"I Just Ran Four Million Regressions" for Backcasting Turkish GDP Growth

"I Just Ran Four Million Regressions" for Backcasting Turkish GDP Growth

Title:

"I Just Ran Four Million Regressions" for Backcasting Turkish GDP Growth

Number:

15/33

Author(s):

Mahmut Günay

Language:

English

Date:

November 2015

Abstract:

In this paper we backcast Turkish GDP growth with bridge equations. In the backcasting models, we consider indicators from production, international trade, consumer and firm surveys, employment, inflation and real exchange rate. We use a systematic search process for finding the combination of variables in the bridge equations that make the least backcast error in the period under investigation. We find that using information from different blocks of data in a bridge equation improves backcasting performance. Our results points out to the importance of using timeliness advantage of soft indicators, such as PMI, effectively. Similar to other studies in the literature, average of backcasts of models makes less backcast error than individual models.

Keywords:

GDP Forecasting, Bridge Equations, Forecast Combination

JEL Codes:

C22; E37

"I Just Ran Four Million Regressions" for Backcasting Turkish GDP Growth
CENTRAL BANK OF THE REPUBLIC OF TURKEY Head Office Anafartalar Mah. İstiklal Cad. No:10 06050 Ulus Altındağ Ankara Phone : (+90 312) 507 50 00 | Fax : (+90 312) 507 56 40