University Press, 1985).↩
Rowe’s original paper (“Biometrical Genetic Models of Self-Reported Delinquent Behavior: A Twin Study,” Behavioral Genetics, Vol. 13, 1983, pp. 473–489) may have misled Herrnstein on this point. Rowe reports correlating the amount of time twins spent together with their similarity on his index of antisocial behavior in four subsamples: identical male twins, identical female twins, fraternal male twins, and fraternal female twins. Only one of these correlations was “statistically significant.” But this does not mean the three remaining correlations were zero, or even that they were very small. It just means that the three remaining correlations are based on so few cases that they could have arisen as a result of sampling error, even if the true correlations for all such twins were zero. Yet as Herrnstein surely knows, the fact that a relationship could be traceable to sampling error does not necessarily mean that it is traceable to this kind of error. It just means that the sample in question is too small to justify any strong conclusion either way.
In this case, however, the problem of small sample size can be cured by pooling the data for all four kinds of twins. Professor Rowe was kind enough to make the original correlations available to me, and when one takes them all together rather than one at a time they are too large to be attributable to chance. Professor Rowe has also calculated the correlation in question for his full sample, and it is much too large to attribute to chance. Here as elsewhere, the mischief done by social scientists’ use of the term “statistically insignificant” is incalculable. Even Harvard professors seem unable to remember what it really means.↩