Today, there is a growing attention drawn to the impact of the changes in consumer and corporate sentiment on the economy. Hence, this study aims to see whether the sentiment index actually helps to forecast the economy’s direction. In Korea, as well as the advanced countries, there have been studies on whether the BSI resources provide useful information. Kim Jhong-wook (2000) pointed out that under rapidly changing economic conditions, the information drawn from the BSI could be helpful.

To understand the reason for growing usefulness of sentiment indices such as BSI, this study focuses on the possibility of information signal of sentiment indices becoming more accurate. Shim Sang-Dal (2001) suggests empirical evidence of the improved function of stock prices as an economic signal brought on by the progress in integrating domestic stock market and the global market in line with the openness of the capital market after the foreign exchange crisis. The evidence is that since the crisis, the
correlation between stock prices and consumption has grown and also even the consumption of the middle- or low-income classes that do not hold large amounts of shares seems to be responsive to the stock prices, as that of the high-income class. If the sentiment index is influenced by stock prices, the index’s function of estimating the economy might have been improved, too. In this case, the BSI could be useful in both periods of highly volatile economy after the crisis and relatively stabilized economy today,
and also, there might be room for the improved estimation of the economy by using the stock price itself.   

This study attempts to review the possibility that the estimation of the industrial production growth rate—which moves coincidently with the economy—could be improved not only for the post-crisis period but also for the recent trend.

One of the main objectives of this study is not only to find out whether the BSI or stock prices are useful to predict the economy, but also to seek out a methodology that could enhance the prediction power so that they provide actual support to more accurate prediction of the future direction of the economy. Considering the fact that the economic structure of Korea changes rapidly, this study selects the Bayesian Time Varying VAR—which might work better than fixed coefficient—and also attempts the Seasonal
Random Walk Prior as an alternative because of the possibility that the Random Walk Prior—whish is derived from Doan, Litterman, and Sims (1984) and used as a prior to the distribution of initial coefficient—could not produce accurate results when the time-series data has seasonal factors. Along with this, the study also makes sure that changes in working days due to holidays are reflected and adjusted to data in advance.

When using the BSI and the time-series data of stock prices—including current month—applied by working day adjustment and the Seasonal Random Walk Prior, the prediction of the six month-industrial production growth rate is significantly improved. The out-of-sample estimation of the latest period from January 2001 to September 2002 showed that the prediction error of one-month absolute average went down to 2% from 3.5% in AR (2) model. Its prediction of the performance of industrial production which recorded
high growth in October after the slowdown in September turned out relatively accurate.

Such an improvement, of course, was supported more by the adjustment of working days than by the use of the BSI. The working days adjustment alone can reduce the one-month prediction of error by 1%.

The reason that the BSI is useful in predicting the economy is because the BSI data is significantly influenced by herd behavior and also likely to be affected by the overall economic signals such as stock prices. All of these imply that policy authority including monetary authority need to pay a keen attention to the movement of stock prices and BSI data, because it is highly possible that these indices serve as leading indicators of the economy.

 

발간사

요 약

제1장 서 론

제2장 기존연구의 검토

제3장 데이터 및 상관관계 분석
　1. 경기와 산업생산증가율
　2. 예비분석

제4장 단순자기회귀모형 분석

제5장 베이지안 변동계수모형분석
　1. 베이지안 변동계수모형
　　가. 순차적(Recursive) 벡터모형
　　나. 계절주기 임의보행 프라이어(Seasonal Random Walk Prior)
　2. 조업일수 사전조정
　3. 예측력 평가결과
　　가. 비교기준 모형의 설정
　　나. BSIA의 유용성평가
　　다. BSIA와 주가지수와 선행지수의 예측력 비교
　　라. 사전제약(Prior)별 예측력 비교
　　마. 예측력 검정결과의 요약

제6장 요약 및 결론

참고문헌

부 록
　가. [變動係數 벡터 自己回歸모형]
　　1) 모형의 구조
　　2) 칼만연산(Kalman Filter)과정
　　3) 초기값에 대한 베이지안 사전제약
　　4) 變形最犬度 基準(Pseudo Likelihood Criterion)에 따른 최적모수선택