recognize, in a measure, the benefits of coöperation. The companies as well as legislators and officials have made mistakes, but many in both groups seem now to see a method of procedure which affords a solution of the difficulty. One of the principal agencies of late years in bringing about a more rational treatment of the problem involved has been the National Convention of Insurance Commissioners, which at first encouraged the study of public questions and recently has made some progress toward concerted state action, the necessity for which is shown by the variety of laws in the preceding table. Perhaps through this association some of the inconsistencies of legislation may be removed.

Eliminating from consideration state insurance, which is not yet seriously considered as a general method of procedure but rather as a last resource not yet required by the exigencies of the case and the problem of acquiring uniformity of state action, a partial answer to which is furnished by the National Association of Insurance Commissioners, two phases of the question of what agency shall make fire insurance rates are apparent.

1. Shall competition or coöperation be admitted as the basis for ratemaking? This question appears to have been answered in favor of coöperation. This paper has attempted to outline the history of the conflict between these two ideas, to indicate the reasons for the ineffectiveness of legislation intended to enforce competition and to show that a second and different problem has now appeared.

2. Is the public interest better served by allowing the insurers to fix rates under regulatory statutes or by delegating price-fixing powers to state officials under state rating acts? This question has not become a distinctly recognized issue because it is possible in the United States for two radically different methods of regulation to exist for some time in different sections of the country without any definite decision in favor of either. It is probable, however, that in the near future the last-mentioned problem will become one of the most important in fire insurance.

PROBLEMS OF FIRE INSURANCE RATEMAKING

BY ROBERT RIEGEL, PH.D.,

Instructor in Insurance, Wharton School of Finance and Commerce, University of Pennsylvania.

EARLY RATING METHODS

The first theory of rating risks to be covered by fire insurance was a classification of buildings (which were for some time the only property insured) into groups, each risk included in the group taking the group rate. In 1681 there were two such classes in England, brick or stone buildings and wooden buildings, and premiums were proportionate to rental values. For each one pound annual rental 2 s. 6 d. was charged for insuring a brick or stone building and double this amount for a wooden building. The classification system thus begun continued for many years and a relic still remains in the minimum or class rates which are now promulgated for groups of properties closely resembling each other, the results of specific rating for these not being commensurate with the expense involved. The inability to insure personal property and the calculation of premiums on a rental basis have long since disappeared, and rates, furthermore, are no longer applied to rental values. In 1706 a London office classified risks in three categories, common, hazardous and doubly hazardous and, turning attention now to the United States, a Massachusetts company in 1798 devised a sixfold classification of buildings. The classification was in accord with construction, roof and walls and a contents classification comprised hazardous and non-hazardous groups, the rates for contents depending in part upon the character of the containing building. A combination of companies in New York in 1826 agreed upon a building classification of eight groups and a division of contents into four groups, ranging from non-hazardous to specially hazardous.

These brief comments upon the history of rating methods are merely intended to illustrate the inevitable tendency of early underwriters to classify risks for the purpose of arriving at rates. Prior to 1870, for example, all risks in the state of Pennsylvania were rated on the classification basis, such classification being based on whether the building was of brick or frame construction and the presence

or absence of fire protection. The less the knowledge of fire hazards the greater the readiness to classify and group, or to put the matter in another way, the more thorough the analysis of hazard and the appreciation of differences the greater became the number of groups and the difficulty of assigning a risk to a particular group. The early application of the classification method is of particular interest because of the recently developed demand for a reversion to that principle of ratemaking.

THE DIMINISHING USEFULNESS OF CLASSIFICATION

Between 1835 and 1866 many companies introduced the practice of recording their experience with each of a large number of classes of risks with the object of deriving some information relative to the necessary rates upon such classes and the custom continued for many years after its usefulness for this purpose had disappeared. With an increased appreciation of the multitude of differences in construction, use and situation of buildings and the nature and qualities of substances the classifications necessarily multiplied, but the latter could not keep up with the changes taking place in the risks themselves. Modern conditions made the classification method of rating increasingly difficult. With increased knowledge and accessibility of material building construction became more varied and to place a building in a definite class became a more difficult problem. The introduction of machinery as a factor in production created a greater variety of occupancies and processes and an almost universal use of power generated in various ways. Heating and lighting methods were also radically modified. It is significant that what a prominent student of rating methods calls "the first real schedule" was devised in 1868 and that a great development in schedules had taken place by 1890; for the nineteenth century was prolific in inventions which radically altered industrial and trade conditions. Whole new industries were created by the telegraph and telephone, used commercially after 1844 and 1876, respectively; the practical application of coal gas production in 1798; the use of coal tar products after 1861 and other inventions too numerous to mention. Electricity and steam were more fully developed as sources of power. During this century the dynamo and electric light came into general use following 1870, the Howe printing press was brought out about 1845, the Bessemer process was invented in

1855, and there had appeared even earlier the power loom, the cotton gin, gas lighting, and knitting machines. From 1836 to 1867 the number of patents issued annually in the United States increased from 109 to 13,026. The alterations thus brought about created a multiplicity of hazards which made the old classification ideas almost useless for the purpose of rating.

A few instances may be cited of the difficulties encountered in attempting to classify risks. A group, to be of any value for ratemaking purposes, must consist of a sufficiently large number of risks of essentially the same hazard; but so many and unlike were the elements of fire hazard that it became impossible to obtain sufficient like risks to form a valid class. Then there was the difficulty of classifying a building containing diverse occupancies and congregated hazards. Thirdly, with the increasing impracticability of classification developed an ever-growing "miscellaneous" group of heterogeneous and unrelated risks. Furthermore, since the classifications were comparisons of premium income and losses on different classes of risks, and since both premium rates and losses were almost if not entirely independent variables, differences in results were attributable to either and conclusions reached were often erroneous.

Sufficient has been said to show that classification of this kind had outlived its usefulness as a method of measuring fire hazard. It became instead a weapon of competition, almost prima facie evidence of lack of coöperation. Each company jealously guarded the results shown by its records and considered them as a part of its stock in trade, indicating the classes of risks which had proved profitable in the past and serving as a guide for future policy. The following is an example of ten years' classified experience (1900-1909) of a prominent fire insurance company, showing that with such a variety of risks included in one class and with losses attributable to many different causes, such a record is not available as a basis for rates:

1 Reproduced from E. G. Richards, "Classification-Discrimination," an address before the Insurance Society of New York, 1913.

For some time rates were adapted to conditions through the medium of inspectors. But since the judgments of individual inspectors varied and no evidence of the justice of a rate thus made ever existed, not to mention that the same inspector would often rate the same hazard differently at two different times, individual judgment was superseded by a system based on combined judgment, now designated the "schedule" system. This has been called "guesswork" but it is as far from that as from science. It would be more accurate to call the schedule process one of "estimation." The advantages of the schedule may be briefly summarized as follows:

1. It attempted an analysis of fire hazard into component parts. Thus even the early schedule of 1868 previously mentioned subdivided construction into walls, roof, floors and floor openings and distinguished heating methods, lighting methods, processes, fire extinguishing devices, etc., as elements of hazard.

2. A schedule furnishes a list of the conditions which the inspector can look for, diminishing the probability of overlooking defects or good features. Thus he is enabled to view the risk with a guide-book at hand, so to speak, indicating the places of interest.

3. A combination of judgments was obtained in determining the charges to be made for the particular features of hazard distinguished by the analysis. It might at least be expected that a smaller percentage of error would be present under such conditions than would appear in the single estimate of any individual inspector. This factor became of increasing importance with the progress of schedule making.

Since no compilation of the various early schedules has ever been made it is impossible to trace in detail the development of the