A self-organizing map analysis of survey-based agents׳ expectations before impending shocks for model selection: The case of the 2008 financial crisis

Abstract

A self-organizing map analysis of survey-based agents' expectations before impending shocks for model selection: The case of the 2008 financial crisis ABSTRACT By means of Self-Organizing Maps we cluster fourteen European countries according to the most suitable way to model their agents' expectations. Using the financial crisis of 2008 as a benchmark, we distinguish between those countries that show a progressive anticipation of the crisis and those where sudden changes in expectations occur. We compare the forecasting performance of Artificial Neural Networks to that of three different time series models and find that neural networks outperform time series models when there are brisk changes in expectations before impending shocks. Conversely, in countries where expectations show a smooth transition towards recession, ARIMA models outperform neural networks. This result demonstrates the usefulness of clustering techniques for selecting the most appropriate method to forecast expectations according to their behaviour.

Clustering and forecasting inflation expectations using the World Economic Survey : the case of the 2014 oil price shock on inflation targeting countries

“…Then the results are evaluated by a competitive function that produce a wining neuron (Best Matching unit). The weights are updated according a learning rule, equation (1), and the neuron's neighborhood are updated too. See the…”

“…The training stage for each iteration consists in adjusting the weighs of the winning neuron and its neighbors by using the learning rule. This process guarantees similarity between the inputs 1 In the NAR-NN Model to perform multi-steps forecasts the network is transformed into a recurrent network after their parameters were trained as a feed forward network. Kohonen suggested to use rectangular and hexagonal neighborhoods.…”

“…The objective of this article is twofold: firstly, we rely on Claveria et al (2016) (1) , and make use of the Self-organizing maps, SOM, as a clustering technique on the agent's expectations for some targeting countries to discriminate them after the oil's shock of 2014 between soft or brisk behaviors. Then, we predict the inflation expectations evaluating, at different horizons, the forecasts from different models specifications of the Non-linear Autoregressive Neural network (NAR-NN).…”

Tracking Economic Growth by Evolving Expectations Via Genetic Programming: A Two-Step Approach

“…The link between survey expectations and quantitative data at the aggregate level has been widely investigated (Abberger, 2007;Dua, 1998, 1992;Bergström, 1995;Berk, 1999;Bovi, 2013;Bruestle andCrain, 2015, Bruno, 2014;Claveria et al, 2007;Claveria et al, 2016aClaveria et al, , 2017bDees and Brinca, 2013;Driver and Urga, 2004;Graff, 2010;Hansson et al, 2005;Jean-Baptiste, 2012;Kauppi et al, 1996;Leduc and Sill, 2013;Lee, 1994;Lehmann and Wohlrabe, 2017;Mittnik and Zadrozny, 2005;Nardo, 2003;Nolte and Pohlmeier, 2007;Pesaran and Weale, 2006;Qiao et al, 2009, Rahiala andTeräsvirta, 1993;Robinzonov et al, 2012;Smith and McAleer, 1995;Sorić et al, 2013;Vermeulen, 2014;Wilms et al, 2016). Since survey data are approximations of unobservable expectations, they inevitably entail a measurement error.…”

Forecasting Inflation Expectations from the CESifo World Economic Survey: An Empirical Application in Inflation Targeting

“…This paper examines the data on inflation expectations from the WES for 16 inflationtargeting countries. Then, by making use of Self-Organizing Maps (SOM) we cluster agents' expectations for these countries to classify them either as "soft" or "brisk" based on the speed of their expectations change after the oil shock of 2014 (Claveria, Monte and Torra, 2016).…”