Big Data and Official Statistics^†

Abstract

The infrastructure and methods for developed countries' economic statistics, largely established in the mid-20th century, rest almost entirely on survey and administrative data. The increasing difficulty of obtaining survey responses threatens the sustainability of this model. Meanwhile, users of economic data are demanding ever more timely and granular information. "Big data" originally created for other purposes offer the promise of new approaches to the compilation of economic data. Drawing primarily on the U.S. experience, the paper considers the challenges to incorporating big data into the ongoing production of official economic statistics and provides examples of progress towards that goal to date. Beyond their value for the routine production of a standard set of official statistics, new sources of data create opportunities to respond more nimbly to emerging needs for information. The concluding section of the paper argues that national statistical offices should expand their mission to seize these opportunities.

Open e-commerce 1.0, five years of crowdsourced U.S. Amazon purchase histories with user demographics

“…As an example, consider the Annual Retail Trade Survey, conducted by the U.S. Census Bureau. Between 2008 and 2018, response rates fell from 82% percent to 64% 41 . In the Technical Validation section of this paper we demonstrate how purchasing data from an e-commerce giant can provide comparable statistics to the Retail Trade Survey.…”

“…At the same time, traditional data collection through government surveys, which produce statistics and datasets for public use, is in decline. In particular, response rates to important surveys conducted by the U.S. Census Bureau and related agencies have fallen sharply in recent years 41 . Between 2013 and 2023 the response rate for the Current Population Survey (CPS), the source of U.S. statistics on employment, fell by 19%, and the response rate for the Consumer Expenditure Survey, which produces important data on consumer spending and is used to estimate inflation, fell by 15% 42 .…”

“…Between 2013 and 2023 the response rate for the Current Population Survey (CPS), the source of U.S. statistics on employment, fell by 19%, and the response rate for the Consumer Expenditure Survey, which produces important data on consumer spending and is used to estimate inflation, fell by 15% 42 . Economists have described how prices data collected through new strategies–web scraping, crowdsourcing, purchasing from data aggregators–can be incorporated into official economic indicators to compensate for declining response rates 41 .…”

The perils of pre-filling: Lessons from the UK’s Annual Survey of Hours and Earning microdata

“…In a time when users are demanding more timely and more disaggregated information, and while the budgets for many statistical agencies have been stagnant or declining, prefilling answers on a survey can offer a number of benefits [6]. Providing prefilled answers to questions which are fixed or just change irregularly can reduce the burden to the respondents, cut processing costs, while also potentially improving the quality of the data.…”