Actuarial Aspects of Human Lifespans 15 



GoMPERTZ, B. (1825). Phil. Trans., 115, 513. 



Makeham, W. M. (1867). J. Inst. Actu., 13, 325. 



Ogborn, M. E. (1953). J. Inst. Actu., 79, 170. 



Pearson, K. Karl Pearson's Early Statistical Papers. Cambridge 



University Press (1948). 

 Perks, W. (1932). J. Inst. Actu., 63, 12. 

 Phillips, E. W. (1935). J. Inst. Actu., 66, 17. 

 Phillips, E. W. (1954). J. Inst. Actu., 80, 289. 

 Rich, C. O. (1940). J. Inst. Actu., 70, 314. 

 Thiele, T. N. (1871). J. Inst. Actu., 16, 313. 



DISCUSSION 



Danielli: What is the basic mechanism underlying this shift of the 

 death curve to the right — that is to say, why was there an improve- 

 ment in mortality? 



Benjamin: I cannot hazard a guess. All I could deal with was the 

 observed deaths, and it is a fact that the peak has shifted to the 

 right. This might be because, as Phillips has suggested, successive 

 generations are tending to attain a sort of ideal curve of deaths — 

 that is, a very sharp peak, even further over to the right than has yet 

 been observed. Or it may simply be that more people are in fact 

 surviving to a constant natural lifespan, which means that the curve, 

 instead of being pulled to the left by what Clarke called anticipated 

 deaths, is allowed to grow up more on the right-hand side. 



Rockstein: Does the initial portion of your curve correspond closely 

 with that for data in other countries ? There seems to be a rather 

 high mortality rate for males in Britain during the early years. 



Benjamin: I have not yet made much comparison with the life- 

 tables of other countries. I should not have thought it was unusual 

 for the Western countries, at least for Western Europe. 



Rotblat: The peak for the anticipated deaths seems to become 

 sharper in the 1950-52 curve. Why should this happen ? I would 

 have expected this peak to become flatter and spread over the whole 

 span of life, rather than sharper. If you used a skewed distribution 

 rather than a symmetrical one perhaps you would not get this 

 sharp peak. 



Benjamin: I am very grateful that you have made that point 

 because this sharpening of the peak in the anticipated deaths is in 

 fact phoney; it is simply an accident of the simplified type of 

 analysis and has no meaning so far as I can see. One can do as 

 Clarke did: make an arbitrary assumption about the proportion of 

 deaths which are senescent and so avoid this assumption of a sym- 

 metrical distribution. But my difficulty is that I do not know where 



