Royal Economic Society Annual Conference 2019, University of Warwick:
Economic Measurement with Big Data
The opportunities to improve economic measurement by leveraging big data are manifold. Big administrative datasets may partially substitute for traditional surveys, improving both the timeliness and granularity of economic statistics. Business and consumer transactions data yield new sources for measuring prices, consumption and economic networks. Satellite images might improve a diverse range of economic statistics, including measures of economic activity and natural capital. Information posted on-line by businesses can be used to develop flexible economic classification systems and internet search activity might inform conjunctural statistics. Yet, the incorporation of big data into the mainstream production of economic statistics remains in its infancy. This session presents three novel applications of big data and data science techniques for the purposes of economic measurement, and discusses the challenges these present.
Session Chair: Chris Giles (Financial Times)
An open and data-driven taxonomy of skills extracted from online job adverts By Jyldyz Djumalieva and Cath Sleeman (Nesta, ESCoE)
Presented by: Jyldyz Djumalieva
In this work we offer an open and data-driven skills taxonomy, which is independent of ESCO and O*NET, two popular available taxonomies that are expert-derived. Since the taxonomy is created in an algorithmic way without expert elicitation, it can be quickly updated to reflect changes in labour demand and provide timely insights to support labour market decision-making. Our proposed taxonomy also captures links between skills, aggregated job titles, and the salaries mentioned in the millions of UK job adverts used in this analysis. To generate the taxonomy, we employ machine learning methods, such as word embeddings, network community detection algorithms and consensus clustering. We model skills as a graph with individual skills as vertices and their co-occurrences in job adverts as edges. The strength of the relationships between the skills is measured using both the frequency of actual co-occurrences of skills in the same advert as well as their shared context, based on a trained word embeddings model. Once skills are represented as a network, we hierarchically group them into clusters. To ensure the stability of the resulting clusters, we introduce bootstrapping and consensus clustering stages into the methodology. While we share initial results and describe the skill clusters, the main purpose of this paper is to outline the methodology for building the taxonomy. (Slides)
Discussant: Sandra McNally (Professor of Economics, University of Sussex)
Large-scale real world financial transaction microdata for national and local economic indicators By Sebnem Oguz and Louisa Nolan (UK Office for National Statistics)
Presented by: Louisa Nolan
Financial transaction data has long been a holy grail for economic statistics. It has the potential to significantly improve our understanding of the economy at a granular level. This presentation will describe the ground-breaking collaborative project carried out by the Office for National Statistics and a major financial institution. The project explores how we can use payments and financing data to understand the night-time economy, the relationship between big-ticket household spending and economic growth, and business supply chains. We will present our progress on these three strands, discuss the challenges of using big data for economic statistics, and show new economic insights derived from this rich data source.
Discussant: Garry Young (Director of Macroeconomic Modelling and Forecasting, NIESR)
Making text count for macroeconomics: what can UK daily newspapers tell us about the future of the economy? By Arthur Turrell (Bank of England), Eleni Kalamara (King’s College London), Chris Redl (Bank of England), George Kapetanios (King’s College London, ESCoE) and Sujit Kapadia (European Central Bank)
Presented by: Eleni Kalamara
Recently, economists have sought to explain the psychological importance of narratives as drivers of sentiment and uncertainty. Those are, in turn, thought to drive agents’ decisions. Newspapers are likely to be an important channel for the propagation of economic narratives. We use the text of hundreds of thousands of articles from popular UK daily newspapers to find out whether news is just noise, or whether it contains useful signals about the future direction of macroeconomic and financial variables. Newspaper text is a form of ‘naturally occurring’, high frequency, big data and we show a range of methods to transform text into time series which can be used in economic analysis. We run a comprehensive horse race of the different text analysis methods and determine which could best serve as indicators for policymakers, and as inputs into forecasts. We also ask which newspapers contain more predictive power about the future direction of the economy.
Discussant: Grant Fitzner (Chief Economist, UK Office for National Statistics)