Probability Sampling using Internet Surveys
How can we obtain a probability sample using internet surveys
Often times, online surveys that we see on social media platforms uses a non-probability sampling method. This is different from traditional surveys that use a probability sampling method.
Probability sampling method is a method of selecting a sample wherein each element in the population has a known, non-zero chance of being included in the sample: otherwise it is non-probability sampling. — Almeda et.al
Non-probability sampling method does not use any randomization mechanism. Rather, the sample was chosen subjectively. There is no objective way of assessing the reliability of results without making assumptions that are oftentimes difficult to verify. Inference to a more general population can be made using a probability sampling method but not with a convenience sample.
Despite the main disadvantage of non-probability sampling, this is often times used because it is less expensive and easier to implement. Especially these days that surveys are being done online, often times on social media platforms.
These online surveys have no randomization mechanism used in selecting the sample. The respondents self-select themselves. Most of the time there is no way to control who answers the online questionnaire. The sample consists of elements that are most accessible or easiest to contact. This is an example of a non-probability sampling method using convenience sampling.
But how can we perform a probability sampling method through an online survey that can still enable us to verify the reliability of the results?
In the book of Matthias Schonlau, Ronald D. Fricker, Jr., Marc N. Elliott called Conducting Research Surveys via E-mail and the Web, they discussed the following online probability sampling methodologies:
1. Sample from a closed population list
We have a closed population list if we specify a target a population within organizations that maintain a list of their members and apply a randomization mechanism from this list. We can then send the online forms to their email addresses or contact them to answer the online form. Generally, it is easy to construct a sample frame which then makes it feasible to draw a probability sample.
2. Sample from a general population
If there is no list available for a closed population, a general population may be preferred. Residents of Quezon City is an example of a general population. This is more difficult because oftentimes a list of email addresses or social media accounts with a wide coverage to serve as the sample frame is not usually available.
3. Pre-recruited panel
Lastly, aprerecruited panel may be selected. This is a group of potential respondents who will answer the forms repeatedly. We may recruit a panel of individuals to participate in ongoing surveys. Panelists will receive a number of surveys a month. Sampling is controlled such that panelists are not given more than one survey on a given topic in a three-month period.
But what if a convenience sample will suffice because I don’t need to use inferences beyond the sample at hand?
We can use the following methodologies:
1. Uncontrolled instrument distribution
This method is “uncontrolled” survey because anyone with access of the form can fill them out, as many times as they want. There are ways to try to control this but respondents can easily find a way to answer the survey multiple times. However, this still does not change the fact that the sample constitutes a convenience sample.
2. Systematic sampling of Web site visitors
Systematic sampling is getting the nth person from a sample frame that is ordered in some way. Sampling every nth visitor of a website constitutes a probability sample if one defines the target population as “visitors to this particular Web site.”
3. Volunteer panel
The volunteer panel method relies on assembling a group of individuals who have volunteered to participate in future surveys.
In summary,
To use a probability sampling method or not depends on the researchers’ objective. If we want to generalize the survey insights to a more general population, one must use a probability sampling method. However, if the researchers will just describe the data from the respondents as they are and will not draw these conclusions to a more general population, convenience sampling may be used.
