MySQL Conversion

• Non-baseline micro results have not been converted to MySQL. This decision was made based on the large number and size of these files, coupled with the likelihood that they will not actually be needed and the fact that they can be easily recreated if necessary. If non-baseline microresults are needed the user just needs to re-execute the run that created the results. If this causes an unreasonable amount of work, please let us know.
  
• Converting the TRIM3 "frame" (i.e. the aspect of the model that handles input and output) from Oracle to MySQL made all existing versions of simulation module code unusable. In all cases, a TRIM programmer can make the code usable by making a few simple changes to the code. However, we only made the latest version of each simulation module's code usable on the MySQL version. Again, this decision was made based on the large number of older versions of code, coupled with the likelihood that they will not actually be needed and the fact that they can be easily made usable if necessary. However, this does mean that if you try to run a simulation that uses any but the latest version of a simulation module's code, you will get an e-mail with the message:

  TRIM-00050: "simulationid" has an error.
  Cannot load "module version" 193

  There are two solutions to this problem. Ideally, you can simply modify your run to use the latest version. However, if it is necessary to use the older version, please notify us and we will make that version available. Note that the latest version of a module on the MySQL version may be later (i.e. a higher number) than that on the Oracle version. In this case, the MySQL version should produce the same results as the Oracle version (the only differences in the versions are those changes necessary to make the code work with the MySQL database).
  
  
• Results produced with the MySQL version may be different from results produced in the Oracle version in the 6th-7th significant digits. This is normal, since C++ is only accurate to the 5th-6th digit. The handling of digits beyond that are sensitive to the particular processor being used, and since the conversion involved moving to a new processor, some differences are to be expected.
  
  
• The MySQL version of TRIM3 is no longer case sensitive. In other words, it will consider a run named "SS2000" to be the same as a run named "ss2000".
  
  
• The tool used to view program rules ("Rules of Simulated Programs") has been simplified. Rather than having to specify each simulation, rule, and year to be displayed, the user simply specifies the desired simulation and the category of rules he/she is interested in. The result is a series of tables, one for each rule in that category with values for all years. These tables can then be copied into worksheets if desired.
  
  
• The tool used for viewing and extracting microdata, as well as the crosstab tool, have been replaced with a new "Microdata Examiner" tool. This takes advantage of a standard set of view/extract/crosstab tools developed for other UI projects, and brings more capabilities to the TRIM navigator. Of particular interest is a feature that lets you create a SAS file directly, rather than having to create an ASCII file first, and then converting that into a SAS dataset.
  
  
• The "Datasets" link to download TRIM data has been dropped, since the increased speed of MySQL makes it no longer necessary to have a special download option for large files. Instead, requests for large amounts of microdata can be handled just as quickly using the new "Microdata Examiner" tool. The other type of data that was available through this link -- summary tables of baseline runs -- should be accessed through the "Summary Tables" tool.

New baseline policy

If a version of the baseline has summary tables and microdata results, then the baseline has been reviewed, realigned if necessary, and is available for use. Otherwise, a program rule correction (or corrections) has been made to the baseline, but the new version has not yet been run and approved for use. Baseline simulations that have summary tables and microdata results are marked with an asterisk on the Simulation Runs web page.

Users should be aware that simulations with the same version number are not necessarily consistent. For example, version 3 of a Food Stamp simulation (FS1999_v3) would not necessarily use version 3 of the SSI simulation (SS1999_v3) as input. To see which SSI simulation was used as input to FS1999_v3, users should select FS1999_v3 on the Simulation Runs web page and click on the "Change Inputs" link.

Medicare baselines for 1993-1998 replaced

In the 1993-1998 Medicare runs that were previously available on Development, the baseline values for these rules were based on the 1992 Census report "Measuring the Effect of Benefits and Taxes on Income and Poverty". For the 1993-1995 baselines, adjustments were made to these 1992 values to generate data for each year. For the 1996-1998 baselines, the values for 1995 were simply reused with no adjustments.

For the 1999 baseline, a new source was used -- the annual HCFA report "Health Care Financing Review Statistical Supplement". The decision was made to rerun the 1993-1998 Medicare baselines with data from this source, and replace the previously available runs. The only effect this change has is in the Medicare result variable "AnnualBenefits", and those portions of the summary tables which show benefit amounts. Since this information is not published as part of the baseline results, and is not used as input by any other simulation (the Medicaid simulation uses simulated Medicare enrollment information, but not benefit amount information), it was decided that having a consistent series of values for these rules was more important than maintaining previous baseline results.

Note: for the 1996-1998 baselines, the names of the Medicare baseline runs were changed from mrbMR1996, pdjMR1997, and pdjMRAlien1998 to MR1996, MR1997, and MRAlien1998, respectively. Users with Medicaid runs for these years should consider changing their run setups to refer to these new names.

For more details on the changes and the new data source, contact the TRIM staff (T3WEB@ui.urban.org).

Medicaid 1996 baseline replaced

• Correcting how the rule DoesAFDCEligQualifyForMedicaid is implemented. This causes an increase in eligibility type #11.
• Correcting how pregnancy is imputed. This causes fewer pregnant women to be found, thereby lowering pregnancy-related eligibility.
• Correcting how the Special State rules are implemented in Tennessee and Hawaii. This increases the eligibility in those states. It also excludes elderly from this eligibility type, thus decreasing eligiblity in Deleware and Oregon.
• Moving calculations from forms to C++ code causes some borderline pct-of-pov cases to switch from eligible to ineligible.
• Changing the AvgMedicaidExp rules (particularly for elderly and disabled). This is due to a change in the method used to adjust enrollment and spending between blind/disabled and elderly. This results in a much lower spending per enrollee for the elderly population.

Note that the many changes and improvements that have been made to the medicaid module over the last couple of years, plus changes in methodologies for computing baseline targets, have resulted in inconsistencies between baselines for 1995 and earlier (which use older versions) and baselines for 1996 and later (which use the latest versions). Consequently, analysts should be wary of making comparsions of baseline results that span 1995-1996 (furthermore, when comparing 1996 to 1997 and later, be sure to use the FYE-aligned version of the 1996 baseline). For details on the inconsistencies, contact the TRIM staff.

SSI Version 3.3 Released

On 9/27/01, version 3.3 of SSI was released to all servers. The changes since version 3.1 are

• the result variable "Part_CitizenshipCategory" is written to both spouses for married couples (previously it was written only to one spouse)
• the variable "Part_CitizenshipCategory2" was added (see dictionary for details)
• table A7 was modified to include all disabled persons under 18 (previously it just included disbaled children under 15)

AFDC Version 5.5 Released

On 9/19/01, version 5.5 of AFDC was released to all servers. This corrects an error in the participation function. Previously, when deciding which months should be considered reporter months, months with the same probabilities of participation were chosen in calendar-year order. This created more reporters in early months than later months. Version 5.5 corrects this so months of reporting are allocated more evenly across the months.

In addition, the following changes were made:

• The variablelist rule AllocFlag665 is now checked to see whether the receipt of AFDC was allocated.
• Add option #5 ("Illinois Method") to DeemAllocUseNeedStandard.
• Add option #6 ("Connecticut Method") and #7 ("Virginia 1998 Method") to "IncomeTestsOption".
• Modified option #11 of EarningsDisregardFormula to use EarningsDisregardFixed2.
• Add option #14 ("Mississippi 1999 Method") to EarningsDisregardFormula.

Medicaid 'no CHIP' baselines for 1997 and 1998 released to Romulus

On 8/27/01, baseline runs "MD1997NoCHIP" and "MDAlien1998NoCHIP" were released to the Romulus (and public) servers. These runs are identical to the baselines on Thetis except CHIP eligibilty was not simulated, and alignment was redone using these modified eligibility estimates. All CHIP-related rules in these runs were set to 0.

PubOrSubsidizedHousing Version 1.5 Released

Changes since version 1.3:

Version 1.4 (revised as following for NSAF):

• Reinstate processing of all units
• Create a new HouseholdWeight output variable for use in forms tables (Returns HouseholdWeight for CPS, FamilyWeight of first weighted family for NSAF)
• Create new NumberPersonsInHh and NumberChildrenInHh output variables for use in forms tables (Count all CPS persons, only persons having FamilyWeight in NSAF)
• In households having at least one person who is not a part of an NSAF social family, allow the imputation of number of bedrooms, counting only persons having a family weight; compare with reported number; use smaller of the two.

• For NSAF data, for non-public/non-subsidized housholds, sum up MonthlyMortgageOrRent, rather than taking it from just one person.
• Deleted temporary output fields that were created for testing/analysis in version 1.2:
  • NSAFEarningsIndex
  • IsSubsidized
  • ReportedRent
• Add per-person result variables:
  • AnnualRentPaidPerPerson
  • AnnualSubsidyReceivedPerPerson
  • MonthlyFairMarketRentPerPerson
  • MonthlyHousingExpensePerPerson
  • MonthlyRentPaidPerPerson
  • MonthlySubsidyReceivePerPerson

Medicaid 1997 and 1998 baselines replaced, version 9.0 released.

On 8/23/01, the 1997 medicaid baseline was replaced on Development and Thetis (none exist on romulus) and the 1998 baseline was replaced on Development and added (for the first time) to Thetis (none exist on romulus). Both these replacements use the latest versions of medicaid (9.0), which incorporates enhancements and corrections, and are thus more accurate than the previous baselines. The old versions of the baselines can be found on Development under the names MD1997Old and MDAlien1998Old. For the 1997 baseline, the more important changes were:

• Correcting how the rule DoesAFDCEligQualifyForMedicaid is implemented. This causes an increase in eligibility type #11.
• Correcting how pregnancy is imputed. This causes fewer pregnant women to be found, thereby lowering pregnancy-related eligibility.
• Adding special state programs to Vermont and Massachusetts, and making the programs in Hawaii and Minnesota more liberal.
• Increasing the age for the Poverty-based rules (set #2) by a year for most states.
• Changing the AvgMedicaidExp rules (particularly for elderly and disabled). This is due to a change in the method used to adjust enrollment and spending between blind/disabled and elderly. Previously we had used 10% and 20% to move "old" disabled enrollees (over 65) and associated spending to elderly. However, when looking at the HCFA data, these percents were not representative of the actual data, so we replaced them with numbers estimated from the HCFA data. The data used for new targets uses 8.3% and 9.5% for adjusting enrollees and spending, respectively. So, this results in a much lower spending per enrollee for the elderly population.

Version 2.6 of ChildCareExp released.

Version 2.6 of ChildCareExp is now available on all servers. Below is a list of changes since version 1.7:

V1_9:
Added copay program rules.
Bug fix: changed 5% deduction (option 3 for program rule EarnedIncomeDeductionFormula from .5 to .05 in the code. (This program rule option was never used, so no runs were affected, and the old code versions were retired anyway, but now at least the bug is fixed.)

V2_0:
Replaced Program Rules InitialEligibilityLimit and ContinuingEligibilityLimit with LowerEligibilityLimit and HigherEligibilityLimit. Added program rules HigherEligibilityReason and HrsWorkOption. Added options to IncomeDisregardFormula and EarnedIncomeDeductionFormula.

V2_1:
Added Table C19.

V2_3:
Updated options for 1999 Copay program rules. Added Table C20. Changed summary tables to reflect new "Assigned a Copayment" column.

V2_4:
Made changes in Expense portion to use NSAF expense variable when applicable.

V2_5:
Changed Table J1 such that families with no income would not show up on the table at all (i.e. percentages do not add up to 100%).

V2_6:
Added AnnualChildCareExpense variable.

Version 3.1 of SSI released.

Version 3.1 of SSI has been released to all servers. This version incorporates a correction to how the allocation flag "Alloc665" is used. In version 3.0, it was looked at whenever the SSI allocation flags were looked at. In 3.1, it is NOT looked at when determining if a person received disability income.

SSI Version 3.0 released.

On 25 May 2001, version 3.0 of SSI was released. The variablelist rule AllocFlag665 is now checked to see whether the receipt of SSI was allocated.

AlienPrep Version 2.0 Released

On 9 May 2001, version 2.0 of AlienPrep was released, incorporating the following changes:

• Previously, the DeemingYears value for pre-reform immigrants was the lesser of the values stored for pre- and post-reform. Now only the pre-reform value is used.
• Previously, a value of 99 for DeemingYearsSSI, DeemingYearsFoodStamps, etc. turned off exemption to sponsor deeming. It now has no special treatment.
• The variablelist rule AllocFlag665 is now checked to see whether the receipt of SSI, AFDC/TANF, food stamps, or Medicaid was allocated.

AFDC Version 4.3 Released

On 21 February 2001, version 4.3 of AFDC was released. This corrects a bug in the previous version, which was causing MFIP processing to be bypassed when BenefitComputationOption=9 was specified. This version also corrects a slight error in the way non-earner units were being handled when BenefitComputationOption=9 was specified. The maximum benefit was being used rather than the payment standard.

AFDC Version 4.2 Released

On 14 February 2001, version 4.2 of AFDC was released to all servers. It differs from 4.1 only in its treatment of NSAF input, bypassing a rare family relationship coding inconsistency appearing in round 2 NSAF data.

PubOrSubsidizedHousing 1.3 Released

On 14 February 2001, version 1.3 of PubOrSubsidizedHousing was released to all servers. it differs from 1.2 only in its treatment of NSAF input, skipping units having no family weight.

FoodStamps Version 4.2 Released

On 14 February 2001, version 4.2 fo FoodStamps was released to all servers. Changes from 4.0 include the addition of alignment table #9, and special treatment of NSAF input, skipping unweighted units.

SSI Version 2.5 Released

On 6 February 2001, version 2.5 of SSI was released to all servers. It differs from 2.4 only in its treatment of NSAF input, correcting the weights used for tables C6, S1, and S2.

SSI Version 2.4 Released

On 1 February 2001, version 2.4 of SSI was released to all servers. It differs from 2.3 only in its treatement of NSAF input, bypassing unweighted NSAF units.

AFDC Version 4.1 Released

On 1 February 2001, version 4.1 of AFDC was released to all servers. It differes form 4.0 only in its treatment of NSAF input. It skips unweighted NSAF units and uses FamilyWeight in lieu of PersonWeight of unit head for NSAF only.

AFDC/TANF 1998 Baseline Replaced

On 31 January 2001, the AFDC/TANF baseline was replaced on all servers. The previous version of the baseline was run using 3.0 of AFDC. The new baseline uses . Only the results for Minnesota and New Mexico were affected by this change.

Existing FederalTax Runs Modified to Use Version 1.5

On 30 January 2001, all existing FederalTax runs on all servers were modified so that if they are run again, they will use version 1.5 of FederalTax. This change was done because older versions of FederalTax are not compatible with the current version of the frame.

FoodStamps Version 4.0 Released

On 22 January, version 4.0 of FoodStamps was released to all servers. By using results from version of AFDC, this version of FoodStamps can implement a preliminary version of an MFIP-like program (i.e., combined cash and food benefits). Note that when using this version of FoodStamps, the variable list rules MonthlySimulatedAFDCPersonType, MonthlySimulatedAFDCEligFor, and PotentialFoodBenefitEligFor must be given values.

AFDC Version 4.0 Released

On 18 January, version 4.0 of AFDC was released to all servers. This version incorporates partial support for Minnesota's MFIP program (combined cash and food benefits). A new version of FoodStamps will be available shortly which, when used with version 4.0 of AFDC, will fully implement the MFIP program.

1993 Input Data Corrected

On 5 January 2001, the 1993 input data on all servers was replaced with a version that corrects the variable FamilySubfamiliesSeparate. This variable divides the household into family units, with related subfamilies counted separately from the primary family. The variable's values consist of four digits. The left two indicate family number within the household, the right two indicate family type. Unfortunately, on the 1993 input file this variable only contained the right two digits (i.e., the family type). The new versions of the 1993 input file correct this. The impact of this change should be minimal, since the only simulation that uses this variable is FoodStamps. The 1993 FoodStamps baseline has not been rerun to reflect this change.