Research

Network participants have long looked at the suitability of traditional methods of analysis to accounting and finance questions. Thus, it is inherent to this network the critical review of existing methodologies in financial analysis, that is complemented by implementing suitable contemporary advances found in the decision making sciences. Furthermore, the network aims to make significant contributions to the theoretical framework of financial analysis, and to inform future research and practice through empirical findings. Inductive financial theory that recognises the role of financial statements as inputs to economic decisions also plays an important role in the research programme, therefore guiding empirical tests of such information inputs.

Properties of Underlying Financial Measures

  • Panel data methods of analysis for analysing the market across time and space. Of specific interest are panel datasets with an assumed infinite number of cross-sections and a finite number of points in time. This combination of dimensions is common in financial analysis (firm-year datasets), especially in large scale multivariate models of accounting variables, stock market informationand economic determinants.

  • Analytical modelling of financial statement information. Financial statement variables are instantaneously co-determined across an articulated set of financial statements (balance sheets at t and t-1, together with income statement at t and cash flow statement at t). A consolidated set of financial statements can be modelled as a matrix of endogenous information, and this poses added complexity in deriving useful econometric models. Thus, when regressing accounting variables on other accounting variables that are extracted from the same set of financial statements, then particular attention must be drawn on the constrained parameterisation of the relationship (through the double entry bookkeeping identity), as well as on choosing the appropriate econometric estimator.

  • The direction and degree of causality between accounting and stock market information. Extant research utilises accounting variables for examining the reaction of the market to this publicly disclosed information. Yet, it may be the case that stock market data creates shocks that are ultimately recognised in the book value of the firm. Also, in this context, the examination of P/E ratio and the functional relationship between accounting returns and economic returns is of particular interest to the network.

  • Identification of a likely range of variation for financial measures and the creation of common-size financial measures. The range of variation is an issue of controversy in the financial literature. Heuristic treatments have included arbitrary trimming, robust estimators and sample selection. The range of variation is also directly related to the creation of common-size measures for analysis via deflation. Deflation changes the range of variation and there is an ongoing debate on the appropriateness of alternative economic deflators and how these may measure the true size of the firm. An interesting issue is the possibility of bounded range of variation that may imply natural/implicit limits in accounting measurements and stock price variability. A bounded range will help interpret extreme observations, and the overall shape of the probability density function of variables directly derived from financial measures and combined with other economic data.

  • Probability density functions for accounting variables and financial ratios. The distribution of earnings has created a new venue of debate in the literature involving incentives of opportunistic earnings management, signalling and other incentives. The distribution of earnings is commonly assumed to be normal in the limit, and attribute observational non-normality due to managerial incentives and other irregularities. However, the earning distribution, and to extent, all other accounting variables distributions are constraint by nature and express pressures that cannot be considered to produce a normal distribution (e.g. taxation, transitory items, asymmetric responsiveness to profits and losses etc). Also, financial ratio distributions (e.g. P/E) pose another challenge due to the direction of causality, the (non)stationarity of stock market prices etc.

The Use of Financial Measures in an Analytical Decision Context

  • Corporate restructuring transactions. A performance-based approach is considered for evaluating cost impacts of both acquisition and divestiture transactions. The basic hypothesis to be investigated is that such transactions result in measurable improvements in the cost structure and, by extension, improve in the performance of firms engaging in these transactions. In this regard, the work extends prior performance studies utilising financial statement information in the form of ratios. The rationale is that stock market inefficiencies may be responsible for the results found in price reaction studies. Accordingly, a direct comparison of pre and post transaction operating performance will reveal the extent of 'real' economic gains from the transaction. A panel data approach is applied for differentiating between deterministic and stochastic trends in financial ratios at the firm level.The hypothesis that restructuring transactions affects the cost base of the firm would be confirmed by verifying non-stationarity in the components of  relevant financial ratios.

  • Two and three state classification problems. Typical two state classifications are credit granting decisions (yes/no) and going concern determinations, whereas the buy/hold/sell recommendation that all equity analysts produce is a three state classification problem. Such decisions are taken only after the application of financial analysis in assessing credit risks, understanding solvency issues, forecasting financial performance, etc.

  • Exploration of statistical decision models in the practice of financial analysis. There are a number of statistical approaches available for use in decision support models, including parametric, semi parametric, Bayesian and non parametric. Specifically, the parametric method of multinomial logistic modelling has been applied to going concern analysis, semi parametric hazards models have been applied to corporate reorganisations and Bayesian and non parametric methods have been applied to stock analysis and credit granting decisions.

  • Economic models of the dynamic behaviour of the firm for discovering the drivers of financial performance. The matching of the variables from the economic models to conventional accounting measurements is an important input into the activities of the network. A particular focus is given to the use of neural networks and tree based methods in classification, with reference to credit granting and going concern analysis. Such non parametric methods provide important benchmarks in the evaluation of statistical decision models, where the lack of distributional assumptions on the input data makes these methods especially useful in handling financial measures with non standard distributions. Furthermore, the continuing investigation of the application of hazard function methods to modelling the outcomes of restructuring activities and stock return behaviour extends the range of binary classification problems in financial analysis. This semi parametric approach not only provides a method of classification, but it also provides timing information which can enhance the quality of financial decision making.

  • Development and evaluation of Bayesian forecasting and quantile estimation methods for Value at Risk (VaR). VaR measures the amount of money when an investment position is at risk of losing with a set probability. VaR forecasting is now mandatory as a risk management tool for all financial institutions. Many competing econometric and time series methods have been used in the literature to forecast VaR, using both parametric and nonparametric methods. Methods that improve on these approaches will have quite some significance and be of importance and interest to the statistical/econometric, forecasting and financial literature, as well as the financial industry itself, especially where VaR forecasting is mandatory.