found that the husband had enlisted for the civil war, and that the wife had taken the companionship of another man within two weeks of his departure. At the end of his term, the husband died within ten days of mustering-out, the wife never having in the mean time returned to her duty or even forsaken her lawless ways. Technically, not having married the second man, this woman was the first one's widow, and therefore within the privileged circle established by statute. The Commissioner indorsed upon the back of her file of papers: "Application refused on grounds of public policy." His decision was overruled on appeal.

This fact and others which might be cited go to show that the Pension Bureau, as conducted under the present administration and some of its predecessors, has been no willing party to the degeneration of the pension list from a pure roll of honor to a mixed catalogue. The blame must rest on Congress and on the clamorous army of claim agents and solicitors who infest the lobbies of that body. To no small extent, moreover, the boldness of the claim agents in making war upon any Commissioner who tries to do his duty, and the inactivity of Congress when reforms are most urgently needed, are traceable to the apparent indifference of good citizens at large. If these were as aggressive in supporting an honest officer as men of the other sort are in assailing him, and if they would insist upon decent pension legislation with half the energy they put into a demand for a tariff schedule or a currency measure, the honest pensioner would soon have cause for pride in his certificate, wholesale perjury would go to a discount, the taxpayers would take fresh heart of hope, and the legacies of an American war now thirty-six years past would cease to be a gaping-stock for the civilized world.

FRANCIS E. Leupp.

STATISTICAL BLUNDERS.

AN epigram attributed to a well-known statistician runs: "Figures won't lie, but fools and liars will figure." This saying well sums up the misuse of statistics. There are both persons who intentionally misuse and misrepresent statistics for their own ends, and those also who misuse them through mere ignorance or carelessness. Statistics are like a sharp knife, a very efficient tool in skilful, honest hands, but a dangerous weapon in the hands of the unskilful or vicious. We do not condemn the knife because, whether through the ignorance or the design of the man who holds it, it creates mischief. Of the two classes, the ignorant and the vicious, the former are the more dangerous, mainly because they are the more numerous. The great and constantly increasing interest in statistics and in the knowledge derived from them has resulted in drawing into this field an army of writers many of whom are unskilled in the use of their tools. This fact has served to discredit statistics and statistical methods of research. We often find in magazines, and even in more permanent forms of literature, the most startling conclusions drawn from an array of figures, the value and meaning of which are entirely misconceived.

Some years ago, a fever of apprehension seized many people through their fear of negro domination in this country. It had been discovered that, on the face of the census returns, negroes had increased more rapidly than whites between 1870 and 1880. The fact is that long before the publication of the figures of the race in 1880 it had been shown conclusively that the census of 1870 was deficient, and that this deficiency was mainly in the enumeration of negroes in the South. Disregarding this discovery, although it had been widely circulated, and ignoring the previous history of the race, which should have led to the opposite conclusion, alarmists rushed into print with the most direful predictions. Books were written on the subject, and the newspapers were filled with it. The amount of trouble and worry which this stupid, careless blunder caused our people is simply incalculable. It took a deal of hard

hammering by those who understood the facts to disabuse the public mind, and even yet one occasionally meets relics of this error.

There is a general belief in the exceptionally great age of negroes. The census age tables are in great part responsible for this, as they show several times as many centenarians among negroes as among whites. Now the fact is that the negro is short-lived. It is doubtful if one ever lived for a century; certainly very few of them have lived for 90 years. Their death rate is nearly double that of whites. The census tables give the ages of people as they are reported to the enumerators; and, as many people do not know their ages, and therefore give them incorrectly, it follows that there must be a large percentage of error. Especially is this true of negroes, and doubly so of aged negroes, who are very prone to exaggerate their ages. The whites, particularly the most highly civilized whites, are the longest-lived people on earth.

Statistics of mortality have always been a fruitful field for the blunderer, who again falls back upon the ill-used census as his source of inspiration. The mortality statistics of the census are derived from two sources: (1) The enumerators' returns, which are obtained in all parts of the country; and (2) the registration returns, obtained from such States and cities as maintain a registration of deaths. The first are everywhere incomplete; the second are fairly complete in some States and cities, but incomplete in others. Accordingly, the enumerators' returns show a death rate far below the rates of most other civilized countries. In that fact the blunderer glories. They show a death rate of negroes but slightly greater than that of whites. This alarms him, for he reasons that the negroes, with their excessive birth rate, must be greatly increasing in number. The returns show inconsistencies in the adult and infantile death rates and in the death rate from various diseases, and the conclusions drawn from them are surprising.

It is the same old story. A little study of the introduction to the statistics of mortality in the census reports would have taught that the figures are confessedly incomplete; that the omissions of deaths are greater in the South than in the North, greater among negroes than among whites, greater among children than among adults, and consequently greater in diseases of children than in those of adults; in short, that the statistics derived from enumerators' returns are wellnigh worthless for all purposes except to prove their own worthlessness.

Turning now to the States and cities where registration is maintained, we find many of them claiming unusually low death rates and arguing therefrom the superior healthfulness of the climate, the water

supply, etc.

One of the States of this Union claimed in 1880 a death rate of but 16 per 1,000 per year. The annual death rate of the United States as a whole is probably between 18 and 19 per 1,000; that of the rural districts being slightly lower, and that of the large cities slightly higher, than these figures. A death rate of 16 per 1,000 was therefore very favorable, and indicated the existence of especially healthful conditions. The State was not satisfied, however, with this excellent showing, and created a Board of Health, which did its work so thoroughly that in 1885 the death rate was reduced to 4.5 per 1,000, and in 1890 to but 4 per 1,000. It is true that an ill-natured, critical person might say that these results were achieved by the omission of threefourths of the deaths, and that the Board of Health, instead of receiving commendation in the Governor's messages, as it did repeatedly, should have been discharged for incapacity. It may be added, in support of this criticism, that the registration returns from that State were condemned as untrustworthy by the United States census of 1890.

The registration returns of the larger cities show similar phenomena. In 1899 the death rates of 136 of our largest cities were published. Of these cities, 23 claimed to have death rates of less than 10 per 1,000; in 65 of them the rate was said to be between 10 and 15 per 1,000; in 43 of them, between 15 and 20 per 1,000; and only 5 cities confessed to a rate exceeding 20 per 1,000. A death rate below 18 in a large city is surprising; if much below, it should arouse suspicion; while the publication of a death rate below 10 is an insult to the intelligence of the public. Probably not one of the cities referred to had a death rate of less than 18 per 1,000, although five-sixths of them reported rates below that figure. There are two methods of obtaining a low death rate: one is to enumerate only a part of the deaths; the other, to estimate the population at too high a figure. When these methods are used conjointly the result is remarkably effective.

Other common fallacies concern our mental, moral, and physical progress. It is believed by many who ought to know better, and is taught by alarmists, that insanity, deafness, blindness, and crime are rapidly increasing, and that we shall soon become a nation of defectives and criminals. Census statistics are quoted in support of this prediction. There is no question but that all these things were indicated on the face of the earlier census returns, and it is equally certain that as a matter of fact none of them is true. Insanity, deafness, blindness, and criminality are not on the increase in our country. It is true that the reports up to 1880 showed a much larger proportion of insane, deaf,

and blind among the population at each succeeding census. But why? Simply because we were able at each succeeding period to obtain a fuller enumeration of these classes, and thus to make a closer approach to the actual facts. Half a century ago defective persons were kept at home, and their deficiencies were scrupulously concealed from public view. In these later days, they are placed in asylums and special schools, and consequently are enumerated and classified in a much larger proportion of cases. The census of 1890 shows a smaller proportion of these defective classes than did the census of 1880, and no doubt represents the true situation.

The explanation of incorrect conclusions regarding criminals is different. The number of criminals depends in great measure upon what the law defines as crimes, and upon the degree of efficiency in the administration of justice. Taking the country as a whole, there are many more statutory crimes than a century or half a century ago. Misdemeanors which now send a man to jail were in former times winked at. On the basis of the number of commitments to jail comparisons have been made in print between the moral condition of Massachusetts and Mississippi, to the disadvantage of the former State. The writer ignored or forgot the differences in the laws of the two States, and the variations in the degree of efficiency with which those laws are enforced. It is probable that a greater number of arrests means a higher, rather than a lower, condition of morals. A recent writer argued that the Northern negroes were much more criminally disposed than were those of the South; basing his argument similarly upon the number of commitments for crime, and thus falling into the same error.

The ten-year period, which in most of our States separates two consecutive enumerations of the people, affords opportunities and temptations to overestimate the population, especially that of growing, progressive cities. The bases of these estimates are variously selected, being in some cases a census of school children, in others the number of names given in the city directory, in others the number of voters registered or the number of votes cast; the total number in each case being multiplied by a number representing the supposed relation of the particular class to the population. There is no harm in this amusement in itself; but when results obtained by such doubtful methods are dignified by the name of statistics, when death rates are based upon them, when the figures of the census even are called in question because they disagree with them, it seems desirable to characterize them properly. In 118 of our large cities the population as thus estimated exceeds the true popu