Research

Terms-of-trade shocks are commonly identified as innovations to the ratio of unit-value-based indices of export and import prices. In emerging market and developing economies (EMDEs), such shocks explain less than 10% of aggregate fluctuations in the country SVARs of Schmitt-Grohé and Uribe (2018), and across countries the variance shares implied by their estimated small open economy model bear no relation to the SVAR estimates, a mismatch they call the terms-of-trade disconnect. I revisit both findings. In EMDEs, where most exports and many imports are commodities, the main source of exogenous variation in the terms of trade is the movement of world commodity prices. This movement raises or lowers export and import prices together, and the ratio discards most of this common movement. I measure terms-of-trade shocks instead as the surprises to country-specific commodity export and import price indices, built from world commodity prices with predetermined trade weights. I find that, measured this way, the shocks account for about 30% of aggregate fluctuations, against 13% when they are measured using a ratio of export to import prices. The surprises that move the two indices together account for about 16% on their own. I then extend the model of Schmitt-Grohé and Uribe (2018) by dividing its exportable and importable sectors into commodity and noncommodity parts and estimate it country by country on the SVAR responses to the commodity-price surprises. I find that the model essentially reproduces the median variance share in the SVARs, 29% against 30%, but across countries its shares are unrelated to the SVAR estimates. The disconnect therefore remains and is at least partly attributable to the model’s transmission mechanism, which assigns the largest effects to large commodity importers, whereas the SVARs find them in net commodity exporters.

Two countries’ trade structures are complementary when what one sells is what the other wants to buy. Complementarity in this sense is conventionally measured by indices computed from realized trade, which leaves such measures without a structural interpretation and unable to explain the flows they are built from. This paper derives a complementarity index from model primitives in a multi-sector Eaton and Kortum (2002) economy. The Structural Complementarity Index pairs an exporter’s technology-driven comparative-advantage profile with an importer’s demand composition, each recovered from a single country’s fundamentals and neither from trade within the pair, so the index cannot be recovered from the bilateral flow it is meant to explain. We then relate it to the Trade Complementarity Index of Michaely (1996), one of the most widely used descriptive measures. The two apply different functionals, a sum of products and a sum of minima, to arguments that in general differ. When home trade shares are uniform across sectors and each exporter is small relative to world sector size, those arguments coincide, and the descriptive index bounds the structural one from above, up to a remainder we bound explicitly. The bound binds with equality exactly under complete Ricardian specialization, the deterministic limit of the model. When trade costs and importers’ supplier access do not vary across sectors, the model yields a structural gravity equation with the Structural Complementarity Index as a regressor, which provides a baseline empirical specification for estimating the effect of complementarity on bilateral trade.

Members of a currency union that are commodity-dependent and poor face a policy problem that combines features existing models treat separately. A world commodity price shock affects them asymmetrically, they cannot adjust the exchange rate, and the common central bank responds to union aggregates that may differ from their own conditions. The one instrument left to them, national fiscal policy, rests on a revenue base drawn largely from commodities, so its capacity moves with the very shock it must offset. I build a New Keynesian currency-union model in which members differ in their exposure to a vector of world commodity prices. Part of a member’s output response to a common commodity price shock is imposed by the union: it arises because the common policy rate responds to the conditions of other members. In a stylized union, this imposed part of the output response equals 15% of the response a net commodity exporter would have under a policy rate responding only to its own conditions, 30% for a net importer, and 12% for a member with zero net exposure. Average member welfare falls as the dispersion of exposure across members rises, and about three quarters of the loss comes from sharing a policy rate set for other members’ conditions. In the thirteen CFA franc zone members covered by the data, with WAEMU and CEMAC each modeled as a separate union, the imposed part of the output response is largest for members that trade few commodities. In an application to cash transfers and import subsidies, the preferred instrument depends on the member’s exposure: at the estimated persistence of commodity prices, the net exporter prefers the subsidy, the member with zero net exposure prefers the transfer, and for the net importer the ranking depends on the size of the responses compared.

This paper develops a nonparametric framework for assessing whether a proposed measure of total factor productivity is consistent with cost-minimizing behavior under a common concave production technology. Allowing for observation-specific Hicks-neutral productivity, we derive necessary and sufficient conditions for rationalizability and characterize the identified set of relative productivity levels. We show that the existence of a rationalizing technology depends only on observed input choices, while relative productivity is partially identified by sharp, finite bounds that can be computed using linear programming. We also provide an exact joint test of a candidate productivity series and obtain a graph-theoretic characterization under constant returns to scale. We apply the framework to 364 U.S. manufacturing industries over 1958–2018. Approximately half satisfy the revealed-preference restrictions on input choices, but constant returns to scale is rejected in every industry over the full sample. Among industries satisfying these restrictions, the published NBER–CES productivity index violates the implied marginal bounds in nearly three-quarters of industry-year observations, and no complete industry-level series passes the joint rationalizability test. These findings raise questions about the consistency of conventional productivity measures with the maintained assumptions of cost minimization, Hicks-neutral technical change, and a stable production technology.

The estimation of impulse responses to shocks across multiple countries is usually of interest in the international economics literature. However, international datasets are often characterized by limited time-series coverage, which undermines the precision of country-specific impulse responses. Researchers often resort to panel regressions to exploit cross-sectional variation and shrink standard errors, but panel regressions impose extensive homogeneity restrictions on the data. This paper develops an R package that generalizes and extends the “Pseudo-Panel Local Projections” strategy described in Berg, Curtis, and Mark (2023) for shrinking impulse-response standard errors without relying on the standard panel regression techniques. While Berg, Curtis, and Mark (2023)’s method relies on constrained estimation of small sets of pseudo-panels based on similar-sized local projection coefficients within horizons, this paper proposes an extension where pseudo-panels are formed by searching for similarities both within and across horizons. This extension increases the chances of obtaining pseudo-panels with enough units to leverage cross-sectional variation and further improve accuracy. An application to the analysis of the effects of commodity terms-of-trade shocks shows that the extended pseudo-panel local projections are better at shrinking impulse response standard errors than Berg, Curtis, and Mark (2023)’s original strategy.

I identify the oil supply shock from the 2026 Iran conflict in a structural VAR of the global crude market and forecast its transmission across 69 economies. About 40 percent of the price increase is supply; the rest is precautionary demand. The paper commits country-specific paths, ordered by oil-import dependence, to be tested against outcomes as they are realized.

This paper estimates the causal effect of trade complementarity on bilateral trade, using a Structural Complementarity Index derived from a multi-sector Eaton–Kortum model. The index pairs each exporter's technology with each importer's demand and enters a structural gravity equation. Estimation uses a global panel that retains zero trade flows, with particular attention to trade among African countries.

African countries trade little with one another. Two common explanations are that they produce similar goods and that the infrastructure linking them is poor. This paper combines both explanations in a market access framework, using a Structural Complementarity Index derived from a multi-sector Eaton–Kortum model together with transport-network data, and asks, country by country, whether improving corridors or diversifying exports would raise intra-African trade more.

Using a global VAR, this study analyzes how war-driven spikes in food, oil and fertilizer prices affect commodity terms of trade, inflation, output and real exchange rates in 44 African countries, predicting prolonged adverse impacts and highlighting Africa’s vulnerability to global shocks and its need for greater self-sufficiency.