Learning to Classify the Consumer Confidence Indicator (in Portugal)

António Bento Caleiro

doi:10.3390/economies9030125

Abstract

The literature paid some attention, some time ago, to the relationship that, supposedly, should exist between the level of consumer confidence and the unemployment rate. This relationship is interesting, both from a scientific point of view, given the inherently subjective character of that level of confidence, but also from the point of view of economic policy, given the importance of the unemployment rate. In this article, that relationship is revisited, using learning models, namely regression and classification trees. Using, for example, the case of Portugal, the unemployment rate presents itself as an adequate classifier of the consumer confidence level. The use of classification trees shows that the separation between low and high values of the consumer confidence indicator is made from an adequate threshold value of the unemployment rate. The use of regression trees shows that the levels of consumer confidence are inversely related to the levels of the unemployment rate. In terms of policy lessons, this confirms that, in the face of economic crises, such as the one we are experiencing, in which confidence levels tend to fall and the unemployment rate increases, the relationship between these two variables cannot be ignored.

Highlights

The severity of the current economic crisis, that many associate with the pandemic outbreak, is generally affecting all countries throughout the world
(an important) part of that severity can be attributed to the lack of knowledge of the relationships that are established between the main economic variables, both at the real and monetary levels. From this point of view, the success of a new economic order depends on the (a)knowledge(ment) of these relationships, namely those that involve, in particular, the level of confidence and the unemployment rate, given the characteristics of the current economic crisis; see van Giesen and Pieters (2019) for further information on these characteristics
The following criteria can be used: Least squares; Least absolute deviations. The use of these criteria, accompanied by a stopping rule—for example, a predeterThis sectionofpresents of the in application of function the methodology in the remined number leaves orthe an results improvement the objective smaller than a certain gression and classification of the level of consumer confidence, according to the unemthreshold—gives rise to the final tree, which should not have as many terminal nodes, i.e., ployment this end, the R package ‘tree’ provided by Ripley (2019) was used

Summary

Introduction

The severity of the current economic crisis, that many associate with the pandemic outbreak, is generally affecting all countries throughout the world. (an important) part of that severity can be attributed to the lack of knowledge (or, at the minimum, nonacknowledgement) of the relationships that are established between the main economic variables, both at the real and monetary levels. From this point of view, the success of a new economic order depends on the (a)knowledge(ment) of these relationships, namely those that involve, in particular, the level of confidence and the unemployment rate, given the characteristics of the current economic crisis; see van Giesen and Pieters (2019) for further information on these characteristics. Since the unemployment rate should be viewed as an explanatory factor of the level of consumer confidence, this fact can confirm, or reinforce, the causal relationship that, allegedly, exists between unemployment and economic growth

Objectives

Results

Discussion

Conclusion