Data Units in FRED®

Introduction

The data accessible through FRED measures many different concepts—for example, the value of overall economic activity or the size of a country’s population. FRED presents the data produced by organizations and individuals “as is”—that is, exactly as provided by the source. This means that each data series is shown in the units reported by the source, whether it be millions of U.S. dollars or number of persons. However, FRED users can select units for the data that help tell the story behind the numbers. This article provides a description of the choices of data units, their common use, and their interpretation. It also describes the steps for creating custom data transformations.

Selecting Units

FRED users can change the units of each data series from the default reported by the source.¹ By visiting the FRED website and navigating to any data series, users can click on “Edit Graph”, select “Edit Line 1”, and click on the “Units” drop-down menu to see the different options available. The FRED website describes the formulas used to calculate each of those units in its Help webpages.² As the following examples illustrate, the storytelling purpose of a particular data visualization dictates the choice of units.

Change is calculated as the value of the data in the current period minus their value in the previous period. When the U.S. Bureau of Labor Statistics (BLS) releases its monthly figures on the thousands of persons estimated to be employees on payrolls, news media outlets frequently highlight their change in value to describe their evolution. In the Figure 1 FRED graph, the source-reported data units of “Thousands of Persons, Seasonally Adjusted” are displayed as “Change, Thousands of Persons”.³

Note that “Seasonally Adjusted” and its counterpart “Not Seasonally Adjusted” are types of data units calculated by the data source. Seasonal adjustment of data involves accounting for the predictable changes in their value brought about by the calendar year. For example, employment figures regularly peak during the last quarter of the year due to high retail sales activity.⁴ The procedure to affect that seasonal adjustment is beyond the scope of this article.

Change from Year Ago is calculated as the value of the data in the current period minus their value from a past period, which could be 1 year, 4 quarters, 12 months, 52 weeks, or 365 days ago, depending on the frequency of the data. Because the BLS does not currently adjust its county-level employment figures for seasonal effects, their change from month to month will reflect seasonal patterns and can easily yield a distorted interpretation of the data. A more robust description of monthly changes in county-level employment can be achieved by changing the units to year-over-year values. In the Figure 2 FRED graph, the source-reported data units of “Persons, Not Seasonally Adjusted“ are displayed as “Change from Year Ago, Persons”.⁵

Notice that when calculating the change in the value of a series, from either the preceding period or a year ago, the value is not scaled by the magnitude of the series. That scaling, frequently presented in percent form, provides a helpful reference point for purposes of storytelling. We will discuss that unit choice next.

Percent Change is calculated as the growth rate of the data between the current period and the previous period. It is expressed in hundredths and frequently followed by the symbol “%” Because it is a rate, this unit type allows users to compare changes in data of different magnitudes. For example, we use percent changes to compare the ups and downs in the San Diego County, CA, population (which was 3.3 million people in 2020) with changes in the U.S. population (which was 330.1 million people that same year). In the Figure 3 FRED graph, the source-reported data units of “Thousands of Persons, Not Seasonally Adjusted” are displayed as “Percent Change”.⁶

Percent Change from Year Ago is calculated as the growth rate of the data between the current period and their value from a past period, which could be 1 year, 4 quarters, 12 months, 52 weeks, or 365 days ago, depending on the frequency of the data. When the BLS releases its monthly consumer price index (CPI) figures , news media outlets report their percent change in value from a year ago to describe the evolution of the inflation rate. In the Figure 4 FRED graph, the source-reported data units of “Index 1982-1984=100, Seasonally Adjusted” are displayed as “Percent Change from Year Ago.” This type of unit also helps to describe the growth rate of data that are not seasonally adjusted: Because it represents their year-over-year percent change, it won’t reflect seasonal patterns likely to distort the interpretation of the data.⁷

Compounded Annual Rate of Change is calculated as an exponential growth rate, whether the interval period is 1 year, 4 quarters, 12 months, 52 weeks, or 365 days, depending on the frequency of the data. A compounded annual rate of change would make a figure grow from its beginning value to its ending value in a finite number of steps. This type of unit helps to describe data experiencing regular but unforeseeable fluctuations in value, and it facilitates comparison with other data reported with different frequencies.

For example, the U.S. Bureau of Economic Analysis (BEA) reports the value of real gross domestic product (GDP) four times, or quarterly, every year. It adjusts the value of real GDP to account for seasonal variations, but because some types of economic activity either peak or only take place during specific times of the year (e.g., purchases of school supplies in August or eggnog in December), the BEA also presents an annualized figure—that is, the value of real GDP that would be reached a year from now if overall economic activity were to keep its current pace for three more quarters.

However, describing the year-over-year growth rate of annualized real GDP figures by calculating a simple percent growth rate between quarters is misleading. Instead, a compounded annual rate of change accounts for the cumulative change in economic activity over time, allowing us to compare real GDP growth figures throughout the year. In the Figure 5 FRED graph, the source-reported real GDP data units of “Billions of Chained 2017 Dollars, Seasonally Adjusted Annual Rate” (in green) are presented as “Compounded Annual Rate of Change”. Next to it (in purple) is the source-reported “Percent Change from Preceding Period, Seasonally Adjusted Annual Rate” of real GDP. Both are effectively identical.

Continuously Compounded Rate of Change and Continuously Compounded Annual Rate of Change unit types are calculated as growth rates of an amount that compounds at a continuous rate, rather than at a step-by-step rate. The interest earned on a savings account balance is calculated using a compound formula, which also is the basis of the present-value calculation. Although FRED allows users to present data in those units, applications of their use are beyond the scope of this article.

Index (Scale value to 100 for chosen date) is calculated by using the value of the data on a chosen date (also known as the base date) to rescale all data points in a series. On the base date, the index has a value of 100. This unit type allows users to compare the rates of change in data series of different magnitudes. For example, we use a custom index to compare the changes in two headline figures of consumer price inflation: the CPI and the personal consumption expenditures price index. Because the BLS uses 1982-1984 as the base date for the former and the BEA uses 2012 as the base date for the latter, it is easier to compare the change in their values by choosing a common base date. In the Figure 6 FRED graph, the source-reported data units for both data series are displayed as “Index” with the same base date of February 2020.⁸

Customizing Units

FRED users can customize the units of data by applying formulas. This FRED feature allows users to transform the data and create a new visualization.⁹ The FRED website describes the steps of that process in its Help webpages.¹⁰

For example, to change the order of magnitude of a data series, FRED users can create a custom data transformation by applying a formula to a single series in a line. To transform billions of U.S. dollars into millions, click on “Edit Graph”, select “Edit Line”, type the formula “a×10000”, and click on “Apply”.

In the Figure 7 FRED graph, the World Bank reports China’s gross domestic product (GDP) in U.S. dollars, and the BEA reports U.S. GDP in billions of the same currency. You can easily see both series in the same graph by multiplying the data for the United States by 1,000,000,000. You can achieve the same end by dividing the data for China by 1,000,000,000.¹¹

Also, to change the units provided by the source of the data to per person or to inflation-adjusted units, FRED users can create a custom data transformation by applying a formula to two series in a line. Those unit customizations require adding a new series (series b) to combine with an existing series (series a) and then applying a formula such as “adivided by/b”.

In the Figure 8 FRED graph, data on the real value of GDP reported by the BEA (series a) is divided by the population data reported by the U.S Census Bureau (series b). This is called per person, or per capita, GDP. Because the Census Bureau reports population figures in thousands of persons and the BEA reports real GDP in billions of U.S. dollars, we use the formula “adivided by/b×1000000”.¹²

Similarly, in the Figure 9 FRED graph, data on the nominal value of the federal minimum hourly wage for nonfarm workers reported by the U.S. Department of Labor (in blue) is plotted along its real value (in red). To calculate the inflation-adjusted value of the minimum wage, its nominal value (series a) is divided by the CPI for all urban consumers (all items in a U.S. city average), reported by the U.S Census Bureau (series b). Because the CPI is an index with a value of 100 on the base date, we use the formula “(adivided by/b)×100”¹³

Summary

Different data units can help tell the story behind the numbers in a data series. FRED offers a range of data units to choose from and outlines the steps for creating customized data transformations.