Emerging practices and perspectives on Big Data analysis in economics: Bigger and better or more of the same?

Abstract

Although the terminology of Big Data has so far gained little traction in economics, the availability of unprecedentedly rich datasets and the need for new approaches – both epistemological and computational – to deal with them is an emerging issue for the discipline. Using interviews conducted with a cross-section of economists, this paper examines perspectives on Big Data across the discipline, the new types of data being used by researchers on economic issues, and the range of responses to this opportunity amongst economists. First, we outline the areas in which it is being used, including the prediction and ‘nowcasting’ of economic trends; mapping and predicting influence in the context of marketing; and acting as a cheaper or more accurate substitute for existing types of data such as censuses or labour market data. We then analyse the broader current and potential contributions of Big Data to economics, such as the ways in which econometric methodology is being used to shed light on questions beyond economics, how Big Data is improving or changing economic models, and the kinds of collaborations arising around Big Data between economists and other disciplines.

High-dimensional data in economics and their (robust) analysis

“…The analysis of such data sets may play an irreplacable role in the commercial sphere as well as in the contribution to the development of economic theory (Taylor et al, 2014). Important tasks of the analysis of high-dimensional data with a large number of variables and a small number of observations include linear regression, classification analysis, clustering or time series analysis.…”

“…Schmarzo (2013)), they cannot be analysed by standard statistical or machine learning methods because of their specific size. The importance of Big Data in economics was discussed by Taylor et al (2014). Their analysis is a very complicated data-driven process, because the data of various types may come from a variety of very messy (unstructured) sources, including instrumental variables, high-frequency time series, free text, granular spatial or geographical data or data from social networks.…”

El uso de datos masivos y sus técnicas analíticas para el diseño e implementación de políticas públicas en Latinoamérica y el Caribe

“…Additionally, big data is exhaustive in its scope, as a dataset can contain all the observations of a particular sample, which allows analysis of the entire set [18]. Moreover, most big data is available in real time or as soon as it is generated, which makes it possible to predict the immediate future, or perform "nowcasting" [26]. In terms of depth, big data can provide the "highest resolution" picture, as it captures extensive details.…”

NETIMIS: Dynamic Simulation of Health Economics Outcomes Using Big Data

“…This data is sufficiently large and complex enough to be part of the new frontier of "big data", the subject of much recent literature [5][6][7][8], including in medicine and health [9][10][11]. "Big data" can be seen as data that is available at such a scale that traditional research and analysis methods fail.…”