Does a Strong Dollar Shrink World Trade? Reconciling the Pricing and Financial Channels of Dollar Transmission

Abstract

A weaker currency is supposed to raise exports. For most of world trade it does not, and the reason is now well understood in two separate literatures that have not been reconciled. The first holds that export prices are set in a single dominant currency, so that bilateral depreciation delivers little expenditure switching and what matters is the exporter's currency against the dollar. The second holds that the dollar is a barometer of global credit conditions, so that appreciation contracts the balance-sheet capacity of dollar-funded lenders and reduces the working capital on which trade runs. This article synthesises both literatures, evaluates their empirical claims against each other rather than in isolation, and argues three things. First, the mechanisms are separable on evidence already available, because they carry distinct signatures in destination composition, time profile, and cross-sectional incidence, and the destination test is decisive: the pricing channel predicts no contraction in exports to the United States, while the credit channel predicts contraction there too, and the observed contraction in United States-bound exports is therefore diagnostic. Second, the field's headline estimates are composites of the two mechanisms, weighted by an unmeasured and country-specific mix of invoicing dollarisation and funding dollarisation, which means they are not stable parameters and should not be used as such in forecasting or policy calibration. Third, the policy implications diverge sharply and are currently conflated: local currency settlement initiatives target the pricing channel alone, and in a world where firms may invoice in one currency while funding in another, such initiatives can reduce measured trade exposure to the dollar while leaving balance-sheet vulnerability essentially untouched. The article closes by specifying what evidence would settle the open questions and why existing data fall short of it.