The Dollar and Emerging Market Economies: Channels and Impacts

The paper opens from the observation that, contrary to the conventional textbook trade channel, a strengthening US dollar has been consistently associated with a downturn in real economic activity within emerging market economies. The conventional channel — associated with Mundell (1963) and Fleming (1962) — posits that a local depreciation (equivalently, a dollar appreciation) makes exports relatively cheaper and imports costlier, typically raising net exports and shifting consumers from imported to domestic goods, thereby stimulating real activity. In the authors’ sample the detrended correlations run the other way: GDP and investment correlate -0.76 and -0.79 with the broad real dollar, imports -0.72, real credit -0.43, and, more surprisingly, exports -0.30; the country real interest rate correlates +0.82 with the dollar. Their explanation is that the apparent puzzle reflects the interplay of multiple transmission channels operating over distinct time horizons.

The baseline is a structural panel VAR estimated on quarterly data for sixteen EMEs from 1996 to 2019, identified through a recursive (lower-triangular) structure in which the dollar can have contemporaneous and lagged effects on EME variables but affects other US variables only with a lag. The vector orders US GDP, the US 2-year real interest rate, the VIX, and the broad trade-weighted US real exchange rate ahead of EME GDP, investment, exports, imports, real credit, the country interest rate and the real exchange rate. The stated purpose of this ordering is to isolate dollar shocks that are independent of contemporaneous movements in US GDP, the US 2-year real rate and the VIX; the 2-year real rate is used specifically to account for changes in expectations of future policy beyond the current stance. The countries are Argentina, Brazil, Chile, Colombia, Czech Republic, Hungary, Indonesia, Korea, Malaysia, Mexico, Peru, Philippines, Poland, South Africa, Thailand and Turkey, chosen for well-developed financial markets and at least fifteen years of data history. Estimation uses least squares dummy variables, justified on the grounds that T ≫ N so LSDV has better finite-sample properties than GMM and the Nickell bias is less important; lag length p = 2 is chosen by AIC, with bootstrap error bands.

A 10% positive shock to the dollar leads to contractions in GDP, investment and real credit and an increase in sovereign risk in EMEs, with the contraction in investment and real credit particularly strong. Exports show a negligible effect and even a small initial contraction, which the authors read as consistent with the interaction between the dominant currency paradigm and the working capital channel of trade fluctuations; imports experience a strong contraction. They note that they examine exports and imports in levels rather than net exports or the current account scaled by GDP, acknowledging that scaling by aggregate expenditure would more cleanly isolate expenditure-switching from general equilibrium forces, but arguing that the separate responses of exports and imports remain informative about transmission mechanisms. In placebo tests replacing the dollar with the yen, pound or euro ordered after the dollar in the Cholesky decomposition, they find no statistically significant effects on EME outcomes.

A forecast-error variance decomposition attributes approximately 14% of GDP and 18% of investment fluctuations in EMEs to dollar innovations, rising to around 28% for credit to the non-financial private sector. Dollar innovations explain about 17% of the variance of forecast errors for imports but only about 8% for exports — a gap the authors align with the dominant currency paradigm’s prediction that dollar movements exert negligible effects on exports. The share for the country interest rate is relatively minor, at approximately 11%. These shares build up over the first year: for GDP the contribution rises from 4% at a one-quarter horizon to 13%–14% from eight quarters onward.

No — estimating the panel SVAR separately for peggers and floaters yields overlapping confidence bands and no statistically significant differences across regimes. Floaters exhibit neither the smaller GDP contraction nor the more favourable net-export response that Mundell–Fleming would predict. The authors draw a methodological inference from this: because the exchange-rate classification itself does not generate the heterogeneity, the pooled baseline SVAR is likely to conflate distinct transmission mechanisms rather than reveal them, which motivates the exposure-factor approach. A complementary local-projections exercise re-estimating the exposure model separately within each regime group supports the interpretation that monetary policy credibility, rather than the exchange rate regime, drives the relevant margin of heterogeneity.

The four channels are dominant currency pricing, global value chain integration, balance-sheet vulnerabilities, and lack of monetary policy credibility, each proxied by a country-level exposure index. Dominant currency pricing — the widespread practice of dollar invoicing in trade, which makes trade prices sticky in dollars rather than local currency, mitigating dollar movements’ effects on exports while amplifying their effect on imports — is captured by the share of exports invoiced in dollars. Global value chain integration — as the import content of exports rises, imports and exports move together, mitigating exchange-rate effects on the trade balance — is captured by the import content of exports. Balance-sheet vulnerabilities — in economies with substantial dollar-denominated debt, appreciation expands the domestic-currency value of liabilities relative to assets, weakening balance sheets and tightening financial conditions through both borrower deterioration and higher funding costs and credit risk for local banks relying on cross-border dollar lending — is captured by the share of credit to the non-financial private sector. Monetary policy credibility — countries with less anchored inflation expectations exhibit considerably higher exchange-rate pass-through, potentially leading to procyclical monetary tightening — is measured using average inflation rather than the inflation-expectations anchoring index, because the latter is too sparse to support the balancing procedure; South Africa is excluded from this exercise for the same reason, leaving fifteen EMEs.

Following Iacoviello and Navarro (2019), the authors estimate a nested-form local projection in which an orthogonalized dollar shock is interacted with each of the four re-centred exposure measures, so that the shock coefficient captures the response at median exposure and each interaction coefficient is the marginal effect of moving that exposure from the median to a higher percentile. The dollar shock is the residual from regressing the broad US real exchange rate on its own four lags and on current and lagged US GDP, the US 2-year real rate and the VIX — a procedure the authors describe as analogous to a Cholesky identification ordering the broad dollar last, as in Christiano, Eichenbaum and Evans (2005). Exposures are re-centred by the distance between their 50th and 95th percentiles, interacted with the shock, and then orthogonalized recursively so each additional exposure is orthogonal to those preceding it. Because the underlying exposure data are sparse, the annual panel is first balanced by Multivariate Imputation by Chained Equations and then converted to quarterly frequency via a state-space model. Using changes in the broad dollar index instead of the orthogonalized shocks yields qualitatively similar but quantitatively larger impulse responses, which the authors attribute to endogenous responses to US developments that the orthogonalization is designed to purge.

The monetary policy credibility channel exhibits the fastest transmission dynamics, with peak effects on GDP and investment within the first few quarters: when credibility is low (higher average inflation), GDP contracts by an additional 0.2–0.3 percentage points in the first two quarters after the shock, while investment falls by about 0.5–0.6 percentage points over the same horizon. The authors describe these magnitudes as economically significant, being up to twice as large as the effects estimated at median credibility levels. They align this rapid transmission with the literature emphasising how incomplete credibility can amplify external shocks through quick adjustments in market expectations and risk premia, and characterise the fast-acting but temporary nature of these effects as new evidence on the dynamic role of credibility, extending previous work focused on contemporaneous policy constraints and cyclical responses.

The balance sheet channel demonstrates strong state dependence with particularly pronounced effects in the medium term: the negative impact on GDP is most evident between 5 and 10 quarters after the shock, maintaining around -0.3 percentage points and remaining statistically significant through the 15-quarter horizon, while investment effects intensify from -0.2 to -0.5 percentage points over the same period. The authors present this temporal pattern as extending traditional balance-sheet channel models by revealing that currency mismatches are especially impactful over medium-term horizons with cumulative effects. Where existing work emphasises how dollar appreciation tightens financial conditions through bank lending, their results show that balance-sheet deterioration triggers a prolonged cycle of declining investment and activity that persists well beyond the initial financial tightening — a medium-term amplification they note is consistent with firm-level evidence that the negative effect of local-currency depreciation on investment not only persists but intensifies over time.

When dominant currency pricing exposure is high, imports show an immediate negative response of about -0.2 to -0.3 percentage points in the first year, consistent with direct pass-through effects, while export dynamics are initially statistically insignificant and turn positive only after eight quarters, reaching approximately 0.5 percentage points. The overall GDP response follows a similar evolution, moving from an initially muted impact to a positive effect of 0.2 percentage points after two years, while investment oscillates between no effect and a decline of around -0.3 to -0.5 percentage points throughout the period. The authors interpret the delayed export response as aligning with mechanisms in which financial frictions and export entry costs create gradual trade adjustments following exchange-rate movements, and suggest that similar dynamics operate even for moderate dollar fluctuations when dominant currency pricing is prevalent. They present this sign reversal as extending work focused on contemporaneous trade elasticities, and conclude that dominant currency pricing creates richer dynamic patterns than static models predict, with competitive gains materialising only after one to two years.

The global value chain channel demonstrates no statistically significant effects on macroeconomic variables. Moving global value chain exposure from the 50th to the 95th percentile of supply chain integration, GDP and investment fluctuate around zero with confidence bands that consistently include zero throughout the horizon; trade flows show similarly muted responses, with point estimates suggesting some movement in exports and imports but effects that are not statistically distinguishable from zero at conventional significance levels throughout the 15-quarter horizon. The authors align this pattern of muted responses with findings that trade intensity and market share are the prime determinants of low aggregate exchange-rate pass-through and of the exchange rate disconnect observed in the data.

Three robustness exercises leave the conclusions unchanged. First, to address potential feedback from EMEs to the dollar, the authors substitute the Real Advanced Foreign Economies dollar index for the Real Broad dollar index; the similarity of the resulting impulse responses suggests to them that feedback effects from EMEs on the dollar play a minor role. Second, because the post-Global-Financial-Crisis period saw global risk proxies begin to co-move strongly with the dollar, they re-estimate the baseline local projection restricting data to the pre-GFC period (excluding Indonesia for data availability); estimates for EME variables consistently align with the baseline, which they read as indicating that dollar appreciations remain contractionary even when the underlying drivers of dollar movements differ. Third, since the subgroup analysis showed commodity exporters typically experience more severe contractions, they add commodity prices alongside the dollar in the local projection framework and obtain impulse responses closely aligned with the baseline.

The paper does not identify the fundamental sources of dollar fluctuations, which the authors note the “exchange rate disconnect” literature documents are largely disconnected from macroeconomic fundamentals over short horizons. Their stated objective is narrower: to understand how EMEs react to broad real dollar movements and which country characteristics matter most for transmission. They control for the VIX, the US 2-year real interest rate and US real GDP to mitigate endogeneity concerns, and acknowledge that additional fundamentals could further strengthen identification, resting instead on the pre-GFC robustness analysis to argue that dollar appreciations remain contractionary even absent co-movement between the dollar and risk proxies. They also note that their linear model cannot capture nonlinearities such as export effects from large depreciations via market reallocation.

The authors argue their findings challenge the conventional Mundell–Fleming framework in two ways — the traditional trade channel plays a quantitatively small role compared with financial channels, and dominant currency pricing and global value chain integration fundamentally alter how exchange rates affect trade flows — but they explicitly decline to conclude that EMEs would be better off with fixed or heavily managed exchange rates. They find fewer benefits from exchange rate flexibility than traditional models would suggest, but note evidence that free-floating EMEs are more insulated from international risk spillovers than EMEs with managed floats. Instead they highlight the importance of building strong institutions to reduce financial vulnerabilities, particularly through credible monetary policy frameworks and macroprudential measures to manage vulnerabilities arising from foreign-currency borrowing. They also draw a methodological implication: because channels dominate at different horizons, approaches that average effects over fixed windows risk conflating mechanisms operating on different timescales, and models treating these mechanisms as static miss the dynamics through which dollar shocks propagate.