as well known, extremely elastic, among those who are called 'moderate' there are many different degrees. In consequence, it does not seem to me permissible to place all the so-called moderate in one common group—they differ greatly from each other—and to argue from the experience of such a heterogeneous group that every degree of moderation, for instance such as a strictly temperate use of alcohol must be injurious to health and shorten the duration of life. The statistics from the life insurance companies in question do not seem to me to furnish any evidence upon which such a conclusion can be based.

"The data from the United Kingdom Temperance and General Provident Institution have been especially thoroughly worked over and published in detail. They therefore permit a closer scrutiny from different points of view, and there has been no lack of criticism from the standpoint of statistics and actuarial technique. I wish here to present some of the objections which have been made from authoritative sources.

"Westergaard objects: 'The mortality in the division for absolutists is thus notably lower than in the general division. Meanwhile, one cannot without further question regard the figures as expressing a sub-mortality on the part of the absolutists. The absolutist group increases rapidly. For the years 1866-1870 one expected in it only somewhat more than half as many cases of death as in the general group; for the years 1896-8 the numbers were almost equal. As is well known, the mortality during the first years after taking out insurance is comparatively low. At the same time this source of error can scarcely explain the entire difference. According to the calculations mentioned above, the level of mortality during the first five years of insurance is about one-fifth lower than later on. If we now deduct from the 7,656 cases of death which were to be expected in the absolutist group about 20 per cent, there remains 6,125, a number which, however, notably exceeds the one observed; namely, 5,383, and places in the shade the experience from the general group, in which for the same period the cases calculated amounted to 10,455, and those really occurring to 1,065. Besides, one can hardly assume that all the members of the general division have been insured more than five years and all the members in the absolutist division less than five years.'

"The investigation by the United Kingdom Temperance and General Provident Institution has been subjected to a very thorough

criticism by Andrae. He remarks first that among the insured women the difference in favor of absolutists-if one notes that the sub-mortality at middle age was partly counter-balanced by the submortality at other age periods-is so insignificant that a favorable or special standing within life insurance on part of the absolutely abstinent woman as compared with the non-absolutists is not justified on the basis of the experience at hand. The relative submortality among the absolutist women was only 3 per cent. He makes other tests of how far one may assume that both divisions contained equally good risks from the beginning. He points to two circumstances which seem to him especially important in judging the statistics under consideration from the actuarial point of view. He compares both among absolutists and non-absolutists the mortality of men who were insured for life with that of men who took out insurance on the payment of a premium at a certain age. Among non-absolutists the relation between the two different groups of insurers was the usual, as known from other experiences: the mortality among the latter was about 26.5 per cent more favorable than among the former (those insured for life). Among the absolutists it was wholly different. Here the surprising result appears that the sub-mortality among those who were insured for a shorter time as compared with the other group was on the average only 6 per cent; consequently, about 20 per cent less than among nonabsolutists and thus in line with normal experience. The submortality in question is meanwhile regarded as conditioned only by the self-selection of the insurers. Those who themselves expect

a long life on account of a close knowledge of their health conditions. stand to gain through insurance by the payment of the premium at a certain age and the very best risks as a rule seek this kind of insurance and in consequence of this the normal sub-mortality is found. in this group. The disproportionate low sub-mortality among the absolutists insured for a shorter period as against those insured for life is consequently to be explained by the fact that the latter exercise a much more effective self-selection; that is to say, that among them, thanks to certain interests, an unusually high percentage of the best risks is to be found which otherwise only insurance for a shorter time brings with it. It also lies close to hand that these interests are in the first instance to be sought in the high participation of gain offered absolutists and also possibly in the effective recommendation of the company to absolutists.

"The analysis which Andrae gives in regard to the greater duration of life among the absolutists insured in the company under consideration he believes to be confirmed by the surprising relation between the mortality of men and women among the absolutists. Among these, the women of all ages show a greater mortality than the men and their super-mortality increases prior to the 54th year of life very greatly. In the general division, on the other hand, the relation between the mortality of men and women shows reasonable correspondence, as the mortality among the women is lower than among the men. For each 100 cases of death calculated according to experience among men, there occurred among the abstaining women 118.5 and among the non-abstaining 94.6. In the age period 20 to 34 years, the mortality among abstaining women reached 200.00; in the age period 35 to 44 years 173.8; and in age period 45 to 54 years 130.3; while among the non-abstaining women the respective number in the corresponding age classes were 119.2, 111.3 and 90.4.

"Andrae sums up the results of his investigations as follows: 'If, as we have done, the sub-normality of the absolutists as compared with that of the non-absolutists is ascribed at least in large measure to the influence of self-selection, and if one adds to this that such self-selection is less prominent among women than among men, then the abnormal relation between the mortality of men on the one side, and of women on the other, is explained: The percentage of good and very good risks is, for the reasons given, greater among men than among women. But if it is insisted that the reason for the greater duration of life on the part of the absolutists. is exclusively or predominantly their abstinence from alcohol or qualities prolonging the duration of life which in especial degree result from this abstinence, then the above mentioned unusual relation becomes incomprehensible. How far a small part of the sub-normality among the absolutists possibly may be explained through one of the last-named causes, cannot be decided on the basis of the experience material at hand and therefore remains statistically unproved. In order to arrive at a decisive result on this point, it would be necessary, according to my opinion, to begin with life insurance companies which deal with absolutists and non-absolutists in exactly the same manner, but which nevertheless make it possible in a reliable way to distinguish the two groups by aid of the questions put to them at the time insurance was taken out. Such

material, however, cannot be obtained in the near future and the question therefore remains open for the present.'

"I am not capable of judging Andrae's criticism from the point of view of insurance technique. It seems to me, however, that in the present condition of the question one must reckon with the possibility that the sub-normality among absolutists, as compared with the others in the above named life insurance statistics, at least in part may be due to the dissimilar relation of both groups to the use of spirituous drinks. But for the reasons mentioned, I cannot find that the investigations show absolutists in general to be superior to the truly moderate in the matter of duration of life.

"Some non-European life insurance companies seem to show an experience different from the English, and to have reached results pointing in the opposite direction. Holitscher, from whom I have obtained the data, says: 'In compiling the material from a company in New Zealand, the absolutists were found to be somewhat worse placed; a similar result was gained by a company in Canada and one in the Cape Colony. If, in accordance with the experience of the English companies, it were a hard and fast condition that absolutists, solely because of their abstinence from spirituous drinks, might expect a duration of life greater by 25 per cent., it is striking that this does not appear from the last mentioned statistics, but quite the contrary.'

"Meanwhile, the statistics just referred to have not been so fully published that they permit a closer critical scrutiny, and it is therefore not possible to say anything final about the weight of the evidence. It has been objected that they concern too small numbers and too short a time to be taken into consideration. This would also seem to be true of the statistics from the Cape Colony which relate only to 865 insured absolutists and 2,792 insured non-absolutists.

"Also from Sweden there is at hand a comparatively extensive investigation by Ekholm. In the Swedish Life Insurance Company the question has been asked applicants since 1897, 'Are you an absolutist or not?' And thus it has been made possible to investigate separately the mortality among absolutists and non-absolutists.

"The statistics prepared by Ekholm cover the years 1897-1906 and relate to about 35,000 persons, of whom somewhat more than half were absolutists. The whole number of deaths was only

496, and 252 of them occurred among absolutists and 244 among the non-absolutists. The absolutists showed on an average 6 per cent less mortality than the rest. If the material is divided into four groups according to age and time of insurance, it appears that the difference in favor of the absolutists rises according to age and the length of the time of insurance. In the group containing persons under 44 years of age, the relative sub-normality of the absolutists was only 2 per cent., but among the older persons 26 per cent.

"Ekholm adds the following remarks to his analysis of results : 'One may argue about the hygienic significance of this result. To me the assumption seems entirely natural that the difference in favor of the absolutists is explained by their abstinence from alcoholic beverages. It seems to me, therefore, that the results indicate absolutism as a thoroughly healthful mode of life which has a certain significance when these absolutists are compared with a selection of healthy and moderate non-absolutists. It is clear that at least a part of the latter have consumed more alcohol than is good for them. To many the results here related may seem to be sufficient proof that even a moderate use of alcoholic beverages is injurious to health. But a closer scrutiny of the question shows that it is not quite as simple as this, for in the end everything depends on what is understood by moderate use.'

"The investigations hitherto at hand are to this extent incomplete, that the life insurance statistics deal solely with two categories, absolutists and non-absolutists. But in order to solve the question completely it is also necessary to know how great the consumption of alcohol on the part of the individuals among non-absolutists has been. Ekholm relates that Swedish life insurance companies have already begun to collect statistics in order to determine this question more thoroughly, as those who apply for insurance are requested to answer the question, how much spirits they consume daily. On the basis of this, the insured are divided into different groups according to the magnitude of the daily consumption of alcohol. By this method of procedure it is expected that it will be possible to determine how large the optimum of alcohol consumption is, and to compare this with zero. Ekholm believes that the solution of the question in this way will require a long time, but that we must be patient, for there is no short cut which surely brings one to the goal."