Migration by Residence Classification 



Estimates of 1940 and 1950 state and county populations were immedi- 

 ately available from the respective censuses. However, two factors combine 

 to limit the accuracy and completeness of the rural farm, rural nonfarm and 

 urban data. They include: 1. The change in census definitions with respect 

 to the rural-urban residence. 2. The estimation necessary to make the 1950 

 "old urban" definition of age classes coincide with the 1940 and 1950 regu- 

 lar age classes. 



With respect to these factors it was necessary to use the supplementary 

 figures in Table 50 of the 1950 census. Upon advice of the Population and 

 Housing Division of the Bureau of the Census, the distribution by five- 

 year age groups of the rural farm and rural nonfarm was based upon the 

 distribution of the total rural population in Table 50. 



In order to complete the estimates of migration by rural-urban resi- 

 dence, the cumulative deaths of the rural farm, rural nonfarm, and urban 

 segments had to be estimated by assuming that deaths would be proportion- 

 ate to the populations estimated above. 



Appendix II 



Assessing Relative Weight of Natural Causes and Migration 



In the course of analyzing the mobility of the population of New Hamp- 

 shire from 1949-1950, the need for assessing the relative weight of natural 

 causes and migration arose. Many of the measures of population change are 

 set up as ratios so that the dependent variable is a function of two or more 

 independent variables. Under these conditions, it appeared that use might 

 be made of partial derivatives in securing these weights. 



To measure the amount of change in New Hampshire population, an 

 expected population for 1950 had been computed based on the 1940 census 

 and state vital statistics assuming no migration. 



For any of the measures then 1950 minus 1940 is the actual change in 

 the measure; 1950E minus 1940 is the change due to natural causes; and 

 1950 minus 1950E would represent the net change in the measure due to 

 migration. This was not entirely satisfactory since it was impossible to tell 

 which component of the measure was most important in inducing the change. 



The changes were then analyzed in the following manner: 



Expected 1950a — 1940a = a 

 Expected 1950b — 1940b = b 

 1950A - 1950 Expected A = A 

 1950B - 1950 Expected B = B 



18 



