Hon. Carroll D. Wright gave the first of his series of lectures on wage statistics last night in Harvard 1, the subject being "Methods and Difficulties in Collecting Statistics of Wages." The speaker after being introduced by Professor Taussig said in part:

The wages question is perhaps the most important branch of political economy and its subjective treatment is attained by the use of statistics. But it is essential that the user of statistics should understand the proper method for constructing them, and the attendant difficulties. There are four great sources of error which are to be avoided in securing statistics. First, observation is not trustworthy; there must be reliable figures and they should cover a considerable space of time. Second, the distinction must be kept in mind between "average" and "arithmetical mean." Third, in dealing with percentages the basic numbers must be carefully considered. Fourth, there is great danger in the use of concomitants; that is, arguing that one thing results from another because it is contemporaneous.

The difficulties of obtaining correct data are almost insurmountable. In the first place all figures must be obtained by personal inspection, for a manufacturer answering a letter of inquiry would be sure to fall into one or more of the fatal errors enumerated above. The proper method of constructing a schedule is to show the wages paid to each employee through a period of years, but here the statistician is confronted with serious problems. The number of people employed is always difficult to find; and even when that is done he may be unable to compute the duration of the labor because the time varies with different industries, and in different departments of the same industry. Or it may be that wages have been paid to a group of laborers for work by the piece and here the statistician knows neither the number of men nor the time.

These difficulties are aggravated by the varying rates of wages which are paid in the same manufacturing concern. In some mills, for instance, weavers are paid from 60 cents to $1.49 a day, there being perhaps fifty rates between the two extremes. Calculations can be made here only on the basis of average. The United States Census deals only with the statistics of wages relating to manufacturing but even here the method used is faulty and is a makeshift for the correct system, which is impossible of consumtion. The problems of collecting railroad wage statistics are also perplexing on account of the great variety in the grades of labor employed. The railroad managers in giving out their figures have been accustomed to throwing all the grades about a thousand in number, into five classes and giving the average for each class. Henry C. Adams, Statistician of the Interstate Commerce Commission, is working toward dividing their scale of wages far more accurately into a large number of classes. But even here the inherent difficulties of the problem are increased by the innovation of paying employees "by the run" instead of by the time.

The problem for the statistician is a difficult one. He must use no a priori methods but must analyze accurately a great mass of facts and give the public results that can be depended upon.