The Value of Weather Information During the Early Years of the Weather Bureau

Erik D. Craft

University of Richmond

The United States Congress established a national weather organization in 1870. Previous attempts by private individuals and quasi-public organizations had been unsuccessful in selling weather information or raising funds to support a forecasting service. Changes in the seasonality and amount of cargo insurance and cargo shipping rates, as well as an analysis of Great Lakes cargo and hull losses, provide evidence of the value of storm warnings on the Great Lakes. Nearly half of the Great Lakes storm warning display stations were closed during the fall of 1883 on account of appropriations reductions due to an embezzlement scandal in the Army Signal Service. This natural experiment permits the econometric estimation of the value of storm warning display locations on the Great Lakes and the calculation of a rate of return minimum bound for weather service expenditures during the Weather Bureau[[Otilde]]s founding period.

I. Introduction

The United States Congress established a national weather organization in 1870 when it instructed the Secretary of War to organize the collection of meteorological observations and forecasting of storms on the Great Lakes and Atlantic Seaboard. In addition to preparing storm warnings for mariners, the United States Army Signal Service soon expanded forecast services to include agricultural and commercial interests.

The Smithsonian Institution as well as individual entrepreneurs and scientists had attempted unsuccessfully to distribute weather information and forecasts. As early as 1848, Alexander Jones, future general agent of the Associated Press, tried to sell weather reports for 12 and 1/2 cents to 25 cents per city per day [Alter 1949].

The Smithsonian Institution under the guidance of its Secretary Joseph Henry sought to resolve the debates of the 1830s and 1840s regarding the nature and understanding of storms. Averaging 6% of total expenditures per year from 1849 until 1874, the Smithsonian made a greater commitment to meteorological studies than to any other field of study [Fleming 1990]. Most observers for the Smithsonian[[Otilde]]s network were voluntary and reported only monthly. Henry enlisted, however, the support of telegraph companies to transmit some reports gratis to Washington, where a national weather map appeared between 1850 and 1856, depending on the source. The Smithsonian intended Òto give warning of important atmospheric changes to our seaboard,Ó in 1865, but a January fire delayed the project indefinitely.

During the fall of 1869, the Cincinnati Observatory astronomer Cleveland Abbe received daily telegraphic reports from a small group of observers spread throughout the Midwest [Alter 1949]. The Cincinnati Chamber of Commerce supported his network for three months during which time Abbe offered infrequent forecasts. In addition to providing reports and forecasts to his supporters at the Chamber of Commerce, Abbe sought to build a list of paying subscribers. After the three-month trial, Abbe was unable to maintain enough paying subscribers to support a service offering forecasts.

During the same fall, the Milwaukee natural scientist, Increase A. Lapham, presented his Congressman Halbert E. Paine with a list of shipping losses on the Great Lakes for the navigation seasons of 1868 and 1869. The losses were estimated at $3.1 and $4.1 million, respectively. By another account, the value of vessels totally lost (not including cargo or partial losses) on the Great Lakes during the November 16-19, 1869 storm exceeded $420,000. The standard historical interpretation ascribes these Lapham loss figures, as well as Congressman Paine[[Otilde]]s introduction to meteorology in college and his Civil War connections, to the Congressional resolution signed by President Ulysses S. Grant on February 9, 1870.

A reduction in the Army Signal Service budget in fiscal year 1883 on account of an embezzling scandal caused the number storm warning stations on the Great Lakes during the fall of 1883 to fall by nearly one half, thereby offering a natural experiment with which to estimate the value of the storm warning locations. If severe weather forecasts on the Great Lakes were valuable transportation inputs, one would also expect insurance and shipping rates to change after the introduction of the new service. Specifically, the differences between rates during peaceful summer months and relatively stormy fall months should diminish.

Although much amateur interest explores the history of individual ships and wrecks on the Great Lakes, previous historians and economists have avoided an analysis of the changing secular losses and their causes. This paper seeks to fill this gap in the history of the Great Lakes, as well as to provide credible evidence of the high rate of return from one early Federal program.

II. Shipping And Weather Information On The Great Lakes After The Civil War

The Signal Service storm warning network system formally began operation on October 23, 1871 with flag displays at eight ports on the Great Lakes and sixteen ports on the Atlantic Seaboard. The ports on the Great Lakes at which cautionary signals were displayed initially were Buffalo, Chicago, Cleveland, Detroit, Grand Haven, Milwaukee, Oswego, and Toledo. A cautionary signal was flown Òwhenever the winds are expected to be as strong as twenty-five miles per hour, and to continue so for several hours, within a radius of one hundred miles from the station.Ó In the first year of operation, ending June 1, 1872, 354 cautionary signals were flown on both the Great Lakes and Atlantic Seaboard, approximately 70% of which the Signal Service verified as correctly forecast.

The tonnage of United States vessels on the Great Lakes grew modestly in the early 1870s, early 1880s, and the late 1880s. United States vessel steam tonnage increased rapidly in the late 1880s as iron ore shipments expanded in the Lake Superior region. Anecdotal evidence regarding the increasing efficiency of loading and unloading technology and of fueling facilities for steam vessels suggests that shipping capacity on the Great Lakes grew faster than tonnage capacity. Such technological advances, though, are described as occurring predominantly after the mid 1880s. At least in the late 1890s, common judgement held that a steamer could do two and one quarter times the work of a sailing vessel on account of speed. Sailing vessel tonnage statistics from around 1884 to 1895 are misleading, as ÒMany of the vessels classed as sailing vessels have really been transformed into barges, for they are now habitually towed.Ó [U.S. Congress Serial 3679]. By the mid 1990s, actual sailing vessels are said to have disappeared from Lake Superior. The use of iron and steel for hulls does not grow significantly until the early 1880s and comprised only 10% of all steam tonnage in 1885 on the Great Lakes.

In 1883, a decrease in the Signal Service budget caused the Army to close nearly half of its storm warning display stations. The Federal government had reduced appropriations on account of embezzlement charges against the disbursing officer of the Signal Service. This variation in the number of stations providing storm warnings offers an unusual opportunity to estimate their effectiveness and value. The forthcoming analysis assumes that the level of effective weather information corresponds in a consistent manner with the number of locations potentially displaying storm signals on the Great Lakes during the fall months.

Optimistic discussion of the newly formed meteorological network in 1871 and explanations of the signals in 1874 in Barnets Coast Pilot provide evidence that independent transportation authorities had faith in the benefits of the service from its inception. Barnets Coast Pilot was a prominent guide to Great Lakes masters and sailors. It listed, for example, exact sailing directions for standard routes, locations of prominent lights, and navigation rules.

III. Marine Insurance

From at least 1855, a well-organized cartel of marine insurers existed on the Great Lakes, known initially as the Board of Lake Underwriters and subsequently as the Inland Lloyds. Minutes of early annual meetings detail explicit plans to survey and classify ships by seaworthiness annually and to station inspectors at strategic locations so that damaged vessels and cargos would be inspected quickly. The Board of Lake Underwriters encouraged cartel rate adherence with the implicit threat of removal of the above reconnaissance service and ship ratings.

The existence of a cartel makes the interpretation of insurance premium price data difficult. The more stable the Board of Lake Underwriters insurance cartel was during the 1870s, the more confident one can conclude that changes in insurance prices reflect changes in risk. Strengthening the confidence with which one may use such price data is the fact that, unlike the cost curves for many monopolies or cartels, approximately constant marginal costs in the insurance industry with an assumed linear demand curve imply that changes in marginal costs would lead to smaller adjustments in insurance prices. Under these basic assumptions, any reduction in insurance prices underestimates the reduction in real costs.

Cargo insurance data displays extreme sensitivity according to season. Rates drop noticeably from 1862 to the early 1870s and again from the early 1870s to 1879 and 1880. For example, the premiums measured in cents per $100 grain insured from Chicago to Buffalo for October 1 through October 14 were: 1.71 in 1862, 1.575 in 1871, 1.404 in 1872 and 1873, and 0.60 in 1880. The variance of the seasonality of the rates drops from 1871 to 1873. The lower rates during the fall seasons after 1871 suggest that insurers may have realized that shipping was safer during these months after the storm warning system began operation in late October 1871. Unfortunately, newspaper reports during the 1870s and early 1880s record much discussion of the varying effectiveness of the cartel. In 1881, cargo insurers of grain from Chicago to Buffalo began or resumed a pool. An example of the monthly division of risks and premiums is printed in the June 9, 1881 issue of the Chicago Inter Ocean. Unclear rebate policies and infrequently reported price quotes constrain the interpretation, and quotes for insurance on grain shipments during the period October 1 through October 14 in the years 1880 and 1881 give an indication of the volatility of this insurance market. The rates were 0.60 and 1.50, respectively.

Public sources may not represent the market rates accurately. However, existing evidence supports the argument that the storm warnings were valuable. The increase in insurance rates for the summer months during the early 1870s may indicate the waxing of the effectiveness of the insurance cartel. If this were the case, then the reduced seasonal variation in premium rates is evidence that the relative risk of delivery of grain from Chicago to Buffalo during the fall months, when weather factors endangered ships most often, diminished substantially.

IV. Grain Shipments From Chicago To Buffalo

If storm forecasts proved valuable to lake shippers, one expects the price of shipping grain from Chicago to Buffalo to diminish. But it is also true that many other factors could be expected to have lowered shipping costs during the 1870s. Weekly price data and monthly quantity data collected by the Chicago Board of Trade indicate that the absolute level and the percentage premium of the summer-fall price differential falls over time, especially after 1872 and 1873, the first two years of full fall operation of the Army Signal Service[[Otilde]]s storm warning system.

Consider first the hypothetical effect on shipping prices of doubling the capacity of vessels by increasing the number of trips per year from better dock technology. If we assume all dock and operating costs remained constant and storms do not cause shipping damage during the summer months, summer shipping prices would fall by 50% before reaching a new equilibrium. Fall shipping prices would fall by a lower absolute amount because damages due to storms would rise. Improved loading and unloading technology could be expected then to increase the absolute summer-fall price differential, as well as the percentage fall premium. Similar logic holds for increased capacity and efficiency of ships due to size and speed, if safety is unaffected by marginal increases in the size of sailing vessels. The creation of deeper channels permits the use of larger vessels. Other internal improvements such as improved lighting of dangerous passageways and the construction of safe harbors will lower shipping costs, but the affect on the summer-fall price differential is uncertain.

Alternatively, imagine all technology except storm forecasting remaining constant on the Great Lakes. Since severe storms are much rarer during summer months, storm warnings would have a correspondingly greater effect on the fall shipping price. Under reasonable assumptions, storm warnings should lead to greater percentage decreases in fall shipping rates than summer rates. The average fall percentage premium drops from 61% during the period 1868 to 1873 to 33% over the years 1874 to 1889.

V. Analysis of Cargo and Hull Losses

A consistent series of cargo and hull loss data reported by the Chicago Inter Ocean and later in conjunction with the Great Lakes trade publication Marine Record begins in 1873 and ends in 1887. The primary challenge of the analysis is distinguishing the influence of weather information from other factors that could have reduced losses during the same period. Examples of these other factors are: internal improvements, construction of lighthouses, establishment of life-saving stations, and general technological change. Technological change can be manifest in the ships, in the process of recovering damaged ships, and in loading and unloading ships. One expects improvements in steam-power technology to reduce the probability of ships exploding or burning. Deepening harbors, creating safe harbors on coastlines where no safe ports existed, and deepening critical waterways such as the St. Clair Flats between Lakes Huron and Erie are examples of internal improvements. Holding other factors constant, one hypothesizes that losses will diminish as tonnage capacity rises, because vessels can transport a larger share of the commodities during the safer summer season. Measures of the severity of weather and the level of commerce should have a positive effect on shipping losses.

A standard linear regression analysis may not be appropriate. First, as weather information improves and becomes more widely available, the effect of the weather on losses should diminish. In the limit, with perfect weather information and distribution, the coefficient on a severity of weather variable would be zero. Estimation of a semilog specification is the proposed response. This specification gives an untransformed right-hand side variable the following interpretation. Holding all other variables constant, losses are increased or decreased by a constant percentage per unit of the independent variable. In other words, in a year in which losses would be high for other reasons, such as severe weather, a given number of warning stations will be responsible for a larger absolute reduction in losses than in an otherwise normal year. Similarly, in a year in which losses are low due to factors such as a wide network of storm warning stations, inclement weather will be responsible for a small absolute increase in losses. These interpretations follow directly from the econometric model specification.

Regression coefficients indicating the value of Great Lakes storm warning locations remain relatively stable as statistically insignificant variables are removed and the specifications are altered. Different specifications consistently yield statistically significant coefficients implying each additional storm warning location on the Great Lakes decreased losses by over 1% from what they otherwise would have been.

The subtraction of actual losses from the expected losses without storm warnings yields estimates of savings averaging over one million dollars during the early years and rising to around three million dollars near the end of the period.

VI. Social Rate of Return and Conclusion

Craft [1995] presents an overview of the Army Signal Service weather service during its first twenty years with specific emphasis on its storm warning service on the Great Lakes. By 1880, combined United States and Canadian meteorological service budgets stabilized at one million dollars.

In any particular year, the calculated savings in reduced losses to vessels and cargo may be quite plausible, whereas the sum of all the reductions in losses over many years is, in general, not an accurate measure of the social value of reduced transportation costs on the Great Lakes. If vessel owners know or expect weather information to lower the probability of damage to their property or the property they transport, the existence of a storm warning system lowered the expected cost of transportation on the Great Lakes. A lower equilibrium cost of transportation services implies a higher equilibrium quantity. Therefore safer passage on the Great Lakes increases the level of commerce on the Great Lakes.

Therefore, when one seeks to calculate the social rate of return to the creation of a storm warning system on the Great Lakes, avoided losses are the relevant benefits to be summed. These benefits correspond to the increased consumer surplus resulting from lowered transportation costs. After present value and consumer surplus adjustments, I estimate 109% as the minimum bound on the rate of return to weather information including the expenditures of the entire United States Army Signal Service and Canadian Meteorological Service from 1870 through the first half of fiscal year 1888. These estimated rates of return include neither the value of weather information on the Atlantic seaboard nor in any other context in the United States other than transportation on the Great Lakes during this same period.


Alter, J. Cecil, ÒNational Weather Service Origins,Ó Bulletin of the Historical and Philosophical Society of Ohio. VII: No. 3 (1949),139-185.

Chicago Board of Trade, Statement of the Trade and Commerce of Chicago (Chicago: various publishers, 1868-1890). Chicago Daily Inter Ocean, (Chicago: Chicago Inter Ocean Publishing Co., 1874-1885).

Chief Signal Officer, Report of the Chief Signal Officer (Washington: GPO, 1871-1890).

Craft, Erik, The Provision and Value of Weather Information Services in the United States During the Founding Period of Weather Bureau (Ph.D. diss., University of Chicago, Chicago, 1995).

Fleming, James Rodger, Meteorology in America, 1800-1870 (Baltimore: Johns Hopkins University Press, 1990).

Mansfield, John B., History of the Great Lakes (Chicago: J. H. Beers and Co., 1899).

The Marine Record, (Cleveland: A. A. Pomeroy, 1883-1890).

United States Congress, House, Transportation on the Great Lakes of North America, 55th Congress, 2nd Sess. Doc. 277, Serial 3679.

United States Department of Treasury, Annual Statement of the Chief of the Bureau of Statistics on the Commerce and Navigation of the United States (Washington: GPO, 1868-1890).

Whitnah, Donald R., A History of the United States Weather Bureau (Urbana: University of Illinois Press, 1961).