A framework for the evaluation and use of alternative data in the Consumer Expenditure Surveys

• Published in: Monthly Labor Review pp. 1 - 11
• Main Authors: Erhard, Laura, McBride, Brett, Safir, Adam
• Format: Journal Article Trade Publication Article
• Language: English
• Published: Washington Superintendent of Documents 01.02.2021
• Subjects: Computational linguistics; Consumer Economics; Consumer research; Consumer spending; Cost Indexes; Costs; Data collection; Diaries; Economic Research; Employment Level; Error of Measurement; Estimates; Evidence Based Practice; Expenditures; Government (Administrative Body); Households; Information technology; Investigations; Language Processing; Marketing research; Measurement Objectives; Natural language interfaces; Natural language processing; Polls & surveys; Private Sector; Program Administration; Strategic Planning; Strategic planning (Business); Surveys; United States > US
• ISSN: 0098-1818; 1937-4658

[...]data collection costs have been increasing because of an erosion over time of respondents' willingness to participate in the CE and the additional time and effort required to contact potential respondents and secure their cooperation. [...]diminishing data collection resources created by increasing costs without commensurate budget increases result in fewer survey participants and less data on expenditures collected in the survey, which negatively affects the quality of the CE data. For ease of discussion, we grouped alternative data into the following categories on the basis of the data source: (1) administrative data or administrative records data, which the Office of Management and Budget describes as "data collected by government entities for program administration, regulatory, or law enforcement purpose"5; (2) consolidated data (e.g., data from credit card companies, data aggregators, or other private sector companies); and (3) operational data that are used to conduct routine agency activities but often are not available for research or statistical uses (e.g., the Statistics of Income program of the Internal Revenue Service transforms tax data into derived records from tax returns that are thus not subject to usual destruction requirements).6 Alternative data also can be organized by the forms they take, ranging from structured data (e.g., most of the federal administrative data produced) to semistructured data, such as those downloaded from the internet, and finally to unstructured data (e.g., open response text data requiring some type of language processing). The CE program must ensure that data from each alternative source meet the following criteria: (1) they are consistent with the CE program's core measurement objectives and are representative of the target population; (2) they meet BLS requirements for data continuity-a sudden loss of an alternative data source cannot cause a disruption in production schedules, and the data elements and structure of alternative datasets cannot cause a sudden and urgent reworking of BLS information technology infrastructures; and (3) they uphold the agency's ability to be transparent.8 Of note, this article focuses exclusively on alternative data sources.