The majority of our government research is still conducted by telephone. Telephone polling has traditionally been considered more representative of community views compared to online surveys for several reasons:

Sampling bias: Online surveys often suffer from self-selection bias, as respondents choose to participate voluntarily. This can lead to overrepresentation of certain groups who are more likely to respond, such as those with strong opinions or specific interests. Telephone polling uses random sampling techniques to select participants, reducing bias and providing a more accurate reflection of the community.

Data quality: Telephone polling allows for real-time interaction between the pollster and the respondent, ensuring that questions are understood correctly and clarifications can be made if needed.

Demographic representation: Telephone polling enables pollsters to control for demographic factors, such as age, gender, or region, during the sampling process. This helps ensure that the final sample is proportionally representative of the community’s population. In contrast, online surveys often rely on self-reported demographics, which may not be as accurate or representative.

It is worth noting that the landscape of survey research is evolving, and online surveys have gained popularity due to their convenience and cost-effectiveness. Efforts are being made to address the limitations of online surveys, such as improving sampling techniques and employing statistical weighting methods to adjust for demographic biases. However, telephone polling, when properly conducted, still remains a widely recognised method for achieving representative community views.

Given declining use of landlines, these days 70+% of our calls are made to mobile phones. The company supplying our phone sample is a well-known and highly respected supplier of randomised land lines and mobile numbers to the market research industry.

They provide a national directory of more than 9 million mobile numbers and 5.7 million residential numbers, updated monthly and verified regularly for consistency. While some research providers utilise the Electronic White Pages (which is no longer being updated) or random digit dialling (which can have a low level of accuracy with regard to the exact location and mobile vs. land line nature of the number) our testing across many LGAs has found our supplier to offer reliable and representative numbers.
Most importantly, we can access geographically specific and age-specific mobile telephone numbers, to improve reach (in particular) among renters and younger residents.

While many different research sectors have moved to online research over the past decade, the majority of research seeking a random and representative sample of the community is still conducted by phone. This has been primarily driven by:

1. The need for an “opt-out” methodology that reaches the quiet majority rather than just “the usual suspects”. Online and social media surveys (known as “self-selecting” or “opt-in” polling) are often dominated by those with strong or polarised views, rather than accurately reflecting a holistic community perspective. Phone research is still popular with many residents, who are happy to give 10-15 minutes of their time when asked to assist their local Council with feedback.
2. The inability of online panel providers to provide sufficient sample sizes at the LGA level. While panels provide quick and cost-effective research outcomes at the metro/regional, state or national level, they can seldom provide more than 50-100 residents in any given LGA – which is insufficient for the needs of most Councils.
3. Growing access to geographically specific and age specific mobile numbers. This has overcome the issue of declining landline use that once posed a threat to phone research. And it has meant that research conmpanies have excellent access to harder-to-reach groups such as younger residents and renters.

Telephone polling enables pollsters to control for demographic factors, such as age, gender, or region, during the sampling process. This helps ensure that the final sample is proportionally representative of the community’s population. In contrast, online surveys often rely on self-reported demographics, which may not be as accurate or representative.

The optimal sample size is typically derived from a combination of five factors:

1. Random sampling error required for the overall sample.
2. Random sampling error required for any sub-samples (e.g., by age or gender).
3. What looks credible or persuasive to external stakeholders.
4. Cost and time in field.
5. (in smaller LGAs) Available phone sample.

Taverner research conducts Council customer satisfaction surveys for over 40 NSW Councils. Our sample sizes range from:

• 200-250 for small-population (<10,000 population) LGAs
• 300-400 for medium sized LGAs (10-50,000 population)
• 500-600 for larger Councils (>50,000).

It’s important to note that random sampling error varies very little once sample size moves beyond 400, regardless of population. Hence, while we often run larger sample sizes for the bigger Councils, this is generally driven more by factors 2 and 3 (above) than it is by the overall random sampling
The vast majority of NSW regional Councils hence use a sample size of n=400, as this provides a good balance between acceptable RSE, perceived rigour, and cost. Those with populations of less than 5,000 often use smaller sample due to constraints with availability of phone sample/records. And those with larger populations (50,000-plus) are generally driven more by perceptions of rigour than they are by actual RSE.