NET EARNINGS IN DISTRICT I 319 
The first stage in the process secured measures of disper- 
sion for ratios in 1924 which were distributed above or below 
their averages (those for all banks, those for banks classified 
by size of city of location, and those for banks classified by 
amounts of earning assets) by different percentage amounts. The 
second stage, for banks with classified percentage amounts of 
dispersion from their average in 1924, determined the net average 
amount of dispersion from their average in 1925. If ratios of 
net earnings are above or below their average in 1924, they are 
in general above or below their average in 1925. But some of 
those above the 1924 average are below, and some of those below 
the 1924 average are above their 1925 average. Accordingly, 
when, for each dispersion group in 1924, the average amount of 
deviation in that year is compared with the net average disper- 
sion in 19235, it is found that the dispersion is less in 1925 for 
ratios which were either above or below their average in 1924. 
The amounts of difference in dispersion, in the two years, vary 
directly with the degree of dispersion in 1924. Ratios which are 
above as well as those which are below the average in 1924 tend 
to regress or move toward the average in 1925. The plus and 
minus quantities in the different parts of the table indicate by 
signs and amounts the regression or movements of the ratios 
back to type.’ 
The foregoing methods of measuring the directions of change 
with respect to the size of the ratios, while making use of three 
separate standards of reference to determine relative size, and 
identifying as changes differences in dispersion in 1924 and 1923, 
secure substantially uniform results. If measurements of both 
directions and amounts of change between 1924 and 1925 are 
made without reference to district-, city-, and volume-group 
averages in these years, is a tendency of regression, analogous 
to that already noticed, present? Answer to the question is found 
in the signs and amounts in Table 188, the method of treating 
the net earnings ratios being the same as that already employed 
in the analysis of ratios of gross earnings.® The table shows 
that, with the exception of the banks in cities with population of 
20,000 to 40,000, it was the ratios which were relatively low 
5 For a fuller description of the meaning of the signs and amounts secured from 
such an analysis, see pages 273-275. 
See page 27%.