Rollback Scenarios for SNAP (also applied to WIC & NSLP)
Full rollback to 2019 levels → 17% decrease in benefits per person. (source here from AO). 17% comes from rolling back benefits to the 2019 TFP baseline and adjusting for inflation using CPI. In this case, the rollback assumes that benefits are reverted to pre-pandemic (2019) levels and then trended forward to account for how food prices have risen between 2019 and 2024. This inflation-adjusted rollback reflects a scenario where SNAP benefits are returned to the 2019 formula and adjusted for food price changes, resulting in smaller cuts than the Urban estimates. I created this rollback based on earlier conversations with AO where it was discussed that TFP would be revered to 2019 levels, so I pulled TX av SNAP benefits and then adjusted forward.
TFP rollback → 21% decrease based on UI’s artificial reduction here. This reflects scenario modeled by the Urban Institute and CBPP, in which only the 2021 TFP reevaluation is reversed, returning SNAP benefits to their pre-2021 baseline while retaining inflation adjustments. This rollback would remove the modernization of the SNAP benefit formula but still allow for standard annual inflation updates moving forward. This scenario does NOT assume the full undoing of inflation-linked benefit increased. https://www.urban.org/data-tools/snap-cuts-state-county-gaps — This scenario reflects a static rollback.
Minimum rollback→ 10% decrease for comparison’s sake
Note: AO provided thisGAO report which mentions a 21% increase in program costs. This study wasn’t necessarily about rollbacks, but rather program inefficiencies. It coincidentaly aligns with the UI reduction estimate.
Medicaid Cut Scenarios
Based on EPI analysis here. Assumes an $880 billion cut to Medicaid/cut of roughly 10.7% to the programs spending going forward (CBO). See also herefor value of Medicaid and hh income. Data on proposed cuts from EPI here. For scenarios, reference Table 1 which provides an estimate of how much the proposed Medicaid cuts would cost the bottom two-fifths of household incomes in every state. It also includes estimates for the middle fifth of U.S. households, who also suffer a smaller, but not trivial, cut to their household incomes as well.
Methodology:
SNAP: To determine which average benefit per person to use, first I pulled the 2019 and 2024 national/state level summaries zip file from the USDA website here.
2019 average benefits per person: $124.64
2024 average benefits per person: $188.19 (this number checks out with CBPP Texas profile here)
Calculating impact per program was a simple mathematical formula across the three different scenarios. Not complex. The monthly per person benefit was multiplied by the reduction percentages, then scaled to an annual loss per person: annual loss = snap benefit per person * reduction scenario *12. A similar method was applied for hh calculation.
An adjustment was made for high cost food counties (40% higher meal costs) was included to estimate additional losses in counties with higher food costs.
Medicaid: To estimate the impact of Medicaid reductions, I used data from EPI to generate scenarios.
Medicaid recipients were divided into three income groups based on EPI estimates of household income distribution. Each group experiences different Medicaid loss rates, based on EPI estimates (see image above). So, calculation was medicaid loss per person = income level * loss rate.
Reduction: Instead of using fixed reduction percentages (like SNAP, etc.), Medicaid losses were modeled using income-based loss rates, which better reflect how medicaid cuts impact different income groups.
Note: The per capita approach was considered, but I didn’t use it because it assumes federal Medicaid growth caps rather than estimating immediate financial loss for the enrolled. I reviewed thisKFF analysis. Their analysis is useful for understanding federal budget constrains and state fiscal adjustments over a decade or more, but the EPI model is better suited to determine how much money a Medicaid recipient in Texas would lose per year.
Data Sources
Total Texas SNAP recipients (2024): 3,193,000. Source: here
Average SNAP benefit per person per month: $188(2024). Source: same as above
Average SNAP benefits in 2019 national: $123 (CBO)
# --- Medicaid Data ---medicaid_total_tx <-4135780income_levels <-c(Low =19307, Second =46351, Middle =76045)medicaid_loss_rates <-c(Low =0.074, Second =0.017, Middle =0.004)tx_medicaid_distribution <-c(Low =0.60, Second =0.25, Middle =0.15)medicaid_loss_per_person <- income_levels * medicaid_loss_rates# --- Individual Medicaid Loss by Quintile ---medicaid_loss_by_quintile <-data.frame(Income_Group =names(income_levels),Income = income_levels,Medicaid_Loss_Per_Person = medicaid_loss_per_person) %>%mutate(across(where(is.numeric), ~formatC(.x, big.mark =",", format ="f", digits =2)))print(medicaid_loss_by_quintile)
Income_Group Income Medicaid_Loss_Per_Person
Low Low 19,307.00 1,428.72
Second Second 46,351.00 787.97
Middle Middle 76,045.00 304.18
# --- Total Medicaid Statewide Loss (EPI Quintiles) ---total_medicaid_loss_tx <-sum(medicaid_total_tx * tx_medicaid_distribution * medicaid_loss_per_person)cat("Total Statewide Medicaid Loss (EPI Quintile-based):", formatC(total_medicaid_loss_tx, big.mark =",", format ="f", digits =0))
Total Statewide Medicaid Loss (EPI Quintile-based): 4,548,735,772
Individual Level Scenarios
AO wanted individual level scenarios
Note: Medicaid losses were estimated based on household incomes as recommended by the Economic Policy Institute (EPI), applying loss percentages derived from household income quintiles. Separate calculations for elderly household members were intentionally omitted to avoid double counting, as household income already encompasses all members.
Note: I went back to giving everyone the lower end of the quintile per the EPI Medicaid analysis. I went back and forth on this one because we want to be able to capture what the loss is for larger households. So, while the EPI threshold “smoothes” out the household size, I originally wanted to start from the assumption that the 19307 was for a single parent household income. Then, I had added a multiplier (x1.5) to adjust the income upwards for the two parent household incomes, but to still keep them in the bottom quintile. KW reviewed and SS also agree that we should keep at 19307. I maintain that a superficial adjustment is the best way to go. SS suggested pulling PUMS but the loss rates are calibrated for those quintile averages, so diverging from that base might create methodological inconsistencies. See the medicaid assumptions in prior chunks.
