Page 82 - clra62_0019-(GIPE)

76 ECONOMIC INQUIRIES AND STUDIES

Conclusion.
The conclusion of this long review may be very
shortly stated. All the leading branches of statistics
without exception, when examined, give numerous
illustrations of the dangers of taking the figures relating
to them from dictionaries or works of reference at hap-
hazard for international comparison, as jf the figures
called by the same names in different countries meant
the same things, or the units had the same valueS. On
the contrary, from the simplest figures as to population
and area, through the more complex figures as to the
moral qualities of communities indicated by statistics
like those relating to education and crime, down to the
stilI more complex figures relating to production, trade
and wealth, the same tale is told as to the necessity for
constant watchfulness lest things that are really unlike
be put together as if they were like. The moral is what
was stated at the outset, that the figures as such may
be right enough, though there are many difficulties as
to the data themselves to be faced in statistics, but the
exact meaning of the figures called by the same name,
when place and circumstances are different, may require
a great deal of elucidation. Perhaps some may think
that the difficulties are so great as to make it hopeless
to handle most statistics in such a way as to reach any
conclusion. This is, however, by no means the case.
When care is taken true conclusions begin to appear,
and a picture is obtained of the general conditions of
communities in the mass which would otherwise be
unattainable. The negative results which are the effect
of the criticism applied to the rough and ready methods
of amateur statisticians are also valuable and important.
There are so many errors about respecting the condition
of most communities, partly derived from, and partly
nursed by, the rash use of statistics with a more or less
conscious bias towards a desired conclusion, that it
clears the air to have a demonstration of the impossibili ty
of these errors b(:ing proved to be true. When one