STATISTICS: R. PEARL 
85 
the case of the whites, the statistics are undoubtedly more accurate, and one 
would certainly not be justfied in going so far as does Clark^ in denying the 
existence of centenarians except in the very rarest cases. 
From the unmodified data in table 1, the constants shown in table 2 have 
been deduced. 
Before commenting on the data of table 2, it is desirable to present the re- 
sults of another method of approach to the matter. That method is to take 
the number of deaths for each year over 100 from the column of a properly 
graduated mortality table, and calculate from such figures the means and 
standard deviations. The mortality tables which have been used for this 
TABLE 2 
Constants for Deaths of Centenarians in the Registration Area in 1916 
GROUP 
MEAN 
STANDARD DEVIATION 
years 
years 
103 
19 ^ 
0 
17 
3.02 ± 0.12 
White females 
102 
42 ± 
0 
12 
2.49 ± 0.09 
White, both sexes together 
102 
75 ± 
0 
10 
2.76 ± 0.07 
Colored, both sexes together 
104 
45 ± 
0 
18 
4.89 ± 0.13 
TABLE 3 
Deaths of Centenarians from United States Life Tables 
dx OF 100,000 BORN ALIVE 
X TO X -\- 1 
White males 
White females 
Negro males 
Negro females 
100-101 
13 
19 
13 
34 
101-102 
8 
11 
10 
26 
102-103 
5 
6 
7 
18 
103-104 
3 
4 
4 
13 
104-105 
1 
2 
3 
9 
105-106 
1 
1 
1 
5 
106-107 
* 0 
1 
1 
3 
107-108 
0 
0 
1 
2 
108-109 
0 
0 
0 
1 
109-110 
0 
0 
0 
' 1 
purpose are Glover's^ 1910 American mortality tables. The distributions of 
death which these tables give for white and colored are exhibited in table 3. 
As would be expected the distributions in table 3 are much smoother than 
those of table 1. Furthermore, their range is less extended at the upper end. 
The constants for the distributions in table 3 are exhibited in table 4. 
From these tables the following points are to be noted: 
1. As was to be expected the mean ages at death are sensibly lower, and 
the standard deviations smaller, when the graduated data of the life tables 
are used instead of the raw statistics. 
