ssics. In such cases
as these the citing of direct evidence brings on you difficulties of a
different kind from those you face when you are establishing a single,
simple fact. Here you will usually depend on two main sources of
evidence: statistics, and the evidence of recognized authorities on the
subject.

30. Statistics. Statistics, which are collections of figures, are
notoriously treacherous. On many important subjects, such, for example,
as the practical effect of the elective system, it is impossible to get
them; and on many other subjects, such as the effects of a protective
tariff, they must be had in so enormous masses, if they are to be
trusted at all, that only profound students can handle them. Where the
facts are complicated, and interests are tangled, moreover, many sets of
figures may enter into the question, as notably in the case of a tariff;
so clearly is this difficulty now recognized that Congress has
authorized a tariff board made up of distinguished students of economics
and men of long experience in dealing with tariff matters to collect and
study the facts and make recommendations based on them. Similarly, with
the investigation into the liquor question made fifteen years ago by the
Committee of Fifty: the whole question had been so tangled by assertion
and counter-assertion that it became desirable to have an investigation
into the facts by men of recognized ability and impartiality.[20]

In general, to use statistics safely you need a wide acquaintance with a
subject, especially where the question is in any way mixed up with men's
feelings, whether through politics or not. All the statistics we have
make dead against great armaments and preparation for war; yet while
human nature is what it is, necessary prudence seems to require every
nation of any size to have them. A very little human nature will upset a
very great body of statistics. Furthermore, in most human affairs
results are produced by a multiplicity of causes; and though statistics
may throw light on three quarters of all the causes that are potent in
any given case, yet the other quarter which are irreducible to definite
statement may wholly alter the result. If you are using statistics in
your argument, therefore, as evidence of some large and complex fact,
you should usually justify them to some extent by showing that there are
no counteracting forces which they do not cover.

With this precaution, however, statistics are the f