Abstract: The grass-roots government, as a government agency that people often come into contact with and interact with in daily life, is the " nerve terminal " in the government management system. Among them, the people's trust in the government plays an important role in government management. Therefore, it is of great practical significance to understand the current situation of people's trust in the grass-roots government for the improvement and perfection of grass-roots government management. Based on the previous studies, it is found that the measurement of grass-roots government trust can be divided into direct measurement and indirect measurement. As a social sensitive topic, government trust is difficult to eliminate social desirability bias by direct measurement. At the same time, due to the complexity and abstractness of the concept of grass-roots government trust, it is difficult to avoid differential item functioning of different groups of people by using direct measurement methods. In previous studies, indirect measurement was also used to measure the trust of grass-roots government. However, due to the unclear division of grass-roots government institutions and the diversification of trust concepts, the dimensions of grass-roots government trust investigated by various research institutes are different. In view of the importance of trust in the grass-roots government and the deficiencies in the measurement methods of previous studies, the study conducted an online questionnaire survey and analysis of 647 residents in Zhejiang Province, and used "anchoring vignettes" to correct differential item functioning and social desirability bias in the self-evaluated the grass-roots government trust scale, so as to explore the current status of people's trust in the grass-roots government and its influencing factors. Non-parametric methods and CHOPIT model were used to analyze the research data. The results show that: (1) After using the "anchoring vignettes", the total ratio of "trust" and "complete trust" is 33.54%, which is 36.01% lower than before correction, which verifies the existence of differential item functioning and social desirability bias; (2) After correction, the trust level of people's grass-roots government decreased, but the overall level is still at a medium level; (3) The people's grass-roots government trust level is affected by age, monthly income level, political outlook, and education level. Based on this, in the future government trust measurement or survey, it is necessary to guard against the result exaggeration caused by deviation, and improve the evaluation method of government trust, so as to improve the effectiveness and accuracy of government trust measurement. There are two deficiencies in this study. First of all, in previous studies, the self-assessment questions were all in the form of one, and in this study, the self-assessment questions were also in the form of one. Using a self-assessment question does not correspond to the specific measurements of the three dimensions in the research one by one. Therefore, the CHOPIT model analysis in the study only corrects the self- assessment of the trust of the grass-roots government as a whole. Secondly, there are still many challenges in the application of anchoring vignettes. There is no consensus among researchers on the test method of vignette equivalence.

Key words: grass-roots government, government trust, anchoring vignettes

摘要： 由于政府信任概念的复杂性及其评价的敏感性，传统一般化的政府信任的测量难以避免评价尺度偏差或社会期望偏差。研究对647名浙江省居民进行网络问卷调查并分析，运用虚拟情境锚定法校正偏差，从而探究民众对基层政府信任现状以及影响因素。结果表明：（1）使用“虚拟情境锚定法”后，“信任”和“完全信任”的总比例为33.54%，较校正前降低了36.01%，即验证了评价尺度偏差与社会期望偏差的存在。（2）校正后的民众基层政府信任水平下降，但整体仍处于中等水平；（3）民众的基层政府信任水平受到年龄、月收入水平、政治面貌、教育程度不同程度的影响。

关键词: 基层政府 政府信任 虚拟情境锚定法