Model Specification Errors

Hi everyone,

Let’s say you are modeling a bank’s (just one bank in the industry) customer deposits using a multivariate regression (over a period of time). One of the independent variables you are considering is money supply (M3). A component of Money supply (M3) is bank deposits (for the industry). Would the model be correctly specified if you include money supply as a variable? My confusion is coming from the fact that the M3 forecast is an industy forecast and will implicitly include the bank’s deposits. Howver, determining beta of a stock using regression also has a simialr thing (market returns will include returns on the stock). I’m trying to avoid “forecasting the past” error.

You are using a macro variable (M3 supply) to predict on a micro variable (deposits on a specific bank). However, if M3 supply statistic variable published is lagged several months, then you may encounter the forecasting the past error. I would lag the M3 variable to fit times tho. Check the data source.

On the macro vs micro variable, is it an issue? Would it be better to use the macro variable to forecast industry deposits then use the bank’s market share(in deposits) to get the individual bank deposits?

No, all the way around indeed. It is a sound model. It is like forecasting company sales by using GDP as a regressor, or using aggregate demand (consumption). Totally valid.

Thank you. I really appreciate your input.