Sample Design for Online Surveys

Sample design is the idea of planning a survey sample which represents the larger population, so that we can draw inferences about the wider population. This is the norm for most surveys where it is simply not possible to study the entire population.

Random Sampling is where a survey is conducted with a random selection from the available sample. For example, sending an online survey to a database of email contacts is likely to generate a random sample. The same can be said for panel samples where survey invitations can be sent randomly to the available panelists who broadly represent the national population.
National Representative Sample Design
But random samples are not necessarily the same as a Representative Sample. If we are looking to conduct a nationally representative sample of the UK population, we need to ensure that it reflects the national proportion of age, gender and region. For example, the region of London accounts for 12.2% of the UK population, and therefore should account for the same proportion of our sample to be representative of the London share of the UK population. We should take the same approach for age, gender and region. Researchers commonly speak of National Representative sample, or "Nat Rep" sample which is representative of the population on the criteria of age, gender and region.
General Population Versus National Representative Sample Design
A common approach for reaching panel respondents is not "Nat Rep", but what some researchers call General Population ("Gen Pop") sample. When panel companies are selecting sample to send survey invitations, they ideally should make the sample selections against national representative demographic criteria. However, in reality the completed survey sample will not be representative of the population unless steps are taken to control the sample. (See Quota Samples and Weighting Data, below). In this case, the sample would be "Gen Pop" and not "Nat Rep". For example, a "Gen Pop" survey of 2,000 respondents could have disproportionately higher completes for women and have fewer responses for men over 55 than is the case for the true national population. "Gen Pop" surveys are common where a large sample is required quickly, and where it is not essential to make strict observations about the needs and perceptions of demographic groups across the nation.
Quota Samples
If we apply Quota limits against our sample, we can construct a sample which is broadly representative of the national population. Commonly in polls and consumer surveys, we would use quotas to ensure a representative number of completes against age, gender and region. There are different types of quota but the frequently used type is called Independent Quotas.
Weighted Data
If we have conducted a survey where the data collected is not drawn from an adequately representative national sample, there may be a case to adjust the data where under represented and over represented sample groups are calibrated so that the results represent the true population. Weighting the data tables is often undertaken for polls and other surveys where it is necessary to report on national and regional differences. There are practical limits to what can be achieved by weighting data - if the sample representativeness is very poor, than weighting can compound the inaccuracy of the data rather than fix it. Arguably, it is better to control the sample through quotas than adjust the results after the data has been collected through weighting techniques. Nonetheless, weighting is a useful technique and a widely used option for researchers.
Availability of Nationally Representative Panels
Countries with an advanced level of internet penetration combined with a culture where digital and online communication is widespread among all age groups, tend to have a greater choice of nationally representative research panels. For example, the UK, Western Europe and USA. The picture is very different in Asia where it is often more difficult to achieve good representation of regions, gender and age. Commonly in Asia, women can be under represented and people over 45 and particularly 55+ are under represented.

Surveygoo Books Available on Amazon and iBooks Store

Stacks Image 33853
Visit Amazon to buy | More info on Book
Stacks Image 33858
Visit iBooks Store to download Free

Start Using Our Survey Tool Now!

See Price Plans