Time-dependency in data collection

posted Dec 18, 2019, 4:32 AM by Bogdan Voicu   [ updated Dec 23, 2019, 2:33 AM ]
Bogdan Voicu, our colleague, published a short research note in the Harmonization Newsletter. The results question the capacity of cross-sectional and longitudinal data to produce comparable estimates if the time of the year when data was collected differs across surveys.

More precisely, depending on the period of the year when one collects data, national-level aggregated estimates may differ, and therefore comparability accords surveys can be affected.

Bogdan employs the Romanian EVS and WVS surveys from 2017 and 2018, which were collected over a 6-months span, and tests whether there is a time-invariance, when controlling for basic predictors. Four types of estimates are used, depending on their level of measurement and complexity of aggregation procedures. They tap for religious belief, life satisfaction, political trust, and postmaterialism. Among them, only religiosity turns sensitive to time of data collection, higher levels being recorded around Christmas and Easter. Consistent seasonality (unreported by existing literature) is therefore observed.

Impact of time of interview on religious belief
controlling for age, gender, education, and employment status,
as well as a dummy for the survey (EVS vs. WVS).
In 2018, Orthodox Easter was in the beginning of April

(please note that the respective issue of the Harmonization Newsletter includes an excellent review of weighting issues in comparative surveys – written by Dominique Joye, Marlène Sapin, and Christof Wolf; a typical debate on mixed-mode challenges, written by Patricia Hadler and Agnès Parent-Thirion with respect to Combining Cognitive Interviews and Web Probing; along with several harmonization examples from various research teams across the World).
Comments