The Reasons for Sampling
There are good reasons why sample surveys are often undertaken in place of censuses. In one way or another, all the reasons typically cited for preferring sampling to census taking have something to do with reducing the cost of getting a given type of information or with increasing the quantity or quality of information received with a given cost.
First, the cost of collecting and processing data is lower the fewer the elementary units to be contacted. This becomes a crucial consideration whenever the number of relevant elementary units is large. Desired information can be gained at a fraction of what a census would cost, provided the sample is properly selected. For example, when the government gathers its monthly unemployment data, it will not do that by questioning the more than 31 million members of the labor force, but by contacting a mere 15,000 households.
Second, a census is sometimes physically impossible (that is, infinitely costly), as when the number of elementary units is infinitely large or when some of them are totally inaccessible. Any process that is expected to operate indefinitely under identical conditions, for example, generates an infinite number of outcomes. Thus, a census can never observe all the outcomes of this process. It can never record all the defective memory chips likely to be produced by a new and ongoing production process or all the effects ever to be produced by a new drug. In such situations, sampling is inevitable.
Thirds, a census is senseless (that is, again infinitely costly) whenever the acquisition of the desired information destroys the elementary units of interest. Consider questions about the lifetime of batteries or about the quality of flash cubes produced by a firm. If every battery or flash cube were tested, all of the output would be used up and the answers to the original questions would be useless.
Fourth, a census is senseless whenever it produces information that comes too late. Consider political opinion polling undertaken prior to an election. A census of many millions of registered voters (or even repeated censuses in the face of rapidly fluctuating preferences) would take too long to yield results; only sampling can provide the desired information in time.
Fifth, for a given cost, sampling can provide more detailed information than a census. Consider how many more questions the government could ask those 15,000 households if it spent the same amount of money, or even a quarter of the money, that it would have spent on a more-than-31-million-person census of the entire labor force.
Sixth, sampling can provide more accurate data than a census. This seems paradoxical but is true because fewer statistical workers are needed, and they can be better trained and more effectively supervised. Thus, for a given cost, higher-quality information is received.
If you want to know more about sampling and types of sampling, please subscribe below to get notified when I post my new article.
Thank you for reading.