A Comparison of Monetary Anchor Options for Commodity-Exporters in Latin America and the Caribbean Jeffrey A. Frankel H

1. A Comparison of Monetary Anchor Options for Commodity-Exportersin Latin America and the CaribbeanJeffrey A. Frankel Harpel Professor, Harvard University Workshop on Myths and Realities of Commodity Dependence: Policy Challenges and Opportunities for Latin America and the Caribbean,  World Bank, Sept. 17-18, 2009. Thanks to Daniella Llanos for excellent research assistance.

2. The need for a monetary anchor in LAC • Inflation rates went very high in the early 1980s, to hyperinflation in some cases. => The need for a nominal anchor was plain to see. • In a non-stochastic model, any nominal variable is as good a choice for anchor as any other nominal variable. • But in a stochastic model, not to mention the real world, it makes a big difference what is the nominal variable to which the monetary authorities publicly commit. (Rogoff, 1985) • Should it be • the money supply? • Exchange rate? • CPI? • Other alternatives?

3. How do LACA economies differ? (while acknowledging heterogeneity within the region) • lower monetary credibility than in industrialized countries; • procyclical finance; • supply shocks, and in particular • terms of trade volatility, arising from • concentration in agric. & mineral commodity exports • which are subject to volatile prices on world markets.

4. Fashions in international currency policy • 1980-82: Monetarism(target the money supply) • 1984-1997: Exchange rate targets for emerging markets • 1998-2001: The corners hypothesis • 1999-2008: Inflation Targeting (+ currency float) -- The new conventional wisdom • among academic economists • at the IMF • among central bankers.

5. The era of the exchange rate anchor • In the successful stabilization programs of the 1980s and early 1990s, the exchange rate was usually the nominal anchor. • Chile’s tablita, • Bolivia’s exchange rate target, • Argentina’s convertibility plan, • Brazil’s real plan, … and others.

6. End of the exchange rate target era • The currency crises of 1994-2002 led to the abandonment of exchange rate targets in favor of more flexible regimes, if not outright floats. • Often, cherished exchange rate anchors were abandoned under the urgent circumstances of a speculative attack • including Mexico & Argentina. • A few made the jump to floating preemptively, before a crisis could hit • Chile & Colombia. • Only a few smaller countries responded to the ever rougher seas of international financial markets by moving the opposite direction, to full dollarization • Ecuador, under pressure of crisis; & • El Salvador, out of longer-run motivations. • On a 30-year time span, the trend has been toward increased flexibility.

7. What to use as nominal target?Five candidates to fill the vacancy • The gold standard is obsolete. • Monetarism? • Enthusiasm for monetarism died down by the late 1980s, • because M1 targets had proven too restrictive in the big industrialized countries. • A surprising number of LACA countries continue officially to list money supply as their anchoring variable • Argentina, Guyana, Jamaica, & Uruguay. • One doubts in practice that they keep money in declared ranges. • Nominal income targeting offsets velocity shocks, the bane of M1 targets. • Exchange rate targeting never disappeared (“fear of floating”) • But Inflation Targeting is the new reigning champion.

8. Inflation Targeting (IT) got the job. • It was a fresh young face, • coming with an already-impressive resume of recent successes in wealthier countries . • Emerging market countries followed suit. • Three South American countries officially adopted IT in 1999, in place of exchange rate targets: • Brazil, • Chile, & • Colombia. • Mexico had done so earlier, after the peso crisis of 1994. • Peru followed in 2002, switching from official money targeting. • Guatemala has officially entered a period of transition to inflation targeting, under a law passed in 2002.

9. In some ways, Inflation Targeting has worked well. • It apparently anchored expectations and avoided a return to inflation in Brazil, for example, despite two severe challenges: • the 50% depreciation of early 1999, as the country exited from the real plan, and • the similarly large depreciation of 2002, when Lula first pulled ahead in the polls. • Giavazzi, Goldfajn, & Herrera(2005);Mishkin(2004)

10. What is the definition of IT? • It is hard to argue with IT when defined broadly: “choose a long run goal for inflation and be transparent.” • But something more specific is implied by the term. • The price target is virtually always the CPI (though sometimes “core” rather than “headline” CPI). • This paper considers other price indices that are possible alternatives to the CPI for the role of nominal anchor. • The narrow definition of IT would have central bank governors committing each year to a goal for the CPI, and then putting 100% weight on achieving that objective to the exclusion of all others.

11. I am not talking about how flexible to be, or rules vs. discretion • Some proponents make clear that they are talking about something broader: flexible inflation targeting, under which the central bank puts some weight on the output objective rather than everything on the inflation objective – as in a Taylor Rule -- over the one-year horizon. • This study does not deal with the eternal question how much weight to place in the short term on a nominal anchor, such as a price index, vs. GDP. • The central focus is, rather: whatever weight is to be placed on a nominal anchor, what should be that nominal anchor?

12. I claim the past few years, particularly the global crisis, have put strains on Inflation Targeting, much as the events of 1994-2001 earlier put strains on exchange rate targeting.

13. Three other kinds of nominal variables have forced their way into the attentions of central bankers, beyond the CPI. • One, the exchange rate, never really left • certainly not for the smaller countries. • A 2nd category, asset prices, is the most relevant recently in industrialized countries. • The financial upheaval of 2007-08 with the US sub-prime mortgage crisis has forced central bankers to re-think their focus on inflation to the exclusion of equity & real estate prices. • But a 3rd category, prices of agricultural & mineral products, is particularly relevant for LAC countries. • The high volatility of commodity prices in the past decade, culminating in the price spike of 2008, has resurrected the desirability of a currency regime that accommodates terms of trade shocks. • This 3rd challenge to CPI-targets gets the most attention in my study.

14. In practice, IT proponents agree central banks should not tighten to offset oil price shocks • They want focus on core CPI, excluding food & energy. • But • food & energy consumption do not cover all supply shocks. • Use of core CPI sacrifices some credibility: • If core CPI is the explicit goal ex ante, the public feels confused. • If it is an excuse for missing targets ex post, the public feels tricked. • The threat to credibility is especially strong where there are historical grounds for believing that government officials fiddle with the consumer price indices for political purposes. • Perhaps for that reason, IT central banks apparently do respond to oil shocks by tightening/appreciating, • as the following correlations suggests….

15. LAC Countries’ Current Regimes and Monthly Correlations of Exchange Rate Changes ($/local currency) with $ Import Price Changes Table 1 IT coun-tries show correl-ations > 0.

16. Why is the correlation between the $ import price and the $ currency value revealing? • The currency of an oil importer should not respond to an increase in the world price of oil by appreciating, to the extent that these central banks target core CPI . • If anything, floating currencies should depreciate in response to such an adverse terms of trade shock. • When these IT currencies respond by appreciating instead, it suggests that the central bank is tightening monetary policy to reduce upward pressure on the CPI.

17. Wanted ! • New candidate variable for nominal target. • Variable should be: • simpler for the public to understand ex ante than core CPI, and yet • robust with respect to supply shocks. • “Robust with respect to supply shocks” means that the central bank should not have to choose ex post between two unpalatable alternatives: • an unnecessary economy-damaging recession or • an embarrassing credibility-damaging violation of the declared target.

18. One variable that fits the desirable characteristics is nominal GDP. • Nominal income targeting is a regime that has the desirable property of taking supply shocks partly as P and partly as Y, without forcing the central bank to abandon the declared nominal anchor. • A popular proposal among macroeconomists in the 1980s. • Some critics claimed that nominal income targeting was less applicable to developing countries because of lags and statistical errors in measurement. • But these measurement problems have diminished. • Furthermore, developing countries are more vulnerable to supply shocks than are industrialized countries => the proposal is more applicable to them, not less. McKibbin & Singh (2003).

19. But nominal income targeting has not been seriously considered since the 1990s, either by rich or poor countries. • Thus it is not be analyzed in this paper. • Fortunately, nominal income is not the only variable that is more robust to supply shocks than the CPI. • I favor some alternative price indices for targets. • To understand the argument, one must first recognize the importance of the external accounts in developing countries: • Small countries are more dependent on trade than big countries. • Those specialized in mineral & agricultural products, as LAC, experience more volatile terms of trade, vs. industrialized countries. • Emerging market countries tend to experience pro-cyclical finance, • not the finance of theory, which willingly smoothes adverse trade shocks. • Often international capital, if anything, exacerbates trade swings.

20. Trade shocks • If the supply shocks are terms of trade shocks, then the choice of CPI to be the price index on which IT focuses is particularly inappropriate. • Alternative? An output-based price index such as the PPI, GDP deflator, or an export price index. • The important difference is that • import goods show up in the CPI, but not in the output-based price indices, • and vice versa for export goods: they show up in the output-based prices but much less in the CPI.

21. Terms of trade volatility is particularly severe for commodity exporters, • which includes most countries in LAC. • Table 2 reports the leading export commodity for each of 20 LAC countries, and the standard deviation of the $ price of that commodity on world markets. • Natural gas & oil are the most variable. • But the prices of aluminum, bananas, coffee, copper, & sugar all show standard deviations above .4; • => price swings of + or - 80% occur 5% of the time.

22. Major Commodity Exports in LAC countriesand Standard Deviation of Prices on World Markets Source: Global Financial Data

23. An exchange rate target is still a valid option, • -- even a peg or dollarization for very small, open countries meeting the OCA criteria. • Larger countries that “float” often find it useful to intervene somewhat, to dampen fluctuations around a central parity. • Need to weigh advantages of fix vs. float. • If a country adopts an exchange rate target, $ ? € ? basket ?

24. Inflation Targeting • based on the CPI (standard IT) ? or • based on other price indices • PEP: Peg the Price of the leading export mineral or agricultural commodit(ies) • PEPI: Target a comprehensive export price index • Target PPI (Producer Price Index)

25. Proposal toPeg theExport Price (PEP) • Intended for countries with volatile terms of trade, particularly those specialized in the production of mineral or agricultural commodity exports. • Proposal: The authorities peg the currency to a basket or price index that includes the price of their leading commodity export (oil, gold, copper, coffee…), rather than to the $ or € or CPI. • My claim is that the regime combines the best of both worlds: • The advantage of automatic accommodation to terms of trade shocks, together with • the advantages of a nominal anchor and integration.

26. How would it work operationally, say, for an oil-exporter? • Each day, after noon spot price of oil in NYC is observed, S($/barrel), the central bank announces the day’s exchange rate, according to the formula: • E (peso/$) = fixed target price P (peso/barrel) / S($/barrel). It intervenes in $ to hold this exchange rate for the day • The result: P (peso/barrel) is indeed fixed from day to day.

27. Does floating give the same answer? • True, commodity currencies tend to appreciate when commodity markets are strong, & vice versa: • Australian, Canadian & NZ $(e.g., Chen & Rogoff) • South African rand(e.g., Frankel, 2007) • Chilean peso and others • But • Some volatility under floating appears gratuitous. • Floaters still need a nominal anchor.

28. A less radical form of the proposal:PEPI, for Peg the Export Price Index • Some have responded to the PEP proposal by pointing out a side-effect of stabilizing the local-currency price of the export commodity in question: destabilizing the local-currency price of other export goods. • For most countries, no good is more than half of exports. • Moreover, countries may wish to encourage diversification away from traditional mineral or agricultural export. • For them, the strict version of PEP is not appropriate. A moderated version is more suitable: • PEPI: Target a broad index of export prices, rather than the price of only one export commodity. .

29. A proposal more moderate still: you can call it IT, but target PPI instead of CPI • I.e., target a broad index of all domestically produced goods, whether exportable or not. • The central bank in practice could not hit a PPI target, or a broad export price index, exactly, • in contrast to the way it could hit exactly a target for the exchange rate, the price of gold, • or even the price of a basket of 4 or 5 ag. or mineral commodities. • There would instead be a declared band for the PPI target, which could be wide if desired, just as with the targeting of the CPI, money supply, or other nominal variables. • Open market operations to keep the price index inside the band could be conducted in terms of either foreign exchange or domestic securities.

30. Regimes Historical reality Counterfactuals $ peg € peg SDR peg CPI target PEP (Peg Export Price) PEPI (Peg Exp.P.Index) PPI target Countries Argentina Bolivia Chile Colombia Costa Rica Ecuador Guatemala Honduras Jamaica Simulations of prices under 7 monetary anchors, alternatives to historical reality • Mexico • Nicaragua • Panama • Peru • Paraguay • El Salvador • Trinidad • Uruguay • Venezuela

31. Define the CPI & PPI each as weighted averages of prices in four sectors, in logs: CPI = wntg Pntg + wcx Pcx + wm Ppm + wotg Potg PPI = vntg Pntg + wcx Pcx + wm Ppm + wotg Potg • Pntg≡ price of nontraded goods in local terms. We assume that, at a horizon of less than 1 year, these prices are sticky. • Pcx ≡ price of exports of leading mineral/agr. commodity in local terms. We define these TG prices to equal the actual historically observed world $ prices, times the exchange rate, which will differ depending on the monetary regime assumed. • Ppm ≡ price of petroleum product imports, determined again as actual world dollar price times the simulated exchange rate. • Potg ≡ price of other traded goods, e.g., manufactures. • We estimate the weights for each country.

32. Estimation for each country of weights in national price index on 3 sectors: non tradable goods, leading commodity export, & other tradable goods Argentina is relatively closed; Mexico is relatively open. As expected. The leading export commodity usually has a higher weight in the country’s PPI than in its CPI, as expected.

33. Price of copper in Chile --Simulations under 6 alternative anchors Chile • Any nominal target would have eliminated high 1970s inflation. • PPI target stabilizes domestic copper price better than does CPI.

34. Table 4: Variability of Export Prices under Alternative Currency Regimes(a) Standard Deviation of Nominal Export Prices

35. Historical $ € SDR commodity CPI PPI peg peg peg peg peg peg

36. Begin with a hypothetical peg to the $. • Notice: same as historical peg in the case of Panama. • In the other cases, we can simulate precisely what the price of soy, copper, etc. would have been in terms of pesos under the counterfactual, • by using the historical series for the exchange rate between the peso and the $: • if the peso historically depreciated against the $ by 1% in some given month, we know the price of soy would have been lower by precisely 1% if the peso had instead been pegged to the $. • In general, the $ pegs would have produced far more stable prices in domestic terms. • This is true of all six nominal anchors, and simply illustrates the tremendous price instability experienced in the 1970s & 80s. • Changes in prices are also more stable under the anchors (Table 4b).

37. The next two columns of Table 4a show what variability of the commodity export prices would have been under an SDR peg or € peg, respectively. • Variability of the domestic price of the commodity export is often lower under the € peg than under the $ peg: for natural gas & oil; iron & steel; copper, aluminum & gold; bananas & sugar; and soy & beef. • A point often missed by observers who read too much into the invoicing of international commodity trade in $: • While the use of the $ may introduce some dollar-stickiness in the very short run, it does not carry over to the medium run. • When the effective foreign exchange vale of the $ rises, $ prices of these commodities tend to fall rather quickly. • The offset is not fully proportionate; but the point is that the prices are not more stable in terms of $ than in terms of €. • Table 4(a) shows that in some cases (soy, coffee & beef), the basket offered by the SDR would stabilize commodity prices better than either the $ or €.

38. The last two columns: comparison of effects of a PPI target & a CPI target, intended as the unique contribution of this study. • The comparison in terms of ability to stabilize domestic prices of the principle export commodities appears in Table 4(a). • Usually the standard deviation of the domestic price of the export commodity is lower under the PPI target than under the CPI target. • In a few cases, it is less than half the size: • Jamaica for aluminum and Uruguay for beef.

39. Stabilizing domestic prices with respect to the export commodity is far from the only criterion that should be considered in comparing alternative candidates for nominal anchor. • Another one is stabilizing domestic prices of other tradable goods. • A valid critique of PEP and PEPI is that it transfers uncertainty that would otherwise occur in the real price of commodity exports into uncertainty in the real price of non-commodity exportables and importables. • This critique is particularly relevant if diversification of the economy is valued. • In Table 5 we conduct simulations, under the same seven alternative regimes, for the domestic prices of import goods.

40. We conclude with the logic that stabilizing the relative price of commodity exports is not much of an accomplishment if it comes at the expense of a corresponding destabilization of the relative price of other traded goods. • Table 6 examines the implications of the alternative regimes for an objective function that is a weighted average of the standard deviation of the real price of the commodity export and the standard deviation of the real price of the import good.

41. Real Prices** Historical Dollar Comm. CPI PPI SDR Peg Euro Peg Regime Peg Peg Target Target ARG 0.661 0.491 0.503 0.486 0.241 0.756 0.679 BOL 0.538 0.443 0.457 0.486 0.138 0.488 0.448 BRA 0.522 0.456 0.442 0.426 0.187 -- -- CHL 0.510 0.485 0.489 0.470 0.298 0.840 0.696 COL 0.456 0.485 0.482 0.490 0.000 1.123 0.974 CRI 0.420 0.368 0.383 0.385 0.242 -- -- ECU 0.456 0.485 0.482 0.490 0.000 -- -- GTM 0.510 0.588 0.600 0.585 0.383 -- -- GUY 0.922 0.581 0.579 0.557 0.383 -- -- HND 0.533 0.588 0.600 0.585 0.383 -- -- JAM 0.338 0.383 0.401 0.403 0.212 0.870 0.483 MEX 0.479 0.485 0.482 0.490 0.000 0.975 1.030 NIC 0.511 0.588 0.600 0.585 0.339 -- -- PAN 0.312 0.368 0.383 0.385 0.206 -- -- PER 0.444 0.485 0.489 0.470 0.171 0.420 0.429 PRY 0.413 0.455 0.475 0.466 0.312 0.743 0.716 SLV 0.750 0.588 0.600 0.585 0.383 -- -- TTO 0.383 0.443 0.457 0.486 0.179 -- -- URY 0.504 0.455 0.475 0.466 0.312 0.793 0.525 VEN 0.429 0.485 0.482 0.490 0.000 -- -- Table (6c): Standard Deviation of Real PricesExport Price SD & Import Price SD Averaged Average of leading commodity export standard deviation & oil price standard deviation. Real prices of TGs in general would be more stable under PPI target than CPI target.

42. The commodity price peg (PEP) wins the competition to reduce relative price variability • by a fairly substantial margin • when we look at the level of nominal prices (Table 6a) • or the level of real prices (Table 6c) ; and • by a smaller margin when we look at nominal price changes(Table 6b). • The three currency pegs are again similar to each other, showing less price variability than the historical regime but more than the commodity peg. • The PPI target usually gives less relative price variability than the CPI target. • Looking at real price variability in Table 6c, • the only exception is Peru; • the gain is substantial in the case of Jamaica and Uruguay, • smaller for the others.

43. Summary of results • Table 1: ever since 1999, when Brazil, Chile & Colombia switched from exchange rate targets to CPI targets, they have experienced a higher correlation between the $ price of their currencies and the $ price of oil imports. • => language about core CPI notwithstanding, the monetary authorities in the IT countries have responded to increases in oil import prices by contracting monetary policy enough to appreciate their currencies. • The production-based price targets (PEP, PEPI, PPI target) would not have this problem.

44. Summary of results, continued • Comparison of 6 alternative nominal targets according to how they would affect the variability of real tradables prices: • commodity exports in Table 4, • imports in Table 5, • and both together in Table 6. • Some conclusions are predictable. • First, according to the simulations the currency anchors offer far more price stability than does the historical reality. • Second, PEP perfectly stabilizes the domestic price of export commodities, by construction.

45. Summary of results, concluded • The more interesting findings: comparison of a CPI target and a PPI target as alternative interpretations of inflation targeting. • The results show that the PPI target generally delivers more stability in the prices of traded goods, especially the export commodity. • This is a consequence of the larger weight on commodity exports in the PPI than in the CPI.

47. Appendices • Alternative nominal anchors • The current choices of LAC countries • Achilles heels of 6 candidate variables • Fixed vs. floating exchange rates • Pros and cons • Criteria for the choice • What is the market failure addressed? • Graphs of nominal & real log export prices, simulated under 7 alternative regimes

48. LAC Countries’ Current Regimes Source: IMF De Facto Classifications of Exchange Rate regimes and Monetary Policy approach (http://www.imf.org/external/np/mfd/er/2006/eng/0706.htm) and Global Financial Data.

49. 6 proposed nominal targets and the Achilles heel of each:

50. Appendix 2 • Advantages of fixed exchange rate • Advantages of floating • Criteria to add up costs & benefits • The corners hypothesis