Quota sampling methods allow a sample to be carried out within a very short space of time, compared to random sampling. This can be very practical in the real world, if data needs to be obtained very quickly; i.e. the reaction of the general public to a particular event.
Another advantage of quota sampling over random sampling is that it allows the researcher to have control over the sample. For example, a random sampling method is not likely to produce a sample which represents all ethnicity groups in the proportion they exist in the population. Quota sampling allows the researcher to adjust into its sample groups of interest in the proportion they occur in the population, making the comparison of the sample to the population more certain. This is something random sampling doesn’t guarantee
(Tansey,2007).
Disadvantages of Quota Sampling
There are many disadvantages that are associated with quota sampling. Firstly, using quota sampling ultimately limits the statistical inferences that can be made with the data. Because there is no randomness in the selection process of sampling units, this makes it impossible to calculate standard errors for the sample data. This limitation means that results taken from a random sample have more potential for statistical analysis as opposed to quota sampling.
Another disadvantage of quota sampling is that most of the time, the sample will be divided into economic/social class groups. If the judgement of who belongs to which class is left completely up to the interviewer, then there is potential for bias. There is no statistical theory based on dividing a population according to economic/social class, so results can be difficult to interpret. There is also the possibility of differing opinions of class between the interviewers if there are more than one working on the sample, therefore inferences made from the sample will have to be made very carefully (Moser & Stuart, 1953).
We have to also consider the possibility of bias erupting within the quotas at an individual level, leading to misrepresentations of the population. Take for instance; there is a quota for over 60 year olds. If the interviewer only finds people who range from 60 years old to around 65, then there is no representation of people who much older than 65 (Moser & Stuart, 1953).
Another example may be that the interviewer doesn’t like to travel to certain places, therefore only interviews people from a certain area, leading to selection bias in the sample. Random sampling doesn’t have this problem; “Random selection is the only selection mechanism in large-n studies that automatically guarantees the absence of selection bias (Epstein & King, 2002).”
Quota sampling also has the problem of non-response bias, a form of selection bias (selection bias is a non-sampling error). If somebody refuses to be a part of a study, then quota sampling allows the interviewer to go and find the next person who is willing, which results in data that is not wholly representative of the population. The reason for this is, that non-respondents probably have certain characteristics, and because the data obtained from the sample will not represent them at all (it will only be representative of respondents), the data is prone to bias.
Other problems include the assumption that the researcher has full knowledge of the population’s characteristics, which is not always possible. Secondly, although the researcher may classify quotas based on characteristics of the population which are of interest to the researcher, the data at the end may not be representative of other characteristics which are important or should be taken into account when trying to make inferences about the population as a whole (Tansey, 2007).
Quota Sampling in the Real World
Quota sampling was very popular amongst political pollsters until around 1948, which was the year when the results for a presidential election in the US were predicted using quota sampling. The polls predicted that Thomas E.Dewey would win, however, it turned out that his rival Harry Truman defeated him in the real election. The polls over-estimated the number of votes Thomas E.Dewey would receive, and this is was due to the problems surrounding quota sampling mentioned previously (Mendenhall, Lyman Ott, Scheaffer, 2006).
The Washington State Public Opinion Laboratory also carried out its own polls separately before the election using both probability sampling and quota sampling:
The Washington state poll of 1948
Source: Scheaffer, Mendenhall Lll, Lyman Ott : Elementary Survey Sampling, page 15
An article of the Journal of the Royal Statistical Society included a study where a random sample and a quota sample were carried out simultaneously to compare the results after, and see which method gave the most precise result. The conclusion of the study was that there was not any significant difference between the results of the two sampling methods, and that academics have too easily made quota sampling a redundant method to use (Moser & Stuart, 1953).
To conclude, there is never going to be a complete dismissal of a particular sampling method. Quota sampling has no theoretical structure; however practicality outweighs its negatives. If a researcher is looking for, “results derived from theoretically safe sampling methods,” then it is safe to say that quota sampling is out of the question (Moser & Stuart, 1953). If there are time and cost constraints to a researcher, then quota sampling can be convenient.
WORD COUNT: 1235
References
Epstein, L & King, G. (2002). The Rules of Inference. University of Chicago Law Review. 69 (winter), p1-209.
FAO Corporate Document Respository. Sampling in Marketing Research. Available: http://www.fao.org/docrep/w3241e/w3241e08.htm. Last accessed 6th December 2010.
Lyman Ott, Mendenhall, Scheaffer, (2006). Elementary Survey Sampling. 6th ed. Canada: Thomson.
Moser & Stuart. (1953). An Experimental Study of Quota Sampling. Journal of the Royal Statistical Society. 116 (4), p412-413 & p387-340.
Tansey, O. (2007). Process Tracing and Elite Interviewing: A Case for Non-Probability Sampling. PS: Political Science & Politics. 40 (4), p770.