Are the revised Forecasts for Romania more accurate than the Preliminary ones during the Crisis?

• Author: Simionescu, M.
• Position: Institute for Economic Forecasting of the Romanian Academy, Bucharest
• Pages: 243-248

Bulletin of the Transilvania University of Braşov

Series V: Economic Sciences • Vol. 7 (56) No. 2 - 2014

ARE THE REVISED FORECASTS FOR

ROMANIA MORE ACCURATE THAN

THE PRELIMINARY ONES DURING

THE CRISIS?

Mihaela SIMIONESCU1

Abstract: The research question of this study is: are the revisions of an

anonymous Romanian forecaster more accurate than the initial predictions

for the current year? Therefore, in order to provide a scientific response to

this question, several macroeconomic variables were considered for the

crisis period (2009-2013): real GDP growth, inflation rate, unemployment

rate and exchange rate. The forecasts are provided for the current year in

two versions: spring variant and autumn one. For 2009, the real GDP

growth and unemployment rate revisions brought higher errors. Indeed, the

crisis start was not precisely anticipated by the expert. For unemployment

rate in each year belonging to the period 2009-2012 the revisions showed

higher errors. According to predictions accuracy indicators, the forecaster

was one which best predicted the exchange rate during 2009-2013. For this

variable, all the accuracy measures have the lowest values. All the

predictions for all the variables outperformed the naïve forecasts. According

to the Harvey-Leybourne-Newbold test, there are not significant differences

in accuracy between the initial and the revised forecasts of the real GDP rate

and the inflation rate between 2009-2013.

Key words: forecasts, accuracy, error, revisions.

1 Institute for Economic Forecasting of the Romanian Academy, Bucharest.

1. Introduction

The main objective of this research is t o

check if the revisions of the expert

forecasts improve the degree of accuracy.

Therefore, the traditional forecasts

accuracy measures and the modified

Diebold-Mariano test for small sample

were employed, taking into account that

only the predictions during the economic

crisis were analyzed (horizon: 2009-2013).

Recent studies have as objective the

comparisons between forecasts accuracy [1].

Gorr and Schneider employed the receiver

operating characteristic curve for predictions

at microeconomic level. The accuracy

measure was the partial surface under the

ROC graph. The authors recommended the

use of complex uni-variate methods for

operations-level predictions with large or

acceptable changes [5].

Other studies compared the inflation

predictions of professional forecast ers to

those made by the academic environment.

For assessing the forecasts accuracy, the

author employed random walk, ARIMA