This is a very thoughtful book. As a non-economist I appreciate the care and labor that has gone into explaining the discipline to outsiders.
While not dwelling on the suffering caused by the hasty privatization of assets, the radicalization of the Washington consensus, and the deregulation of derivatives, Rodrik argues that these policy prescriptions did not necessarily follow from economic science. He argues that economists had available to them models that should have deflated optimism about these radical policy prescriptions. These models should have made more economists alert to imperfect information and competition, coordination failures, second best perversities, principal-agent misalignments, etc.
These elegant models specify causal mechanisms and can be empirically tested. Rodrik is more than successful in defending the epistemological value of the economists' practice of model building and empirical testing, though one of course can differ about the critical assumptions that should be built into them. Rodrik himself raises tough questions about the limits of empirical testing.
If economists are to be faulted, says Rodrik, it should be for allowing discussion to be dominated by those economists that do not admit that economics has produced several and even competing models and that the best economists struggle to determine which ones fit best in specific situations; the economists who came to dominate discussion also often laced their policy prescriptions with extra-economic value judgments as if these followed from the scientific choosing and testing of models.
Rodrik argues that there are no general, all-encompassing theories in economics that provide universally-valid causal accounts and answer tough moral questions. This not what the best economics provides, and Rodrik argues persuasively that it is exactly because of this that economics has real value as a science. For example, Rodrik explains that economists cannot determine a priori what the employment effects of a rise in the minimum wage would be; it depends, he argues, on a technical analysis of the nature of competition in the industries most impacted, though here Rodrik here does not really address the popular concern that a higher minimum wage would not reduce employment as much as cost bosses and big shareholders a few extra nights on an extravagant vacation or a few extra feet on the length of their yachts.
While Rodrik seems to be following Jon Elster's specification of the hypothetico-deductive method in explaining how models should be tested (like Elster, Rodrik says a successful model should be able not only to account for a specific explanandum via a clearly delineated mechanism, it should also be able to account for other phenomena while the rival, rejected models should be shown to have implications that are not observed), I am left uncertain as to exactly why economists seem to have worked, for the most part, with incorrect models--in other words, with why so many of them exercised poor judgement in their choice of models when making sense of transition economies, economic development and the proliferation of new financial instruments.
Rodrik focuses on three recent developments that have worked their way into the mainstream in spite of their interdisciplinary nature--randomized controlled trials (of which he seems skeptical like recent Nobelist Angus Deaton), behavioral economics (which, it seems to me, is being overhyped for its ability to account for asset bubbles) and the new institutionalism of Acemoglu and Robinson (which has been the subject of stimulating debate).
It does not seem to me that this work vindicates model building as much as it shows the importance and limits of statistical work, the use of psychological experiments to understand human motivation, and the use of comparative historical methods to test ideas. If this is the best work that economics producing, it would suggest that economics is advancing to the extent that it integrates itself with the social sciences and proceeds as an interdisciplinary enterprise.