A. O. POWYS 
269 
children, we are obliged to call to our assistance another table in the census 
which shows the number of wives with each sized family and the ages at marriage, 
but not the duration of marriage. In the manner thus shown — having regard 
to the age at marriage — the distribution of families of 11 and more children 
has been effected, and to this extent the columns of the last three durations in 
Table XXIV. are vitiated, though the averages obtained cannot possibly be 
affected to any api^reciable extent. This table shows the distribution of fertility 
to wives married at under 20 years for various durations of marriage within the 
reproductive period. 
The average issue for under 5 years' duration of marriage is evidently too 
small and the percentage of women without issue too large, owing to the large' 
niimber of women married under 1 year to be fair criteria of the results of a 
mean duration of 2^ years' marriage. The results of this duration were obtained 
from the census and were found to be 422 women with 468 children (56 had 0, 
270 had 1, 90 had 2, and 6 had 3), or an average of I'll each with 13"3 per cent, 
without issue. As, however, the result was 1'32 in the case of deaths of married 
women, which I think is more probably the truth in view of the fact that so many 
of those marrying under 20 being undetermined in the ages 20 and 21 — 25, I 
adopted this value in the calculation of the correlation and regression straight 
line, with the following general results : 
Meati Age at Marriage I41j4l5 years. 
„ Size of Family 4-6967 children. 
1 1 ■ • (Duration of Marriage 7 9585 years. 
btandard Deviation ,. „ .. " -.oo^i i-ii 
[Size of Family 2-3841 children. 
Correlation "99790 ± 00002 
Coetiicient of Regression "29894 
Regression Straight Line (Origin at Marriage. Unit of a; = 1 year) 
2/= -3198 + -29894 ic, 
where x = duration of marriage, and y the number of offspring. 
The statistics and regression straight line are plotted on Fig. 15, on which 
is also plotted for the purpose of comparison the regression straight line as 
determined by deaths. 
The methods of tabulation of the census do not admit of tracing the marriages 
contracted between intervals of 5 years, say, between 20 and 25, but does admit 
of them being dealt with as under 25 years at marriage*. The information for 
* It might at first sight be tliought possible to have effected the desired computations for quin- 
quennial periods. For the results are ascertainable for marriages contracted as under each quinquennial 
age : as for instance, under age 20, under age 25, under age 30, &c., and the dilference between any two 
succeeding ages would appear to give the necessary results. Thus, take the duration of marriage 5 and 
under 10 years, marrying under age 20 there were 3190 such women who had 7825 children ; marrying 
under age 25 there were 18,0S6 such women who had 47,653 children; but it is not true that the difference 
between these two series, viz. 14,896 women having 39,828 children married between the ages 20 and 25, 
for many (and it is impossible to say how many) of the latter were married at ages under 20. The 
number shown as marrying under 20 are those we are certain of from the returns but are not all. The 
same applies to all the groups. 
