Market Predictability - page 3

The main goal of this article is to provide an answer to the question: “Does anything forecast exchange rates, and if so, which variables?”. It is well known that exchange rate ‡fluctuations are very di¢ cult to predict using economic models, and that a random walk forecasts exchange rates better than any economic model (the Meese and Rogo¤ puzzle). However, the recent literature has identi…fied a series of fundamentals/methodologies that claim to have resolved the puzzle. This article provides a critical review of the recent literature on exchange rate forecasting and illustrates the new methodologies and fundamentals that have been recently proposed in an up-to-date, thorough empirical analysis. Overall, our analysis of the literature and the data suggests that the answer to the question: "Are exchange rates predictable?" is, "It depends" –on the choice of predictor, forecast horizon, sample period, model, and forecast evaluation method. Predictability is most apparent when one or more of the following hold: the predictors are Taylor rule or net foreign assets, the model is linear, and a small number of parameters are estimated. The toughest benchmark is the random walk without drift.

The predictability of monthly stock market index returns is considered, with special emphasis on possible outliers in the data. The data consists of indices from fifteen OECD countries. Three models are used for predicting these series. A random walk model is used as a benchmark, which any other model should be able to improve on. A basic autoregressive (AR) model is then compared with an ARoutlier model, where dummy variables are added for detected outliers. It is assumed that if such outliers are taken into account in the model, the predictions should become better compared to the basic AR model, in which any potential outliers are ignored. Four criteria are used to compare the predictions, a number of which are made both one and two steps ahead. The results indicate that for autocorrelated series (ten out of a total of fifteen series) the outlier model does indeed provide the best predictions.

Regulatory stress tests have become a key tool for setting bank capital levels. Publicly disclosed results for four rounds of stress tests suggest that as the stress testing process has evolved, its outcomes have become more predictable and therefore arguably less informative. In particular, projected stress losses in the 2013 and 2014 stress tests are nearly perfectly correlated for bank holding companies that participated in both rounds. We also compare projected losses across different scenarios used in the 2014 stress test and find surprisingly high correlations for outcomes grouped by bank or by loan category, which suggests an opportunity to get more information out of the stress tests through greater diversity in the scenarios used. We discuss potential implications of these patterns for the further development and application of stress testing

Market efficiency theory suggests that market is rational and provides correct pricing. That is, the current prices of securities are close to their fundamental values because of either the rational investors or the arbitragers’ buy and sell action of underpriced or overpriced stocks. On the other hand, observed market anomalies have a challenge for this argument. They claim that irrational investment activities and the arbitrage opportunities’ being limited in markets cause some market anomalies that are inconsistent with efficient market hypothesis. The most commonly seen anomalies are the “volume”, “volatility”, “cash dividends”, “equity premium puzzle”, and the “predictability”. This work is a literature survey, and its main objective is to deal with efficient market theory and market anomalies in order to examine the question “Are markets rational or not"?

In this paper, we present a few simple market-timing strategies that appear to outperform the “buy-and-hold” strategy, with real-time data from 1970 to 2000. Our focus is on spreads between the E/P ratio of the S&P 500 index and interest rates. Extremely low spreads, as compared to their historical ranges, appear to predict higher frequencies of subsequent market downturns in monthly data. We construct “horse races” between switching strategies based on extremely low spreads and the market index. Switching strategies call for investing in the stock market index unless spreads are lower than predefined thresholds. We find that switching strategies outperformed the market index in the sense that they provide higher mean returns and lower variances. In particular, the strategy based on the spread between the E/P ratio and a short-term interest rate comfortably and robustly beat the market index even when transaction costs are incorporated.

Are excess returns predictable, and if so, what does this mean for investors? In classic studies of rational valuation (e.g. Samuelson (1965, 1973) and Shiller (1981)), risk premia are constant over time and thus excess returns are unpredictable.2 However, an extensive empirical literature has found evidence for predictability in returns on stocks and bonds by scaled-price ratios and interest rates

The predictability in stock returns is an interesting issue by both practitioners and academicians. Many predictor variables have been employed in those literatures including dividend-price ratio, earnings-price ratio, firm size, interest rate and also other economic and fundamental variables. However, those empirical results did not explicitly indicate which variables should be included as predictors in predictive regression and there are concerned about data over fitting. In particular, Bossaerts and Hillion (1999) investigated predictability power on stock returns by separating data to in-sample and out-of-sample. They confirmed in-sample return predictability but failed to confirm out-of-sample return predictability. Their empirical results suggested that predictive variables used in regression should be treated as time varying factors. Historical data should be separated to prior and posterior distribution in order to reflect past information and expectations. Posterior probability of data should be conditional on information variables. Recently, new methodology is employed for return predictability namely Bayesian Model Averaging (BMA). This model takes into account the conditional probability of posterior data. Most of previous empirical results regarding predictability on stock returns focused on developed market. For emerging market, there were a few studies about model uncertainty and none had applied Bayesian model to predictive regression. Therefore, it leads to the research question that “As one of the emerging capital market members, does Thailand stock market have return predictability power by applying Bayesian methodology?”

Can we obtain excess profits from financial market by forecasting financial time series? This question is directly related to whether the financial market prices are predictable. After running several tests, many financial economists have accepted the random-walk hypothesis, which states that financial market prices are completely random because of the efficiency of the market and thus the prices of the financial market cannot be predicted 1. However, other economists and investors believe that prices may move in trends and past prices can be used to forecast future price changes to some degree 2. Empirical studies have provided some evidences that there are long-term or short-term dependent relationships in financial time series, which offer promises for predictability 3,4. As a model-independent measurement, approximate entropy
ApEn has been proposed to characterize the irregular degree of financial time series. ApEn assigns a nonnegative number to a given sequential data, with larger values corresponding to greater apparent serial randomness or lower predictability and with smaller values corresponding to more regularity or higher predictability 5. As shown in Ref. 5, ApEn values of Dow Jones index and Hang Seng
Hong Kong index are significantly different from that of random series. From the viewpoint of investment, large number of models has been proposed to provide investors with more accurate forecasts. Time-series models, based on conventionally statistical methods, were very popular in constructing various kinds of financial market prediction models in the past 6. Without relying too much on specific assumptions and error distributions, artificial neural networks
ANNs have been demonstrated to be successful research models to forecast, detect, and summarize the structure of financial variables 7. Recent years, results of some forecast exercises suggested that support vector machine
SVM might be the most promising method for predicting financial time series 8,9. Though there are a great number of papers investigating the predictability of financial market, we still do not fully understand the relationship between the predictable degree and the information structure of financial time series.

This paper studies predictability of currency returns over time and the extent to which it is captured by trading rules commonly used in currency markets. We consider the strategies that an investor endowed with rational expectations could have pursued to exploit out-of-sample currency predictability and generate abnormal returns. We find a close relation between these strategies and indices that track popular technical trading rules, namely moving average cross-over rules and the carry trade, implying that the technical rules represent heuristics by which professional market participants exploit currency mispricing. We find evidence that such mispricing reflects initially wrong investors’ beliefs (wrong priors), but information is efficiently processed as it becomes available. Predictability is highest in the mid ’90, subsequently decreases sharply, but increases again in the final part of the sample period, especially for the Euro and other emerging currencies.