Search

This article studies correlations and dynamic interactions between real stock returns and inflation in the UK for 1830–2000. The BLS test suggests that an endogenous break point exists in 1970, and therefore the pre- and post-break periods are required to be analysed separately. According to the empirical results, for the post-break period, unpredictable stock returns present little correlation to unpredictable inflation, and an increase in stock returns has an insignificant effect on inflation. Impulse response analyses also demonstrate that a positive shock to inflation does not have a negative impact on stock returns. These results are in contrast to the well-known empirical results for the pre-break period.

Korean Abstract: 본 연구에서는 월별 자료를 이용하여 자금시장과 주식시장, 그리고 외환시장이 동태적으로 서로에게 어떤 영향을 미치는지를 살펴보았다. 먼저 축약형 벡터자기회귀모형(VAR)을 최소자승추정법(OLS)으로 추정한 후 Q'Q=QQ'=I의 성격을 갖는 Givens 행렬을 이용하여 경제이론모형에 기반을 둔 부호제약조건을 충족시키는 충격반응함수들을 구하였다. 다음으로 이 충격반응함수들로부터 중위수, 5번째 백분위수, 95번째 백분위수의 충격반응함수들을 도출한 후 촐레스키 분해를 통해 구한 충격반응함수와 비교하였다. 실증적 분석결과는 통상적인 촐레스키 분해에 근거한 충격반응분석이 이들 시장간의 연계성을 설명하는 데 한계를 가지고 있음을 보여준다. 또한 주식시장과 외환시장간의 연계성은 이들 시장과 자금시장간의 연계성에 비해 높을 뿐만 아니라 단기금리의 상승 충격은 케인즈언 모형의 예상과 달리 원/달러환율을 상승시킨다. 최근으로 올수록 주식시장 충격이 외환시장에 미치는 영향은 커지는 반면 외환시장 충격이 주식시장에 미치는 영향은 작아진다. English Abstract: The paper investigates the dynamic interlinkage between the money, stock, and foreign exchange markets from impulse response analyses using monthly data. First, it obtains the impulse response functions which satisfy sign restriction conditions based on economic theories using a Givens matrix after it estimates a reduced-form VAR model by OLS. Next, it derives median, 5 percentile, and 95 percentile impulse response functions from them and then compares them with the impulse response functions based on Choleski decomposition. The empirical results show that the impulse response analysis using Choleski decomposition has limits to explain the interrelations between three markets. The interlinkage between stock and foreign exchange markets is relatively higher, compared to that between these markets and the money market. A positive shock to the money market rate increase won/dollar exchange rates and this result is different from the anticipation of the Keynesian theory. The effect of shock to the stock market o the foreign exchange market recently becomes stronger. On the contrary, an impact of the latter on the former becomes weaker.

By estimating a structural-form GJR model, this paper finds that contemporaneous interactions exist between the stock markets of the U.S., Japan, and Hong Kong. However, it is determined that Hong Kong stock returns react minimally to increases in Japan stock returns.

Out-of-sample performance of exchange rate volatility model depends on criteria used to measure it. The linear GARCH models cannot generally outperform the nonlinear models in the RMSE criterion. Furthermore, the nonparametric kernel model is best in the MAE criterion.

This paper investigates joint dynamics in three Asian emerging foreign exchange markets. The empirical results indicate that domestic and foreign currency-specific shocks have a significant impact on exchange rate changes and their volatility in the region. It also suggests that exchange rate changes and their volatility in these Asian countries respond asymmetrically to positive and negative residuals. But these effects on exchange rate changes become weaker after the Asian currency crises of 1997 came to an end.

The temporal aggregation eï¬ fect on autocorrelation functions for the squares of exchange rate changes is theoretically and empirically analyzed. As thedata frequency decreases, autocorrelation functions converge to zero. The paper also compares weighting schemes and out-of-sample performances of several competing models - GARCH(1, 1), homoskedastic, kernel, flat rolling regressions, and Foster and Nelson’s weighted rolling regressions models. Temporal aggregation generally aggravates out-of-sample performances of GARCH(1,1) models and weighted rolling regressions, compared with other models. Low-frequency GARCH(I, I) models derived from high-frequency GARCH(1, 1) models are not worse than direct low-frequency GARCH(1, 1) models.