Page 295 - Special Topic Session (STS) - Volume 3
P. 295
STS544 Paolo F. et al.
Nowcasting and the production of economic activity indicators in real time
have been the focus of a growing literature. An early work related to the
tracking of economic conditions in real time by creating new high-frequency
indicator is Aruoba, Diebold, and Scotti (2009). The nowcasting literature is
interested in estimating an existing economic indicator (usually quarterly GDP
growth) in real-time. Few examples drawn from the nowcasting literature are
Giannone, Reichlin, and Small (2008) and Evans (2005) among many others.
In this study, we combine firm-level datasets and machine learning
techniques, as well as traditional statistical models which can deal with large
datasets, to provide faster estimates of Finnish real economic activity, both at
the quarterly and monthly frequencies. The monthly series we target is the
Trend Indicator of Output (TIO) , published by Statistics Finland, while the
1
quarterly series is GDP. For both series we compute nowcasts of the year-on-
year growth rate. In addition, we examine the predictive power of a novel
dataset based on traffic volumes' measurements. The use of novel data
sources, such as firm-level turnovers data and traffic measurements, in
combination with the use of a wide array of machine learning techniques
provides the main contribution of our study to the nowcasting literature.
Our approach of combining predictions obtained by using a large set of
machine learning algorithms, based on firm-level data, is able to provide
accurate estimates of monthly economic activity growth, with revision errors
that are in line with the ones of Statistics Finland, while shortening the
publication lags by 30 days. The resulting early estimates of the monthly
indicator are used to compute three nowcasts of GDP year-on-year growth.
The first two nowcasts provide good accurancy, even though there are some
notable revision errors. However, the estimates produced after the end of the
quarter are very accurate, while providing a 45 days reduction in the
publication lag. We conduct a similar analysis using truck traffic volumes'
measurements, and find satisfactory results, albeit inferior to the ones
obtained from firm-level data.
The remainder of this paper is divided as follows: in Section 2 we discuss
some of the large set of models adopted in the analysis, in Section 3 we
describe our target indicators and data sources. In Section 4 we report the
results and Section 5 concludes.
2. Methodological Aspects
Given that the main contribution of this study is the use of novel data
sources, we keep the description of the models adopted brief. This section
does not cover comprehensively the techniques we use for two reasons: firstly,
the sheer number of statistical models and machine learning techniques
1 A description of this indicator is available at http://www.stat.fi/til/ktkk/index_en.html
284 | I S I W S C 2 0 1 9