Description
Monitoring fieldwork indicators and switching protocols in planned phases, using response propensity models, to control cost and nonresponse bias during data collection.
Instead of fixing one protocol and hoping for a representative result, responsive design plans phases in advance, each with indicators that trigger a switch. During fieldwork, response by subgroup, productivity, budget spent and representativeness are monitored daily, and a response propensity model is refitted from paradata and observations. When a phase reaches capacity (further effort no longer changes estimates), the team switches protocol: first chasing never-contacted or high-propensity cases, then balancing response across domains, and finally subsampling the remaining difficult cases and giving them a new, more expensive protocol such as a larger incentive. Adaptive design pre-specifies different protocols for different groups. The approach needs timely paradata and flexible field operations; the NSFG found 12 per cent of the sample consumed 37 per cent of call attempts, which is the waste it targets.
Attributes
- Discipline
- Research Design & Rigor
- Loss type
- Unmigrated
- Stage
- Post-launch
- Data
- Quantitative
- Effort
- Weeks
Steps
- Plan the phases and the indicators and thresholds that will trigger a switch between them.
- Monitor response by subgroup, productivity, cost and representativeness daily.
- Fit a daily response propensity model from paradata (contact history, timing, prior behaviour).
- When a phase reaches capacity, switch effort: target never-contacted or high-propensity cases, then balance response rates across domains.
- In a final phase, subsample the remaining hard cases and apply a new protocol such as a larger incentive or a different mode.
- Weight for the subsampling and report phase-level results.
Sources
- The SAGE Handbook of Survey Methodology (Wolf et al., 2016)
Also known as
- Responsive design
- Adaptive survey design
- Phase capacity
- Two-phase sampling for nonresponse
- Daily response propensity modelling
Categories
- Research Operations
- Research Design & Rigor