Using non-parametric search algorithms to forecast daily excess stock returns

Are the learning procedures of genetic algorithms (GAs) able to generate optimal architectures for artificial neural networks (ANNs) in high frequency data? In this experimental study,GAs are used to identify the best architecture for ANNs. Additional learning is undertaken by the ANNs to forecast daily excess stock returns. No ANN architectures were able to outperform a random walk,despite the finding of non-linearity in the excess returns. This failure is attributed to the absence of suitable ANN structures and further implies that researchers need to be cautious when making inferences from ANN results that use high frequency data.

The sensitivity of US banks' stock returns to interest and exchange rate changes

Purpose – The purpose of this paper is to investigate the impact of foreign exchange and interest rate changes on US banks’ stock returns. Design/methodology/approach – The approach employs an EGARCH model to account for the ARCH effects in daily returns. Most prior studies have used standard OLS estimation methods with the result that the presence of ARCH effects would have affected estimation efficiency. For comparative purposes, the standard OLS estimation method is also used to measure sensitivity. Findings – The findings are as follows: under the conditional t-distributional assumption, the EGARCH model generated a much better fit to the data although the goodness-of-fit of the model is not entirely satisfactory; the market index return accounts for most of the variation in stock returns at both the individual bank and portfolio levels; and the degree of sensitivity of the stock returns to interest rate and FX rate changes is not very pronounced despite the use of high frequency data. Earlier results had indicated that daily data provided greater evidence of exposure sensitivity. Practical implications – Assuming that banks do not hedge perfectly, these findings have important financial implications as they suggest that the hedging policies of the banks are not reflected in their stock prices. Alternatively, it is possible that different GARCH-type models might be more appropriate when modelling high frequency returns. Originality/value – The paper contributes to existing knowledge in the area by showing that ARCH effects do impact on measures of sensitivity.

Risk, seasonality and the asymmetric behaviour of stock returns

For some time there has been a puzzle surrounding the seasonal behaviour of stock returns. This paper demonstrates that there is an asymmetric relationship between systematic risk and return across the different months of the year for both large and small firms. In the case of both large and small firms systematic risk appears to be priced in only two months of the year, January and April. During the other months no persistent relationship between systematic risk and return appears to exist. The paper also shows that when systematic risk is priced, the size of the systematic risk premium is higher for large firms than for small firms and varies significantly across the months of the year.

Breaking down the non-normality of daily stock returns

This paper investigates whether the non-normality typically observed in daily stock-market returns could arise because of the joint existence of breaks and GARCH effects. It proposes a data-driven procedure to credibly identify the number and timing of breaks and applies it on the benchmark stock-market indices of 27 OECD countries. The findings suggest that a substantial element of the observed deviations from normality might indeed be due to the co-existence of breaks and GARCH effects. However, the presence of structural changes is found to be the primary reason for the non-normality and not the GARCH effects. Also, there is still some remaining excess kurtosis that is unlikely to be linked to the specification of the conditional volatility or the presence of breaks. Finally, an interesting sideline result implies that GARCH models have limited capacity in forecasting stock-market volatility.

Short and long memory in stock returns data

The properties of an iterative procedure for the estimation of the parameters of an ARFIMA process are investigated in a Monte Carlo study. The estimation procedure is applied to stock returns data for 15 countries. © 2012.

Market quality changes in the London Stock Market

This paper examines the impact that the introduction of a closing call auction had on market quality at the London Stock Exchange. Using estimates from the partial adjustment with noise model of Amihud and Mendelson [Amihud, Y., Mendelson, H., 1987. Trading mechanisms and stock returns: An empirical investigation. Journal of Finance 42, 533–553] we show that opening and closing market quality improved for participating stocks. When we stratify our sample securities into five groups based on trading activity we find that the least active securities experience the greatest improvements to market quality. A control sample of stocks are not characterized by discernable changes to market quality.

The effect of universal futures on opening and closing stock market prices

Purpose – In 2001, Euronext-Liffe introduced single security futures contracts for the first time. The purpose of this paper is to examine the impact that these single security futures had on the volatility of the underlying stocks. Design/methodology/approach – The Inclan and Tiao algorithm was used to show that the volatility of underlying securities did not change after universal futures were introduced. Findings – It was found that in the aftermath of the introduction of universal futures the volatility of the underlying securities increases. Increased volatility is not apparent in the control sample. This suggests that single security futures did have some impact on the volatility of the underlying securities. Originality/value – Despite the huge literature that has examined the effects of a futures listing on the volatility of underlying stock returns, little consensus has emerged. This paper adds to the dialogue by focusing on the effects of a single security futures contract rather than concentrating on the effects of index futures contracts.

Explaining volatility and serial correlation in opening and closing returns:a study of the FT-30 components

In this paper the performance of opening and closing returns, for the components of the FT-30 will be studied. It will be shown that for these stocks opening returns have higher volatility and a greater tendency towards negative serial correlation than closing returns. Unlike previous studies this contrasting performance cannot solely be attributed to differences in the trading mechanism across the trading day. All the stocks used in our sample trade thought the day using a uniform trading mechanism. In this paper, we suggest that it is differences in the speed that closing and opening returns adjust to new information that causes differences in return performance. By estimating the Amihud and Mendelson (1987) [Amihud, Yakov, & Mendelson, Haim (1987). Trading mechanisms and stock returns: An empirical investigation, Journal of Finance, 62 533-553.] partial adjustment model with noise, we show that opening returns have a tendency towards over-reaction, while closing returns have a tendency towards under-reaction. We suggest that it is these differences that cause a substantial proportion (although not all) of the asymmetric return patterns associated with opening and closing returns. © 2005 Elsevier Inc. All rights reserved.

An empirical investigation of the UK stock market response to the implementation of SSAP 20

This study presents an empirical investigation of the UK stock market response to the im-plementation of the UK Statement of Standard Accounting Practice (SSAP) No. 20 “Foreign Cur-rency Translation” (issued in April 1983). Such an empirical investigation has not yet been under-taken for the UK. Our results show that the stock market generally appeared to have anticipatedthe implementation of SSAP 20. For the aggregate set of adopters, we found a positive stock mar-ket response in the official year of adoption, reflecting the appreciation of the income-stabilisingeffects of the standard. This paper also presents a cross-sectional analysis that tests for a relation-ship between the stock returns and the accounting measures of those firms that adopted SSAP 20.We found a significant relation between the stock returns and the related accounting measures inthe actual adoption period of the aggregate set of adopters. This study generally focuses on theinterpretation of the financial impacts of the various accounting choices of firms within their fi-nancial and economic environments.

The leverage effect in the UK stock market

This study seeks to explain the leverage in UK stock returns by reference to the return volatility, leverage and size characteristics of UK companies. A leverage effect is found that is stronger for smaller companies and has greater explanatory power over the returns of smaller companies. The properties of a theoretical model that predicts that companies with higher leverage ratios will experience greater leverage effects are explored. On examining leverage ratio data, it is found that there is a propensity for smaller companies to have higher leverage ratios. The transmission of volatility shocks between the companies is also examined and it is found that the volatility of larger firm returns is important in determining both the volatility and returns of smaller firms, but not the reverse. Moreover, it is found that where volatility spillovers are important, they improve out-of-sample volatility forecasts. © 2005 Taylor & Francis Group Ltd.

On the existence of visual technical patterns in the UK stock market

The aim in this paper is to replicate and extend the analysis of visual technical patterns by Lo et al. (2000) using data on the UK market. A non-parametric smoother is used to model a nonlinear trend in stock price series. Technical patterns, such as the 'head-and-shoulders' pattern, that are characterised by a sequence of turning points are identified in the smoothed data. Statistical tests are used to determine whether returns conditioned on the technical patterns are different from random returns and, in an extension to the analysis of Lo et al. (2000), whether they can outperform a market benchmark return. For the stocks in the FTSE 100 and FTSE 250 indices over the period 1986 to 2001, we find that technical patterns occur with different frequencies to each other and in different relativities to the frequencies found in the US market. Our extended statistical testing indicates that UK stock returns are less influenced by technical patterns than was the case for US stock returns.

Mean reversion in the short horizon returns of UK portfolios

This paper will show that short horizon stock returns for UK portfolios are more predictable than suggested by sample autocorrelation co-efficients. Four capitalisation based portfolios are constructed for the period 1976–1991. It is shown that the first order autocorrelation coefficient of monthly returns can explain no more than 10% of the variation in monthly portfolio returns. Monthly autocorrelation coefficients assume that each weekly return of the previous month contains the same amount of information. However, this will not be the case if short horizon returns contain predictable components which dissipate rapidly. In this case, the return of the most recent week would say a lot more about the future monthly portfolio return than other weeks. This suggests that when predicting future monthly portfolio returns more weight should be given to the most recent weeks of the previous month, because, the most recent weekly returns provide the most information about the subsequent months' performance. We construct a model which exploits the mean reverting characteristics of monthly portfolio returns. Using this model we forecast future monthly portfolio returns. When compared to forecasts that utilise the autocorrelation statistic the model which exploits the mean reverting characteristics of monthlyportfolio returns can forecast future returns better than the autocorrelation statistic, both in and out of sample.

Differences between theoretical predictions and operational performance in a stock control application

Over the past decade, several experienced Operational Researchers have advanced the view that the theoretical aspects of model building have raced ahead of the ability of people to use them. Consequently, the impact of Operational Research on commercial organisations and the public sector is limited, and many systems fail to achieve their anticipated benefits in full. The primary objective of this study is to examine a complex interactive Stock Control system, and identify the reasons for the differences between the theoretical expectations and the operational performance. The methodology used is to hypothesise all the possible factors which could cause a divergence between theory and practice, and to evaluate numerically the effect each of these factors has on two main control indices - Service Level and Average Stock Value. Both analytical and empirical methods are used, and simulation is employed extensively. The factors are divided into two main categories for analysis - theoretical imperfections in the model, and the usage of the system by Buyers. No evidence could be found in the literature of any previous attempts to place the differences between theory and practice in a system in quantitative perspective nor, more specifically, to study the effects of Buyer/computer interaction in a Stock Control system. The study reveals that, in general, the human factors influencing performance are of a much higher order of magnitude than the theoretical factors, thus providing objective evidence to support the original premise. The most important finding is that, by judicious intervention into an automatic stock control algorithm, it is possible for Buyers to produce results which not only attain but surpass the algorithmic predictions. However, the complexity and behavioural recalcitrance of these systems are such that an innately numerate, enquiring type of Buyer needs to be inducted to realise the performance potential of the overall man/computer system.

Liquidity and portfolio optimisation

This thesis presents research within empirical financial economics with focus on liquidity and portfolio optimisation in the stock market. The discussion on liquidity is focused on measurement issues, including TAQ data processing and measurement of systematic liquidity factors (FSO). Furthermore, a framework for treatment of the two topics in combination is provided. The liquidity part of the thesis gives a conceptual background to liquidity and discusses several different approaches to liquidity measurement. It contributes to liquidity measurement by providing detailed guidelines on the data processing needed for applying TAQ data to liquidity research. The main focus, however, is the derivation of systematic liquidity factors. The principal component approach to systematic liquidity measurement is refined by the introduction of moving and expanding estimation windows, allowing for time-varying liquidity co-variances between stocks. Under several liability specifications, this improves the ability to explain stock liquidity and returns, as compared to static window PCA and market average approximations of systematic liquidity. The highest ability to explain stock returns is obtained when using inventory cost as a liquidity measure and a moving window PCA as the systematic liquidity derivation technique. Systematic factors of this setting also have a strong ability in explaining a cross-sectional liquidity variation. Portfolio optimisation in the FSO framework is tested in two empirical studies. These contribute to the assessment of FSO by expanding the applicability to stock indexes and individual stocks, by considering a wide selection of utility function specifications, and by showing explicitly how the full-scale optimum can be identified using either grid search or the heuristic search algorithm of differential evolution. The studies show that relative to mean-variance portfolios, FSO performs well in these settings and that the computational expense can be mitigated dramatically by application of differential evolution.