This thesis is concerned with applying frontier methods in econometric theory, to revisit relevant economic questions concerning productivity, technology and innovation. The focal point of this thesis is to try to employ econometric techniques that may allow us to depart from the usual assumptions of linearity or additivity of the model, as is usually found in economic literature. Therefore, a binding common ground of all the chapters of this thesis is relaxing these restrictive assumptions, in economic topics relevant to productivity and innovation, in micro and macro level. In this framework, recent advancements in nonparametric econometric theory allow relaxing such model restrictions. The first chapter estimates a production function at firm level that allows departing from the standard hypothesis of Hicks-neutral technical change. Simultaneously, it is coping with the endogeneity of innovation, the latter being considered a measure of TC. Parametric specifications that allow non-Hicks neutral technical change are derived. The chapter also presents testable conditions, for common parametric approximations, under which Hicks neutrality holds. Cobb-Douglas specifications are estimated adopting IV methods for heterogeneous effect of innovation on productivity. The empirical results reject Hicks neutrality towards the presence of a capital-saving TC. Finally, this chapter serves as a link with the nonparametric approaches developed in the following ones. In the second chapter, the issue of localization of technical change is addressed using the firm level dataset of the previous chapter. Recent advances on generalized kernel instrumental regression are exploited in order to revisit the relationship between innovation and productivity. This allows to highlight the possible presence of a localized effect of an endogenous innovation variable and, thanks to smoothing discrete variables, also to account for fully heterogeneous technologies across sectors. Such issues are extremely relevant from both a theoretical and a policy oriented perspective but cannot be addressed by adopting common parametric approaches. The issue of the predictive performances of this nonparametric estimator when compared to some parametric alternatives is also addressed. The results i) indicate that the proposed nonparametric estimator performs better than parametric ones and ii) reveal some relevant patterns that can only be detected using the nonparametric estimator. The third chapter revisits the issue of international R&D spillovers by using nonparametric methods, and tests the validity of the main results provided in the literature with respect to the possible existence of nonlinearities, threshold effects and non-additive relations. It considers a sieve estimation of a panel data model of technology diffusion among countries, paying attention to the issue of error cross sectional dependence. The adopted semiparametric approach is an extension of the parametric factor model by Pesaran (2006). The comparison between the parametric and the semiparametric approach reveals a better performance of the latter. From an economic viewpoint, the results show new evidence with respect to the benefits of countries from domestic and foreign R&D. Finally, the last chapter of the thesis is a review of the nonparametric kernel regression. Following, mainly, the work of Li and Racine (2007), it summarizes the key features of least squares cross validation and the kernel regression using local constant, local linear or local polynomial approaches. Moreover, the nonparametric IV kernel regression is presented, along with the relevant topics of ill-posed inverse problems and regularization methods. This chapter serves also as an informal “appendix” of the previous chapters, especially for the ones concerning kernel regression, because it provides useful insights of the underlying methodologies.

Il presente lavoro si propone di applicare metodi di avanguardia nella teoria econometrica ed, in quest’ottica di rivisitare una serie di questioni economiche relative ai temi della produttività, della tecnologia e dell’innovazione. Il primo capitolo della tesi si concentra sul contesto parametrico. La letteratura mainstream sull’econometria della produttività e sul cambiamento tecnologico (CT) assumono l’additività del CT, introducendo cosı la neutralità à la Hicks derivante dal CT all’interno del modello. In questa prospettiva, all’ interno del capitolo si stima una funzione di produzione a livello di azienda che permette di allontanarsi dalla ipotesi usuale di neutralità à la Hicks relativa al CT. In parallelo, il capitolo affronta il tema dell’ endogenita dell’innovazione la quale è considerata una misura della CT. Le specificazioni parametriche che consentono un cambiamento tecnico non Hicks-neutral sono derivate. Il capitolo presenta anche delle condizioni verificabili per approssimazioni parametriche comuni, in base alle quali vale la neutralità Hicks. Le specificazioni Cobb-Douglas sono stimate adottando i metodi IV per un effetto eterogeneo dell’innovazione sulla produttività. I risultati empirici rifiutano la neutralità à la Hicks rispetto alla presenza di progresso tecnico capital-saving. Il secondo capitolo è connesso ai temi espressi all’interno del primo in una pluralità di aspetti. Il tema della localizzazione del cambiamento tecnologico viene affrontata utilizzando la serie di dati sulle aziende del capitolo precedente. A tal proposito, al fine di riconsiderare il rapporto tra innovazione e produttività, sono impiegati i recenti progressi sulla regressione kernel strumentale generalizzata. Tale esercizio permette di evidenziare l’eventuale presenza degli effetti di localizzazione di una variabile di innovazione endogena e, grazie alla ”smoothing” delle variabili discrete, di tenere conto delle tecnologie completamente eterogenee tra i diversi settori. Tali questioni sono estremamente rilevanti sia da una prospettiva teorica sia in termini di policy implications, ma non possono essere affrontate con l’adozione di approcci parametrici comuni. Il capitolo affronta anche il tema delle prestazioni predittive degli stimatori non parametrici rispetto ad alcune alternative parametriche. I risultati i) indicano che lo stimatore non parametrico proposto produce risultati migliori di quelli parametrici e ii) evidenziano alcuni modelli rilevanti che possono essere rilevati solo utilizzando lo stimatore non parametrico. Il terzo capitolo riesamina la questione delle ricadute internazionali delle attività di ricerca e sviluppo utilizzando metodi non parametrici, e verifica la validità dei risultati principali forniti nella letteratura rispetto alla possibile esistenza di non linearità, effetti di soglia e di relazioni non additive. In quest’ottica, esso considera una stimasieve di un modello di dati panel di diffusione tecnologica tra i paesi, focalizzandosi in particolar modo sulla questione della dipendenza dell’errore sezione trasversale. L’approccio semiparametrico adottato è un’estensione del modello fattore parametrico da Pesaran (2006). Il confronto tra l’approccio parametrico e semiparametrico rivela una prestazione migliore di quest’ultimo. Dal punto di vista economico, i risultati mostrano nuove prove rispetto ai benefici dei programmi di ricerca e sviluppo nazionali ed esteri per i paesi. Infine, l’ultimo capitolo della tesi è una revisione della regressione kernel non parametrica. Seguendo, soprattutto, il lavoro di Li e Racine (2007), riassume le caratteristiche chiavi di ”least squares cross validation” e della regressione kernel usando approcci costanti locali, lineari locali o polinomi locali.

Essays in nonparametric econometrics with applications to the economics of productivity and innovation

GIOLDASIS, Georgios
2018

Abstract

This thesis is concerned with applying frontier methods in econometric theory, to revisit relevant economic questions concerning productivity, technology and innovation. The focal point of this thesis is to try to employ econometric techniques that may allow us to depart from the usual assumptions of linearity or additivity of the model, as is usually found in economic literature. Therefore, a binding common ground of all the chapters of this thesis is relaxing these restrictive assumptions, in economic topics relevant to productivity and innovation, in micro and macro level. In this framework, recent advancements in nonparametric econometric theory allow relaxing such model restrictions. The first chapter estimates a production function at firm level that allows departing from the standard hypothesis of Hicks-neutral technical change. Simultaneously, it is coping with the endogeneity of innovation, the latter being considered a measure of TC. Parametric specifications that allow non-Hicks neutral technical change are derived. The chapter also presents testable conditions, for common parametric approximations, under which Hicks neutrality holds. Cobb-Douglas specifications are estimated adopting IV methods for heterogeneous effect of innovation on productivity. The empirical results reject Hicks neutrality towards the presence of a capital-saving TC. Finally, this chapter serves as a link with the nonparametric approaches developed in the following ones. In the second chapter, the issue of localization of technical change is addressed using the firm level dataset of the previous chapter. Recent advances on generalized kernel instrumental regression are exploited in order to revisit the relationship between innovation and productivity. This allows to highlight the possible presence of a localized effect of an endogenous innovation variable and, thanks to smoothing discrete variables, also to account for fully heterogeneous technologies across sectors. Such issues are extremely relevant from both a theoretical and a policy oriented perspective but cannot be addressed by adopting common parametric approaches. The issue of the predictive performances of this nonparametric estimator when compared to some parametric alternatives is also addressed. The results i) indicate that the proposed nonparametric estimator performs better than parametric ones and ii) reveal some relevant patterns that can only be detected using the nonparametric estimator. The third chapter revisits the issue of international R&D spillovers by using nonparametric methods, and tests the validity of the main results provided in the literature with respect to the possible existence of nonlinearities, threshold effects and non-additive relations. It considers a sieve estimation of a panel data model of technology diffusion among countries, paying attention to the issue of error cross sectional dependence. The adopted semiparametric approach is an extension of the parametric factor model by Pesaran (2006). The comparison between the parametric and the semiparametric approach reveals a better performance of the latter. From an economic viewpoint, the results show new evidence with respect to the benefits of countries from domestic and foreign R&D. Finally, the last chapter of the thesis is a review of the nonparametric kernel regression. Following, mainly, the work of Li and Racine (2007), it summarizes the key features of least squares cross validation and the kernel regression using local constant, local linear or local polynomial approaches. Moreover, the nonparametric IV kernel regression is presented, along with the relevant topics of ill-posed inverse problems and regularization methods. This chapter serves also as an informal “appendix” of the previous chapters, especially for the ones concerning kernel regression, because it provides useful insights of the underlying methodologies.
MUSOLESI, Antonio
File in questo prodotto:
File Dimensione Formato  
thesis_gioldasis_FINAL.pdf

accesso aperto

Descrizione: thesis_gioldasis_FINAL
Tipologia: Tesi di dottorato
Dimensione 2.83 MB
Formato Adobe PDF
2.83 MB Adobe PDF Visualizza/Apri

I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11392/2488198
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact