Reason to add measurement error

Dear Johannes,

Your "A Guide to Specifying Observation Equations for the Estimation of DSGE Models " has mentioned 3 reasons to add measurement error.

Could I ask if there is a 4th reason----no model concept/variable that actually maps data, so just use the observable data as the proxy…like RED paper “Investment shocks and the relative price of investment”, the authors match spread data to MEI shock, using following observable equation, spread_obs=scale parameter* MEI shock+ Measurement error.

Their measurement error seems for the gap between actual data and proxy data? Is this a 4th reason to add Measurement Error?

Many thanks in advance,


It’s a matter of personal interpretation. I would argue that this is a special case of the main reason for using measurement error: the actual object you want to measure is only poorly measured, here in the form of a proxy. I have updated the Guide accordingly. Thanks.