Same-Year County Spending Averages Combine Routine Prevention Budgets With Outbreak Emergency Funds
Splitting emergency outbreak fund surges from steady base prevention budgets across county years.
Health spending and county death rates move both ways when an outbreak hits. State and federal relief grants flow to counties where death rates jump. Splitting outbreak grant surges from base care shows how steady local health funds save lives over time.
Across 943 pre-pandemic California county-years (2003 to 2019), routine-budget models isolate 2.38 fewer deaths per 100,000 per $10/resident annually ($420,168 per life saved, 32.4:1 return), building to -4.32 over 13-year cumulative horizons ($231,481 per life, 58.8:1 return) and -5.31 in multi-year routine-budget models.[1][2] During 2020 to 2023, emergency COVID-19 grants surged 64.2% and show a +2.47 slope because relief funds followed pandemic death spikes, pulling the pooled 2003 to 2023 average to -0.25.[2] Checking both the -2.38 annual floor and the -4.32 to -5.31 cumulative slopes protects multi-year grants from early cuts.[3]
Table 1. Splitting Outbreak Feedback From Annual and Cumulative Routine-Budget Effects
Comparing 2020-2023 pandemic surge feedback against the pre-pandemic annual county panel, 13-year cumulative differences, and 13-year routine-budget IV models.
| Budget Measure |
Pandemic Surge
COVID-19 Surge Window
|
Annual Baseline
Pre-Pandemic Annual Panel
|
13-Yr Span
13-Year Cumulative Change
|
Routine-Budget IV
Multi-Year Formula Check
|
|---|---|---|---|---|
| Panel Sample & Method | COVID-19 PanelN = 220 | Pre-COVID PanelN = 943 | 13-Yr Cumulative ModelN = 57 | 13-Yr Routine-Budget IVN = 57 |
| Active Path in Model | Two-WayDeaths ←> Surge Aid | Routine Budget -> Annual Health | Sustained Budget -> 13-Yr Health | Instrumented Budget -> 13-Yr Health |
| Outbreak Surge Handling | +64.2% emergency grant surge | Separates 2020-2023 COVID era | 13-yr pre-LGRS cumulative change | Variance-based routine-budget check |
| Mortality Shift per 100,000 | +2.47Pandemic Surge | 2.38 fewer deaths / 100Kannual | 4.32 fewer deaths / 100K13-yr | 5.31 fewer deaths / 100Kroutine IV |
| Cost per Life ($10 Scale)* | Not definedpositive slope | $420,168 per life saved | $231,481 per life saved | $188,324 per life saved |
| Return per $1 Spent ($13.6M Life Value)* | Reads as 0:1reverse feedback | 32.4:115.0:1 Low to 49.3:1 High | 58.8:127.2:1 Low to 89.4:1 High | 72.2:133.5:1 Low to 109.9:1 High |
| Statewide Value at 39M Pop ($10 Scale)* | Surge aid tracks death waves | 928 lives$12.6B, annual | 1,685 lives$22.9B, 13-yr | 2,071 lives$28.2B, routine IV |
Table 1 Notes & * Modeled Estimates
Expand notes
Table 1 Notes (Mode B Panel Estimates; Mode A Valuations): Pandemic 2020-2023 TWFE OLS is +2.47 deaths per 100,000 per $10/capita (β = +0.247 per $1, SE = 0.075, p = 0.0010, N = 220), while pooled 2003-2023 TWFE OLS is -0.25 per $10 (β = -0.025, SE = 0.067, p = 0.710, N = 1,163). Pre-pandemic 2003-2019 TWFE OLS is -2.38 per $10 (β = -0.238, SE = 0.078, p = 0.0022, wild cluster bootstrap p = 0.007, N = 943; urban -2.59, p < 0.0001; smaller/rural -2.42, p = 0.0136). Single-regime 13-year cumulative long-difference OLS (2003-2016, N = 57) is -4.32 per $10 (SE = 1.54, p = 0.0052; 16-year 2003-2019 excl. 5 LGRS outliers is -5.91, p = 0.0330). Single-regime 13-year Long-Difference Lewbel 2SLS (2003-2016, N = 57) is -5.31 per $10 (SE = 1.67, p = 0.0015, first-stage F = 111.20, Hansen J p = 0.2132; 16-year 2003-2019 Long-Difference 2SLS is -11.49, p = 0.0009, F = 13.09).
Column Scope. COVID-19 Surge Window. County Panel 2020-2023, N = 220. Pre-Pandemic Annual Panel. 2003-2019 Subsample, N = 943. 13-Year Cumulative Change. Single-Regime 2003-2016, N = 57. Multi-Year Formula Check. 13-Yr Routine-Budget Model, N = 57.
Model Crosswalk. Plain table labels map to standard study models. Raw Yearly Average / Simple Avg is an unadjusted OLS model where routine and outbreak budgets mix. County & Year FE is a Two-Way Fixed Effects (TWFE) panel model. Routine-Budget Analysis / 21-Year County Budget Model is a Two-Stage Least Squares (2SLS / Lewbel 2012) model that isolates routine base budget shifts. County Board Budget Rules uses Brown (2014) county board rules.
Two-Ruler Benchmark Summary. Mode B Panel (57 Counties, N = 943 Pre-COVID & N = 220 COVID-19 County-Years, 2003 to 2023). Ruler 1: Pandemic Emergency Surge Window (2020-2023): +2.47 Deaths / 100K (4.85-Point Sign Flip). When pandemic death waves hit in 2020 to 2023, emergency grants pushed county health spending up 64.2% ($99.65 to $168.70/capita). The 2020-2023 slope is +2.47, sitting 4.85 points above the -2.38 pre-COVID slope and attenuating pooled 2003-2023 TWFE to -0.25. Ruler 2: Routine-Budget Horizons (Annual TWFE & 13-Yr Cumulative): -2.38 Annual | -4.32 Cumul. | -5.31 2SLS / 100K. Tracking pre-pandemic routine budgets isolates 2.38 fewer deaths/100K annually ($420,168/life, 32.4:1 BCR). Over 13-year cumulative horizons, chronic-disease prevention builds to -4.32 in OLS ($231,481/life, 58.8:1 BCR) and -5.31 in 2SLS ($188,324/life, 72.2:1 BCR).
Method & Data Audit (TooEarlyToSay.com): Public Health Spending Regime Analysis →
Table 2. Managing Budget Cycles When Outbreak Surges Cloud Base Gains
How separating reactive outbreak aid across calm and pandemic windows guides county budget rules.
| Choice Factor | Pandemic Surge Window | Pre-Pandemic Annual Baseline | 13-Year Cumulative Horizon | 13-Yr Routine-Budget IV |
|---|---|---|---|---|
| Feedback & Sample Path | Emergency COVID-19 grants rose 64.2% ($99.65 to $168.70/capita) as death rates spiked. | Calm-year budgets test steady disease control in both urban and rural counties. | Sustained 13-year changes capture compounding heart and chronic disease gains. | 13-year routine-budget IV corrects for residual feedback with strong variance-based instruments. |
| Why 1-Year Audits Misread This Stage | Pooling 2020-2023 emergency grants with routine budgets attenuates the full-panel slope to -0.25. | Single-year models capture only 55% of the 13-year cumulative mortality reduction. | Short one-year pilots stop chronic disease and high blood pressure control before multi-year gains mature. | Single-year IV has weak first stages, whereas 13-year routine-budget IV is strongly identified. |
| Budget Tradeoff | Cutting base budgets after emergency grants expire removes staff needed for chronic disease control. | Even in Year 1 alone, the -2.38 pre-pandemic baseline yields a 32.4:1 return. | Over 13-year cumulative horizons, the -4.32 slope yields a 58.8:1 return. | At the -5.31 routine-budget IV slope, a statewide $10/resident outlay implies ~2,071 lives saved. |
| Best-Fit Budget Rule | Split Outbreak Aid From Base Budgets. Separate emergency pandemic grants from routine prevention before scoring county results. | Use -2.38 (32:1) as Annual Floor. Anchor conservative 1-year budget memos in the -2.38 pre-pandemic baseline (32.4:1 return). | Report Cumulative Multi-Year ROI. Pair the annual floor with the -4.32 13-year cumulative return (58.8:1) in board reports. | Use Multi-Year Grant Cycles. Fund county health grants in multi-year blocks so chronic-disease prevention can compound. |
Table 2 Notes: Estimator Details, Pandemic Regime Shift, Exposure Horizons & Provenance
Expand notes
• Estimator Details & Robustness: In the audited 57-county panel (Value_of_Public_Health/paper/ajph_submission.tex), the primary 2003-2019 pre-COVID TWFE OLS specification (N = 943) yields 2.38 fewer deaths per 100,000 per $10/capita (β = -0.238, SE = 0.078, p = 0.0022, wild cluster bootstrap p = 0.007). The effect holds in both large urban counties (>500,000 residents: -2.59, p < 0.0001) and smaller/rural counties (≤500,000 residents: -2.42, p = 0.0136), and in the single-regime 2003-2016 SCO schedule (-2.62, p = 0.0128).
• Pandemic Regime Reversal & Multi-Year Long Differences: In the system Mortality_it = β·Spending_it + γ·X_it + α_i + τ_t + ε_it, the 2020-2023 COVID-19 emergency window (N = 220) saw a 64.2% spending surge ($99.65 to $168.70 per capita) directed toward pandemic mortality waves, producing a positive slope of +2.47 per $10 (β = +0.247, SE = 0.075, p = 0.0010) and a 4.85-point regime shift above the pre-pandemic baseline (-2.38). Over multi-year cumulative horizons, 13-year long-difference OLS (2003-2016, N = 57) yields -4.32 per $10 (SE = 1.54, p = 0.0052), 16-year long-difference OLS (2003-2019 excl. 5 LGRS boundary outliers, N = 49) yields -5.91 per $10 (SE = 2.77, p = 0.0330), and 13-year Long-Difference Lewbel (2012) 2SLS yields -5.31 per $10 (SE = 1.67, p = 0.0015, first-stage F = 111.20, Hansen J p = 0.2132; 16-year Long-Difference 2SLS = -11.49, p = 0.0009, F = 13.09).
• Unit Convention ($1 per Capita Regression Tables vs. $10 per Capita Policy Scale): In Value_of_Public_Health/paper/ajph_submission.tex (Tables 2-3 and Online Appendix Table A3), regression slopes β are reported per $1 per capita (-0.238 annual TWFE, -0.432 13-year long difference, -0.531 13-year Long-Difference 2SLS) alongside the policy-scaled effect per $10 per capita (10 × β = -2.38, -4.32, and 5.31 fewer deaths per 100,000). Because both scales represent the exact same underlying ratio ($10 × 100,000 / 2.38 = $1 × 100,000 / 0.238 = $420,168 per statistical life saved), cost per life saved and Benefit-Cost Ratios (32.4:1 annual, 58.8:1 13-year cumulative, 72.2:1 Long-Difference 2SLS at $13.6M VSL) are invariant to whether the table is read per $1 or per $10.
• Data Provenance: Panel estimates (+2.47 COVID-19 surge, -0.25 pooled 2003-2023, -2.38 pre-COVID annual TWFE, -2.42 smaller/rural, -2.59 urban, -4.32 13-year cumulative OLS, -5.91 16-year cumulative OLS, -5.31 and -11.49 Long-Difference Lewbel 2SLS) and VSL inputs come from src/frontend/js/lha-calculator.js, public/tools/lha-calculator/method.html, and Value_of_Public_Health/paper/ajph_submission.tex.
• Extended Cell Context. Feedback & Sample Path (Pre-Pandemic Annual Panel). Calm-year budgets (2003-2019) test steady disease control in both urban (-2.59) and rural (-2.42) counties. · Why 1-Year Audits Misread This Stage (Pre-Pandemic Annual Panel). Single-year models capture only 55% of the 13-year cumulative mortality reduction (-2.38 vs. -4.32). · Budget Tradeoff (Pre-Pandemic Annual Panel). Even in Year 1 alone, the -2.38 pre-pandemic baseline yields a 32.4:1 return ($420,168 per life saved). · Budget Tradeoff (13-Year Cumulative Change). Over 13-year cumulative horizons, the -4.32 slope yields a 58.8:1 return ($231,481 per life saved). · Budget Tradeoff (Multi-Year Formula Check). At the -5.31 routine-budget IV slope, a statewide $10/resident outlay implies ~2,071 lives saved ($188,324 per life).
Splitting Outbreak Aid From Steady Prevention Gains
Every public health budget works inside a feedback loop during public health emergencies. When pandemic death waves hit in 2020 to 2023, federal and state relief grants boosted county health spending by 64.2% ($99.65 to $168.70 per capita), producing a positive +2.47 slope per $10 (p = 0.0010). Because emergency aid chased pandemic death spikes, pooling 2020 to 2023 with routine years attenuates the full-panel county slope to -0.25 (p = 0.710).
Health economics separates routine pre-pandemic prevention budgets (2003 to 2019, N = 943) from emergency outbreak surges and tracks how chronic-disease prevention compounds over time. Annual within-county panel models isolate 2.38 fewer deaths per 100,000 per $10 ($420,168 per life saved, $32.40 return per $1), holding in both urban (-2.59) and smaller/rural (-2.42) counties. Over 13-year cumulative horizons, the effect nearly doubles to -4.32 in cumulative differences ($231,481 per life, $58.80 return per $1) and -5.31 in 13-year routine-budget IV models ($188,324 per life, $72.20 return per $1). Showing boards both the annual baseline and the multi-year cumulative return helps keep effective programs funded across budget cycles.
Data Sources, Model Inputs, Notes & Endnotes
(Numbered endnotes [1]–[5], state data sources & math rules)
Expand details
Case Study Focus: System Dynamics & Feedback · Needs-Driven Spending & Simultaneity (Mode B Panel)
Numbered Endnotes & Main Sources
- [1]California County Public Health Spending Panel (2003-2023): California State Controller County Financial Transactions linked to CDPH Vital Statistics death records across 57 counties. N = 1,163 county-years (N = 943 in the 2003-2019 pre-pandemic window; N = 779 in 2003-2016 single-regime; N = 220 in the 2020-2023 COVID-19 window).
- [2]Needs-Driven Feedback vs. Routine Budget Isolation: Lewbel, A. (2012), Journal of Business & Economic Statistics, 30(1), 67-80 provides the heteroskedasticity-based IV framework. The CAPHE California 57-county LHA panel finds -2.38 deaths/100K per $10/capita in 2003-2019 annual TWFE OLS ($420,168/life, 32.4:1 BCR), -4.32 in 13-year cumulative long-difference OLS ($231,481/life, 58.8:1 BCR), and -5.31 in 13-year Long-Difference Lewbel 2SLS (first-stage F = 111.20, $188,324/life, 72.2:1 BCR), versus +2.47 in the 2020-2023 pandemic surge window (p = 0.0010) and -0.25 in pooled 2003-2023 TWFE (p = 0.710).
- [3]Independent Governance Check & Valuation: Brown, T. T. (2014), Economics & Human Biology, 13, 34-45 finds protective effects of county health spending in California. HHS ASPE 2025 VSL Guidelines sets a $13.6M central value.
- [4]CDPH Vital Statistics Mortality Data
- [5]HHS ASPE 2025 VSL Guidelines
Mode Classification (Mode B Panel Estimates; Mode A Valuations in Table 1): All regression slopes (+2.47 COVID-19 window, -0.25 pooled 2003-2023, -2.38 pre-pandemic 2003-2019 TWFE OLS, -2.42 smaller/rural, -2.59 urban, -4.32 13-year cumulative OLS, -5.91 16-year cumulative OLS, and -5.31 / -11.49 Long-Difference Lewbel 2SLS) and 57-county California population baselines are observed Mode B inputs from src/frontend/js/lha-calculator.js, public/tools/lha-calculator/method.html, and Value_of_Public_Health/paper/ajph_submission.tex. Table 1 valuations apply the HHS 2025 VSL schedule ($6.3M Low, $13.6M Central, $20.7M High) to those observed slopes.
Primary References & Data Sources: Lewbel, A. (2012), JBES, 30(1), 67-80; Brown, T. T. (2014), Economics & Human Biology, 13, 34-45; California State Controller County Financial Transactions; CDPH Vital Statistics Mortality Data; and HHS ASPE 2025 VSL Guidelines.
Companion Method & Code Guides (TooEarlyToSay.com)
- Public Health Spending: When Emergency Aid Flips the Sign. Across 57 California counties from 2003 to 2023 (1,163 county-years), each dollar per capita corresponds to -0.238 fewer deaths per 100,000 in 2003 to 2019 versus +0.247 in 2020 to 2023.
Common Questions & Method Notes
- What is "reverse causation" in public health budgets?
- It happens when the health outcome (death or disease rates) drives changes in the budget input (public health spending). During the 2020-2023 COVID-19 pandemic, emergency relief grants raised county public health budgets by 64.2% in response to death spikes, producing a +2.47 slope per $10 (p = 0.0010) that masks the -2.38 pre-pandemic protective effect when the two regimes are pooled.
- How do the $1 per capita regression coefficients and $10 per capita policy figures relate?
- In the working paper (ajph_submission.tex), regression slopes β are reported per $1 per capita (-0.238 annual TWFE, -0.432 13-year cumulative OLS, -0.531 Long-Difference 2SLS) alongside the scaled impact per $10 per capita (-2.38, -4.32, and 5.31 fewer deaths per 100,000). Both express the exact same cost-effectiveness ratio ($420,168 per life saved annually and $231,481 over 13-year cumulative horizons).
- Does county public health spending still pay off if a board uses only the Year 1 annual baseline?
- Yes. Using the conservative 2003-2019 annual TWFE estimate of 2.38 fewer deaths per 100,000 per $10/capita yields $420,168 per life saved and a 32.4:1 benefit-cost ratio at the $13.6M HHS central VSL (15.0:1 at the $6.3M Low VSL and 49.3:1 at the $20.7M High VSL).