Survey Research in Times of Big Data

La encuesta es la técnica de investigación predominante en la investigación en Ciencias Sociales. Sin embargo, la aparición de otras fuentes de datos como las publicaciones en redes sociales o los datos generados por GPS suponen nuevas oportunidades para la investigación. En este escenario, algunas voces han defendido la idea de que, debido a su menor coste y la velocidad a la que se generan, los big data irán sustituyendo progresivamente a los datos de encuesta. Sin embargo, este optimismo contrasta con los problemas de calidad y accesibilidad que presentan los big data como la fata de cobertura de algunos grupos de la población o el acceso restringido a alguna de estas fuentes. Este artículo, a partir de una revisión profunda de la literatura de los últimos años, explora como la cooperación entre los big data y las encuestas resulta en mejoras significativas de la calidad de los datos y una reducción de los costes. Nowadays, while surveys still dominate the research landscape in social sciences, alternative data sources such as social media posts or GPS data open a whole range of opportunities for researchers. In this scenario, some voices advocate for a progressive substitution of survey data. They anticipate that big data, which is cheaper and faster than surveys, will be enough to answer relevant research questions. However, this optimism contrasts with all the quality and accessibility issues associated with big data such as the lack of coverage or data ownership and restricted accessibility. The aim of this paper is to explore how, nowadays, the combination of big data and surveys results in significant improvements in data quality and survey costs.

Going digital: added value of electronic data collection in 2018 Afghanistan Health Survey

Background Through a nationally representative household survey in Afghanistan, we conducted an operational study in two relatively secure provinces comparing effectiveness of computer-aided personal interviewing (CAPI) with paper-and-pencil interviewing (PAPI). Methods In Panjshir and Parwan provinces, household survey data were collected using paper questionnaires in 15 clusters, and OpenDataKit (ODK) software on electronic tablets in 15 other clusters. Added value was evaluated from three perspectives: efficient implementation, data quality, and acceptability. Efficiency was measured through financial expenditures and time stamped data. Data quality was measured by examining completeness. Acceptability was studied through focus group discussions with survey staff. Results Survey costs were 68% more expensive in CAPI clusters compared to PAPI clusters, due primarily to the upfront one-time investment for survey programming. Enumerators spent significantly less time administering surveys in CAPI cluster households (248 min survey time) compared to PAPI (289 min), for an average savings of 41 min per household (95% CI 25–55). CAPI offered a savings of 87 days for data management over PAPI. Among 49 tracer variables (meaning responses were required from all respondents), small differences were observed between PAPI and CAPI. 2.2% of the cleaned dataset’s tracer data points were missing in CAPI surveys (1216/ 56,073 data points), compared to 3.2% in PAPI surveys (1953/ 60,675 data points). In pre-cleaned datasets, 3.9% of tracer data points were missing in CAPI surveys (2151/ 55,092 data points) compared to 3.2% in PAPI surveys (1924/ 60,113 data points). Enumerators from Panjsher and Parwan preferred CAPI over PAPI due to time savings, user-friendliness, improved data security, and less conspicuity when traveling; however approximately half of enumerators trained from all 34 provinces reported feeling unsafe due to Taliban presence. Community and household respondent skepticism could be resolved by enumerator reassurance. Enumerators shared that in the future, they prefer collecting data using CAPI when possible. Conclusions CAPI offers clear gains in efficiency over PAPI for data collection and management time, although costs are relatively comparable even without the programming investment. However, serious field staff concerns around Taliban threats and general insecurity mean that CAPI should only be conducted in relatively secure areas.

Towards Remote Inspections of FPSO's Using Drones Instrumented with Computer Vision and Hyperspectral Imaging

The Class Society DNV has performed production surveys in enclosed spaces using drones since 2016, demonstrating cost savings and increased personnel safety. The goal is to develop autonomous inspection drones to reduce the need to enter tanks and enable remote inspection. The vision is a drone that can fly by itself, track where it is, and spot rust and cracks, and measure steel thickness. We expect that drone-assisted remote inspection will reduce survey costs for the clients and be a major safety improvement for surveyors. Several drone capabilities are required to enable visual close-up inspection and non-destructive testing in enclosed, GPS-denied, and poorly lit environments. In this study, we report the most recent status from an ongoing research project, including several industry partners. We highlight technical challenges and preliminary results on drone navigation functionalities, computer vision for detection of cracks, and the use of hyperspectral imaging to detect and classify the chemical composition of coatings, rust, and other use cases.

Going Digital: Added Value of Electronic Data Collection In 2018 Afghanistan Health Survey

Background: Through a nationally representative household survey in Afghanistan, we conducted an operational study in two relatively secure provinces comparing effectiveness of computer-aided personal interviewing (CAPI) with paper-and-pencil interviewing (PAPI). Methods: In Panjshir and Parwan provinces, household survey data were collected using paper questionnaires in 15 clusters, and OpenDataKit (ODK) software on electronic tablets in 15 other clusters. Added value was evaluated from three perspectives: efficient implementation, data quality, and acceptability. Efficiency was measured through financial expenditures and time stamped data. Data quality was measured by examining completeness. Acceptability was studied through focus group discussions with survey staff.Results: Survey costs were 68% more expensive in CAPI clusters compared to PAPI clusters, due primarily to the upfront one-time investment for survey programming. Enumerators spent significantly less time administering surveys in CAPI cluster households (248 minutes survey time) compared to PAPI (289 minutes), for an average savings of 41 minutes per household (95% CI: 25 – 55). CAPI offered a savings of 87 days for data management over PAPI.Among 49 tracer variables (meaning responses were required from all respondents), small differences were observed between PAPI and CAPI. 2.2% of the cleaned dataset’s tracer data points were missing in CAPI surveys (1,216/ 56,073 data points), compared to 3.2% in PAPI surveys (1,953/ 60,675 data points). In pre-cleaned datasets, 3.9% of tracer data points were missing in CAPI surveys (2,151/ 55,092 data points) compared to 3.2% in PAPI surveys (1,924/ 60,113 data points). Enumerators from Panjsher and Parwan preferred CAPI over PAPI due to time savings, user-friendliness, improved data security, and less conspicuity when traveling; however approximately half of enumerators trained from all 34 provinces reported feeling unsafe due to Taliban presence. Community and household respondent skepticism could be resolved by enumerator reassurance. Enumerators shared that in the future, they prefer collecting data using CAPI when possible.Conclusions: CAPI offers clear gains in efficiency over PAPI for data collection and management time, although costs are relatively comparable even without the programming investment. However, serious field staff concerns around Taliban threats and general insecurity mean that CAPI should only be conducted in relatively secure areas.

Monitoring abundance of aggregated animals (Florida manatees) using an unmanned aerial system (UAS)

Imperfect detection is an important problem when counting wildlife, but new technologies such as unmanned aerial systems (UAS) can help overcome this obstacle. We used data collected by a UAS and a Bayesian closed capture-mark-recapture model to estimate abundance and distribution while accounting for imperfect detection of aggregated Florida manatees (Trichechus manatus latirostris) at thermal refuges to assess use of current and new warmwater sources in winter. Our UAS hovered for 10 min and recorded 4 K video over sites in Collier County, FL. Open-source software was used to create recapture histories for 10- and 6-min time periods. Mean estimates of probability of detection for 1-min intervals at each canal varied by survey and ranged between 0.05 and 0.92. Overall, detection probability for sites varied between 0.62 and 1.00 across surveys and length of video (6 and 10 min). Abundance varied by survey and location, and estimates indicated that distribution changed over time, with use of the novel source of warmwater increasing over time. The highest cumulative estimate occurred in the coldest winter, 2018 (N = 158, CI 141–190). Methods here reduced survey costs, increased safety and obtained rigorous abundance estimates at aggregation sites previously too difficult to monitor.

How to Increase Survey Response and Manage the Postal and Face-to-Face Field Work of Participant Recruitment during the Covid-19 Pandemic: A Note with Preliminary Evidence from Immigrant-Origin and Native Voters in a German Metropolis in Spring 2021

This note presents preliminary evidence from a postal and face-to-face recruitment field for a telephone survey during the Covid-19 pandemic. Drawn from the city register of Duisburg, a metropolis of 500,000 inhabitants, voters eligible for the September 2021 Bundestag elections are the target population. They are stratified into four groups by onomastic classification: native Germans, Germans of Turkish Descent, Germans with a (Post-) Soviet background and Germans of any other immigrant origin. This note presents ten measures of optimising the recruitment of individuals from these four groups under the constraints of an all-encompassing high-incidence pandemic with distance measures and curfews in place, with closed shops and schools and reduced other care facilities. In addition, the report puts forward statistical analyses of some measures applied in an experimental design and estimates the overall and per-person-recruited costs. Overall, the pandemic forced us to make drastic changes to research design and field organisation and to invest much more time and money than planned. Especially overburdened postal deliveries and restrictive policies for public places made the endeavour most difficult. So far, the ongoing fieldwork has produced a response rate of 8.2 % (only individuals willing to participate in telephone interview) with 11.0 % as a likely expected outcome. One individual recruited to take part in the telephone panel survey costs an estimated 53-71 €.Early-bird incentives produced a positive response (meaning the willingness to partake in the survey) of 5.5 % compared to 4.0 % among those without early-bird incentives. The most effective measure to boost overall positive response among initial non-responders was in-person canvassing with unconditional incentive (21.3 %) instead of postal reminders (9.9 %). Canvassing a smaller group of people in person proved to be as costly as reminding a larger group of people in writing: we estimated costs of 60-61 € per additionally recruited survey participants through either channel.

Evaluating Alternative Flight Plans in Thermal Drone Wildlife Surveys—Simulation Study

The rapidly developing technology of unmanned aerial vehicles (drones) extends to the availability of aerial surveys for wildlife research and management. However, regulations limiting drone operations to visual line of sight (VLOS) seriously affect the design of surveys, as flight paths must be concentrated within small sampling blocks. Such a design is inferior to spatially unrestricted randomized designs available if operations beyond visual line of sight (BVLOS) are allowed. We used computer simulations to assess whether the VLOS rule affects the accuracy and precision of wildlife density estimates derived from drone collected data. We tested two alternative flight plans (VLOS vs. BVLOS) in simulated surveys of low-, medium- and high-density populations of a hypothetical ungulate species with three levels of effort (one to three repetitions). The population density was estimated using the ratio estimate and distance sampling method. The observed differences in the accuracy and precision of estimates from the VLOS and BVLOS surveys were relatively small and negligible. Only in the case of the low-density population (2 ind./100 ha) surveyed once was the VLOS design inferior to BVLOS, delivering biased and less precise estimates. These results show that while the VLOS regulations complicate survey logistics and interfere with random survey design, the quality of derived estimates does not have to be compromised. We advise testing alternative survey variants with the aid of computer simulations to achieve reliable estimates while minimizing survey costs.

Comparing the Ability of Regression Modeling and Bayesian Additive Regression Trees to Predict Costs in a Responsive Survey Design Context

Responsive survey designs rely upon incoming data from the field data collection to optimize cost and quality tradeoffs. In order to make these decisions in real-time, survey managers rely upon monitoring tools that generate proxy indicators for cost and quality. There is a developing literature on proxy indicators for the risk of nonresponse bias. However, there is very little research on proxy indicators for costs and almost none aimed at predicting costs under alternative design strategies. Predictions of survey costs and proxy error indicators can be used to optimize survey designs in real time. Using data from the National Survey of Family Growth, we evaluate alternative modeling strategies aimed at predicting survey costs (specifically, interviewer hours). The models include multilevel regression (with random interviewer effects) and Bayesian Additive Regression Trees (BART).

The European Values Study 2017: On the Way to the Future Using Mixed-Modes

The European Values Study (EVS) was first conducted in 1981 and then repeated in 1990, 1999, 2008, and 2017, with the aim of providing researchers with data to investigate whether European individual and social values are changing and to what degree. The EVS is traditionally carried out as a probability-based face-to-face survey that takes around 1 hour to complete. In recent years, large-scale population surveys such as the EVS have been challenged by decreasing response rates and increasing survey costs. In the light of these challenges, six countries that participated in the last wave of the EVS tested the application of self-administered mixed-modes (Denmark, Finland, Germany, Iceland, the Netherlands, and Switzerland). With the present data brief, we will introduce researchers to the latest wave of the EVS, the implemented mode experiments, and the EVS data releases. In our view, it is pivotal for data use in substantive research to make the reasoning behind design changes and country-specific implementations transparent as well as to highlight new research opportunities.

Using Time Series Models to Understand Survey Costs

Survey costs are an understudied area. However, understanding survey costs is critical for making efficient decisions about cost-error trade-offs, as well as to accurately project future costs. In this article, we examine a measure of survey costs—per interview costs—over time in a repeated cross-sectional survey. We examine both measurement issues and variability in costs. The measurement issues relate to the classification of various costs into the appropriate time period. We explore several issues that make this process of classification difficult. A time series analysis is then utilized to examine the trends and seasonality in per interview costs. Under the assumptions of our model, after removing the trend and seasonality components, the remainder is variation in costs. This approach allows us to treat survey cost estimates in a manner similar to any other survey estimate.