Gambar halaman
PDF
ePub

extent as at times to destroy the usefulness of the inquiry. In fact, the various changes that have already occurred in the dates of the agricultural censuses have left statisticians very much in the dark as to the comparability of certain items. However, changing the date to November 30 would exert a relatively small influence on comparability as compared with the more radical change to April 1, and would involve certain advantages which will be mentioned hereafter.

CONFUSION IN CROP STATISTICS LIKELY TO RESULT FROM CHANGING THE DATE TO APRIL 1

The taking of a census on April 1 is likely to cause great confusion in the enumeration of fall-planted crops because of confusing the acreage harvested the previous year with that planted in the previous year, but harvested in the year enumeration is made. This confusion frequently appears in schedules received by the Department of Agriculture in the early spring. If the census is taken on April 1 in the Southern States, farmers will be supposed to be reporting on wheat, rye, oats, and barley planted a year and a half previously and harvested for nearly a year. The new crop planted the previous fall will be approaching harvest or actually harvested on April 1, and there is almost sure to be a mix up in the figures reported.

CONFUSION IN LIVESTOCK ENUMERATION RESULTING FROM CHANGE OF DATE

Changing the date of the census to April 1 will also result in a very great deal of confusion in the enumeration of livestock.

Census figures of livestock are inventories giving the numbers of the different species by classes as of a given date. Their value as showing the situation of the livestock industry depends upon their adequacy, accuracy, and comparability with previous inventories. Any material change in the date of the inventory tends to decrease the value of the information. Any change in date that destroys the comparability of the various classes of the different species tends further to reduce the value of the inventory.

The Department of Agriculture has been issuing estimates of livestock on farms as of January 1 since 1867, and all statistics having to do with livestock production, marketing, slaughter and movement are organized on a calendar year basis. Since these estimates of the department are the official figures of livestock numbers in intercensal years, it is highly important that this series be continued on its present basis, which gives inventory numbers at the beginning of the calendar year.

In addition to these estimates of inventories the department has undertaken to estimate the amount and value of livestock production by States for calendar years. These estimates include balance sheets by States showing the number on hand at the beginning of the year, the number born and the number brought into the State as items of increase; and the number marketed, the number slaughtered locally and on farms, and death losses as items of decrease. From these balance sheets the value of production is computed. This work requires dependable inventory figures at the beginning and end of the year.

To be of great significance in showing changes in livestock production inventory figures for different species should be by classes. This need was recognized by the Census Bureau before the 1900 enumeration was made, and from 1900 to 1925 the enumeration of livestock has been by classes. Any change in the census date that destroys the comparability of these classes destroys largely the value of the enumeration.

To understand how changes in dates destroy the comparability of census enumerations the case of milk cows may be considered. In 1900 the inquiry covered "cows kept for milk 2 years old and over" on June 1. This would include all cows milked and heifers 2 years old June 1 that were being milked but not heifers over 2 years that had not freshened. The heifers included would be largely 24 to 27 months old.

In 1910 the inquiry covered "cows and heifers kept for milk, born before January 1, 1909." This classified as milk cows all animals over 151⁄2 months of age, and the resulting figures were not at all comparable with the 1900 figures.

In 1920 the census tried to separate cattle into dairy cattle and beef cattle, and the figure of dairy cows 2 years old and over January 1, was considered as the figure representing milk cows. A similar attempt was made in 1925, but the figures of dairy cows 2 years old and over January were not at all comparable to the similar figures for 1920, due to change in the form of inquiry and to changes

85244-28-18

in the cattle situation. Neither the 1920 nor the 1925 figures of all cows (both for milk and not for milk) are comparable with those of 1910, but they are somewhat comparable with those of 1900.

The estimates of the department cover cows and heifers kept for milk 2 years old and over, January 1, and a separate estimate is made of heifers 1 to 2 years old being kept for milk cows. These estimates are based upon the 1925 census figures of " 'cows milked" in 1924 related to total cows and heifers 2 years old and over January 1.

If, however, the census date is changed to April 1, in 1930, no arrangement of questions can get figures comparable to the present estimates. Even if the schedule should be made to include heifers born the previous year (1929) to be kept for milk cows as was done in 1910, this would only include heifers that were under 1 year, that is, heifer calves, January 1. To get a figure covering the department's estimate of heifers 1 to 2 years old, January 1, being kept for milk cows, a separate question would have to be asked as to heifers born in 1928. Figures covering animals included in the present estimate of cows and heifers 2 years old and over for milk, would require a question on such animals born before January 1, 1928. Most persons familiar with livestock reports and livestock records will agree that the inclusion of such dates would lead to endless confusion and give results of little value as showing actual classification. Since there is a slaughter of cows and heifers probably exceeding 2,000,000 head between January 1 and April 1, a large part of which are dairy cows, and an unknown number of deaths, figures of numbers April 1, even if it could include the same classification, would not be comparable with a January 1 estimate.

An

With hogs, an April 1 enumeration, regardless of the manner in which the schedule was arranged, would not get figures at all comparable with January 1 numbers. The 1920 and 1925 schedules of January 1, obtained the number of sows and gilts for breeding, the number of pigs under 6 months, which gives a complete enumeration of the fall pig crops of the previous year; and other hogs over 6 months, which gives an enumeration of the spring pig crop of the previous year, less the slaughter during the last three months of the previous year. April 1 enumeration only gets sows and gilts for spring farrow, which will be somewhat comparable with a similar figure January 1, and April 1 which will be only a part of the pigs born in the spring of 1930 and no indication of total spring crop of that year, and all other hogs which will be made up of a remnant of the spring crop of 1929 and a part of the fall crop of 1929, but in no case an indication of the size of either crop. The 1920 enumeration, if made as of April 1, will not show in any manner what changes in hog production have taken place from 1920 to 1930. There is a total slaughter of over 20,000,000 hogs, and an unknown death loss between January 1 and April 1, which hogs were on farms January 1, and will not be on April 1.

A similar showing can be made for other species of livestock. For example, a January enumeration of sheep will show around 2,500,000 sheep and lambs in Colorado, while an April 1 enumeration would show only about 1,000,000 headthe difference is due to the 1,500,000 lambs on feed January 1 which will be largely marketed by April 1. Nebraska, Iowa, Illinois, and other heavy lambfeeding States would show large decreases for the same reason.

In short, a careful study of the situation leads to the conclusion that an April 1 enumeration of livestock in 1930 will be of little value as showing the trend of livestock production from 1920 to 1930, except with horses and mules, will be of little use to the Department of Agriculture for checking estimates of numbers January 1, will bring further confusion into the series of census enumerations and will be entirely unsatisfactory to all interests connected with livestock production, marketing, and slaughter who have occasion to use census figures for studying conditions in the industry.

In general, agricultural statisticians have recognized the importance of taking the livestock census as nearly as possible at the time of the year when the livestock crop of the year have been largely disposed of and before the birth of a new crop. It has generally been recognized that this date was at or near the first of the year. In planning the world census of agriculture, which it is hoped will be taken in 1930 and concerning which favorable replies had been received from 51 governments by March 1, 1926, it was agreed by the various statisticans concerned that the best date for this world census would be January 1, and that so far as practicable an effort would be made to obtain agreement on that date.

IV. LOSS OF ACCURACY DUE TO LAPSE OF TIME

Without question, the change of the date of the agricultural census to April 1 will seriously affect the accuracy of data reported on for the preceding crop year. In the analysis of data compiled from schedules in the Division of Crop and Livestock Estimates the tendency of farmers to forget some of the details of their farming operations is so common that we have come to use a term for describing this phenomenon which we call "memory bias." Not only has it become necessary to recognize the existence of this bias and to try to make allowance for it in connection with the preparation of estimates, but it has actually been found necessary to measure the bias. We are advancing no theory as to the underlying causes of this bias, nor do we criticise farmers for their apparent lack of memory concerning farm operations.

We make no implication that they are either ignorant, uneducated, careless, or negligent. It is the viewpoint of the staff of the Division of Crop and Livestock Estimates that we must take the reports as they come to us and so treat them that they may yield as accurate and correct a foundation for the preparation of estimates as lies in our power to make. It may be well to present a few examples of memory bias.

In connection with its estimates of the production of cotton the Bureau of Agricultural Economics makes inquiry of its crop reporters as to the average yield per acre of cotton on December 1 of each year and also on March 1 of the succeeding year. The average yield per acre as reported by the very same list of cotton-crop reporters totaling about 6,000 is as follows:

TABLE 1.-Yield of cotton reported to Department of Agriculture for 1925 and 1926 crops, in pounds per acre

[blocks in formation]

It will be noted that there is a material spread between the average of the reports at the earlier and later dates.

In Table 2 is presented a summary of the reports of 837 farmers in Wisconsin, relating to acreages and livestock on their farms on May 15, 1921. In the first column are shown the totals reported by these farmers on that date. A year later they were asked to report these same items, not only as of May 15, 1922, but also for a year earlier, that is, May 15, 1921. For example, 837 identical reporters reported on May 15, 1921, that they had 13,747 acres of corn planted on that date. A year later they reported that they had 14,090 acres on their farms a year preceding the time when the report was made. Attention is directed to the fact that for not a single crop are the totals alike. Examination of these data indicates that some rye was remembered a year later as wheat, and considerable timothy was remembered as clover. The memory bias of one year on dairy cows was only 0.4 of 1 per cent but for dairy heifers it was 22.9 per cent. (Table 2.)

TABLE 2.-Acres of crops and numbers of livestock on farms of 837 identical crop reporters in Wisconsin on May 15, 1921

[blocks in formation]

A study of 784 identical farms for the preceding year, that is, relating to May 15, 1920, and as reported a year later, showed an almost identical situation with respect to memory bias. In Table 3 similar comparisons are given showing historical and current replies for sheep as reported by Utah farmers for three years, and in Table 4 similar comparisons for Utah sheep as between July and January reports. In this case the farmers were asked to report concerning sheep shorn six months or more previously. Much the same situation is noted. (Tables 3 and 4.)

TABLE 3.-Number of sheep reported by Uiah farmers, January 1

COMPARISONS OF CURRENT REPORTS OF INVENTORY NUMBERS OF TOTAL SHEEP AND LAMBS-JANUARY 1 WITH THOSE MADE FROM MEMORY A YEAR LATER

[graphic][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][merged small][subsumed][subsumed][merged small][subsumed][subsumed][merged small][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][subsumed][merged small][merged small][merged small]

TABLE 4.-Number of sheep reported shorn by Utah farmers DIFFERENCES BETWEEN NUMBERS REPORTED SHORN CURRENTLY (JULY) AND HISTORICALLY (FOLLOWING JANUARY)

[blocks in formation]

Current reports used as base. Number shorn not asked in July, 1924. hence only two years' data.

-1.6

The

These are merely a few illustrations of memory bias among farmers. farmer is asked to recall certain acreages and numbers of livestock which were on his farm a year earlier. Undoubtedly the extent of memory bias for a shorter period would be less than for a period of one year, but nevertheless may be considerable. The Bureau of Agricultural Economics once attempted to secure quarterly reports from operators of interior mills and elevators, requesting reports on stocks at the opening and close of the quarter, receipts and shipments during the quarter. Examination of a considerable number of schedules indicated that even the elevator operators, who presumably keep some sort of record of quantities of grain in store at all times, submitted reports relating to the same date which were not consistent.

The general conclusions which have been drawn and which are used constantly by the Bureau of Agricultural Economics in interpreting the reports of crop correspondents may be stated as follows:

1. When there has been but little actual change from one date to the next there is ordinarily relatively little memory bias in reports on most items, such as horses, cows, acreage of corn, wheat, etc. While there are many individuals who do not report the correct figures for the preceding date, the number who report too many is largely offset by the number who report too few. Ordinarily the memory bias on wheat, hay, horses, or cows is less than 1 per cent.

2. When a considerable period of time has elapsed the errors of those remembering too many is much less exactly offset by the errors of those remembering too few. Thus, during 1927 the number of cows in New York State increased slightly, and instead of the usual memory bias of remembering one-half of 1 per cent, too many reporters remembered 1.8 per cent too many. During the same year the number of horses declined and, in reporting for the previous year, crop correspondents remembered 2.1 per cent too few.

3. Memory bias is much greater in items of minor importance, or which change materially from month to month. For example, in reporting poultry on hand a year previous, crop correspondents apparently remember too few by 6 or 8 per cent. In the case of catch crops or others which are grown only occasionally, the memory bias is often quite heavy.

4. Memory bias is much heavier in the case of items for which there is no exact definition. For example, the momory bias on all corn as a whole is probably between 1 and 2 per cent in New York, but the memory bias on corn harvested for green feed or fodder only may exceed 20 per cent. The bias on oats is usually about one-half of 1 per cent, but the memory bias on grains cut green for hay is probably around 15 per cent.

5. On certain items there are certain twists of memory which are so frequently encountered that little reliance can be placed on the memory of crop reporters. Among these are such items as dairy heifers and other young stock, including colts, lambs, and pigs.

One of the most important results of statistical analysis of samples in the Bureau of Agricultural Economics has been the repeated proof of the necessity of comparability as between inquiries. The results of various inquiries have been endangered so often by even slight changes in wording, by changes in time of inquiry, or by changes in the type of farmer to whom the inquiry is sent that

« SebelumnyaLanjutkan »