The purposes of this work are to develop several models for the estimation of reputational losses as consequences of operative events and to study how and how much the type of the event have an impact on these estimations. In particular we focused on four companies belonging to the Oil&Gas sector. In the first part of this work an analysis on non catastrophic events is performed and focused on a particular company. Four different models are suggested for the evaluation of reputational losses and moving-window local model is chosen in which are considered also data of Price Sensitive events, i.e. events that do not belong to reputational ones but that are able to condition share value too, in order to diversify effects that do not come from reputation. In this model the window around the date of the event, defining the local dataset used to estimate the loss, is made by two Price Sensitive events near the reputational event. Estimations computed from this model are used to perform weighted ANOVA in order to identify which types of events have more impact on reputation. Category and Location of the events are much more significant than other features. In particular, dividing Category into three groups taken from so called Sustainability Principles (Economic, Environmental and Social), results led to the conclusions that this company seems to be exposed to environmental or economical problems. After repeated same analysis using simple local model instead of moving-window local model on other competitor companies, A, B and C, results are compared in order to understand if every company has his own influent type of events. In particular a joint dataset is created with all relative losses and features. Location and Category are still significant and Category effect depends on which company is selected. The second part of this thesis analyzes another dataset about catastrophic events which considers operative losses amounts. The aim is to develop an analysis which involves estimations of relative reputational losses, by simple local model, losses amounts and features of the events using ANCOVA modeling, in order to find what kind of relation exist between reputational losses and losses amounts for every type of events. However small size of the dataset creates problem in data fitting.
Questa tesi presenta come obiettivi lo sviluppo di modelli di stima delle perdite reputazionali accusate da società appartenenti al settore industriale dell’Oil&Gas in seguito ad eventi di perdita operativa e lo studio di quali tipologie di eventi hanno maggiore influenza sulle stime stesse. Nella prima parte dell’analisi riferita ad eventi reputazionali non catastrofici si è focalizzata l’attenzione su una compagnia e sono stati proposti quattro diversi modelli per la stima della perdita reputazionale percentuale del singolo evento. Dopo aver evidenziato i pro e i contro la scelta è ricaduta sul modello locale a finestra mobile in quanto sembra il più adatto a descrivere i dati a disposizione. Attraverso le stime ottenute è stato sviluppato un modello ANOVA in modo da quantificare gli effetti che hanno diversi tipi di evento. Dopo aver ripetuto le analisi per tre ulteriori società concorrenti ne sono stati paragonati i risultati attraverso la creazione di un dataset congiunto. La seconda parte di questa tesi tratta di eventi catastrofici di cui si possiedono anche i valori monetari della perdita operativa accusata. Si è sfruttato il modello locale semplice per la stima delle perdite relative a partire dalle quali è stata poi svolto un modello ANCOVA in modo da studiare la relazione tra le diverse tipologie di evento e la perdita monetaria percentuale.
Sviluppo di modelli di stima per l'analisi del rischio reputazionale nel settore Oil & Gas
BELLONI, CLAUDIA
2012/2013
Abstract
The purposes of this work are to develop several models for the estimation of reputational losses as consequences of operative events and to study how and how much the type of the event have an impact on these estimations. In particular we focused on four companies belonging to the Oil&Gas sector. In the first part of this work an analysis on non catastrophic events is performed and focused on a particular company. Four different models are suggested for the evaluation of reputational losses and moving-window local model is chosen in which are considered also data of Price Sensitive events, i.e. events that do not belong to reputational ones but that are able to condition share value too, in order to diversify effects that do not come from reputation. In this model the window around the date of the event, defining the local dataset used to estimate the loss, is made by two Price Sensitive events near the reputational event. Estimations computed from this model are used to perform weighted ANOVA in order to identify which types of events have more impact on reputation. Category and Location of the events are much more significant than other features. In particular, dividing Category into three groups taken from so called Sustainability Principles (Economic, Environmental and Social), results led to the conclusions that this company seems to be exposed to environmental or economical problems. After repeated same analysis using simple local model instead of moving-window local model on other competitor companies, A, B and C, results are compared in order to understand if every company has his own influent type of events. In particular a joint dataset is created with all relative losses and features. Location and Category are still significant and Category effect depends on which company is selected. The second part of this thesis analyzes another dataset about catastrophic events which considers operative losses amounts. The aim is to develop an analysis which involves estimations of relative reputational losses, by simple local model, losses amounts and features of the events using ANCOVA modeling, in order to find what kind of relation exist between reputational losses and losses amounts for every type of events. However small size of the dataset creates problem in data fitting.| File | Dimensione | Formato | |
|---|---|---|---|
|
2013_12_Belloni.pdf
Open Access dal 25/11/2014
Dimensione
9.26 MB
Formato
Adobe PDF
|
9.26 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/88642