Asymmetric Exchange Rates and Unofficial Exchange Rate Interventions: The Case of South Korea

Abstract

Applying Milton Friedman's ‘plucking’ model of output fluctuations, we investigate the behavior of the Korean won/dollar exchange rate using a state-space model with Markov switching, which incorporates both symmetric and asymmetric shocks. We find that the Korean won/US dollar exchange rate rarely falls below its trend, but is plucked upward from time to time by transitory shocks. This asymmetry suggests that the monetary authority unofficially intervenes in the foreign exchange market to support its own target level from below. Further evidence from changes in reserve assets indirectly supports our finding.

Keywords:

• Won/dollar exchange rate
• asymmetry
• government intervention

JEL CLASSIFICATIONS :

• C11
• E60
• F31

Acknowledgements

I would like to thank Chang-Jin Kim for kindly providing his Gauss code and Taeyoon Sung, Hyuk Jin Ha, and anonymous referees for their comments. I also thank Dayoung Kim for her research assistance. Any remaining errors are mine. The usual disclaimers apply.

Notes

¹Because of these market interventions in an allegedly market-determined exchange rate system, we call them ‘unofficial’ in the title.

²It is natural to expect that these upper and lower levels can change over time. Our methodology, which follows shortly, can address this issue.

³Readers who are interested in the methods and effects of foreign exchange intervention may refer to the following extensive survey, ‘Foreign Exchange Market Intervention in Emerging Markets: Motives, Techniques and Implications’, by the Bank for International Settlement, BIS Paper No. 24, published in 2005.

⁴This idea was later confirmed in Friedman (1993) in a less formal way.

⁵We also report the results when it follows an AR (1) process.

⁶We assume that σ _w is not state-dependent just for simplicity. Our estimation result below shows that the estimates of σ _u and σ _v are not different across two states and we do not expect a state-dependent σ _w to make the estimates of parameters radically different.

⁷A typical measurement equation takes the following form

  where z _t is a vector of exogenous or predetermined variables and e _t is a random shock.

⁸For estimation, one needs to calculate Pr[S _t =j, S _t−1=i|Ψ _t−1] and Pr[S _t =j|Ψ _t ]. For these and other details of the approximate MLE, refer to Kim and Nelson (1998, 1999).

⁹Asymmetric GARCH models can partially address this issue. For example, one can model that a positive shock has less effect on the conditional variance.

¹⁰We also assign a negative initial value for τ in estimation to see the effect of initial values. The result is not affected by different initial values.

¹¹In addition to the opportunity cost of reserve assets spent on government interventions, moral hazard with respect to market development can be another cost to these kinds of market interventions.

¹²If we calculate the average won/dollar exchange rates without the year 2008, the average devaluation increases to 23.3%.

¹³The title of the report is Report to Congress on International Economic and Exchange Rate Policies, available at http://www.treasury.gov/resource-center/international/exchange-rate-policies/Documents/

¹⁴However, we do not believe that the transitory component reflects all the effects of market interventions. The trend component will be, at least partially, affected by market interventions.