Household Heterogeneity and the Transmission of Foreign Shocks

The paper is motivated directly by Hungary’s experience in the run-up to and aftermath of the Global Financial Crisis, when households borrowed heavily in foreign currency (particularly to finance home purchases) and then suffered severe consumption contractions once the currency depreciated sharply during the 2009 sudden stop, a channel documented empirically by Verner and Gyongyosi (2018) (Sec. 1). The authors note that at least thirteen European countries (Austria, Iceland, Estonia, Latvia, Lithuania, Hungary, Poland, Slovenia, Croatia, Serbia, Bulgaria, Romania, and Ukraine) had more than 20% of household debt denominated in foreign currency in the 2000s. While domestic market incompleteness had already been integrated into closed-economy macro models, the paper argues no existing quantitative open-economy framework combined that household heterogeneity with the standard international-macro apparatus needed to study currency-denominated debt revaluation across the wealth distribution – the gap the HANKSOME framework is built to fill.

The model nests a Bewley-Huggett-Aiyagari household block – with idiosyncratic labor productivity risk and a borrowing constraint – inside a small open economy with capital, monopolistically competitive firms subject to Rotemberg price-adjustment costs, and a union that sets wages following Hagedorn, Manovskii and Mitman’s approach to aggregating heterogeneous household preferences (Sec. 2). Crucially, households’ nominal asset portfolios can include bonds denominated in domestic currency, bonds denominated in foreign currency, and real bonds denominated in units of the home good, with arbitrage across these asset classes pinning down Fisher and covered-interest-parity conditions (eq. 8-10); this portfolio structure lets the paper independently vary both the aggregate share of debt denominated in foreign currency and how that debt is distributed across the wealth distribution, which turns out to be the key to the paper’s results.

The model is calibrated at a quarterly frequency to match Hungary’s household debt-to-income ratio, wealth-to-GDP ratio, exports and net foreign asset position, and, distinctively, the cross-sectional distribution of household leverage from the Household Finance and Consumption Survey (HFCS) administered by the European Central Bank (Sec. 3). Key targets include a net-wealth-to-quarterly-GDP ratio of 8.26, a capital-to-output ratio of 9.70, an export share of value added around one-third, a net-foreign-asset-to-quarterly-GDP ratio of -2, and, distinctively, the HFCS-based finding that Hungarian households in the bottom wealth quintile held zero net wealth but roughly 6% of the economy’s aggregate gross assets (and a corresponding gross debt position equal to about 24% of average household yearly labor income) – the calibration target that lets the model reproduce Hungary’s specific pattern of leverage concentrated among the wealth-poor.

Simulating an unanticipated interruption to a decade-long current account expansion – calibrated to match the timing and magnitude of Hungary’s 2009 reversal – the paper finds that under a flexible exchange rate, consumption and investment contract sharply while output actually rises on impact (Sec. 5). The mechanism is that the tightening borrowing constraint and the fall in household net worth from debt revaluation induce households to work more (a labor-supply wealth effect), and this expansion in labor supply, combined with a sharp deterioration in the terms of trade that boosts foreign demand for the now-cheaper home good, generates an output boom even as domestic absorption collapses – though the paper notes the terms-of-trade decline is large enough that the foreign-currency value of domestic output still falls. This decoupling of output and consumption dynamics is central to the paper’s argument that the welfare cost of floating in this setting is not the classic “contractionary devaluation” story.

Under a fixed exchange rate, output does contract, because binding nominal rigidities are no longer offset by exchange-rate adjustment, but the resulting milder terms-of-trade depreciation limits the revaluation of households’ foreign-currency debt, producing a smaller, not larger, fall in consumption relative to the flexible-rate case (Sec. 5). The paper computes that, in consumption-equivalent-variation (CEV) terms, a fixed exchange rate regime would have raised utilitarian welfare by 0.47% relative to the flexible regime actually followed, with more than 92% of households benefiting; even stripping out the utility value of the additional leisure generated by the peg’s recession, the welfare gain remains a positive 0.21% CEV and 74% of households still prefer the fixed regime. The paper frames this explicitly as a welfare-based rationale for empirically observed “fear of floating” that does not depend on the depreciation itself being contractionary for output.

A stripped-down experiment (Sec. 6.1-6.2) compares three benchmarks subjected to the same credit-supply contraction: a “real benchmark” where household debt is denominated in home goods (so there is no debt-revaluation channel and a flexible exchange rate exactly replicates the frictionless allocation); a flexible-exchange-rate economy with foreign-currency debt; and a fixed-exchange-rate economy with foreign-currency debt. In the real benchmark, the terms of trade and nominal exchange rate depreciate to offset nominal rigidities, but since debt does not revalue, the resulting consumption decline is comparatively mild. Once debt is redenominated into foreign currency, the same depreciation that closes the flexible economy’s output gap now also revalues debt and sharply amplifies the fall in household net worth and consumption – “the depreciation of the nominal exchange rate in the presence of foreign currency debt leads to an amplification of the responses of all endogenous variables.” Under the fixed-rate counterfactual with foreign-currency debt, by contrast, output contracts due to binding nominal rigidities, but the resulting milder terms-of-trade adjustment means “a more moderate revaluation of foreign-currency debt,” so household consumption falls by less than under the flexible regime even though output contracts by more – the paper’s clearest illustration that the welfare-relevant channel is debt revaluation, not the output gap itself.

Holding the economy’s aggregate leverage fixed but concentrating more of that debt among households with zero or low net worth substantially deepens the aggregate consumption contraction and the terms-of-trade decline under a flexible exchange rate, because those wealth-poor households have the highest marginal propensities to consume, so a given aggregate debt revaluation translates into more spending cutbacks the more it is concentrated among them (Sec. 6.3, Fig. 11). The paper is explicit that “in the absence of heterogeneity in the marginal propensity to consume – as would be the case in the representative agent economy – the three economies would deliver the same aggregate responses,” so the differences documented across the low-, middle-, and high-poor-leverage calibrations exist purely because of household heterogeneity and market incompleteness. Under a fixed exchange rate, by contrast, the leverage distribution has “negligible effects” on aggregate outcomes, because the milder terms-of-trade movement limits debt revaluation for everyone regardless of who holds the debt.

In the empirically relevant scenario (flexible exchange rate, foreign-currency debt), highly leveraged households in the bottom wealth quintile suffer the largest welfare losses – willing to forgo more than 10% of lifetime consumption to avoid the credit-supply contraction in the high-poor-leverage calibration – while wealthier households, who by construction hold less debt whenever poorer households hold more (since aggregate leverage is held fixed across the experiments), suffer smaller losses or even gains (Sec. 6.4, Fig. 12-13). Mapping out combinations of household leverage and the foreign-currency debt share, the paper finds that low-wealth households unambiguously prefer the fixed exchange rate regime across essentially the whole parameter space it considers (partly because a fixed rate also supports the real wage, benefiting even unlevered poor households), while wealthier households prefer the flexible regime only when either the leverage of the poor or the foreign-currency debt share is comparatively low.

No – computing welfare for a hypothetical “as-if” representative household using the HANKSOME model’s own aggregate consumption path shows that this representative agent behaves much like the median household and would frequently choose the flexible exchange rate regime even in cases where the population-wide utilitarian welfare metric – which accounts for the heterogeneous incidence of debt revaluation across the wealth distribution – favors the fixed regime instead (Sec. 6.4, Fig. 14). The paper presents this as direct evidence that a representative-agent model can reach a different, and in this context less protective, policy conclusion purely by construction, because it cannot see the concentrated welfare losses borne by highly-leveraged, low-wealth households that drive the aggregate utilitarian comparison in the full heterogeneous-agent model.

The paper explicitly distinguishes its debt-revaluation mechanism from two related but distinct strands of literature: the balance-sheet/contractionary-devaluation tradition (Diaz-Alejandro 1963; Céspedes, Chang and Velasco 2004; Frankel 2005), which typically ties the costs of devaluation to output contractions via corporate balance sheets, and the original “fear of floating” account of Calvo and Reinhart (2002), which attributes central banks’ reluctance to devalue to a credibility-driven fear of triggering dollarization of liabilities (Sec. 1.1). In this paper’s model, foreign-currency household debt is a cause, not an averted consequence, of the reluctance to devalue, and the mechanism operates through the household sector rather than corporate balance sheets; moreover, because output actually rises under the flexible regime in the paper’s calibrated experiment, the welfare cost of floating cannot be attributed to devaluation being contractionary for output at all – the paper’s title for this result is that the model “can rationalize the observed ‘fear of floating’ … even in the absence of contractionary devaluations.” The paper also contrasts its findings with Fornaro (2015) and Ottonello (2013), whose collateral-constraint-based frameworks instead favor a more volatile exchange rate during external crises, and with Cugat (2019), whose limited-heterogeneity model (sector-specific income risk rather than debt revaluation) does not find a comparable case for fixed exchange rates.