How do sector exports respond to the real effective exchange rate? Mexico — Machines

Panel REER elasticity estimates for Mexico by HS section, a cross-country heterogeneity view for Machines, and an FX-shock scenario table anchored on the published long-run export price elasticity from Hooper, Johnson and Marquez (2000). Methodological caveats are spelled out on each figure: REER is jointly determined with trade flows, so cross-sectional fixed-effects slopes are descriptive only and the simulator uses the literature benchmark for point estimates.

Sector-level pass-through: HS-section elasticities under two-way fixed effects

For each HS section we estimate ln(X_ict) = α_ic + λ_ct + β_c ln(REER_it) + u_ict across the 93-country IMF IFS panel with annual coverage 1995–2025. Double-demeaning sweeps out country-within-section and year-within-section fixed effects, so β_c is identified off within-country deviations from global year-specific trends. Standard errors are classical OLS-on-demeaned residuals and should be read as lower bounds: clustering by country or block bootstrap would typically widen them (Bertrand, Duflo and Mullainathan 2004, QJE). We report these as transparency on the data; the identified elasticities from HJM 2000 are preferred for the simulator below.

Country heterogeneity in the selected sector

A large literature documents that FX pass-through into prices and quantities varies systematically with country characteristics. Campa and Goldberg (2005, REStat) “Exchange Rate Pass-Through into Import Prices” document heterogeneity in short- and long-run pass-through across 23 OECD countries, correlated with trade openness and the macro-policy environment. Gopinath and Itskhoki (2010, AER) “Frequency of Price Adjustment and Pass-Through” show that the frequency of price adjustment by exporters is a structural determinant: low-frequency adjusters pass through less.

FX shock simulator

The scenario table applies the Hooper-Johnson-Marquez (2000) aggregate long-run export elasticity for MEX (G-7 mean as fallback) to a set of real exchange rate shocks spanning a one-decile band. The HJM elasticity is a log-linearised long-run response: %ΔX ≈ β · ln(1 + shock). We apply it uniformly to the selected sector; heterogeneity across sectors is real and non-trivial (Imbs and Mejean 2015 AEJ:Macro estimate Armington sectoral elasticities of substitution in the 2–6 range, generally higher for tradable goods and lower for services and commodities), but mapping their σ to an export supply elasticity requires a demand-side closure we do not impose here.

Recent history: does the overlay match?

The acid test is whether the model tracks reality. We overlay the actual REER path for Mexico against actual Machines exports over the mexico peso real appreciation 2020-2024. Both series are indexed to 100 at the first year with complete coverage. If the HJM-class elasticity is right in magnitude, a large REER move should show a visible opposite-signed export move over three to five years (Bahmani-Oskooee and Ratha 2004 document a typical J- curve lag of this length).

Distribution of country-level slopes: what does the panel look like?

How does the country-specific pass-through for Machines distribute across the 93-country IMF REER panel? The histogram below bins the Figure 2 per-country slopes into 0.5-wide buckets and overlays two benchmarks: the G-7 mean from Hooper-Johnson-Marquez (2000) at β = -0.90, and zero (the null of no FX pass-through). Following Imbs and Mejean (2015), we expect the panel mass to sit above the HJM benchmark (i.e. less negative), reflecting attenuation toward zero from the joint endogeneity of REER and exports within a country-year. A cleanly identified elasticity requires either instrumental variation (Fitzgerald 2012, AER) or structural assumptions; the raw descriptive histogram still informs the size of the attenuation gap that the simulator must correct for.

Lag structure: 1-year vs 3-year REER pass-through by income tier

The J-curve says the export response to a real depreciation builds over time: small or wrong-signed in year 1, larger and correctly signed three to five years out (Bahmani-Oskooee and Ratha 2004). Does this lag structure vary with country income? The dominant-currency-pricing framework (Gopinath 2015, NBER 21646) predicts that low- and middle-income exporters that price in USD face a flatter and more delayed response, because their invoice currency decouples prices from the bilateral nominal rate. We stratify the 93-country IMF REER panel by 2018 GNI per capita (PPP; World Bank income tiers) and run a two-way fixed-effects regression of ln(total exports) on lagged ln(REER), separately for horizons h = 1 and h = 3.

Threshold: is pass-through nonlinear above 15% depreciation?

A widely-documented nonlinearity in REER pass-through says the coefficient is steeper for large moves — above roughly 15% annual depreciation the response is closer to textbook long-run elasticities, while small moves within the threshold band are noisy or near-zero. Bussière (2013, Oxford Bull. Econ. Stat. 75(5): 731–758) and Bussière, López & Tille (2015, JIE 97(1): 21–33) identify this asymmetry in G-7 import and export price pass-through; it survives invoicing-currency controls. The regression below partitions country-year first-differences into three regimes defined by log-change cutoffs: big depreciation (ΔlnREER ≤ ln(0.85), i.e. ≥15% real depreciation), big appreciation (ΔlnREER ≥ ln(1.15)), and small moves in between; then estimates Δln(total exports) on ΔlnREER separately within each regime.

REER volatility and estimated pass-through: where is identification cleanest?

A standard signal-extraction argument (Devereux & Engel 2003, RES 70(4): 765–783) says that the within-country slope of log exports on log REER is most precisely identified in countries whose REER moves a lot relative to other shocks: when the regressor has high sample variance, attenuation toward zero from co-movement with productivity and demand shocks is mechanically smaller. The figure below scatters the per-country slope from Figure 2 against the country’s REER volatility (sample standard deviation of annual Δln REER over 1995–2025). Countries on the right of the plot are where the cross-section reading of the export elasticity is most credible; countries clustered on the left are where the slope is dominated by simultaneity and the HJM-class benchmark in Figure 3 is the better prior.

Policy read, 2025 Article IV context

The IMF Integrated Policy Framework (IMF 2020) treats the REER-export elasticity as a primitive input: a country with a high-magnitude elasticity needs less FX intervention to absorb terms-of-trade shocks because expenditure-switching does the work, while a low-elasticity economy with dominant-currency-priced trade (Gopinath 2015, NBER 21646) may legitimately lean on reserves or capital-flow measures. Three readings follow for Mexico. First, Figure 1’s within-β for Machines at 0.25 must be read as a lower bound; the HJM-anchored simulator in Figure 3 is the central estimate. Second, Figure 2’s openness-elasticity pattern is the Campa-Goldberg (2005) empirical regularity: small-open economies pass through more, and MEX’s position flags whether FX-driven export adjustment is a first-order channel in Article IV scenario analysis. Third, the Figure 5 distribution says the panel mass sits closer to zero than the HJM benchmark, so naive cross-country averages understate the true elasticity; treasuries should use HJM or Imbs- Mejean (2015) sectoral Armington estimates as priors and calibrate the within-sample slope only as a conservative floor.

How treasury and macro advisory use this

• FX scenario planning. Figure 3 translates a board-level REER assumption into a headline export-value impact for Machines, with both a literature-anchored benchmark and a within-sample lower bound so that range-based planning is explicit.
• Country-specific exposure reviews. Figure 2 flags whether the client country sits in a high-elasticity or low-elasticity part of the cross section for this sector; an exporter with a more elastic profile is a higher-beta trade partner for the counterparty’s own FX positioning.
• Ex-post validation. Figure 4 is the history check that every scenario product should ship with: if the model’s implied elasticity cannot make sense of the last decade of actual moves for the selected country, use it only with explicit caveats.

References

• Bahmani-Oskooee, M., & Ratha, A. (2004). “The J-curve: a literature review.” Applied Economics 36(13): 1377–1398.
• Bertrand, M., Duflo, E., & Mullainathan, S. (2004). “How much should we trust difference-in-differences estimates?” Quarterly Journal of Economics 119(1): 249–275.
• Campa, J. M., & Goldberg, L. S. (2005). “Exchange rate pass-through into import prices.” Review of Economics and Statistics 87(4): 679–690.
• Gopinath, G., & Itskhoki, O. (2010). “Frequency of price adjustment and pass-through.” American Economic Review 100(1): 304–336.
• Hooper, P., Johnson, K., & Marquez, J. (2000). “Trade elasticities for the G-7 countries.” Princeton Studies in International Economics No. 87. Princeton University Press.
• Imbs, J., & Mejean, I. (2015). “Elasticity optimism.” American Economic Journal: Macroeconomics 7(3): 43–83.
• Meese, R., & Rogoff, K. (1983). “Empirical exchange rate models of the seventies.” Journal of International Economics 14(1-2): 3–24.