Dropping variables versus use of proxy variables in linear regression

Dropping variables versus use of proxy variables in linear regression