Combined mobile-phone and social-media sampling for web survey on social effects of COVID-19 in Spain

Tax Compliance Determinants: Empirical Evidence From Croatia

“…For this purpose, 2 subsamples were examined -one by mobile phone invitations (from a phone book), and the other through social network invitations (Facebook and Instagram). Targeting two specific sociodemographic groups, these subsamples were combined to achieve the greatest possible representativeness of the sample in relation to population parameters (inspired by Rinken et al, 2020). A total of 862 responses were valid and accepted for this study, representing a response rate of 60.9%.…”

Enhancing estimation methods for integrating probability and nonprobability survey samples with machine‐learning techniques. An application to a Survey on the impact of the COVID‐19 pandemic in Spain

“…This new estimation technique was designed to analyze the data obtained in a web survey on the effects of the COVID‐19 pandemic in Spain (ESPACOV Survey) that used a mixed multiphase sampling design inspired by the responsive approach (Groves & Heeringa, 2006 ). This survey was designed, implemented, and funded by the Institute for Advanced Social Studies at the Spanish National Research Council (IESA‐CSIC) (Rinken et al., 2020 ). There were two editions of the survey: the first one was fielded from April 4 through April 11, 2020 in the fourth week of the lockdown, that in Spain began on March 14.…”

“…An in‐depth explanation and justification of this methodology is provided in Rinken et al. ( 2020 ). The combined use of SMS invitations and an RDD (random digit dialing) sampling frame minimizes sampling and coverage problems for collecting web survey data in countries where unsolicited text messages are allowed and there is a high coverage of smartphones (Kim & Couper, 2021 ) as it is the case in Spain.…”

“…In other cases, where small subpopulations are the group under study and probability sampling is unlikely to meet sample size requirements (Disogra et al., 2011 ; Robbins et al., 2021 ), complementing a small probability sample with a larger nonprobability one may improve the efficiency of the estimates (Wiśniowski et al., 2020 ). Nonprobability samples may also be designed as a mean for compensating for known biases in probability‐based web survey samples by purposely targeting respondent profiles that tend to be underrepresented in these surveys as it is the case in the Survey on the impact of the COVID‐19 pandemic in Spain (ESPACOV) that motivates this paper (Rinken et al., 2020 ). The most rigorous uses of these designs entail the integration of data from both probability and nonprobability samples to produce a single inference that compensates for biases observed in both kind of samples.…”

“…Most surveys that integrate probability and nonprobability samples simply pool the samples and make inference using the Horvitz–Thompson or Hàjek estimator assuming the entire sample is probabilistic (Rinken et al., 2020 ). This method is rarely appropriate because usually nonprobability samples are not distributed proportionally with respect to demographic or other relevant subgroups in the population.…”