ON THE NATURE OF INSURANCE. 23 



as a caution against lightly admitting a change of pre- 

 miums, on the authority of any small number of facts. 

 But more particularly should they be attended to in the 

 formation of new varieties of contingency offices, the 

 chances of which have not yet stood the test of ex- 

 perience. 



But there are reasons why the premiums of an in- 

 surance office need not be so high as the very limited 

 number of data in their tables might seem to require. 

 If the fluctuations from the average, which are within 

 the most cautious definition of reasonable probability _, 

 were all to be encountered at once, or might be en- 

 countered at once, it is difficult to say what premiums 

 should be considered as too high. But this cannot be 

 the case, unless, indeed, a pestilence should single out 

 the members of an insurance office, or an earthquake 

 should, by one extraordinary event, swallow them all up 

 in the place where, by a most remarkable coincidence, 

 they were all assembled together. Such extreme cases 

 are not worth consideration ; and we may take the 

 chances of life and death as distributed over a large 

 number of years. In the meanwhile the surplus fund 

 increases at compound interest; and the problem is, not 

 whether a given number of lives will, on the whole,, 

 drop so much before the predicted time that a given 

 fund will be destroyed, but whether this can happen so 

 fast, that it will outrun the increase of the fund at 

 compound interest. If, indeed, there were compound 

 mortality to set against compound interest ; that is, if 

 the number of deaths must become larger from year to 

 year, or if the rate of mortality were increasing, the 

 fear of such a result might be entertained ; but all 

 experience is on the other side, and tends to show that 

 the value of life is increasing, instead of decreasing. 



The tables of an insurance office must be considered 

 as collections of limited data, the premiums deduced 

 from which are increased by a percentage, to meet the 

 possible fluctuations of mortality. As soon as these 

 tables are formed, and the directors have published 



