Two areas still need further examination in the ecological study of inequality and mortality. First, the evidence for the relationship between income inequality and mortality remains inconclusive, particularly when the analytic unit is small (e.g., county in the U.S.). Second, most previous studies are cross-sectional and are unable to address the recent diverging patterns whereby mortality has decreased and income inequality increased. This study aims to contribute to both topic areas by studying the relationship between inequality and mortality via a spatiotemporal approach that simultaneously considers the spatial structure and the temporal trends of inequality and mortality using county panel data between 1990 and 2010 for the conterminous U.S. Using both spatial panel random effect model and spatial panel fixed effect models, we found that (a) income inequality was not a significant factor for mortality after taking into account the spatiotemporal structure and the most salient factors for mortality (e.g., socioeconomic status); (b) the spatial panel fixed effect model indicated that income inequality was negatively associated with mortality over the time, a relationship mirroring the diverging patterns; and (c) the significant spatial and temporal fixed effects suggested that both dimensions are critical factors in understanding the inequality-mortality relationship in the U.S. Our findings extend support to the argument that income inequality does not affect mortality and suggest that the cross-sectional findings may be a consequence of ignoring the temporal trends.

中文翻译：

---

从不同的角度看：使用空间面板模型检查美国各县的不平等与死亡率之间的关联


在不平等和死亡率的生态研究中，有两个领域仍需要进一步研究。首先，收入不平等与死亡率之间关系的证据仍然没有定论，特别是当分析单位较小时（例如美国的县）。其次，以前的大多数研究都是横向的，无法解决最近死亡率下降和收入不平等加剧的不同模式。本研究旨在通过时空方法研究不平等与死亡率之间的关系，利用 1990 年至 2010 年美国毗邻地区的县面板数据同时考虑不平等和死亡率的空间结构和时间趋势，从而为这两个主题领域做出贡献。面板随机效应模型和空间面板固定效应模型，我们发现（a）考虑到时空结构和最显着的死亡率因素（例如社会经济地位）后，收入不平等不是死亡率的显着因素； (b) 空间面板固定效应模型表明，随着时间的推移，收入不平等与死亡率呈负相关，这种关系反映了不同的模式； (c) 显着的空间和时间固定效应表明，这两个维度都是理解美国不平等与死亡率关系的关键因素。我们的研究结果支持了收入不平等不影响死亡率的论点，并表明横截面研究结果可能是忽视时间趋势的结果。