Report on Revision to Handbook Method Unemployment Estimation

Bureau of Labor Statistics
Local Area Unemployment Statistics
Research and Methods Branch
Report on Revision to Handbook Method Unemployment Estimation
March 2014

Executive Summary
The Local Area Unemployment Statistics (LAUS) Program proposes implementing an
updated approach to the estimation of substate unemployment. The first proposal is to
employ the primary substate unemployment estimation methodology, known as the
Handbook Method, at the level of individual counties (and equivalents) in States outside
of New England. Currently, the Handbook Method is executed at the level of Labor Market
Areas (LMAs), which consist of one or more counties. This proposal requires no changes
to the calculations constituting the Handbook Method, since the data required as inputs
already exist at the county level. By employing the Handbook Method at the county level,
substate estimates will better represent local conditions by using more locally available
data.
The second proposal involves updating the estimation of non-covered agricultural
unemployment. Twelve-month-ago monthly agricultural sector data from the Quarterly
Census of Employment and Wages (QCEW) will replace the annual average QCEW data
currently used to approximate the level of household employment participating in the
Unemployment Insurance (UI) Program. Also, a seasonal factor within the formula created
from national Current Population Survey (CPS) data in the 1970s and 1980s will be
eliminated. These changes will improve the ability of this subcomponent of unemployment
to represent local seasonal patterns.

1

Objective of the Research
Over the last several years, the LAUS Program, in cooperation with State partners, has
conducted extensive research into improving the ability of States to correctly classify and
allocate continued claims from the Unemployment Insurance Program to substate areas.
With the adoption of new technology and practices for the classification and allocation of
continued claims, the LAUS Program and its State partners now have confidence in the
accuracy of this input at greater levels of geographic specificity.
Given the importance of continued claims to the estimation of unemployment in the
Handbook Method, the LAUS Program investigated the possibility of expanding the use
of the Handbook Method to the county level. Reliable data were found to exist at the
county level to allow the use of the Handbook Method at this greater level of geographic
specificity. Research also revealed the desirability of reducing the use of secondary
methods of estimation when locally available data could be utilized to a fuller extent.
The LAUS Program also investigated several possibilities for improvement in the
estimation of the non-covered agricultural unemployed, an atypical method used by request
for areas within some States. First, research into the trends of these estimates revealed
unusual seasonal patterns where relatively high seasonal unemployment coincided with
seasonal periods of relatively high agricultural employment. State partners suggested that
these patterns may not reflect actual economic conditions in those areas. Also, the LAUS
Program sought to replace the seasonal factors used in the formula, which do not vary from
year to year or between different geographical areas, with data that better reflect the recent
experience of agricultural employment at the local level.
Description of the Current Methodology
The Handbook Method estimates monthly unemployment for an area using available
information without the expense of expanding a labor force survey like the Current
Population Survey (CPS). Handbook unemployment estimates are computed as the sum
of total continued claims (Line 8), unemployed exhaustees (Line 9), non-covered
agricultural unemployment (Line 10), unemployed re-entrants into the labor force (Line

2

13), and unemployed new entrants into the labor force (Line 15). This sum is then captured
in Line 16, total Handbook unemployment.
Handbook Unemployment
Line

Description

5

UI Claims

6

UCFE Claims

7

Rail Road Claims

8

Total Claims (lines 5 + 6 + 7)

9

Unemployed Exhaustees

10

Non-covered Agricultural Unemployment

11

Unemployed excluding Entrants (lines 8 + 9 + 10)

12

Re-entrants Ratio

13

Re-entrants

14

New Entrants Ratio

15

New Entrants

16

Total Unemployment (lines 11 + 13 + 15)

Lines 5-7: Covered Unemployment
The covered category represents unemployed individuals that are covered by UI benefits
programs. This category is further broken down into two groups: (1) those currently
receiving UI benefits and (2) those who have exhausted their benefits. The number in the
former group is obtained directly from an actual count of UI claimants for the reference
week. These are represented in Lines 5, 6, and 7 for regular State UI, UCFE, and Railroad
claims, respectively. Because claimants within each State’s UI system are associated with
a postal address, claimants can be reliably assigned to a county or equivalent for the
purposes of LAUS estimation.
Line 9: Exhaustees
Line 9 represents unemployed exhaustees. These are individuals who have exhausted all
regular UI benefits to which they are entitled and are no longer tracked by the State’s UI
system. When a claimant exhausts their benefits a final payment is indicated in their claim

3

record. Each week these final payments are pooled and survived based on duration of
unemployment data from the CPS.
Although exhaustee data are available at the county level, the current method counts them
at the LMA level and then allocates them to the LMA component counties based on each
county’s share of the LMA’s total continued claims. This redistribution of the exhaustee
data can distort the county counts of exhaustees.
This is particularly evident when a layoff occurs in one county in which the claimants later
exhaust their UI benefits. In this case the county that experienced the layoff has a larger
share of continued claims at the time of the layoff and thus a larger share of the LMA’s
pool of exhaustees. This results in the other county or counties making up the LMA having
a smaller shares of continued claims and being allocated a smaller share of the pool of
exhaustees at the time of the layoff. Later, once the claimants in the county affected by the
layoff become exhaustees, there tends to be a sharp decline in that county’s share of
continued claims that proportionally reduces that county’s share of the LMA’s pool of
exhaustees. Meanwhile, the other counties in the LMA tend to receive a larger share of the
pool of exhaustees, because their share of the LMA’s total continued claims rises as their
counts of continued claims remain relatively unchanged.
The calculation of Handbook unemployment at the county level rather than at the LMA
level would alleviate this phenomenon by maintaining exhaustee pools at a more local
level. In the case of the example above, the Handbook Method’s estimate of unemployed
exhaustees will better reflect the actual number of exhaustees in a county by confining
claims resulting from a layoff to the county in which the claimants reside.
Line 10: Non-covered Agricultural Unemployment
Line 10 designates the component of the Handbook Method used to the estimate the level
of non-covered agricultural unemployment at the LMA level. “Non-covered” signifies an
individual’s not being eligible for benefits through the Unemployment Insurance Program.
Eighteen States have requested and been approved to use this optional methodology.

4

The current methodology utilizes the following inputs provided for each LMA which is
assigned to a State that has been approved to use the methodology:
A01:

The LMA’s most recent annual average covered agricultural establishment employment
estimate from the QCEW, as of the beginning of the current production year

L03:

The current monthly level of total agricultural employment (i.e. both covered and noncovered) for the LMA as estimated by the Handbook Method before benchmarking to
model-based control totals

r:

The current monthly experienced unemployment rate excepting non-covered agriculture
for the LMA as estimated by the Handbook Method before benchmarking to model-based
control totals

w:

One of twelve monthly factors, fixed across all areas and all years, based on a seasonal
relationship created from national data on agricultural and wage and salary unemployment
rates from the CPS in the period from 1977 to 1982, as listed below:
January
February
March
April
May
June
July
August
September
October
November
December

0.934
0.934
0.972
0.791
0.606
0.586
0.623
0.638
0.662
0.676
0.900
1.014

These inputs are combined in the formula below to produce an area’s Line 10 estimate:

10

03

01 ∗
1

This formula exhibits several properties which act as implicit assumptions imposed on the
estimate of the level of non-covered agricultural unemployment. Research into some of
5

these properties led to the updates proposed in this document. First, the use of the annual
average QCEW input (A01) assumes that the level of covered agricultural employment
does not vary throughout the year. Total agricultural employment (L03), however, can and
does vary from month to month. The difference between the two, intended to approximate
the level of non-covered agricultural employment, is therefore partly the function of an
interaction between a seasonally varying input and an input that is fixed for the duration of
the year.
The seasonal factor (w) partly serves to compensate for this discrepancy in seasonality
between A01 and L03. It also serves to differentiate the rate of unemployment among the
non-covered agricultural labor force from that of the experienced labor force, r. However,
it does so using a relationship founded in national-level data from several decades ago,
imposing the assumption that this seasonal relationship holds across time and diverse local
geographies. LAUS examined currently available CPS data to evaluate the possibility of
updating this variable, but national monthly estimates of this population are statistically
unreliable. This unreliability is even more pronounced at lower levels of geography.
Lines 12-15: Entrants
The entrant unemployment category is not estimated directly, because unemployment for
these persons is not immediately preceded by a period of employment, which is required
to receive UI benefits. The CPS produces monthly estimates of entrant unemployment, but
these estimates suffer from statistical volatility due to the small sample sizes involved at
the State level and below. Therefore, estimates for new entrants and re-entrants are derived
from weighted 5-year averages based on current and historical State data from the CPS.
These averages are then allocated to all LMAs based on each LMA’s percentage share of
its State’s age-specific population. For new entrants, the LMA’s percentage share of the
population aged 16 to 19 years is used. For re-entrants, the LMA’s percentage share of the
population aged 20 years and over is used. These estimates, when summed with those from
the covered unemployment category, produce what is known as an LMA’s Handbook
unemployment estimate.

6

In the case of interstate LMAs (LMAs incorporating areas in more than one State), the part
of the LMA specific to a given State (known as an intrastate part) is allocated a portion of
the State’s average entrant unemployment using the method above. This estimate of
entrant unemployment is then summed with that of other intrastate parts to create the
estimate of entrant unemployment for the LMA.
Additivity Adjustment
LMA Handbook unemployment estimates are then controlled to Statewide unemployment
estimates in a process known as additivity adjustment. All Handbook unemployment
estimates specific to a State are summed, and then an individual area’s share of that sum is
used to proportionally allocate the State’s total official not seasonally adjusted
unemployment. The allocated amount is the official not seasonally adjusted level of
unemployment.
Disaggregation
To calculate estimates at the county level in the case of an LMA that consists of more than
one county, additivity-adjusted official LMA-level unemployment estimates are
disaggregated. The preferred technique is the known as the Population-Claims method.
This method utilizes a county’s proportion of its LMA’s UI claims data to allocate a share
of the LMA’s additivity-adjusted covered unemployment to the county. Additivityadjusted new entrant and re-entrant unemployment are allocated based on the county’s
share of the teenaged and adult populations, respectively.
This method is counterintuitive in the case of entrant unemployment among interstate
LMAs. As noted above, entrant unemployment is allocated to the intrastate parts of
interstate LMAs prior to the step of disaggregation. The distribution of a given population
is not identical to the distribution of unemployment among that population, yet the current
method use of shares of population both to disaggregate entrant unemployment from States
to the intrastate parts of an interstate LMA and then to disaggregate the (now additivityadjusted) interstate LMA’s entrant unemployment into its intrastate parts. This process
complicates estimation with an additional step without bringing additional information on
local circumstances.
7

Development and Evaluation of Alternative Methods
County-Level Estimation Using the Handbook Method
Research noted several drawbacks to the current methodology of Handbook Method
estimation at the LMA level. First, slowly changing exhaustee counts in Line 9 are
currently allocated by quickly changing continued claims, even though the necessary data
for directly estimating exhaustees at the county level now exists. Second, counterintuitive
disaggregation methods are used in the case of LMAs which cross State boundaries. These
disaggregation methods involve interstate disaggregation ratios being applied to all types
of unemployment, even entrant unemployment, which is first estimated by allocating Statespecific data to the intrastate parts of interstate LMAs.
Calculating Handbook unemployment at the county level instead of the LMA level solves
these issues. The vast majority of data needed for Handbook estimation is available at the
county level. Currently, values are often recomputed as counties are summed to the LMA
and then disaggregated back to the county level. This results in discrepancies between
unemployment disaggregated to the county level using current methods and a directly
estimated equivalent for a given county. Directly estimating county-level unemployment
in the Handbook Method will render substate estimates more reflective of locally available
data of comparable reliability to that used in the current LMA-level estimation.
Line 10
There are multiple limitations to any effort to measure the non-covered agricultural
unemployed. First, no direct estimates of this population currently exist. Research by the
LAUS Program relies on CPS national estimates of the unemployment of persons formerly
employed in unincorporated agriculture, an approximation of the concept sought. Even
these national estimates display high sampling error. Second, no direct estimates of noncovered household agricultural employment exist. Participation versus non-participation
in the UI Program is not a component of the CPS Questionnaire, and QCEW measures the
employed participating in the UI Program only as they exist on establishment payrolls.
Establishment employment differs from household employment in several ways, including
the treatment of multiple job holders and the distinction between place of work and place

8

of residence. Any method for the estimation of non-covered agricultural unemployment,
particularly at the local level, will necessarily be indirect given available data.
The concentration of non-covered agricultural unemployment in LMAs of low population,
as estimated by the current methodology and asserted by State partners utilizing the
methodology, is another important aspect of the estimation of Line 10. While Line 10 is a
small proportion of unemployment as estimated using the Handbook Method in aggregate,
Line 10 has on occasion constituted a quarter or more of unemployment in a subset of areas
during some months.
Given the limited data available and the importance attributed to Line 10 estimation by the
LAUS Program’s State partners, the LAUS Program considered alternative uses of existing
data sources. To avoid the issues raised by the use of annual average QCEW data, LAUS
obtained access to QCEW data for internal use in estimating Line 10, non-covered
agricultural unemployment. Monthly QCEW employment in the agricultural sector at the
LMA level is often suppressed in publication to protect the confidentiality of respondents
in the program. LAUS, in consultation with the QCEW program, has come to the
conclusion that present security measures and the manipulation of the data involved in the
estimation of substate unemployment are sufficient to protect respondent-identifiable
information potentially present in monthly QCEW agricultural data.
By using year-ago monthly data to approximate covered agricultural employment, the
residual meant to approximate non-covered agricultural employment (represented in the
formula as (L03- A01)) will gain a plausible and local seasonality. LAUS therefore took
the step of removing the seasonal factor w, which had no basis in data that was either
current or local. The proposed formula is presented below, with A01 now representing the
year-ago monthly QCEW agricultural employment estimate:

10

03

01 ∗

The formula now imposes the assumption that the presumed unemployment rate among
the subset of the labor force associated with non-covered agriculture will be identical to
that of the remainder of the experienced labor force as estimated by the Handbook Method.
9

However, the LAUS Program and its State partners are aware of no reliable, current local
data to inform a different unemployment rate among this population.
Recommendations
The LAUS Program recommends the use of the new approach of estimating unemployment
at the county level. The data is currently available at the county level, and the new
approach solves the misallocation and disaggregation issues in the estimation of
unemployed exhaustees and unemployed entrants. These changes in methodology should
be incorporated into the 2015 redesign.
The LAUS Program recommends the use of the new approach to estimating non-covered
agricultural unemployment within the Handbook Method employment as explained above.
Using a monthly QCEW input will improve the seasonal properties of the estimate.
Removing the uniform national seasonal adjustment factor will remove an assumption
imposed on local data which could neither be rendered current nor local. These changes
in methodology should be incorporated into the 2015 redesign.

10