Quantifying the frequency at which volcanic eruptions of different size occurs is impor- tant for hazard assessment. Volcanic records can be used to estimate the recurrence rate of large-magnitude eruptions (magnitude ≥4), but recording biases that impact data complete- ness complicate analysis. To overcome these biases, we conceptualize the volcanic record as a series of individual and unique time series associated by a common behavior. Thus, we approach issues of completeness on a volcano-by-volcano basis and use a hierarchical Bayes- ian approach to characterize the common frequency-magnitude (f-M) behavior for different groups of volcanoes. We identify variations in the f-M relationship between different volcano types and between different volcanic arcs. By accounting for systematic under-recording in the volcanic record, we also calculate the global recurrence rates for large-magnitude erup- tions during the Holocene, which are similar to previous estimates. However, higher recur- rence rates for smaller-magnitude events are observed, which is a result of our adjustments for data completeness. Quantifying how the f-M relationship varies between different groups of volcanoes provides an opportunity to understand how the tectonic setting in uences f-M behavior, which is important to quantify long-term regional volcanic hazard.