Latin America liberalized in line with standard trade theory (lower tariffs, let uncompetitive import-competing firms shrink, expect labor to move to higher-productivity export sectors), but because new competitive industries failed to rise rapidly, displaced labor did not go to more productive sectors—producing growth-reducing structural change.

If markets in poor economies worked well and governments imposed no artificial impediments to labor mobility, wages and the marginal value product of labor would equalize across sectors; the persistence of large gaps is the puzzle to be explained.

Just as rich countries use patents to let innovators capture monopoly rents and overcome the knowledge-market imperfection, developing-country governments should subsidize the pioneering investor in products new to their own economy because those investments generate learning spillovers (e.g., demonstrating Jamaica can run call centers) that the pioneer cannot internalize.

Because new export industries in developing countries (call centers, canned pineapple, cut flowers) compete globally with thin margins and cannot earn supernormal profits, the pioneer who reveals the country's viability for that industry cannot recoup the social surplus, so without subsidy these high-spillover industries fail to emerge.

In El Salvador, despite extensive liberalization, privatization and stabilization with a sound contractual environment, nothing except garments took off (and only due to a special US trade privilege); investors asked what they would do with $25 million said they would rather put it in Miami, illustrating that the binding constraint was market failures associated with low income, not government failure.

Most gains from further liberalization of trade in goods have been exhausted (wasting political capital at high cost to legitimacy), whereas the gains from liberalizing the temporary movement of people (e.g., expanding temporary work visa schemes) are huge—comparable to where trade in goods stood in the 1950s.

Rodrik largely agrees that in normal circumstances cheap Chinese imports are a welfare-enhancing terms-of-trade gift to US consumers, but argues that at ~9-10% US unemployment, if there is a Keynesian demand-side element, the external deficit being a 'gift' to consumers is only partly true—paralleling Keynes himself turning to protectionism in the 1930s.

The consequences of globalization for developing countries depend on the manner in which they integrate into the global economy: China, India and some Asian countries saw productivity, employment and growth-enhancing structural change, while in Latin America and sub-Saharan Africa labor moved in the wrong direction toward less productive activities, notably informality.

In Malawi the mining sector is so capital-intensive that output per worker matches the US economy as a whole, but mining has limited employment-absorption capacity—you cannot put the whole labor force into it—so workers who cannot enter mining fall back into agriculture and petty trade where productivity is a tiny fraction of mining's.

In many Latin American manufacturing sectors, productivity rose through technological upgrading and capital accumulation rather than employment growth (a 'rationalization' of production), and the labor displaced moved to less productive informal services rather than to more productive parts of the economy.

Following Arthur Lewis's dual-economy model, economic development is the process by which labor moves from traditional low-productivity sectors (typically traditional agriculture) to more modern productive sectors (typically urban industry as the leading modern sector).

The inter-sectoral variation (coefficient of variation) in labor productivity is huge in developing countries—especially the poorest—but shrinks dramatically over the course of development, so in rich countries workers displaced from manufacturing go to sectors with only slightly lower productivity.

Government failures—red tape, corruption, weak property rights and poor rule of law—impose disproportionate costs on the modern parts of the economy (which need inputs, customs, tax compliance and contracts) compared to traditional petty trade, which is why the modern sector remains underdeveloped.

Market failures—coordination failures (complementary inputs, infrastructure, cargo service all needing to exist simultaneously, a chicken-and-egg problem) and learning/cost-discovery externalities (the pioneer investor reveals whether the cost structure supports an industry but bears private costs while later entrants free-ride on the positive signal)—prevent the rise of new productive industries in poor economies.

China opened up 'at the margin'—keeping existing state enterprises and their supports in place while creating new incentives (special economic zones, export subsidies) for new export-oriented investment—thereby turning toward world markets without pulling the rug from under medium-productivity industries and avoiding premature deindustrialization, whereas Latin America deindustrialized prematurely.

China's manufacturing employment share steadily increased and only recently began to shrink, whereas Latin America experienced premature deindustrialization—manufacturing employment began shrinking too early relative to cross-country development benchmarks.

There is no better industrial policy than an undervalued (competitive) currency; Latin American countries (notably Mexico) systematically got this wrong by letting currencies float and opening capital accounts, causing appreciation that kills new investment at the margin in tradable industries, while Asian countries carefully prevented appreciation.

Currency undervaluation is superior to selective industrial policy because it subsidizes all tradable industries across the board (a 10% undervaluation subsidizes them all by 10%) without requiring the government to pick winners, and because micro-level industrial policies cannot undo a 20% currency overvaluation caused by capital inflows.

Latin America gets locked into a pattern of specialization intensive in natural resources and primary products, which generate little employment and a lot of political rent-seeking, which in turn is bad for governance and politics—a self-reinforcing 'lousy dynamic'.

Part of the US's anemic post-recession labor market recovery is structural: workers pulled into housing (carpenters, electricians, drywall applicators) between roughly 1995 and 2005 must now reallocate, and growth will be slow until that structural adjustment is accomplished—though there is likely also a Keynesian demand-side element given how high unemployment rose.