How volatile are East Asian stocks during high volatility periods?

1. How volatile are East Asian stocksduring high volatility periods? A workshop paper byCarlos C. BautistaCollege of Business AdministrationUniversity of the PhilippinesEconometric forecasting & high-frequency data analysisInstitute for Mathematical Sciences, National University of SingaporeSchool of Economics & Social Sciences, Singapore Management UniversityApril-May 2004

2. Abstract This study reports the estimates of the magnitude of volatility during abnormal times relative to normal periods for seven East Asian economies using a rudimentary univariate Markov-switching ARCH method. The results show that global and regional events like the 1990 gulf war and the 1997 Asian currency crisis led to high volatility episodes whose magnitude relative to normal times differ from country to country. Country-specific events like the opening up of country borders in the mid-1990s are also observed to lead to high volatility periods. Additional insights are obtained when volatility was assumed to evolve according to a three-state Markov regime switching process.

3. Objectives 1. The study attempts to determine the extent of volatility in the stock markets of 7 East Asian economies. • This is done by computing a volatility index.The index is a measure relative to “normal” volatility within the country. • The Markov-regime-switching ARCH by Hamilton & Susmel (1994) is used to compute the index. 2. The study attempts to relate the results obtained with the events that transpired during the period covered by the study.

4. Countries, data and period covered by the study • COUNTRIES: Hong Kong, Indonesia, South Korea, Malaysia, Philippines, Thailand, Singapore • DATA SOURCE: Stock price data from DATASTREAM, and country institutions • FREQUENCY: Weekly data from 1988 to 2003

5. Brief review of related literature • ARCH modeling • Engle (1982) • Bollerslev (1986) • Markov-switching • Hamilton (1989) • Lam (1990) • Markov-switching ARCH • Hamilton & Susmel (1994) • Gray (1996) • There are many other published articles on the first two topics but not as much on the third.

6. The volatility state is assumed to be the outcome of an unobserved first-order K-state Markov process, which is described by transition probabilities: P[s(t) = j|s(t - 1) = i] = pij g(s) is a variance factor that serves to scale the ARCH process. s = 1, 2, …, K, is the state variable that indexes the “volatility regime.” a normalization is imposed such that g(1) = 1 and g(s)  1 for s = 2, …, K. Hence, state 1 is the base state. Switching ARCH model

7. Table 1Stock returns – Descriptive statistics

8. Table 22-state regime-switching ARCH regressions

9. Fig 2.1: Hong Kong – 2 state model • Index = 4.2 • High volatility is shown from the start of the 1997 Asian crisis until 2001 • Shaded areas in all diagrams cover the start of the 1997 Asian crisis up to the end of the period under study

10. 1988 to 1990 data were obtained from ‘pusat data pasar modal,’ Gadjah Mada University Index = 15.4 The spike in 1988 marks the start of liberalization when a package of laws were passed to stimulate the market.This accounts for the high value of the index. Fig 2.2: Indonesia – 2 state model

11. Index = 3.9 The 2 state model was not able to capture the opening of the Korean stock market to foreign investors in January 3, 1992 This led to attempts to estimate the 3 state model Fig 2.3: South Korea – 2 state model

12. Index = 5.1 Clear effects of the gulf war, early 90s liberalization and the 1997 Asian crisis ARCH lags were added until a significant coefficient was obtained Fig 2.4: Malaysia – 2 state model

13. Index = 3.6 The earliest among the countries to revert to the low volatility state after the 1997 Asian crisis Other high volatility states:1989 coup attempt (failed)1990 fiscal crisis1993 liberalization This is an update of Bautista (2003) Fig 2.5: Philippines – 2 state model

14. Index = 5.4 Clear influence of the gulf war Intermittent high volatility after the Asian crisis Fig 2.6: Singapore – 2 state model

15. Index = 4.7 High volatility state in 1996, many months before the crisis Clear effects of the gulf war and early 90s liberalization Fig 2.7: Thailand – 2 state model

16. Table 33-state regime-switching ARCH regressions

17. Table 4Transition probability matrices, 3-state model

18. Fig 3.1: Hong Kong – 3 state model • Index, moderate = 3.1Index, high = 9.7 • Very short high volatility episode at the start of the crisis

19. Fig 3.2: Indonesia – 3 state model • Index, moderate = 2.6Index, high = 34.0 • A very calm market many years before the 1997 Asian crisis; Has not reverted to this state since.

20. Fig 3.3: Korea – 3 state model • Index, moderate = 2.2Index, high = 5.5 • Moderate volatility in 1992 reflects the opening of the stock market. • Moderate volatility prior to and after the 1997 Asian crisis

21. Fig 3.4: Malaysia – 3 state model • Index, moderate = 3.1Index, high = 14.5 • Moderate volatility prior to the 1997 Asian crisis

22. Fig 3.5: Philippines – 3 state model • Index, moderate = 1.9Index, high = 5.2 • High volatility a few months before the crisis; A result that also appeared in Bautista (2003)

23. Fig 3.6: Singapore – 3 state model • Index, moderate = 3.1Index, high = 14.1 • Short high volatility period at the start of the 1997 Asian crisis;Moderately volatile thereafter.

24. Fig 3.7: Thailand – 3 state model • Index, moderate = 2.9Index, high = 9.4 • Moderate volatility in 1996; can be compared with the 2 state result.

25. Closing remarks • This paper has examined the magnitude of stock market volatility relative to normal periods in seven Asian economies. • The paper showed that the volatility patterns that emerge in the estimation adequately characterize events that transpired in the East Asian region. It also to some extent highlighted unique characteristics of the individual country markets. • For the Philippines, Malaysia, Thailand and Korea, the computed probabilities of moderate and high volatility states might be a good indicator of a forthcoming crisis.

26. The End