Heuristic Optimization Methods for Dynamic Panel Data Model Selection: Application on the Russian Innovative Performance

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Innovations, be they radical new products or technology improvements are widely recognized as a key factor of economic growth. To identify the factors triggering innovative activities is a main concern for economic theory and empirical analysis. As the number of hypotheses is large, the process of model selection becomes a crucial part of the empirical implementation. The problem is complicated by the fact that unobserved heterogeneity and possible endogeneity of regressors have to be taken into account. A new efficient solution to this problem is suggested, applying optimization heuristics, which exploits the inherent discrete nature of the problem. The model selection is based on information criteria and the Sargan test of overidentifying restrictions. The method is applied to Russian regional data within the framework of a log-linear dynamic panel data model. To illustrate the performance of the method, we also report the results of Monte-Carlo simulations. Terms of use: Documents in EconStor may

Labour Market Performance in OECD Countries: A Comprehensive Empirical Modelling Approach of Institutional Interdependencies

“…In particular, we apply heuristic optimization techniques as advocated by e.g. Savin and Winker (2012), Acosta-González and Fernández-Rodríguez (2007) and Kapetanios (2007). These are able to deal with non-smooth, discrete optimization problems.…”

“…Recently, Acosta-González and Fernández-Rodríguez (2007), Kapetanios (2007) and Savin and Winker (2012) have shown that heuristic optimization techniques based on information criteria as objective function deliver promising results in selecting regressors in different model selection set-ups of multiple regression models. 7 Kapetanios (2007) documents an outperformance of Simulated Annealing (SA) and Genetic Algorithms (GA) as two classical heuristic methods over PcGets.…”

“…7 Kapetanios (2007) documents an outperformance of Simulated Annealing (SA) and Genetic Algorithms (GA) as two classical heuristic methods over PcGets. The closest to our approach is the model selection set-up by Savin and Winker (2012) who apply heuristics to a (dynamic) panel model and identify a genetic algorithm as being the best model selection strategy for this type of model. Hence, we apply a genetic algorithm to select relevant institutional interactions.…”

“…We choose an information criterion as loss function in order to derive a high model fit by simultaneously penalizing for overparametrization. As the Bayesian Information Criterion (BIC) is consistent and seems to deliver superior results in model selection as pointed out by Kapetanios (2007) and Savin and Winker (2012), we select BIC as our target function. To avoid model misspecifications, we add a penalty to the objective function if the selected model has serially correlated errors based on the test developed by Born and Breitung (2013).…”

“…8 The heuristic optimization methods are mostly designed in line with the suggestions of Savin and Winker (2012) for dynamic panel model selection. Genetic algorithms (GA) update the whole set of solutions simultaneously (see also table 1 for a pseudocode).…”