Report on Revision to Handbook Method Employment Estimation

substate_emp.pdf

Local Area Unemployment Statistics Program

Report on Revision to Handbook Method Employment Estimation

OMB: 1220-0017

Document [pdf]
Download: pdf | pdf
Bureau of Labor Statistics
Local Area Unemployment Statistics
Research and Methods Branch
Report on Revision to Handbook Method Employment Estimation
March 2014

Executive Summary
The Local Area Unemployment Statistics (LAUS) program proposes implementing an updated
approach to the estimation of substate “handbook method” employment. Non-agricultural wage
and salary employment will be disaggregated to the county level outside of New England and the
Minor Civil Division (MCD) level in New England to correspond with the geographic level at
which the other components of handbook method substate employment and unemployment will be
input for estimation. Also, the threshold for inclusion of commuter areas in dynamic residency
adjustment ratio (DRR) calculations will be changed from 100 employed persons to 10% of
residence area commuter employment. This change is being made to streamline estimation
operations and to avoid the inclusion of spurious commutation areas. The methodologies for
estimating “all-other” and agricultural employment have been completely revised. This
methodology update is necessitated by the discontinuation of the decennial Census “long” form
questionnaire. The new methodology uses more timely data captured by the Current Population
Survey (CPS) and the American Community Survey (ACS) to replace Census “long” form data.
For all-other employment, an advantage of this new methodology is that it no longer relies on the
correlation between wage and salary employment and all-other employment. While this
relationship holds true for most areas, the assumption that these two measures move in tandem is
not always borne out in the data. This methodology revision also improves upon the current
approach to estimating agricultural employment at the level of agricultural regions defined by the
USDA Farm Labor Survey. Estimation of agricultural employment tailored to the State level is
possible using the revised methodology based on CPS and ACS data.

1

Introduction
The “handbook method” is used by the LAUS program to generate estimates for 2,350 intrastate
and interstate Labor Market Areas (LMAs). With this approach, various labor force components
are individually measured and subsequently added to produce handbook labor force estimates for
the LMAs. Once the handbook estimates for all the LMAs within a State have been created, they
then are controlled to the official State labor force estimates produced by the time-series models to
become the official LAUS estimates.
Substate handbook employment estimates are computed as the sum of Line 1 (Non-agricultural
Wage & Salary Employment), Line 2 (All-other Employment), and Line 3 (Agricultural
Employment). This sum is then captured in handbook Line 4, total Handbook employment.
Figure 1. Handbook Employment Line Items
Employment 
Line 

Description 

1 

Non‐agricultural Wage & Salary Employment 

2 

All‐other Employment 

3 

Agricultural Employment 

4 

Total Handbook Employment (lines 1 + 2 + 3) 

As part of a general evaluation of the LAUS methodology, the methodology for estimating nonagricultural wage & salary employment was selected for modification for the following reasons:
Decennial Census journey-to-work employment inputs are no longer being produced; The existing
threshold for including commutation areas in dynamic residency adjustment ratio calculations
included many unrealistic commuting areas; and new employment and unemployment inputs will
be created at the county level (MCD level in New England) instead of the LMA level to create
more accurate estimates at finer levels of geographic detail.
All-other employment accounts for the employment not captured either in the handbook estimate
for non-agricultural wage and salary employment (line 1) or agricultural employment (line 3).
All-other employment accounts for self-employed persons who work in their own unincorporated
business, unpaid family members who work for a business owned by a family member; and
private household workers. The methodology for estimating all-other employment was selected
 

2
 

for modification for the following reasons: The Census “long” form all-other employment inputs
are no longer available; the assumption that there is a positive correlation between non-agriculture
wage and salary employment (NAWS) and all-other employment does not always prove to be
accurate, which indicates that the step-3 ratio may introduce error resulting from indexing change
in all-other employment to change in non-agriculture wage and salary employment; and the strata
designed to account for local labor market conditions are assigned for a period of at least 10 years,
making them less likely to reflect current conditions as result.
The methodology for estimating agricultural employment was selected for modification for the
following reasons: The Census “long” form all-other employment inputs are no longer available;
the employment inputs for agricultural regions adopted from the Farm Labor Survey are not
always representative of the trends and seasonality of the component States.
Objective of the Research
The goal of this research into the LAUS Handbook employment methodology was to find new
data sources for journey-to-work employment, non-agricultural wage and salary employment, allother employment, and agricultural employment and to develop an approach that would provide
more accurate measures of the components of substate handbook method employment.
Line 1: Non-agricultural Wage & Salary Employment
The Handbook method calculation of non-agricultural wage and salary employment begins with
input data that pertain to jobs by place of work. Because employment estimates from these sources
are based on the location of the establishment, these "place-of-work" estimates must be adjusted to
reflect the place-of-residence concept used in the CPS survey of households. Resident
employment includes workers living and working in the same area and also those who work in
other areas within commuting distance. Estimates of resident employment should, therefore,
reflect employment changes in those related commutation areas as well. In 2005, LAUS
introduced dynamic residency ratios (DRRs) to provide this adjustment. Multiple residency
adjustment ratios were produced, using Census 2000 county-to-county worker commuting data
and March/April 2000 total nonfarm job estimates. Separate residency adjustment ratios were
developed for each estimating area and up to four additional labor market areas into which at least
 

3
 

100 residents commuted to work. Ratios for each of the commuting areas are multiplied by their
respective monthly nonfarm jobs estimates to produce estimates of estimating area residents who
work in each of the commuting areas. Separate commuting area estimates are summed to create a
total of the resident nonfarm wage-and-salary employed for the estimating area. This adjustment
also accounts for multiple jobholding and unpaid absences in the payroll employment estimates.

When the LAUS implemented dynamic residency ratios, an important consideration in
implementing this change was that it would be at least a decade until the DRR commuting areas
would be updated using journey-to-work data from the next decennial census. For this reason, a
relatively low threshold for the inclusion of commuting areas was applied in order to capture as
much change in area commuting patterns over a long time frame as possible. Up to five total
commuting areas with 100 or more commuters each from the residence area (50 or more
commuters in New England) were included in residence area DRR calculation. We no longer face
the time frame associated with decennial Census journey-to-work employment estimates, as ACS
releases journey-to-work employment estimates every five years for the purpose of the OMB
metro and micro area geography updates. It has been determined that 10% of total commuter
employment is a preferable threshold for including commutation areas in a residence area’s DRR
calculation. This new threshold achieves a desirable balance between operational streamlining and
 

4
 

capturing as much economic information as necessary. This change was based on the increased
frequency of availability of these commutation data from ACS combined with the appearance of
many superfluous and unreasonable commuting areas when the current threshold was applied to
the ACS journey-to-work data. The list of DRR areas will be updated each time a new ACS
commutation employment dataset is released.
LAUS estimation is currently undertaken at the LMA level for substate areas calculated using the
handbook method. Establishment-based Non Agricultural Wage and Salary employment estimates
and labor-management disputants are input at the LMA level. A dynamic residency ratio is then
applied to these inputs to convert the establishment-based employment to a residency employment
basis. As part of the methodology revision, all MCDs in New England and counties outside of
New England will be treated as Handbook method areas. This change will bring about a
distinction between the geography of Handbook Method areas and the geography of
establishment-based employment inputs. Developing handbook line 1 estimates at the county and
MCD levels will be accomplished by building a disaggregation step into the line 1 calculation.
Line 1 values will be calculated in an intermediate step for multi-county or multi-MCD areas.
These intermediate handbook line 1 values will then be disaggregated to the county- or MCD-level
using ACS non-agricultural wage and salary employment ratios derived from the most recent ACS
five-year dataset. Essentially, the calculation of Handbook line 1 will be the same as it is currently
but with the added step of disaggregating the multi-county or multi-MCD area line 1 value to the
county- or MCD-level. This change in the calculation of line 1 employment estimates will bring
about geographic consistency with the other components of handbook employment and
unemployment estimation, which will also be input at the county or MCD level. In addition,
developing handbook method inputs at a more granular geographic level will allow better
operational flexibility in future updates to the geographic definitions of labor market areas as
counties (and MCDs in New England) are the basic component in OMB definitions of LMAs.
Line 2: All-other Employment
Description of the Current Methodology
The current approach to estimating all-other employment is based on the assumption that there is a
correlation at the national level between changes in non-agriculture wage and salary employment
 

5
 

and all-other employment. The original analysis which led to this estimating methodology was
based on an examination of the relationship between all-other employment and wage-and-salary
employment in the Nation as a whole and in a randomly selected sample of areas using the 1940
and 1950 Census data. It was found that in both the areas and the Nation, the relative change in
wage-and-salary employment was accompanied by a proportional relative change in all-other
employment. In other words, slow wage and salary growth was accompanied by slow all-other
employment growth, and rapid wage and salary growth was accompanied by rapid all-other
employment growth. It was also found that the proportional relative changes in all-other
employment in the areas and in the Nation were very close to each other. This meant that the
relative change in area all-other employment could be derived given the relative change in area
wage and salary employment and the ratio of the relative national change in all-other employment
to the relative national change in wage-and-salary employment.

Analyses utilizing data from subsequent Censuses corroborated the findings of the original study.
However, discrepancies between individual areas, on the one hand, and areas and the Nation on
the other, proved quite common and pointed out the need for area adjustment. The CPS sample
expansion of the 1970's provided additional geographic detail on all-other employment and
allowed the opportunity for analysis and testing of differences in the proportionality factor
between States.

Following each Census, the relative change in wage and salary employment divided by the relative
change in all-other employment was calculated and reviewed. Clusters of
States with similar proportionality constants were grouped into strata. Four strata were defined
following the 1980 Census and three were defined following the 1990 and 2000 Censuses. By
grouping States into strata based on their ratio of relative change, it was found that LMA all-other
employment estimates could be improved. Specifically, using the proportionality factor for Statebased strata to estimate all-other employment for LMAs significantly reduced the range of error in
estimating all-other employment.

These ratios of relative change are called “Step 3 Ratios” and are created and distributed on a
monthly basis. States with similar relative changes in non-agriculture wage and salary
 

6
 

employment and all-other employment are grouped into one of three strata. Once the State strata
are established, LMAs are assigned to one of the three State strata based on the LMA’s relative
change in non-agriculture wage & salary employment and all-other employment.

Three steps are then taken to estimate all-other employment: First, the LMA’s change ratio of
nonagricultural wage & salary employment is created using the decennial Census base estimate
and the reference month estimate. In the second step, this ratio is used to extrapolate the LMA’s
Census all-other employment base estimate through the reference month. In the third step, the
extrapolated LMA estimate is adjusted using the Step 3 ratio corresponding to the assigned State
stratum.

For example, if the LMA’s Census all-other employment count is 10,000 and the change ratio
between the LMA’s Census and the current non-agriculture wage & salary employment is 15%,
the extrapolated estimate is 11,500 (10,000*.15). If the Step-3 ratio is .98, the LMA all-other
employment estimate is 11,270 or (11,500*.98).

One issue with this methodology is its reliance on a decennial Census input that becomes
progressively more outdated as time elapses from the previous decennial Census. In the attempt to
produce an all-other employment component of handbook employment estimation that reflects
current economic conditions in LMAs, change in non-agricultural wage and salary employment is
used as a proxy for change in all-other employment. Furthermore, LMAs are differentiated within
three separate strata to account for differing relationships between change in all-other employment
and change in non-agricultural wage and salary employment. While this method mostly proves
effective, subsequent analysis using ACS all-other employment estimates has highlighted that for
certain areas, the relationship between changes in all-other and non-agricultural wage and salary
employment does not hold true.
Development and evaluation of alternative methods
The CPS and the ACS were identified as the most appropriate sources for the all-other
employment data that were previously obtained from the decennial Census long-form survey. For
LAUS purposes, both ACS and CPS offer differing strengths and drawbacks. The goal was to
 

7
 

utilize the strengths and to mitigate the drawbacks of each data source. For instance, the CPS data
are only available at the State level, while the ACS data are available at the detailed geographic
level needed for LAUS substate Handbook method estimation. Also, the CPS data are current and
are available on a monthly basis, while the required ACS data are available on a yearly basis in the
form of 5-year estimates.
Since the monthly CPS data are only available at the State level and the ACS provides more
geographically detailed data, the ACS data are used to distribute the CPS data to substate areas. To
do this, the ACS all-other employment estimate for a given area is divided by the sum of ACS allother employment for all areas within the State. The resulting ratio for a given area is referred to
as the “ACS share”. The ACS share is expressed as:
	

	

	

	

Where:	
	area	estimate	of	ACS	all‐other	employment	

∑

	 	sum	of	areas’	estimates	of	ACS	all‐other	employment

The ACS shares of all-other employment are relatively stable from year to year and are used to
disaggregate CPS monthly statewide all-other employment to the area level. While Handbook line
2 all-other employment was previously calculated at the LMA level, the ACS data allow for the
development of these inputs at the county level outside of New England and at the MCD level for
New England. The precedent for using ACS data to disaggregate CPS all-other employment
comes from the handbook methodology used to estimate new entrant and reentrant unemployment.
This method assigns a portion of the CPS statewide new entrant and reentrant unemployment to
individual areas based on a population-specific ratio derived for the specific area.
Research showed that CPS all-other employment estimates at the State level tend to be volatile
month-to-month and are not suitable for direct use. To mitigate the volatility of the CPS monthly
statewide all-other employment estimates and obtain inputs more suitable for handbook
estimation, five years of CPS data for a given month are used to develop weighted-average
estimates. This allows the current month’s CPS estimate to gain strength from prior year
estimates while retaining the seasonality of the reference month. (Figure 2. shows the weights
 

8
 

used, where “y” is the current year.) For consistency, the sum of 5-year State weighted averages is
controlled to the current monthly national CPS estimate of all-other employment.
Figure 2.
Year 
y 
y – 1 
y – 2 
y – 3 
y – 4 

Weight 
0.40 
0.25 
0.20 
0.10 
0.05 

Using a weighted average of statewide CPS all-other employment and the area ACS share to
generate the handbook method area all-other employment estimate is expressed as follows:
All‐Other	Emparea	 	 CPSwto	*	ACSshareo *	CPSr	
Where:	
CPSwto	 	Weighted	average	of	the	given	month’s	CPS	all‐other		
																employment	for	the	area	
ACSshareo			 	ACS	share	of	all‐other	employment	for	the	area	
CPSr			 	Ratio	for	controlling	sum‐of‐State	weighted	averages	
															to	national	CPS	all‐other	employment

ACS data provide estimates for all-other employment at a more detailed level of geography and
allow for the elimination of the Step 3 ratios, which introduce error into the estimate when allother employment moves differently from non-agriculture wage and salary employment.

Research findings show that the new methodology generally improves handbook method all-other
employment estimates. For example, for some areas the new methodology produces estimates that
much more closely approximate the most recent corresponding ACS all-other employment
estimates. ACS data are a suitable benchmark in this instance, since they are the only available
direct estimates of all-other employment at the detailed area level to be published since the Census
2000 long form data was made available. Overall, for most handbook method areas the
incorporation of all-other employment estimates based on the new methodology has a limited
impact on total handbook employment and the corresponding unemployment rate due to the
 

9
 

limited proportion of all-other employment in the majority of handbook-estimated areas.
Occasional large differences in labor force estimates between the current and the new
methodology appear to be caused by improvements over the shortcomings of the old
methodology, which sometimes deviated from the more current ACS direct estimates of all-other
employment.
Line 3: Agricultural Employment
Description of the Current Methodology
Unlike the non-agricultural Handbook employment estimates, which split employment by class of
worker—wage and salary (line 1) and “all-other” (line 2)—the agricultural Handbook employment
estimate encompasses all classes of worker—wage and salary, self-employed, and unpaid
family—in a single estimate. This is accomplished by applying a monthly change factor to a
decennial base of agricultural employment obtained from 2000 Census long form data. The
following formula shows the calculation for each LMA.
L03   =   ( C05 )   x   ( Change factor ) 
Where: 
Variable 

Description (LSS Plus Variable ID) 

L03 

Handbook Agricultural Employment 

C05 

Census Agricultural Employment (C05)  

Change 
factor 

Agricultural Employment Monthly Change 
Factor (G01 – G21) 

Development of the Agricultural Employment Methodology

Prior to the incorporation of 2000 Census data into the Handbook methodology, the procedure for
agricultural employment estimation utilized information from the 1990 Census, the Current
Population Survey (CPS), and the Department of Agriculture’s Farm Labor Survey (FLS). As of
2002, the FLS ceased to provide information for all farm workers and began limiting its quarterly
 

10
 

publication to information for hired workers only. Because hired workers account for only 35 to
50 percent of all agricultural workers, the FLS data became an inadequate benchmark for
Handbook agricultural employment estimation.

To be congruent with the CPS definition of employment, the self-employed and unpaid family
workers must be included in addition to hired workers. Because of this, FLS data are no longer
used. Currently, unpublished monthly estimates of agricultural employment from the CPS are
used in lieu of FLS data.

Agricultural Regions

The Agriculture Department, through the FLS, designated twenty-one estimating regions. Fifteen
of the regions were creating by grouping States that have similar agricultural activities, while six
others each comprise only one State. Though LAUS no longer uses FLS data, the Handbook
methodology continues to utilize the FLS agricultural regions. The regions are listed in the
following table. 
 
Region 
Number 
1 
2 
3 
4 
5 
6 
7 
8 
9 
10 
11 
12 
13 
14 
15 

Agricultural Region 
Northeast I 
Northeast II 
Appalachian I 
Appalachian II 
Southeast 
Florida 
Lake 
Corn Belt I 
Corn Belt II 
Delta 
Northern Plains 
Southern Plains 
Mountain I 
Mountain II 
Mountain III 

State(s)* 
CT, ME, MA, NH, NY, RI, and VT 
DE, MD, NJ, and PA 
NC and VA 
TN and WV 
AL, GA, and SC 
FL 
MI, MN, and WI 
IL, IN, KY, and OH 
IA and MO 
AR, LA, and MS 
KS, NE, ND, and SD 
OK and TX 
ID, MT, and WY 
CO, NV, and UT 
AZ and NM 
 

11
 

16 
17 
18 
19 
20 
21 

Pacific 
California 
Hawaii 
Michigan 
Minnesota 
Wisconsin 

OR and WA 
CA 
HI 
MI 
MN 
WI 

 
* Alaska, the District of Columbia, and Puerto Rico are not included in any agricultural estimating 
region.   
 

Agricultural Employment Monthly Change Factors

A change factor created at the agricultural region level from the component States’ annual average
CPS agricultural employment estimates is used annually to rebase to the LMA’s decennial Census
agricultural employment estimate. This ratio is then multiplied by the annual change factor from
the previous year in order to move the decennial Census area estimate forward. The annual change
factor is calculated as the CPS annual average agricultural employment for the agricultural region
in the recently completed year (y) divided by the CPS annual average agricultural employment for
the agricultural region from the prior year (y-1) multiplied by the annual change factor from the
previous year.
 
AACPS (y) 
 

Annual change factor =

Annual change factor (y‐1)
AACPS (y‐1)

 

A current production month change factor is then created by applying the annual change factor to
a ratio of the reference month CPS agricultural employment for the agricultural region divided by
the July CPS agricultural employment of the previous completed year.
 
Ref month CPS

 

Monthly factor = Annual change factor   
 

 

12
 

July CPS

Each month the current change factor produced using the above formula is applied to the LMA
agricultural employment estimate from the 2000 Census to arrive at the current month’s total
agricultural employment estimate for the LMA.
Development and evaluation of alternative methods
The CPS and the ACS were identified as the most appropriate sources for the agricultural
employment data that were previously obtained from the decennial Census long-form survey. For
LAUS purposes, both ACS and CPS offer differing strengths and drawbacks. The goal was to
utilize the strengths and to mitigate the drawbacks of each data source. For instance, the CPS data
are only available at the State level, while the ACS data are available at the detailed geographic
level needed for LAUS substate Handbook method estimation. Also, the CPS data are current and
are available on a monthly basis, while the required ACS data are available on a yearly basis in the
form of 5-year estimates.
Since the monthly CPS data are only available at the State level and the ACS provides more
geographically detailed data, the ACS data area used to distribute the CPS data to substate areas.
To do this, the ACS agricultural employment estimate for a given area is divided by the sum of
ACS agricultural employment for all areas within the State. The resulting ratio for a given area is
referred to as the “ACS share”. The ACS share is expressed as:
	

	

	

	

Where:	
	area	estimate	of	ACS	agricultural	employment	

∑

	 	sum	of	areas’	estimates	of	ACS	agricultural	employment

The ACS shares of agricultural employment are relatively stable from year to year and are used to
disaggregate CPS monthly statewide agricultural employment to the area level. While Handbook
line 3 agricultural employment was previously calculated at the LMA level, the ACS data allow
for the development of these inputs at the county level outside of New England and at the MCD
level for New England. Developing handbook method inputs at a more granular geographic level
will allow better operational flexibility in future updates to the geographic definitions of labor
 

13
 

market areas. The precedent for using ACS data to disaggregate CPS agricultural employment
comes from the handbook methodology used to estimate new entrant and reentrant unemployment.
This method assigns a portion of the CPS statewide new entrant and reentrant unemployment to
individual areas based on a population-specific ratio derived for the specific area.
Research showed that CPS agricultural employment estimates at the State level tend to be volatile
month-to-month and are not suitable for direct use. To mitigate the volatility of the CPS monthly
statewide agricultural employment estimates and obtain inputs more suitable for handbook
estimation, five years of CPS data for a given month are used to develop weighted-average
estimates. This allows the current month’s CPS estimate to gain strength from prior year
estimates while retaining the seasonality of the reference month. (Figure 2. shows the weights
used, where “y” is the current year.) For consistency, the sum of 5-year State weighted averages is
controlled to the currently monthly national CPS estimate of agricultural employment.
Figure 2.
Year 
y 
y – 1 
y – 2 
y – 3 
y – 4 

Weight 
0.40 
0.25 
0.20 
0.10 
0.05 

Using a weighted average of statewide CPS agricultural employment and the area ACS share to
generate the handbook method area agricultural employment estimate is expressed as follows:
Agricultural	Emparea	 	 CPSwta	*	ACSsharea *	CPSr	
Where:	
CPSwta	 	Weighted	average	of	the	given	month’s	CPS	agricultual		
																employment	for	the	area	
ACSsharea		 	ACS	share	of	agricultural	employment	for	the	area	
CPSr			 	Ratio	for	controlling	sum‐of‐State	weighted	averages	
															to	national	CPS	agricultural	employment

 

14
 

ACS data provide estimates for agricultural employment at the detailed level of geography
required for substate handbook estimation and CPS weighted averages provide State-specific
monthly control totals for the application of ACS area agricultural employment shares. This
allows for the preservation of State-specific trend and seasonality in agricultural employment, and
enables us to eliminate reliance on the agricultural regions inherited from when FLS data was still
used in estimating handbook line 3.

Research findings show that the new methodology generally improves handbook method
agricultural employment estimates. For example, for some areas the new methodology produces
estimates that much more closely approximate the most recent corresponding ACS agricultural
employment estimates. ACS data are a suitable benchmark in this instance, since they are the only
available direct estimates of all agricultural employment at the detailed area level to be published
since the Census 2000 long form data was made available. Overall, for most handbook method
areas the incorporation of agricultural employment estimates based on the new methodology has a
limited impact on total handbook employment and the corresponding unemployment rate due to
the limited proportion of agricultural employment in the majority of handbook-estimated areas.
Occasional large differences in labor force estimates between the current and the new
methodology appear to be caused by improvements over the shortcomings of the old
methodology, which sometimes deviated from the more current ACS direct estimates of
agricultural employment.
Recommendations
The LAUS program recommends the use of the new approaches to estimating the various
components of handbook method employment enumerated above. Changes to the estimation of
line 1 reflect the increased availability of journey-to-work employment data and ensure the
geographic consistency of the input level for handbook method estimation. State data from the
CPS represents the most current source of all-other and agricultural employment and the use of
ACS shares of all-other and agricultural employment prove to be the best method to effectively
distribute these data to the States’ LMAs. These changes in methodology should be incorporated
into the 2015 redesign.

 

15
 


File Typeapplication/pdf
File TitleReport on Revision to Handbook Method Employment Estimation
Subjecthandbook employment, 2015 LAUS redesign
AuthorU.S. Bureau of Labor Statistics, Local Area Unemployment Statist
File Modified2014-08-14
File Created2014-08-14

© 2024 OMB.report | Privacy Policy