Sergio Rebelo | Macro Paper Warehouse

Government finance in the wake of currency crises

This paper asks two questions: how do governments actually pay for the fiscal costs associated with currency crises, and what do different financing methods imply for post-crisis rates of inflation and depreciation? The authors note that the classical view – that governments print money to finance ongoing or prospective deficits, so that seigniorage should rise sharply after a crisis – is inconsistent with the data on six recent currency crises (Introduction, Table 1). To address these questions they build a small open-economy general-equilibrium model, populated by a representative infinitely-lived agent who consumes tradable and nontradable goods, in which a currency crisis is triggered by prospective government deficits rather than by ongoing money creation. Unlike standard models, which typically assume seigniorage is the government’s only source of depreciation-related revenue, this model gives the government a menu of such revenue: seigniorage, “debt deflation” (the fall in the dollar value of nonindexed local-currency debt), and an “implicit fiscal reform” (an automatic fall in the dollar value of government transfers and purchases that are not indexed to the exchange rate). The model also incorporates nontradable distribution costs (retailing, wholesaling, transportation) needed to sell tradable goods at retail, and brief nominal stickiness in the price of nontradables, both of which the authors show are needed to generate a first-year rate of depreciation that substantially exceeds the first-year rate of inflation, as observed in the data. Calibrated to Korean data, the model can account for high depreciation together with low inflation, with seigniorage covering less than 15 percent of Korea’s fiscal costs. The authors then use the model in conjunction with fiscal data to construct detailed case studies of how Korea (1997), Mexico (1994), and Turkey (2001) actually financed the costs of their currency and banking crises. They find that, in all three countries, seigniorage has been a minor source of financing to date (7-10 percent of net crisis costs), debt deflation has exceeded seigniorage, and large declines in the dollar value of government transfers – and, in Korea and Mexico, this was the single most important channel – were central sources of revenue, though none of the three countries had, as of the data available (through 2001-2002), fully financed the fiscal costs of its crisis. The authors caution that their precise quantitative decompositions rely on debatable counterfactual assumptions about what fiscal aggregates would have done absent the crisis, but express confidence in the qualitative patterns, and note the paper is silent on why different governments chose different financing strategies.

Large Devaluations and the Real Exchange Rate

This paper asks why large devaluations are typically followed by large, persistent declines in the real exchange rate (RER) rather than being quickly offset by domestic inflation. The authors argue that the primary force behind this pattern is slow adjustment in the price of nontradable goods and services, not slow adjustment in the price of goods that are actually imported or exported (Section 1, Introduction). Using data from five large devaluation episodes – Argentina (2001), Brazil (1999), Korea (1997), Mexico (1994), and Thailand (1997) – they show that at-the-dock import and export prices move roughly in line with the nominal exchange rate, so relative purchasing power parity (PPP) is a reasonable description of “pure-traded-goods” prices even during large devaluations; it is the retail price of nominally tradable goods, and the price of nontradable goods and services, that adjust slowly. The key insight the authors offer for why the traditional tradable/nontradable CPI split is misleading is that the retail price of a tradable good embeds two nontradable components – distribution costs (wholesaling, retailing, marketing, and local transport, estimated at roughly 50 percent of the retail price) and “local goods” that are conventionally classified as tradable but are in practice produced and consumed only domestically. Once the CPI basket is re-decomposed to account for these components, a series of “price-accounting” exercises show that the resulting predicted inflation rates are close to actual post-devaluation inflation in each country, and an Engel (1999) decomposition of the RER confirms that the great majority of the RER’s post-devaluation movement reflects the price of nontradables relative to pure-traded goods, not movements in the relative price of pure-traded goods themselves. A detailed case study of Argentina’s 2001 devaluation, using disaggregated CPI data, a purpose-built survey of prices in Buenos Aires, and supermarket scanner data, corroborates these findings and documents that retail-goods prices adjust far more frequently than services prices, and that there was substantial “flight from quality” toward cheaper goods after the devaluation. The authors test the robustness of their findings against two large real-appreciation episodes (Argentina 1991-95, Mexico 1988-94), four medium-size devaluations (Finland, Italy, Sweden, and the UK in 1992), and small business-cycle-frequency exchange-rate fluctuations in ten OECD countries (1971-2001), finding that the nontradables-relative-to-pure-traded-goods channel remains important in the medium and small cases too, though relatively less dominant than in the large-devaluation episodes, where fluctuations in pure-traded-goods prices instead account for most of the movement.

Markups Across Space and Time

Anderson, Rebelo, and Wong study the behavior of markups in the retail sector across regions and over time, using a combination of firm-level Compustat data and product-level scanner data from two large retailers — one operating over 100 stores across U.S. states (quarterly data from 2006 Q1 to 2009 Q3, covering roughly 3.6 million SKU-store pairs across 79 product categories) and one operating hundreds of stores across Canadian provinces (quarterly data from 2016 Q1 to 2018 Q4, covering 15.6 million item-store pairs across 41 product groups). Markups are measured using gross margins — sales minus cost of goods sold as a fraction of sales — computed at the product level using the replacement cost for every item. This measurement approach is appropriate for retail because cost of goods sold accounts for over 80 percent of total retail firm costs, making it a reliable proxy for marginal cost. The replacement cost data, available at the store level, is the cost used by managers in actual pricing decisions, distinguishing these datasets from typical scanner data that contain only average costs.

On the Fiscal Implications of Twin Crises

This paper, a companion analysis to Burnside, Eichenbaum, and Rebelo’s “Prospective Deficits and the Asian Currency Crisis,” studies how governments actually pay for the fiscal costs of “twin” currency-and-banking crises and what different financing choices imply for post-crisis inflation. The motivating puzzle is that the classical first-generation view of currency crises – that governments print money to finance deficits, a view especially appealing for twin crises because bank bailouts create large fiscal costs that are not typically financed by big explicit fiscal reforms – has an important empirical shortcoming: it predicts high post-crisis inflation, whereas many large devaluations are in fact followed by only moderate money growth and inflation. The paper lays out five strategies a government can combine to pay for a twin crisis’s fiscal costs: an explicit fiscal reform (raising taxes or cutting spending), explicit default on debt, printing money for seignorage, deflating the real value of outstanding nonindexed nominal debt through the devaluation itself (“debt devaluation”), and an implicit fiscal reform that lets inflation erode the real value of government outlays that are fixed, at least temporarily, in nominal terms (such as civil-service wages). Working with a reduced form of the Burnside-Eichenbaum-Rebelo (2001) model – a Cagan-style money-demand specification, a government budget constraint, a debt-threshold exit rule for abandoning the fixed exchange rate, and an assumption about post-devaluation monetary policy – the authors show that a benchmark version in which purchasing power parity (PPP) holds and all liabilities are fully indexed predicts far too little devaluation and far too much inflation relative to what actual crises show. Introducing nonindexed government liabilities (pre-crisis nominal bonds and nominal spending commitments) lets debt devaluation and implicit fiscal reform substitute for seignorage, reducing the model’s implied inflation, but this extension alone still cannot generate devaluations as large as those observed. Eliminating PPP – by adding nontradable goods (with sticky short-run prices), costs of distributing tradable goods, and the resulting wedge between the exchange rate and the CPI – breaks the one-to-one link between devaluation and inflation and lets the model reproduce both the large devaluations and the comparatively moderate inflation actually seen after crises. Applying the model to Mexico’s 1994 crisis and Korea’s 1997 crisis, the authors estimate that Mexico has financed, and is likely to continue financing, the bulk of its roughly 15-percent-of-GDP fiscal cost through seignorage – consistent with Mexico’s relatively high post-crisis inflation – while Korea, whose crisis is estimated at roughly 24 percent of 1997 GDP, is likely to rely mostly on a combination of future implicit and explicit fiscal reforms, consistent with Korea’s much lower post-crisis inflation. The authors acknowledge that the model overstates first-year Korean inflation (about 15 percent modeled versus roughly 7 percent actual) and cannot account for the difference between Korea’s exchange-rate overshooting pattern and the absence of overshooting in Mexico.

Prospective Deficits and the Asian Currency Crisis

This paper argues that the 1997 Asian currency crisis is best explained not by the standard first-generation account of ongoing fiscal deficits monetized until reserves run out, nor by a purely self-fulfilling multiple-equilibrium panic, but by large prospective government deficits – implicit bailout guarantees to failing banking systems that agents came to expect would eventually be financed, at least in part, by seignorage revenues. The paper states its motivating puzzle plainly: the governments of the crisis countries (Indonesia, Korea, Malaysia, the Philippines, and Thailand) were running either surpluses or small deficits in the years before the crisis, so the conventional deficit-based story does not fit. To formalize the alternative, the authors build a continuous-time, perfect-foresight small open economy model with a representative agent, a government subject to an intertemporal budget constraint, and a cash-in-advance money demand; at time zero agents learn that future government transfer payments (the bailout) will permanently rise after some future date, and the government’s present-value budget constraint implies that a speculative attack – the abandonment of the fixed exchange rate – becomes inevitable regardless of the government’s initial foreign reserves or debt position. The government is assumed to abandon the peg according to a debt-threshold rule, and the paper’s key methodological move is to distinguish the date of the attack from the (generally later) date at which the government actually implements the new, revenue-raising monetary policy. Calibrated separately to 1996 Korean and Thai data, the model generates a speculative attack (t*=2.27 years for Korea, t*=2.42 for Thailand) that occurs after agents learn of the higher prospective deficit but before the new monetary policy is put in place, so an observer looking only at past deficits, money growth, and inflation would see no warning signs and might wrongly conclude the crisis was a multiple-equilibrium, self-fulfilling event. The paper further shows that the government can delay the attack by borrowing more (raising the debt threshold), but only at the cost of higher future inflation once the attack occurs, so that under the model’s assumptions (no nominal rigidities) the policy that minimizes distortion is to abandon the fixed rate as soon as the news arrives. Two pieces of cross-country empirical evidence are offered in support of the mechanism: non-performing-loan-based estimates showing large increases in governments’ implicit liabilities concentrated in Korea and Thailand, and stock-market-based measures showing that the value of the financial sector had been declining, in both absolute and relative terms, well before the currency crisis in Korea, Thailand, and to a lesser extent Malaysia and the Philippines. The authors are explicit that the analysis abstracts from nominal rigidities, treats the state of the banking system as revealed to agents all at once rather than gradually, models the exit decision with a simple debt-threshold rule rather than an explicitly political process, and does not address how the collapse of financial intermediaries itself affects future output and tax revenue.

When Is the Government Spending Multiplier Large?

This paper argues that the government-spending multiplier can be much larger than one when the nominal interest rate is constant – most naturally because the zero lower bound on nominal interest rates binds – and that the larger the fraction of new spending that arrives while the rate is stuck at zero, the larger the multiplier. The authors build the argument in three steps. First, in a simple new-Keynesian model without capital in which the central bank follows a Taylor rule, the government-spending multiplier is modest, generally close to or a little above one for plausible parameters, because a fiscal expansion that raises output and expected inflation triggers an interest-rate response that crowds out private spending. Second, when the nominal rate is instead held constant – the natural case being a large, temporary rise in the representative household’s discount factor (a stand-in for a rise in the desire to save) large enough to push the zero bound into a binding state – an increase in government spending instead counteracts a self-reinforcing deflationary spiral: higher expected inflation lowers the real interest rate, which raises private spending, output, and inflation further, in contrast to the Taylor-rule case where the same channel runs the other way. In their baseline calibration the zero-bound multiplier is 3.7, roughly three times the standard multiplier, and the multiplier is shown analytically to be larger precisely in economies where the output cost of being stuck at the zero bound is itself larger; it is also larger the longer government spending is expected to arrive while the bound still binds, and shrinks sharply if the spending instead arrives only after the bound stops binding. Third, quantifying the mechanism in the medium-scale, estimated DSGE model of Altig, Christiano, Eichenbaum, and Lindé (2011), which adds sticky wages, habit formation, variable capital utilization, and investment adjustment costs to the simple model, the authors find an impact multiplier of roughly 1.6 and a peak multiplier of about 2.3 when spending is elevated for twelve quarters while the zero bound binds, compared with less than one under an active Taylor rule; the same model, hit with a discount-factor shock and a financial-intermediation-cost shock calibrated to 2008-2010 credit spreads, reproduces the broad shape of the 2008-2010 collapse in output, consumption, investment, and inflation and the fall of the federal funds rate to zero. The paper further shows that financing the spending increase with distortionary labor-income or household-borne capital-income taxes raises, rather than lowers, the multiplier while the zero bound binds, the opposite of the standard result away from the bound, and that the same logic implies it can be socially optimal to raise government spending very substantially – in the authors’ calibration to roughly 30 percent of GDP from a steady-state 20 percent – when the zero bound binds. The authors are explicit that their analysis abstracts from political-economy considerations bearing on the difficulty of reversing new spending programs, and that reduced-form estimation of the zero-bound multiplier from historical data is largely infeasible because multipliers differ so much across regimes and government spending itself typically rises endogenously in response to the same shocks that make the bound bind.