Please use this identifier to cite or link to this item:
Etter, Richard
Graff, Michael
Year of Publication: 
Series/Report no.: 
KOF Studien 22
In 2009, KOF was mandated to analyse the business tendency surveys (BTS) conducted on a monthly basis by the Banco Central de Reserva del Perú (BCRP) and to assess the potential to develop coincident and/or leading indicators for core economic variables of the Peruvian economy. The analysed BCRP BTS data start in 2002 and end in 2010. They cover manufacturing, construction, commerce, services and total economic activity (GDP). The time span of eight years and the sectoral coverage allowed us to extend our search to comparably sophisticated indicators. The selection process of indicators comprised semantic analyses, cross-correlation analyses as well as turning point analyses (referring to a quality index suggested by the authors). As with other BTS data, the BCRP data revealed a trade-off between stability at the margin and the signal-to-noise ratio. Low pass filtering of the indicators increases the signal-to-noise ratio stability, but makes end points, i.e. the most up-to-date observation of the series, prone to filter-induced re-visions. Also, we faced the usual choice whether to refer to a smaller bundle of variables (usually on firm-specific items) or to a larger variable set, where over-fitting is less likely. Altogether, given the reference series considered, we identified a number of leading indicators for the GDP growth rate, but hardly any coincident BTS data. We hence were able to suggest a composite leading indicator for the dynamics of GDP. This indicator is given by the fist principal component of our selection of leading indicators. In sample, it shows a lead of five months before the reference series, which amounts to about six months before the first official data release dates.
Persistent Identifier of the first edition: 
Document Type: 
Research Report

Files in This Item:
347.31 kB

Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.