A Quantitative Model for the Integrated Policy Framework

That emerging market central banks which adopted inflation targeting nonetheless kept using FX intervention and capital flow tools, while advanced economy central banks largely did not (Section 1, pp. 1-2). The setup notes that inflation targeting, “first introduced in New Zealand in 1990 and then in many other advanced economies,” “was found to be successful in stabilizing both inflation and real aggregates (Svensson, 2010),” and that emerging and developing adopters “on average… outperformed comparators with other monetary policy frameworks.” Yet “unlike their advanced economy counterparts, many EME central banks with IT frameworks have continued to rely on foreign exchange interventions (FXIs) in their monetary policy operations, and some also use capital flow management tools (CFMs),” especially “during episodes of volatile capital flows (Hoffmann et al., 2019).” The dilemma is stated concretely: “lowering interest rates to deter capital inflows may lead to domestic credit booms and increase financial stability risks, while contractionary monetary policy during a capital outflow episode may further undermine economic activity and in so doing exacerbate the outflows.”

Vegh and Vuletin’s (2013) finding that roughly half the emerging and developing sample met outflow shocks with tighter policy, against countercyclical responses in small open advanced economies. “Vegh and Vuletin (2013) provided empirical evidence that – for about half of the emerging market and developing countries in their sample – monetary policy responses to adverse capital flow shocks were procyclical, and outflow shocks were often met with tighter monetary policy. By contrast, in small open advanced economies domestic considerations dominate and adverse external shocks are usually met with monetary policy easing, which cushions the effects of external tightening on domestic demand” (p. 1). The authors report the standard interpretation without adopting it as their own mechanism: “It has been argued that this disparity in responses reflects that EME central banks often pursue exchange rate and financial stability goals in addition to inflation stabilization, and that multiple objectives require multiple instruments (Ghosh et al., 2016).” Their own motivation is narrower and empirical: “the substantial empirical evidence showing that exchange rate changes tend to have larger and more persistent effects in emerging market economies than in advanced economies,” which “appears partly attributable to relatively less well-anchored inflation expectations in many EMEs” (p. 2, fn. 1).

Four extensions to the Galí-Monacelli core, each aimed at empirical realism rather than tractability (Section 2, pp. 4-5). First, “a broader array of real and nominal rigidities, including habit persistence in consumption, sticky wages and prices, and imperfect passthrough of exchange rate changes to traded goods prices,” with local currency pricing as benchmark but producer and dominant currency pricing also available (Gopinath et al., 2016). Second, “the possibility that some agents form inflation expectations adaptively to capture imperfect credibility of monetary policy,” which the authors call “important in accounting for how exchange rate changes may have large second round effects on inflation.” Third, incomplete markets, “so that domestic agents must borrow or invest in a foreign currency-denominated bond, in contrast to the assumption of full international risk sharing in Galí and Monacelli (2005).” Fourth, Gabaix (2016) discounting in the aggregate demand and supply equations, which “mitigates the forward guidance puzzle, so that policies that operate by affecting long-horizon forward rates have less traction in stimulating aggregate demand and inflation today.” The model is solved in Dynare.

A threshold UIP premium, a depreciation-sensitive borrowing spread, and the effective lower bound – together they make initial conditions determine the severity of a shock (Section 2, p. 5; Section 4, pp. 20-22). “First, we allow for a nonlinear UIP risk premium specifying that the return required by investors on domestic borrowing rises sharply when net foreign liabilities exceed a threshold level. Second, we assume that the spread facing domestic borrowers rises nonlinearly as the home currency depreciates. This feature, in the spirit of (Bruno and Shin, 2018), proxies for how depreciations may lead to a sharp tightening of domestic financial conditions if there is significant unhedged foreign currency debt. Lastly, we allow for the possibility that policy rates may be constrained by the effective lower bound.” The economic point of the threshold structure is that it reproduces a stylised crisis pattern: the premium “only rises modestly as net foreign liabilities increase” under normal risk tolerance, but “jumps much more for highly indebted economies during periods of market stress in which risk tolerance declines.” The authors tie this to episodes: “in the run-up to a crisis, spreads are often low even for economies with high vulnerabilities but spreads in vulnerable economies rise sharply during the crisis once risk tolerance declines. This pattern was apparent, for instance, around the time of the euro area sovereign debt crisis as well as during the recent COVID crisis.”

A real exchange rate threshold of 22.5, set deliberately high so that depreciation alone does not move spreads (Section 4.1, pp. 21-22). The spread follows a logistic form in the gap between the actual real exchange rate and a shocked threshold. “The calibration of q-bar = 22.5 implies that even a substantial real exchange rate depreciation relative to the steady state will not induce much of a rise in the private borrowing spread unless the stochastic risk appetite shock is also sufficiently negative.” The justification is explicitly empirical: “This feature helps match empirical evidence for EME corporate spreads which suggests that although large exchange rate depreciations make a sharp runup in private borrowing spreads (proxied empirically by corporate spreads) more likely, it is nevertheless often the case that the exchange rate may depreciate substantially without triggering a big increase in spreads.” The nonlinear parameters are not free-handed: they are pinned down through “essentially a simulated method of moments exercise (see e.g. McFadden, 1989)” matching “unconditional persistence, unconditional and conditional volatilities, and mean and max values” for sovereign dollar-bond spreads across selected emerging markets.

Only the price and wage formation parameters, plus the presence of the two balance sheet channels. Everything else is held identical (Section 2.4, pp. 14-16). “For tractability, the calibration is identical for advanced and emerging market economies with two key exceptions. First, we allow the parameters governing price and wage formation to differ, to capture less-well anchored inflation expectations in EMEs relative to AEs. Second, the EME formulation of the model allows for two key nonlinear balance sheet channels.” The advanced economy calibration draws on the estimated-DSGE literature (Lindé et al., 2016; Del Negro et al., 2013; Campbell et al., 2012): Phillips slopes of 0.005 for prices and 0.02 for wages, with indexation of 0.5 and 0 – equivalent, in a homogeneous Calvo model, to re-optimisation probabilities of 0.92 for firms and 0.87 for unions. The emerging market sets both indexation parameters to 0.75 “to capture the greater degree to which inflation expectations depend on realized inflation and wages,” with slopes raised to 0.02 and 0.03 – equivalent to re-optimising “on average once per year.” Shared parameters include a discount factor of 0.995 (a 2 percent annualised steady-state real rate, 4 percent nominal at 2 percent inflation), an intertemporal substitution elasticity of 1, habit of 0.8, Frisch elasticity of 0.5, capital share of 0.3, government share of 20 percent, trade elasticities of 0.8 and an import share of 0.2. Trade rigidity does differ by type: import and export firms adjust prices “on average every 3 (5) quarters” in emerging (advanced) economies. The Taylor rule uses an inflation coefficient of 1.5 and an output gap coefficient of 0.0625 – “0.25 at an annual rate… modestly lower than in the standard Taylor rule (0.50).”

Because of expectation anchoring: the advanced economy can look through a transient inflation rise, the emerging market cannot (Section 3.1, pp. 17-18). The shock is an investor risk tolerance decline “tantamount to a UIP risk premium shock,” AR(1) with persistence 0.9, “scaled so that the real exchange rate depreciates by 10 percent in the AE economy.” For the advanced economy, “the exchange rate depreciation stimulates net exports, causing output to rise. The combination of higher output and higher import prices causes inflation to rise, but with well-anchored inflation expectations monetary policy can ’look through’ the transient rise in inflation and focus on output. All told, the shock looks very similar to a ‘standard’ aggregate demand shock.” For the emerging market, “inflation expectations are less well anchored than in the AE, and the exchange rate depreciation has large and persistent effects on inflation. This induces the central bank to raise real (and nominal) interest rates, which crowds out domestic demand. Output contracts slightly on balance, while inflation rises substantially.” The authors flag that this comparison understates the problem, since the balance sheet channel is switched off for it: “Had our simulation allowed for a balance sheet channel in which the exchange rate depreciation increased private borrowing spreads, output would contract even more.”

No – the two rules trade inflation against output, and neither delivers both. Comparing the benchmark Taylor rule with a more aggressive inflation-stabilisation rule under the same shock: the accommodative case “keeps output relatively stable but allows the exchange rate depreciation to have large and persistent effects on both CPI inflation and domestic inflation, potentially undermining monetary policy credibility,” while “tighter policy – in which the central bank responds more vigorously to both domestic and imported inflation – limits the size and persistence of the increase in inflation but generates a sharper output contraction of about 1.5 percent” (Section 3.2, pp. 18-19). The conclusion drawn is the paper’s justification for its subject: “conventional monetary policy performs rather poorly in simultaneously achieving a central bank’s inflation and output objectives if inflation expectations are not well-anchored. This provides a rationale for using additional policy instruments (FX intervention or capital controls) to reduce the risk of a de-anchoring of inflation expectations, and thereby allowing monetary policy to focus more on output stabilization.”

A great deal: the same shocks produce a several-times-larger depreciation and a considerably larger output contraction in a high-liability economy (Section 4.2, pp. 25-27). The scenario pairs a decline in global risk tolerance (AR(1) of 0.85, “a half-life of about 5 quarters”) with a fall in foreign import demand, applied identically to a “vulnerable EME” with high net foreign liabilities and a “less vulnerable EME” with low ones. “Figure 4 shows that the vulnerable EME experiences a depreciation of its real exchange rate that is several times larger than in the less vulnerable EME. While the depreciation provides a larger boost to net exports in the vulnerable EME, its output nonetheless experiences a considerably larger contraction. The bigger output decline both reflects that the depreciation triggers a larger rise in private borrowing spreads, and that the larger rise in inflation in the vulnerable EME requires monetary policy to be much tighter.” The authors are careful to attribute the disparity to initial conditions rather than structure in this experiment, and note it is a lower bound: “the poorer tradeoff for the more vulnerable EME is driven by the larger exchange rate depreciation – and thus due to weaker initial conditions – the disparity between the EMEs would be even larger if we allowed for structural differences such as less well-anchored inflation expectations in the more vulnerable EME.”

Both limit the depreciation and thereby relieve the stagflationary bind – but they borrow from the future, and the cost grows with the persistence of the shock (Section 4.2, pp. 27-28). “Either of these policies reduces the depreciation of the exchange rate, which allows policy to be more accommodative by lessening inflationary pressure, and also mitigates the rise in private borrowing spreads. The upshot is that these policies allay the ‘stagflationary’ effects of the UIP risk premium shock, reducing inflation while boosting output and exerting even larger expansionary effects on domestic demand.” The intertemporal cost is spelled out: “the stronger exchange rate and domestic demand under FXIs or CFMs translate into less trade balance improvement, and, consequently, a somewhat bigger deterioration of net foreign liabilities. The higher net foreign liabilities in turn account for why the UIP risk premium runs a bit higher in the longer-run under these policies than under the baseline.” The scope condition on persistence is explicit and load-bearing: “their ability to improve near-term inflation-output tradeoffs makes them appealing in the case of fairly transient shocks… Their appeal is somewhat diminished, however, as the shock becomes more protracted. While CFMs and FXIs still improve output-inflation tradeoffs in the near-term, they weaken the turnaround in domestic borrowing that is needed to lower risk spreads in the longer-term. Because investor risk appetite remains low, risk spreads run considerably higher at longer-horizons, pushing output below the level that would prevail absent such interventions and inflation higher.” The two tools are not identical: the upward pressure on the premium “is slightly higher for CFMs than FXIs given that the latter directly depress the risk premium… whereas CFMs do not (instead, they make domestic assets more attractive by raising their after-tax rate of return).”

Yes, and the paper prices it honestly as insurance with a premium. “In addition to improving tradeoffs in episodes of financial tightening, these policies may play an insurance role by limiting an endogenous buildup of risk associated with higher foreign indebtedness. This may be achieved by responding to ‘favorable’ risk premium shocks (associated with surging capital inflows) that cause the exchange rate to appreciate and private borrowing spreads to fall… it means keeping the economy away from high debt levels that make the economy vulnerable to a sudden sharp rise in spreads.” The cost is stated in the same breath: “Even so, this insurance does come at some cost: in particular, such policies preclude some of the borrowing that would benefit households if the high global risk tolerance state were to persist” (Section 4.2, pp. 28-29).

A distributional picture: the nonlinearities produce fat left tails in real activity, and systematic rules shrink them (Sections 4.3-4.4, pp. 29-36). The setup uses six shocks over T = 20,000 simulated observations, with the linear and nonlinear models fed identical shocks “so that any differences in the resulting simulated distributions reflect differences in shock propagation mechanisms.” Most shocks are AR(1) with root 0.95 (0.8 for the monetary shock), with standard deviations of 1.5 for consumption and government spending, 2 for foreign demand, and 0.25 annualised for the monetary shock. Because the model has no steady-state spreads, the authors “add the median spreads observed in our EM economies (117 basis points for corporates and 205 basis points for the government) to the simulated data” specifically “to test whether our risk premium specifications are reasonable and do not generate sharply negative sovereign spreads.” A validation check is reported against the data rather than asserted: the unconditional correlation between the UIP premium and private spreads “is 0.35, which is close to the median value of 0.39 in the data,” and the authors note the relationship “is not particularly tight,” “consistent with the data.” The resulting skews: “Domestic absorption has a sizeable negative left skew which stretches to -15 percent. The output gap also has a sizeable, though somewhat smaller, skew to the left that extends roughly to -10 percent,” the difference being that “the trade balance works as an automatic stabilizer.” Under a combined FXI/CFM rule, “these policy rules markedly reduce the likelihood of large real exchange rate depreciations… This in turn mitigates the rise in private borrowing spreads, and the need for sharp policy tightening to control inflationary pressures. These policy reactions have the effect of noticeably reducing the tail risk of a large output contraction.”

Deliberately restrained: they respond only to the non-fundamental component of the premium, not to the premium itself (Section 4.4, pp. 33-35). The FX purchase rule responds to the innovation in private borrowing spreads and to the gap between net foreign liabilities and their shocked threshold, with benchmark coefficients of 5 and a value tied to the threshold parameter; the outflow-CFM rule responds to the nonlinear premium innovation relative to its linearised counterpart, with coefficients of 5 and 2.5. The restraint is the design point: “the specification implies that the central [bank] does not attempt to offset all fluctuations in the UIP risk premium: the central bank only responds to the part of the premium that depends on the non-fundamental debt limit shock and not the fundamental part that is only driven by net foreign liabilities… In this sense the rule is ‘minimalistic’ and implies that the CB only intervenes when either private or the countries exchange rate premium are off relative to fundamentals and if they are not the central bank only uses its traditional policy tool.” Two caveats travel with the results. The rules still leave the longer-run tradeoff in place: “as was the case in Figure 5 these interventions delay the adjustment towards lower net foreign liabilities which implies that the UIP risk premium eventually becomes slightly more elevated.” And feasibility is assumed, not shown: “An implicit assumption is of course that the central bank has sufficient reserves and determination to follow-through on the rules.” One practical finding is that the CFM rule is cheap to run: “very small adjustments are needed for outflow CFMs according to our model if a rule clarifying how they will be adopted is credibly communicated.”

That FX purchases can substitute for exhausted rate policy, illustrating Svensson’s “foolproof way” (Section 5, pp. 37-39). The exercise uses the linearised advanced economy model with both risk premium shocks set to zero – a scope condition worth noting, since the balance sheet channels that drive the rest of the paper are switched off. Following the fiscal multiplier literature (Christiano et al., 2011; Erceg and Lindé, 2014), the baseline combines negative consumption-preference and export-demand shocks (AR(1) of 0.95) with a positive risk appetite shock (0.85) that appreciates the currency. “Inflation falls below zero, output falls over 10 percent below baseline, and the policy rate is pinned at zero.” FX purchases work through three channels: “The exchange depreciation induced by the FXI cushion the blow of the recessionary shocks by raising net exports and stimulating output. Moreover, because the central bank doesn’t react to higher inflation by raising policy rates, real interest rates decline which crowds in domestic demand initially. Furthermore, and importantly, the weaker exchange rate implies that core CPI inflation remains much closer to its 2 percent target during the crisis.” The extension to emerging markets is explicitly conditional: they “may derive even more of a boost to inflation and output,” but only “in the absence of any material effects on either the UIP risk premium or on the private borrowing spread,” and the conclusion applies “at least for EMEs with indebtedness well below the debt limit (i.e., the EME in question is not vulnerable to a risk-off shock).”

They explicitly decline to. The paper contains one of the clearer self-limiting statements in this literature. “While our model highlights that FXIs and CFMs can improve policy tradeoffs under certain conditions, additional considerations – not fully addressed by our model – must be taken into account when deciding whether or not to utilize these tools in practice. Thus, our model results should not be taken as an unqualified endorsement of the use of these tools either in general or in specific situations” (Section 1, p. 3). Two reasons are given. The first is a modelling limitation: “while our model (in effect) assumes perfect foresight, decisions about whether to use these tools must often be made under considerable uncertainty about the underlying shocks and the efficacy of policy actions. For instance, 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 requires the equilibrium real exchange rate to depreciate.” The second is about long-run costs outside the model: “Policies that stabilize the exchange rate may impede the development of hedging markets, and encourage an excessive buildup of foreign currency debt that may amplify vulnerabilities; and frequent resort to CFMs may deter foreign investment and slow financial market development.” The authors position the model as an input rather than an answer: “A full analysis of these potential benefits and costs is essential for determining appropriate policy, though we see our modeling framework as providing important input into making that assessment.”

Four, and the first three are visibly the agenda of the successor paper. The conclusion proposes extending the model “to include macroprudential policy, for example, by adding a reduced-form housing and/or banking sector”; considering “how fiscal policy could affect the appropriate use of IPF tools”; moving to “a multilateral setting to assess the potential spillovers to other economies of using IPF tools”; and better understanding “the empirical transmission of IPF tools, and how transmission varies with structural characteristics.” On the last, the paper reports work in progress: “The model is currently being estimated for a group of emerging and advanced small open economies” (Section 6, pp. 39-40). The successor WP/21/292 delivers the multilateral extension via a two-country structure, and cites this paper as the related earlier project. The authors also flag the calibration’s own limits, for instance that the 20 percent import share “is lower than in many small open economies,” justified because “our model does not allow imports to be used as inputs into export good production.”