Michigan experienced a significant measles outbreak in 2019 amidst rising rates of non-medical vaccine exemptions (NMEs) and low vaccination coverage compared with the rest of the United States. There is a critical need to better understand the landscape of non-vaccination in Michigan to assess the risk of vaccine-preventable outbreaks in the state, yet there is no agreed-upon best practice for characterizing spatial clustering of non-vaccination, and numerous clustering metrics are available in the statistical, geographic, and epidemiologic literature. We used school-level NME data to characterize the spatiotemporal landscape of vaccine exemptions in Michigan from 2008-2018 using Moran’s I, the Isolation Index, Modified Aggregation Index, and the Theil Index at four spatial scales. We also used thresholds of 5%, 10%, and 20% non-vaccination to assess the bias incurred when aggregating vaccination data. We found that aggregating school-level data to levels commonly used for public reporting can lead to large biases in identifying the number and location of at-risk students, and that different clustering metrics yielded variable interpretations of the non-vaccination landscape in Michigan. This paper shows the importance of choosing clustering metrics with their mechanistic interpretations in mind: be it large- or fine-scale heterogeneity, or between-and-within group contributions to spatial variation.