Barbara M' Fraumeni

1
Muskie School of Public Service
Ph.D. Program in Public Policy
Implementation with a Categorical Approach
Barbara M. Fraumeni Muskie School of Public
Service, USM, Portland, ME the National Bureau
of Economic Research, USA China Center for Human
Capital and Labor Market Research Central
University of Finance and Economics Bejing, China
June 19, 2009
2
Restarting an Old Project
Muskie School of Public Service
Ph.D. Program in Public Policy

• Primary project objective is to streamline
• But also to
• Update Jorgenson-Fraumeni estimates
• Put the project in production mode

3
Why Streamline?
Muskie School of Public Service
Ph.D. Program in Public Policy

• To substantially reduce the time and data needed
  to construct Jorgenson-Fraumeni (J-F) human
  capital accounts
• To facilitate use of commonly available data sets
• To increase the likelihood that more countries
  will construct J-F accounts
• Currently Australia, Canada, New Zealand, Norway,
  United Kingdom, United States, and Sweden have
  such accounts

4
Muskie School of Public Service
Ph.D. Program in Public Policy

• Comparable,
• multi-country data sets
• are the ultimate goal

5
Human Capital From Indicators Indexes to
Accounts (Fraumeni 2008)Data Starting Points
Muskie School of Public Service
Ph.D. Program in Public Policy

• OECDs Education at a Glance
• Lisbon Councils Human Capital Index
• EU and WORLD KLEMS
• Cao, Ho, Jorgenson, Ruoen, Ximing

6
Human Capital From Indicators Indexes to
Accounts (Fraumeni 2008)OECD
Muskie School of Public Service
Ph.D. Program in Public Policy

• 30 member countries 6 partner countries
• Many types of indicators
• 500 pages, 150 tables
• Educational attainment
• Types relative levels of expenditures
• Relative earnings by educational attainment
• Vocational programs
• Years in school
• Underlying data is available

7
Human Capital From Indicators Indexes to
Accounts (Fraumeni 2008)Lisbon Council
Muskie School of Public Service
Ph.D. Program in Public Policy

• Exists for 13 EU countries and 12 Central
  Eastern European states
• Human Capital Index as a source of perhaps useful
  data
• Expenditures on formal education
• Opportunity cost of
• Parental education
• Adult education
• Learning on the job

8
Human Capital From Indicators Indexes to
Accounts (Fraumeni 2008)Lisbon Council
Muskie School of Public Service
Ph.D. Program in Public Policy

• Demography employment measure estimates the
  number of people who will be employed in 2030 by
  looking at economic, demographic, and migratory
  trends
• Unknown if Lisbon Council data would be made
  available

9
Human Capital From Indicators Indexes to
Accounts (Fraumeni 2008)KLEMS
Muskie School of Public Service
Ph.D. Program in Public Policy

• Probably a source of categorical labor data on
• Hours worked
• Employees
• Wages
• Exist for about half of the EU countries
• Other countries, such as
• U.S.
• Australia
• Canada

10
China KLEX (CHJRX, forthcoming)
Muskie School of Public Service
Ph.D. Program in Public Policy

• To be published in the June Review of Income and
  Wealth
• Contact Yue Ximing of Renmin University
• Productivity by industry and across all
  industries
• Has all of the required labor data, however by
  broad age categories
• 16-34
• 35-54
• Over 54

11
Desirable Properties of an Account
Muskie School of Public Service
Ph.D. Program in Public Policy

• Two of three volumes, values and prices
• Common numeraire, such as a base currency
• Related to a generally accepted macro measure
• Include inputs and outputs
• Complete system, internally externally

12
Desirable Properties of an Account
Muskie School of Public Service
Ph.D. Program in Public Policy

• Short-run
• First Three Objectives
• Longer-run
• Last Two Objectives

13
Starting Points
Muskie School of Public Service
Ph.D. Program in Public Policy

• Construct human capital investment stocks
• Formal education
• Later
• Informal vocational education
• Training
• Adult education

14
Starting PointsDemographic Information
Muskie School of Public Service
Ph.D. Program in Public Policy

• Population by age, sex, highest education level
  attained, and labor force participation
• Population births immigrants emigrants
  deaths
• Education at a Glance categories
• Single year of age information when possible for
  all individuals lt25
• Forward-looking survival rates

15
Starting PointsEducation
Muskie School of Public Service
Ph.D. Program in Public Policy

• Enrollment
• Given categorical data, could be imputed by using
  a typical enrollment sequence
• Similar to what the Canadians have done
• Issues
• All students might not start school at the same
  age
• Drop-outs
• Failure to advance
• Skipping grades
• Special education
• Students going part-time
• Advanced degrees

16
Starting PointsTime Use
Muskie School of Public Service
Ph.D. Program in Public Policy

• Minimum for broad categories
• Work
• School (homework time?)
• Maintenance
• Sleep
• Leisure?
• Second priority - child care
• Third priority
• Care of others
• Health improvement/maintenance
• Complete time use accounts

17
Valuation of Time
Muskie School of Public Service
Ph.D. Program in Public Policy

• Market
• Wage rate
• After-tax?
• Nonmarket - ?

18
Valuation of TimeNonmarket
Muskie School of Public Service
Ph.D. Program in Public Policy

• Opportunity Cost
• Marginal value of time?
• Replacement cost
• Generalist
• Specialist
• Quality-adjusted specialist?

19
Valuation of TimeNonmarket
Muskie School of Public Service
Ph.D. Program in Public Policy

• Zero value time?
• Ignoring consumption value of primarily human
  capital investment activities?

20
Fraumeni Simplified MethodCan Categorical
Estimation Work?
Muskie School of Public Service
Ph.D. Program in Public Policy
21
Jorgenson-Fraumeni Data Required for Lifetime
Income
Muskie School of Public Service
Ph.D. Program in Public Policy

• By sex, individual years of age, and education
  (either highest level attained by individual
  level or enrollment by individual level)
• Population
• Enrollment
• Employment
• Labor compensation
• Hours of market work
• Survival rates
• All but survival rates are contemporary
  information

22
Equation Notation(categorical notation in
capital letters)
Muskie School of Public Service
Ph.D. Program in Public Policy

• mi(s,a,E) lifetime market income
• nmi(s,a,E) lifetime nonmarket income
• ymi(s,A,E) yearly (current) market income
• ynmi(s,A,E) yearly (current) nonmarket income
• g real rate of growth in labor income
• r discount rate
• sr(s,A) survival rate to one year older
• s sex
• a(A) age, by single year of age, e.g., age 0, 1,
  2, ...74, 75
• e(E) highest level of education attained, by
  individual level of education from grade 1, 2,
  ..., through at least one year of graduate school
• older(OLDER) age 1, e.g., being one year older

23
Five Stages
Muskie School of Public Service
Ph.D. Program in Public Policy

• Stage 1 No school or work, ages 0-5
• Stage 2 School, but no work, ages 6-15
• Stage 3 School and work, ages 16-34
• Stage 4 Work only, ages 35-74
• Stage 5 Retirement, zero income

24
Equations for Ages 0-5
Muskie School of Public Service
Ph.D. Program in Public Policy

• mi(s,a,E) sr(s,OLDER) mi(s,older,E)
  (1g)/(1r)
• nmi(s,a,E) sr(s,OLDER) nmi(s,older,E)
  (1g)/(1r)
• Where E1

25
Equations for Ages 0-5 Question
Muskie School of Public Service
Ph.D. Program in Public Policy

• Survival rate from birth to age 1 is normally
  lower than any other year

26
Equations for Ages 35-74
Muskie School of Public Service
Ph.D. Program in Public Policy

• mi(s,a,E) ymi(s,A,E) sr(s,OLDER)
  mi(s,older,E) (1g)/(1r)
• nmi(s,a,E) ynmi(s,A,E) sr(s,OLDER)
  nmi(s,older,E) (1g)/(1r)

27
What Is Yearly Nonmarket Income?
Muskie School of Public Service
Ph.D. Program in Public Policy

• Assumed 10 hours per day is spent sleeping and in
  maintenance
• Assumed students spend 1600 hours per year in
  formal education activities
• 14 hours per day available to go to school,
  perform market work and engage in nonmarket
  activities, including nonmarket work and leisure
• Opportunity cost methodology is used to value
  nonmarket activities

28
Equations for Ages 35-74Backwards Recursive
Muskie School of Public Service
Ph.D. Program in Public Policy

• Using future known information creates a problem
  when the future is beyond the data availability
  period
• However, a cohort approach can introduce business
  cycle effects

29
Equations for Ages 6-34
Muskie School of Public Service
Ph.D. Program in Public Policy

• mi(s,a,E) ymi(s,A,E) sr (s,OLDER)
  senr(s,A,ENR) mi(s,older,E1)
• (1 - senr(s,A,ENR)) mi(s,older,E)
• (1g)/(1r)
• nmi(s,a,E) ynmi(s,A,E) sr (s,OLDER)
  senr(s,A,ENR) nmi(s,older,E1)
• (1 - senr(s,A,ENR)) nmi(s,older,E)
• (1g)/(1r)

30
Equations for Ages 6-34
Muskie School of Public Service
Ph.D. Program in Public Policy

• Individuals do not move from to a higher age
  category to a higher education category
  necessarily at the same time
• Tracking different number of years of discounting
  before lifetime income of higher education
  category is realized
• Tracking different number of years of using a
  categorical survival rate

31
Jorgenson-FraumeniVolumes
Muskie School of Public Service
Ph.D. Program in Public Policy

• Translog indexes, where the weights are shares of
  lifetime income or investment
• Volumes in the index are population or numbers of
  persons enrolled in school
• For example
• Lifetime income volume (year,s,a,e)
• .5 share(year,s,a,e) share(year-1,s,a,e)
• lnpopulation(year,s,a,e)
  lnpopulation(year-1,s,a,e)

32
Streamlining Issues
Muskie School of Public Service
Ph.D. Program in Public Policy

• Population needs to be imputed by single year of
  age population(s,a,E), during estimation
• Enrollment questions in general
• In what grade are individuals of certain ages
  enrolled?
• Drop-outs who return to school
• Undergraduate and graduate education

33
With Only Categorical Data
Muskie School of Public Service
Ph.D. Program in Public Policy

• Deriving population by individual year of age is
  critical
• B(s) is the number of persons born (of age 0)
• Pop(s,1,1) is categorical population for age
  category 1 (ages 0-5) and education category 1
  (grade 8 or less completed)
• Population(s,a,1) is population by single year of
  age for education category 1 (grade 8 or less
  completed)
• Sr(s,1) is the average one-year rate of survival
  of individuals in age category 1 (ages 0-5) t
• B(s) ? age 0 population(s,0,1)
• Sr(s,1)B(s) ? age 1 population(s,1,1)
• Sr(s,1)2B(s) ? age 2 population(s,2,1)
• Sr(s,1)3B(s) ? age 3 population(s,3,1)
• Sr(s,1)4B(s) ? age 4 population(s,4,1)
• Sr(s,1)5B(s) ? age 5 population(s,5,1)
• Pop(s,1,1) S (n0 to 5) sr(s,1)n B(s) to be
  controlled

34
With Only Categorical DataEnrollment
Muskie School of Public Service
Ph.D. Program in Public Policy

• Population dating implications for enrollment
• Drop-outs during the school year as opposed to at
  the end of the school year

35
With Only Categorical Data
Muskie School of Public Service
Ph.D. Program in Public Policy

• General enrollment imputation questions
• Are essentially all individuals enrolled at
  certain ages?
• Are there certain points when not-enrolled
  changes?
• Can then by assumption allocate students to
  certain grades (or not enrolled)
• One grade or two grades choice
• Controlling
• Drop-outs never return?
• Certain points when not enrolled changes?

36
Recent Papers
Muskie School of Public Service
Ph.D. Program in Public Policy

• Wei (2008 2 papers) for Australia, every 10
  years 1981-2001
• Market only
• Ages 18-65
• Includes work experience investment
• Linear combination of lifetime labor incomes of
  older cohorts between Census years
• Greaker and Liu (2008) for Norway, 2006
• Market only
• Ages 15-67

37
Recent Papers
Muskie School of Public Service
Ph.D. Program in Public Policy

• Gu and Wong(2009) for Canada, 1970-2007
• Market only
• Ages 15-74
• Christian (2009) for U.S., 1994-2006
• Both market and nonmarket
• Ages 0-80

38
Categorical Issues in Recent Papers
Muskie School of Public Service
Ph.D. Program in Public Policy

• Wei (2008) for Australia, every 10 years 1981-
  2001
• Categories of educational qualifications
• Single year of age information
• Greaker and Liu (2008)
• Enrollment by individual year of age, but by
  grade level categories
• Labor force highest level of educational
  attainment by individual year of age by the same
  grade level categories
• Data on left years to complete an education
  category determines number of years of
  discounting
• Data on enrollment by individuals of any age

39
Categorical Issues in Recent Papers
Muskie School of Public Service
Ph.D. Program in Public Policy

• Gu and Wong (2009)
• Population is by sex, single year of age, and by
  category of highest educational level completed
• Enrollment by sex, single year of age (15-on),
  enrollment grade assumed
• Christian (2009)
• School age population (0-34) is by sex, single
  year of age and by individual grade enrolled from
  1994-on
• Before 1993, population by individual number of
  years of education completed
• From 1993-on, population by degrees or
  certifications achieved, detail imputed using
  enrollment data
• Older population (35-80) is by sex, individual
  year of age, and category of highest education
  level attained

40
Christian (2009) Highest Level Attained
Imputation
Muskie School of Public Service
Ph.D. Program in Public Policy

• Uses historical enrollment data to impute highest
  individual grade level of education attained by
  individual grade level
• Not enrolled characteristics for those under 18
  imputed from characteristics of those enrolled
• Highest education level completed is current
  enrolled grade level minus one for those enrolled
  in high school or below

41
Christian (2009) Highest Level Attained
Imputation
Muskie School of Public Service
Ph.D. Program in Public Policy

• Those over 17 who are in the zero highest level
  of education attained category are assumed to
  remain in that category
• Highest education level attained by individual
  grade levels for those over 17 who are in
  enrolled in college at the undergraduate level
  imputed from historical trends, then smoothed
  using a quadratic formula
• Smoothing for those who highest level attained is
  undergraduate college
• 3-year moving average
• Smoothed enrollment rates capped at 95

42
Christian (2009) Highest Level Attained
Imputationfor Those in Graduate School
Muskie School of Public Service
Ph.D. Program in Public Policy

• Those with graduate degrees are ignored
• Smoothing across age within years using a moving
  average across three consecutive ages
• Smoothing done a total of 3 times, including
  after bounding and other adjustments

43
ChristianAging Investment in Education
Muskie School of Public Service
Ph.D. Program in Public Policy

• A 17 year old in order to complete another year
  of education (say the 12th grade) must also age
  by one year to age 18
• Alternatively, one can posit what the lifetime
  income difference would be for an 18 year old who
  has completed the 12th grade and one who has not

44
Christian - Dropouts Investment in Education
Muskie School of Public Service
Ph.D. Program in Public Policy

• Drop-outs do return with some probability for
  which there might be data
• True particularly for college
• If this is not taken into consideration, as it
  raises future lifetime income, then investment in
  education is larger than it would be otherwise
• But assuming the maximum number of individuals
  continue on may counteract this effect

45
Christian Imputation Methodology???
Muskie School of Public Service
Ph.D. Program in Public Policy

• Christian

46
Migration
Muskie School of Public Service
Ph.D. Program in Public Policy

• J-F Not dealt with
• Gu and Wang Not dealt with
• Christian Imputed, but really just a residual

47
Muskie School of Public Service
Ph.D. Program in Public Policy

• Institutional population not dealt with
• Military was, but not in updates

48
Illustrativeand Preliminary Numbers
Muskie School of Public Service
Ph.D. Program in Public Policy
49
Ph.D. Program in Public Policy
Muskie School of Public Service
50
Ph.D. Program in Public Policy
Muskie School of Public Service
51
Ph.D. Program in Public Policy
Muskie School of Public Service
52
Ph.D. Program in Public Policy
Muskie School of Public Service
53
Yearly Nonmarket IncomePer Capita Equation
Muskie School of Public Service
Ph.D. Program in Public Policy

• ynmi(s,a,E)14752-hours(s,A,E)/pop(s,A,E)-1300
  senr(s,A,E) cmp(s,A,E) (1 tax)(1-taxam)

54
Ph.D. Program in Public Policy
Muskie School of Public Service
55
Ph.D. Program in Public Policy
Muskie School of Public Service
56
Ph.D. Program in Public Policy
Muskie School of Public Service
57
Ph.D. Program in Public Policy
Muskie School of Public Service
58
Ph.D. Program in Public Policy
Muskie School of Public Service
59
Ph.D. Program in Public Policy
Muskie School of Public Service
60

Muskie School of Public Service
Ph.D. Program in Public Policy
61
Muskie School of Public Service
Ph.D. Program in Public Policy
62
Acknowledgements
Muskie School of Public Service
Ph.D. Program in Public Policy

• Mun Ho and other researchers associated with Dale
  Jorgenson
• For the labor data base with categorical labor
  compensation, hours, and employees
• Michael Christian of the University of Wisconsin
• For 1990-2003 population (and enrollment rates)
  by single year of age and by individual level of
  highest educational level attained (individual
  level of enrollment)
• Sarah Kopack, my Muskie School 10 hour per week
  Graduate Assistant
• For pulling through when needed in spite of
  essentially not knowing how to use Excel as of
  early September

63
Christian (2009) results
Muskie School of Public Service
Ph.D. Program in Public Policy

• Gives a sense of what J-F new estimates might be
• Total human capital stock
• 1986 J-F (1992) 286 trillion
• 1994 Christian (2009) about 425 trillion
• 2006 Christian is 738 trillion
• 2006 Christian size comparisons
• Total is over 50 times GDP
• Total is over 15 times nonhuman capital stock
• Market is over 15 times GDP

64
Christian (2009) results
Muskie School of Public Service
Ph.D. Program in Public Policy

• Market comparisons
• 2006 Christian 212 trillion
• 2001 Fraumeni 117 trillion
• Starts at age 16
• Individuals never go to school in the future
• Nonmarket shares for both Christian and J-F hover
  around 70

65
Christian (2009) results
Muskie School of Public Service
Ph.D. Program in Public Policy

• Gives a sense of what J-F new estimates might be
• Total human capital stock
• 1986 J-F (1992) 286 trillion
• 1994 Christian (2009) about 425 trillion
• Nonmarket shares for both Christian and J-F hover
  around 70

66
Christian (2009) results
Muskie School of Public Service
Ph.D. Program in Public Policy

• Rates of growth of total human capital stock
  volumes
• 1948-1986 J-F 1.6
• 1994-2006 Christian 1.1
• Population growth differences
• Nonhuman capital growing much more rapidly
• 1948-1986 3.6
• 1994-2006 2.6

67
Gu and Wong (2009) Results1970-2007
Muskie School of Public Service
Ph.D. Program in Public Policy

• Rates of growth of volumes
• Human stock 1.7, nonhuman stock 2.8
• Human investment .4, nonhuman investment 3.9
• Contribution to growth in human capital stock
  volume
• 1.5 due to growth in working-age population
• .2 is due to rising educational levels
• After early 80s
• Education contributed .8
• Aging reduced stock growth by .5

68
Gu and Wong (2009) Results1970-2007
Muskie School of Public Service
Ph.D. Program in Public Policy

• Ratio of human capital stock and investment to
  nonhuman capital stock and investment declines
  over time
• Stock 5.6X in 1970, 4X in 2007
• Investment 5.6X in 1971, 2X in 2007
• Human capital stock relative to GDP
• 16 times GDP in 1970
• 11 times GDP in 2007

69
Greaker and Liu (2008) Results2006
Muskie School of Public Service
Ph.D. Program in Public Policy

• Human capital stock is 8 times GDP in 2006
• Lots of attention paid to comparisons by
  educational level and by gender

70
Initial Steps Towards Human Capital Accounts for
all Countries
Muskie School of Public Service
Ph.D. Program in Public Policy

• J-F market only
• Formal education

71
Start With
Muskie School of Public Service
Ph.D. Program in Public Policy

• Turin joint Fondazione Giovanni Agnelli /OECD
  workshop in November 2008
• Researcher from Norway, Gang Liu, will be working
  at OECD for 10 months as of Oct. 1 to work on
  human capital accounts
• Proposal to OECD Committee meeting in June
• Perhaps a discussion in connection with a
  European Foundation Centres Governing Council
  and Management Committee meeting
• China Center for Human Capital and Labor Market
  Research project Great news!

72
Longer-term Objectives
Muskie School of Public Service
Ph.D. Program in Public Policy

• Inputs and outputs
• Parents, teachers, schools, books and other
  educational materials, social capital
• Quality adjusted
• Students (covered)
• Full nonmarket account
• Investment in health
• Educating, nurturing, caring for others
• Embodied within an experimental set of GDP
  accounts

73
Longer-term ObjectivesOther than Formal Education
Muskie School of Public Service
Ph.D. Program in Public Policy

• Early childhood education
• Vocational training
• On- and off-the-job training
• Adult education
• Informal education

74
Longer-term ObjectivesVery Long-term
Muskie School of Public Service
Ph.D. Program in Public Policy

• Mobility
• Opportunity

75
Timetable Proposed in Turin
Muskie School of Public Service
Ph.D. Program in Public Policy

• J-F for comparability across countries
• J-F for 5 additional countries within 18 months
• Gather again in Turin
• J-F market accounts for most, if not all, OECD
  countries within 5 years