David Li and Xianglin Li, AXA Financial
Charles Wang, Acadian Asset Management
Ren-Raw Chen, Rutgers University and Jingzhi Huang, Penn State University
Peng Liang, Yale University
Xiaoling Zhang, Federal Reserve Board
Chunsheng Zhou, Department of Finance, University of California, Riverside
C.K. Zheng, Morgan Stanley Dean Witter
Edward Qian and Stephen Gorman, Putnam Investments
Charles Wang, Putnam Investments
Dongwei Su and Belton M. Fleisher
Department of Economics, The Ohio State University
Dongwei Su, Department of Economics, The Ohio State University
Lin Guo, Boston College & Suffolk University
Ehud I. Ronn and Changneng Xuan, University of Texas at Austin
Thomas J. Chemmanur & Y. Helen Liu: Columbia University
(April 1996) Dongwei Su and Belton M. Fleisher Department of Economics, The Ohio State University E-mail: su.24@osu.edu WWW Homepage: http://www.econ.ohio-state.edu/Su/dsu1.html This paper studies the nature and causes of volatility of stock-market returns in China. We specify an empirical model to capture the effect of market information, market regulation, and IPO underpricing on changes in daily stock-market returns. In addition to gaining a deeper understanding of the behavior of Chinese stock markets, an important purpose of our study is to derive useful policy implications for stock-market regulatory authority. In our basic model, time-series of changes in stock-market returns are regressed on both local and global information variables. Local variables include the lagged change in six-month Chinese treasury bond rate, lagged dividend yield, and lagged exchange rate change. Global variables include the world stock-market yield (MSCI world index return) in excess of the 30-day U.S. Treasury Bill rate, the Hong Kong Hang Seng index returns in excess of the 30-day U.S. Treasury Bill rate, the change in term structure spread (U.S. ten-year Treasury bond yield minus the 30-day U.S. Treasury Bill rate), and the daily change in 30-day U.S. Treasury Bill rate. Our empirical model of the Chinese stock market is then estimated under three basic alternative formulations of the error generation process. The three processes are : (1) Generalized Autoregressive Conditional Heteroskedasticity, or GARCH(1,1) process; (2) Semi-parametric Autoregressive Conditional Heteroskedasticity, or SPARCH process; (3) Asymmetric GARCH process. Under the GARCH process, we take into consideration three error distributions, namely, Gaussian normal distribution, standardized-t distribution, and generalized error distribution (GED). All the parameters are estimated using maximum likelihood. after comparing the goodness-of-fit of all the specifications, asymmetric GARCH fits the data best. We then apply the best fitted specification to the study of policy implications. We focus on three issues: (1) Has the removal of limits on daily price changes increased the stock return volatility? Shall the government reintroduce the daily price limits? (2) Do stock returns across different geographic markets and across different types of shares tend to respond to common news and factors? Shall the government permit dual listing of shares and eliminate multiple share categories? (3) What is the relationship between IPO underpricing and stock returns volatility? What can the government do about it?
(April 1996) Dongwei SU, Department of Economics, The Ohio State University E-mail: su.24@osu.edu or visit his WWW homepage WWW Homepage: http://www.econ.ohio-state.edu/Su/dsu1.html One of the most important characteristics of China's new stock markets is the difference in IPO pricing between A shares and other non-Chinese shares (B, H, N shares). A shares are only listed in China and can only be purchased by Chinese citizens. Non-Chinese shares can only be owned by foreign investors. All of the non-Chinese shares have IPO prices that are less than those of the corresponding A shares, despite the fact that they have the same yield and voting rights as the Chinese A shares. This study develops and tests a one-period capital asset pricing model (CAPM) to explain the difference in IPO pricing for A and non-Chinese shares under one-way domestic ownership restrictions, assuming price-making behavior of Chinese firms. The model is an extension to Stulz and Wasserfallen's model (The Review of Financial Studies, 1995). Two propositions are derived from the model. The first proposition is for the existence of price discrimination. The second proposition establishes sufficient conditions for A shares premia. The paper shows that A share premia depend on: (1) domestic and foreign investor's absolute risk aversion coefficients; (2) the risk-free rates across countries; (3) the correlation of non-Chinese shares returns with international financial factors; (4) the beta-coefficients of non-Chinese shares with respect to the global financial factors. The model is then tested using data on Chinese stock market, which consists of daily market indices for A shares and B shares, daily stock prices for all listed companies in both security exchanges in China, as well as information on dividend distribution, stock splits, rights issues and other pertinent accounting information.
(Presented at the Third Annual Conference of the Chinese Finance Association, September 1996, New York, New York) Lin Guo, Boston College & Suffolk University E-mail: guoli@bcvms.bc.edu This paper is the first to model empirically a FSLIC-insured institution's length of insolvency. The length of insolvency is defined as the time interval between an institution's starting date of becoming insolvent and its beginning date of being resolved. Analyzing the determinants of the timing of insolvency resolution bears directly on the cost efficiency of capital forbearance policies. We assume that a deposit institution maximizes its net worth, and that the top regulatory manager minimizes the costs of resolving insolvent institutions under various regulatory constraints. Using this framework, we develop the following testable hypotheses: (1) if regulators are faithful agent of taxpayers, ceteris paribus, they should first resolve those institutions with lower capital ratios and higher percentages of speculative investments to minimize the resolution costs; and (2) economic, political, bureaucratic, informational and legal constraints faced by regulators prevent them either from resolving insolvent institutions promptly or from following the social-cost minimizing sequence. A two-step hazards model is employed to estimate the conditional relation between insolvency and resolution and to establish statistically what variables can explain the length of insolvency. The first equation models the evolution of a thrift's net worth. The second equation is a hazard regression that models the timing of insolvency resolution. The net worth ratio enters the second equation as an explanatory variable. It is natural to suppose that the depth of insolvency would influence decisions about the length of time that insolvency is tolerated. This triangular model allows us to separate the determinants of insolvency from those determining the timing of the resolution. As a first step in tracking changes in resolution strategies over time, we examine insolvency resolutions year by year. The sample covers insolvency resolutions from October 1985 to September 1989. There were 318 FSLIC-assisted resolutions during this period. To analyze selection bias, we go beyond the resolved subsample by including insolvent thrifts that were unresolved in the sample period. The length of insolvency is censored for these unresolved institutions. We develop evidence that regulators did not conform to the benchmark strategy that an "unconflicted" agent would follow in timing insolvency resolutions. Results are inconsistent with the hypothesis that regulators simply seek to minimize the losses to the insurance fund. First, authorities do not first resolve the institutions with lowest net worth ratios and highest percentages of speculative investments. Second, regulators also gave more forbearance to larger, older and District 7&9 thrifts. These findings are consistent with the idea that the timing of insolvency resolution is affected by personal, political and bureaucratic constraints faced by regulators. In sum, the results lend support to the view that the timing of insolvency resolution was subject to layered incentive conflicts between: (a) private stakeholders in deposit institutions and managers of federal insurance agencies; (b) managers of federal insurance agencies and government officials who appoint and oversee them; (c) government officials and the taxpayers they represent.
(Presented at the Third Annual Conference of the Chinese Finance Association, September 1996, New York, New York) Ehud I. Ronn and Changneng Xuan, University of Texas at Austin E-mail: shuan@uts.cc.utexas.edu This paper demonstrates the desirability of vega-hedging as a solution to the potential problem of model misspecification. The many and varied providers of volatility-dependent OTC derivative securities (such as options) typically hedge their exposure to the underlying asset by maintaining a delta-neutral position; some augment their hedging by ``managing" their vega exposure. Both hedges require the specification of the stochastic process for the underlying instrument. Since traders can never be sure that the process they have postulated is the ``true" model, we show that vega-hedging is a desirable attribute in reducing tracking error (or ``daily P\&L variation") for such traders. This paper explains why option traders, unsure of the accuracy and reliability of their option pricing model, often vega- and delta-hedge simultaneously, instead of pursuing the delta-hedging-only policy implied by the Black-Scholes formula. We conclude that time series-based attempts to estimate the parameters of the underlying assets' stochastic processes is a misplaced activity, that regulators should encourage broker-dealers to adopt vega-hedging strategies, and that organized exchanges should list long-dated volatility-dependent instruments to provide proper price-discovery and vega-hedging capability.
(Presented at the Third Annual Conference of the Chinese Finance Association, September 1996, New York, New York) Thomas J. Chemmanur & Y. Helen Liu: Columbia University E-mail: yliu@groucho.gsb.columbia.edu The information content of dividends in U.S. equity market has been well documented in previous empirical literature. However, an interesting question remains: Are these effects peculiar to the United States or are they prominent in economies such as Hong Kong where the tax regime and institutional features are significantly different? In this paper, we conduct a comparative empirical study of divident policy in Hong Kong and the United States. We find that dividend payout ratios vary a lot more in the U.S. than in Hong Kong. A test of the classical Lintner model shows that dividends are highly correlated with current year profits in Hong Kong, while they are highly correlated with previous year dividends in the U.S. U.S. firms exhibit a much stronger dividend smoothing effect than Hong Kong firms. there is no significant underperformance or outperformance relative to the market of Hong Kong equities one year before, during and after dividend cuts or dividend increases; while there is severe underperformance by dividend cutting stocks and significant outperformance by dividend increasing stocks in the year prior to and during dividend change year for U.S. firms. A logit regression demonstrates that the significant one-year lagged economic factors for explanining dividend cut decision in he U.S. are dividend yield, size, market-to-book ratio and abnormal return, while in Hong Kong the only significant factor is dividend yield. Our empirical study suggests that Hong Kong firms adjust dividends more dynamically commensurate with earnings. Dividends transform information of firm prospects in the U.S. but dividend signaling seems to be minimal in Hong Kong. issues.