Data Sharing October 2003 FESAC by HC120727154814


									Exploring Differences in Employment between
     Household and Establishment Data

Katharine G. Abraham
 University of Maryland and NBER
John C. Haltiwanger
 University of Maryland and NBER
Kristin Sandusky
 U.S. Census Bureau
James R. Spletzer
 U.S. Bureau of Labor Statistics

                  2008 World Congress (May 15, 2008)
• Each month, the U.S. Bureau of Labor Statistics
  releases data on current employment. The data
  come from two different surveys:
   – the Current Population Survey (CPS), also known as the
     household survey, and
   – the Current Employment Statistics survey (CES), also known as
     the payroll survey or the establishment survey.

• Despite their different definitions, samples,
  estimation procedures, and concepts, the CPS and
  CES employment series track well over long periods.
• However, at times, their rates of growth and decline
  can differ significantly.
CES & CPS data used in previous chart

Adjusted CPS Household Survey
• Published every month by the BLS
• Make the CPS look more like the CES
• Subtract CPS employment not in scope to CES:
  – Unincorporated self employed (9½ million in 2004)
  – Agricultural employment, unpaid family workers, private HH
    workers, workers on unpaid absences (5 million in 2004)
• Make the CPS on a jobs concept similar to CES:
  – Add multiple jobholders (7 million in 2004)
CES & CPS data used in previous chart
CES (Establishment Survey)                CPS (Household Survey)
•   Monthly sample survey of 160,000      •   Monthly sample survey of
    businesses (400,000 worksites)            approximately 60,000 households
•   Designed to measure employment        •   Designed to measure employment
    with industrial and geographical          and unemployment with
    detail                                    demographic detail
•   Employment measures the number        •   Employment measures the number
    of nonfarm payroll jobs                   of employed persons
•   Reference period is the pay period    •   Reference period is the week that
    that includes the 12th of the month       includes the 12th of the month
•   Employees of all ages are included    •   Workers aged >16 are included
•   Self-employed are excluded            •   Self-employed are included
•   Multiple jobholders are counted for   •   Multiple jobholders are counted
    each job                                  once
•   Agricultural sector and private       •   Agricultural sector and private
    household workers are excluded            household workers are included
•   Workers on leave without pay are      •   Workers on leave without pay are
    excluded                                  included
•   Employment estimates are              •   Employment estimates are
    benchmarked annually to universe          controlled to estimates of the civilian
    counts                                    noninstitutional population

• As is evident from the previous chart,
  employment from the establishment and
  household surveys had diverging trends
  during the late 1990s and early 2000s

• During the past 60 years, this late 1990s and
  early 2000s discrepancy was unprecedented
  in size and duration
 Ratio of establishment survey employment to household survey
 nonagricultural wage and salary employment, 1948-2004
1.06                                                                                                         1.06

1.05                                                                                                         1.05

1.04                                                                                                         1.04

1.03                                                                                                         1.03

1.02                                                                                                         1.02

1.01                                                                                                         1.01

1.00                                                                                                         1.00

0.99                                                                                                         0.99

0.98                                                                                                         0.98
    1948   1952   1956   1960   1964   1968   1972   1976   1980   1984   1988   1992   1996   2000   2004
• Although many theories about the recent CPS-
  CES divergence have been put forth, complete
  explanations have never been found
  –   Sampling error
  –   Household population controls
  –   Missed births in the CES
  –   Age minimum (16+) in CPS
  –   Persons with more than two jobs
  –   2nd civilian jobs of armed forces
  –   Employment of noninstitutional population
  –   Welfare to work programs
  –   Foreign commuters
  –   Undocumented immigrant workers
• Other possible explanations for the CPS-CES
  divergence (continued)
  – Worker classification
     • Measurement error in self-employment
  – “Off-the-books” employment
     • Employers not reporting workers that would be measured as
       employed in the CPS
  – Marginal employment (informal jobs)
     • Short duration or low earning jobs that are not reported by
       household survey respondents
  – Job Changing
     • If a person changes jobs during the survey reference period,
       both jobs would be counted in the CES but only once in the

• Our research strategy is to match CPS microdata to
  UI wage records for the same individuals to learn
  whether differences in household versus employer
  reports of individuals’ employment histories can offer
  an explanation
• Our research today focuses on:
   – “Off-the-books” nonstandard employment
   – Marginal (low-earnings) short-duration jobs

• These explanations would need business cycle
  properties in order to explain the divergence
   – We believe this is plausible
      Measurement Framework

Linked CPS – UI    Individual holds job in
wage records       UI wage records
                         Yes                 No

holds job    Yes          X1                 X2
in CPS
             No           X3                 X4
    Measurement Framework

Number of persons employed in CPS & UI

CPS number of persons employed
    X1 + X2

UI wage records number of persons employed
     X1 + X3

Difference (CPS - UI) number of persons employed
      X2 – X3
    Measurement Framework
Linked CPS – UI      Individual holds job in
wage records         UI wage records
microdata               Yes       Yes        No
                       1 Job    2+ Jobs
Individual Yes
                       Y1         Y2        Y3
holds job 1 Job
in CPS
                       Y4         Y5        Y6
           2+ Jobs

          No           Y7         Y8        Y9
    Measurement Framework

Number of jobs in CPS & UI

Assume: multiple jobholders in CPS hold (1+m) jobs
       multiple jobholders in UI hold (1+n) jobs
       m>1, n>1

CPS number of jobs:
    Y1 + Y2 + Y3 + m(Y4 + Y5 + Y6)

UI number of jobs:
     Y1 + Y4 + Y7 + n(Y2 + Y5 + Y8)
 Measurement Framework

Our goal: estimate each of the components
underlying the CES-CPS employment trend
discrepancy and examine their cyclical

For example, (X2 - X3)
   X2: employed in CPS but not in UI wage records
       “Off-the-books” nonstandard employment
   X3: employed in UI wage records but not in CPS
       Marginal short duration jobs
     Measurement Framework

“Off-the-books” nonstandard employment        X2 countercyclical
arrangements that are not reported by           X1 procyclical
employers shrinks in relative importance      Y4 countercyclical
during business cycle expansions                Y5 procyclical

Marginal (low earnings) short duration jobs
that are not reported by household survey       X3 procyclical
respondents grow in number during               Y2 procyclical
business cycle expansions

Increases in the job-changing rate during
business cycle expansions leads to relative
increases in employment counts                  Y2 procyclical
                  Data: CPS

• Monthly household survey that collects information
  about the labor force status of those aged 16+
• Survey conducted in person or by telephone
• Approximately 60,000 households interviewed each
  month, with a single respondent generally reporting
  for all members of the household
• Households are in the sample for 4 months, out for
  8 months, and in for another 4 months
• Survey sample in each month represents the civilian
  non-institutionalized population
                  Data: CPS

 Vitally important to our research is the ability to
  match individuals in the CPS to various
  administrative datasets at the U.S. Census

 Approximately 70-80% (the exact number varies
  by year) of individuals in the March supplement
  have what is called a Protected Identity Key
  (PIK). The PIK is a unique internal Census
  Bureau identification number for the individual.
        Data: UI Wage Records
•   Administrative data are necessary to operate the
    State Unemployment Insurance (UI) programs
    –   UI provides unemployment benefits to eligible workers
        who are unemployed through no fault of their own
    –   In general, UI benefits are based on a percentage of an
        individual’s earnings over a recent 52 week period
    –   To determine benefits for UI claimants, States need an
        individual’s earnings history
    –   Wage records are the employer-reported administrative
        data underlying the State UI programs
•   Wage records are essentially a quarterly universe
    of employed persons on nonfarm payrolls
    –   3 data elements: SSN, UI number, quarterly earnings
           Data: UI Wage Records

•   UI wage records “related” to CES employment
    – Employers subject to State UI laws are required to
      file two forms every quarter:
       1) Quarterly Contributions Report (QCR)
          The foundation of the QCEW data, 9 million estabs
       2) Quarterly Wage Record (QWR), 136 million employees

    – CES employment is benchmarked annually to the
      QCEW employment counts
       •   We believe that comparisons of CPS to UI wage records
           for the same individuals will be potentially informative
           about reasons for CPS versus CES discrepancy
        Data: Analysis Sample
• Census LEHD program has UI wage record data
  for 17 states from 1996 to present
• CPS restrictions:
  – Because a quarterly CPS employment variable is needed for an
    “apples to apples” comparison to UI wage records, we limit the
    sample to CPS respondents who responded to the January,
    February, and March basic CPS
  – Because we will be matching CPS microdata to the UI wage
    records, we limit the sample to CPS respondents with a PIK
  – Propensity score methods used to adjust the weights for both of
    these sample restrictions

• Finally, need to adjust CPS employment series for
  population adjustments
         Data: Analysis Sample

Constructing CPS quarterly employment records
• Individuals who report any wage and salary job over
  the quarter are categorized as employed in CPS
• Information on multiple jobs held simultaneously and
  job changes used to categorize people as holding one
  job or two plus jobs in CPS
   – Most certain a job change has occurred if question asked
     and answered directly, but not always asked
   – Multiple job question asked every month, but class of
     second job asked only in outgoing rotation group
   – Results displayed for more restrictive of two criteria
             Empirical Results

Overall %          Individual holds wage & salary
Row %              job in UI wage records
Column %
                         Yes             No

Individual              52.5%           11.5%
holds        Yes        82.1%           17.9%
wage &                  94.4%           25.9%
salary job               3.1%           32.9%
in CPS       No          8.7%           91.3%
                         5.6%           74.1%
             Empirical Results

Overall %            Individual holds wage & salary
Row %                job in UI wage records
Column %
                        Yes: 1 job    Yes: 2+ jobs

Individual                83.9%            7.8%
holds                     91.5%            8.5%
             1 job
wage &                    95.7%           63.5%
salary job   Yes:          3.7%            4.5%
in CPS       2+           45.4%           54.6%
             jobs          4.3%           36.5%
             Empirical Results

• Previous two tables show:
  – Substantial discrepancies in employment status:
     • 17.9% of CPS workers not working in UI
     • 5.6% of UI workers not working in CPS
  – Even larger discrepancies for multiple jobholder
    status (conditional on working in both datasets):
     • 45.4% of MJH workers in CPS not MJH in UI
     • 63.5% of MJH workers in UI not MJH in CPS
  – These discrepancies are large enough to be
    relevant for the 1-2% topside employment trend
    discrepancy between the CES and CPS
          Summary & Next Steps
• The results presented are preliminary and should be
  considered work in progress. We have found:
   – Substantial discrepancies in employment status and the number
     of jobs in a quarter when matching CPS and UI wage records for
     same people
   – We have modeled these discrepancies as “off the books”
     nonstandard employment measured in the CPS but not in the UI
     wage records {X2}, and marginal short duration jobs measured in
     the UI wage records but not in the CPS {X3}
   – The off-diagonals are related to characteristics of workers and
     jobs in ways that we believe make sense (not presented here)
• Our next step is to look at whether the discrepancies
  move in such a way that they could account for CPS
  versus CES discrepancies (work in progress)

To top