SERIES CORRELATED WITH NET EARNINGS 257
It is apparent at once, from an inspection of this table, that
ratios of total expense to earning assets are not independent of
—that is, remain the same with—variable ratios of loans and
discounts as measured. Indeed, the types of relations between
ratios of loans and discounts and of total expense are the same
as those between ratios of loans and discounts and of gross earnings.
 Accordingly, with both gross earnings and total expense
positively correlated with loans and discounts, and with total expense
 indifferently correlated with deposits, the relation between
gross earnings and total expense is in doubt when approached
in this indirect manner. But if high gross earnings go with high
total expense and low gross earnings with low total expense, as
is the case when the ratios themselves are directly correlated
with respect to the district and yearly variations and percentage
changes from year to year, the manner in which net earnings
are related to both of them, and to the series correlated with
gross earnings and with total expense, is not self-evident. Further
study is necessary to establish these relations.

TABLE 151

NATURE OF CORRELATION IN PERCENTAGE DEVIATIONS AND YEAR-TO-YEAR
 CHANGES IN DIFFERENT SERIES PAIRED

SERIES CORRELATED

Independent Variables

Dependent Variables

Gross Earnings to Earning
Assets. ..................
Total Expense to Earning
Assets

Total Expense to Earning
Assets... ...............
Gross Earnings to Earning
Accets

NATURE oF CORRELATION

Differences

from
District
Averages

Changes
‘rom
Vear
to
Year

Differences

from
Country's
Yearly
Averages

Positive . Positive
Pocitive + Positive

Positive

Positive

But first, attention should be directed to the type of relations
between variable gross earnings and variable total expense. The
nature of the correlation is shown in Table 151. If, as shown in
this table, high or low ratios of gross earnings, respectively, accompany
 high or low ratios of total expense, how are ratios of net
earnings affected by these sympathetic changes? Conflicting evidence
 bearing on this question is found in Table 152, in which is
summarized the nature of the relations between variable gross