There are four issues in interpreting the results of a
First, while the model is simple, it is no easy task to obtain appropriate measures. Almost always in public sector applications, the measures of program benefits are poor estimates of something that does not easily lend itself to measurement. The measurement error introduced by this estimation process could, in some cases, reverse the conclusions to be drawn; it certainly clouds the issue in most cases.
Second, it is important to compare only roughly comparable programs. The mathematics of the model will not warn you if you are comparing programs that have nothing in common. Feed a computer garbage, it will merrily give garbage back to you. The problem is that the interesting questions frequently involve programs which are not comparable. Morris Hill's article (1968) is particularly instructive in this matter.
Third, the analyst should expect to be frustrated by the inability to obtain quickly the necessary data. It would be nice if you had the leisure to examine programs in detail; usually the decision-maker needs the answer yesterday. Some quantitative analytical tools are designed to use secondary data, data which is already on the shelf, gathered by the census bureau or the demographer's office. Benefit/cost analysis does not lend itself to that approach; each analysis is unique. This makes time for data collection all the more precious.
Finally, the analyst should realize that what is offered here is a general model. It will probably need to be expanded and modified each time to suit the specific setting. This was alluded to in the discussion of discounting; but it is applicable to all elements of the analysis. As long as the user loads and then removes the template disk, the model in memory can be freely modified with no damage to the original template. The modified templates can even be saved (preferably on a different disk and with a different name) for future consultation.
© 1996 A.J.Filipovitch
Revised 11 March 2005