The Current Employment Statistics (CES) program currently publishes diffusion indexes to measure how widely national employment changes are spread across industries over 1-, 3-, 6-, and 12-month time spans. Diffusion indexes help us understand whether a change in employment may be caused by smaller employment changes in many industries or by large changes in a few industries.1 We calculate an overall index from 258 employment series (primarily 4-digit NAICS industries2) covering all nonfarm payroll employment in the private sector.
To derive the index, each industry is assigned a value of 0, 50, or 100, depending on whether its employment showed a decrease, no change, or an increase, respectively, over the time span. We then calculate the average (mean) of these values, and this percent is the diffusion index number. A diffusion index of 50 would show that the same number of component industries had increasing employment and decreasing employment, while an index higher than 50 would suggest more industries were increasing employment than decreasing over the index time span.3
BLS has developed experimental diffusion indexes of total nonfarm employment over 1-, 3-, 6-, and 12-month time spans, first at the statewide level for all 50 states and the District of Columbia, then as an aggregate of 387 Metropolitan Statistical Areas (MSAs).4 Diffusion indexes may be useful in understanding state and local variation of employment changes. For example, an unusually low reading for the index during a period of job gains would indicate that they are concentrated in a few areas, perhaps signaling a geographic shift in the economy. Conversely, if the index is reading unusually high during a period of job loss, the losses may be concentrated in certain states or areas, as might happen following a natural disaster or other localized event.
CES time series for states and areas are a hybrid of historical data benchmarked using a complete census count of employment, primarily from the Quarterly Census of Employment and Wages (QCEW), and sample-based data estimated from the CES survey. The two portions of the series tend to show different seasonal patterns. For that reason, we derive seasonal factors for each portion independently, a process known as "two-step seasonal adjustment." (This process is not used for national CES series, because of a difference in benchmarking techniques.) When a 12-month span of a time series contains both sample-based and population data, the over-the-year changes in the seasonally adjusted and not seasonally adjusted series may differ by a great deal; the 12-month change in the not seasonally adjusted series may still contain significant residual seasonality. For this reason, 12-month diffusion indexes for CES State and Area data are constructed using the seasonally adjusted components. This differs from the procedure for national data, which uses not seasonally adjusted data in the 12-month diffusion indexes.
The availability of seasonally adjusted data is an issue when constructing an index of metropolitan areas. At present, BLS publishes only total nonfarm employment on a seasonally adjusted basis for MSAs. However, periodic redefinitions of areas and the nature of the two-step process mean that, for some periods of time, no seasonally adjusted data may be available for many MSAs. Following the revised delineation of metropolitan statistical areas published in March 2015, 91 metropolitan areas were not published on a seasonally adjusted basis because of a lack of sample-based histories.
Figure 1 shows the 1-month diffusion index for the 50 states and D.C. The index shows a clear cyclical pattern, but with a lot of volatility. There are only 2 months where the index reaches 100 (where all states are gaining employment), and zero months where the index reaches 0. There is usually at least one state that is contrary to the majority of states in any given month. For example, during the 2007–09 recession, North Dakota posted job gains in 11 of 18 months. The index is volatile, but a clear downward trend appears around the start of 2015. In part, this reflects job losses in oil-dependent states such as Alaska, North Dakota, Oklahoma, and Wyoming. These states saw net job losses in 2015 and 2016 following a steep drop in oil prices in the second half of 2014. Most of the states have gained jobs in 2017 and 2018.
The data available are experimental. BLS would appreciate any feedback on the usefulness of the data or the methods we use to develop the data. We will use this information to help us decide whether to make the data official. Send us comments through the feedback form at https://data.bls.gov/forms/sae-experimental.htm.
The data are available in an Excel file with 10 total worksheets. The first sheet contains data for the four All States Diffusion Indexes (1-month, 3-month, 6-month, and 12-month spans). The sixth sheet contains data for the four All MSAs Diffusion Indexes. The remaining eight worksheets contain graphs for each individual data series.
The first and sixth sheets containing the data each have ten columns. The variables are defined below:
|State_FIPS_Code||2-digit code equal to '00' for All States Diffusion Index series and '99' for All MSAs Diffusion Index series|
|Area_FIPS_Code||5-digit code equal to '00000' for All States Diffusion Index series and '99999' for All MSAs Diffusion Index series|
|Series_Code||8-digit series code equal to '00000000'. All Diffusion Index series are calculated using data at the total nonfarm level|
|DI_1month||Values for the 1-month span Diffusion Index|
|DI_3month||Values for the 3-month span Diffusion Index|
|DI_6month||Values for the 6-month span Diffusion Index|
|DI_12month||Values for the 12-month span Diffusion Index|
|Recession||Indicator of whether or not month was during a recession, equal to 100 if during a recession and 0 otherwise. These data are included for graphing purposes only, and are not related to the calculation of the diffusion index data. All recession dates determined by the NBER.5|
The data file linked below contains data from January 1991 to December 2018.
SAE Diffusion Index Dataset (.xlsx file)
1 See Getz and Ulmer, 1990. https://www.bls.gov/opub/mlr/1990/article/diffusion-indexes-an-economic-barometer.htm
2 CES estimates are categorized by ownership and industry. The Quarterly Census of Employment and Wages (QCEW) assigns respondents an ownership code — private or public with public ownership further divided into federal, state, or local. Respondents are then assigned a North American Industry Classification System (NAICS) code. NAICS codes group establishments into industries based on the activity in which they are primarily engaged. Establishments using similar raw material inputs, similar capital equipment, and similar labor are classified in the same industry. More information about NAICS codes in the CES State and Area program is available at https://www.bls.gov/sae/additional-resources/details-on-the-conversion-to-the-2017-north-american-industry-classification-system-naics-from-2012-naics.htm.
4 This includes all Metropolitan Statistical Areas and New England City and Town Areas (NECTA) published by BLS on a seasonally adjusted basis. As of the 2017 benchmark, only data for the Enid, Oklahoma, MSA is not available seasonally adjusted. The count of MSAs does not include Metropolitan Divisions, NECTA Divisions, or Nonstandard areas.
Last Modified Date: January 22, 2019