Please use this identifier to cite or link to this item: http://hdl.handle.net/1946/10534
In this paper, a large factor model is utilized with the aim of isolating the medium-to-long-run component (MLRC) from a large data set which could potentially describe the underlying data generating process, in this case gross domestic product (GDP). The central idea is to extract a few common factors from all the data set with as much variance as possible. To do so, the use of a particular kind of principal components analysis is employed, which is specifically designed to extract from the data set the common information of the data. More precisely, the linear combinations of the data. Removal of the short-run dynamics from a stationary time series to isolate MLRC can be done with a simple bandpass filter.
Quarterly GDP is not timely, it suffers from heavy short-run component and is under constant revision. The main objective of this paper is to construct a timely indicator which could better describe the current economic state than GDP does. For this to be reliable, the data set is made contemporaneous and synchronized in a way that all time series in the data set are describing the same process at the same time.
Further, the data set is on a monthly basis which makes it even more timely than the quarterly GDP. Therefore, the Icelandic Coincident Indicator (ICI) is on a monthly basis which is an estimate of the MLRC of GDP, in real time.
The main findings are that ICI captures the underlying data generating process of GDP rather nicely, it is smooth with less volatility but still manages to describe the average growth of GDP.
Keywords/lykilorð: Fjármálahagfræði, Business Cycle, Factor módel, Spektral greining, spálíkan, DataMarket.