Short‐term forecasts of economic activity: Are fortnightly factors useful?

• Author: Libero Monteforte,Valentina Raponi
• Date: 01 April 2019
• DOI: http://doi.org/10.1002/for.2565
• Published date: 01 April 2019

Received: 14 April 2018 Revised: 4 September 2018 Accepted: 21 November 2018

DOI: 10.1002/for.2565

RESEARCH ARTICLE

Short-term forecasts of economic activity: Are fortnightly

factors useful?

Libero Monteforte1,2 Valentina Raponi3,4

1Economic Outlook and Monetary Policy

Directorate, Banca d'Italia, Rome, Italy

2Macroeconomic analysis department,

Ufficio Parlamentare di Bilancio, Rome,

Italy

3Imperial College Business School,

Imperial Colege London, London, UK

4MEMOTEF, Sapienza, Università di

Roma, Rome, Italy

Correspondence

Valentina Raponi, Imperial College

Business School, South Kensington

Campus, Ayrton Road, Kensington,

London SW7 2AZ, UK.

Email: v.raponi13@imperial.ac.uk

Abstract

A short-term mixed-frequency model is proposed to estimate and forecast Ital-

ian economic activity fortnightly.We introduce a dynamic one-factor model with

three frequencies (quarterly, monthly, and fortnightly) by selecting indicators

that show significant coincident and leading properties and are representative

of both demand and supply. We conduct an out-of-sample forecasting exercise

and compare the prediction errors of our model with those of alternative models

that do not include fortnightly indicators. Wefind that high-frequency indicators

significantly improve the real-time forecasts of Italian gross domestic product

(GDP); this result suggests that models exploiting the information available

at different lags and frequencies provide forecasting gains beyond those based

on monthly variables alone. Moreover, the model provides a new fortnightly

indicator of GDP, consistent with the official quarterly series.

KEYWORDS

factor models, forecasting, mixed frequency data, Kalman filter, state-space models, temporal

disaggregation

1INTRODUCTION

Estimating the current state of the economy and fore-

casting future values of macroeconomic variables are

extremely important tasks. Institutions need to improve

the timeliness of policies, and market makers need to

anticipate asset price changes. The most important indica-

tor of economic activity, gross domestic product (GDP), in

Italy is released quarterly, with a delay of about 30 days.1

Therefore accurate and timely predictions of GDP are nec-

essary to gain an insight into the current and future state

of the economy.

The increasing search and availability of nonstructured

data sets, based on big data and experimental data, suggest

1Note that before May 2018 Italian GDP data availability was even less

timely, with a publication lag of 45 days.

that high-frequency series should contain additional infor-

mation about the business cycle and therefore should be

considered relevant for both macroeconomic nowcasting

and forecasting. The problem in using these data is that

they are typically available at different frequencies and

with a ragged edge structure. This requires the use of

models able to incorporate this heterogeneity in terms of

frequency, number of variables, and time durations. In

particular, taking advantage of indicators available in real

time requires an efficient tool in order to face two main

challenges: first, how to handle the mixing-frequency fea-

tures of the available data, matching, for example, daily

financial data with monthly variables and other quarterly

indicators; the second issue concerns how to extract use-

ful information—that is, how to identify the main com-

mon components from the cross-section of the available

indicators.

Journal of Forecasting. 2019;38:207–221. wileyonlinelibrary.com/journal/for © 2018 John Wiley & Sons, Ltd. 207