世界银行-记录时间差距：基于智能手机和回忆的时间使用测量方法的比较研究（英）-2024.2-47页_2mb

Summary

This report presents the findings from a large-scale, randomized controlled experiment conducted in Malawi to compare real-time smartphone-based time use data collection with traditional recall-based methods. The key insights are:

1. Methodological Differences:

   • The study employed two arms: a pictorial smartphone app (TimeTracker) for real-time self-reporting and a 24-hour recall diary administered by interviewers, supplemented by a 7-day stylized recall module.
   • The smartphone method revealed higher participation in employment and unpaid work but lower engagement in leisure and self-care compared to the recall arm during specific times of the day.
   • The 24-hour recall diary used 15-minute intervals leading to overestimation of time for activities lasting less than 15 minutes (affecting ~30% of activities) and an underestimation of activities and their distribution throughout the day (especially after 6 pm). Recall fatigue was suggested as a contributing factor.
   • The 7-day recall module, while an improvement over the 24-hour recall for some activities, still resulted in significant overestimation for employment and unpaid work, likely due to further aggregation over a longer period and potential reliance on memory or proxy recall.

2. Gender Gaps:

   • While the magnitude of gender gaps (in time allocation to activities) remained substantial, they were generally narrower in the smartphone arm for most activities, particularly in unpaid work participation among men.
   • Substantial underreporting (especially of nighttime activities) occurred in the recall arm, affecting both men and women, with nuances in patterns based on time of day.

3. Time Use Patterns:

   • The smartphone method captured greater intraday variation, revealing shifts away from productive activities in the afternoon/evening in the recall arm, leading to a loss of information.
   • The smartphone method showed a higher incidence of simultaneous activities (multitasking), providing a more detailed view of time allocation, compared to the recall method which was limited to two activities per 15-minute interval.

4. Implications:

   • Recall-based methods provide an incomplete and potentially misleading representation of time use.
   • The smartphone method's detailed picture allows for better measurement of opportunity costs, multitasking, and time allocation nuances linked to factors like energy access (electricity).
   • The TimeTracker app offers a promising tool that can be adapted for deployment in different contexts using redeployed smartphones.

5. Future Directions:

   • Replication studies in diverse contexts are needed to verify the findings.
   • Guidance is required on reducing the duration needed for smartphone data collection while retaining analytical value.
   • Exploration of methods like within-survey imputation for sub-samples could be considered to potentially reduce costs.

Conclusion: Real-time, smartphone-based time use data collection generates richer, likely more accurate, data compared to traditional recall-based methods, particularly regarding activity participation, intraday variation, multitasking, and electricity effects. While the latter offers benefits, further research is essential before widespread scaling up, especially in low- and middle-income countries.