Click for next page ( 194


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 193
APPENDIX Examples of Retirement-Income-Related Projection Models This appendix describes examples of three types of retirement income models: microsimulation models, cell-based models, and intergenerational models. It lists the major references for information about the model. MICROSIMULATION MODELS CORSIM (Caldwell, 1993; Caldwell et al., 1993; see also LOGISIL, 1994~: Dynamic microsimulation model of people and households; projects life histories for people of all ages, year by year; first version developed in 1986-1988 as adaptation of DYNASIM2 (see below); rebuilt in 1989-1993; written in C lan- guage; operates on desktop personal computers; being rewritten to also run on parallel-processing supercomputers. Processes modeled include fertility, immi- gration, mortality; first marriage, remarriage, divorce, custody of children, leav- ing home, education level; weeks worked and earnings; employment-related trans- fer income, welfare-related transfer income, pension-related transfer income; Social Security payroll taxes, federal and state income taxes, property taxes, estate taxes; family earned income, family transfer income, family asset income; consumption, savings; home ownership, market value, mortgage debt; ownership of other assets, market value, debt; asset changes from savings, asset changes from appreciation, asset transfers at death, asset transfers at divorce, income from assets; smoking, alcohol, diabetes. Being expanded and enhanced by the Cana- dian government as DYNACAN; additions include simulation of the Canadian public and private pension systems; enhancements include making it possible to compare a baseline program and a policy alternative in one instead of two runs 193

OCR for page 193
94 ASSESSING POLICIES FOR RETIREMENT INCOME and more flexibility in specifying output for policy analysis and validation pur poses. DYNASIM2 (Dynamic Simulation of Income Model) (Johnson and Zedlewski, 1982; Johnson et al., 1983; Zedlewski, 1990~: Dynamic microsimulation model of people and households; projects life histories for people of all ages, year by year; first version completed in 1975; redesigned version completed in the early 1980s with elements of original DYNASIM, the PENSIM model developed by James Schulz to simulate private pension alternatives, and other features for analyzing retirement-income-related policy issues; written in FORTRAN; oper- ates on mainframes and minicomputers; recently rewritten to operate on desktop personal computers. Processes modeled include death, birth, marriage, divorce, disability, leaving home, education level, migration; labor force participation, annual hours of participation, hourly wage, whether unemployed, proportion of labor force hours unemployed; job change, industry, pension coverage, plan par- ticipation; pension eligibility, type of plan, benefit formula, plan constants, ben- efit computation; Social Security retirement benefit eligibility, retirement benefit computation, disability benefit, spouse benefit, children's benefit; participation in Individual Retirement Account (IRA), accumulations, distribution; whether retiring from job, whether accepting new job; Supplemental Security Income (SSI) eligibility, benefits, participation; federal income taxes, Social Security payroll taxes. PIMS (Pension Insurance Management System) (Holmer, 1993~: Dynamic and stochastic microsimulation model of private employers with defined benefit pension plans insured by the Pension Benefit Guaranty Corporation (PBGC); written using the object-oriented capabilities of the C++ language; input data on firms (from Compustat) and sponsored plans (from 5500 forms) organized as a relational database; developed to operate on a personal computer platform. Monte Carlo methods are used to characterize uncertainty about the future course of the economy, industries, and individual firms, as well as uncertainty about future asset returns. The PIMS economy module generates future values for each em- ployer of asset-debt ratios, employment, and market-value equity, and then uses the economy and firm variables to compute a bankruptcy probability using a logit equation estimated with pooled corporate data. The plan module simulates changes in age-service matrices of active plan participants, terminated vested participants, and retired participants. The plan module also simulates each firm's annual contributions to sponsored plans, plan benefits, and several measures of plan liabilities. The PBGC module simulates risk-based pension insurance pre- mium income and expenses involved in pension insurance claims resulting from the past and future bankruptcy of firms that sponsor one or more underfunded plans.

OCR for page 193
EXAMPLES OF RETIREME - -INCOME-RE~TED PROJECTION MODELS 195 PRISM (Pension and Retirement Income Simulation Model) (Kennel! and Shells, 1986, 1990; see also Rivlin and Wiener, 1988~: Dynamic microsimulation model of people and households; projects life histories for people alive in base year (1979), year by year; developed in 1980 for the President's Commission on Pension Policy; long-term-care module added in 1986; written in FORTRAN; operates on mainframe computers; recently converted to operate on desktop per- sonal computers. Processes modeled include death, birth, marriage, divorce, education level, disability; annual hours of work, hourly wage; job change, indus- try, pension coverage, pension plan assignment; decision to retire and accept pension, decision to retire and accept Social Security; IRA adoption, contribu- tions; employer pension benefit computation; Social Security retirement benefit eligibility, retirement benefit computation, disability benefit, spouse benefit, children's benefit; IRA distribution; SSI eligibility, benefits, participation; fed- eral and state income tax, Social Security payroll tax. The long-term care module uses the basic PRISM model, with some modifications, to project family struc- ture, employment, income, assets, and private health insurance coverage for the elderly; the module simulates disability status of the elderly, their use of and expenditures for nursing home and home care services, and their accumulation and spending down of assets to gain Medicaid eligibility. Treasury Individual Income Tax Simulation Model (OTA Model) (Cilke et al., 1994~: Static microsimulation model of taxpayers and families; written in FOR- TRAN; operates on minicomputers; includes a two-stage static routine to update and project the database for a total of 10 years (first stage applies growth rates on each dollar amount to reflect actual and projected per capita real growth and inflation; second stage adjusts the weights of family heads to hit aggregate targets for different variables). Models federal and state income tax and Social Security payroll tax liabilities under current law and alternatives, based on samples of individual income tax returns filed with the IRS. Wolf et al. Model for Simulating Life Histories of the Elderly [under develop- ment] (Wolf et al., 1995; McNally and Wolf, 1996~: Dynamic microsimulation model of people and households being developed with funding from the National Institute on Aging; will initially develop longitudinal histories for people aged 48 and older in 1988; will focus on kinship networks and functional (disability) status; will project total income for each family but will not, at least initially, identify separate sources of income, such as pensions. Progress has been made in creating the synthetic starting population for the simulation runs and in develop- ing a capability to project kinship availability. Progress is also being made in developing multi-equation models of income dynamics, labor market and retire- ment behavior, and marriage and divorce, with data from the Panel Study of Income Dynamics and a multivariate random-effects specification, and in simu

OCR for page 193
196 ASSESSING POLICIES FOR RETIREMENT INCOME rating disability dynamics on the basis of the Grade of Membership (GoM) model developed by Kenneth Manton, Eric Stallard, and others at Duke University. CELL-BASED MODELS AARP [American Association of Retired Persons] Solvency and Individual Return Model (SIR) (Cohen end Beedon, 1994a, 1994b): Cell-based model of the Social Security system; develops year-by-year projections for 75-year period. Uses the Social Security Office of the Actuary's intermediate demographic and economic assumptions for simulating the effects of changing the tax or benefit formulas for workers who retire at different times and with high, medium, or low wages over their work life; can also simulate the effects of privatization schemes. Outputs include years for workers to recover contributions, ratio of benefits to contributions, and Old-Age and Survivors Insurance and Disability Insurance (OASDI) trust fund reserves as a percent of outgo. Macroeconomic-Demographic Model (MDM) (Anderson, 1984, 1990a, 1990b): System of large cell-based models linked to a macroeconomic growth model; originally developed in 1979 for the President's Commission on Pension Policy and the National Institute on Aging to address interactions of Social Security and the private pension system; subsequently expanded to simulate the effects of population aging on the health care system. System models pertain to population growth, the labor market, pension benefits, family formation, consumer expendi- tures, housing demand, health care expenditures, and health care benefits; macro- economic model includes two goods (investment and consumer goods) and two factor inputs (labor and capital services). Schieber and Shoven Model (Schieber and Shoven,1993~: Cell-based model of the funding status of the private pension system; develops year-by-year projec- tions for 75-year period of assets, benefits, contributions, investment income, net inflow (current and real), and total payroll for defined benefit and defined contri- bution plans, separately for private employer, state and local, and federal plans. Uses the Social Security Office of the Actuary's projections of the population by age, sex, and work force participation for each year; distributes the work force into private, state and local, and federal employment by tenure and pension participation status; accounts for mortality, job leaving, job entrance, and job change; projects employer and employee contributions using the Social Security Office of the Actuary' s assumptions about real wage growth. Social Security Actuarial Cost Model (Board of Trustees [OASDIi, 1996~: Cell-based set of models of the funding status of the Social Security system; develops year-by-year projections for 75-year period of payroll taxes, investment income, and benefits for retirement, survivors, disability, and other components

OCR for page 193
EXAMPLES OF RETIREME - -INCOME-RE~TED PROJECTION MODELS 197 of the Social Security system. Develops projections of the population by age and sex on the basis of assumptions about fertility, mortality, and net immigration; develops projections of labor force participation for men by 5-year age group and marital status and for women by 5-year age group, marital status, and presence of children; projects earnings and payroll tax contributions on the basis of assump- tions about real wage growth; projects benefits on the basis of projected earnings and assumptions about retirement age; projects investment income on the basis of assumptions about interest rates. Projections are usually developed for high-cost, intermediate, and low-cost scenarios. SSASIM (Holmer, 1995a, 1995b): Dynamic and stochastic cell-based model of the Social Security system; written using the object-oriented capabilities of the C++ language; input data on starting values of population and economic vari- ables as well as policy and behavioral assumptions organized as a relational database; developed to operate on a personal computer platform. Monte Carlo methods are used to characterize uncertainty about the future course of thirteen key demographic and economic input variables used in the Social Security actu- arial cost model, as well as uncertainty about future asset returns. The model's logical structure is similar to that of the Social Security actuarial cost model except that numerous economic feedback effects are modeled and that program- related risks are explicitly represented using Monte Carlo methods. Operating in a non-stochastic mode, SSASIM can replicate each of the three scenarios pre- sented in the Trustees Report. The original development of SSASIM was spon- sored by the 1994-95 Social Security Advisory Council to support the evaluation of alternative asset allocation policies for the trust fund; subsequent model en- hancements, which will permit evaluation of individual account reform propos- als, are being sponsored by the Employee Benefit Research Institute. INTERGENERATIONAL MODELS Aaron, Bosworth, and Burtless Model (Aaron et al., 1989~: Computable gen- eral equilibrium model in which labor supply and private saving patterns are based on observed profiles and are assumed to be exogenous. Future labor supply patterns are based on the demographic projections of the Social Security Office of the Actuary. Capital accumulation is derived from the identity that relates saving and investment; an assumed production function determines the relative returns to capital and labor. Model incorporates a detailed characteriza- tion of the U.S. Social Security system, which it was developed to analyze. Auerbach and Kotlikoff Model (Auerbach and Kotlikoff, 1987; Auerbach, Kotlikoff, and Skinner, 1983; Auerbach et al., 1989~: Computable general equi- librium model with three sectors: household (with 75 overlapping 1-year genera- tions), production, and government. For each sector a system of nonlinear equa

OCR for page 193
198 ASSESSING POLICIES FOR RETIREMENT INCOME lions relates endogenous behavioral variables (e.g., consumption or labor supply) to predetermined economic variables and taste and technological parameters. Individual model components are fairly simple, but the interactions are complex. By solving for the economy's general equilibrium transition path, the model simulates the major feedback effects between policy and demographic changes and changes in the time paths of wages, interest rates, labor supply, and the capital stock. Model has been used to address such issues as how much Social Security contribution rates must be increased to maintain current benefit levels; effects on national saving rates and real wages of changing population age structure; effects on international capital flows of changes in saving rates and real wages; effects on overall well-being of people in different generations of economic changes associated with demographic transition; effects on economic performance and generational welfare of reductions in Social Security benefits. Differs from the Aaron, Bosworth, and Burtless model in modeling house- hold saving and labor supply behavior as endogenous, based on an optimizing life-cycle model and the assumption of perfect foresight. Imrohoroglu, Huang, and Sargent Model (Imrohoroglu, Huang, and Sargent, 1994~: Computable general equilibrium model; assumes households want to smooth consumption and insure against lifetime uncertainty but have access to restricted set of assets and risk-sharing arrangements; assumes households are identical when first formed but that luck makes their wealth and consumption diverge as they age; specifies preferences, technologies, information and demo- graphic structures, and government policies in ways that permit rapid computa- tion of decision rules; models the effects of transitions between Social Security policies (e.g., from an unfunded system to a fully funded system). Differs from the Auerbach and Kotlikoff model in allowing households to face uncertainty about preferences, income, and life-span.