Federal Reserve Economic Data: A History

Introduction

Economists and monetary policymakers regularly highlight the importance of trustworthy and reliable data for decision-making (Clarida, 2018; Trichet, 2004; Williams, 2019). The discipline itself has taken a data-centric “empirical turn” (Hamermesh, 2013) and several data-related expected proficiencies of economics majors are succinctly identified in the scholarship of teaching and learning (Hansen, 2012). As researchers and as instructors, we work and we teach in the age of data. In what follows, we will describe the contributions of the Federal Reserve System to the collection, production, and distribution of economic data. First, we present a chronological account of related efforts, many of which were groundbreaking. Next, we describe the origins and evolution of the data information services provided by the FRED online database, its partnership with sources, and the profile of its users. Next, we review the data-related educational outreach efforts of the Federal Reserve Bank of St. Louis, presenting some preliminary evidence of their impact. A summary and several reflections on the future of economic data and education conclude the article.

Existing literature on government data often omits the contributions of the Federal Reserve. It also fails to give a sense of what data were available when. Although there are a number of histories of statistical analysis (e.g., Stigler, 1986) and the use of statistical data and mathematical analysis in the early years of the economics profession (Morgan, 1990; Schumpeter, 1954), there is little research on the state of economic indicators in the United States prior to the Great Depression. Duncan and Shelton’s well-known 1978 study of the “revolution” in U.S. government statistics focuses on the increasing professionalization and statistical sophistication beginning in the 1920s. Many historical studies of federal statistics and statistical work are focused on specific agencies, organizations, committees, or datasets (exemplified in the multiple book-length histories of gross domestic product).

Available histories of federal data production show that although statistics were becoming popular in the United States long before the founding of the Fed and its innovative data production (Mitchell, 1919; On the Hundredth Anniversary of the “Commercial & Financial Chronicle, 1939; Mason et al., 1990; Charles & Giraud, 2013), that data were patchy, delayed, unreliable, and/or poorly suited for economic analysis (Carter & Sutch, 1995; Davis, 2004; Kuznets, 1934; Rockoff, 2019; Romer, 1986; Smiley, 1983). Furthermore, the literature shows that economic analysis of statistical data was still in its infant stages in 1913 (Craver & Leijonhufvud, 1987; Klein, 1950, p. 123; Morgan, 1990, p. 8; Stigler, 1986), and the economics profession at that time did not view statistical data and mathematical analysis as primary tools of analysis (Cherrier, 2017). As a new institution, the Federal Reserve was therefore in a unique position to set new precedents in the provision of statistical information.

U.S. Government Statistical Data and the Federal Reserve System

At the time the Fed began to publish its first data (banking data in 1914 and manufacturing production data in 1919), essentially no standardized U.S. principal economic indicators were regularly and consistently made available by any entity (see Mitchell, 1919). As Duncan and Shelton (1978) assert, the newly created Fed’s data production “created a strong force for the improvement of business cycle statistics in general” (p. 8). The oldest U.S. headline indicator, the consumer price index, was available only as narrowly focused or regional indexes prior to 1919 (Bureau of Labor Statistics, 2014a). The Board’s monthly manufacturing production index was launched in January 1919, and the Bureau of Labor Statistics first regularly published nationwide cost of living reports—semiannually—in 1919 (Rockoff, 2019, p. 151). The BLS had begun work to create a cost of living index in the 1890s but publication of data was erratic, and the best method of computing price indexes was still under debate (Rockoff, 2019, p. 150). Wesley Clark Mitchell had published his book on business cycles in 1913, but the National Bureau of Economic Research (NBER) would not be founded until 1920 and would not publish its first business cycle dates for another decade (NBER). Interest in efficiency of federal statistical data collection and output throughout this period culminated in the 1922 Report of the Bureau of Efficiency and its recommendations for improvements in statistical work, but that report had little immediate impact (Duncan & Shelton, 1978, p. 11). Additional statistical publications launched, including in 1921 the Survey of Current Business, then a publication of the Department of Commerce (Department of Commerce, 1921). The Survey featured a wide variety of government and commercial sources for its data. Table 1 presents a brief chronology of some early data releases discussed in this article.

OPEN IN VIEWER

Table 1.

Early U.S. Government Statistical Data Production.

Date of first issue	Regular indicator or publication	Published by	Agency founded
1919	Manufacturing production index	Federal Reserve	1913
1921 a	Consumer price index	Bureau of Labor (later Bureau of Labor Statistics)	1884
1921	Survey of Current Business	Department of Commerce	1903
1929	Business cycle dates	NBER	1920
1940	National unemployment rate (Monthly Report on the Labor Force)	Bureau of Labor Statistics	1884
1935 b	National income (later GNP, later GDP)	Department of Commerce (later Bureau of Economic Analysis)	1903

Note. NBER = National Bureau of Economic Research (National Bureau of Economic Research, n.d.).

a

Initial report published in 1919, with data estimated back to 1913. Annual releases began in 1921; quarterly in 1935; and monthly in 1940 (Bureau of Labor Statistics, 1966, p. 2). ^b National income was reported annually as of 1935 and monthly as of 1938. This evolved into GNP in 1942 (Card, 2011, p. 554).

In contrast to its peers, the Federal Reserve began producing and compiling banking data from the moment business commenced in November of 1914, requesting from the newly opened Reserve Banks “statistical records which the Federal Reserve plans to conduct in its offices as a part of its system of information” (Federal Reserve Board, 1914a, 1914b). In the first year of operations, the Federal Reserve compiled reports of capital stock; condition reports of member banks (from c. February 1915); and data on the gold settlement fund (from March 1915). “The work of the Division of Reports and Statistics and that of the Division of Audit and Examination are so closely related that it is essential that close cooperation shall exist between the two” (Federal Reserve Board, 1914b). Beginning in May 1915, the Federal Reserve Board began to publish the Federal Reserve Bulletin, which would become the central statistical publication of the Fed and “the most important monthly publication for financial statistics in the Nation” (Duncan & Shelton, 1978, p. 8). The Bulletin, which was distributed free of charge to the twelve Federal Reserve Banks and member banks of the System, included the reports mentioned above, narrative reports on business conditions in the 12 Federal Reserve Districts, and other relevant information on the banking industry (Federal Reserve Board, 1915). The data in the Bulletin were designed to work in tandem with existing federal financial statistics, such as those from the Office of the Comptroller of the Currency and the annual Statistical Abstracts which became a product of the new Department of Commerce in 1913 (Duncan & Shelton, 1978, pp. 8–10).

Additional data were added to the Bulletin over time, while the 12 Federal Reserve Banks began to publish their own monthly reports of business and financial conditions for each district. ¹ Then, in June 1918, the Board published its first set of macroeconomic data. Those consisted of two pages showing indexed movement of wholesale prices between 1914 and 1918 (Federal Reserve Board, 1918a), relying heavily on the price data gathered by the Bureau of Labor Statistics (Federal Reserve Board, 1918b). This feature was the first of a series of monthly overviews of trends in various economic indicators, primarily focused on prices and commodity production.

Throughout the 1920s, the Federal Reserve expanded the statistical data it made available to the public, including precursors to key interest rate publications and retail sales (Board of Governors of the Federal Reserve System, 2019), while other statistical agencies within the federal government were reducing budget lines for statistical data and shrinking those programs, despite growing interest in statistical data and its potential use for macroeconomic analysis (Bureau of Labor Statistics, 2014a, 2014b). By the dawn of the 1930s, the Federal Reserve System, both the Board and the Banks, was one of the more consistent publishers of economic data and information in what was a fairly sparse field with no single, comprehensive, publicly available picture of the economy (Johnston & Podleski, 2020; Landefeld et al., 2008): The improved and standardized cost of living was relaunched in 1935 (Bureau of Labor Statistics), Kuznets’s single study of national income (which would give rise to GNP and later GDP) was published in 1934 (Kuznets, 1934), and the Census of Unemployment would not be published until 1937 (Card, 2011).

During World War II and in the postwar years, the Fed’s statistical data production grew. Data on consumer saving and retail credit began to be published as part of the Fed’s 1941 mandate from President Roosevelt to “curb” consumer credit (Roosevelt, 1941). In the early 1940s, the Fed also began to publish data on bond yields and prices and made available the nearly comprehensive volume Banking and Monetary Statistics, 1914–1941 for general publication. In the introduction to that book, the Board’s research director E.A. Goldenweiser wrote that “[These statistics] should also inspire confidence. They are an augury that credit policy can be based in the future, as in the past, on fact rather than on fancy” (Board of Governors of the Federal Reserve System, 1943). Fed researchers strove to continuously improve the public data it produced and bring it into line with the new statistical standards of the day (Jarmin, 2019; Jones, 1954).

In the 1950s—and with increasing speed in the 1960s—the Fed’s publication of data also grew along with the expansion of economic concepts; ² but unlike their counterparts in other federal statistical agencies, the creators of Federal Reserve statistical data did not regularly disseminate their data to the general public. Unlike the publications of the Bureau of Labor Statistics and, later, the Bureau of Economic Analysis, Fed data publications were not included in the Federal Depository Library Program, which makes available government publications to the public through municipal and academic libraries across the country (Stierholz, 2016b).

At the same time, what was considered a key economic indicator was changing. New kinds of research and data analysis were flourishing in some parts of the Federal Reserve System, particularly at the Board and at the Atlanta and St. Louis Reserve Banks (Meltzer, 2014 p. 81). The St. Louis Fed emerged in the early 1960s as the epicenter of Fed “monetarist” economics and made regular dissemination of money stock data a key tool in communicating their ideas (Federal Reserve Bank of St. Louis, 2019; Hafer & Wheelock, 2001, p. 8). Throughout the 1960s, under the leadership of Homer Jones, the research director of the Federal Reserve Bank of St. Louis, the St. Louis Fed began its own standalone data publication, U.S. Financial Data, in January 1966 (Federal Reserve Bank of St. Louis, 1991a), followed by additional data publications (Poole, 2006, p. 86). These data publications were mostly “rehash” of data produced elsewhere, but “rehash[ed]” by St. Louis staff for distribution, as Homer Jones characterized it in the November 18, 1967, Business Week article on their work.

The History of FRED

The development of the FRED database can be traced back to May 17, 1961, when the first issue of a newsletter entirely devoted to economic data was mailed (Jones, 1961). Totaling 21 pages, it consisted of a cover letter, 19 pages of tabulated monthly time series data for three different monetary aggregates (money supply, money supply plus time deposits, and money supply plus time deposits and short-term government securities), and a concluding page with three graphs of the 3-month moving average of the data (Stierholz, 2019). The nature of the published data has been construed as part of the monetarist debates of the early 1960s (Friedman, 1976; Hafer & Wheelock, 2001; Poole, 2006), yet the underlying public service provided by the free distribution of economic data has endured.

Interest in the content grew steadily and additional statistical releases were developed by the Research staff and added to the print newsletter. For a detailed timeline, see https://fraser-sdu.stlouisfed.org/timeline/stl-fed-data. Over time, technological advancement allowed for the distribution of data through new means. On April 18, 1991, the Federal Reserve Economic Data (FRED^®, https://fred.stlouisfed.org/) service was launched through a free electronic bulletin board. This information service allowed personal computers to connect to a St. Louis Fed computer via modem and download 30 different series of data (Federal Reserve Bank of St. Louis, 1991b). In 2 years, that number grew to 300 (Federal Reserve Bank of St. Louis, 2014). Demand for FRED was robust. By 1995, when a website was launched to download text files with data via the internet, 865 series were available. As shown in Figure 1, the volume of user traffic on the website has consistently outpaced the number of data series added to FRED.

OPEN IN VIEWER

Figure 1.

Number of data series in FRED and volume of web traffic over time.

Educational Outreach

Educational institutions acknowledged the research value of the FRED data service immediately (Federal Reserve Bank of St. Louis, 1991b). Its pedagogical use was developed soon after; and, over time, the scholarship of teaching and learning in economics embraced the importance of enhancing quantitative literacy as a complement to economic literacy. The Journal of Economic Education, the flagship journal in the field even solicited examples of using FRED in the classroom (McGoldrick, 2014). The call specifically mentioned the ease of access to data that FRED users enjoy (McGoldrick, 2014, p. 169) and has been answered, to-date, by Mendez-Carbajo (2015), Mendez-Carbajo and Asarta (2017), and Staveley-O’Carroll (2018).

Building on this growing interest among educators, between September 2017 and May 2019, a small group of economic education specialists and FRED data scientists facilitated 15 separate professional development workshops in different Canadian and U.S. cities. Varying in length between long sessions (2 hours) and half days (5 hours), the workshops introduced participants to a series of data visualization features available through the FRED portal. Those features were demonstrated in an educational context, where the workshop facilitators guided the participants through a series of data search, graph-building, and graph-reading exercises.

A total of 425 instructors (faculty and graduate students) participated in the workshops. Of those, 294 completed a 20-question survey, which represents a 69% response rate. Most of the survey respondents (84%) taught economics and the total of their annual student enrollment was 70,589 persons. Slightly less than one-half of the instructors (46%) acknowledged using FRED for teaching, while slightly more than one-half (54%) did not. The most frequent (90%) instructional use of FRED among the former group was as a lecturing resource. A fair amount of self-selection among participants in professional development events could be expected. Nevertheless, slightly more than one-half (56%) of the instructors surveyed at the FRED workshops considered themselves to be at-par with their colleagues when it came to teaching with data and few (22%) had published scholarship on teaching and learning.

Besides the questions capturing workshop participants’ demographic information, the survey contained a pre- and post-survey about instructors’ interest in and self-efficacy related to teaching with data. Specifically, in the pre-survey instructors used a Likert-type scale to state their degree of agreement with the statement: “Bringing data into my teaching is difficult.” In the post-survey, instructors used the same scale to state their degree of agreement with the statement: “Bringing data into my teaching is easier than I thought.” Figure 2 shows the median values of the participants’ pre- and post-workshop teaching self-efficacy organized in two groups: instructors currently teaching with FRED and instructors not currently teaching with FRED.

OPEN IN VIEWER

Figure 2.

Average self-efficacy related to teaching with data among workshop participants.

Prior to participating in the professional development workshop, instructors already using FRED in their teaching reported higher self-efficacy related to teaching with data than instructors not using FRED. At the conclusion of the professional development workshop, both groups of instructors reported increased self-efficacy related to teaching with data, with more pronounced gains among those instructors not currently using FRED in their teaching ⁴ . Finally, besides the change in perceived instructional self-efficacy among workshop participants, professional development opportunities of this nature can have measurable spillover benefits on instructors’ actual abilities. The work of Thornton and Vredeveld (1977) documents gains in instructors’ understanding of economics by being exposed to curricular materials they can use in their own classrooms. Thus, it can be argued that in the process of participating in a workshop about how to teach with FRED data, instructors became more data literate themselves.

Conclusions and Looking Ahead

The drive to collect economic data is at least as old as the Federal Reserve System itself. Trustworthy and timely quantitative and qualitative information is the basis for effective policymaking in pursuit of Congress’ mandate to the U.S. central bank. This need is not only pressing but also evolving. For example, Bernanke and Boivin (2000) show how large data sets can be incorporated into the study of monetary policy and improve forecast accuracy. At the same time, Smets (2018) outlines the challenges of translating new big digital data sets into concrete policy implications. Poole (2006) argues data distribution is a valuable service to all economic agents. While the extra cost of making the information used by policymakers available to the general public is modest, it promotes institutional transparency and accountability. Moreover, under a paradigm of rational expectations, informational asymmetry between policymakers and private market participants undermines the good operation of the economy. Only when businesses, households, and the Central Bank access the same information dataset can their expectations align and a stable equilibrium emerge.

Almost continually since its launch in 1991, FRED has increased both the volume of information it makes accessible to its users while also broadening its range of data information services. FRED’s time series data visualization options are frequently refined through enhancements to its graphs, and it has added a geographical information system to display cross-sectional data, an archive of data revisions, and a repository of Federal Reserve historical documents. Moreover, it has diversified the modes of access to its database, attracting both high-volume users through its application programming interface and smart phone and tablet users through its mobile applications.

The general direction for FRED’s future growth is broadly drawn on three paths (Federal Reserve Bank of St. Louis, 2019). First, in terms of driving its data collection and dissemination, FRED aims to add more international and socioeconomic data. Those data will help contextualize its comprehensive coverage of U.S. macroeconomic topics. Second, in terms of facilitating the discoverability of a constantly growing volume of data, FRED aims to improve the user search experience. That will include returning more targeted initial results from any given query and offering intuitive filters to refine them. Third, in terms of helping users make more of the available data, FRED aims to improve the data consumption experience. That will include options to add supplemental content relevant to data series or collaborate with data providers to ensure FRED presents those data in the most useful way possible.

The value of FRED as an asset in economic education cannot be overstated. Mendez-Carbajo (2020) argues for developing the range of data-literacy skills central to the discipline through the instructional use of FRED. The data services accessible through FRED allow students to master the professional-level practices of searching, accessing, and analyzing real-time economic information. These intellectual competencies can and should be developed across the curriculum, increasing their level of sophistication as the student progresses through the major. Thus, when students are introduced to robust data visualization and citation practices in lower division courses, they can develop habits of mind well suited to undertake accurate and replicable statistical analysis work in upper division courses. And they will be able to trust that FRED will continue to be a reliable source of information along the way.