Rural roads and national welfare : are "local" methods of evaluation satisfactory?

Do 'local' methods of evaluation, such as partial equilibrium analysis at market prices or estimation of shadow prices, provide reliable assessments of a large rural roads programme's social profitability? Consider a small open economy with one city and a rural hinterland, two traded goods, two non-tradables, two specific factors and mobile labour. The wage in some urban employment is regulated. Revenue is raised by a tariff or an excise on the imported good. Theory and model calibration with numerical examples establish that local methods perform rather dismally. With the equivalent variation yielded by general equilibrium analysis as benchmark, the first-order partial equilibrium method grossly underestimates a programme's net benefit. Shadow prices derived on the assumption that all economic activity takes place at the border - a wholesale neglect of space - yield absurd underestimates. Two spatially sensitive variants of shadow pricing fall well short of remedying them.