On the Fiscal Implications of Twin Crises

First-generation currency-crisis models, which are especially well suited to explaining twin banking-currency crises, predict high post-crisis inflation because they assume the fiscal costs of a bank bailout are financed by printing money – but in reality, “many large devaluations are followed by moderate rates of money growth and inflation” (Section 1). The authors frame this as motivating three questions: “how do governments actually pay for the fiscal costs of twin banking-currency crises? … what are the implications of different financing methods for post-crisis inflation rates? … can the inflation predictions of first generation type models be reconciled with the data?” (Section 1).

The paper lists explicit fiscal reform, explicit default, seignorage, debt devaluation, and implicit fiscal reform as the five available strategies, noting that “in a world of forward looking economic agents, different mixes of these strategies have different implications for both the severity of a currency crisis and for post-crisis inflation rates” (Section 1). Verbatim, the strategies are “(i) implement an explicit fiscal reform by raising taxes or reducing spending; (ii) explicitly default on outstanding debt; (iii) print money to generate seignorage revenues; (iv) deflate the real value of outstanding nonindexed nominal debt; or (v) engage in an implicit fiscal reform by deflating the real value of government outlays that are fixed, at least temporarily, in nominal terms (e.g. civil servant wages or social security payments)” (Section 1).

The model is explicitly a reduced form of the earlier Asian-crisis paper, stripped to its essential elements so the financing-mix question can be studied directly: “we analyze these implications using a version of the model in Burnside, Eichenbaum and Rebelo (2001) in which a currency crisis is triggered by prospective government deficits… we reduce the model to its essential elements: a money demand specification, a government budget constraint, a rule for exiting the fixed exchange rate regime, and an assumption about the nature of monetary policy after the devaluation” (Section 1). As in the earlier paper, the attack date is pinned down by a debt-threshold exit rule, and the government’s post-crisis money growth is set to satisfy its intertemporal budget constraint (Sections 2-3).

With PPP holding and no nonindexed government liabilities, the benchmark calibration to Korean data implies inflation of about 35 percent in the crisis year and 20 percent in steady state, with the rate of devaluation exactly equal to the rate of inflation – both far larger, and far more tightly linked, than what is typically observed (Section 3, “A Numerical Example”). The authors state directly that “the model predicts counterfactually large rates of inflation after a crisis” and that, because PPP holds by construction, “the rate of inflation coincides with the rate of exchange rate depreciation,” which “is inconsistent with the evidence” since “rates of devaluation are typically much larger than the corresponding rates of inflation” after real-world attacks (Section 3).

Introducing pre-crisis nominal bonds and nominal spending commitments lets debt devaluation and implicit fiscal reform substitute for some of the seignorage otherwise required, which lowers the money-growth rate needed to satisfy the government’s budget constraint and, in turn, lowers steady-state inflation (Section 4.1). In the baseline numerical example (nonindexed debt equal to 5 percent of GDP), adding nonindexed debt alone cuts steady-state inflation from 20.0 percent to 16.1 percent; with a much larger nonindexed debt stock (B = 0.5), the effect is far bigger – “the rate of inflation would be 15.5 percent in the first year after the currency crisis and 2.1 percent thereafter,” with the government raising “only 14.6 percent of the fiscal cost of the crisis… from seignorage revenues” and the rest from debt deflation. Adding a modest implicit fiscal reform on top of the baseline nonindexed-debt case (nonindexed government spending equal to only “about 2 percent of GDP”) “pays for over 50 percent of the cost of the crisis,” cutting long-run inflation further, from 16.1 percent to 6.1 percent (Section 4.1). The authors caution, however, that “for the countries involved in the Asian crisis of 1997 and Mexico there was not enough nonindexed debt for this to be a complete resolution of the problem” (Section 4.1).

By eliminating purchasing power parity through nontradable goods (whose prices are held fixed for several months after the crisis) and costs of distributing tradable goods, which together open a wedge between the exchange-rate depreciation and CPI inflation (Section 4.2). With nontradables alone, first-year inflation falls to 17.7 percent (steady-state 4.0 percent) while the modeled devaluation rises to 35.4 percent in the first year; adding distribution costs pushes the wedge further, so that – contrasted with the benchmark model’s 34.9 percent first-year/20 percent steady-state inflation, with devaluation exactly equal to inflation – the extended numerical example implies “first year inflation roughly equal to 14 percent while the currency devalues by over 50 percent,” with steady-state inflation of only “1 percent,” which the authors summarize as showing “this version of the model can account for large devaluations without generating grossly counterfactual implications for inflation” (Section 4.2).

The paper estimates Mexico’s fiscal cost at roughly 15 percent of 1994 GDP, finds that seignorage (6.3 billion dollars), debt devaluation (8.4 billion dollars), and net fiscal reforms (3.5 billion dollars) together account for about 18.2 billion dollars (4.3 percent of 1994 GDP) raised to date, and concludes the remainder will most likely be financed by continued money creation, consistent with Mexico’s historically higher post-crisis inflation (Section 5.2). The authors calculate that financing the rest of a larger estimated shortfall (using Caprio and Klingebiel’s higher cost estimate) would require the monetary base to grow at “an annual rate of 21.2 percent” from 2001 onward, and conclude that “absent any sign of fiscal reforms, it seems quite likely that the bulk of the costs will be covered via explicit seignorage revenues,” implying continued relatively high Mexican inflation (Section 5.2).

The paper estimates Korea’s fiscal cost at roughly 24 percent of 1997 GDP (114.4 billion dollars), finds that seignorage (5.2 billion), debt devaluation (13.7 billion), and fiscal reforms net of recession costs (9.7 billion) together cover only about 28.6 billion dollars (6 percent of GDP) to date, and argues the large remaining shortfall is likely to be financed mainly through future implicit and explicit fiscal reforms rather than money printing (Section 5.3). Consistent with this, a numerical experiment in which a future explicit reform covers “roughly 16 percent of GDP or 66.7 percent of the fiscal cost of the crisis” generates a modeled steady-state inflation rate of only 1.6 percent alongside a realistically large 59.9-percent first-year won depreciation, which the authors offer as an explanation for why “a year after the crisis, inflation in Korea became extremely low” even though the devaluation was large (Section 5.3).

The authors flag that the model overstates Korea’s actual first-year inflation (about 15 percent modeled versus roughly 7 percent observed) and cannot explain why the Korean won exhibited an overshooting pattern in its exchange rate that the Mexican peso did not (Section 5.3). They attribute the inflation overstatement to three possible factors – abstracting from the severe Korean recession, measurement problems in the Korean CPI (including a possible “flight from quality” toward lower-quality substitutes not captured in the index), and the fact that roughly one-fifth of the CPI’s weight is in government-controlled nontradable prices such as medical care and education – and explicitly leave the depreciation-overshooting puzzle as “an important area for future research” (Section 5.3).

The authors explicitly note that a crisis financed mainly by devaluing the real value of nonindexed government debt, with little future money growth or inflation, is “closely related to the work of Cochrane (2001), Sims (1994) and Woodford (1995) on the fiscal theory of the price level” (Section 2). This connection is drawn directly from the government intertemporal budget constraint (2.6): if the government has no nonindexed liabilities, a bank bailout “would have to be financed entirely via seignorage revenues,” but if it does hold nonindexed nominal debt, a one-time devaluation can satisfy part of the constraint by revaluing that debt, “which would have potentially very different implications for money growth and inflation” than the pure-seignorage case (Section 2).

The paper concludes that the mix of financing strategies a government uses – not just whether it prints money – determines both the size of the devaluation and the path of post-crisis inflation, and that this financing-mix lens can reconcile prospective-deficit-driven crisis models with the empirical fact that large devaluations are often followed by only moderate inflation (Section 6, “Conclusion”). The authors restate their model’s four key features – a crisis triggered by prospective deficits, outstanding nonindexed government debt whose real value can be reduced through devaluation, government liabilities not indexed to inflation, and nontradable goods and distribution costs that break PPP – and conclude that, applied to their two case studies, “the Mexican government is likely to pay for the bulk of the fiscal costs of its crisis through seignorage revenues… [while] the Korean government is likely to rely more on a combination of implicit and explicit fiscal reforms” (Section 6).