Congress provided funds in fiscal 1970 to initiate the program of multiple-frame sample surveys with primary emphasis on cattle and hogs. Sound management practice dictates that the total multiple-frame sampling program be put into operation in successive steps over time, and in fiscal 1971 Congress provided additional funds to expand this program to a total of 16 important livestock States. These funds enabled the Agency to begin quarterly multiple-frame hog surveys in the 10 Corn Belt States accounting for nearly 74 percent of the U.S. hog and pig inventory and semiannual multiple-frame cattle surveys in 16 States accounting for about 52 percent of the U.S. cattle inventory. Agency experience with multiple-frame sampling on a research and operational basis indicates that it is a sound methodology to improve the accuracy of estimates for cattle and hogs as well as many other commodities.

LIVESTOCK ESTIMATES

Dr. TRELOGAN. The livestock industry continues to experience a trend toward greater specialization and increased size of operation requiring increasingly larger outlays of commercial credit. This emphasizes the importance of reliable data regarding livestock numbers and prospective production.

The Department is aware of the critical demand for improved accuracy of livestock estimates. This became acutely evident in early 1967 owing to the necessary 2-percent upward revision of cattle estimates, following the 1964 census of agriculture. The revision was within the normal range to be expected using the methods then being employed. However, the industry insists this degree of accuracy is no longer satisfactory and more reliable estimates are needed. This can be accomplished only by newer, more accurate, sampling and estimating procedures. To follow the usual statistical course of obtaining greater accuracy by simply expanding the size of the existing area sample would incur costs too high to be acceptable. The inclusion of 13 additional States in the multiple-frame sampling program constitutes a second stage of program expansion essential for correction of a serious problem in estimating cattle and hog numbers. Current plans are to expand to a fully operational program in all remaining important producing States by the end of fiscal 1975, subject to availability of funds.

MULTIPLE-FRAME SAMPLING TECHNIQUE

The multiple-frame sampling technique grew out of our search for a more economical way to obtain improved accuracy. This technique features probability sampling, utilizing a combination of list and land area survey procedures. The area sampling component is already provided. It consists of the sampling segments included in the current June and December enumerative surveys made possible by appropriations for the agency, during the period 1961 to 1967.

The new element in this method is the list portion which consists of a probability sample drawn from a list of producers that is as complete as is practical to develop in each State. The procedure requires that data be obtained from all farmers selected in the sample. Response is initially invited through direct mailing before resorting to the more costly data collection methods of telephone and personal contacts. The employment of probability sampling insures that the sample is representative of the entire group. This new approach is designed to provide reliable, independent estimates at the lowest cost.

The objective of the proposed multiple-frame probability surveys is to provide cattle and hog estimates with sampling errors of 1 percent or less for the 29-State area. Individual State estimates are expected to have sampling errors of about 3 percent for the most important cattle or hog producing States and 5 to 10 percent for less important States. This compares with present estimates that have sampling errors of about 2 percent at the U.S. level and 5 to 8 percent for major livestock States.

Improvements in the estimates of total inventories will automatically give a proportional reduction in sampling errors for the specie classes. This advantage may, in fact, be more important than gains in precision of total estimates for species of livestock. Much of the controversy following the cattle revisions in 1967 centered on the amount of change between the beef and dairy types and the amount of change in class components of the beef cattle estimate, especially cows and heifers.

IMPROVED DATA COLLECTION, PROCESSING AND DISSEMINATION

CAPABILITIES

An increase of $300,000 is requested to allow the Agency to improve its capability in the transmission, processing and dissemination of data through (1) the acquisition of modern rapid data transmission computer terminals linking 10 State statistical offices with large scale computers, and, (2) the acquisition of facsimile data communication equipment for each of the Agency's 44 State statistical offices to permit rapid transmission of survey data and published data in hard copy form to and from the Washington, D.C., headquarters.

Increased use of complex survey sampling methods such as the multiple-frame livestock survey places more reliance on large scale computers for processing survey data into detailed estimates in order to meet report due dates. Utilizing modern technology, survey data can be moved rapidly from field offices to central processing points, thus reports can be released on a more timely basis. This objective can be achieved most effectively by following a phased approach of placing rapid data transmission computer terminals in about 10 State statistical offices each year over a 4-year period.

This approach includes plans for the systematic training of personnel to acquire skills needed to develop and utilize necessary generalized data processing systems and computer programs.

Rapid data transmission computer terminals would be acquired for each of 10 State statistical offices to be selected from those currently operating multiple-frame livestock surveys. Priority would be given to those States not having access to electronic computers.

The objective to link all conterminous State offices to large scale computers would be consistent with Federal Government directives encouraging the sharing of large scale computers and the use of common computer programs and would also coincide with the ADP plans of the Department of Agriculture.

The increased funds requested would also allow the service to place a facsimile sending-receiving device operable over regular telephone lines in each of the Agency's conterminous State statistical offices and Washington, D.C. to permit rapid transmission of survey data and

published data in hard copy form. Frequently State statistical offices are in the undesirable position of being unable to service the public with data contained in national reports on the day of their release. This condition exists because the estimates released by the Crop Reporting Board, most of them at 3 p.m. Washington time, are not received through current Postal Service methods until 1 or more days later. Much of the data is subject to speculative use and should be available to everyone at about the same time. Certainly farmers should be able to get this information promptly.

RESEARCH ON REMOTE SENSING

An increase of $40,000 is requested to provide electronic data processing services for development of advanced mathematical techniques. needed to determine possible uses of remote sensing data for improving the accuracy and timeliness of crop acreage and yield estimates. Extensive data processing by large scale computers would be employed to run correlations between ground truth data and interpretations of remote sensor readings taken under different conditions.

Remote sensing technology has developed the capacity to collect large volumes of data on land use and crops. However, the signals or imagery recorded by sensors must be subjected to complicated interpretation and processing to achieve useful results. Mathematical formulae for converting remote sensing data must be developed which are economically and statistically feasible. Benefits accruing from use of remote sensing would be greater objectivity of measurements, more geographic detail in estimates, and reduction of reporting burden on farmers.

In closing, I would like to emphasize that for agricultural statistics to play their proper role, they must maintain a reputation for accuracy and objectivity. Doubt about the reliability of an estimate can be as damaging to agricultural interests as a lack of information. A Government supply estimate should, therefore, be the single best estimate that can be derived. We constantly strive to accomplish this goal and the items of increase presented in this budget request would make a significant contribution toward our success in this effort.

This concludes my statement. Members of my staff and I shall be happy to respond to your questions.

Mr. WHITTEN. Dr. Trelogan, I thank you for a very comprehensive statement. This committee has dealt with you and I trust will be dealing with you and your associates for years to come.

We on the committee have a knowledge of the need to be accurate, and being accurate is often not enough. Of course, as you point out, at best you can take only a very small sample without having a tremendous expense. The ability to draw proper conclusions from that sampling comes with years of experience and trial and error, and of necessity must be like that.

In the area in which you operate much depends on foreign policy, whether we keep our commodities moving in world trade. And much depends on other factors such as weather, corn blight, and other things of that sort.

We had testimony earlier that the new cotton crop, from 30 to 50 percent of it, had already been sold in advance of being planted. To