It is well to mark also, in passing, that the prices of some or many of those articles which occupy a very important place in all calculations of the cost of living of the wage earners--the articles of food and clothing, and shelter--may change in a different measure, or even in a different direction from the prices of the other commodities which compose the general price level. This possibility is the most genuine as regards food prices. Movements of food prices, and, indeed, of the prices of all agricultural products, are apt over short periods to be determined by weather conditions rather than by the industrial events which govern the general price movement. Mr. W. C. Mitchell in his book on business cycles studied the relation between the movements of retail food prices (the figures ordinarily used in cost of living investigations) and general business conditions during the 1890-1910 period in the United States. He writes in conclusion that "these figures (i.e., of 30 retail food prices) indicate a certain correspondence between retail prices and business conditions. In 1893, indeed, the thirty foods rose slightly instead of falling, but they declined during the dull years which followed the panic, and rose again when prosperity returned. The rise was slow until 1900-02; it became slow again in 1902-04; but rapid in 1905-07. The panic of 1907 came too late in the autumn to exercise much influence upon the average retail price level of that year. On the whole, this series reflects the course of business cycles better than might have been expected. For the supply of vegetables and animal foods varies in an arbitrary fashion determined by the weather, and the demand for staple foods is less affected by prosperity and depression than that for the more dispensable commodities."[39] Even over periods of some duration there may be a marked difference between the movement of food prices and other prices. 3.--Changes in the general level of prices must have prior causes, but they, themselves in turn cause economic disturbance. They give a tilt to the whole industrial system which manifests itself in the outcome of distribution. The effects upon the distribution of the product of an upward movement of prices are ordinarily different from those produced by a general decline in prices. It is well to begin with the first case--a period of a rise in the general price level. To give an accurate analysis of the successive interactions by which an upward movement in the general price level, once stimulated, asserts itself, is both a delicate and lengthy task. It cannot be attempted here.[40] It suffices to note the ordinary distributive results of the process; with the important reservation, however, that they do not occur in the measure that the rise is occasioned by a general reduction in the productivity of industry such as might be caused by war.