Contractionary Currency Crashes In Developing Countries

1
Contractionary Currency CrashesIn Developing
Countries

• Mundell-Fleming Lecture, Nov. 5, 2004Fifth IMF
  Annual Research Conference.Forthcoming, IMF
  Staff Papers, 2005.
• IMF Institute, May 27, 2005.
• Jeffrey A. FrankelHarpel Professor
• KSG, Harvard University

2
Why are devaluations so costly?

• Political costs
• Economic costs
• If the answer is contractionary effects of
  devaluation,
• what is the mechanism?
• And what can a country do to minimize them?

3
After a devaluation, leaders lose their jobs

• Twice as often within 1 yr. (30)
• as in control group (14) in 1960s.
• R.Cooper, 1971. (Criterion is 10 deval.)
• Updated to 1970-2003,
• within 12 mo.s of devaluation (27), vs,
  otherwise (21)
• within 6 mo.s (19), vs. as otherwise (12).
• Difference is statistically significant at 0.5
  level
• (Criterion is 25 deval, incl. 10 acceleration.)

4
Premier 2/3 more likely to lose his job within 6
mos. of currency crash
1970-2003 Within 6 mos. Of devaluation Other periods
Change in premier 36 (19.1) 812 (11.6)
No change in premier 153 (81.0) 6192 (88.4)
Total cases 189 7004
5
Fin.min.s / governors are even more likely to
lose office in yr.
6
Political price of devaluations in developing
countries
7
Citations (1972-2003)

• Of Cooper (1971) 84
• Of Mundell (1963) 319
• Of Fleming (1962) 257

8
Possible reasons currency crashes correlate with
change of leadership

• Elections cause devaluations, rather than the
  other way around (tho our timing is post-deval.)
• Devaluation as proxy for unpopular IMF austerity
  programs
• The government is perceived as having broken a
  promise
• High economic costs associated with devaluations.

9
Conditioning on IMF program does not affect
leader turnover

• Dummy variable defined for initiation of IMF
  program within 3 months of currency crash
• does not raise frequency of leader job loss,
  relative to other devaluations.
• Thus conditioning on the IMF dummy variable has
  no discernible effect on frequency of leader job
  loss within 6 mo.s
• 21.1 with an IMF program, vs.
• 21.9 without.
• Both similar to overall rate of job turnover
• following devaluations (22.8) in full sample --
• still almost double the 11.6 rate in normal
  times.

10
Executives twice as likely to lose their jobs if
the government had said no devaluation.
Leader was changed within 12 mos unchanged within 12 mos Frequency of changes
Of 6 promises (whether by premier, fin min, or CB) 4 2 2/3
Of 15 with no prior promises not to devalue 5 10 1/3
Total 21 case studies 9 12
11
Executives gt twice as likely lose their jobs if
the government had said no devaluation.
Leader was Changed within 6 mo.s Unchanged within 6 mo.s Frequency of changes
Of 6 promises (whether by premier, fin min, or CB) 3 3 .5
Of 15 with no prior promises not to devalue 3 12 .2
Total 21 case studies 6 15
12
Narrowing down source of political costs of
devaluation

• As noted, IMF programs make no difference.
• Even in those cases where no assurances had been
  given over preceding month, rate of job loss
  (20) still gt no devaluation cases (11.6)
• (or 33 at 12-month horizon gt 20.5) .
• Thus, although broken promise effect is there,
  political costs must also reflect economic pain.

13
Why did output fall sharply in many recent crises?

• Excessive expenditure-reduction (Sachs
  Stiglitz)?
• Excessive fiscal contraction
• Excessive monetary contraction (high i gt
  default)
• Versus what alternative?
• More expenditure-switching instead?
• Devaluations were very large as it was.
• Magical relaxation of external finance
  constraint?
• Who pays?
• Doesnt change the graph
• But if devaluations are contractionary,
  
• internal balance line slopes the other way
• New intersection is hard to find
• May lose output regardless the policy mix.

14
Textbook model right combination of i E
should attain new adverse external finance
constraint, without recession
15
1990s version External balance line slopes down,
like internal balance line
16
Possible contractionary effects of devaluation

• Some require rapid passthrough, from exchange
  rate to prices of either
• imported inputs
• consumer imports, so W
• All TGs, so the distribution of income
• CPI, so M/CPI
• Other effects do not need passthrough
• Balance sheet effect

17
The passthrough coefficient has fallen in
developing countries.

• Slow/incomplete passthrough of exchange rate
  changes has long been a property of US market
  other large rich countries.
• After large devaluations in Asia and other
  emerging market countries from 1994 (Mexico) to
  2001 (Argentina), most observers feared
  correspondingly large rises in local currency
  prices.
• It never happened.

18
Passthrough is greatest for prices of imports at
dock, but less for retail and CPISource
Frankel, Parsley Wei (2004) effect within one
year
19
Passthrough for less developed countries gt for
rich, historically. Source Frankel, Parsley
Wei (2004) effect within one year
20
Passthrough to prices of 8 narrow imported
products

• Coefficient initially higher for developing
  countries (.8) than rich (.3), in 1990
• Downward trend in coefficient during decade
• Significant, regardless of income level
• Twice as fast for developing countries as for
  rich
• Partly explained, for rich, by rising real wages
• Speed of adjustment (ECM)
• Initially higher for developing countries than
  rich
• Significant downward trend for developing c.s
  (only)

21
(No Transcript)
22
(No Transcript)
23
Balance sheet effect is most important of the
contractionary devaluation effects

• Analytical literature on balance sheet effect
  includes
• Kiyotaki Moore (1997), Krugman (1999), Aghion,
  Banerjee Bacchetta (2000), Cespedes, Chang
  Velasco (2000, 2003), Chang Velasco (1999),
  Caballero and Krishnamurty (2002), Christiano,
  Gust Roldos (2002), Dornbusch (2001), Mendoza
  (2002), Calvo, Izquierdo, Talvi (2003), Cavallo
  (2004), Eichengreen Hausmann (1999) and many
  others.
• Empirical evidence of its output cost includes
• Cavallo, Kisselev, Perri and Roubini (2002)
• Guidotti, Sturzenneger and Villar (2003)

24
Foreign-denominated debt in crises x real
devaluation gt Output loss Source M.Cavallo,
K.Kisselev, F.Perri, N.Roubini, 2002.
25
Some suggested influences on the balance sheet
effect

• Eichengreen Hausmann (1999) made the structural
  inability to borrow in local currencies famous
  under the name original sin.
• Is it historical fate?
• Some argue that the culprit is adjustable pegs.
• Short run During the procrastination phase,
  composition of debt typically shifts to .
• Medium run Trade openness affects vulnerability
  to currency crashes.

26
Procrastination phase period after reserves
peak, but before the speculative attack

• Typically lasts 6-13 mo.s Frankel Wei (WB,
  2005)
• E.g., Mexico, 1994
• Four ways to gamble for resurrection
• Officials announce wont devalue
• Run down reserves
• Shift composition of debt to short-term
• Shift composition to -denominated
• Each ploy may gain time, but raises cost if
  devaluation comes.

27
Exhaustion of Mexicos ReservesUp to December
1994 Crisis Data source IMF International
Financial Statistics.
28
Mexicos shift to -linked debt, in 1992-94
Data source Mexican Ministry of Finance and
Public Credit.
29
Mexicos shift to short-term debt in 1992-94
Data source Mexican Ministry of Finance and
Public Credit.
30
Lesson

• After inflows cease, adjust to new external
  balance early
• while it is still possible to maintain internal
  balance
• Rather than procrastinating by running down
  reserves shifting to s.t. debt.
• Because after the crash, any combination of high
  i devaluation may be contractionary.

31
In long run, how does openness to trade affect a
countrys vulnerability?

• Some say openness raises vulnerability to sudden
  stops
• Vulnerability to real shocks, financial shocks
  loss of trade credit
• Empirical support
• Milesi-Ferretti Razin (1998, 2000)
• Others say it reduces vulnerability
• Sachs (1985) explains Asian resilience, vs.
  Latin America
• Trade is a hostage that reassures creditors
• Eaton Gersovitz (1981), Rose (2002)
• If m is lower, cost of adjustment to given
  cut-off is higher
• measured by necessary expenditure-reduction at
  given prices
• or by necessary devaluation (with corresponding
  contractionary effects)
• Argentina is favorite example of toxic
  interaction of low trade/GDP and balance sheet
  effect
• Guidotti, Sturzenegger Villar, (2003), Calvo
  Talvi (2004), Desai Mitra (2004), Cavallo
  (2004), and Secy. Paul ONeill.
• Empirical support
• Calvo, Izquierdo Mejia (2003) Edwards
  (2004a,b)

32
Countries of currency crashes tend to be less
open to trade, especially those with sudden stops
as well
33
Countries that are less open to trade are more
prone to sudden stops currency crashes
34
Endogeneity of trade/GDP

• Possible channels of endogeneity of openness
• Via income richer countries tend to liberalize
• Part of a more general reform strategy driven by
  pro-globalization philosophy or Washington
  Consensus forces.
• Experience with crises -- the dependent variable
  -- may itself cause liberalization, via an IMF
  program.
• Feedbacks betw. trade financial openness.
  Aizenman (2003)
• Solution Use gravity as IV for trade
• Does Openness to Trade Make Countries More
  Vulnerable to Sudden Stops, Or Less? Using
  Gravity to Establish Causality Cavallo
  Frankel (2004)

35
Predicting sudden stops, while treating
endogeneity of tradeSource Cavallo Frankel
(2004)
Ordinary probit IV (gravity)
Trade openness -0.53 (0.259) -2.45 (0.813)
Foreign Debt/GDP t-1 -0.080 (0.217) 0.196 (0.275)
Liability Dollarization t-1 (0.316 (0.195) 0.591 (0.256)
CA/GDP t-1 -4.068 (1.297) -7.386 (2.06)
observations 778 1062
36
Predicting currency crashes, while treating
endogeneity of tradeSource Cavallo Frankel
(2004)
Ordinary probit IV (gravity)
Trade openness -0.37 (0.21) -0.97 (0.59)
Foreign Debt/GDP t-1 0.35 (0.18) 0.57 (0.23)
Liability Dollarizationt-1 0.04 (0.20) -0.11 (0.18)
observations 798 1196
37
S Stops crashes rise also when it is
gravity-fitted openness that is low
38
Conclusion

• An increase in trade openness of 10 percentage
  points decreases likelihood of a sudden stop
  (definition of Calvo, et al.) by approximately
  32.
• E.g., moving from Argentinas trade share (.2) to
  Australias (.3)
• CA/GDP is the other strongest predictor
• Increase in openness also decreases the
  likelihood of currency crash, defined as 25
  increase in exchange market pressure (exchange
  rate reserves)
• Foreign debt/GDP or Reserves/imports become the
  other strong predictors
• Also some evidence that openness reduces the
  output cost associated with crises

39
Take-aways

• Leaders are twice as likely to lose office after
  a devaluation.
• Passthrough has declined sharply.
• To minimize contractionary effects of
  devaluation, countries should
• avoid weakening of their balance sheets during
  procrastination phase of debt cycle
• in the long term, open to trade.

40
The highway analogy

• Superhighways are useful, get you where you want
  to go quicker. But accidents at high speeds are
  more likely fatal. -- Merton
• Sudden stops
• Its not the speed that kills, its the sudden
  stops Dornbusch
• ? sharp disappearance of private capital
  inflows, reflected (esp. at 1st) in reserve
  declines, (soon) in disappearance of previously
  large CA deficit - Calvo
• Superhighways Modern financial markets get you
  where you want to go fast, but accidents are
  bigger, and so more care is required. -- Merton
• Is it the road or the driver? Even when
  multiple countries have accidents in the same
  stretch of road, their own policies are also
  important determinants its not just the fault
  of the system. Summers
• Contagion is also a contributor to multi-car
  accidents.

41
Highway analogy, continued

• Moral hazard -- G7/IMF bailouts reduce impact of
  given crisis
• in the LR undermine the incentive for investors
  borrowers to be careful. Like air bags and
  ambulances.
• But to claim that moral hazard means we should
  abolish the IMF would be like claiming that
  drivers would be safer with a spike in the center
  of the steering wheel column Mussa.
• Optimal sequence Highway off-ramp should not
  dump high-speed traffic into center of a village
  before streets are paved, intersections
  regulated, pedestrians learn to walk on
  sidewalks. So a country with a primitive
  domestic financial system perhaps should not open
  to the full force of international capital flows
  before domestic reforms prudential regulation.
  
• Slowing down There may be a role for controls on
  capital inflow (speed bumps posted limits).
• Reaction time How driver reacts in short
  interval between appearance of hazard and moment
  of impact (speculative attack) influences
  outcome. Adjust, rather than procrastinate by
  using up reserves switching to short-term
  debt JF
• Routine defensive driving Keeping high reserves
  and an economy open to trade is like leaving
  ample following-distance.