How Statistics Are Manufactured: The Experiences of a Census-Taker
THROUGHOUT the civilized world statistics have come to be regarded as public guide-posts, not mere monuments to public whim. The decennial national inquisition conducted by the census agents can be justified only on the ground that some important public end is to be subserved. That end is the guidance of the statesman. Statist, statesman,and statistics are of the same derivation, and refer to public servants. Statistics have in fact always been distinguished from mere facts, in that statistics, instead of being dead and unrelated, are intended to suggest action, social control of future contingencies, social mastery of the forces whose working they chronicle.
In spite of the avowed incompleteness and inaccuracy of former census returns, whether relating to the vitality or to the industrial activity of our population, it must be admitted that the average citizen gives to our official statistics absolute credence. Cynical as we profess to be with reference to the expert’s motives or hypnotic skill, nevertheless we are worshipers of the tables whose meaning he is thought to pervert. Rarely is the accuracy of these tables themselves challenged. On the contrary, we submissively accept them as the major premise of the political syllogism, and divide into parties for the discussion of the conclusions. The wag’s remark that one should always state, in applying, whether Republican or Democratic statistics are wanted, does not imply that the individual Republican or Democrat questions the infallibility of his own statistics. Thus it happens that, just as we accept as final the population enumeration and base upon it a distribution of influence in national legislation, so upon industrial statistics we base our political and economic policies.
Acting upon this principle, the government questioned in 1900 about 640,000 manufacturing establishments on the assumption that an analysis of the returns would throw light upon the relation to the public welfare of protection, the gold standard, strikes, trusts, expansion, bounties on shipping, immigration,higher education, factory legislation, and the like. The facts gathered four years ago have been classified and compiled at Washington. Immediately upon the advance publication of totals, showing an increase of over 40 per cent since 1890, the political syllogism was exploited, and the figures were said to warrant this conclusion, quoted from a Philadelphia newspaper: “This is not at all in accord with the theory that only great business concerns any longer have any chance of success in this country.” It is not extravagant to predict that the material contained in the four volumes of manufactures statistics recently published will form the basis for political controversy for the next two decades.
It is the object of this paper to indicate as briefly and correctly as possible the nature of a large portion of the material. If it should appear that the writer does not share the superstitious reverence for official statistics, let it not be inferred that he wishes to depreciate the importance of the decennial census or of the statistical method. It certainly is not desirable to diminish public interest in our statistics. It has been a very difficult and expensive matter — over one billion dollars for the last census — to collect and compile the voluminous reports of the last two censuses. Yet it is perfectly obvious that the value of future statistics must depend upon candid recognition of the defects of past statistics. The chief statistician himself, in four pages of Part I, has warned those who might prepare campaign editorials and historical summaries of certain limitations, declaring candidly: “The science of statistics, as applied to manufactures, is yet in its infancy.”
The true character of these statistics is probably best known to the census agents. They helped to manufacture the statistics of manufactures. They have, therefore, an intimate knowledge of both the raw material and the workmanship which have gone into the finished product recently published by the census office. The writer had one month’s experience as special agent in two districts in the city of Philadelphia. Statements were taken from about one hundred and fifty manufacturing establishments. Acquaintance with the daily experiences of a score of special agents working in the various districts of Philadelphia assures me that the conditions hereafter described are not exceptional. On the contrary, the average intelligence of the two districts cited is considerably above that of most of the districts in any large city.
Nor is it probable that conditions in Philadelphia specially favored inaccuracies. On the contrary, it is most likely that nowhere in the country was a more conscientious effort made to secure reliable results. The Chief Special Agent, Mr. George S. Boudinot, had not only years of experience as manager of commercial agencies, but had won the confidence of the census authorities, for whom he had conducted the bureau of manufacturing statistics in 1890. Furthermore, Mr. Boudinot employed every university or specially trained man whom he could, having on the staff many men experienced in the service of mercantile agencies. It would be interesting to hear from Washington with reference to the qualifications demanded of special agents in other special districts. It will appear from the experiences here recounted that the task assigned to the census agent was too great for any training less than that of a magician or a fortune-teller.
The schedule asked for an exhaustive statement of the preceding year’s business. Special industries, such as sawmills, chemical factories, woolen and steel factories, and the like, were reported on schedules calling for minute details with reference to the specific industries, such as the quantity of chemicals, of various grades of lumber, or qualities of wool. The vast majority of industries, however, were reported on “schedule three.” In “ schedule three,” therefore, will be found the principal source of errors.
This called for (1) the capital and its distribution into buildings, land, machinery, tools and implements, stock and cash on hand, and bills receivable. (2) The greatest and least number of persons employed during the year, including men, women, and children over sixteen, whether piece workers, ordinary employees, superintendents, managers, clerks, salesmen, salaried officers of corporations, proprietors and firm members. (3) The total account of wages or salaries paid during the year. (4) The average number of wage earners employed during the year, including piece workers, but excluding all who draw salaries. It is obvious that this information furnishes a check on the statement of wages paid during the year, as well as on that contained in (5), the number of months in operation on full time, on three fourths time, on half time, on one fourth time, and the number of months idle. (6) A statement of the quantity and cost of materials purchased under the following classifications: Raw materials, partially manufactured, fuel, mill supplies, all other materials, and the amount of freight paid on goods purchased. (7) Miscellaneous expenses, including rent of works, rent of power and heat, amount paid for taxes, rent of office, interest, insurance, internal revenue tax and stamps, ordinary repairs of buildings and machinery, advertising and all other sundries. Under this head is placed the amount paid for contract work. (8) The kind, quantity, and value of goods manufactured, the amount received for custom work and repairing, including a statement of the value of the manufactured products for the preceding business year. (9) Lastly are statements as to the kind, amount, and ownership of power used.
The manufacturer is to answer these questions under penalty not to exceed a fine of $10,000 and one year’s imprisonment. His signature on each schedule certifies that the information therein contained is complete and correct to the best of his knowledge and belief.
In the case of establishments which keep detailed book accounts there is obviously no obstacle to an accurate report, except the almost ubiquitous desire of the proprietor to overstate his expense account, or understate his product, or both. The chief statistician has declared that there has undoubtedly been a great increase in the number of establishments which keep adequate records of the year’s business. Just what the increase has been we do not know, nor will the present census enlighten us as to the proportion of its material that is based upon book accounts. The census expert did not explain why it was impracticable to ascertain this proportion. It is hard for the casual observer to understand wherein lies the impracticability which Professor Willcox mentions. After a business man has told his age, present and past marital relations, size and sanity of his family, the purity of his blood, his capital, expense account, wages paid to children employed, product and profit of two years’ business, it is hardly conceivable that he would object to state whether the abstract to which he subscribes is from his ledger or from his memory.
Judging from my own acquaintance with these reports, not one in ten is based upon ledgers. In one month’s experience not one report was taken entirely from books. In the one sworn statement that came nearest to fulfilling these requirements it was necessary to add $25,000 to the product in order to bring it within the limits of reasonableness. In my two districts not ten per cent pretended to keep even partial accounts. Bills payable, with duplicates, are filed on a wire hook, the assets are in the till. The past is buried and forgotten. Cigar factories furnished exceptions so far as concerned amounts of tobacco used, cigars made, and taxes paid. Even here, however, the same internal revenue book once told two entirely different stories in the hands of the two members of the firm. A colleague worked in a district where unbusiness-like methods were so general that he was not surprised to be called in as arbiter in a discussion between a manufacturer and her son as to the number of children in the family. He was able to do what cannot be done with most census returns, — suggest a lining up along the wall for purpose of roll call. The present paper deals with returns based almost entirely upon memory or guesswork, under the censorship of the special agent.
Our instructions were to record every productive industry, no matter how trifling its product, nor how infinitesimal its capital. Industry was productive when it resulted in changing the form of a commodity. To change place or ownership is exchange, to change form is production. There can be no doubt in the minds of special agents that an unnecessarily numerous army of manufacturers are engaged in changing form on a small scale. To make sure that these small factories would not be overlooked or scorned, the government paid by the day, not by the average product. The agent was expected to be on the outlook for signs of productive workmanship. To guaranty that he would not pass by some productive industry in ignorance that it belonged in the category of manufacturing establishments, the government published a list.
Among the many types of manufacturers which the agent, unwarned, might have passed by, may be mentioned the following: A tailor who makes suits; the custom branch of a department store; a tailor who contracts to have suits made; a tailor who runs a sweatshop; a tailor who “sweats;” a tailor who only patches and sews buttons; a washerwoman who cleans clothes with benzine; the German plumber who repairs clothes at night ; the dressmaker who furnishes only her labor and needles; the dry goods merchant who converts unsalable cambric into aprons or doll clothes. likewise a blacksmith is a manufacturer, as are the jobbing carpenter, mason, plumber, locksmith, cycle-repairer, lamp-mender, wick-fitter, shoemaker, milliner, cooper, paperhanger, plasterer, painter, photographer, printer. An undertaker who trims coffins is a manufacturer, because he, like the others above mentioned, changes the form of the products which pass through his hands. Little wonder that one agent found place for a crematory on his schedule!
Supplementary occupations furnish a still more perplexing list of census factories. One is surprised to learn that the following are manufacturers: the day laborer who works at night, six weeks out of the year, sharpening skates; others who work off and on at spare moments after supper repairing bicycles, mending shoes, rolling cigars, cooling candy, turning ice cream freezers. The housewife who takes in sewing during September and October; the dry-goods firm that sends its shopworn remnants out to be made up; the hardware company with a capital of $100,000 which makes five dollars’ worth of blank keys, or earns ten dollars trimming and repairing lamps; these are manufacturing concerns. A department store counts for as many manufacturing establishments as it has departments that repair, renovate, make, or change the form of commodities.
The difficulty of obtaining a report of reasonable accuracy is obvious. Even if complete book accounts were kept, it would be practically impossible to determine the proper relation between expense and profit of such enterprises as above enumerated. The agent had nothing In guide him but his own discretion and the necessity of returning a consistent schedule. The method of procedure may be indicated by citing as typical schedules those for dressmakers, night workers, bakers, and supplementary manufactories.
Out of a large number of dressmaking establishments reported by me, not one was in a special building, nor did one keep books. Oftener was the establishment conducted by one or more members of a family, who thus contributed to the family income by working off and on during the year upon such custom as was attracted by the sign in the window. The problem was, then, to estimate the portion of the expenses of a private residence which should be charged to the manufacturing of dresses. In most cases, the materials were furnished by patrons,— a fact which the census office was quite reluctant to recognize. The employees were for the most part members of the same family, or perhaps unpaid “learners.” In many cases the only capital invested was one sewing machine, which would have been in the house even if no custom work were done. The agent learns from experience that it is futile to permit the dressmaker to think over the schedule. This will result in a very imperfect and impossible schedule, or perhaps in the firm decision not to “answer those impudent questions, because her lawyer brother says it is not necessary.”
It is also unwise to ask questions in the order on the schedule, for the fabric of guesses is certain crazy patchwork by the time the tenth blank is filled out. Therefore, the agent learns to begin by obtaining some idea of the number of months the “factory” is employed on full time and the number of sewing machines used. If the number of employees seems to correspond to the number of sewing machines, the agent has the basis for a guess at the product, He does not ask what it is. On the contrary, he suggests $500, $1000, $850, and then, like the Dutch auctioneer, deducts until he and the dressmaker come to a practical and amicable agreement. Or perhaps he suggests the number of dresses made, in the absence of positive proof to the contrary, the manufacturer accepts, often with unfeigned dismay, the a priori, inevitable product which is shown to meet all of the requirements of mathematical accuracy and statistical harmony.
Given the product, it is necessary to determine expense. The manufacturer would not continue in business unless it was profitable. By subtracting this reasonable profit from product, the agent has left the gross cost of conducting the business. He begins to inquire. How much rent would you pay, if you paid rent ? How much salary would you receive, if you received a salary ? How much would you have paid your sister, if she had been some one’s else sister ? How much material would you have bought, if you had furnished material ? How much would fuel and light have cost, if you had paid for fuel and light ? How much capital would the sewing machine represent, if it were yours and employed only in manufacturing? “If you had a brother, would he like green cheese?”
Not less enigmatical is the schedule of the day laborer who conducts manufacturing some hours of some nights of some months of the year. How many hours or nights or months he will not know. How many lawn mowers or skates he may have sharpened, how many sewing machines or bicycles repaired, or the amount earned by pressing trousers or cleaning coats or soling shoes or cooling candy, he cannot tell. Infinitely less intimate is he with the expense of conducting what he is astonished to hear is a manufactory. He has always thought his profit from this extra industry was pure gain. He is disappointed to learn that his profit must be only reasonable, according to the modest minimum standard of the census office. His disappointment changes to wonderment, however, as he observes the ease with which the agent has convinced him, by the Socratic method, that he pays rent and wages and profits to himself as landlord, laborer, and captain of industry, each bearing to the other a relation of in exorable consistency.
Typical of supplementary industries is the large retail hardware store which makes ten dollars’ worth of keys in the course of twelve months. Having listed this establishment, it is clear that the product sets a limit at once upon the inquiry. The agent must prepare the schedule so as to leave a reasonable profit on key manufacturing. What portion of the $50,000 capital is to be charged to key making? What portion of the $2000 rent ? What share of the tinkering clerk’s salary of $100 a month ? Shall the business be reported as running twelve months a year on one fourth time, or one month on one fourth time ? How much may reasonably be charged to fuel, light, and taxes, or to salary of firm members?
Similar to this case is that of the retail dry-goods firm with a capital of $100,000, which manufactures, from unsalable calico, aprons to the value of $12 a year, or doll dresses to the value of $7. How shall unsalable calico be appraised ? how many laborers shall be recorded ? Here again is the repetition of the problem of adjusting expenditures to product.
Another type of industry which presents numerous difficulties is the bakery. The returns show 14,917 bakers, with a product of $176,000,000. In almost every bakery are combined a manufactory in the basement or kitchen, a ground floor retail shop which sells bread, cake, pies, ice cream, and the like, and a residence. A moment’s consideration will show how natural it is for the actual bakery to be treated as a factory in making up the expense column, and as a retail store in the product column. From this situation doubtless arose the conviction expressed by a certain census official that “any baker who says he does not make a profit of 100 per cent is lying.”
In determining the expense we had no instructions regarding the apportionment of capital among kitchen, store, and residence. Nor did we compute seriously the value of the services of wife and children. The product, in practice, was discovered somewhat as follows: The baker gives the capital invested, which the agent distributes reasonably. The baker then guesses at the average number of barrels of flour used per week. A barrel of flour should make 270 loaves of bread. The agent, knowing this, can compute the number of loaves per year and the reasonable cost of furnishing salt, butter, etc. The baker is asked about the number of pounds of cake, cookies, etc., to which items five to fifteen per cent of the flour will go, on the average. It the agent adds this product to the bread without first having deducted fifteen per cent of the flour, he has overstated by $1000 to $4000 the product of a bakery which consumes twenty barrels of flour each week.
The total value of the baker’s product depends entirely upon the agent. He may and probably does consider that all of the bread and cake and ice cream are actually sold, and at the maximum market price. The facts would seem to prove the inaccuracy of each assumption. One baker declared that he bought no fuel, and when asked with what he kept a fire in the oven, snarled “Bread!” He partially made good his claim by showing piles of hard bread. Another, who did a thriving business, was astounded at the agent’s extravagant showing of profit. He later found the discrepancy in the fact that a large portion of the product was sold at a sacrifice, and from six to ten per cent was waste. Others agreed that bread brought more when fresh than when stale, and that not all of the product found buyers. It is doubtful, therefore, if a strict separation of factory income from retail income and a careful statement of sums actually received, based upon quantity sold, not manufactured, would have convinced the census officials that the reasonable minimum profit for bakers was 100 per cent. In fact, the final returns show a profit of but 35 per cent, including interest on the capital invested.
These cases are sufficient to indicate the difficulties which confront the census agent, assuming that he is honest, in his desire to ascertain the truth. They indicate furthermore what splendid opportunities are afforded to the agent whose sense of humor, local patriotism, party loyalty, or desire to rank high among his colleagues, constrains him to look at industrial statistics as puzzles to be skillfully and tastefully dovetailed one with another. And they must also demonstrate the necessity of relying upon the agent for the facts, rather than upon the manufacturer. To the latter the process of filling out a schedule is equivalent, to a seminar in business accounting or a laboratory experiment at Monte Carlo. He is astonished to find out how little he knows about his own business, but having had a product named for him, and a reasonable profit of which he never dreamed, he is prepared for any number of surprises in matters of detail. It is all like the education of Alice in Wonderland. His business grows on the income side and shrinks on the cost side, and against his conviction he signs his name to a statement that to the best of his knowledge he has done a reasonably profitable business.
It maybe suggested that these cases are unimportant and obviously exceptional. As to the latter point, only the census office can give a satisfactory answer. Unfortunately, its answer must be incomplete, for no provision was made for denoting incidental or supplementary manufacturing. The only incentive to mention the fact was the fear that otherwise the schedule would be returned as not showing reasonable relations, item for item. As to the importance of these minor establishments, that depends upon their number and also upon the use to which the returns will be put. It needs no argument to show how necessary minute classification is to insure a reliable application of industrial statistics to a discussion of the trust problem. For practical public purposes not one of the abovementioned establishments, except possibly the bakeries, belongs to the category of manufactories, however necessary they are to a statement of the nation’s productivity.
The concentration of responsibility in the hands of the census agents would seem to make possible a degree of uniformity of standards not practicable among manufacturers. So far as the last census was concerned, the central office made no provision for uniformity of standards, except to provide supervisors to judge the reasonableness of a schedule. The agents were left to their own discretion. The result will be, therefore, hundreds of thousands of schedules based upon varying principles of division and adjustment. In his independent helplessness the agent could know but one rule,— that is, to maintain proper proportions. The manufacturer insists upon overstating expenses, — the harmony is restored and the business properly described by the agent’s pushing up receipts. The agent employs, therefore, the sliding scale of the physician who sends his bill to the grocer after receiving the latter’s monthly statement.
The central office is responsible for the perfect working of this sliding scale. Schedules were time and again returned as unsatisfactory, because they showed that the traditional reasonable profit was not being earned. In time of prosperity manufacturers make money. They would not continue in business if they did not make money. A baker says he is losing capital. A plumber says he is living on rents. A blacksmith maintains that he suffered an accident last year which makes it necessary for him either to continue his patronage at a loss of $600 or to give up his business entirely. These showings were mistrusted by the census office. The agent had to protect himself by appending signed explanations of what is ordinarily accepted as a commonplace of business experience, namely, that a goodly proportion of business experiments fail.
Undoubtedly the feeling was widespread among agents that the object of the census was not to ascertain the truth, the whole truth, and nothing but the truth, but rather to prove a thesis. The moral tone of the agents suffered depreciation, even in a few days. The sliding scale reigned supreme, while the slowest and most ignorant and most venal man on the staff was permitted to set the standard of a day’s work for Uncle Sam. The epicurean cynicism and abandon of the staff seemed to justify the warning of a noted statistician to the writer: “If you ever expect to work with statistics, the less you know of their origin, the better.”
The difficulties are numberless, but the count is in, and must be used as effectively as possible. Luckily, too, the original material is in charge of skilled statistical experts. By minute and candid classification the inaccuracies may be weighted. The a priori reliable statistics may yet be sifted more carefully from the a priori unreliable, and the possibility of building upon the inaccuracies of this census in all future statistical computations may be minimized.
A little more time and expense in connection with the analysis of statistics of manufactures for 1900 will save much outlay for the next census, and many mistakes in legislation and administration. It is manifestly poor economy for Congress to spend millions in collecting less complete statements than circumstances permit. Perhaps it might pay for Congress to demand confessions from various special agents, to be used as hints to the experts and stimuli to the permanent census bureau.