[论文解读] Surveyor Gender Modifies Average Survey Responses: Evidence from Household Surveys in Four Sub-Saharan African Countries

Noah Haber, Paul Jake Robyn+5 more|arXiv (Cornell University)|Oct 3, 2018

Survey Methodology and Nonresponse参考文献 17被引用 4

一句话总结

本研究调查了在撒哈拉以南非洲四个国家的住户调查中，调查员性别如何影响受访者的回答。基于22,835次访谈的合并数据，研究发现女性调查员能显著提高资产拥有权（比值比1.094）和生殖健康事件（如死产，比值比1.29；流产，比值比1.17）的报告率，表明调查员性别对回答准确性有显著影响，应在调查设计中加以控制。

ABSTRACT

Relatively little is known regarding the influence of surveyor traits on respondent behavior. In this paper, we assess the extent to which survey gender modifies average survey responses in the context of health-focused household surveys. We pool data from four recent health-focused household surveys using both male and female surveyors: Burkina Faso (2014), Cameroon (2012), Central African Republic (2012), and Republic of Congo (2014). In all surveys, surveyors were pre-assigned to households based on an initial household listing. We compare responses given to male and female surveyors across three domains: household characteristics, child mortality and reproductive health. Multivariable regression with enumeration area fixed-effects were used to estimate response differentials. A total of 22,835 household surveys were analyzed. The proportion of interviews conducted by female interviews varied between 9 percent in Central African Republic and 52 percent in Cameroon. Female surveyor gender increased the odds of reporting asset ownership by 9.4% (OR 1.094, 95% CI: 1.024,1.169) and increased the odds of reporting a pregnancy-related event by 25% (OR 1.246 (95% CI: 1.12,1.393). Being interviewed by a woman increased the odds of respondents reporting a stillbirth by 29% (95% CI: 1.118,1.492), and the odds of reporting a miscarriage by 17% (95% CI: 1.072,1.284). Substantial heterogeneity in gender-specific reporting was found across the four countries. We did not find evidence that the gender of the participant modified the effect of surveyor gender for household items. Our results suggest that surveyor gender is highly predictive of survey responses. For health surveys, female surveyors are likely to receive more accurate and consistent responses. More generally, social distance between interviewers and interviewees should be minimized in large scale surveys.

研究动机与目标

• 考察调查员性别是否影响撒哈拉以南非洲国家住户调查中的回答。
• 评估调查员性别差异对健康导向调查中数据质量的影响程度。
• 评估调查员与受访者之间因性别而产生的社会距离是否改变报告行为。
• 确定调查员性别的影响是否因国家或受访者特征而异。

提出的方法

• 对布基纳法索、喀麦隆、中非共和国和刚果共和国四国近期的四项健康导向住户调查数据进行合并分析。
• 调查员根据初始住户名单预先分配至住户，确保按性别实现随机分配。
• 采用具有普查区固定效应的多变量回归模型，估算调查员性别对回答差异的影响。
• 结果按三个领域进行分析：家庭特征、儿童死亡率和生殖健康。
• 计算比值比（ORs）及其95%置信区间，以量化调查员性别对报告的影响。

实验结果

研究问题

• RQ1 调查员性别是否显著改变撒哈拉以南非洲地区住户调查中的平均回答？
• RQ2 调查员性别如何影响对死产和流产等敏感生殖健康结果的报告？
• RQ3 调查员性别的影响在不同国家是否一致，还是因背景而异？
• RQ4 受访者的性别是否调节调查员性别对回答模式的影响？
• RQ5 调查员与受访者之间的社会距离（通过性别中介）在多大程度上影响调查准确性？

主要发现

• 女性调查员使资产拥有权报告的几率提高9.4%（比值比1.094，95%置信区间：1.024–1.169）。
• 由女性调查员访谈时，报告妊娠相关事件的几率高出24.6%（比值比1.246，95%置信区间：1.120–1.393）。
• 由女性调查员访谈时，受访者报告死产的几率高出29%（比值比1.290，95%置信区间：1.118–1.492）。
• 由女性调查员访谈时，报告流产的几率提高17%（比值比1.170，95%置信区间：1.072–1.284）。
• 在四个国家中观察到性别相关报告效应的显著异质性，表明影响具有情境依赖性。
• 未发现受访者性别会调节调查员性别对家庭相关报告的影响。