SERIES CORRELATED WITH EXPENSES 22°
2. TOTAL EXPENSE

Let us begin our analysis by determining the percentage
amounts and the directions by which the yearly district ratios
in each of a number of series deviate from their seven-year average
 levels, the classification of the districts being made according
to the directions and percentage amounts by which their ratios of
total expense deviate from their own seven-year average levels.
This is the conventional form of analysis found in earlier chapters.
 That is, what is now done, in the case of total expense, is
(1) to classify the percentage amounts by which the ratios in
each district each year deviate from the respective district averages
 for the years 1919-1925, (2) to total the frequencies, and
(3) to compute for each dispersion group the average amount of
deviation or difference. Such a summary is found in the stub
classes (lines) of Table 138. For the districts having the respective
 classified deviations of ratios of total expense, the net
percentage amounts of deviation in a number of series from their
district averages for the period 1919-1925 are then calculated.
Such percentage amounts by sign are shown in the caption classes
(columns) of Table 138.
It will be observed that the basis of selecting the districts in
order to determine the net amount of deviation is the nature and
percentage amount of their deviation in the matter of total expense.
 In earlier chapters where an analogous method was followed,
 the basis of selecting the districts, in order to determine
the type and net percentage amounts of dispersion of ratios of
total expense, for instance, was the classified percentage amounts
of dispersion with respect to loans and discounts, deposits, and
50 on. Our interest then was in the types and net amounts of
deviation in total expense associated with variable amounts of
deviation in other data; our interest now is in the types and net
amounts of deviation in other data associated with variable
amounts of deviation in total expense, the amounts, in all cases,
being measured on a percentage basis.
But what are the findings, given in Table 138, from such an
analysis? Briefly, they are as follows: (1) In years and districts
in which ratios of total expense to earning assets are high or low
—that is, above or below the district levels for the period 1919-