The CARES Project

Community priorities & engagement in Pennsylvania

The CARES Survey

Survey Content Areas and Items

The CARES Survey seeks to collect community- and locality-focused data that describes the quality of life in communities across Pennsylvania. This type of information is different than the more traditional socio-economic and demographic data collected by the U.S. Decennial Census and American Community Survey. However, all three surveys complement one another and provide rich data to tell a more complete story about local conditions and life in Pennsylvania. The CARES Survey will collect qualitative and quantitative data on the following topics:

  • Perceptions of local community, local governance, and university engagement
  • Preferences for university-community engagement projects
  • Likelihood of participation in different engagement project activities
  • Barriers and supports to engagement participation
  • Local priority issues
  • Socio-demographic characteristics

The survey will use a combination of close-ended, open-ended, Likert-response scale, and indexed items to gauge stakeholder input. Specific survey sections and items will be continually sourced, created, and refined throughout the development and pilot testing process.

Stakeholder Populations & Sampling

The CARES Survey will routinely collect data from internal (university) and external (non-university) stakeholders. Potential survey respondents will be selected using both random and non-random sampling procedures in order to get data that is as informed and representative of the target populations as possible. Sampling procedures and sample sizes will be determined using The Tailored Design Method (Dillman, Smyth, & Christian, 2014) and in consultation with survey research specialists.

Internal (university) stakeholder samples will include:

  • Faculty members (random)
  • Undergraduate students (random)
  • Graduate students (random)
  • Administrators (non-random)
  • Relevant staff (non-random)

External (non-university) stakeholder samples at the county level

  • General residents in each locality (random)
  • Municipal and county elected leaders (non-random or census)
  • Sector-specific panels – curated sets of public/private leaders in business, educational, environmental, health, and social service sectors/organizations (non-random)

Some survey items will be collected across all stakeholder groups (samples), while others will be limited to a single or only a few groups. Appendix A shows a preliminary stakeholder-survey item matrix detailing which content areas/items will be collected from each stakeholder group.

Survey Contact and Collection

The CARES Survey will be facilitated using a mixed-mode survey strategy to contact potential participants and collect data. Specific survey protocols will be developed according to the principles of The Tailored Design Method (Dillman, Smyth, and Christian, 2014). Survey contact and collection modes for each stakeholder sample will be tested and refined over time. The goal is to find the most cost-effective way to achieve a high response rate given the available contact information.

The contact and data collection protocol for each stakeholder group will be based on the amount of contact information that exists for each group’s sampling frame – the population from which a sample is drawn. The sampling frame does not always cover the total population of interest and the amount of associated contact information can vary. The postal (physical address) and electronic (email address) contact information of university stakeholders should be readily available, but that information will be more limited for non-university stakeholders.

University and select non-university stakeholder samples will be contacted using both postal and electronic mail and given successive opportunities to respond with a paper or online survey instrument. In order to obtain the most representative data from general Pennsylvania residents, the study must contact a sample of resident households through postal mail, as a statewide email-based sample frame does not exist. Residents will be given successive opportunities to respond on paper or online. The success and utility of the CARES Survey and Database depends on collecting reliable data from local residents representative of their larger county populations. Therefore, while this population may be the most costly to survey, it is the most essential.

Data Distribution Schedule & Costs

The total cost of conducting the CARES Survey will depend on:

  • Desired geographic scale of application (e.g. county, regional, or state population);
  • Desired statistical precision of generalizing sample findings to the geographic population;
  • Completed sample size needed to generalize findings to scale with desired precision;
  • Number of surveys distributed to achieve completed sample size given a response rate;
  • Per person cost of the survey mode (contacting persons and collecting responses);
  • Survey schedule (how often data is collected and over what time period)

The cost of conducting the CARES Survey will require a balance of geographic scale and statistical precision. The geographic scale of data collection will need to match the geographic or operational scale of users (e.g. county, county cluster or region, or state levels). For example, data that is representative of the broader state population will not be as useful to individuals and programs operating at a smaller regional or county level. Users of the CARES Database must be able to appropriately generalize the sample results to their own geographic territory and target populations. One way to reduce the cost of survey distribution is to decrease the level of precision by increasing the margin of error in one’s estimation of the population value. Increasing the margin of error reduces the completed sample size necessary to generalize findings from the sample to the larger population. However, as the degree of precision is relaxed to reduce cost, a resulting estimate may be an inappropriate figure from which to make policy and program decisions. Thus, it will be important to determine what practitioners and policymakers view as an acceptable margin of error. For example, if the true population value was 65%, would users be comfortable making decisions based on estimates of 62-68% (a margin of error of ±3%) or 60-70% (±5%)? Four recommended data distribution schedules and their costs are presented in Appendix B for consideration.

Skip to toolbar