THE TECHNIQUE OF THE SURVEY 



THE SAMPLING PLAN 



To intei-view every hunter and fislierman in the United States 

 would be an extremely difficult and costly undertaking. For- 

 tunately, with modern sampling methods, this is not necessary. 

 Highly accurate results can be obtained for the United States as a 

 whole from a comparatively small sample. In such sampling, it 

 is the distribution of the sample rather than its size that makes it 

 truly representative of the total population. 



Since information was desired from nonlicensed as well as li- 

 censed hunters and fishermen, a sample drawn from State license 

 records would not suffice. To obtain the needed information, a 

 representative sample of households throughout the United States 

 was needed. The method used in selecting this sample for the 

 fishing and hunting survey is known as "area probability sam- 

 pling." It is the method used by the Bureau of the Census and 

 by most national survey organizations and is considered the best 

 sampling procedure yet developed. 



Under the probability sampling method used in this survey, 

 every household in the United States had a known chance of being 

 included in the sample. The households included were selected 

 by statistical formulas rather than by persoiuil choice. 



This selection began with the population records of the 1950 

 census. The 1950 population was divided into segments, and 

 within each segment a large number of small areas ("cells") were 

 assigned, each with specified boundaries but with no specified 

 number of homes. Tlie survey thus became self-adjusting for 

 increases in population since 1950. Where new homes had been 

 built since 1950, increasing an area's population, the sample took 

 note of this increase. With this automatic adjustment, the 

 sample for the whole country indicated a total of 48,389,000 

 households in 1955, containing a total of 118,^66,000 persons 12 

 vears old and older. 



On the average, 1 out of every 2,135 households was included 

 in the sample. This provided a sample of about 20,000 liouse- 

 holds in 1,000 neighborhoods in 48 States. In each of these 

 households all persons 12 years old or older who had fished or 

 liunted in 1955 were eligible for interview. 



Although 1 in 2,1.'55 represented the average chance that a 

 household would fall in the sample, the chances actually varied 

 from as high as 1 in 834 for some groups of homes to as low as 

 1 in 5,579 for others. Tlie reason for sampling different groups 

 of homes at different rates was increased efficiency. When all 

 homes are very much alike, a sample of a few will give a reliable 

 picture of the whole. Where homes differ widely, a larger sample 

 is needed to be sure that the whole range is represented adecjuately. 

 Where fishing and bunting were very popular activities, homes 

 tended to be fairly uniform in that regard. For purposes of the 

 survey, this meant that the lowest sampling rate was needed 

 where fishing and hunting were most common and the highest 

 rate was needed where they were least common. 



Four steps led the interviewer to the door of the sample 

 household : 



1. The United States was divided into 16 parts. 



2. A statistical formula picked some communities from each part. 



3. A statistical formula pick<-d some neighborhoods from each 

 comnnmity. 



4. A statistical formula picked some iiouses from each neigli- 

 borhood. 



"Statistical formula" is a short name for the scientific exercise 

 of impersonal chance. Tiie operation of these formulas, coupled 

 with patterns of human behavior which repeat themselves over 

 and over again, makes it possible to find out what happens in 

 millions of homes bv examining oidv a few thousand. 



46 



