ADVERTISEMENTS:
This article throws light upon the three main types of non-probability sampling used for conducting social research. The types are: 1. Accidental Samples 2. Quota Samples 3. Purposive or Judgement Samples.
Non-Probability Sampling: Type # 1. Accidental Samples:
In accidental sampling, the researcher simply reaches out and picks up the cases that fall to hand, continuing the process till such time as the sample acquires a desired size. The researcher, for example, may take the first 150 persons he meets on any one of the pedestrian paths of a street, who are willing to be interviewed or to provide the kind of information he is seeking.
Similarly, a welfare officer, wanting to make certain generalizations about the factory workers may study the workers of a particular department in the factory where he is working.
ADVERTISEMENTS:
A journalist, wanting to know how the ‘people’ feel about an issue may choose to interview cases conveniently available from different walks of life, e.g., teachers, workers, taxi-drivers, retail shopkeepers, housewives and others who are presumed to reflect public opinion.
In such a sample, there is, of course, no other way of estimating the bias (difference between the average sample value and the true population value) except by doing a parallel study with a probability sample or by undertaking a complete census.
If one uses an accidental sample, one can only hope and pray that he is not being too grossly misled by his sample findings which constitute the basis for estimating the state of the ‘population.’
This does not mean, however, that accidental samples do not have any place in scientific research. This type of sampling, besides being economical and convenient, can also afford a basis for stimulation of insights and working hypotheses.
ADVERTISEMENTS:
Where too much accuracy is not needed or where pre-occupation is with tentative clues to hypothesis-formulation (as in exploratory studies), the procedure of accidental sampling is quite useful.
Non-Probability Sampling: Type # 2. Quota Samples:
One of the most commonly used methods of sampling in marketing researches and election polls is the method of quota sampling. The basic objective of quota sampling is the selection of a sample that is a replica of the ‘population’ with respect to which one would wish to generalize.
Quota sampling, by and large, affords the insurance that diverse elements in the ‘population’ will be included in the sample and that these elements will be taken account of in proportions in which they obtain in the population.
Suppose, we are sampling from a ‘population’ of girl- students comprising the total number of girls studying in co-educational institutions and those studying in institutions for girls only. Suppose, there is a sharp difference between the two sub-populations in respect of the characteristics we wish to measure.
This being so, the results of the survey would almost certainly give out an extremely misleading picture of the total ‘population’, if we did not include an adequate proportion of girls studying in the co-educational institutions.
The quota sampler anticipating such possible differences between sub-groups will try to ensure the inclusion in his sample of enough number of cases from each stratum to afford a reliable picture of the total ‘population.’
Quota sampling usually proceeds in three Steps:
(1) The population is classified in terms of properties known or assumed to be pertinent to the characteristics being studied.
(2) The proportion of the population falling into each class is determined on the basis of the known, assumed or estimated composition of the population in respect of the above.
ADVERTISEMENTS:
(3) Lastly, each observer or interviewer is assigned a quota of respondents. The responsibility of selecting the respondents or subjects is theirs. The quotas are so fixed that the total sample observed or interviewed would reflect the proportions among the classes as determined in the previous step (i.e., 2).
Since the observer or interviewer has the final say in the selection of the subjects, the selection of items/cases depends on the judgement of the interviewer/observer. It often happens in practice, however, that the various components of the sample are not in the same proportion as the corresponding strata in the population.
The interviewers might not have followed their instructions correctly and faithfully. Disproportion between the samples and properties of population are more likely to occur, especially, in respect of the less manifest traits which not have been included as a part of the specifications for interviewer’s/observer’s/quotas.
It should be noted that the inadequacy in the sample can be corrected during the analysis by weighing the different strata in terms of their proportions in the population (involving multiplication or division of results by the appropriate corrective figures).
Thus, the critical requirement in quota sampling is not that the various strata in the population be sampled in their correct proportions; it is rather that there should be enough cases from each stratum to make possible an estimate of the population-value and secondly that we know the proportion of each stratum in the total ‘population.’
If these two conditions are met, the estimates of values for the various strata can be combined to give an estimate of the total population value.
Despite these precautions in the course of selecting the sample and corrections effected during the course of analysis, quota sampling may produce serious errors since it involves, undeniably, an accidental sampling procedure. A part of the sample in any particular class constitutes an accidental sample of the corresponding stratum of the population.
Data for fixing quotas are generally drawn from previous census results and certain contemporary sources. When drastic changes are taking place in society, the estimated quotas may be seriously in error and produce misleading results.
A great deal depends on the interviewer/observer’s judgement in sampling. In general, the observer or interviewer may be assumed to fill his quotas in a manner that suits his own convenience. The interviewer is more likely to select people similar to himself in many respects.
Thus, the stratum-wise samples may not be representative of the strata in the population. The interviewer/observer is seldom so well-informed compared to the researcher (if the two are different) hence left to choose samples by himself, he is likely to introduce two biases, (a) of classification of subject and (b) non- random selection.
The result of quota sampling may often not be seriously in error but whether or not they are, is extremely difficult to establish. We have no assurance that the quota sample will give reliable results within a certain limit of tolerance. And since random sampling, hence the probability principle, is not involved at any stage, the errors of the method cannot be determined by statistical procedures.
Mathematical corrections may be effected if there are disproportions in samples from various strata. But this step depends on our advance knowledge about the true proportions of strata in the ‘population.’
For certain populations, one does not just know this and here the only control that an investigator can avail of is the sampling process itself. There is by now enough experience with quota sampling so that its vulnerability to certain types of biases can be controlled.
Non-Probability Sampling: Type # 3. Purposive or Judgement Samples:
The basic assumption behind judgement or purposive sampling is that with the exercise of good judgement and appropriate strategy one can handpick the ‘right’ cases to be included in the sample and thus develop samples that are satisfactory in relation to one’s research needs.
A common strategy of purposive sampling is to pick cases that are judged to be typical of the population in which one is interested. The selection of elements proceeds under the assumption that errors of judgement in the selection will tend to counter-balance each other.
In other words, when practical considerations pose serious hazards in the way of adopting, probability sampling, the researcher looks for a subgroup which is typical of the ‘population’ as a whole (in respect of some characteristic that he is interested in).
The sub-group is the ‘barometer’ of the ‘population.’ Observations are restricted to this sub-group and conclusions from these observations are generalized to the total ‘population.’ For example, a researcher interested in the effect of rural electrification on traditional social institutions may choose as his sample a particular village where electrification has been effected, say, about a couple of years back.
He makes his observations in this village and believes that what obtains here would also obtain with very little variation in other villages that have also been electrified. There is, however, no demonstrable basis for such a belief, it may ultimately turn out to be ill founded.
Judgement or purposive sampling is very precarious, because much stronger assumptions must be made about the population and sampling procedure than are required while employing probability sampling. Secondly, sampling errors and biases cannot be computed for this type of samples since sampling procedure does not involve probability sampling at any stage.
Data secured on the basis of judgement or purposive samples, at best, point to certain hypotheses but in general they cannot be used as a basis for the statistical testing of hypotheses. Thus, judgement sampling has great utility in exploratory or formulative studies aiming as they do at obtaining insights that would help posing problems or formulating hypotheses for research.