For once, it is not me saying it but Dixon, Rimmer and Waschik in their "Evaluating the effects of local content measures in a CGE model: Eliminating the US Buy America(n) programs" paper:
"Like many countries, the U.S. implements local content policies. Through these policies, the U.S. government attempts to stimulate employment, especially in the manufacturing sector, by favoring U.S. contractors for public sector projects (Buy American regulations) and by insisting that these contractors themselves favor domestic suppliers of inputs such as steel (Buy America regulations). We refer to these policies collectively as Buy America(n). Enforcement of the policies is via complex legalistic processes and often contractors to the U.S. government adopt a cautious approach by favoring U.S. suppliers even when this may not be strictly legally required. In these circumstances, it is not possible to provide a definitive model-based quantification of the effects of Buy America(n). Nevertheless, as demonstrated in this paper, a detailed CGE analysis can give valuable guidance concerning the efficacy of these policies. In an illustrative simulation we find that scrapping Buy-America(n) would reduce U.S. employment in manufacturing but boost employment in the rest of the economy with a net gain of about 300 thousand jobs. Even in the manufacturing sector, there would be many winning industries including those producing machinery and other high-tech products. Employment would increase in 50 out of 51 states and 430 out of 436 congressional districts."
I particularly like the final sentence.