There is much truth and singnificance in the old adage in empirical economics, "garbage in equals
garbage out". It does not matter how powerful a given statistical technique or
mathematical tool may be, it cannot overcome problems that fundamentally reside in
the data themselves. Which variables to use? Are the selected variables consistent
with the concept and phenomena they are supposed to capture? What sources should
be used for which variables? What is the reliability of data that are compiled? Are
there measurement errors and outliers? These are questions that an applied
economist or econometrician embarking on a productivity measurement exercise
should always ask themselves. A comprehensive assessment of the alternatives and
data sources and a preliminary data editing exercise should precede any full-scale
empirical analysis involving DBA or econometric estimation SFA models.