This paper examines the effect of combat exposure on homelessness in surviving deployed veterans. We assess combat exposure in 50,522 Millennium Cohort Study (MCS) participants by combining self-reports of witnessing death through the 2011–2013 survey cycle with military deployment records since September 2001. We use participants’ 2014–2016 MCS survey cycle reports to assess subsequent homelessness. We accommodate uncertainty surrounding limited data with an information theoretic, semi-parametric Generalized Maximum Entropy model. We estimate, on average, a single combat exposure increases the probability of homelessness by about 0.40 percentage points (27% relative to the mean probability) and multiple combat exposures increase the probability of homelessness by about 0.57 percentage points (38% relative to the mean probability). Our model also sheds light on pre-deployment characteristics associated with combat exposure resilience; estimates indicate veterans with poorer pre-deployment mental or physical health and veterans under the age of 30 years are less resilient to the effects of combat exposure on homelessness. Cost calculations with model estimates suggest combat exposure contributed to 4600 veterans experiencing homelessness and $54 million in related public spending.

本文研究了战斗暴露对幸存的已部署退伍军人中无家可归者的影响。通过将2011-2013年调查周期中目击者的自我报告与2001年9月以来的军事部署记录相结合，我们评估了50522名千年队列研究（MCS）参与者的战斗暴露。我们使用参与者的2014-2016年MCS调查周期报告来评估后续工作无家可归。我们利用信息理论的半参数广义最大熵模型来适应有限数据周围的不确定性。我们估计，一次战斗暴露平均会使无家可归的可能性增加约0.40个百分点（相对于平均概率为27％），而多次战斗暴露会使无家可归的可能性增加约0.57个百分点（相对于平均值为38％）可能性）。我们的模型还揭示了与战斗暴露弹性相关的部署前特征；估计数表明，部署前的心理或身体健康较差的退伍军人以及30岁以下的退伍军人对战斗暴露对无家可归者的影响抵御能力较弱。根据模型估算进行的成本计算表明，战斗暴露使4600名经历了无家可归的退伍军人和相关公共支出达到了5400万美元。