each precinct seemed to yield a good index or measure of the degree of conservatism of each precinct. These six measures were: state-wide single tax; an amendment reorganizing the whole state government somewhat after the model of the English government; an eight-hour day on public works; a measure giving the mayor power to control street speaking; an amendment requiring a majority of the votes cast at the election for the adoption of constitutional amendments ( a majority of

TABLE IV

CORRELATION BETWEEN THE INDEX OF CONSERVATISM AND THE PERCENTAGES OF VOTERS IN FAVOR OF THE
EIGHT-HOUR LAW FOR WOMEN, BY PRECINCTS

Percentages voting on the conservative side of six measures, by precincts

-27

-30-33

-36-39

-63

-66-69-72-75 Total

6

9

votes cast on the particular amendment was all that was required at the time) and a similar amendment on the adoption of all initiative measures. That such a measure of the conservatism of the precincts is a good one is shown by the coefficients of x, in the regression equations which are listed at the close of the paper (Table V).

Table III shows that the more conservative precincts have larger percentages of women voting in them than the radical precincts have; the divergence is not great, the coefficient of correlation being +0.29. And Table IV shows that the conservative precincts voted against the eight-hour day for women while radical precincts voted for it, the correlation being -0.77. Granting these facts, our problem is to find whether precincts with large percentages of women voting, voted in greater proportion against the eight-hour bill than did precincts with small percentages of women voting, the factor of conservatism being kept constant. This the coefficient of partial correlation shows and on this bill it is 0.30, not so large as when the factor of conservatism was not kept constant, (-0.408). In other words, when we eliminate the factor of conservatism of the precincts, we find that precincts with large percentages of women voting still vote in larger proportion against the eighthour measure than the precincts with small percentages of women voting.

I

The same method of analysis has been applied to the other 25 measures that we have just considered in connec

1 Conservatism is a general term and fluctuates somewhat in meaning. If each of the 26 measures had been voted on in 1912, when women did not vote, there could have been a separate variable to be held constant for each of the 26 correlations. This would have been more nearly ideal than a single variable of conservatism. However, only one measure was voted on at both election periods, the abolition of the death penalty. Using the vote of this measure as the third variable in finding the partial correlation coefficient for this measure, we find it to be 0.04, which is the same as the partial correlation coefficient (-0.01) when the third variable was the general index of conservatism.

'The interpretation of the partial correlation coefficients is somewhat more accurate on measures which elicited great interest and a large vote, partly for the reason that the issues were better known and also because the total number of women voting on the less well-known measures was somewhat smaller, the exact number on each measure being unattainable. However, a theoretical examination of the possible error shows that it could not have been very large.

tion with the eight-hour day and the results are shown in Table I.'

WASHINGTON, D. C.

WILLIAM F. OGBURN AND INEZ GOLTRA.

1 The results shown in Table I practically conclude the discussion of method. However, it seems desirable to add a few comments which will be of interest chiefly to the reader with a knowledge of the method of correlation. The regression equations are found in Table V. In regard to the linearity of the regression equations, they are all clearly linear, except the regressions of X on Y and Z on Y. In these cases the differences between the correlation ratio squared and the correlation coefficient squared are about 0.12, and the errors of these differences are about 0.04. The fact that ratios have been correlated may lead to a suspicion of "spurious" correlation. There is no "spurious" correlation due to ratios, however. This conclusion is based on the analysis in Yule's paper in the Journal of the Royal Statistical Society for June, 1910, "On the Interpretation of Correlations between Indices or Ratios."

X = Percentage of voters in favor of measure, by precincts.

Y = Percentage of those voting that are women, by precincts.

PRICE FIXING IN THE UNITED STATES DURING

THE WAR

III1

1. Chief Purposes in Price Fixing

RICE fixing in the United States was almost entirely the

PR

product of war conditions. In general, there were three

chief purposes in fixing prices: (1) to secure production of needed commodities; (2) to prevent social unrest by checking profiteering, coördinating food prices and wages and stabilizing industrial conditions; (3) to assure government economy, both in buying munitions and in a fiscal sense.

In order to appraise the measure of success and the results of the price-fixing program, it is desirable to present a more detailed statement of the purposes of the price fixers. The con

1 Concluded from the POLITICAL SCIENCE QUARTERLY, vol. xxxiv, pp. 262-289. Supplementing the first installment of this article, published in the March issue of the POLITICAL SCIENCE QUARTERLY, the following facts, which have come to the writer's attention, are worthy of presentation:

(1) The price of silver was fixed by Congressional act of April 28, 1918, which, in dealing with monetary problems, provided for the stabilization of the prices and the encouragement of the production of silver. The act provided for the melting of 350,000,000 silver dollars and the sale of the bullion at $1.00 per fine ounce, and for the purchase of an equivalent quantity of silver from the mines and reduction works of the United States at $1.00 per ounce. As a result the price rose to approximately $1.00. This amounted to an indirect fixing of a minimum price.

(2) The Food Administration, about February, 1918, fixed a price of 99 per pound for arsenic. This was done by agreement with the producers, who stated that the price would yield a fair profit. The price of arsenic had risen from 49 before the war to 159 in February, 1918.

(3) The price of ferro-manganese was fixed for the 70% grade by informal agreement at $250 delivered.

(4) The prices of several chemicals were "fixed", chiefly for government purchase, among them being the following: Carbon tetrachloride, liquid chlorine gas (7%1⁄2o), phenol (28 per lb.), picric acid, and formaldehyde (164 lb.).

(5) The price of toluol was fixed at $1.50-1.55 per gal.

(6) The War Industry Board, in June, 1918, announced to the naval stores industries that the advance in prices of rosin and turpentine would have to cease.

(7) The Council of National Defense, in May, 1917, by agreement with the leading packers, "pegged" the price of hides for government requirements at the existing market rate.