To add one more thing to what kleinbl00 said below, there's another apt saying that applies here: "All models are wrong, but some are useful". I don't necessarily agree with that assessment for all cases, but the point is that a model is an attempt to put some physical phenomenon into mathematical terms. The level of accuracy depends on several factors. By far the most important factor is the assumptions upon which the model is based. There's no such thing as a model that can account for all variables, conditions, and contingencies. The best we can hope for is that the major contributing factors are accounted for so that whatever question we pose to the model (as one can think of a model as a fancy magic 8 ball) an answer is returned that agrees with reality to within our known error. This is impossible if your fundamental assumptions are incorrect, no matter whether we're talking economics, ecology, or electronics. The author's point, in a nutshell, is that when we view the government's contribution to GDP as a simple credit/debit balance sheet, the model is automatically incorrect, and it's not even useful. In fact, it's down right destructive, given that it produces a false sense of being able to cut modest amounts from certain programs without dramatically affecting the end result (see for example, my home state of MI, who spends more on prisons that higher education, willfully ignorant of the fact that there's almost no ROI for prisons for anyone but the shareholders in the prison business).