Media‐Based Sentiment Indices as an Alternative Measure of Consumer Confidence

• Date: 01 December 2020
• DOI: http://doi.org/10.1111/saje.12261
• Author: Johann F. Kirsten,Hanjo Odendaal,Monique Reid
• Published date: 01 December 2020

South African Journal of Economics Vol. 88:4 December 2020

doi: 10.1111/saje.12261

409

© 2020 Economic Society of South Africa

MEDIA-BASED SENTIMENT INDICES AS AN ALTERNATIVE

MEASURE OF CONSUMER CONFIDENCE

HANJO ODENDAAL*,† , MONIQUE REID† AND JOHANN F. KIRSTEN†,‡

Abstract

In this paper, we consider the feasibility of constructing online sentiment indices, using large

amounts of media data, as an alternative to the conventional survey method used to create the

consumer confidence index in South Africa. A clustering framework is adopted to provide an

indication of possible candidate sentiment indices constructed from a combination of different

text sources and dictionaries that best mimic the traditional survey-based consumer confidence

index from the South African Bureau for Economic Research (BER). The results conclude that it is

possible to create an index using sentiment analysis using online editorial data that does resemble

the BER’s consumer confidence index. The different media-based sentiment indices (MSI) show

a significant level of correlation and co-movement with the BER’s CCI. Impulse responses and

cross-correlation functions indicate that the MSI could potentially lead the survey-based method

up to two quarters. Furthermore, Granger-causality tests show that the media-based indices are

good predictors of future consumer confidence index values. The results provide motivation for

further study on the use of sentiment-based techniques and online media data sources to track

consumer confidence within an emerging market such as South Africa.

JEL Classification: B41, C52, C55, C83

Keywords: Big data, sentiment analysis, consumer confidence

1. INTRODUCTION

As the analogue era slowly fades into the “video-cassette” or “floppy disk” of yesteryear,

the new digital age is generating information at an ever increasing rate. Data are being

generated at higher volumes and appearing in a host of different formats. Much of this

information is a by-product of economic activity, rather than from physical sampling or

surveying. The private sector is developing innovative ways to generate and collect exten-

sive datasets which can be converted into new revenue streams, without too much direct

production cost – a “collect it if you can” mentality.

Economists are also embracing the data revolution, complimenting these develop-

ments by rethinking how this kind of data could be employed within everyday economic

analysis. New types of data are being generated faster and with far greater scope and

coverage. One of the influential early examples of how online data collection on a large

scale was used within economics, to complement more traditional methods, is the Billion

Prices Project (BPP) of Cavallo (2013). Under the BBP, prices from hundreds of online

* Corresponding author: Department of Economics, Stellenbosch University, South Africa.

E-mail: Hanjo.oden@gmail.com

† Department of Economics, Stellenbosch University

‡ Bureau for Economic Research (BER), Stellenbosch University

South African Journal

of Economics

410 South African Journal of Economics Vol. 88:4 December 2020

© 2020 Economic Society of South Africa

stores, spanning over 50 countries, are collected daily. In this way, it offers an alternative

to traditional price indices with the advantage of being a real-time measure that is also

available at a far higher frequency.

Although this approach offers various advantages relating to real-time availability

and granularity, there are some disadvantages. The first concerns the issue of weighting.

Datasets on products collated through online collection contain no information on quan-

tities being sold. To overcome this, official consumer expenditure surveys or other sources

have to be used for any expenditure-weighted application. Second, despite collecting vast

amounts of data from various retailers, the price of some goods and especially services are

not listed online. This biases the subsequent index towards price movements in the goods

market.

In this paper, we construct and assess text-based indices that function as a proxy for

economic sentiment found in online news media. In using news media data in the anal-

ysis, the paper aims to overcome standard pitfalls such as personal privacy concerns,

volume of text to analyse, as well as the difficulty of obtaining reliable historical data,

associated with social media data. We aim to contribute to the understanding of the

feasibility of using these indices in an emerging market economy such as South Africa.

The paper is the first of its kind within an emerging market context. In emerging mar-

kets, especially South Africa with its diverse socio-economic landscape, the question of

whether news can wholly act as an alternative consumer confidence indicator is an in-

teresting one. For countries with a diverse socio-economic landscape, it could be argued

that the opinion in the news is not a proxy for the attitude of the population as a whole.

The survey-based business and consumer confidence indices developed by the Bureau of

Economic Research of South Africa (BER) are the most widely quoted in South Africa.

Their indices are based on the famous confidence index of the University of Michigan

(UM). Research has confirmed that the confidence index developed by UM has helped

to forecast macroeconomic outcomes after having controlled for a host of economic fac-

tors, such as disposable income and past personal consumption expenditure (Bram and

Ludvigson, 1997; Souleles, 2004; Curtin, 2007).

A second contribution of the paper is to suggest a framework by which researchers

and practitioners can develop a monthly index that could act as a potential alternative

to the traditional survey-based consumer confidence index using multiple online media

channels and dictionaries. Sentiment analysis forms part of a larger field called compu-

tational linguistics and although consumer confidence and sentiment can be thought of

as interchangeable concepts, in this paper sentiment refers to the constructed sentiment

score or polarity as derived by computational analysis. These indices can be incorporated

as complementary or alternative measures within national statistics as an indicator for

consumer sentiment. By incorporating daily news and editorial content, the index aims

to capture market information from a plethora of different channels, embodying a shared

view of economic agents (as represented by the authors of the media articles), that extend

beyond the opinions of select professionals.

This unstructured information set offers higher dimensionality and higher frequency

than survey-based methods. The data and approach allow for the analysis of economic

fluctuations through a bottom up modelling approach. A time series clustering technique

is used to identify which of the subset of indices created (each relying on different combi-

nations of dictionaries and online news sources), best reflects the current BER consumer