How should we measure the return on public investment in a VAR?

A new method of empirically computing the macroeconomic returns to public investment is proposed. Pereira's (2000) technique is modified, and a measure which accounts for both public and private investment costs is suggested. An empirical application to US data shows that differences between alternative ways of measuring rates of return are non-trivial - taking into consideration the full investment effort halves estimated returns when partial public costs only are considered.