MLP ensembles improve long term prediction accuracy over single networks

This work describes an award winning approach for solving the NN3 Forecasting Competition problem, focusing on the sound experimental validation of its main innovative feature. The NN3 forecasting task consisted of predicting 18 future values of 111 short monthly time series. The main feature of the approach was the use of the median for combining the forecasts of an ensemble of 15 MLPs to predict each time series. Experimental comparison to a single MLP shows that the ensemble increases the performance accuracy for multiple-step ahead forecasting. This system performed well on the withheld data, having finished as the second best solution of the competition with an SMAPE of 16.17%.     

Measuring Market Conduct in the Brazilian Cement Industry: A Dynamic Econometric Investigation

Indirect assessments of market conduct have become widespread in the New Empirical Industrial Organization—NEIO literature. Recently, Steen and Salvanes (1999, Int J Ind Organ 17:147–177) provided a flexible dynamic econometric formulation of the approach advanced by Bresnahan (1982, Econ Lett 19:87–92) and Lau (1982, Econ Lett 10:93–99). The present paper considers a similar approach applied to regional cement markets in Brazil under more favorable data availability, and attempts to address part of the concerns that usually emerge with respect to the NEIO literature. In particular, issues pertaining to structural stability and the control for the number of competing firms are addressed. The evidence clearly indicates non-negligible and distinct market power in different regions and also distinct conduct patterns in the short and long run.      

Should I say it in English? Exploring language effects on print advertising among Latin American bilinguals

ABSTRACT

Despite the extensive research conducted on language standardization in advertising across several countries, little attention has been given to the use of English versus Spanish and code-switching when advertising to Latin American bilingual consumers. We propose that stereotypes about English speakers and code-switching have potential to help determine which language is most effective in print advertising. The results of experiments conducted in Chile, Ecuador, and Mexico show that the effects of language-related stereotypes on the persuasiveness of English ads vary across different countries. In the case of Chile, English may be persuasively superior, depending on the favourableness of the stereotype of individuals that code-switch. Differently, print ads in English are functionally equivalent, in terms of ad attitudes, to Spanish and code-switching ads in Mexico, and superior in Ecuador, regardless of the favourability of the language-related stereotypes. Suggestions for advertising practice and future research are offered.     

Management forecast credibility and underreaction to news

In this paper, we first document evidence of underreaction to management forecast news. We then hypothesize that the credibility of the forecast influences the magnitude of this underreaction. Relying on evidence that more credible forecasts are associated with a larger reaction in the short window around the management forecasts and a smaller post-management forecast drift in returns, we show that the magnitude of the underreaction is smaller for firms that provide more credible forecasts. Our paper contributes to the literature by providing out-of-sample evidence of the drift in returns documented in the post-earnings-announcement drift literature, with the credibility of the news being one explanation for the phenomenon.     

Regional relationships between inherent coffee quality and growing environment for denomination of origin labels in Nariño and Cauca,Colombia

International markets are increasingly signaling demand for quality-differentiated coffee, which the Colombian Coffee Growers Federation (FNC) proposed to exploit to identify those regional coffees that would fulfill the requirements to be classified as denomination of origin. The objective of this study was to develop and implement a sound, robust and repeatable approach with and for the FNC to identify regional causal relationships between coffee quality and environmental characteristics as bases for labels of denomination of origin. Environmental differences between coffee-growing areas in the departments of Cauca and Nariño were statistically significant for several characteristics, including the number of dry months, annual precipitation and diurnal temperature range. The dominant varieties (Caturra and Colombia) did not show major differences in quality attributes, and were pooled for the analyses with the environmental data. There are significant differences in biochemical and sensorial product characteristics between the two departments. The spatial patterns in product characteristics exhibit a non-random, regionally-changing structure that is related to those in the environmental data. The generated results provided ample evidence to support the application for regionally-based denominations of origin. Recommendations were derived to help mainstreaming the developed approach and thereby facilitate policy decisions for its use in other geographies and with other crops. Furthermore, the importance of systematic interdisciplinary institutional collaboration for large-scale denomination of origin projects was corroborated for food policy dialogue and decision making. It seems plausible that producers of high-quality products within other commodities are likely to follow the FNC in seeking denomination of origin for their goods. The presented approach is crucial to facilitate policy.