MODELING UNCERTAIN FORECAST ACCURACY IN SUPPLY CHAINS WITH POSTPONEMENT

We examine a situation where a manufacturer operates in a two‐mode production environment. The first mode could involve overseas vendors and manufacturing facilities. If additional units are later required, the company must use its second mode—more expensive last‐minute domestic vendors and manufacturing sites. We develop a new methodology for analyzing the impact of forecast accuracy on the decision to postpone production. We examine the interaction of forecast accuracy, shortage vs. holding costs, transportation costs and the cost of postponing production in the supply chain of a single product facing uncertain demand. Our model can be used to analyze the cost of important changes, such as increasing forecast accuracy, reducing the cost of backorders, lowering the cost of delaying production, or lowering transportation costs. Our model allows a firm to understand its overall cost structure so that it can accurately evaluate the impact of improved forecast accuracy and lowered costs in the context of postponement.     

Comparing forecasts from fixed and variable coefficient models: The case of money demand

In this paper we introduce a class of tentatively plausible, fixed-coefficient models of money demand and evaluate their forecast performance. When these models are reestimated allowing all coefficients to vary over time, the forecasting performance improves dramatically. Aside from offering insights about improved methods of analyzing time series data, the most promising direct use for point estimates derived from time-varying coefficients is as an aid in calibrating proposed models of the kind discussed here.     

Forecasting and turning point predictions in a Bayesian panel VAR model

We provide methods for forecasting variables and predicting turning points in panel Bayesian VARs. We specify a flexible model, which accounts for both interdependencies in the cross section and time variations in the parameters. Posterior distributions for the parameters are obtained for hierarchical and for Minnesota-type priors. Formulas for multistep, multiunit point and average forecasts are provided. An application to the problem of forecasting the growth rate of output and of predicting turning points in the G-7 illustrates the approach. A comparison with alternative forecasting methods is also provided.     

Empirical evidence on individual, group and shrinkage seasonal indices

This paper provides empirical evidence on forecasting seasonal demand using both individual and group seasonal indices methods. The findings show that the group seasonal indices methods outperform the individual seasonal indices method. This paper also offers empirical results from comparing two shrinkage methods with the group seasonal indices methods. The theoretical rules developed by the authors for choosing between group seasonal indices and individual seasonal indices produce more accurate forecasts than do published rules for choosing between shrinkage methods, when measured by the MSE, and are competitive when measured by the symmetric MAPE.     

Shrinkage estimation of semiparametric multiplicative error models

Within models for nonnegative time series, it is common to encounter deterministic components (trends, seasonalities) which can be specified in a flexible form. This work proposes the use of shrinkage type estimation for the parameters of such components. The amount of smoothing to be imposed on the estimates can be chosen using different methodologies: Cross-Validation for dependent data or the recently proposed Focused Information Criterion. We illustrate such a methodology using a semiparametric autoregressive conditional duration model that decomposes the conditional expectations of durations into their dynamic (parametric) and diurnal (flexible) components. We use a shrinkage estimator that jointly estimates the parameters of the two components and controls the smoothness of the estimated flexible component. The results show that, from the forecasting perspective, an appropriate shrinkage strategy can significantly improve on the baseline maximum likelihood estimation.     

Forecasting monthly and quarterly time series using STL decomposition

This paper is a re-examination of the benefits and limitations of decomposition and combination techniques in the area of forecasting, and also a contribution to the field, offering a new forecasting method. The new method is based on the disaggregation of time series components through the STL decomposition procedure, the extrapolation of linear combinations of the disaggregated sub-series, and the reaggregation of the extrapolations to obtain estimates for the global series. Applying the forecasting method to data from the NN3 and M1 Competition series, the results suggest that it can perform well relative to four other standard statistical techniques from the literature, namely the ARIMA, Theta, Holt-Winters’ and Holt’s Damped Trend methods. The relative advantages of the new method are then investigated further relative to a simple combination of the four statistical methods and a Classical Decomposition forecasting method. The strength of the method lies in its ability to predict long lead times with relatively high levels of accuracy, and to perform consistently well for a wide range of time series, irrespective of the characteristics, underlying structure and level of noise of the data.     

The information content of regional employment data for forecasting aggregate conditions

We consider whether disaggregated data enhance the efficiency of aggregate employment forecasts. We find that incorporating spatial interaction into a disaggregated forecasting model lowers the out-of-sample mean squared error from a univariate aggregate model by 70% at a two-year horizon.     

An enhanced approach for Trend Impact Analysis

Decision makers in governments, corporations and institutions all need to forecast the future. Usually, traditional quantitative forecasting techniques are applied for this purpose. But the limitation of such methods is well known since all quantitative methods that are built solely on historical data (whether time-series or causal methods) produce forecasts by extrapolating such data into the future ignoring the effects of unprecedented future events that could cause deviation from the original surprise-free forecast if they were to occur. In the meanwhile, pure qualitative methods that don't utilize historical data miss its sound foundation. In the field of future studies, attempts are often made to combine quantitative and qualitative approaches using various hybrid methods such as Trend Impact Analysis. This paper introduces an advanced algorithm to enhance Trend Impact Analysis that adds another level of sophistication to the current algorithm. This advanced algorithm takes into account not only the impact of unprecedented future events' occurrences on the future trend, but also the different severity degrees with which the event might occur. This idea of severity degrees is novel, and its implementation is the main contribution of this paper.     

Predictive financial models of the euro area: A new evaluation test

This paper investigates the predictive ability of financial variables for euro area growth. Our forecasts are built from univariate autoregressive and single equation models. Euro area aggregate forecasts are constructed both by employing aggregate variables and by aggregating country-specific forecasts. The forecast evaluation is based on a recently developed test for equal predictive ability between nested models. Employing a monthly dataset from the period between January 1988 and May 2005 and setting the out-of-sample period to be from 2001 onwards, we find that the single most powerful predictor on a country basis is the stock market returns, followed by money supply growth. However, for the euro area aggregate, the set of most powerful predictors includes interest rate variables as well. The forecasts from pooling individual country models outperform those from the aggregate itself for short run forecasts, while for longer horizons this pattern is reversed. Additional benefits are obtained when combining information from a range of variables or combining model forecasts.