The Editor: The Increase of Ignorance 



179 



0, wc know that there is no constant 

 relation, no resemblance, between the 

 things we arc correlating. If it is 

 unity, we know the correlation is 

 absolute ; that if one factor changes ever 

 so little, the other must change in a 

 corresponding amount. Generally, a 

 correlation above .5 is considered dis- 

 tinctly high. 



It must be said that the picture which 

 we will get of Pittsburgh is only an 

 impressionistic one. The data available 

 are coarse, and as we cannot treat a 

 smaller unit than a whole ward, we 

 cannot use any great refinement of 

 method. It will be all the easier, how- 

 ever, to see the general outlines of the 

 ])icture, and I think that they will be 

 accurate. 



With this preliminary warning we 

 examine the connection between birth- 

 rate and infant mortality, our measure 

 of the latter being made by dividing the 

 number of deaths under one year in a 

 ward by the number of births in the 

 same ward for the same year (1912), 

 while the birth-rate in the various wards 

 was computed by dividing the total 

 number of births per ward in 1912 by 

 the total niimber of females in that ward 

 in the same year. This is not an 

 accurate measure, because the proportion 

 of unmarried women may be larger in 

 some wards than in others, but it is 

 the best available. Arranging the 

 twenty-seven wards in the order of 

 their birth-rate, from the highest to the 

 lowest, and then ranking them for infant 

 mortality (1912) in the same way, we 

 find the correlation between the ranks ^ 

 in respect of these two facts is only .307. 

 This is much lower than one would have 

 anticipated, and in comparison with 

 its probable error^ is of little significance. 

 We must conclude that in 1912 there 

 was little or no connection between 

 birth-rate and infant mortalitv in Pitts- 



burgh, contrary to what one would 

 expect who has heard the birth control 

 propagandists insisting that a high 

 birth-rate means a high rate of infant 

 mortality. Here we find that we cannot 

 judge, by knowing how a ward stands 

 in respect to birth-rate, whether or 

 not it will have higher, or lower, rank 

 in respect to infant mortality. The 

 two facts have little connection — it is 

 evident that each is influenced by many 

 causes, which may be stronger in one 

 case than in the other. 



INCREASE OF POPUL.\TION 



As the first year of life is b\- far the 

 most fatal one, we can reach a measure, 

 rough but sufficient for our purpose, 

 of the rate of first-year increase of 

 various parts of the population, by 

 deducting the number of deaths (1912) 

 under one year of age from the number 

 of births in the same year. We will 

 make this comparable by dividing the 

 figure so obtained by the population of 

 the ward, and for convenience we will 

 call the resulting rate the net increase, 

 although it is in fact a percentage. 



The net increase, we find, depends 

 almost absolutely on the birth-rate, the 

 coefficient being .978; but it has little 

 relation to the infant mortality, with 

 which it is correlated to the extent of 

 only .245 with a large probable error. 

 That is, the birth-rate largely deter- 

 mines which ward shall show the 

 greatest increase of population, the 

 infant mortality exercising little effect 

 in this case. The point is important, 

 because we are sometimes told that 

 even though the superior classes of a 

 community have a low birth-rate, 

 yet the infant mortality among them 

 is so much lower than that of the slums, 

 that in the end they manage to hold their 

 own. We shall see still more striking 

 c\4dencc, a Httle later, that this was 



' The correlation coefficient was corrected by means of Pearson's tal)le (Draper's Comj)any 

 Research Memoirs, Biometric Series IV, 1907, p. 18). As indicated above, I am under no illusioris 

 about the exactness of the method of ranks, but it is the only one applicable to Miss Scorer's 

 data. The probable error is relatively large because of the small number of cases (twenty-seven). 

 I must add that I am not responsible for the accuracy of the original data, but only for the correla- 

 tions. I believe the data to be accurate for the present purpose, but I have not placed much weight 

 on the absolute values reached. The relative values are the important ones frcm a eugenic 

 standpoint and as far as I can see they are accurate enough at least to show the general trend. 

 I am indebted to Dr. Sewall Wright for counsel and criticism in this study. 



- See complete table of coefficients and probable errors at the end of this paper. 



