Content uploaded by J. Scott Armstrong
Author content
All content in this area was uploaded by J. Scott Armstrong
Content may be subject to copyright.
A preview of the PDF is not available
A Comparative Study of Methods for Long-Range Market Forecasting
Abstract
The following hypotheses about long-range market forecasting were examined:
H 1 Objective methods provide more accuracy than do subjective methods.
H 2 The relative advantage of objective over subjective methods increases as the amount of change in the environment increases.
H 3 Causal methods provide more accuracy than do naïve methods.
H 4 The relative advantage of causal over naïve methods increases as the amount of change in the environment increases.
Support for these hypotheses was then obtained from the literature and from a study of a single market. The study used three different models to make ex ante forecasts of the U.S. air travel market from 1963 through 1968.
These hypotheses imply that econometric methods are more accurate for long-range market forecasting than are the major alternatives, expert judgment and extrapolation, and that the relative superiority of econometric methods increases as the time span of the forecast increases.
Supplementary resources (3)
... Market forecasting for novel technologies is challenging [63][64][65][66][67] even in the case of much less disruptive technologies and more stable economic and social conditions. When forecasting the market potential for QTs, it is important to take better into account financial risks, and include societal risks, as they could impact the QTs adoption and growth potential not less than the purely technical risk. ...
Quantum technologies (QTs) hold the promise to transform a wide range of industries, such as computing, communications, finance, healthcare, defense, space, and beyond. Of the various QTs, the most relevant is presently quantum computing (QC), of significant projected market potential, with some estimates forecasting it to reach many billion dollars over the next few years. There are, however, risks to factor, as highlighted in this perspective, the most relevant technical, economic, and societal risks. Those financial and societal are projected to become increasingly relevant phased with the progressing solution of the technical issues. The synergy with artificial intelligence comes with further opportunities as development can be faster and more effective, but also increases risks. AI presents numerous opportunities, but it also comes with several risks and challenges. Some of the major risks associated with AI include bias and fairness, transparency and explainability, job displacement, security concerns, ethical concerns, privacy issues, lack of regulation and standards, and exponential growth and unintended consequences. Balancing the advantages of AI-driven quantum technologies with associated risks presents a significant challenge. It necessitates careful consideration of ethical, economic, and technical aspects to ensure that these technologies are developed and deployed in a manner that is beneficial and equitable for everyone.
... Hardly withstanding the emerging issue of parameter instability, the field of forecasting continues to rely on the out-of-sample forecasting procedures that were first discussed by Armstrong & Grohman (1972) over five decades ago. A crucial consideration remains unaddressed with traditional out-ofsample forecasting procedures: they rely on model recalibration. ...
... A previsão de vendas é essencial para o direcionamento dos recursos financeiros da empresa em estoques e, de acordo com Alalawin et al (2020), prever a demanda do consumidor com precisão é um dos principais objetivos das empresas para atender bem os seus consumidores, para que eles tenham satisfação e confiança na aquisição do produto ou serviço que a empresa oferece. Armstrong & Grohman (1972) em seu estudo de comparação de métodos de previsão de longo prazo, concluíram que métodos causais são mais acurados que métodos näive de previsão de demanda, a superioridade dos métodos causais aumenta à medida que o horizonte de tempo aumenta, e a regressão linear múltipla é classificada como método de previsão causal por Pedhazur (1982 Conforme Corrêa & Corrêa (2013), as previsões de médio prazo maiores que três meses, não possuem característica de repetir a mesma quantidade do passado como as previsões de curto prazo, mas é esperado que suas relações de vendas com outras variáveis que influenciam no comportamento, continuem contribuindo para o futuro e os autores recomendam a utilização de modelos causais para as previsões de médio prazo. ...
... A variation on the rolling-window design outlined by Armstrong and Grohman (1972) was used to generate forecasts for each city using each method. As Figure 3 illustrates, separate forecasts were produced using three years of training data beginning in each successive month. ...
... Therefore, the following multi period evaluation setups are introduced as opposed to the fixed origin setup. Armstrong and Grohman (1972) are among the first researchers to give a descriptive explanation of the rolling origin evaluation setup. Although the terms rolling origin setup and tsCV are used interchangeably in the literature, in addition to the forecast origin rolling forward, tsCV also allows to skip origins, effectively rolling forward by more than one step at a time (analogously to the difference between a leave-one-out CV and a k-fold CV). ...
Recent trends in the Machine Learning (ML) and in particular Deep Learning (DL) domains have demonstrated that with the availability of massive amounts of time series, ML and DL techniques are competitive in time series forecasting. Nevertheless, the different forms of non-stationarities associated with time series challenge the capabilities of data-driven ML models. Furthermore, due to the domain of forecasting being fostered mainly by statisticians and econometricians over the years, the concepts related to forecast evaluation are not the mainstream knowledge among ML researchers. We demonstrate in our work that as a consequence, ML researchers oftentimes adopt flawed evaluation practices which results in spurious conclusions suggesting methods that are not competitive in reality to be seemingly competitive. Therefore, in this work we provide a tutorial-like compilation of the details associated with forecast evaluation. This way, we intend to impart the information associated with forecast evaluation to fit the context of ML, as means of bridging the knowledge gap between traditional methods of forecasting and adopting current state-of-the-art ML techniques.We elaborate the details of the different problematic characteristics of time series such as non-normality and non-stationarities and how they are associated with common pitfalls in forecast evaluation. Best practices in forecast evaluation are outlined with respect to the different steps such as data partitioning, error calculation, statistical testing, and others. Further guidelines are also provided along selecting valid and suitable error measures depending on the specific characteristics of the dataset at hand.
... The first 3 weeks of the available dataset are used as a warm-up period on which the model is trained only. For the remaining dataset, a rolling-origin scheme (Armstrong and Grohman 1972) is applied. This is an iterative approach that simulates how new information becomes incrementally available for training over time in an operational setting. ...
A wealth of forecasting models is available for operational weather forecasting. Their strengths often depend on the lead time considered, which generates the need for a seamless combination of different forecast methods. The combined and continuous products are made in order to retain or even enhance the forecast quality of the individual forecasts and to extend the lead time to potentially hazardous weather events. In this study, we further improve an artificial neural network based combination model that was recently proposed in a previous paper. This model combines two initial precipitation ensemble forecasts and produces exceedance probabilities for a set of thresholds for hourly precipitation amounts. Both initial forecasts perform differently well for different lead times, whereas the combined forecast is calibrated and outperforms both initial forecasts with respect to various validation scores and for all considered lead times (+1h to +6h). Moreover, the robustness of the combination model is tested by applying it to a new dataset and by evaluating the spatial and temporal consistency of its forecasts. The changes proposed further improve the forecast quality and make it more useful for practical applications. Temporal consistency of the combined product is evaluated using a flip-flop index. It is shown that the combination provides a higher persistence with decreasing lead times compared to both input systems.
This study introduces a coalesce forecasting model tailored for flood-prone regions, specifically focusing on Bihar, India. Research has revealed significant disparities in rainfall patterns across various zones such as Tirhut, Patna, and Munger zones experiencing greater mean rainfall than Bhagalpur and Kosi. To evaluate the forecasting capabilities, coalescing methods were applied which includes the autoregressive integrated moving average (ARIMA), exponential smoothing state space (ETS), neural network autoregressive (NNAR), and seasonal-trend decomposition. Moreover, Loess (STL) methods, and trigonometric seasonality, Box‒Cox transformation, ARMA errors, and trend and seasonal components (TBATS) were also employed to contrast the benchmark models such as the seasonal naïve, naïve, and mean methods. These methods were evaluated using error evaluators such as residual error, root mean square error (RMSE), mean absolute error (MAE), mean absolute scaled error (MASE), and autocorrelation of errors at lag 1 (ACF1) to determine the performance of these techniques. Additionally, statistical tests, such as the Box–Pierce and Box–Ljung tests, supported these findings. Among the error evaluators and forecasting models, the ETS and NNAR models remain the top choices for Saran-Tirhut-Bhagalpur and Munger-Magadh-Kosi, respectively, effectively capturing rainfall patterns and minimizing residual errors, as indicated by low RMSE values. Moreover, ARIMA and TBATS remain the top choices for Patna, Purnia and Darbhanga, respectively, followed by ETS model. In addition, the STL model secured the second position for Saran, Tirhut, Bhagalpur, and Purnia zones. This research emphasizes the importance of understanding regional rainfall dynamics for effective flood risk management and climate adaptation strategies. This study provides valuable tools for water resource management and agricultural planning in Bihar amidst climate variability challenges. It advocates for rainfall trend analysis followed by forecasting to achieve more precise water resource management and planning.
Many post‐processing methods improve forecasts at individual locations but remove their correlation structure. However, this information is essential for forecasting larger‐scale events, such as the total precipitation amount over areas like river catchments, which are relevant for weather warnings and flood predictions. We propose a method to reintroduce spatial correlation into a post‐processed forecast using an R‐vine copula fitted to historical observations. The method rearranges predictions at individual locations and ensures that they still exhibit the post‐processed marginal distributions. It works similarly to well‐known approaches, like the “Schaake shuffle” and “ensemble copula coupling.” However, compared to these methods, which rely on a ranking with no ties at each considered location in their source for spatial correlation, the copula serves as a measure of how well a given arrangement compares with the observed historical distribution. Therefore, no close relationship is required between the post‐processed marginal distributions and the spatial correlation source. This is advantageous for post‐processed seamless forecasts in two ways. First, meteorological parameters such as the precipitation amount, whose distribution has an atom at zero, have rankings with ties. Second, seamless forecasts represent an optimal combination of their input forecasts and may spatially shifted from them at scales larger than the areas considered herein, leading to non‐reasonable spatial correlation sources for the well‐known methods. Our results indicate that the calibration of the combination model carries over to the output of the proposed model, that is, the evaluation of area predictions shows a similar improvement in forecast quality as the predictions for individual locations. Additionally, the spatial correlation of the forecast is evaluated with the help of object‐based metrics, for which the proposed model also shows an improvement compared to both input forecasts.
Many researchers appear to operate under the impression that causal models lead to more accurate forecasts than those provided by naive models (or “projections”). This study was based on the premise that causal models lead to better forecasts than do naive models in certain situations. The key element of these situations is that there are “large changes.” One situation where large changes might be expected is that of long-range forecasting — and, in particular, long-range forecasting for international markets. Recent improvements in the quality and availability of international data have substantially reduced the cost of developing causal models in this situation. A study of camera markets in seventeen countries indicated that the margin of superiority of causal models over naive models is of great practical importance.
Economic statisticians do not enjoy an untarnished reputation for accurate forecasting. We have managed, over the years, to come up with some memorable failures. While we have also had our share of successes, they are not as well remembered nor as numerous as we should like. Recently, however, we have begun to pay more attention to the record, and a substantial body of evidence on forecasting performance has accumulated. In this paper I propose to review this record, try to arrive at a balanced appraisal, and offer some suggestions for improvement.
Long-term forecasts for cars, selected consumer durables and energy
- O Herlihy
O'herlihy, C. et al. (1967), Long-term forecasts for cars, selected consumer durables and energy, National Institute Economic Review, 40 (May), 34-61