Association for Public Policy Analysis & Management

During the 2014-2015 academic year, an estimated 1.3 million public school students in the U.S. were counted as homeless, meaning they “lack a fixed, regular, and adequate nighttime residence” (National Center for Education Statistics). Evidence from previous studies indicates that even after controlling for other socio-economic indicators of disadvantage, homelessness is associated with lower academic achievement, absenteeism, and drop out (Cowen, 2017; Miller, 2011). Homeless students are also at greater risk of suffering from chronic health conditions and behavioral problems (Miller, 2011).

Despite theoretical and empirical connections between homelessness and critical educational and social outcomes, we know surprisingly little about homelessness in the U.S. One explanation for this disconnect is that national household surveys are ill-equipped to capture the complex living situation of families experiencing homeless and housing instability (Desmond, 2016). National surveys are likely to underestimate the true prevalence of homelessness making it difficult to assess homelessness as an independent predictor of social and economic outcomes.

Analysis of publically available data in Michigan finds that just over half (55%) of homeless high school students in Michigan graduated in four years. This rate was significantly lower than both the rate for economically disadvantaged students (68%) and the average graduation rate for the state as a whole (80%). One in five homeless students dropped out of high school (20%), a rate six percentage points higher than that of economically disadvantaged students (14%) and 11 percentage points higher than the statewide average of 9%.

This study will further use a multi-level discrete-time event history analysis to estimate the risk of experiencing homelessness among Michigan school students over the course of their schooling. We will leverage administrative data from the state of Michigan that tracks individual students from grades K-12 over the period 2003 to 2016. In addition to person-level predictors, our study will incorporate a rich set of contextual factors such as neighborhood socio-economic characteristics and foreclosures rates. Rates of homelessness over time will be examined for key socio-demographic groups.