Data Source and Processing
We obtained records for GARS plan participants via a FOIA request. Our data includes individual-level records for all plan active, inactive and beneficiary participants as-of May 13, 2015. Using the individual-level records we aggregate the data by age and compute for each age: the count of participants, the average level of earned benefits (annualized) and the percentage of participants who are tier 1. For survivors and beneficiaries we use the stated benefits. For active and inactive participants we calculate the earned benefits using the most recent salary and years of experience.
Forecasting Survivors and Beneficiaries
There are no tier 2 GARS beneficiaries so we use just the count and average benefits data to forecast the population and average benefits at each date. In each period we apply mortality assumptions and increase benefits by the input benefits growth rate. When a beneficiaries dies we assume 3/4 of a survivor enters the population at 2/3 of the average benefits of the original beneficiary.
Forecasting Active and Inactive Participants
We forecast tier 1 and tier 2 active and inactive participants separately since there are significant differences benefit levels, benefit growth rates and retirement rate assumptions between the two populations. Since these populations are not currently collecting benefits we forecast the number of future beneficiaries and their average level of benefits in each period of the forecast. To forecast the population we apply the mortality rate, and for participants over 55 the retirement rate. Prior to retirement for actives we increase the earned benefits by the input salary growth rate, for inactives we apply no increase before retirement.
When active or inactive participants retire we calculate the count of retirees and the average level of benefits at each age. Then for these beneficiaries we apply mortality assumptions, forecast survivors and grow benefits based on the tier of the participant.
Computing the Actuarial Liability
We compute both the nominal and present value of all benefits due by year using the input discount rate. We calculate the liability for each participant type separately using the expected number of beneficiaries and expected average benefits in each period.
Amortizing the UAAL
To amortize the unfunded liability we solve for the present value of the actuarial liability subtract the actuarial value of assets and solve for the level payment. In the level payment calculation we use the input discount rate and the input number of periods. If the number of periods is set to 0 we assume the UAAL will be paid off immediately at it’s current present value. If the discount rate is set to 0 we assume the nominal liability will be paid off over the specified number of periods.
Data Source and Processing