调查模式对数据质量的影响：来自尼日利亚的实验证据

Investment Rating - The report does not provide a specific investment rating for the industry. Core Insights - The study investigates the impact of survey mode (in-person vs. phone) on data quality in rural Nigeria, revealing that phone responses differ from in-person responses by 17-18% at the median across various outcome measures [4][13][57]. - The findings indicate that mode effects are significant and economically meaningful, with 70% of outcomes showing positive mode effects, suggesting that respondents tend to provide more affirmative or frequent responses via phone [13][60]. - The research highlights the importance of understanding and mitigating measurement errors in phone surveys to improve data quality in low- and middle-income countries [17][19]. Summary by Sections Introduction - High-quality socioeconomic data is essential for research and development interventions, with phone surveys becoming increasingly important in low- and middle-income countries, especially post-COVID-19 [9][10]. - Concerns about data quality in phone surveys have emerged, necessitating a deeper understanding of mode effects and their implications for economic research [10][11]. Experimental Design and Data - The study utilized a randomized survey experiment involving 937 households in rural Nigeria, comparing responses from both phone and in-person interviews [12][33]. - The experimental design allowed for the identification of mode effects while controlling for various confounding factors [39][45]. Main Estimates of Mode Effects - The results show that responses collected via phone differ significantly from in-person estimates, with the median mode effect at 18% for between-respondent estimates and 17% for within-respondent estimates [57][58]. - Mode effects vary by topic, with the largest effects observed in health-related questions, such as health expenditure incidence and COVID-19 vaccination rates [58][60]. Respondent-Level Mode Effects and Heterogeneity - The within-respondent design revealed substantial heterogeneity in mode effects, with many respondents providing inconsistent answers across modes [64][70]. - More educated respondents were found to be less susceptible to mode effects, indicating that education level plays a role in response consistency [70][71]. Conclusion - The study emphasizes the need for further research to understand the drivers of mode effects and to develop strategies for mitigating measurement errors in phone surveys [20][61].