In many of the studies we reviewed, it is common practice to use an observed variable to proxy for a relevant variable that could not be observed.