206 BIOLOGY OF DEATH 



Now this is a statistically insignificant net correlation, 

 being less even than 3 times its probable error. It means 

 that, when the birth rate and artificial feeding rate are 

 held constant, differences in the infant death rate are 

 not sufficiently influenced or determined by differences 

 in the poverty rate to lead to a coefficient of correlation 

 significantly different from zero, so far as Bavarian 

 populations are indicative. 



This result is further confirmed by an analysis which 

 Greenwood and Brown made of Heron's London mate- 

 rial, showing that in that case 



r 34 .i==.19.13 



This coefficient means that the differences in infant 

 mortality rate in the different districts of London, when 

 the birth rate is made constant, are not associated with 

 differences in poverty between the same districts to an 

 extent sufficient to lead to a correlation coefficient sensi- 

 bly different from zero. 



Finally, Stevenson has, since the appearance of 

 Hersch's paper, studied the same problems on the basis 

 of the London data, for the sake of comparison with 

 the results from Paris. He takes as the index of eco- 

 nomic status the number of domestic servants (of both 

 sexes) per 100 of population, and has examined the death 

 rates from all causes, infant mortality, and tuberculosis 

 for the identical years that Hersch used. The results 

 are set forth in Table 24a. 



Commenting on the facts regarding general mortality 

 from all causes in London, Stevenson says: 



"These bear an altogether different aspect from the Parisian figures. 

 Whereas the latter increase so regularly with poverty that the highest 

 rate for any district in one group never exceeds the lowest for any district 

 in the next poorer group, in London the gradation, even for the groups 

 themselves, is irregular, the lowest death-rate not being returned for the 



