census of the population, taken on January 28, 1897, prove satisfactory and if they are analysed in detail, they will greatly facilitate the taking of an industrial census. As long as there are no such censuses it can only be a question of registering some of the big industrial establishments. It must be conceded that the present system of collecting and processing statistical information on such big establishments ("factories and workers" in the prevailing terminology) is unsatisfactory in the highest degree. Its first shortcoming is the division of factory statistics among various "departments" and the absence of a special, purely statistical institution that centralises the collecting, checking, and classifying of all information on all types of factories. When you analyse the data of our present-day factory statistics you find yourself on territory that is intersected in all directions by the boundaries of various "departments" (which employ special ways and means of registration, and so on). It sometimes happens that these boundaries pass through a certain factory, so that one section of a factory (the iron foundry, for example) comes under the Department of Mines and Metallurgy, while another section (the manufacture of ironware, for example) comes under the Department of Commerce and Manufactures. It can be understood how this makes the use of the data difficult and into what errors those investigators risk falling (and fall) who do not pay sufficient attention to this complicated question. With regard to the checking of the information, it must be said in particular that the Factory Inspectorate will, naturally, never be in a position to check the extent to which all information supplied by all factory owners corresponds to reality. Under a system of the present-day type (i.e., under which the information is not gathered by means of a census conducted by a special staff of agents but by means of questionnaires circulated among factory owners), the chief attention should be paid to ensuring that the central statistical institution have direct contact with all factory owners, systematically control the uniformity of the returns, and see to their completeness and to the dispatch of questionnaires to all industrial centres of any importance -- that it thus prevent the fortuitous inclusion of dissimilar data, or different applications and interpretations of the programme. The second basic shortcoming of present-day statistics
lies in the fact that the programme for the gathering of information has not been elaborated. If this programme is prepared in offices and is not submitted to the criticism of specialists and (what is particularly important) to an all round discussion in the press, the information never can be in any way complete and uniform. We have seen, for example, how unsatisfactorily even the basic programmatic question what is a "factory"? -- is being solved. Since there is no industrial census, and the system employed is that of gathering information from the industrialists themselves (through the police, the Factory lnspectorate, etc.), the concept "factory" should most certainly be defined with complete accuracy and limited to big establishments of such size as to warrant our expectation that they will be registered everywhere and in their entirety without omissions. It appears that the fundamental elements of the definition of a "factory establishment" as at present accepted have been quite well chosen: 1) the number of workers employed in the establishment to be no fewer than 15 (the question of separating auxiliary workers from factory workers in the true sense of the word, of determining the average number of workers for the year, etc., to be elaborated); and 2) the presence of a steam-engine (even when the number of workers is smaller). Although extreme caution should be exercised in extending this definition, it is an unfortunate fact that to these distinguishing characteristics have been added other, quite indeterminate ones. If, for instance, the bigger establishments employing water power must not be omitted, it should be shown with absolute accuracy what establishments of this type are subject to registration (using motive power of not less than so many units, or employing not less than a certain number of workers and so on). If it is considered essential to include smaller establishments in some branches, these branches must be listed very precisely and other definite features of the concept "factory establishments" must be given. Those branches in which "factory" establishments merge with "handicraft" or "agricultural" establishments (felt, brick, leather, flour milling, oil pressing, and many others) should be given special attention. We believe that the two characteristics we have given of the concept "factory " should in no case be extended, because even such relatively big
establishments can scarcely be registered without omissions under the existing system of gathering information. A reform of the system may be expressed either in partial and insignificant changes or in the introduction of full industrial censuses. As far as the extent of the information is concerned, i.e., the number of questions asked the industrialists, here, too, a radical distinction has to be made between an industrial census and statistics of the present-day type. It is only possible and necessary to strive for complete information in the first case (questions on the history of the estabIishment, its relations to neighbouring establishments and the neighbourhood population, the commercial side of affairs, raw and auxiliary materials, quantity and type of the product, wages, the length of the working day, shifts, night-work and overtime, and so on and so forth). In the second case great caution must be exercised: it is better to obtain relatively little reliable, complete, and uniform information than a lot of fragmentary, doubtful information that cannot be used for comparisons. The only addition undoubtedly necessary is that of questions on machinery in use and on the amount of output.
In saying that our factory statistics are unsatisfactory in the highest degree, we do not by any means wish to imply that their data are not deserving of attention and analysis. Quite the contrary. We have examined in detail the shortcomings of the existing system in order to stress the necessity for a particularly thorough analysis of the data. The chief and basic purpose of this analysis should be the separation of the wheat from the chaff, the separation of the relatively useful material from the useless. As we have seen, the chief mistake made by Mr. Karyshev (and many others) consists precisely in the failure to make such a separation. The figures on "factories" are the least reliable, and under no circumstances can they be used without a thorough preliminary analysis (the separate listing of the bigger establishments, etc.). The number of workers and the output values are much more reliable in the grand totals (it is, however, still nccessary to make a strict analysis of which productions were included and in which way, how the output value was computed, etc.). If the more detailed totals are taken, it is possible that the data will prove unsuited for comparison and their use condu-
cive to error. The fables of the reduction of the number of factories in Russia and of the number of factory workers (relative to the population) -- fables that have been so zealously disssminated by the Narodniks can only be explained as due to the ignoring of all these circumstances.
As far as the analysis of the material itself is concerned, it must undoubtedly be based on information on each separate factory, i.e., card-index information. The cards must, first and foremost, be grouped by territorial units. The gubernia is too big a unit. The question of the distribution of industry is so important that the classification must be for individual cities, suburbs, villages, and groups of villages that form industrial centres or districts. Further, grouping by branches of industry is essential. In this respect our latest factory statistical system has, in our opinion, introduced an undesirable change, causing a radical rupture with the old subdivision into branches of industry that has predominated right from the sixties (and earlier). The List made a new grouping of industries in twelve sections: if the data are taken by sections only, we get an excessively broad framework embracing branches of production of the most diverse character and throwing them together (felt cloth and rough felt, saw mills and furniture manufacture, notepaper and printing, iron-founding and jewellery, bricks and porcelain, leather and wax, oil-pressing and sugar-refining, beer-brewing and tobacco, etc.). If these sections are subdivided in detail into separate branches we get groups that are far too detailed (see Mikulin, op. cit.), over three hundred of them! The old system that had ten sections and about a hundred branches of production (91 in the Directory for 1890) seems to us to have been much happier. Furthermore, it is essential to group the factories according to the number of workers, the type of motive power, as well as according to the amount of output. Such a grouping is particularly necessary from the purely theoretical standpoint for the study of the condition and de velopment of industry and for the separation of relatively useful from useless data in the material at hand The absence of such a grouping (necessary within the territorial groups and the groups of branches of production) is the most significant shortcoming of our present publications on factory statistics, which allow only "average figures" to be determined,
quite often absolutely false and loading to serious errors. Lastly, grouping under all these headings should not be limited to a determination of the number of establishments in each group (or sub-group) but must be accompanied by a calculation of the number of workers and aggregate output in each group, in establishments employing both machine and hand labour, etc. In other words, combined tables are necessary as well as group tables.
It would be a mistake to think that such an analysis involves an inordinate amount of labour. The Zemstvo statistical bureaus with their modest budgets and small staffs carry out much more complicated work for each uyezd; they analyse 20,000, 30,000 and 40,000 separate cards (and the number of relatively big, "factory" establishments throughout the whole of Russia would probably not be more than 15,000-16,000); moreover, the volume of information on each card is incomparably greater: there are several hundred columns in the Zemstvo statistical abstracts, whereas in the List there are less than twenty. Notwithstanding this, the best Zemstvo statistical abstracts not only provide group tables under various headings, but also combined tables, i.e., those showing a combination of various features.
Such an analysis of the data would, firstly, provide the requisite material for economic science. Secondly, it would fully decide the question of separating relatively useful from useless data. Such an analysis would immediately disclose the fortuitous character of data on some branches of industry, some gubernias, some points of the programme, etc. An opportunity would be provided to extract relatively full, reliable, and uniform material. Valuable indications would be obtained of the way in which these qualities can be assured in the future.