Reeves et al. Respond to "Harnessing Housing Natural Experiments"

We appreciate the interest our study (1), which took advantage of a natural experiment design to better ascertain causality than conventional observational studies could, has generated. In their accompanying commentary, Bentley et al. (2) suggest that misclassification of depression symptoms (e.g., someone with depression reporting that they are not depressed) may explain our findings if misclassification rates changed due to the reduction in the Local Housing Allowance, one component of the United Kingdom's Housing Benefit (HB) for people in the private rental sector. Misclassification can bias estimated effects in 2 ways. First, if misclassification is nondifferential over time, it will dilute associations, tending to underestimate effect sizes. This is likely to result from measurement error. Second, differential misclassification can impact the direction of findings. This can occur, for example, if the reduction in the HB itself leads to changes in how people report symptoms of depression. This is what Bentley et al. suggest may have happened and generated a spurious correlation.