|
on Forecasting |
| By: | Roberto Patuelli (Department of Spatial Economics, Vrije Universiteit Amsterdam); Aura Reggiani (Department of Economics, University of Bologna, Italy); Peter Nijkamp (Department of Spatial Economics, Vrije Universiteit Amsterdam); Uwe Blien (Institut für Arbeitsmarkt und Berufsforschung (IAB), Nuremberg) |
| Abstract: | In this paper, a set of neural network (NN) models is developed to compute short-term forecasts of regional employment patterns in Germany. NNs are modern statistical tools based on learning algorithms that are able to process large amounts of data. NNs are enjoying increasing interest in several fields, because of their effectiveness in handling complex data sets when the functional relationship between dependent and independent variables is not explicitly specified. The present paper compares two NN methodologies. First, it uses NNs to forecast regional employment in both the former West and East Germany. Each model implemented computes single estimates of employment growth rates for each German district, with a 2-year forecasting range. Next, additional forecasts are computed, by combining the NN methodology with Shift-Share Analysis (SSA). Since SSA aims to identify variations observed among the labour districts, its results are used as further explanatory variables in the NN models. The data set used in our experiments consists of a panel of 439 German districts. Because of differences in the size and time horizons of the data, the forecasts for West and East Germany are computed separately. The out-of-sample forecasting ability of the models is evaluated by means of several appropriate statistical indicators. |
| Keywords: | networks; forecasts; regional employment; shift-share analysis; shift-share regression |
| JEL: | C23 E27 R12 |
| Date: | 2006–02–17 |
| URL: | http://d.repec.org/n?u=RePEc:dgr:uvatin:20060020&r=for |
| By: | Jacob Bikker; Laura Spierdijk; Roy Hoevenaars; Pieter Jelle van der Sluis |
| Abstract: | Often, a relatively small group of trades causes the major part of the trading costs on an investment portfolio. For the equity trades studied in this paper, executed by the world’s second largest pension fund, we find that only 10% of the trades determines 75% of total market impact costs. Consequently, reducing the trading costs of comparatively few expensive trades would already result in substantial savings on total trading costs. Since trading costs depend to some extent on controllable variables, investors can try to lower trading costs by carefully controlling these factors. As a first step in this direction, this paper focuses on the identification of expensive trades before actual trading takes place. However, forecasting market impact costs appears notoriously difficult and traditional methods fail. Therefore, we propose two alternative methods to form expectations about future trading costs. The first method uses five ‘buckets’ to classify trades, where the buckets represent increasing levels of market impact costs. Each trade is assigned to a bucket depending on the probability that the trade will incur high market impact costs. The second method identifies expensive trades by considering the probability that market impact costs will exceed a critical level. When this probability is high, a trade is classified as potentially expensive. Applied to the pension fund data, both methods succeed in filtering out a considerable number of trades with high trading costs and substantially outperform no-skill prediction methods. The results underline the productive role that model-based forecasts can play in trading cost management. |
| Keywords: | market impact costs; forecasting; institutional trading; trading cost management. |
| JEL: | G11 G23 C53 |
| Date: | 2006–03 |
| URL: | http://d.repec.org/n?u=RePEc:dnb:dnbwpp:095&r=for |
| By: | Ilias Lekkos (EFG Eurobank); Costas Milas (Keele University); Theodore Panagiotidis (Loughborough University) |
| Abstract: | This paper explores the ability of factor models to predict the dynamics of US and UK interest rate swap spreads within a linear and a non-linear framework. We reject linearity for the US and UK swap spreads in favour of a regime-switching smooth transition vector autoregressive (STVAR) model, where the switching between regimes is controlled by the slope of the US term structure of interest rates. We compare the ability of the STVAR model to predict swap spreads with that of a non-linear nearest-neighbours model as well as that of linear AR and VAR models. We find some evidence that the non-linear models predict better than the linear ones. At short horizons, the nearest-neighbours (NN) model predicts better than the STVAR model US swap spreads in periods of increasing risk conditions and UK swap spreads in periods of decreasing risk conditions. At long horizons, the STVAR model increases its forecasting ability over the linear models, whereas the NN model does not outperform the rest of the models. |
| Keywords: | Interest rate swap spreads, term structure of interest rates, factor models, regime switching, smooth transition models, nearest-neighbours, forecasting. |
| JEL: | C51 C52 C53 E43 |
| Date: | 2006–03 |
| URL: | http://d.repec.org/n?u=RePEc:lbo:lbowps:2006_6&r=for |
| By: | Raj Aggarwal; Brian M. Lucey; Sunil K. Mohanty |
| Abstract: | An important puzzle in international finance is the failure of the forward exchange rate to be a rational forecast of the future spot rate. It has often been suggested that this puzzle may be resolved by using better statistical procedures that correct for both non-stationarity and nonnormality in the data. We document that even after accounting for non-stationarity, nonnormality, and heteroscedasticity using parametric and non-parametric tests on data for over a quarter century, US dollar forward rates for horizons ranging from one to twelve months for the major currencies, the British pound, Japanese yen, Swiss franc, and the German mark, are generally not rational forecasts of future spot rates. These findings of non-rationality in forward exchange rates for the major currencies continue to be puzzling especially as these foreign exchange markets are some of the most liquid asset markets with very low trading costs. |
| Keywords: | flight-to-quality, contagion, multivariate GARCH |
| JEL: | F31 G14 F47 G15 |
| Date: | 2006–04–05 |
| URL: | http://d.repec.org/n?u=RePEc:iis:dispap:iiisdp123&r=for |
| By: | Millimet, Daniel (SMU); Henderson, Daniel (SUNY-Binghamton) |
| Abstract: | Despite the solid theoretical foundation on which the gravity model of bilateral trade is based, empirical implementation requires several assumptions which do not follow directly from the underlying theory. First, unobserved trade costs are assumed to be a (log-) linear function of observables. Second, the ad-valorem tax equivalents of trade costs are predominantly assumed to be constant across space, and to a lesser extent time. Maintaining consistency with the underlying theory, but relaxing these assumptions, we estimate gravity models ?in levels and logs ?using two data sets via nonparametric methods. The results are striking. Despite the added flexibility of the nonparametric models, parametric models based on these assumptions offer equally or more reliable in-sample forecasts (sometimes) and out-of-sample forecasts (always), particularly in the levels model. Moreover, formal statistical tests fail to reject the theoretically consistent parametric functional form. Thus, concerns in the gravity literature over functional form appear unwarranted, and estimation of the gravity model in levels is recommended. |
| Keywords: | Gravity Model, Bilateral Trade, Border Effect, Currency Union, Generalized Kernel Estimation |
| JEL: | C14 C33 F14 |
| Date: | 2006–04–05 |
| URL: | http://d.repec.org/n?u=RePEc:smu:ecowpa:0517&r=for |