A Quantitative Microfounded Model for the Integrated Policy Framework

That emerging market central banks with inflation-targeting frameworks keep using FX intervention and capital flow management tools, and keep naming volatile capital flows as a leading concern, while their advanced economy counterparts largely do not (Section 1, p. 1). The authors cite Hoffmann et al. (2019) for the persistence of these practices and Carstens (2019) for the concern being “one often conspicuously absent from the communications of their AE counterparts,” and state their aim as “to shed light on these intriguing asymmetries, quantifying the underlying monetary tradeoffs and analyzing how the use of FXI and CFMs can affect them” (p. 1).

First, an adverse capital outflow shock that depreciates the currency cannot be met with easing; second, sudden stops are both more frequent and more severe in emerging markets. The dilemma is attributed to emerging markets “with less well-anchored inflation expectations, substantial net foreign liabilities, unhedged FX exposures and shallower FX markets,” and is “broadly in line with evidence in Ghosh et al. (2017)” (p. 1). On the first element: the depreciating shock “cannot be met with monetary policy easing, which would otherwise cushion the effects of external tightening on domestic demand,” because inflationary and financial stability concerns both cut against it. The authors ground this in observed behaviour: it is “arguably, why monetary policy responses to adverse capital flow shocks were procyclical, i.e., associated with a tightening, in about half of the emerging market sample in Vegh and Vuletin (2013),” whereas in small open advanced economies “domestic factors appeared to dominate, translating into countercyclical responses” (fn. 1). On the second, citing Bianchi and Mendoza (2020), sudden stop episodes “are not only more frequent in EMEs, but also significantly more severe than in most AEs” (p. 1).

A full New Keynesian structure with two microfounded financial frictions, price and wage rigidity, indexation and pass-through, and a two-country rather than small-open-economy setup. The authors’ claim to contribution is explicitly about joint modelling: existing analyses “typically rely on stylized setups, which do not jointly feature the financial, nominal and real frictions that are embedded into our model,” and while stylized models are “very useful for isolating and explaining particular mechanisms, their ability to capture key aspects of emerging market economies relevant for monetary policy, such as pass-through of exchange rate shocks to inflation and economic activity, is limited”; such models also “provide little guidance on the relative importance of the underlying market imperfections” (pp. 1-2). The first friction, from Gabaix and Maggiori (2015), “captures the limited risk-bearing capacity of agents trading in the FX market and gives rise to inefficient fluctuations in the uncovered interest rate parity (UIP) risk premium”; the second, following Arellano-Mendoza, Chari et al., Mendoza, Bianchi and Chang, “is an occasionally binding external debt limit, which can trigger sharp increases in credit spreads, current account reversals and deep contractions in economic activity” (p. 2). Weakly anchored expectations are a proxy, not a literal learning mechanism: “We proxy for how medium-term inflation expectations may be destabilized by exchange rate changes through allowing the latter to potentially exert large and persistent effects on wage and price inflation via indexation mechanisms and high passthrough of exchange rate changes to import prices” (p. 2). The Kimball aggregator serves two purposes – it generates the asymmetric inflation response, as in Harding et al. (2021) on the missing-deflation puzzle, and “it also lowers the sensitivity of prices to marginal cost for a given degree of price stickiness” (fn. 3).

Separately for a representative emerging market and a representative advanced economy, with the financial-friction parameters pinned to published estimates (Section 3, pp. 21-23). From the authors’ own dataset, the median advanced and emerging economy have similar government purchase shares, import content of consumption, and reserves-to-GDP, but “while the net foreign asset (NFA) position in AEs is roughly balanced, it is typically negative in EMEs, making them vulnerable to swings in external financing conditions,” and real interest rates and inflation are higher in emerging markets, “translating into a lower probability of hitting the effective lower bound” (p. 21). Specific values: the debt limit parameter is set so that “in the steady state the ratio of debt to annual GDP is 12 percentage points below the constraint,” which “ensures that… the emerging market economy stays in the constrained regime for about 3 percent of time, which is a typical target used in the sudden stop literature (Bianchi and Mendoza, 2020).” The FX-shallowness parameter Gamma is 0.06 for the emerging market and 0.02 for the advanced economy, the former “so that… purchases of foreign currency worth 10% of GDP generate a roughly 15% exchange rate depreciation, in line with evidence presented by Adler et al. (2019).” Home-ownership shares of financiers and portfolio investors are both 0.75, which “allows us to obtain an unhedged FX exposure of a median EME equal to around 16% of GDP, in line with IMF (2021),” and banks are fully domestically owned in the baseline (p. 23). The intertemporal substitution elasticity is 1, consistent with log utility (p. 23). The authors flag a caveat about applying the calibration to any particular country: “there is considerable within-group heterogeneity both in advanced and emerging market economies” (fn. 13).

Yes, within estimated confidence bands, and it reproduces two asymmetries. The nominal-rigidities calibration is chosen to “closely match the empirical evidence on distinct reactions of AE and EME aggregates to an exogenous, 10 percent exchange rate depreciation,” and the model impulse responses “sit comfortably within the estimated ranges for both AEs and EMEs” when plotted against the 90 percent confidence intervals of Brandao-Marques et al. (2021) (pp. 22-23). Two features are reproduced: “contractionary depreciations in EMEs and a much larger response of inflation in this group of countries”; and, via the Kimball aggregator, “larger reactions of prices and output to an exchange rate depreciation compared to an appreciation of the same (absolute) size” (p. 23). The authors also note that this positive skewness in CPI “is not inherited from possibly asymmetric responses of the exchange rate, as the distribution of changes in the latter is close to symmetric” (fn. 14).

Almost entirely because of the nominal-rigidities parameters, not because of FX mismatch – at least while the debt constraint is slack (Section 4.1, pp. 26-27). The shock makes portfolio investors less willing to hold home bonds, follows an AR(1) with persistence 0.95, and is scaled so the advanced economy’s real exchange rate depreciates about 10 percent; for that comparison the authors temporarily set aside differences in FX market depth and rule out hitting the debt limit. In the advanced economy “the exchange rate depreciation stimulates net exports, generating an expansion in economic activity,” inflation rises but “with well-anchored inflation expectations, monetary policy can ’look through’ the transient rise and focus on output,” and “all told, the shock looks very similar to a standard aggregate demand shock” – indeed with Gamma = 0 it “would have no effects on such an economy at all” (p. 26). In the emerging market the depreciation’s “large and persistent effects on inflation” induce tightening that “crowds out domestic demand, while higher stickiness of export prices in foreign currency moderates an increase in foreign demand (as in Gopinath et al., 2020). As a consequence, output contracts while inflation rises in parallel” (p. 27). The decomposition is explicit and matters for interpretation: “the disparate responses of output and inflation in AEs and EMEs are almost entirely due to differences in parametrization of the nominal rigidities block… In contrast, FX mismatches, represented by the negative net foreign asset position in EMEs, do not amplify the contractionary nature of exchange rate depreciations. This, however, is only the case because the debt constraint is slack. As we show subsequently, had adverse balance sheet effects caused the economy to hit the debt limit, then output would have contracted even more substantially” (p. 27).

Mostly from smoothing inefficient UIP-premium fluctuations, not from preventing sudden stops – the paper establishes this by comparison, not assertion (Section 6.3, pp. 38-39). The rule sets the FXI feedback coefficient on the UIP premium to 0.5. It works on two channels: “it calls for deploying FXIs to lean against inefficient movements in the UIP premium induced by swings in investor risk appetite,” reducing exchange rate, inflation and trade-balance volatility; and it “has a precautionary flavor, as the fall in the volatility of the UIP premium helps reduce the probability of financial stress.” In their setup “the FXI policy rule completely eliminates ‘sudden stops’.” The decisive comparison: the rule “would continue to improve welfare even if the precautionary rationale were entirely absent. This follows from noting that its benefits, on both measures considered, significantly exceed those associated with completely eliminating the debt limit,” which “underscores the fact that there are substantial benefits of smoothing inefficient fluctuations in the UIP risk premium, which are costly irrespective of whether or not they occasionally precipitate a crisis” (p. 39). The authors also record that the welfare benefits of strengthening monetary autonomy “turn out to be largest in environments characterized by substantial balance sheet FX mismatches, low FX market liquidity, and weakly anchored inflation expectations” (p. 4).

Substantial gains, but with a notable share coming from taxing foreigners and from the terms-of-trade effect rather than purely from crisis avoidance (Section 6.3, pp. 40-41). The rule leans against the accumulation of net foreign liabilities, implemented with a feedback coefficient of 0.01. It “significantly reduces the frequency of ‘sudden stops’ and brings more stability to economic activity,” and its welfare gains “exceed those associated with eradicating sudden stops.” The authors are candid about the composition: “Most of these additional gains can be attributed to extra revenue collected by taxing foreign investors whenever net foreign liabilities rise above their steady state level. Moreover, by decreasing the average level of foreign debt, the rule also compresses the UIP premium and hence makes the real exchange rate stronger on average, thus addressing the terms of trade externality discussed in the NOEM literature (Corsetti et al., 2010)” (pp. 40-41). They also flag when this tool is especially attractive: “economies that are near the debt limit, but without a sufficient stock of reserves to implement FX sales to allay downward exchange rate pressures in a sudden stop event” (p. 4).

Complements, and especially so near the debt limit. Even though the FXI rule already eliminates sudden stops, “using it in conjunction with the aforementioned CFM rule, which one could naturally motivate by overborrowing, generates even larger gains… these two rules combined make for a potent policy mix. Their complementarity is particularly effective when the economy is close to its external debt limit. In such circumstances, the CFM rule facilitates gradual decumulation of net foreign liabilities by keeping the exchange rate weak, while the FXI rule protects the economy against abrupt portfolio capital outflows that could otherwise disrupt this smooth adjustment by triggering a ‘sudden stop’” (p. 41). The authors add that the case for CFMs is understated by their model: synergies “could be even larger in the presence of sectoral vulnerabilities, which CFMs may be better suited to deal with,” and “problems with identifying the ‘portfolio component’ of capital flows, as well as a stochastic debt limit, would both make it harder for FXIs to eliminate financial stress so efficiently, increasing the relative role of CFMs” (p. 41).

An FX intervention rule that fires only on the borrowing limit binding is, on balance, welfare-reducing in the stochastic environment – it wins on impact and loses over the ergodic distribution (Sections 5-6.3, pp. 4, 42). The rule props up the exchange rate by selling reserves when the constraint binds, and “by doing so, it attenuates the spike in credit spreads once the ‘sudden stop’ occurs, making these episodes less painful. Even so, this prevents the needed adjustment in external debt and, by making the exchange rate stronger, it undermines the economy’s external competitiveness, slowing down accumulation of net foreign assets during the recovery phase.” The result is that the rule “makes the economy more vulnerable to financial crises in an ergodic sense, reducing the average level of output while amplifying its volatility and skewness,” so “both the utility-based welfare criterion and the simple loss function point to, on balance, detrimental effects of such a rule” (p. 42). The general statement in the introduction is the same: rules “that respond only to financial stress episodes can actually be detrimental to welfare in a stochastic environment, and may exacerbate macroeconomic costs as measured by a standard quadratic loss function,” so “it is hence important for the FXI and CFM policies to be active also when financial conditions are favorable, and in particular to lean against overheating when investors’ risk appetite is high” (p. 4).

If the economy is unlikely to be pushed back into crisis, or if sudden stops are sufficiently disruptive – the authors explicitly show the sign can flip (Section 6.4, pp. 42-43). “FXI and CFM interventions can be beneficial during ‘sudden stops’ if future shocks are unlikely to drive the economy back into crisis; and more generally, the welfare benefits of using these tools in a crisis depend both on the distribution of shocks, initial conditions, and structural features of the economy.” Concretely: “if sudden stops were relatively more disruptive for the economy, which we can proxy by assuming a sufficiently high share of foreign ownership of domestic banks, then the welfare effects associated with rules only responding to spreads could flip sign and become positive.” Their methodological upshot is stated as such: “it is critical to use stochastic simulations to assess the benefits of such policies, as their net benefits may be smaller than suggested by the on-impact effects” (p. 43). Two implementation caveats follow: the FXI rule assumes the central bank “is able to precisely determine the UIP premium component attributable to portfolio investors’ actions,” and if it cannot, outcomes are “somewhat inferior,” raising the role for CFMs; and if the UIP premium is hard to estimate, a rule responding directly to the exchange rate delivers “some (but smaller) welfare gains… but only if portfolio flows account for a sufficiently large share of fluctuations in this variable, so that the benefits of reducing fluctuations in the UIP premium outweigh the losses from preventing efficient exchange rate adjustments in response to other shocks” (p. 43).

That sterilized FX purchases can provide domestic stimulus at the effective lower bound, illustrating Svensson’s “foolproof way” (Section 7, pp. 43-45). The baseline is an advanced economy hit by negative home and foreign preference shocks (AR(1) roots of 0.95) plus a positive risk-appetite shock (root 0.90) that puts appreciation pressure on the currency: “inflation falls below zero, output falls by about 8 percent, and the policy rate is pinned at zero.” The intervention is persistent (AR(1) root 0.95) and large – “amounting to 20% of annual trend GDP in the first quarter” – which the authors tie to the advanced economy’s deep FX markets (Gamma reset from 0.06 to 0.02). Three channels operate: the induced depreciation raises net exports; because the central bank does not raise rates until the lower bound stops binding, real rates fall and crowd in domestic demand; and “the weaker exchange rate keeps CPI inflation closer to its target during the crisis, implying that the intervention helps stabilize both inflation and output” (p. 44). Under the emerging market calibration with the same initial conditions and FX depth, “the same-sized intervention… would have notably larger stimulative effects on domestic absorption and the output gap,” mainly through “larger second round effects on inflation” – but with an explicit scope condition: “at least for those EMEs with indebtedness safely below the debt limit” (pp. 44-45).

They can be beggar-thy-neighbor and even globally contractionary in the short run, but only when the foreign economy is also at its lower bound. Exploiting the two-country structure with the home economy set to half of the world, the authors report that “in a global liquidity trap, FXI may have sizable beggar-thy-neighbor effects, even to the extent of being globally contractionary in the short run. Interestingly, the outcomes become much more favorable when the foreign economy is not stuck in a liquidity trap, but can instead lean against deflationary pressures by providing conventional monetary stimulus. The output spillovers in that case are very small, as a result of which, the world economy ends up clearly benefitting from FX purchases in the constrained economy” (pp. 45-47). The introduction states the same asymmetry: FXI by a large bloc at the lower bound “can be associated with negative output spillovers in economies whose monetary policy is also constrained, but are fairly innocuous whenever the foreign economy still has ample policy space” (p. 4).

That FXI or CFMs substitute for sound monetary policy, that the model’s frictions are the only relevant ones, or that the tools are costless in the long run. The first warning is a footnote in the introduction: “our results should not be construed to imply that FXIs or CFMs can or should be seen as a substitute for sound monetary policy” (fn. 4). The conclusion lists what is missing and what is coming: the model does not include macroprudential policy, which “could also be extended to include… e.g., by adding a housing sector and/or bank capital,” and the framework is being estimated by Bayesian methods in parallel work (p. 47). Two practical considerations sit outside the model: decisions “must often be made under considerable uncertainty about the underlying shocks and the distance from the unobserved debt limit,” so “using FX sales to support the exchange rate may risk large and potentially destabilizing losses of reserves if the shock is more persistent than anticipated”; and the tools “may have non-negligible longer-term costs. For example, they may impede financial market development and encourage an excessive buildup of foreign currency debt, potentially amplifying vulnerabilities” (p. 47). The authors’ own summary of their contribution is correspondingly hedged – they see the framework “as providing an important input into the associated cost-benefit analysis,” not as settling it.