Understanding machine learning-based forecasting methods: A decomposition framework and research opportunities

Machine learning (ML) methods are gaining popularity in the forecasting field, as they have shown strong empirical performance in the recent M4 and M5 competitions, as well as in several Kaggle competitions. However, understanding why and how these methods work well for forecasting is still at a very early stage, partly due to their complexity. In this paper, I present a framework for regression-based ML that provides researchers with a common language and abstraction to aid in their study. To demonstrate the utility of the framework, I show how it can be used to map and compare ML methods used in the M5 Uncertainty competition. I then describe how the framework can be used together with ablation testing to systematically study their performance. Lastly, I use the framework to provide an overview of the solution space in regression-based ML forecasting, identifying areas for further research.     

A fast and scalable ensemble of global models with long memory and data partitioning for the M5 forecasting competition

This work presents key insights on the model development strategies used in our cross-learning-based retail demand forecast framework. The proposed framework outperforms state-of-the-art univariate models in the time series forecasting literature. It has achieved 17th position in the accuracy track of the M5 forecasting competition, which is among the top 1% of solutions.     

Pricing used books on Amazon.com: a spatial approach to price dispersion

This paper estimates a spatial autoregressive (SAR) model of price dispersion using publicly available internet bookselling data. It uses a semiparametric adaptive estimator that does not require the usual Gaussian assumption of maximum likelihood (ML) estimators. The results suggest that both price competition and seller heterogeneity are key drivers of the observed price dispersion. The paper finds that sellers with large sales volume, newly established sellers and US mainland states-based sellers tend to price lower. The identified significant spatial interaction is evidence of spatial price competition. Controlling for everything else, a seller asks a lower price when large sellers charge relatively high prices, which is also evidence of price-based selling and undercutting.     

Bidding for Firms with Unknown Characteristics

When a region successfully attracts a firm by offering subsidies, the firm often commits itself to performance targets in terms of employment. In this paper, we interpret these firm‐specific targets as a consequence of incomplete information. We analyze a model of two regions that compete for a firm, assuming that the firm's productivity is ex ante unknown. We show that performance targets often induce overemployment in high‐productivity firms, and that tax credits are often superior to lump‐sum payments. Moreover, when regions differ in wage rates, the low‐wage region wins the bid and has a higher surplus than under complete information. Finally, we show that, under incomplete information, bidding might not lead to efficient firm location.     

The evolution of world trade from 1995 to 2014: A network approach

This paper employs network analysis to study world trade from 1995 to 2014. We focus on the main connective features of the world trade network (WTN) and their dynamics. Results suggest that countries’ efforts to attain the benefits of trade have resulted in an intertwined network that is increasingly dense, reciprocal, and clustered. However, these features do not correspond to a linear aggregation of the characteristics of its constituents (trade sectors). Trade linkages are distributed homogeneously among countries, but their intensity (i.e. their value) is highly concentrated in a small set of countries. The main connective features of the WTN were not affected by the 2007–2008 international financial crisis. However, we find that the crisis marks a turning point in the evolution of the WTN from a two-group (led by the US and Germany) to a three-group (led by the US, Germany, and China) hierarchical structure.     

政府间竞争影响城市土地市场化水平的双边效应研究：基于财政竞争和引资竞争的不同作用

研究目的:本文将政府间竞争划分为财政竞争和引资竞争,比较两者对城市土地市场化水平影响的大小,分析政府间竞争与城市土地市场化水平之间的关系。研究方法:基于2003—2016年全国284个地级市的面板数据,利用双边随机前沿模型测算财政竞争和引资竞争对城市土地市场化水平的影响。研究结果:财政竞争对城市土地市场化水平具有正向效应,提高了城市土地市场化水平1.61%;引资竞争对城市土地市场化水平具有负向效应,降低了城市土地市场化水平12.51%;政府间竞争总体上降低了城市土地市场化水平10.90%。时间趋势表明,政府间竞争对城市土地市场化水平的抑制作用呈现波动下降的状态。研究结论:政府间竞争降低了城市土地市场化水平。     

Determinants of the allocative,cost and scope efficiencies of Indian banks

ABSTRACT

This study aims to investigate the impact of competition on determinants of allocative, scope and cost efficiencies of Indian scheduled commercial banks (SCBs). Specifically, the study, analyzes the impact of the second round of licensing on the efficiency of Indian SCBs. This is the first paper to measure scope efficiency of Indian banks and analyze its determinants. A two-stage analysis is performed on a balanced panel dataset of Indian SCBs for the period 1999–2016. In the first stage, the allocative, cost and scope efficiencies for each bank are estimated following the data envelopment analysis approach. In the second stage, internal determinants of the stated efficiency measures are estimated following the system of the generalized method of moments approach. The findings suggest that competition has not resulted in enhancing the efficiency of Indian SCBs. Among factors that can influence efficiency, it is seen that size does matter. Larger banks can enhance the efficiency of SCBs. It is also seen that having more foreign banks improves the overall efficiency of SCBs. However, before embarking on further rounds of licensing, the study posits that market-driven correction to succeed, it is imperative to address sunspots in the form of investor or borrower repression.     

Global energy forecasting competition 2017: Hierarchical probabilistic load forecasting

The Global Energy Forecasting Competition 2017 (GEFCom2017) attracted more than 300 students and professionals from over 30 countries for solving hierarchical probabilistic load forecasting problems. Of the series of global energy forecasting competitions that have been held, GEFCom2017 is the most challenging one to date: the first one to have a qualifying match, the first one to use hierarchical data with more than two levels, the first one to allow the usage of external data sources, the first one to ask for real-time ex-ante forecasts, and the longest one. This paper introduces the qualifying and final matches of GEFCom2017, summarizes the top-ranked methods, publishes the data used in the competition, and presents several reflections on the competition series and a vision for future energy forecasting competitions.     

Doing more and doing better are two different entities: Different patterns of family control and environmental performance

Family firms bear two types of agency costs, including type I and type II agency problems, in corporate environmental practices: (1) Outside executives at family firms hesitate to engage in environmental strategies, which can lead to drops in profits; (2) Controlling families employ opportunistically environmental management to achieve their interests. We argue that a primary cause for the agency problems lies on ineffective internal corporate governance at family firms, which can cause loss of managerial (or power) balance between outside executives and family executives. Our findings show that family firms with ownership and strategic control (FSC), which family executives and outside executives monitor and constrain each other, can achieve the highest environmental performance. Moreover, external controls, including product market competition and provincial environmental regulations, substitute effective internal control of FSC. The environmental performance premium of FSC is more prevalent when the production market competition is lower. Family firms with ownership, operational, and strategic control (FOSC) can achieve higher environmental performance within a province with more stringent environmental regulations.     

M5 competition uncertainty: Overdispersion,distributional forecasting,GAMLSS, and beyond

The M5 competition uncertainty track aims for probabilistic forecasting of sales of thousands of Walmart retail goods. We show that the M5 competition data face strong overdispersion and sporadic demand, especially zero demand. We discuss modeling issues concerning adequate probabilistic forecasting of such count data processes. Unfortunately, the majority of popular prediction methods used in the M5 competition (e.g. lightgbm and xgboost GBMs) fail to address the data characteristics, due to the considered objective functions. Distributional forecasting provides a suitable modeling approach to overcome those problems. The GAMLSS framework allows for flexible probabilistic forecasting using low-dimensional distributions. We illustrate how the GAMLSS approach can be applied to M5 competition data by modeling the location and scale parameters of various distributions, e.g. the negative binomial distribution. Finally, we discuss software packages for distributional modeling and their drawbacks, like the R package gamlss with its package extensions, and (deep) distributional forecasting libraries such as TensorFlow Probability.