Hi all,
I am working on a movers design where the key regressor is the age at which an individual moves (see e.g. Chetty & Hendren, 2018). For some observations I observe the exact timing of the move, but for others I only know that the move occurred within a window of length g_i. In the latter case I assign the midpoint of the window as the move timing. Under the assumption that moves are uniformly distributed within the window, this introduces classical measurement error with mean zero and variance g_i²/12, where g_i varies across observations.
Does anyone know of methods or references that exploit this kind of known, observation-specific error variance to correct for the resulting attenuation bias?
Thanks in advance.
I am working on a movers design where the key regressor is the age at which an individual moves (see e.g. Chetty & Hendren, 2018). For some observations I observe the exact timing of the move, but for others I only know that the move occurred within a window of length g_i. In the latter case I assign the midpoint of the window as the move timing. Under the assumption that moves are uniformly distributed within the window, this introduces classical measurement error with mean zero and variance g_i²/12, where g_i varies across observations.
Does anyone know of methods or references that exploit this kind of known, observation-specific error variance to correct for the resulting attenuation bias?
Thanks in advance.
