The availability of timely, comprehensive and disaggregate indicators has become in the last years one of the most important issues for the economic analysis. Among macroeconomic series, GDP is considered as the one carrying highest informative content, nevertheless the official National Accounts release of its value is published with strong delay. In order to fill this gap many researchers have attempted to get an indirect measure of the state of the economy exploiting disaggregated series related to GDP and available at higher frequency. Among official data for Europe the timeliest data are the Business and Consumer Surveys of the European Commission, published at the end of each month and referring to the same month. Within respect to this advantage in term of timeliness it remains however to understand how to use in quantitative terms the information from Surveys, which is primarily qualitative. To address this issue we present in the first Chapter two new methods for the quantification of the qualitative information provided by firms in the Business Survey. The Spectral Envelope, as well as the Cumulative Logit model with a non linear Kalman filter is applied to get a quantified indicator of the level of production. The main result is that this indicator is high coherent with the cycle of the Industrial production, which is extracted independently. The issue of GDP nowcasting is carried out extensively in the second Chapter, where the focus is the disaggregation in time of the quarterly value added for the Euro area. By using a mixed frequency model that relies on quarterly and monthly series, the monthly indicator for GDP is obtained casting the traditional Stock and Watson (1991) model in state space form, as in Mariano and Murasawa (2003). The estimation is carried out from the output (sectors of activities) and the expenditure side (demand components) of GDP, which both contribute to the final computation on the basis of their precision. The chain linked nature of the National Account series is also considered. The results of this application show that, in contrast with some recent literature, Survey data do not matter for the estimation of GDP. More appropriate investigations on the role of Survey data for the estimation of macroeconomic variables is provided in the third Chapter, where the basic model presented in the previous Chapter is extended to allow for more than one common factor and low frequency cycles. After some analysis of real time data and revisions, the main conclusion is that Survey data in a richest model are indeed useful to produce more accurate estimates and forecast, especially 3 steps ahead of the current released of GDP and when hard data are not available yet. We conclude our analysis on GDP estimation considering a mean-variance coincident index for the US economy. There is general consensus on the fact that most of the macroeconomic series have become less volatile in the last 20 year in US. This issue is addressed by extending the framework developed for Europe in the previous Chapters with a two regimes Garch-type model for the fluctuations of the economy that mimic the dynamic of the so called ``great moderation". The main finding is that the volatility of the coincident index seems to respond well to negative shock affecting the US economy such as wars, oil crises and terroristic attacks. In addition, while the level of the economy is mainly driven by the industrial production, the uncertainty in the real outcomes comes from other series, such as income and employment. Finally by permitting time varying volatility, the in and out of sample forecasts for GDP have confidence intervals that shrink and expand with the degree of uncertainty in the economy.
La disponibilità di informazioni tempestive, disaggregate ed esaustive sullo stato dell'economia è diventata negli ultimi decenni un presupposto fondamentale per l'analisi economica. Fra le serie macroeconomiche di riferimento un ruolo di primo piano è comunemente accordato al PIL, seppur i dati di Contabilità Nazionale vengano pubblicati con notevole ritardo (60gg per l'Europa). Tale esigenza empirica è sfociata nella produzione di molteplici tecniche econometriche, più o meno sofisticate, per ricavare una stima indiretta del PIL tramite l'utilizzo di indicatori ad alta frequenza. Tra le serie mensili di maggior utilizzo, assumono particolare rilievo le Indagini sulle imprese e sui consumatori pubblicate dalla Commissione Europea, alla fine di ogni mese con riferimento al mese in corso. A fronte di un indubbio vantaggio in termini di tempestività, rimane da stabilire in che misura utilizzare per analisi quantitative i dati delle Surveys, raccolti sotto forma di variabili qualitative. In risposta a tale problematica, si presentano nel primo Capitolo due nuovi metodi di quantificazione: la Spectral envelope e un modello logit cumulato basato su un filtro di Kalman non lineare. L'indicatore quantitativo così ottenuto per il livello di produzione in Italia mostra una elevata coerenza con il ciclo della produzione Industriale estratto esogenamente. Nel secondo Capitolo, si affronta la problematica del nowcasting del PIL dell'Area dell'Euro, tramite una procedura di disaggregazione temporale in un modello a frequenze miste, mensili e trimestrali. Il valore mensile del PIL viene ricavato tramite un set di indicatori ad alta frequenza seguendo un modello dynamico fattoriale alla Stock and Watson (1991) espresso in Spazio degli Stati seguendo l'esempio di Mariano e Murasawa (2003). Il modello viene stimato sia dal lato della domanda (componenti di spesa) che dell' offerta (settori di attività economica), combinando quindi i due risultati nella stima finale del PIL in ragione della loro precisione e tenendo conto del nuovo schema di concatenazione di Contabilità Nazionale. Contrariamente ad alcuni recenti risultati in letteratura, in questa applicazione le Surveys non impattano sulla stima del PIL. Tale risultato empirico viene analizzato in maggior dettaglio nel terzo Capitolo, ove il modello di base proposto nel Capitolo precedente viene esteso includendo due fattori comuni e correzione della componente autoregressiva per cicli a bassa frequenza. Tramite alcuni esperimenti su dati in tempo reale con valutazione delle revisioni, si giunge alla conclusione che l'uso delle Surveys migliora la precisione delle stime e delle previsioni, specialmente per un orizzonte temporale medio-lungo (3 passi in avanti rispetto all'ultima release del PIL). In conclusione si presenta la stima di un indicatore di media-varianza per il PIL degli Stati Uniti. Esiste un generale consenso in letteratura nel ritenere che negli ultimi 20 anni le maggiori serie macroeconomiche americane siano divenute meno volatili. Pertanto, il modello sviluppato per l'Europa viene esteso considerando una varianza variabile nel tempo secondo un modello Garch-type a due regimi, che intende replicare il fenomeno noto in letteratura come "great moderation". La flessibilità di tale parametrizzazione permette di ottenere intervalli di confidenza più accurati in fase di stima come di previsione. Inoltre, la volatilità stimata sembra esprimere mostrare come l'economia americana risponda a shocks negativi come guerre, crisi petrolifere e attacchi terroristici. In conclusione si osserva che, mentre il livello dell'economia risulta ancorato alla produzione Industriale, la sua incertezza dipende da altre serie, come il reddito e l'occupazione.
Essays on quantification and disaggregation of time series
Cecilia, Frale
2008
Abstract
The availability of timely, comprehensive and disaggregate indicators has become in the last years one of the most important issues for the economic analysis. Among macroeconomic series, GDP is considered as the one carrying highest informative content, nevertheless the official National Accounts release of its value is published with strong delay. In order to fill this gap many researchers have attempted to get an indirect measure of the state of the economy exploiting disaggregated series related to GDP and available at higher frequency. Among official data for Europe the timeliest data are the Business and Consumer Surveys of the European Commission, published at the end of each month and referring to the same month. Within respect to this advantage in term of timeliness it remains however to understand how to use in quantitative terms the information from Surveys, which is primarily qualitative. To address this issue we present in the first Chapter two new methods for the quantification of the qualitative information provided by firms in the Business Survey. The Spectral Envelope, as well as the Cumulative Logit model with a non linear Kalman filter is applied to get a quantified indicator of the level of production. The main result is that this indicator is high coherent with the cycle of the Industrial production, which is extracted independently. The issue of GDP nowcasting is carried out extensively in the second Chapter, where the focus is the disaggregation in time of the quarterly value added for the Euro area. By using a mixed frequency model that relies on quarterly and monthly series, the monthly indicator for GDP is obtained casting the traditional Stock and Watson (1991) model in state space form, as in Mariano and Murasawa (2003). The estimation is carried out from the output (sectors of activities) and the expenditure side (demand components) of GDP, which both contribute to the final computation on the basis of their precision. The chain linked nature of the National Account series is also considered. The results of this application show that, in contrast with some recent literature, Survey data do not matter for the estimation of GDP. More appropriate investigations on the role of Survey data for the estimation of macroeconomic variables is provided in the third Chapter, where the basic model presented in the previous Chapter is extended to allow for more than one common factor and low frequency cycles. After some analysis of real time data and revisions, the main conclusion is that Survey data in a richest model are indeed useful to produce more accurate estimates and forecast, especially 3 steps ahead of the current released of GDP and when hard data are not available yet. We conclude our analysis on GDP estimation considering a mean-variance coincident index for the US economy. There is general consensus on the fact that most of the macroeconomic series have become less volatile in the last 20 year in US. This issue is addressed by extending the framework developed for Europe in the previous Chapters with a two regimes Garch-type model for the fluctuations of the economy that mimic the dynamic of the so called ``great moderation". The main finding is that the volatility of the coincident index seems to respond well to negative shock affecting the US economy such as wars, oil crises and terroristic attacks. In addition, while the level of the economy is mainly driven by the industrial production, the uncertainty in the real outcomes comes from other series, such as income and employment. Finally by permitting time varying volatility, the in and out of sample forecasts for GDP have confidence intervals that shrink and expand with the degree of uncertainty in the economy.File | Dimensione | Formato | |
---|---|---|---|
CFrale_thesis.pdf
accesso aperto
Dimensione
1.31 MB
Formato
Adobe PDF
|
1.31 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/195098
URN:NBN:IT:UNIROMA2-195098