lected are very little better than rough guesses. The same considerations apply to "value of products," "cost of materials," "expenses," etc. The difficulty is not necessarily one of error in reporting (yet undoubtedly this is an important factor) nor in the accuracy with which such facts might be determined, but rather with the accuracy with which they are determined under the conditions of collection. If nothing more is desired than an indication of trend this may be secured in cases where complete accuracy of detail is wanting, providing errors are distributed uniformly about the average and tend to correct each other, and where sampling is representative. These conditions, however, so seldom maintain (never in the last instances cited) that data compiled ostensibly under these considerations must be used with great care and circumspection for any use where accuracy is important or where vital issues are involved. It is painful to see nice distinctions and weighty conclusions rest upon such questionable support!

On the other hand, secondary statistical data are frequently compiled where absolute accuracy of determination is impossible and where no pretense is made toward completeness. The data at best are estimates. At present no statistical machinery and data are available for an accurate determination of the amount of gold-producing ore in the United States; of the amount in horse-power of our water power resources, or of the amount of standing timber existing in the United States.1 Absolute accuracy is not necessary and no pretense is made of its realization. Of course, there may be accurate and there may be inaccurate estimates,

1 See the interesting report on "The Lumber Industry, Part I, Standing Timber," by The United States Bureau of Corporations, 1913, where methods of estimating the amount of standing timber in various districts and for various woods are described and criticized, pp. 7–10, 45 ff.

and it is always incumbent upon him who uses them to choose those which, all things considered, seem best to meet the requirements which they should possess and to use them as estimates. Essentially accurate conclusions, of course, may be drawn from rough estimates, but in their use the element of danger is so great that caution should always accompany their employment, and sound judgment constantly be invoked to guard against false conclusions being drawn from them.

Moreover, not all phenomena allow of statistical measurement. Numerical frequency may be of no real nor vital significance. The devotion of a people to a principle of right or justice can hardly be measured by the number of those who find no occasion to violate it. Regard for law and order may not be measured by the number of people who remain out of jail. Conversely, the disregard for law is not fully measured by the number of arrests and convictions for a given period. The degree of insanity is not necessarily measured by the number of commitments to insane asylums together with the number of occupants of such institutions. The sacredness with which the marriage institution is regarded is not accurately reflected by the number of divorces granted, nor respect for higher education alone by the number of students enrolled in institutions of collegiate and university rank. It is an error to expect statistical data alone to answer these questions, and it is even a worse error solely to base conclusions respecting them on data which are now extant.

Not less important than the element of accuracy is a sixth consideration, viz., the homogeneity of conditions which the data describe. If violent changes of methods of doing business have resulted during a period of time, and the corporate form of organization has become more common

because of the relative size of the business unit, then it would be inaccurate to base conclusions respecting the proportion of business done under this type of organization at two periods where all business is taken as the basis of comparison. If "future" transactions, in a given market, are supplanting "spot" transactions, and the substitution has caused prices to rule higher or lower, then prices to-day may not be compared in this respect with those characterizing a period when such methods of dealing were not indulged in. If prices to-day are influenced by the practice of retail dealers "protecting" manufacturers by refusing to give price concessions, then present prices are not fully comparable with times when such conditions did not maintain. If price levels are to be compared, it is unfair to make the basis of comparison prices of commodities bought in small quantities with those paid for in wholesale lots. The conditions are not equivalent, and comparisons are invalid until they are reduced to a common denominator. Prices expressed in a depreciated standard can be compared with those made on a gold basis only after a conversion of one has been made into terms of the other.

Not only may statistical data be descriptive of nonhomogeneous conditions (and this fact not be revealed), but they may also greatly differ in composition at different times. Reporting, editing, tabulating, and analyzing may be of widely different degrees of excellence. New forces may have been given recognition, different emphasis may have been placed on different things, different definitions may have been insisted on, new units of measurements or modifications of old ones may have been employed, wider or narrower fields may have been covered, the proportional elements used to make up a total may have changed materially, etc. The presence of these and similar conditions makes

comparisons over long periods difficult, if not exceedingly dangerous. The desire for "comparability" often becomes the controlling factor in statistical computation, and serious omissions, strained interpretations, etc. (all important in the use of the data for a given time) countenanced in order to preserve it. The retention of the "capital" inquiry, in all its crudity, in the statistics of manufacture by the United States Census Bureau is largely out of consideration for the "value of comparisons." The omission until recently on the part of the United States Bureau of Labor Statistics of fifteen commodities formerly used in the computation of an index number of retail prices, raises at least the question of the possibility of comparing the figures before 1907 with those since that date. The various definitions of a "farm," or of an "establishment," or of "manufacturing" used by the United States Census Bureau at different times, make hazardous comparisons over an extended period. Exports and imports, whether expressed in quantity or in value, must always be interpreted in terms of the units of measurement employed.2 The student should always go behind

1 The lack of comparability has been definitely asserted by the Commissioner of the Bureau of Labor Statistics. "Some Features of the Statistical Work of the Bureau of Labor Statistics," Royal Meeker, Commissioner, Publications of the American Statistical Association, March, 1915, pp. 431

441.

2 Most interesting discussions of the difficulties of making international comparisons of import and export statistics, and of the imperfections of our own import and export statistics, are contained in an article by Frank R. Rutter on "Statistics of Imports and Exports," in The Publications of the American Statistical Association, March, 1916, pp. 16-35. Apropos the topic here under consideration the following extracts are of interest:

By virtue of a law passed in 1893 the agent of a railroad company carrying goods to a foreign country by land was made punishable to the amount of $50 for failure to present a manifest to the collector of customs. "The effect of the change in law is reflected in the exports through Buffalo to Canada. From less than $500,000 in 1890 the figures jumped to over $4,000,000 in 1895." Ibid., p. 20.

On the matter of units of measurements and classification, the following

the printed figures and make sure of the units, their interpretation, and the weight assigned to the different factors

quotation is of interest: "The greatest need for the expansion of the classification is found in the case of exports. The most detailed classification of exports now covers less than 600 items, while in the imports for consumption there are about 3,000 distinct items. The chief preventive of an increase in the number of items is the indefinite character of export declarations. So many articles are described merely by general terms that it is out of the question to separate articles frequently of much commercial importance.

"Defects in the present classification, aside from its incompleteness, are the incomparability of the import and export schedules and the failure to conform to current commercial terms. The latter defect is due to the preservation in the tariff of many terms now obsolete, and the necessity of having the statistical classes follow closely the tariff items." Ibid., p. 26.

On the definition of "imports" the author says:

"What is generally understood by the term 'imports'? Legally, an article is imported when landed, whether for immediate consumption or for storage in bonded warehouses. From an economic point of view, however, bonded warehouses may well be regarded as foreign territory. The door of the bonded warehouse is really the economic frontier of the country.

"Since the United States is not a large reëxporting country, the difference between 'imports' and 'imports for consumption' is largely one of time. The instances in which goods are exported from warehouses are few as compared with the instances in which after the lapse of time goods are entered for consumption within the country.

"Perhaps the distinction is most clearly brought out by an illustration. While the last tariff was under discussion wool in large quantities was landed at our ports and stored in bonded warehouses until December 1, 1914, when it could be withdrawn without payment of duty. Was such wool really imported when it was landed or when it was removed from the warehouse?

"On the export side we have a clear distinction between domestic exports and foreign exports. On the import side imports for consumption are most nearly comparable with domestic exports, yet not fully comparable, since free goods are not generally warehoused and may be entered for consumption although intended for reëxportation. To be strictly accurate, dutiable imports for consumption should be compared with domestic exports and free imports with domestic and foreign exports combined." Ibid., p. 28.

"Perhaps the most striking instance of the unfortunate result of our method of valuation is seen in the import prices of rubber. Notwithstanding the improvement of plantation rubber, Para rubber is still quoted at a slightly higher price. In Brazil, however, there is a heavy export duty, which constitutes an important element in the price. This duty is not included in our statistical valuation with the result that the value of India