Policy
Simulation
Group

Pension Projection Models

Enhancements Introduced in Earlier Versions of PSG Models (since 2004)

Go to enhancements introduced during:

Enhancements introduced during 2011:
(return to top of page)

2011-Dec
No enhancements.

2011-Nov
Complete the development and validation of the federal income tax module. The benchmark runs for the PSG models have the income tax module inactive, but the module can be activated to use baseline parameter values by simply setting the GEMINI RUN.use_inctax input database parameter to T. Note that the income tax module can also be activated in PENSIM using the PENSIM RUN.use_inctax parameter, but that income tax results from PENSIM do not reflect income from social security (as do the GEMINI estimates). The GEMINI income tax estimates are computed using social security benefits even when using the free, educational version of GEMINI, so there is no reason to use PENSIM to estimate federal income taxes. When tabulating tax output statistics, be sure to read the output file documentation, especially the need to apply one-half relative weight to income tax statistics for married individuals. Also, note that baseline assumptions, which are located in the PENSIM input database, call for 2010 federal income tax parameters to be projected forward (either by inflation indexing or at its 2010 level). This is unlikely to be what happens in the next few years, so be sure to consider carefully how you specify federal income tax policy in the years beyond 2010.

Completely revise the income tax chapter (Chapter 7) of the PENSIM Overview so that it now describes the estimation and use of imputation equations to impute other income and itemized deduction amounts, the calibration of income tax filing probability function parameters, the calibration of non-refundable savers credit take-up probability function parameters, and the validation tests conducted on the federal income tax module.

Add several new parameters and rename several other parameters in the PENSIM ITEITC table to support calculation of the earned income tax credit when the income tax filing unit receives investment income in addition to earnings. See the revised PENSIM ITEITC table documentation for details.

Add three new parameters to the PENSIM ITBASE table to support the calculation of limits on income-tax itemized deductions. See the revised PENSIM ITBASE table documentation for details.

Add one new parameter to the PENSIM ITAMT table to support the calculation of taxable income under the alternative minimum tax when the taxpayer has itemized medical expenses. See the revised PENSIM ITAMT table documentation for details.

Revise filing probability parameters for those who are EITC eligible in the PENSIM ITNOF table. See the revised PENSIM ITNOF table documentation for details.

Add several new statistics to supplement 4 of the GEMINI .xs? output results files. These new statistics provide information on the size of various imputed income amounts used in the calculation of federal income tax liability. See the revised documentation of the GEMINI .atx output file for details. Be sure to note the requirement for using one-half relative weights when tabulating federal income tax statistics for couples (with statistics for single individuals getting the normal relative weight of one).

2011-Oct
Revise logic that tabulates the statistics on recently-purchased annuities for supplement 5 of the PENSIM .xsa, .xsb, .xsc, .xsd, and .xse output files so that annuities inherited from a deceased spouse are ignored.

Add three new parameters to the GEMINI RUN table that enable the writing and reading of a binary (.iti) file that ensures imputed income tax amounts are the same for individuals across two (or more) simulation runs. Read the revised documentation of the RUN table in the GEMINI input database for details.

Add a new PENSIM table, SCH_IMM, that contains foreign-born schooling parameter offsets that were previously located in the SCH_DYN table. Make the SCH and SCH_IMM tables be in the scope of the run specification language used in the RSF Toolkit. Add documentation for the new SCH_IMM table and add documentation for the other six PENSIM input database tables that contain parameters concerning the timing and nature of schooling events for native-born individuals.

Add and document 25 new PENSIM input database tables containing parameters of federal income-tax-module imputation equations for (a) seven types of income that are not simulated in the PSG models but are subject to federal income taxation, and (b) total itemized Schedule A deduction amounts and their distribution across three Schedule A deduction categories. The 25 new tables include the IMPUTE table, which is a child of the INCTAX table, and 24 child tables of IMPUTE. Refer to the revised documentation of the PENSIM input database for details. The parameters in the new tables are read and amounts for each of the other types of income and each of the itemized Schedule A deduction amounts are imputed, but these imputed amounts are not yet used in the income tax calculations.

2011-Sep
Add to SSASIM a new output file that contains aggregate results for each simulated birth cohort at every age. Details on the new output statistics are in the documentation for the new SSASIM .olc output file.

Add to PENSIM Overview a new chapter that describes the logical structure of the federal income tax module and the results of several validation tests of the income tax module.

Upgrade SQLite database software from version 3.7.4 (dated December 8, 2010) to version 3.7.7.1 (dated June 28, 2011). This enhancement will not cause any changes in results produced by the PSG models.

Upgrade the aws utility used by the AWS Toolkit from version 1.74 (dated December 9, 2010) to version 1.75 (dated July 15, 2011). This enhancement will not cause any changes in results produced by the PSG models.

2011-Aug
No enhancements.

2011-Jul
Enhance PENSIM so that each of the seven annual mutual-fund management fees can differ depending on whether the fund is held in a job-based pension account or in an individual rollover account. In prior versions of PENSIM, each fund had one fee that was charged no matter where the fund was held. This enhancement involves replacing the seven old fee parameters with fourteen new fee parameters. For details see the revised documentation of the ASSETS table in the PENSIM input database. The baseline assumptions are unchanged, so this enhancement does not cause any changes in benchmark run results.

Add a new PENSIM input parameter that can be used to prevent individuals from ever becoming eligible for traditional defined-benefit pension plans (which includes all DB plans except for cash-balance plans). For details on this new PENCHAR.no_tdb_elg parameter, read the revised documentation of the PENSIM input database. The new parameter has a baseline value of false, so this enhancement does not cause any changes in benchmark run results.

Add a new PENSIM input parameter, PAWITH1.arawd_prob, that specifies the decimal probability that a disabled individual will seek to access his/her rollover account when becoming disabled, rather than waiting until age 62. Accessing the rollover account may involve buying an annuity or making annual withdrawals depending on the assumed values of other parameters in the PAWITH1 table. In past versions of PENSIM the implicit PAWITH1.arawd parameter was hard-coded to a value of one (that is, disabled individuals always sought to access their rollover account). The new explicit parameter has a value of one in all benchmark runs before 2011 and has a value of 0.67 in the 2011 benchmark runs. This change in assumption has been incorporated so that the federal income tax revenue generated in 2004 by the taxation of social security DI benefits better matches the actual revenue, and the new assumption is less extreme that the old assumption. Using the new baseline assumption in a simulation of a two percent sample of the 2000 birth cohort produces a 24 percent reduction in pension income derived from DC plans and DB cash-balance plans at age 61. The 24 percent is less than 33 percent because not every individual has a positive balance in his/her rollover account when becoming disabled. Note that this change in the assumed timing of annuity purchases among some of the disabled causes a very small increase in the loading factors that the annuity provider must apply to annuity prices in order to achieve the target 1.08 ratio of the present value of annuity sales revenues and the present value of annuity payment costs. The slightly higher loading factors imply slightly lower annuity payments, which means many of the 2011 benchmark run results are slightly different than in the prior version of the PSG models.

2011-Jun
Add five new benchmark runs to the PSG models (11 for CBA mode, 31 for OLC mode, 51 for ESP mode, 71 for RCS mode, and 91 for ESP+ mode) that use the intermediate-cost assumptions in the 2011 Trustees Report. The new CBA run 11 and OLC run 31 closely replicate the intermediate-cost OASDI projection in the 2011 Trustees Report. The RSF Toolkit and all the analysis toolkits now operate using these new benchmark runs. The old 2010 Trustees Report runs 10, 30, 50, 70, and 90, are unchanged and are still available for use with old run specification files.

The RSF Toolkit Build logic now constructs runs based on the new x1 benchmark runs. In addition, the Build logic has been changed to generate run specification files that set RUN.threads to 8 (instead of 4 in prior versions of the PSG models). Eight threads of execution will require about three gigabytes of memory. If you find that your computer is sluggish when executing an OLC mode run with eight threads, simply change the number of thread to four before executing the run specification file. If you have the memory and at least four CPU cores, then a single-scenario OLC run should take about three minutes to execute when using eight threads.

Revised PENSIM Overview to include a discussion of the new calibration of the DC rollover probability logit equation, and to include new pension validation results for DC contribution rates and DC account balances. Read Overview section C.19 for details on the new rollover parameters, and sections 10.1.2 and 10.1.3 for details on the new validation results. The new validation results confirm earlier validation results that suggest PENSIM is simulating realistic defined-contribution pension behavior.

Revise income tax filing logic and revise/add income-tax-related statistics to supplement 5 in the PENSIM .xsa, .xsb, .xsc, .xsd, and .xse output files. The revised documentation makes it clear that all the income tax liability and credit amounts should be ignored if the new nonfiler statistic is 1. This can happen for one of two reasons: (a) the sample individual is not in an income tax filing unit because is a dependent on another income tax return, which is indicated by an income tax filing status of -1, or (b) sample individual is in a tax unit but does not file a return because of low income. In this second case, other income tax statistics on tax liability and credits may be nonzero, but those amounts represent what would have happened if the individual had actually filed a return, and therefore, should not be included in any income tax tabulations. Add new PENSIM table ITNOF that contains income tax return filing probability function parameters. Consult the revised documentation of the PENSIM input database for details on the new table and its parameters.

Add a new supplement 4 containing federal income tax statistics to the GEMINI .xsa, .xsb, and .xsc output files. Read the revised documentation of these GEMINI output files for details.

Add several federal income tax statistics to the GEMINI .atx output file, and add a new input parameter, STATS.atx_age, that specifies the age for which these income tax statistics are collected. Read the revised documentation of the .atx output file for the details.

Remove statistics 45 through 52 at the end of the GEMINI .sum output file. These eight statistics were lifetime present value and internal rates of return for account-based pension contributions and benefits. Similar statistics for all pensions are still included in the GEMINI .sum output file.

Improve PENSIM logic that determines which individuals are included in cross-sectional survey samples when the survey supplement is retrospective. This change eliminates the PENSIM XS?.postmortem input parameters. Read the revised documentation for the PENSIM .xsa, .xsb, .xsc, .xsd, and .xse output files for details on this logic change, which generally causes small or modest shifts in aggregate results because the individuals whose sample status change include only those who die, immigrate, or emigrate at about the time the retrospective survey is conducted.

Simplify the GEMINI .adq and .bAA output files by including only total employer-sponsored pension benefits for sample individuals and couples. If you do any post-simulation tabulations of the .adq or .bAA output files, read the revised documentation for those output files to see how the file format has changed. The format changes may require modifications of post-simulation tabulation programs that process the .adq or .bAA output files. The out2bage utility that constructs the .bAA output file from the .adq, .sum, and .cpi output files has been revised to work automatically with the new .adq file format.

2011-May
Add to PENSIM logic that simulates the pension account rollover decision for those who emigrate when on a job where they have a positive pension account balance. In prior versions of PENSIM, individuals in this situation were assumed to cash-out their plan balance on that job. This change in emigration logic, which makes the handling of DC plans congruent with the handling of DB plans, generates a small increase in pension benefits among individuals who have emigrated and are no longer residents.

Add to the PENSIM .jeb output file a new statistic that indicates the reason for a job ending. See the revised documentation of the PENSIM .jeb output file for details on this new statistic.

Add a new PENSIM output results file that contains the vested balance in DC and DB-cash-balance plans at the end of each job held by cohort sample individuals. The new output file has the .jeb (for job end balance) file name extension and its contents are documented in the revised documentation. The baseline PENSIM run does not include this output file, so the new OUTPUT.jeb input parameter must be set to T in a run specification file in order to produce the new output file.

Add new parameters to PENSIM ITEITC table that specify an income tax filing probability function that depends on the size of the EITC for which the the individual is eligible. The shape of the function is suggested by results from Dean Plueger, Earned Income Tax Credit Participation Rate for Tax Year 2005, a paper presented at the 2009 IRS Research Conference. The baseline values of the parameters are set so that PENSIM simulates close to the actual number of income tax returns paid an EITC and the actual aggregate dollar amount of EITC paid in tax years 2004 and 2008. Given this income tax filing calibration, the baseline values of the constant take-up probability for the nonrefundable savers credit have be set so that they produce the approximately one billion dollars of credits paid in 2004 and in 2008.

2011-Apr
Add to the GEMINI .sid output file a new statistic that equals the number of retirement years in which the individual's real retirement income (RRI) is zero. Consult the revised documentation of the GEMINI .sid output file for details.

Add to the RSF Toolkit a new Help menu item that changes the screen size of all the PSG toolkits to one of four sizes (the default size is normal; the three other sizes are: smaller, larger, and largest). Revise the Getting Started with the RSF Toolkit documentation to reflect this enhancement.

Revise saver's credit logic in the PENSIM income tax module so that current year pension contribution are offset by pension benefits and withdrawals from both DB and DC pension plans during the current year and the look-back years. In prior versions the offset amount included only withdrawals from DC pension plans. This change leads to a very small reduction in the number of tax filing units eligible for either a nonrefundable saver's credit under current-law or a refundable saver's credit under a reform. This reduction in eligibility occurs only in low-AGI tax filing units receiving DB pension benefits and making DC pension contributions at the same time.

Change the logic of the out2bage utility so that when it merges statistics from the GEMINI .sum and .adq output files to make a cross-sectional sample at the specified age, it includes only sample individuals who are living residents at the specified age. This change facilitates matching of PENSIM output files with the GEMINI .bAA output file generated by the out2bage utility.

Combine in all PENSIM and GEMINI output files the top two levels of schooling into a single category: four-year college degree or graduate degree. This change affects only output statistics; there is no change to the internal logic of the PSG models.

Remove supplement 2 as an option for the PENSIM .xsa, .xsb, .xsc, .xsd, .xse output files. The earnings and pension statistics formerly in supplement 2 can be found in the GEMINI .bAA output file, and the automatic-enrollment statistic formerly in supplement 2 has always been available in the PENSIM .cov output file.

Add an immigration-age-emigration-age code to the PENSIM .xsa, .xsb, .xsc, .xsd, .xse, and .pro output files. See the revised documentation of those output files for details on how the code is calculated and on the location of the new statistic in each output file. This code has been added to PENSIM output files that do not contain immigration and emigration age in order to facilitate merging these PENSIM files with other PENSIM output file and with GEMINI output files that already include both the immigration age and emigration age statistics. This new code statistic replaces the core job-duration statistic in the PENSIM .xs? output files, so the job-duration statistic has been added to supplements 1, 7, and 8.

Change PENSIM OUTPUT table read logic so that it is now required that the OUTPUT.xs? parameter is set to true (T) in order for the parameters in the XS? table to be read. In prior versions of the model the parameters of the five XS? tables were always read even when the OUTPUT.xs? parameter was false.

Add three new output statistics to supplement 5 of each of the five PENSIM cross-sectional-survey output files (.xsa through .xse). The three new output statistics provide details about the federal income tax's saver's credit. For details see the revised documentation of supplement 5 in the five .xs? output files.

Add a requirement in the cross-section survey tables (XSA through XSE) of the PENSIM input database that the xs_age or olc_moment parameter must have a value of 0.001 when the survey uses a prior-year-retrospective supplement. In the current version of PENSIM, supplements 3 through 5 are retrospective. This change, which may require changes from 0.01 to 0.001 in some run specification files, should have no effect of the output files generated by the PENSIM run. The change has been made in order to be able to implement the postmortem feature described in the next item.

Add a new input parameter to each of the five PENSIM cross-section survey tables (XSA through XSE). The new postmortem parameter effects the number of individuals who are administered the survey only if the survey contains a prior-year retrospective supplement. If the postmortem parameter is true for a survey with a retrospective supplement, then individuals who die a day before the scheduled survey are included in the sample. If the value of postmortem is false, they are not in the sample because they are dead at the time of the survey. The baseline value of postmortem is false, in order to maintain consistency with prior versions of PENSIM. Setting postmortem to true would be useful if it is important to include people in their last year of life in a retrospective supplement sample. For details see the revised documentation of the cross-section survey tables (XSA through XSE). The documentation for the .xsa through .xse PENSIM output files describes which supplements have a prior-year retrospective focus.

Add a new PENSIM table ITCCNOR that contains federal income tax parameters for the nonrefundable child credit, and add income tax module logic that uses these parameters to compute credit amounts. Add a new PENSIM table ITEITC that contains federal income tax parameters for the refundable earned income credit, and add income tax module logic that uses these parameters to compute EITC amounts. Add a new PENSIM table ITCCREF that contains federal income tax parameters for the refundable (i.e., additional) child credit, and add income tax module logic that uses these parameters to compute refundable child credit amounts. Add income tax module logic for the 2006 federal telephone excise tax rebate credit and the 2009-10 making-work-pay credit.

2011-Mar
Add to RSF Toolkit the ability to specify real parameter values in scientific notation. So, for example, it is now possible to specify a reform that eliminates the maximum taxable earnings (aka pop the cap) by setting MTE.mte_amount to 9.9e30 as indicated in the revised documentation of the SSASIM MTE table.

Add GEMINI .sum output file dieage statistic to the .bAA output file created by the out2bage output-merging utility. The new dieage statistic is placed at the end of each individual record so that the layout of the old statistics is unchanged. See the revised documentation for GEMINI outputs files for details on the content of the .bAA output file.

Add a PENSIM input database table for policy parameters and income tax logic that enable the simulation of the nonrefundable saver's credit that is part of current federal income tax law.

Add automatic scaling of the size of all PSG toolkit applications based on each user's computer screen resolution. This eliminates the need for the tkprefs file that was previously used to specify user preferences. The revised Getting Started with the RSF Toolkit document reflects the new situation.

Add additional GCC compiler options under Windows to ensure that none of the .exe files included in the PSG models distribution depend on any dynamic link libraries (.dll files). All the .exe files in the prior version of the PSG models did not work without two .dll files that were not included in the model distribution. This problem, which was a side-effect of switching to a newer version of the GCC C++ compiler, has now been resolved. The Linux version of the PSG models was unaffected by this problem.

Add five new parameters to the PENSIM ITBASE table that specify the four combined-income thresholds used in the federal income taxation of social security benefits and whether or not these thresholds are indexed for inflation. Read the revised documentation of the PENSIM input database for details on these new ITBASE table parameters.

Add six new parameters to the PENSIM ITBASE table that specify the parameters of the standard deduction in the federal income tax. Read the revised documentation of the PENSIM input database for details on these new ITBASE table parameters.

Add six new parameters to the PENSIM ITBASE table that specify the parameters of the personal exemption and its phase-out in the federal income tax. Read the revised documentation of the PENSIM input database for details on these new ITBASE table parameters.

Add parameters to the PENSIM ITSCHX, ITSCHY, and ITSCHZ tables that specify the regular federal income tax rate brackets for Schedule X, Schedule Y-1, and Schedule Z, respectively. Read the revised documentation of the PENSIM input database for details on these new ITBASE table parameters.

Add the arc_immyrs parameter to the OUTPUT table in PENSIM input database. This new parameter, which in all previous versions of PENSIM has been hard-coded to be zero, controls the number of years over which annuity providers attempt to immunize their annuity contract payment liabilities. Read the revised documentation of the PENSIM input database for details on this new OUTPUT table parameter.

Expand the range of valid values for the PENSIM INDBEHX.retdelayrs parameter to include negative values, which cause forced retirement before age 62. Read the revised documentation of the PENSIM input database for details on this revised INDBEHX table parameter and associated behavioral logic.

Revise PENSIM so that RUN.sample_pct can be as large as 10 (that is, a one-in-ten sample) in the ESP and RCS modes of operation.

Upgrade the Windows GNU compiler collection (GCC) used to build the PSG models from GCC 3.4.5 to GCC 4.5.0, which causes some insignificant changes in model output (the value of some statistics change in the fifth or sixth decimal digit) and in model performance (a deterministic OLC run executes about one percent faster).

2011-Feb
No enhancements.

2011-Jan
Add to PENSIM cross-sectional output files (.xsa, .xsb, .xsc, .xsd, .xse) a new supplement 8 that contains a subset of the statistics in supplement 1 plus the nominal dollar balance of the DC plan account. Consult the revised documentation for those five output files for details on the statistics included in the new supplement.

Improve the methods used to compute family pension statistics included in the SCF-like supplement 6 of the PENSIM cross-sectional output files (.xsa, .xsb, .xsc, .xsd, .xse).

Fix a minor bug in the logic used to determine the number of deferred annuity contract units owned by sample members' spouses at death before retirement (which are then bequeathed to the surviving sample member). Fixing this bug causes a very small (about seven-tenths of one percent) increase in the average pension benefit when it is assumed that all employee DC plan contributions are used to buy deferred annuity contracts, and smaller percentage increases as the volume of deferred annuity contract purchases declines. This bug fix does not affect any output statistics unless the PENSIM run calls for some purchasing of deferred annuity contracts.

Add a new parameter to the PENSIM PAWITH1 table that can be used to specify an age for buying an immediate annuity contract that is later than the individual's retirement age. Read the revised documentation of the PENSIM input database for details on the new PAWITH1.acp_buyage parameter.

Add a new parameter to the PENSIM PAWITH1 table that can be used to specify a delay in receiving immediate annuity contract payments until a specified age. This new parameter allows the immediate annuity to be like a longevity insurance product. Read the revised documentation of the PENSIM input database for details on the new PAWITH1.ac_f_p_age parameter, particularly the details on how a positive value of the parameter causes rollover account withdrawals to stop beginning at the ac_f_p_age with any remaining balance used to buy an immediate annuity with no payment delay.

Add to the GEMINI .sid output file a new statistic that is the pension rollover account balance at death when the individual has no surviving spouse (the statistic has a value of zero when the individual has a surviving spouse no matter what the size of the balance). For details read the revised documentation of the GEMINI .sid output file.

Add to the PENSIM .cov output file two new statistics that measure the size of the rollover account balance and the size of traditional DB pension benefits at the first (immediate) annuity contract purchase opportunity (facpo). The new statistics include the balance or benefits of the sample individual's spouse if the sample individual is married at the facpo. For details read the revised documentation of the PENSIM .cov output file.

Add a new parameter to the PENSIM PAWITH1 table that can be used to specify an immediate annuity contract purchase fraction that depends on the ratio of a couple's traditional defined-benefit pension benefits to the couple's rollover account balances. Read the revised PAWITH1 table documentation for details on this new acp_adjr parameter.

Upgrade the SQLite database library from the 3.7.2 to 3.7.4 version and the aws utility used by the AWS Toolkit from the 1.70 to 1.74 version, which will have little visible effect on the operation of the PSG models.

Enhancements introduced during 2010:
(return to top of page)

2010-Dec
Replace the pawd_adje parameter in the PENSIM PAWITH1 table with the pawd_adjf parameter that is associated with a female gender dummy variable. Read the revised PAWITH1 table documentation of the pawd_rule parameter for information about how the new parameter affects the rollover account withdrawal rate.

Add a new parameter to the PENSIM RUN table that makes it easier to specify a run in which no marriages are simulated for cohort sample individuals. See the documentation for the new RUN.noweddings parameter in the revised documentation of PENSIM input parameters.

Fix a bug, which was introduced in the 10/1/10 version of PENSIM, that causes immediate annuities to be purchased out of the rollover account balance every year during retirement instead of just in the first year of retirement. This bug did not have any effect on runs using the baseline PENSIM behavioral assumption of full annuitization or on runs that assume no annuitization. Runs that assume partial annuitization will now produce significantly different PENSIM results. The social security results produced by the PSG models are unchanged by this bug fix.

Revise the baseline behavior in PENSIM related to the timing of first access of the rollover account (for buying an immediate annuity, claiming deferred annuity payments, or making annual withdrawals) by widow(er)s. In prior versions of PENSIM, it was assumed that widow(er)s first accessed the rollover account at age sixty even if they were employed. Beginning with this version of PENSIM, widow(er)s who are aged 60+ do not start to access the rollover account as long as they are employed. This revision has no effect on the social security results produced by the PSG models, but does have very small effects on pension benefit results. For example, in a run that simulates a two percent sample of the 2000 birth cohort, the average pension benefit at age seventy rises by about one-quarter of one percent after this revision in the baseline behavioral assumptions.

Upgrade to the 1.70 version of the aws utility program used by the AWS Toolkit that executes the Linux version of the PSG models on Amazon Web Services computers.

2010-Nov
Add a minimize program to the output analyzer programs included in the PENSIM distribution. The minimize program can be used to find automatically the value of one or more PSG model input parameters that minimizes (or maximizes) some statistic that is tabulated from raw model output. For details, see minimize help by using the Output analyzer program documentation item on the Analyze menu of the RSF Toolkit. The minimize program requires both a start file (containing the starting points of the minimization search) and a Tcl program (that, given a point in the relevant parameter space, computes the value of the function being minimized). Sample start files and Tcl programs are available upon request.

Add and revise PENSIM input parameters in the PAWITH1 table that control the value of the fraction of the rollover account balance that is used to buy an immediate annuity and the fraction of the rollover account balance that is withdrawn each year to support retirement income needs. In both cases the adjustment is made based on the specified values of the parameters and the value of the individual's age and life expectancy (or equivalently the individual's mortality rate) relative to the average for all individuals of the same gender and age. Consult the revised documentation of the PAWITH1 table for details.

Add six new statistics to the PENSIM .cov output file. Three of the new statistics provide information about the fraction of the rollover account balance that is used to purchase an annuity at the individual's first opportunity to buy an immediate life annuity. Two of the new statistics provide information about the price paid for the purchased immediate life annuity and the joint-and-survivor survivor benefit fraction of that annuity. And the sixth new statistic is the fraction of the rollover account balance that is withdrawn to support retirement income needs (rather than to buy an annuity) at the individual's first opportunity to buy an immediate life annuity. See the revised documentation of the PENSIM .cov output file for details on these new statistics.

Remove obsolete ACPRICE table from PENSIM. Add missing PENSIM input database documentation for the existing ACCTFEE table.

Fix a long-standing bug in the immediate annuity cost tabulation logic used to produce annuity provider cost statistics in the PENSIM .arc output file. This bug caused an overestimation of annuity prices (especially for women) in prior versions of PENSIM. After the bug fix, immediate annuity prices are about eight percent lower for women, and about one percent lower for men, in a run that assumes full annuitization and uses the same target .arc revenue-cost ratio as used in a pre-bug-fix run.

Change values of immediate annuity price loads for women and men (PAWITH1.ac_load_f and PAWITH1.ac_load_m) in PENSIM baseline runs so that the annuity provider revenue-to-cost ratios in the PENSIM .arc output file are 1.08 in an ESP mode run with a two percent sample of the 2000 birth cohort using 2010 Trustees Report assumptions. These target revenue-to-cost ratios assume that eight percent of annuity payment costs are used to cover administrative and marketing expenses incurred by the annuity provider and also to provide a return on the reserve capital that the annuity provider must hold to cushion against insolvency. Another way to think about an eight percent revenue-to-cost ratio is that the eight percent is equivalent to the present value of twenty years of an annual fee of 75 basis points when using an annual discount rate of eight percent.

Add a deferred annuity contract as a new asset in PENSIM. The characteristics of the deferred annuity contract (DAC) are specified in the new DACCHAR input database table, which is a child table of the ASSETS table. Deferred annuity contract units are purchased using some fraction of employee contributions to employer-sponsored defined-contribution pension plans, with the age- and gender-specific fraction being specified in the new dac_frac_f and dac_frac_m fields in each of the four ACCTAA? tables. The development work involved in adding this new asset is not yet complete, and therefore, no attempt should be made to use this new feature until completion of the development work is announced.

2010-Oct
Add to the PENSIM ITSCREF table two new parameters that control the refundable savers credit take-up probability. Read the revised documentation of the PENSIM ITSCREF table for details. Also, add two new statistics (related to savers credit take-up probabilities) to supplement 5 of the .xs? output files.

Add to the PENSIM PAWITH1 table four new pension rollover account withdrawal behavior parameters that allow a wider range of rollover account withdrawal behavior to be specified.

Add to the Help menu of the RSF Toolkit access to documentation of Tcl and AWK language commands, which can be used to write post-simulation tabulation programs. Also, add access to documentation of SQL language commands that can be used in a run specification file or a Tcl program via the SQLite capability included in the PSG models distribution.

Add to the AWS Toolkit for the PSG models the option to start AWS c1.xlarge spot instances that are priced at about 25 cents per hour most of the time unless there is heavy demand for AWS computers. This typical spot price compares with the 68 cents per hour for c1.xlarge on-demand instances.

2010-Sep
Add to the PSG model input databases five new runs that use intermediate-cost assumptions from the 2010 Trustees Report: run 10 operating in CBA mode, run 30 operating in OLC mode, run 50 operating in ESP mode, run 70 operating in RCS mode, and run 90 operating in ESP+ mode. In addition to the new Trustees Report assumptions, all five of these x0 runs incorporate more recent historical information so that the last known historical information is for 2008 (instead of 2004 as before). The RSF Toolkit has been changed so that the benchmark runs are these new x0 runs and so that the build of new run specifications uses these new x0 runs as baseline runs. If for some reason you want to continue building new rsf using the old unchanged x9 runs as the baselines, then change toolkit preferences on the Help menu of the RSF Toolkit. To convert old rsf to use the new x0 runs as the baseline runs, simply change the x? to x0 in each create model run same as run x? except: statement.

Re-estimate the inflation-asset-returns VAR(2) model, which is described in the revised Section 1.2.5 of the SSASIM Guide, using four additional years of data through 2009. This new version of the VAR(2) model is used to specify stochastic inflation and asset returns in the asset-returns-10.rsf run specification file, which is included in the PSG model distribution.

Specify both CBA-mode and OLC-mode stochastic runs based on 2010 Trustees Report assumptions and the new inflation-asset-returns VAR(2) model. The specifications for these two new runs are in the stoch10cba.rsf and stoch10olc.rsf files, respectively, which also include a summary of the simulated variability in OASDI actuarial balance and cost rate and how these simulated results compare with stochastic results in the current and past Trustees Reports. The stoch10spec.tcl program, which generates assumption variable means and standard deviations for the two rsf, and the stoch10.awk program, which tabulates raw stochastic output statistics from the runs, are also included in the upgrade package.

Refine PENSIM retirement timing logic in situations where the individual's OAI early retirement age (ERA) is greater than 62 and the individual's age is greater than or equal to the individual's normal retirement age (NRA). This change maintains the spike in retirement at the NRA, which was not maintained in the old logic. Both the new and old logic maintain the larger spike in retirement at the higher ERA. This change has no effect on PSG model results under current-law OASDI policy (which sets ERA to 62), but does cause relatively minor changes in results produced by runs that specify reforms that raise the ERA.

Revise PENSIM so that the last retirement age can be as high as 73 rather than 70 in prior versions of the PSG models. This change causes a different stream of random numbers to be used for some individual behavior in PENSIM, which means that the individual-by-individual results generated by old OLC, ESP, and RCS runs will be different when using this version of the PSG models, but aggregate statistics tabulated for reasonably large samples of individuals should not differ by a statistically significant amount. This change has been made to facilitate analysis of OASDI reforms that raise the the early, normal, or last retirement ages.

Add to RSF Toolkit Help menu a new item that checks whether the version of the PSG models you are using is up-to-date or whether there is a newer version. This check involves a simple Internet query, and therefore, the speed of the check depends on the quality of your Internet connection.

Allow the @ symbol (for which the new run id is substituted) to be included in an embedded SQL statement in a run specification file. Revise the Getting Started with the RSF Toolkit documentation to reflect this enhancement.

Upgrade SQLite database source code and utility programs from the 3.6.22 to the 3.7.2 version. This change should have no effect on the operation of the PSG models.

2010-Aug
Add an OLC-mode linkage between PENSIM employment and SSASIM GDP so that changes in employment simulated in PENSIM (caused either by changes in behavioral assumptions or by changes in pension or social security rules) feed back into the calculation of GDP in SSASIM. This new linkage causes minor differences between the GDP estimates generated by SSASIM operating in the CBA mode and the GDP estimates generated by SSASIM operating in the OLC mode. There has always been a PENSIM-SSASIM linkage with respect to taxable earnings; now there is an analogous linkage for GDP.

Convert documentation of output statistics for all three PSG models from text files to HTML files that can be viewed using your default browser. Rename the HTML files that contain documentation of input parameters for all three PSG models, so that they all can be located in the same directory along with the new output statistics documentation files. Beginning with this version of the PSG models, all the documentation files are located in a single directory (the doc subdirectory of the SSASIM executables directory). These changes permit an enhancement of the process of viewing and keyword searching the documentation. All this detailed input parameter and output statistic documentation is being made available on the PSG website as well as continuing to be part of the PSG models installation.

The RSF Toolkit has been enhanced in several ways. First, the menu items for viewing and searching documentation have been simplified and moved to the Help menu. Second, the left-hand menu has been renamed the File menu to reflect the fact that many types of files (besides .rsf files) can be opened, edited, and saved by the RSF Toolkit editor. Third, the Backup and Exit buttons have been moved to the File menu. Fourth, all the menus can now be pulled down using the Alt key, which implies, for example, that you can now quit the RSF Toolkit using the familiar Alt+F,x keystroke sequence. Fifth, the Find capability on the Edit menu has been simplified and is now accessible via the Ctrl+F key. Sixth, the fast way to execute a single run whose rsf is in the editor has been added to the Execute menu and this action is now invoked by the go key, Ctrl+G, rather than by Ctrl+F which is now used to invoke the Find capability. And finally, a new capability, which has been added at the bottom of the Edit menu, allows direct access to the documentation of the table field named on the current line of the editor via the Alt+D keystroke. For details on this new feature, which is analogous to the Alt+V key (formerly the Right-Mouse-Button action) that shows the base run value of the table field named on the current editor line, read either the Information on Edit menu items at the bottom the the Edit menu or the Capabilities section of the Getting Started with the RSF Toolkit document, which has been thoroughly revised to reflect all these changes.

2010-Jul
Add to the SSASIM .sto output results file two new annual statistics: aggregate taxable earnings and gross domestic product (GDP), both expressed in billions of nominal dollars. The two new statistics are written to the .sto output file only when the OUTPUT.sto_dollar parameter is true, which is not the case unless it is set to true in a run specification file. See the revised documentation of the .sto output file for details. Both these statistics have been available for years in other SSASIM output files, but adding them to the .sto file provides a convenience for those who want to compute cost and income rates that are expressed as a percent of GDP (rather than as a percent of taxable earnings).

2010-Jun
Add to SSASIM input database a new field in the OUTPUT table that controls whether or not dollar amount fields for annual income and cost in the OASDI, OASI, and DI trust funds appear in the SSASIM .sto output results file. For details, see the revised documentation of the .sto output file and the documentation of the new OUTPUT.sto_dollar parameter in the SSASIM input database.

Add to SSASIM input database a new field in the CAPMRKT table that specifies the modified duration of the Treasury bond whose nominal interest rate is simulated by the nir assumption variable. This new input parameter has the same value as has been hard-wired in the source code for many years. Making the duration an input parameter allows it to be changed in calibrations of future Trustees Reports without forcing all prior benchmark run results to change. For details, see the revised documentation of the CAPMRKT table in the SSASIM input database.

Add to PENSIM .xsL output results files a new statistic for supplement 1 that indicates whether or not the most recent employee contribution to a DC plan is limited by either plan rules or by pension laws. For details, see the revised documentation of the PENSIM .xsa, .xsb, .xsc, .xsd, and .xse output results files, or search the PENSIM output file documentation for the keyword limit, using the RSF Toolkit.

Increase the number of file types that can be opened, read, edited, and saved by the RSF Toolkit editor to include results (*.res) files and all files (*.*).

Add to PENSIM input database a new field in the STCONT table that specifies whether or not the presence of automatic-enrollment procedures in a DC savings-thrift plan eliminates whatever plan maximums on employee contributions that would have been imputed by the PENSIM pension characteristics imputation model. For details, see the revised documentation of the STCONT table in the PENSIM input database.

2010-May
No enhancements.

2010-Apr
Add extensive 2008 demographic, economic, and social security data to the SSASIM input database so that RUN.year_zero will be able to be changed from 2004 to 2008 beginning with the benchmark runs that use 2010 Trustees Report assumptions. The new 2008 year_zero data have been tested in a series of runs that use 2009 Trustees Report assumptions.

Add three new fields to the SSASIM PRDMRKT table that together provide the capability of specifying an inflation rate bend point after the kink point during the transition to the ultimate inflation rate. This enhancement permits more accurate simulation of nonlinear movements in the inflation rate, as will be necessary when moving the simulation starting year from 2004 to 2008 in benchmark runs that incorporate assumptions from the 2010 Trustees Report. For details, see the revised documentation of the SSASIM input database via the RSF menu in the RSF Toolkit.

Add three new fields to the SSASIM ESTAB table that together provide the capability of specifying a productivity growth rate bend point after the kink point during the transition to the ultimate productivity growth rate. This enhancement permits more accurate simulation of nonlinear movements in the productivity growth rate, as will be necessary when moving the simulation starting year from 2004 to 2008 in benchmark runs that incorporate assumptions from the 2010 Trustees Report. For details, see the revised documentation of the SSASIM input database via the RSF menu in the RSF Toolkit.

Add to the PSG model distribution a run specification file, which is named stoch09olc.rsf, for a stochastic OLC run that produces results that approximate stochastic results reported in Appendix E of Trustees Reports since 2003. This run 239 is an appropriate base run for conducting stochastic analysis of reform provisions, which is the standard social security policy analysis approach at CBO. Each year the stochastic specification will be updated to incorporate new Trustees Report assumptions. Read the heavily commented run specification file for details on assumptions, OLC results, and a summary of Trustees Report results since 2003.

Revise default assumptions in the GEMINI willingness-to-pay calculator and expand the description of the program, which is available by picking the wtpcalc item from the 'Output analyzer program documentation' option on the Analyze menu of the RSF Toolkit.

2010-Mar
No enhancements.

2010-Feb
Add a new GEMINI output results file that reports for each age the aggregate OASDI benefits received and OASDI payroll taxes paid by all native-born individuals in a birth cohort sample. For details on this new .bta GEMINI output results file, read its documentation using the 'View output file documentation' item on the RSF menu of the RSF Toolkit. Read the revised documentation on the GEMINI STATS input database table for information about how to activate the new output file.

Add a new GEMINI output analyzer program that calculates several statistics that value the aggregate OASDI benefit minus aggregate OASDI payroll tax cashflow for all native-born individuals in a birth cohort sample as reported in the GEMINI .bta output results file. This new wtpcalc (willingness-to-pay calculator) program computes not only conventional money's worth net present value (NPV) and internal rate of return (IRR) statistics for the cohort sample, but also uses modern asset pricing methods to price the cohort's aggregate benefit-minus-tax cashflow. In addition, the wtpcalc analyzer program computes the multiplicative scaling factor that, when applied to the cohort's aggregate payroll tax cashflow, produces a zero price for the benefit-minus-scaled-tax cashflow. A scaling factor less than one implies that at birth an individual in the cohort sample finds the value of OASDI benefits to be less than the value of OASDI payroll taxes, while a scaling factor greater than one implies the value of OASDI benefits to be greater than the value of OASDI payroll taxes. The numerical value of the scaling factor indicates the amount by which payroll taxes would have to be decreased or increased to equate the value of OASDI benefits and taxes. The use of modern asset pricing methods permits the relaxation of two restrictions inherent in conventional money's worth methods: (1) the assumption that future annual discount rates are know with certainty and (2) the assumption that the OASDI benefit-minus-tax cashflow is uncorrelated with individual consumption (that is, the insurance value of the OASDI program is ignored). The modern asset pricing methods nest the conventional money's worth methods in the sense that the modern methods produce the same results as the conventional methods when the extreme assumptions of no future discount rate uncertainty and no OASDI insurance value are both made. For details on the new wtpcalc output analyzer program, read its documentation using the 'Output analyzer program documentation' item on the Analyze menu of the RSF Toolkit. To make these calculations transparent, the C++ source code for the wtpcalc program has been included in the GEMINI installation package.

Add a new GEMINI output results file that reports the present value of nominal adult-equivalent couple earnings and the present value of nominal OASDI benefit minus taxes for each native-born individual in a birth cohort sample, thus providing individual detail on the aggregate statistics reported in the new .bta output results file. For details on this new .bti GEMINI output results file, read its documentation using the 'View output file documentation' item on the RSF menu of the RSF Toolkit. Read the revised documentation on the GEMINI STATS input database table for information about how to activate the new output file.

Upgrade SQLite database source code and dynamic-link library to the current 3.6.22 version.

Add to PENSIM input database a new field in the TRACE table that shows during run execution the (positive) id number of the sample individual who marries a non-sample individual with a specified (negative) id number. Read the revised documentation of the PENSIM input database for more information about the new TRACE.indid4spid field.

Revise PENSIM logic that simulates how the saver's credit that is received by a married couple filing income taxes jointly is split between the two individuals. In prior versions of the model, the whole credit was assigned to the individual who is a member of the cohort sample being simulated. Beginning with this version, the credit is split in proportion to the individual's (net) pension contribution. This more realistic approach requires the development of a new saver's credit account for accumulating the credits assigned to non-sample individuals while married to a sample individual and received before they marry a sample individual, investing the accumulated credits, and withdrawing a fraction of the accumulated credits each year following disability or retirement.

2010-Jan
No enhancements.

Enhancements introduced during 2009:
(return to top of page)

2009-Dec
Fix a typo in PENSIM source code that prevented aggregation of threaded PENSIM output generated by multiple-scenario runs that specify .xsc output but not .xsd and .xse output. Thanks to Stan Panis of Advanced Analytical Consulting Group for reporting this bug. This fixes a bug that has been in PENSIM since April 2009 when the PENSIM .xsd and .xse output files were added. This change has no effect on the values of PENSIM output statistics.

Add PENSIM code that turns off the accessing of an individual's rollover account (for the purpose of making withdrawals or buying a life annuity) when that individual recovers from the disability that triggered the start of rollover-account accessing. This change eliminates a bug that has been in PENSIM since September 2007 when disability recovery was added to PENSIM. This change has small effects on pension benefits among the small group of individuals who recover from a spell of disability and who had a positive rollover-account balance during the spell of disability. For example, in the PENSIM run used to identify and fix this bug, only 56 out of 205,397 individuals (0.027 percent) were affected.

Add federal income tax saver's credit statistics to supplement 5 of the PENSIM cross-sectional surveys whose results appear in the .xsa, .xsb, .xsc, .xsd, and .xse output files.

Add federal income tax logic to PENSIM that computes AGI and a refundable saver's credit structurally similar to that proposed in the FY2010 federal government budget. In the current version of the model, all the saver's credit is assumed to be deposited either in the individual's DC pension plan account or in the individual's pension rollover account.

Upgrade the Info-ZIP zip.exe utility included in the PSG models distribution from version 2.31 to version 3.00, and upgrade the unzip.exe utility from the 5.52 to the 6.00 version.

Add the individual's lifetime asset-allocation style to the PENSIM .cov output results file.

Improve PENSIM input database documentation for seven tables that are related to asset returns, especially returns on different corporate equity asset categories.

Add to the RSF Toolkit Execute menu a new item that provides access to the AWS Toolkit, which can be used to control the execution of runs on Amazon Web Services (AWS) computers.

Fix a minor logic bug in the calculation of survivor benefits for foreign-born individuals who emigrate from the United States after establishing their insured status and then die leaving a surviving spouse. This bug was introduced into the PSG models during 2008 when the migration logic of the models was revised extensively to reflect the change in migration methods used to produce estimates in the Trustees Report. This bug fix avoids a rare, but fatal, error in stochastic OLC runs and it causes a minuscule increase in average benefits. The increase in average benefits is so small that there is no change in the value of any of the OASDI solvency statistics in the .ssa output file for a standard deterministic OLC run. The incidence of the bug was so rare that none of the benchmark run results change after fixing the bug.

2009-Nov
Streamline logic of dbrun utility that is employed to use run specification files in the RSF Toolkit. This logic improvement produces a substantial speed up in the creation of model runs from rsf, with the use time dropping from five seconds to one second for a simple RCS mode run specification file employed in the testing of the enhancement.

Standardize the order of the model create-run sections produced by the RSF Toolkit build process. The order of the sections in all run modes is now PENSIM at the top, GEMINI in the middle, and SSASIM at the bottom of the run specification file. This change facilitates comparisons of the contents of two run specification files using the Contrast two files on disk item on the RSF menu of the RSF Toolkit. Run specification files that do not have this new section ordering can still be used by the RSF Toolkit.

2009-Oct
Revise extensively the automated install/upgrade procedure, primarily to make it easier to use by IT staff who are in charge of installing and upgrading the PSG models for users who do not have administrative privileges on their personal computers. Other enhancements to the install/upgrade procedure include: the suggested model installation folder is better aligned with the file structure in Windows Vista and Windows 7, the appropriate environment variables are set in a new way that works on Windows XP, Windows Vista, and Windows 7, and the prompt for the model installation folder has been eliminated when doing an upgrade.

Change the handling of directory naming in the PSG models so as to gain flexibility in locating the model installation folder within the Windows file structure. As users move from Windows XP to Windows 7, the Windows file structure strongly suggests installing the models in the \Users\{username} directory rather than in the root \ directory. This change will affect only those who use database directories other than the default db subdirectory. If you have done this, be sure to read the new documentation on the SSASIM OLC.gemini_dir, ESP.pensim_dir, RCS.gemini_dir fields, on the PENSIM RUN.ssasim_dir field, and on the GEMINI RUN.ssasim_dir, SAMPLE.pensim_dir fields. If you use only the default db subdirectories, there is nothing to do.

Change the default backup folder/directory in the RSF Toolkit so as to gain flexibility in where the backup file is stored within the Windows directory structure. The new backup folder/directory is the parent folder/directory of the SSASIM model installation folder/directory, but you can change that to be anything you want by changing preferences on the RSF Toolkit Help menu.

Upgrade gnuplot graphing program from version 4.0 to 4.2.6 so as to take advantage of bug fixes and new features that will be used by the analysis toolkits in the future. Add new png graphics format option to TFS Toolkit, the trust-fund solvency analysis toolkit. The png graphics format is lower quality than the existing eps or pdf graphics formats, but a png file can be viewed on the screen and printed without Adobe Acrobat Standard or Adobe Acrobat Professional being installed, and it can be inserted into a Microsoft (Word, Excel, PowerPoint) document without going through the hassle Microsoft has made out of importing eps files.

Upgrade database library to SQLite 3.6.19 release, which enforces foreign-key constraints. Revise SQL statements that define the PSG model input database tables so that referential integrity is enforced with triggers rather than with foreign-key constraints, which were parsed but not enforced in earlier SQLite releases. These changes are extensive, but should not be visible to users of the PSG models.

Add INCTAX table, and several INCTAX child tables, to the PENSIM input database. These tables will contain federal income tax policy parameters for years beginning with 2004. Adding a federal income tax capability to PENSIM is a major undertaking that will involve months of work. In the meantime, there will be signs of the construction effort and users should not attempt to set the value of PENSIM RUN.use_inctax to true.

2009-Sep
Add two new fields to SSASIM PRDMRKT table that enable the simulation of a one-time perturbation to the annual inflation rate followed by an offsetting perturbation in the following year.

Add the ability to simulate a payable-benefits policy in stochastic runs, which replaces the old BEND_DI.pay_pct and BEND_RI.pay_pct fields that worked only for deterministic (single-scenario) runs. This new ability is implemented by adding a field to SSASIM RUN table that specifies an optional input file that contains OASI and DI payable-benefits percentages for each year of each scenario. This new field, RUN.paypctfile, contains a single dot character in all the benchmark runs that are used as base runs in run specification files, implying a scheduled-benefits policy. A payable-benefits policy can be implied by specifying the name of a file that has the same format as the new .pay output results file (see next item). See the SSASIM input database documentation for details on the new RUN.paypctfile field and the two simple steps required to translate a .pay output file into a paypct input file.

Add a new output results file to SSASIM that automatically generates payable-benefits percentages for each scenario year. The new .pay output file contains payable-benefits percents for the OASI and DI programs, separately, for each year of each scenario. See the SSASIM input database documentation for details on the two new fields: OUTPUT.pay and OUTPUT.pay_preyrs.

Add a suite of tests that checks for identical output in stochastic runs in all modes of operation regardless of which threading or distributed-processing options are used. Threading of cohort samples in OLC runs has been a feature of the PSG models since 2003, and threading of scenarios in stochastic ESP and RCS runs was added in March 2009. Distributed processing of bundles of scenarios in stochastic ESP, RCS, and OLC runs was added in July 2009. Threading utilizes multiple cores in a single computer, while distributed processing utilizes several computers on a network. The new test suite ensures that runs produce exactly the same results no matter what combination of threading and distributed processing is specified.

Add the ability to edit, in the RSF Toolkit editor, files used as or with post-simulation tab scripts. The three types of files are Windows batch scripts (*.bat files), Tcl programs (*.tcl files), and AWK programs (*.awk files). The RSF editor can now use the full range of editing commands on these three file types. This enhancement allows the preparation in the RSF Toolkit editor of files used as or with tab scripts. See the description of the RUN.tab_script field in the SSASIM input database documentation for details on how to automate post-simulation analysis of the raw output files generated by the PSG models.

Add a new PENSIM table that contains pension account fees for rollover accounts and for automatic-IRA accounts. The new table, which is a child of the PENSIM RUN table, is called ACCTFEE. See the documentation of the ACCTFEE for details on account fees, which are indexed for price inflation. The baseline values of both fees have been set to zero in order to replicate the

2009-Aug
No enhancements.

2009-Jul
Add three new fields to the SSASIM RUN table that together create an optional ability to split multiple-scenario runs into subruns, each one of which simulates a bundle of scenarios that is a subset of the total number of scenarios in the run. The purpose of this new capability is to support distributed processing of subruns on separate computers, with the subrun output results being merged after all the subruns have finished execution. The merged output results are the same as would be produced by the original run if it was not split into subruns. The point of the distributed processing is to execute a very long stochastic model run in less time than it would take to execute on a single computer. For example, using five identical computers to distribute the processing means that five subruns can be executed in about one-fifth the time required for sequential execution of the subruns on one computer. And, if even more computers are used for the distributed processing, the execution time can be reduced dramatically. The three new SSASIM RUN table fields are called dp_active, dp_f_scen (the first scenario in the subrun's bundle), and dp_l_scen (the last scenario in the subrun's bundle), and they are described in the new SSASIM input database documentation. RUN.dp_active is set to false in all the benchmark runs, which means that no changes need to be made in old run specification files in order to continue using the PSG models in the conventional non-distributed-processing way.

Develop two utilities that support distributed processing by automating the tasks of splitting a multiple-scenario run into subruns and merging output from all the subruns. The PSG models distributed-processing capability is a simple application of the Google MapReduce framework. The dp-split utility (the Map step) creates many subrun specification files from the original stochastic rsf. The dp-merge utility (the Reduce step) combines the output from each subrun into a single set of output results files that is equivalent to what would be produced by executing the original rsf. The utilities are short Tcl programs named dp-split.tcl and dp-merge.tcl that are invoked by the dp-split.bat and dp-merge.bat scripts, respectively. All four of these dp-* files are located in the SSASIM executables folder, where they can be read to understand how they work. They can be used in a wide variety of distributed-processing environments, including the distribute processing of subruns on a local area network of computers running the PSG models.

Remove the four lines at the bottom of most SSASIM output files that repeated the header line and showed the mean, standard deviation, and standard error of each output statistic. Revise the logic of all the res output tabulation utilities to handle the new file format.

Remove all-scenario summary lines at the bottom of the PENSIM .arc output file and at the bottom of the GEMINI .arc output file. Revise documentation of these two output files to include instructions for how to compute all-scenario solvency of the annuity provider in multiple-scenario stochastic runs.

Upgrade the Tcl language interpreter used by the PSG models from version 8.5.2 to the first 8.6 beta version.

Correct values for the low-benefit threshold in the SSASIM COHORT table rows used by the 2008TR benchmark runs and by the 2009TR benchmark runs, so that their values correspond to what the COHORT.notes field says they are: Census Bureau poverty lines. PSG model users can set these low-benefit thresholds to any values they want.

Add a new field to each of the SSASIM trust FUND tables. The new field is called firstchkyr and it specifies the first calendar year in which a check is conducted for a relatively flat maturity profile for special-issue Treasury bonds held in the trust fund. This check is conducted only if sprd_mats is true and rmv_by_cpn is false, which are the default values in all the benchmark runs. If the check finds a maturity profile that is not very flat, a warning message (that includes all the special-issue Treasury bonds held by the fund) is printed and run execution continues. If, after studying such a warning message, you are certain everything is being simulated as you want, you can suppress the warning message by setting this field to a year after the last printed warning message. These new fields have been added to replace hard-coded first check years, and therefore, they add flexibility for model users, but do not represent any new substantive capabilities in the PSG models.

Add new field to PENSIM TRACE table that is named coh_b_year. The logic associated with this new field makes it possible to trace an individual or spouse included in a cohort sample born in a specified year during an OLC run.

Change PENSIM asset terminology so that 'life-cycle fund' becomes 'target-date fund' to keep up with English usage trends. This is a change in terminology only; no changes have been made to model logic. All model documentation has been revised to reflect this change. Old run specification files that include the lcf_pct_f or lcf_pct_m fields in any one of the four ACCTAA1-ACCTAA4 tables need to be changed to work with the new version of the PSG models. Simply change every lcf_pct_? to tdf_pct_? in each rsf that you want to use with the new version of PENSIM.

Add two new fields to the PENSIM PENBEHV table that specify a separate stream of random numbers that are used to simulate individual pension behavior related to auto-IRA plans. The number of auto-IRA random numbers drawn for each individual lifetime, PENBEHV.num_aiu, has a value of one in the benchmark runs that do not assume any auto-IRA plans, so when simulating an auto-IRA reform be sure to increase this number above one.

Add a new field to the PENSIM CSTOCK table that specifies whether or not the random number streams that are used to simulate company-stock returns are synchronized across runs that have the same number of sample individuals and job histories. The new field, which is called CSTOCK.sync_rns, is set to false in the benchmark runs. When setting it to true expect at least a ten percent increase in PENSIM run execution times.

2009-Jun
Add four existing PENSIM migration tables (EMIGRA, EMIGDYN, IMMIGRA, IMMIDYN) to the scope of the run specification language and add documentation for all parameters in those four tables.

2009-May
Add five new benchmark runs that use the intermediate-cost assumptions from the 2009 Trustees Report. The new runs are as follows: CBA 19, OLC 39, ESP 59, RCS 79, and ESP+ 99. Given that pension benefits need to be computed in order to calculate social security benefits correctly (given the existence of the WEP/GPO benefit reduction rules that are part of current-law policy), both the OLC solvency run 39 and the RCS distributional run 79 use PENSIM to compute employer-sponsored pension benefits, which are then used by GEMINI to compute social security benefits. The Build process in the RSF Toolkit has been revised to use these 2009 benchmark runs as the base runs when specifying a new run. A one-in-a-thousand-sample-size OLC run, using run 39 as its base, produces OASI, DI, and OASDI summarized cost rates, summarized income rates, and actuarial balances, that are the same (within rounding error) as those shown in the 2009 Trustees Report for the intermediate-cost assumptions.

Add six new fields to the SSASIM POP table that together determine foreign-born emigration rates that vary by years since immigration and by age. The resulting foreign-born emigration rates are used to implement (in CBA mode runs only) the explicit-emigration logic that was introduced in the 2008 Trustees Report. The new migration logic has been used in all other modes of PSG model operation for almost a year. Now all PSG model runs use the new explicit-emigration logic, which is described in the 2008 Trustees Report on pages 76 through 81. The new migration logic will be activated beginning with CBA mode benchmark run 19, which will incorporate 2009 Trustees Report intermediate-cost assumptions. CBA mode benchmark runs that use assumptions from Trustees Reports dated before 2009 continue to use the old migration logic.

Remove the SSASIM reform option to have the social security COLA be computed using the average wage index (AWI) instead of the consumer price index (CPI).

Add logic to the indexing rules for MTE, XTE, and ETE amounts (see the SSASIM MTE table) that suspends the indexing in years for which the (pre-offset) social security benefit COLA is zero.

Upgrade the SQLite programs and source code that are part of the PSG models to the 3.6.14.2 version.

2009-Apr
Add two new cross-sectional surveys to PENSIM output, which brings the total number of cross-sectional surveys that can be conducted in a single run to five.

Add to SSASIM the capability of executing, at the very end of a run of the PSG models, any specified script that tabulates raw output generated by that run. This enhancement provides a convenient way to automate the processing of a run's raw output so that any kind of post-simulation analysis or storage of the raw output files can be conducted. The script can be either a Windows batch file or a Tcl program. This enhancement requires the addition of a new field at the end of the SSASIM RUN table called tab_script.

Upgrade the SQLite programs and source code that are part of the PSG models to the 3.6.12 version. There should be no visible change in the operation of the PSG models because of this change.

2009-Mar
Implement a series of changes that together create a threading capability for stochastic ESP mode and RCS mode runs. This threading capability allows these kinds of runs to make use of all the CPU cores present on a computer, and therefore, to execute much faster.

Alter slightly the way in which inflation rates are computed for years since 1997 and the way they are used to calculate the social security COLA. These improvements cause very small changes in social security results (in CBA, OLC, and RCS modes) that are of no policy significance and cause very small changes to inflation-adjusted pension benefits (in EPS and ESP+ modes).

2009-Feb
No enhancements.

2009-Jan
Add eight new statistics at the end of the GEMINI .sum output results file that are money's worth (present value and real internal rate of return) statistics for account-based employer-sponsored pension plans. Consult the revised documentation of the GEMINI .sum output results file for the definitions of the new pension statistics, which are numbered from (45) to (52). The new statistics are computed using pension benefits and pension contributions (from both employees and employers) for all DB/CA (cash balance) and all DC plans over the cohort individual's lifetime. Note that all traditional DB plans are excluded from the calculations.

Add to PENSIM the ability to simulate automatic-IRA pension plans, like those proposed in J. Mark Iwry and David C. John, Pursuing Universal Retirement Security Through Automatic IRAs: Executive Summary, The Retirement Security Project, April 2007. The automatic-IRA plans have been added as a new fifth type of DC pension plan. This enhancement involves the addition of four new PENSIM input database tables containing the offering requirements and characteristics of automatic-IRA plans, and the addition of new fields in the LAW_CYR policy table to indicate the timing of the automatic-IRA reform and the inflation-indexed level of the maximum IRA contribution amount. Several PENSIM output files have been changed to report the new DC plan type. In addition, several PENSIM enhancements add new pension behavior parameters to allow assumed behavior of automatic-IRA eligibles to differ from that of other DC-plan eligibles.

Expand scope of cross-sectional-survey supplement 3, which is an option in the PENSIM .xsa, .xsb, and .xsc output results files. This supplement now reports information on prior-year rollover decisions associated with all types of DC plans (including automatic-IRA plans) rather than just DC/ST plans.

Add four new pension characteristic parameters for zero-matching DC/ST plans. The new parameters permit the zero-matching DC/ST plans to have automatic-enrollment arrangements or not. The four new parameters are located in the ZMPDBP1, ZMPNOP1, ZMPDBG1, and ZMPNOG1 tables, which correspond to the setting within which the employer offers the zero-matching DC/ST plan. The benchmark runs set the value of these new parameters to false, which has been the hard-wired value since the zero-matching DC/ST plans were added to PENSIM in 2007. When the zero-matching DC/ST plans do have automatic enrollment, the default contribution rate is always three percent and the default investment is always the target-date fund.

Revise the 2008 pension law parameters used in PENSIM benchmark runs. Previous versions of PENSIM used published pension limitation amounts through 2006 and then inflation indexed the 2006 amounts in subsequent years. The new version of PENSIM uses published pension limitation amounts through 2008 and then inflation indexes the 2008 amounts in subsequent years. This enhancement creates some very minor differences in the benchmark run results.

Add the ability to simulate government contributions to DC/ST plans, including the so-called zero-matching DC/ST plans. The government contributions can be either a flat annual amount or a fraction of employee contributions up to an annual limit or a combination of the two (that is, a contribution that has a flat component and a matching component). The policy parameters that specify the nature of the government DC/ST contributions are in a new PENSIM table, GPC_CYR, which is a child table of PENLAW. In conjunction with these new policy options, the PACONT1 table now contains a new employee contribution behavior parameter (gpc_f_o_r) that specifies the rate at which employee contributions are reduced when the government offers a flat contribution. Also, add output statistics related to the new government pension contributions (GPC) to both the PENSIM .cst cost output file and to supplement 1 of the .xsa, .xsb, and .xsc output files. Read the documentation of those output files to get the details of the GPC statistics.

Add two new payroll tax rate parameters and rename one payroll tax rate parameter in both the SSASIM TAXR_DI table (for DI tax rates) and the SSASIM TAXR_RI table (for OASI tax rates). These changes allow the specification of a payroll tax rate that differs across two age groups. The boundary age between the two groups and the payroll tax rates that apply to the two age groups can vary from year to year. Consult the revised documentation of the SSASIM input database tables TAXR_DI and TAXR_RI for details.

Add a new kind of fund to the list of assets in PENSIM, which involves adding two new fields to the ASSETS table and two new fields to each of the four ACCTAAn (account-asset-allocation) tables. The new so-called guaranteed-return fund is somehow engineered to provide a constant real (i.e., adjusted for inflation) rate of return every year. The two new ASSET table fields allow specification of the assumed real rate of return and the fund's fee. And the two new ACCTAAn table fields allow the specification of gender-specific and age-specific portfolio percentages for the guaranteed-return fund that apply to all DC plan accounts as well as to all rollover accounts. There is no attempt in the current version of PENSIM to estimate the cost to the fund of providing such a guaranteed real return. If the guaranteed return is low relative to returns on market assets, then the fund could probably make a long-term profit providing the guaranteed real return. On the other hand, if the guaranteed return is high relative to returns on market assets, then the fund would be likely to incur a long-term loss providing the guaranteed real return. The current version of PENSIM does not estimate the highest level of guaranteed real return that avoids a long-term loss for the fund.

Revise the PENSIM PAINVS table parameters having to do with default investment endorsement effects to allow different endorsement effects in defined-contribution savings-thrift (DC/ST) plans and in automatic-IRA (DC/AI) plans.

Remove the PENSIM PAROLL1 table full_ai_ro field and add the PAROLL2 table is_ai_plan field, which is a binary logit equation coefficient for a dummy variable that is one only when the individual at job end has a positive balance in an automatic-IRA (DC/AI) pension plan. The addition of this parameter allows the rollover behavior of DC/AI participants to differ from that of other DC plan participants.

Upgrade the SQLite database library and utility programs from version 3.6.0 to version 3.6.10, which should have no visible effect on PSG model operations or results.

Enhancements introduced during 2008:
(return to top of page)

2008-Dec
No enhancements.

2008-Nov
Add quarters-of-coverage statistic to GEMINI .sum and .bAA output files.

2008-Oct
Add docage, which is statistic (43) in the GEMINI .sum output file, to the end of each row of the GEMINI .bAA output file, where it becomes statistic (30). Add the percentage annual rate of return on trust-fund assets to the end of each row in the SSASIM .dfc, .dfd, and .dfr output files as statistic (14). Make several minor revisions to the source code that cause very small changes in some output statistics, typically in the fourth or fifth digit. These code changes should cause no substantive change in output from the PSG models.

2008-Sep
Enhance the GEMINI trace capability so that any individual, either cohort sample members (with positive id values) or spouses of sample members (with negative id values), can have detailed annual tracing information written out during either RCS or OLC mode runs. Add two new statistics to the I row of the PENSIM .pen output file, which changes the column numbers of the first-age and last-age statistics at the end of the I row in the .pen output file. Add two new migration-related statistics to the core questionnaire of the GEMINI .xsa, .xsb, and .xsc output files, which changes the column numbers of most of the output statistics in these files. Add one new migration-related statistic at the end the GEMINI .sum output file. Change the value of GEMINI .sum statistic 36 (age at which initial OASDI benefit is received) when no benefit is ever received: the no-benefit age was -1 in prior versions; now the no-benefit age is 999.

Prepare new benchmark runs that incorporate intermediate-cost assumptions and methods from the 2008 Trustees Report. The new Trustees Report projection includes substantial changes in migration assumptions and methods, which account for a sizable improvement in the OASDI actuarial balance. The new benchmark runs distributed with this version of the PSG models simulate migration so that the both annual-total and age-specific counts for both net and gross immigration are close to those in the Trustees Report projection. The migration logic of the PSG models is somewhat different than that of the Trustees Report projection, and the CBO CBOLT model, in three ways:

(a) The logic has been modified so that some immigrants lack sufficient documentation to claim social security benefits, even though they have paid social security taxes when they work in the United States. This enhancement enables the PSG models to simulate, at the micro level, the so-called earnings suspense file, which accounts for trust-fund revenues that cannot be matched to a valid social security number and which constitutes a substantial fraction of the total trust-fund balance. Both OCACT and CBO are assuming in their projections that in each year about 1.3 percent of OASDI payroll tax revenue flows into the suspense file, which is also what the new PSG-model benchmark runs assume. Unlike the other models, the immigrants who contribute the 1.3 percent of taxes are included in the mainstream simulation, but receive no OASDI benefits because of their undocumented status.

(b) Another difference between OCACT (but not CBOLT) and the PSG models is the assumption that the distribution of educational attainment (and hence, earnings) among immigrants differs from that of the native born, with immigrants having somewhat higher proportions of both low and high educational attainment.

(c) The final difference is that the new logic of higher gross immigration and emigration flows begins (as in the real world) decades ago in the PSG models rather than beginning more recently at the start of the simulation period. This happens in the PSG models because all cohorts born since 1935 are simulated from life to death. The implication of this difference is that the composition of new retirees is being affected by the higher assumed migration flows sooner in the simulation period.

Activate the recently-developed WEP/GPO benefit reduction rules in the new benchmark runs that use 2008 Trustees Report assumptions. The pre-2008 benchmark runs continue to not simulate the WEP/GPO rules.

The new migration methods and assumptions, together with the activation of the WEP/GPO benefit reduction rules, cause a widening in the long-standing gap between the OASDI actuarial balance estimated by the PSG models using standard microsimulation methods and the estimate in the Trustees Report. The 2004 gap was 54 basis points (-1.35 versus -1.89, with CBO reporting a CBOLT estimate of -1.36 for that year using Trustees Report assumptions); the 2007 gap was 38 basis points (-1.57 versus -1.95 percent); the 2008 gap is 72 basis points (-0.98 versus -1.70 percent). As a result, the size of the annual actuarial adjustment factor needed to emulate OCACT methods has grown. The new factor values are used in the RSF Toolkit Build process when choosing the emulate OCACT methods option and when the Trustees Report year is 2008. The PSG models still produce the 38 basis point gap when ignoring the WEP/GPO benefit reduction rules and using the old migration assumptions and methods from the 2007 Trustees Report. Note that, as in past years, the PSG model projection using 2008 Trustees Report assumptions closely matches the projections of population, taxable earnings, DI revenue, DI cost, OASI revenue, and OASI beneficiaries in the 2008 Trustees Report. The sole projection difference, as in past years and as found in other microsimulation models, is that projected average OASI benefits for the baby-boom cohorts are lower than in the Trustees Report. If the lifetime earnings of those cohorts were to be assumed to be high enough to generate the average benefit levels projected in the Trustees Report, it would be impossible to match the taxable earnings projection in the Trustees Report and the current balance of the trust funds would be much larger than they are in reality. In other words, the logic of the microsimulation method, which requires that the same lifetime earnings be used to estimate both payroll tax revenues and benefit costs, makes it impossible to match all aspects of the Trustees Report projection.

The new migration methods and assumptions, together with the activation of the WEP/GPO benefit reduction rules, cause a substantial increase in the execution time of OLC and RCS mode runs. This increase in run execution time is caused by the fact that many more migrants are being simulated in a run and because employer-sponsored pension benefits now need to be simulated in order to accurately compute OASDI benefits using the WEP/GPO rules. On a computer that executes a 0.1 sample_pct OLC run using old assumptions and methods in about 1.9 minutes, now takes about 5.3 minutes to execute the same run using new assumptions and methods. In coming months, we hope to be able to develop an imputation scheme that will approximate uncovered pension benefits without running PENSIM in full-pension mode. If we are successful, OLC mode run times should be cut roughly in half.

The RSF Toolkit preferences file has been changed so that the new 2008 Trustees Report benchmark runs are used as the base runs when a run specification file is built by the Toolkit. If you have changed other settings in the preference file, be sure to save those settings so that you can enter them after doing the upgrade. If you want the RSF Toolkit to use the 2007 Trustees Report benchmark runs as the base runs in an rsf (as done in prior versions of the models), simply edit the preferences file using the RSF Toolkit Help menu item called Changing toolkit preferences.

Revise the pension offering assumptions in PENSIM so that the base ESP mode run 58 assume that the prevalence of automatic-enrollment in private-sector DC/ST plans rises from zero to seventy percent over the 1998-2012 period. This is the baseline assumption for the new 2008 benchmark runs, but alternative assumptions can be easily specified using the run specifications language.

2008-Aug
No enhancements.

2008-Jul
Add an extra year of life before accelerated mortality starts in PENSIM runs that support SSASIM OLC mode runs. Add a pension excise tax as a policy option in the PENSIM LAW_CYR table, with the revenue of this tax on employer-sponsored pension distributions being deposited in the OASI trust fund. Add individual behavior parameters to PENSIM PAROLL1 table that can be used to eliminate completely pension account rollovers at job-end. Add two statistics to the PENSIM .cov output file that report the number of lifetime jobs on which an individual vests in a DB (DC) pension plan on a job where that individual's earnings are not subject to social security payroll taxes. Upgrade the PSG models to SQLite 3.6.0 version and to GNU AWK 3.1.6 version. Add SSASIM constraint that the values of pia_frac_1 (the PIA-formula fraction that applies to AIME amounts below the first bend point) in the PIADYNR, PIADYNS, PIADYND tables must be the same.

2008-Jun
Move GEMINI .xsa output file statistic (22) from the supplementary to the core survey questions. Add GEMINI .xsb and .xsc output files that have the same file layout as the .xsa output file. Add field scnsid_ind to STATS table in GEMINI input database with the new logical field controlling whether individual or adult-equivalent-couple benefits and earnings statistics are used to construct the statistics included in the GEMINI .scn and .sid output files. Change GEMINI input database parameter values used in the adult equivalents formula in the RCS mode and ESP+ mode benchmark runs in order to make statistic (41) in the GEMINI .sum output file a more conventional denominator in retirement income replacement rates. Add SSASIM .pfp output file, which contains annual values of PIA formula parameters: the three bendpoints and four formula percents for each of the three eligibility events. Add three certainty-equivalent statistics to GEMINI .sid output file and three certainty-equivalent statistics to GEMINI .scn output file. Add sid_crra field to STATS table in GEMINI input database for specifying the constant relative risk aversion (CRRA) parameter in the expected utility function used to calculate the new certainty-equivalent statistics. Add a new field to both the BEND_DI and BEND_RI tables that allows a change in the method of computing DI and OASI COLAs. Prior model versions always computed annual COLAs using a CPI-indexing method, but now there is the option to use an AWI-indexing method.

2008-May
Revise the source code of the PSG models so that the core models and utilities — but not the various Toolkits — can compile and execute under operating systems other than Windows. So far, this has been done successfully for Mac OS X (10.4+) and Debian 4.0 (32-bit) Linux. The purpose of this work is to gain access to more powerful hardware platforms such as the Mac Pro, which has eight Intel Xeon cores standard, and cloud computing facilities that offer scores of quad-core Linux servers for rent by the hour. Upgrade the Tcl interpreter included with the PSG models to Tcl 8.5.2 version. Upgrade the PSG models to SQLite 3.5.9 version. Add SSASIM INSURED table that includes the main policy parameter that determines whether an individual is fully insured for OASDI benefits. Provide a PENSIM xsall utility that automatically combines the many xs?NNNNN.YYY files produced during an OLC mode run into a single xs?NNNNN.all file that represents a genuine cross-sectional sample of individuals drawn at the survey moment during the survey year. Add to PENSIM the ability to trace the life events experienced by a spouse of a sample individual. Add to PENSIM pension behavior the possibility that individuals cashout the lumpsum equivalent of their vested benefits in career-earnings or terminal-earnings defined-benefit plans when the job ends before retirement, disability, or death. Add GEMINI .sid output file that contains information on each sample individual used to tabulate the scenario statistics contained in the .scn output file, which allows calculation of statistics that measure the variability of social security and pension benefits (a) across stochastic scenarios, (b) across cohort individuals within each scenario, and (c) across retirement years for each individual.

2008-Apr
Add an embedded-SQL capability to the RSF Toolkit's run specification language, which means that now all parameters in the input databases of the PSG models can be manipulated using the run specification language without having to work directly with the databases. Add PENSIM .mig output file that contains aggregate migration and population counts for the simulated birth cohort. Revise PENSIM schooling logic and tables so that native-born and foreign-born individuals can have different educational attainment distributions. Revise the GEMINI AAAF parameter so that it can be a negative number.

2008-Mar
Add ability to specify exemplary cohort individuals (ECI) using the RSF Toolkit's run specification language. Add to SSASIM input database scaled earnings profiles derived by OCACT from 2007 Trustees Report assumptions. Add PENSIM cross-sectional survey supplement 7 that contains statistics that can be used to determine how many individuals work beyond the age at which they are eligible for unreduced (or normal) benefits from the defined-benefit plan covering them on their current job; the new supplement to the PENSIM .xsa, .xsb, .xsc output files has been used by DOL/EBSA in a regulatory impact analysis.

2008-Feb
Revise the baseline PENSIM assumptions concerning the transition probabilities between the positive-annual-contribution state and the zero-annual-contribution state using MINT5 estimation results. Add PENSIM DBWITH1 table parameters to determine gender-specific probabilities of a married individual choosing a single-life annuity as the payout from a traditional defined-benefit pension plan (rather than the default, under federal law, joint-and-survivor payout); the new parameters are set to produce probabilities tabulated from 1992-2000 HRS data shown in the MINT5 final report. Add four earnings statistics to the GEMINI .xsa output file.

2008-Jan
Provide a free educational version of GEMINI and add the ESP+ mode benchmark run to the RSF Toolkit. Add ability to define a range of earnings that is excluded from OASDI taxable earnings, allowing specification of so-called donut-hole payroll tax reform provisions.

Enhancements introduced during 2007:
(return to top of page)

2007-Dec
Add to EDA Toolkit a page containing the numerical value of statistics that appear in the graph; make several other improvements in the EDA Toolkit.

2007-Nov
Add immigration and emigration ages to PENSIM .cov output file. Upgrade the PSG models to SQLite 3.5.2 version.

2007-Oct
No enhancements.

2007-Sep
Add disabled adult children to the PSG models. Add recovery from disability as a life event in PENSIM. Add zero-employer-matching savings-thrift plans to PENSIM using custom tabulations of 1996-97 private-sector and 1998 public-sector Employee Benefit Survey data performed by BLS staff at the request of EBSA. Add to GEMINI cross-sectional output files (.xsa, .xsb, .xsc) new supplements that contain information about beneficiary head counts using SSA beneficiary type definitions. Perform a cross-model validation test comparing beneficiary counts by benefit type in 2060 with estimates for 2060 in the Trustees Report, the results of which show that the PSG models produce estimates that are close to those of OCACT (except in the case of adult SI beneficiaries where the OCACT estimates seem too high given the upward trend in female labor force participation rates).

2007-Aug
No enhancements.

2007-Jul
Add to PENSIM cross-sectional output files (.xsa, .xsb, .xsc) a set of supplementary questions that produce statistics that can be compared to SCF data on defined-contribution balances. Begin adding logic and output statistics for zero-employer-matching savings-thrift plans, which are soon to be introduced into PENSIM. Improve the realism of the Markov process that determines annual transition probabilities in to and out of the state of defined-contribution participants making a zero annual contribution. Calibrate employer probabilities of offering different types of pension plans over the 1975-2005 period using publicly available data. Conduct two validation tests of PENSIM estimates of defined-contribution account balances using 2004 SCF data and 2005 data from the EBRI-ICI 401(k) database, and describe the validation test results in the PENSIM Overview.

2007-Jun
Provide a run specification file for a CBA mode run that produces stochastic results similar to those in the 2007 Trustees Report. Upgrade the PSG models to SQLite 3.4.0 version. Revise the effect of PENSIM INDBEHX retirement delay so that nobody delays retirement beyond age seventy. Add three sets of supplementary questions to core survey questions in the GEMINI .xsa output file.

2007-May
Add new PSG models benchmark runs that use 2007 Trustees Report intermediate-cost assumptions. Change the compiler collection used to build the PSG models from Open Watcom to GNU running under MinGW; the resulting models are slightly larger, but run noticeably faster: one-thousand scenario CBA mode runs are roughly the same speed, OLC mode runs are about twenty-five percent faster, and RCS mode runs are almost thirty percent faster. Add a supplement 5 to the three PENSIM cross-sectional survey output files (.xsa, .xsb, .xsc) that is a March-CPS-style prior-year-retrospective survey on annual earnings and annuity transactions. Add to PENSIM an exemplary cohort individual account saving feature that simulates account contributions, returns, and balance for a single individual, allowing PENSIM to generate simple projections of one individual for comparison with results from other studies (see, for example, the wildly misleading results in the CRS report by P. Purcell and D. Whitman, Retirement Savings: How Much Will Workers Have When They Retire? January 2007). Revise the GEMINI AAAF feature so that the annual actuarial adjustment factor can vary in value across calendar years. Add GEMINI .xsa output file that can be used to generate multiple-cohort cross-sectional survey output for a specified calendar year when GEMINI is operating in OLC mode.

2007-Apr
Change the database technology used by each PSG model so that the hundreds of Microsoft FoxPro files are replaced with a single-file SQLite database. Provide as part of the PSG models installation a Tcl language interpreter (tclkitsh.exe) as well as a Tcl SQLite dynamic link library (tclsqlite3.dll) so that any user can directly manipulate the model databases if the RSF Toolkit run specification language is not sufficient. Add a backup button to the RSF Toolkit. Add social security COLA offset reform to the list of reform provisions available in the RSF Toolkit Build interview. Add a tkprefs file that allows users to specify preferences about the Toolkits: the folder where backups are placed, the Trustees Report year, and the size of the Toolkit window and fonts. Add to PENSIM .pen output file a statistic that reports the couple's end-of-the-year rollover account balance. Add to the PENSIM .cov output file four statistics that report the number of years the individual held full-time jobs, held part-time jobs, was eligible for a defined-benefit pension plan, and was eligible for a defined-contribution pension plan. Add PENSIM INDBEHX table that contains individual behavior parameters are in the scope of the RSF Toolkit run specification language.

2007-Mar
Add to SSASIM AIME table five policy parameters that allow the specification of social security caregiver credit reforms. Add to the RSF Toolkit several new convenience features: multiple run specification file use, find and replace in all run specification files, and renaming of run specification files. Add to PENSIM cross-sectional survey output file a fourth optional supplement that focuses on pension eligibility and vesting a recent job endings. Add to the PENSIM DBVEST table four new parameters that allow the specification of maximum cliff vesting years across all defined-benefit pension plans, with the maximum possibly being different in the private and public sectors and/or possibly being implemented in different years. Add to the PENSIM DCVEST table four new parameters that allow the specification of maximum years to cliff vesting and maximum years to full graded vesting across all defined-contribution pension plans, with the maximum possibly being implemented in different years. Add PENSIM .age output file that contains selected individual statistics at each age in a specified age range. Add to the PENSIM .cov output file several statistics describing the longest job held by the individual. Add to PENSIM ASSETS table a so-called stable-value fund, revise the pension account asset-allocation (ACCTAA?) tables to allow investment in stable-value funds, and revise the cross-sectional survey output to include holdings of stable-value funds. Add to the PENSIM STAENR1 table an additional parameter having to do with automatic enrollment in defined-contribution pension plans. Add to PENSIM PAINVS table new parameters that allow the characterization of an endorsement effect caused by the selection of a default investment under automatic enrollment. Fix a bug in the out2bage utility inadvertently introduced by the source code changes made during February 2007.

2007-Feb
Add SSASIM and GEMINI trace output that confirms that negative PIA reductions at DI-to-OAI conversion are not applied in reforms that would appear to call for such reductions; this substantive change in logic was implemented in January of 2007. Extensively revise the source code of the PSG models to make the code more like C and less like C++, particularly with regard to string formatting and file reading and writing; the source code continues to use fundamental C++ features such as classes, templates, and inheritance. Add roughly two hundred tables in the PENSIM input database to the scope of the RSF Toolkit run specification language. Add the GEMINI TRACE table to the scope of the RSF Toolkit run specification language.

2007-Jan
Provide for the first time to the public a full-featured version of PENSIM that can simulate employer-sponsored pension accumulation and withdrawal. Revise the logic used to compute the maximum taxable earnings (MTE) level from the average wage index (AWI) so that it mimics more closely the actual administrative calculation.

Enhancements introduced during 2006:
(return to top of page)

2006-Dec
Revise the TFS Toolkit so that annual values of the income and cost rate are connected by lines if they are drawn from an actual scenario (rather than representing percentile values, which typically come from different scenarios each year). Provide a resce utility that computes the certainty-equivalent value of the distribution of an output statistic using either expected utility theory or cumulative prospect theory. Add to PENSIM a third cross-sectional survey, so that now up to three different cross-sectional surveys can be conducted in a single ESP mode run.

2006-Nov
Add to the RSF Toolkit Build interview the option of specifying an employer-sponsored pension (ESP) mode run, in anticipation of the public distribution of a pension-simulating version of PENSIM. Provide a Trust-Fund Solvency (TFS) Toolkit that produces a one-page summary that contains several solvency statistics and a time-series graph that shows income and cost rates in either deterministic or stochastic CBA mode or OLC mode runs. Add longevity indexing of the minimum PIA base amount in the SSASIM MINPIA table. Add to PENSIM a second cross-sectional survey, so that now up to two different cross-sectional surveys can be conducted in a single ESP mode run.

2006-Oct
No enhancements.

2006-Sep
Add to the RSF Toolkit the ability to see input database table values used by the base run by clicking a run specification language statement for a table with the right mouse button. Streamline the specification of spouses and kids in ECI mode runs and revise the ECI Toolkit to work with the new structure of the SSASIM input database tables. Add equity return and Treasury bill spread to the set of key SSASIM assumption variables, revise the run specification language to handle the new assumption variables, and revise several SSASIM output files to report their annual values. Add ability to make a one-time change in the wage-indexed level of minimum earnings required to earn a social security quarter of coverage. Add several PENSIM input database tables that allow more detailed specification of assets that can be held in pension accounts.

2006-Aug
No enhancements.

2006-Jul
Revise the RSF Toolkit in several ways and make it the standard interface to the PSG models. Provide an Exemplary Cohort Individual (ECI) Toolkit that automates extraction of output statistics to an Excel spreadsheet. Add a new employer-sponsored pension (ESP) mode of operation to the PSG models in anticipation of the public release of a pension-simulating version of PENSIM. Add to the EDA Toolkit the ability to create two-dimensional color histogram graphs that visualize the joint distribution of two continuous GEMINI output statistics.

2006-Jun
Provide a run specification file for a CBA mode run that produces stochastic results similar to those in the 2006 Trustees Report. Provide an Exploratory Data Analysis (EDA) Toolkit for tabulating and graphing statistics from GEMINI output files. Add GEMINI .cpi output file containing annual values of the CPI index. Revise the out2bage utility used by the EDA Toolkit to generate .bAA output files so that it uses the .cpi output file to convert nominal dollar amounts into real dollar amounts. Add to GEMINI .sum output file two level earnings statistics that are suitable denominators for individual and couple replacement rates.

2006-May
Update SSASIM HISTORY and HISTMDR tables with statistics for recent years and add new HISTMR table, so that now the last historical year in PSG model simulations is 2004 instead of 1997. Provide a run specification file for a CBA mode run that produces stochastic results similar to those in the 2005 Trustees Report. Add PSG model benchmark runs that use intermediate-cost assumptions from the 2006 Trustees Report. Add low benefits-plus-earnings avoidance rate statistic to the GEMINI .sum output file.

2006-Apr
No enhancements.

2006-Mar
Add options for SSASIM to produce .sti and .sto output files in OLC mode runs. Improve logic of DI-to-OAI conversion for cohorts with fractional normal retirement ages, producing a smoother flow of beneficiaries from DI to OAI during years when the normal retirement age is increasing by a fraction of a year. Fix a bug in the calculation of PIA for individuals who experience a disability or death event after age 62 and before the retirement event. Add to SSASIM scaled earnings calculated by OCACT using 2005 Trustees Report assumptions.

2006-Feb
No enhancements.

2006-Jan
Upgrade to a newer version of the Open Watcom C++ compiler to build the PSG models, resulting in roughly a ten percent reduction in run execution time. Change the run specification file PENSIM preamble generated by the RSF Toolkit so that there is no emigration of foreign-born individuals, which is a highly unrealistic assumption, but one that aligns the PSG models with assumptions made in the Trustees Report and in the CBO CBOLT model.

Enhancements introduced during 2005:
(return to top of page)

2005-Dec
Revise PENSIM so that it directly writes binary coh files in the OLC and RCS modes of operation. Add OLC mode logic to PENSIM that forces immediate death of cohort sample individuals and their spouses in the year following the last simulation year in SSASIM. These two changes in the PSG models have no effect on output results, but reduce the execution time for an OLC run by about fifty percent. Fix a bug in the calculation of enhanced PIA (as in the CSSS reform proposals) and the minimum PIA (as in the Graham reform proposal) that was caused by assuming quarters of coverage are counted back to age 21 instead of age 22 (as in those reform proposals). Fix a bug in the calculation of minimum PIA for DI beneficiaries being converted to retired-worker beneficiaries under the Graham minimum PIA reform provision. Add GEMINI .scn output file containing aggregate statistics for the cohort sample in each scenario.

2005-Nov
Add a policy parameter that allows the maximum social security account contribution rate to vary over time.

2005-Oct
Add ability to specify (deterministic or stochastic) macrodemographic and macroeconomic assumption parameters using the run specification language in the RSF Toolkit. Add statistics to SSASIM .pod output file that report the age composition of the population. Add aggregate taxable earnings to the SSASIM .ben output file.

2005-Sep
Continue developing the RSF Toolkit, which will eventually become the user interface to the PSG models. Add policy parameter that enable specification of reforms that involve the scaling down of PIA-formula fractions for the DI eligibility event in a way that differs by eligibility age, by calendar year, and by formula fraction (i.e., in a progressive manner).

2005-Aug
No enhancements.

2005-Jul
Provide the first, incomplete version of the RSF Toolkit, which will serve as the new interface to the PSG models, providing the ability to specify a new model run using a simple run specification language without having to manipulate the model input databases by hand. Add several policy parameters to allow specification of reforms that establish non-linear excess payroll tax rate schedules.

2005-Jun
Add policy parameters that enable specification of reforms in which the bottom two and the top two OASDI PIA fractions decline at different rates (i.e., the progressive indexing of benefits). Add a policy parameter that allows specification of reforms in which disabled-worker beneficiaries are not converted to retired-worker beneficiaries at their normal retirement age. Add a policy parameter that allows specification of reforms in which disabled workers are not allowed access to DI benefits, and must take reduced early retirement benefits, if their age is greater than or equal to the early retirement age. Improve the resclean utility so that it automatically detects output files to delete in all three PSG model working directories.

2005-May
Introduce automatic installation and version upgrade capabilities. Provide initial version of the dbrun-create utility that automates some aspects of creating a new run in the PSG model input databases; this requires some revisions to the structure of the several SSASIM tables. Correct a bug that read the fraction of newborns who are girls as an integer rather than a real value.

2005-Apr
Add benchmark runs to the PSG model input databases that incorporate intermediate-cost assumptions from the 2005 Trustees Report. Streamline the specification of spouses and children of exemplary cohort individuals (ECI) in the SSASIM input database. Add a statistic to the SSASIM .pod output file that reports the population aged 65 or more in each simulation year. Add to PENSIM the ability to reset all random number streams at the start of each scenario, which permits simulating the exact same sample of cohort individuals in each scenario (when the reset parameter is true) or a different sample of cohort individuals in each scenario (when the reset parameter is false); the later is more reliable because there is no chance that the single scenario sample is statistically unrepresentative of the cohort population.

2005-Mar
Revise PSG model logic so that it is possible to specify an OLC mode run that has as many as 900 years in the actuarial test period (rather than the 75 years used in the Trustees Report). Simplify the specification of exemplary cohort individuals (ECI) in the SSASIM input database using OCACT scaled earnings as the default earnings history. Add a new OASI and DI payroll tax rate parameter that allows taxation of earnings above the maximum taxable earnings (MTE) without using the above MTE earnings in OASDI benefit calculations. Add new social security account policy parameter that allows a positive contribution rate on earnings above the MTE. Provide a command-line dbrun-delete utility that deletes from model input databases all rows used by the specified run that are not used by any other runs. Provide a command-line hist2d utility that produces two-dimensional histogram graphs from GEMINI output files.

2005-Feb
Streamline the specification of social security accounts in several ways. Add new DI policy parameter that allows specification of Graham reform proposal for DI PIA calculation when PIA fractions are declining. Revise the assumption about the timing within a calendar year of the payment of OASDI taxes and the receipt of OASDI benefits so that cashflows are assumed to occur at the start of the year, bringing the micro ECI and RCS modes of operation into line with the assumptions used in the macro CBA and OLC modes of operation. Add a GEMINI annuity provider revenue and cost output file that reports the solvency of the social security account annuity provider, which provides statistics needed to determine the pricing of annuities that will ensure the solvency of the provider.

2005-Jan
Add a fourth automatic adjustment method for the OASDI PIA formula fraction that sets this year's fraction equal to last year's fraction, which allows reforms that stop benefit indexing in some future year. Simplify the modeling of social security account clawbacks. Improve the logic used to estimate trust-fund income produced by the income taxation of social security benefits. Revise PENSIM job-start logic so that individuals never start a new job late on New Year's Eve. Add two statistics to the GEMINI .sum output file that report the lifetime present value of OASI and DI benefits received from the DB tier of social security.

Enhancements introduced during 2004:
(return to top of page)

2004-Dec
Add DI-to-OAI conversion logic option required to characterize the Graham proposal. Add two new statistics to the GEMINI .sum output file that record the age at which the first OASDI benefit is received and the dollar amount of that first benefit.

2004-Nov
Add ability to specify various automatic adjustments (or indexing) methods for the fractions used in the OASDI PIA formula, which allows easier specification of the CSSS reform proposals and correct simulation of CSSS plan 2 under stochastic assumptions that imply variation in real wage growth across macroeconomic scenarios. Add ability of not applying the minimum PIA policy to PIA calculations at the disability event, which allows that specification of the minimum PIA policy contained in the Graham reform proposal.

2004-Oct
No enhancements.

2004-Sep
Add ability to increase the number of dropout years in the OASDI AIME formula, which adds flexibility to the longstanding ability to reduce the number of AIME dropout years. Add ability to specify a nonlinear minimum PIA formula as in the most recent Kolbe-Stenholm bill (see OCACT memo dated 2/11/04). Add ability to specify a maximum retired-worker spousal benefit as in the most recent Kolbe-Stenholm bill. Add to the PSG models distribution two new output analysis utilities, mean and pctiles, that quickly tabulate at the Windows command prompt the mean and percentile distributions of any statistic in GEMINI .sum and .bAA output files.

2004-Aug
No enhancements.

2004-Jul
No enhancements.

2004-Jun
Add new PSG model benchmark runs that use 2004 Trustees Report intermediate-cost assumptions.

2004-May
Add to the binary env file generated by SSASIM for PENSIM input OASDI statistics required for offset-integration benefit calculation logic for some defined-benefit pension plans. Add to the GEMINI .sum output file a second OASDI internal rate of return statistic so that the old statistic is computed using individual tax and benefit cashflows and the new statistic is computed using family-size-adjusted couple tax and benefit cashflows. Add age of disability onset and age of disability recovery to GEMINI .bAA output file.

2004-Apr
No enhancements.

2004-Mar
Change the assumption used in calculating summarized cost rates, summarized income rates, and the actuarial balance, concerning the timing of cashflows within each calendar year. Prior versions of SSASIM assumed that cashflows involved in the calculation of these present value statistics occur at the end of the calendar year. But Trustees Report Table IV.B6, which appeared for the first time in the 2003 report (see page 61), implies that the Office of the Chief Actuary assumes that the cashflows occur at the beginning of each calendar year in these present value calculations. The same within-year timing assumption is made beginning with this version of SSASIM. This change in present-value-calculation logic causes small changes in the summarized rates.

2004-Feb
Add SSASIM .ben output file that contains annual OASI and DI beneficiary head counts as well as annual values of the SSA average wage index (AWI), the CPI index, and OASDI maximum taxable earnings (MTE). Update PENSIM input parameters so that OLC mode runs use 2003 Trustees Report intermediate-cost assumptions same as CBA mode runs. Add ability to specify whether or not GEMINI computes an OASDI internal rate of return statistic for each individual, and whether or not the cashflows used in the calculation are individual or couple taxes and benefits. Add to GEMINI an annual actuarial adjustment factor (AAAF) to enable OLC mode runs using 2003 Trustees Report intermediate-cost assumptions to simulate aggregate future benefits that are as high as reported in the 2003 Trustees Report. The AAAF feature allows users to either use standard microsimulation methods and get lower benefits (as in the CBO CBOLT model) or to use the AAAF to emulate the higher benefit shown in the 2003 Trustees Report.

2004-Jan
Add features to support new OLC mode of operation, including a SSASIM population detail .pod output file, and immigration and emigration behavior in PENSIM. Add new OAI policy parameter that controls size of auxiliary child benefits so they can differ from auxiliary spouse benefits. Add present value of taxable earnings statistics to SSASIM .ssa output file. Add statistics to GEMINI .sum output file that adjust couple benefits, taxes, and earnings for family size, using two user-defined parameter values and a family size equivalence formula from a 1995 National Academy of Sciences report. Add this family size equivalent statistic to the GEMINI .adq output file.

This page was last revised on January 27, 2012.

PolicySimulationGroup

Pension Projection Models

Enhancements Introduced in Earlier Versions of PSG Models (since 2004)

Policy
Simulation
Group