A comparison of linear forecasting models and neural networks:an application to Euro inflation and Euro Divisia

Linear models reach their limitations in applications with nonlinearities in the data. In this paper new empirical evidence is provided on the relative Euro inflation forecasting performance of linear and non-linear models. The well established and widely used univariate ARIMA and multivariate VAR models are used as linear forecasting models whereas neural networks (NN) are used as non-linear forecasting models. It is endeavoured to keep the level of subjectivity in the NN building process to a minimum in an attempt to exploit the full potentials of the NN. It is also investigated whether the historically poor performance of the theoretically superior measure of the monetary services flow, Divisia, relative to the traditional Simple Sum measure could be attributed to a certain extent to the evaluation of these indices within a linear framework. Results obtained suggest that non-linear models provide better within-sample and out-of-sample forecasts and linear models are simply a subset of them. The Divisia index also outperforms the Simple Sum index when evaluated in a non-linear framework. © 2005 Taylor & Francis Group Ltd.

A composite leading indicator of the inflation cycle for the Euro area

We evaluate the performance of composite leading indicators of turning points of inflation in the Euro area, constructed by combining the techniques of Fourier analysis and Kalman filters with the National Bureau of Economic Research methodology. In addition, the study compares the empirical performance of Euro Simple Sum and Divisia monetary aggregates and provides a tentative answer to the issue of whether or not the UK should join the Euro area. Our findings suggest that, first, the cyclical pattern of the different composite leading indicators very closely reflect that of the inflation cycle for the Euro area; second, the empirical performance of the Euro Divisia is better than its Simple Sum counterpart and third, the UK is better out of the Euro area. © 2005 Taylor & Francis Group Ltd.

Does money matter in inflation forecasting

This paper provides the most fully comprehensive evidence to date on whether or not monetary aggregates are valuable for forecasting US inflation in the early to mid 2000s. We explore a wide range of different definitions of money, including different methods of aggregation and different collections of included monetary assets. In our forecasting experiment we use two non-linear techniques, namely, recurrent neural networks and kernel recursive least squares regression - techniques that are new to macroeconomics. Recurrent neural networks operate with potentially unbounded input memory, while the kernel regression technique is a finite memory predictor. The two methodologies compete to find the best fitting US inflation forecasting models and are then compared to forecasts from a naive random walk model. The best models were non-linear autoregressive models based on kernel methods. Our findings do not provide much support for the usefulness of monetary aggregates in forecasting inflation.

Wealth shocks, credit-supply shocks, and asset allocation:evidence from household and firm portfolios

We use a unique dataset with bank clients’ security holdings for all German banks to examine how macroeconomic shocks affect asset allocation preferences of households and non-financial firms. Our analysis focuses on two alternative mechanisms which can influence portfolio choice: wealth shocks, which are represented by the sovereign debt crisis in the Euro area, and credit-supply shocks which arise from reductions in borrowing abilities during bank distress. While households with large holdings of securities from stressed Euro area countries (Greece, Ireland, Italy, Portugal, and Spain) de-crease the degree of concentration in their security portfolio as a result of the Euro area crisis, non-financial firms with similar levels of holdings from stressed Euro area countries do not. Credit-supply shocks at the bank level result in lower concentration, for both households and non-financial corporations. Only shocks to corporate credit bear ramifications on bank clients’ portfolio concentration. Our results are robust to falsification tests, and instrumental variables estimation.

Illiquidity shocks and the comovement between stocks:new evidence using smooth transition

This paper extends the smooth transition conditional correlation model by studying for the first time the impact that illiquidity shocks have on stock market return comovement. We show that firms that experience shocks that increase illiquidity are less liquid than firms that experience shocks that decrease illiquidity. Shocks that increase illiquidity have no statistical impact on comovement. However, shocks that reduce illiquidity lead to a fall in comovement, a pattern that becomes stronger as the illiquidity of the firm increases. This discovery is consistent with increased transparency and an improvement in price efficiency. We find that a small number of firms experience a double illiquidity shock. For these firms, at the first shock, a rise in illiquidity reduces comovement while a fall in illiquidity raises comovement. The second shock partly reverses these changes as a rise in illiquidity is associated with a rise in comovement and a fall in illiquidity is associated with a fall in comovement. These results have important implications for portfolio construction and also for the measurement and evolution of market beta and the cost of capital as it suggests that investors can achieve higher returns for the same amount of market risk because of the greater diversification benefits that exist. We also find that illiquidity, friction, firm size and the pre-shock correlation are all associated with the magnitude of the correlation change. © 2013 Elsevier B.V.

Evolution, recurrency and kernels in learning to model inflation

This paper provides the most fully comprehensive evidence to date on whether or not monetary aggregates are valuable for forecasting US inflation in the early to mid 2000s. We explore a wide range of different definitions of money, including different methods of aggregation and different collections of included monetary assets. We use non-linear, artificial intelligence techniques, namely, recurrent neural networks, evolution strategies and kernel methods in our forecasting experiment. In the experiment, these three methodologies compete to find the best fitting US inflation forecasting models and are then compared to forecasts from a naive random walk model. The best models were non-linear autoregressive models based on kernel methods. Our findings do not provide much support for the usefulness of monetary aggregates in forecasting inflation. There is evidence in the literature that evolutionary methods can be used to evolve kernels hence our future work should combine the evolutionary and kernel methods to get the benefits of both.

Wealth shocks, credit-supply shocks, and asset allocation:evidence from household and firm portfolios

We use a unique dataset with bank clients’ security holdings for all German banks to examine how macroeconomic shocks affect asset allocation preferences of households and non-financial firms. Our analysis focuses on two alternative mechanisms which can influence portfolio choice: wealth shocks, which are represented by the sovereign debt crisis in the Eurozone, and credit-supply shocks which arise from reductions in borrowing abilities during bank distress. We document het- erogeneous responses to these two types of shocks. While households with large holdings of secu- rities from stressed Eurozone countries (Greece, Ireland, Italy, Portugal, and Spain) decrease the degree of concentration in their security portfolio as a result of the Eurozone crisis, non-financial firms with similar levels of holdings from stressed Eurozone countries do not. Credit-supply shocks at the bank level (caused by bank distress) result in lower concentration, for both households and non-financial corporations. We also show that only shocks to corporate credit bear ramifications on bank clients’ portfolio concentration, while shocks in retail credit are inconsequential. Our results are robust to falsification tests, propensity score matching techniques, and instrumental variables estimation.

Inflation convergence in Europe

Convergence has been a popular theme in applied economics since the seminal papers of Barro (1991) and Barro and Sala-i-Martin (1992). The very notion of convergence quickly becomes problematic from an academic viewpoint however when we try and formalise a framework to think about these issues. In the light of the abundance of available convergence concepts, it would be useful to have a more universal framework that encompassed existing concepts as special cases. Moreover, much of the convergence literature has treated the issue as a zero-one outcome. We argue that it is more sensible and useful for policy decision makers and academic researchers to consider also ongoing convergence over time. Assessing the progress of ongoing convergence is one interesting and important means of evaluating whether the Eastern European New Member Countries (NMC) of the European Union (EU) are getting closer to being deemed “ready” to join the European Monetary Union (EMU), that is, fulfilling the Maastricht convergence criteria.

Inflation forecasting, relative price variability and skewness

This article presents out-of-sample inflation forecasting results based on relative price variability and skewness. It is demonstrated that forecasts on long horizons of 1.5-2 years are significantly improved if the forecast equation is augmented with skewness. © 2010 Taylor & Francis.

Does money matter in inflation forecasting?

This paper provides the most fully comprehensive evidence to date on whether or not monetary aggregates are valuable for forecasting US inflation in the early to mid 2000s. We explore a wide range of different definitions of money, including different methods of aggregation and different collections of included monetary assets. In our forecasting experiment we use two nonlinear techniques, namely, recurrent neural networks and kernel recursive least squares regressiontechniques that are new to macroeconomics. Recurrent neural networks operate with potentially unbounded input memory, while the kernel regression technique is a finite memory predictor. The two methodologies compete to find the best fitting US inflation forecasting models and are then compared to forecasts from a nave random walk model. The best models were nonlinear autoregressive models based on kernel methods. Our findings do not provide much support for the usefulness of monetary aggregates in forecasting inflation. Beyond its economic findings, our study is in the tradition of physicists' long-standing interest in the interconnections among statistical mechanics, neural networks, and related nonparametric statistical methods, and suggests potential avenues of extension for such studies. © 2010 Elsevier B.V. All rights reserved.