This repository integrates multiple county-level datasets:

• Mortality Data:
  National Vital Statistics System (NVSS) Multiple Cause of Death files (2010–2022), restricted to suicide deaths identified using ICD-10 codes: X60–X84, Y87.0. Deaths were aggregated annually and standardized per 100,000 population using Census denominators.

• Social Connectedness:
  Meta’s Social Connectedness Index (SCI) (2020), quantifying relative probabilities of Facebook friendship ties between counties.

• Sociodemographic and Economic Covariates:

  • Agency for Healthcare Research and Quality (AHRQ) community-level indicators (2010–2020).
  • American Community Survey (ACS) 5-year estimates for years not covered by AHRQ.
  • Variables include: population density, age distribution, racial/ethnic composition, median household income, unemployment rate, educational attainment, and limited English proficiency.

• H1: A one–standard deviation (1-SD) increase in the SCI-weighted average suicide mortality rate in socially connected counties is positively associated with suicide mortality in the focal county, controlling for spatial exposure.
• H2: A one–standard deviation (1-SD) increase in ERPO (Extreme Risk Protection Order) social exposure is negatively associated with suicide mortality in the focal county.

The SCI-weighted average suicide mortality rate in counties socially connected to county i at time t is:

s_{-it} = Σ_{j ≠ i} w_{ij} y_{jt}

where the weights are:

w_{ij} = [n_j × SCI_{ij}] / Σ_{k ≠ i} [n_k × SCI_{ik}]

• y_{jt}: suicide mortality rate in county j at time t
• n_j: population of county j
• SCI_{ij}: Social Connectedness Index between counties i and j

The spatially weighted average suicide mortality rate in counties geographically close to county i at time t is:

d_{-it} = Σ_{j ≠ i} a_{ij} y_{jt}

where the weights are:

a_{ij} = [1/d_{ij}] / Σ_{k ≠ i} [1/d_{ik}]

• y_{jt}: suicide mortality rate in county j at time t
• d_{ij}: great-circle distance between centroids of counties i and j

The ERPO Social Exposure for county i at time t is:

ERPO_Social_Exposure_{it} = Σ_{s(i) ≠ s(j)} 1[ERPO in state s(j)]_t × [SCI_{ij} / Σ_h SCI_{ih}]

where:

• 1[ERPO in state s(j)]_t: indicator if state j has an ERPO law at time t
• SCI_{ij}: Social Connectedness Index between counties i and j
• The sum is over all counties j in a different state than i

The ERPO Spatial Exposure for county i at time t is:

ERPO_Spatial_Exposure_{it} = Σ_{s(i) ≠ s(j)} 1[ERPO in state s(j)]_t × [1/d_{ij} / Σ_{k ≠ i} (1/d_{ik})]

where:

• d_{ij}: great-circle distance between centroids of counties i and j
• The sum is over all counties j in a different state than i

The regression model for social and spatial influence is:

y_{it} = ζ₁ s_{-it} + ζ₂ d_{-it} + ζ₃ᵗ X_{it} + μ_i + φ_t + ε_{it}

where:

• y_{it}: suicide mortality in county i, year t
• s_{-it}: SCI-weighted average suicide mortality rate in socially connected counties
• d_{-it}: spatially weighted average suicide mortality rate in geographically close counties
• X_{it}: time-varying covariates
• μ_i: county fixed effects
• φ_t: year fixed effects
• ε_{it}: error term

The regression model for ERPO social and spatial exposure is:

y_{it} = θ₁ ERPO_Social_Exposure_{it} + θ₂ ERPO_Spatial_Exposure_{it} + θ₃ᵗ X_{it} + φ_i + γ_{st} + ε_{it}

where:

• ERPO_Social_Exposure_{it}: ERPO social exposure for county i at time t
• ERPO_Spatial_Exposure_{it}: ERPO spatial exposure for county i at time t
• X_{it}: time-varying covariates
• φ_i: county fixed effects
• γ_{st}: state-by-year fixed effects
• ε_{it}: error term

• Statistical significance threshold: (p < 0.05).
• Results are reported with cluster-robust standard errors at the state level.

• nvss_indirect_exposure_2010_2022.R
  Contains implementation of the regression models testing main hypotheses (equation 3, equation 8).

• sensitivity_analysis_with_age_adjusted_suicide_.R
  Implements robustness checks using age-adjusted suicide mortality as the dependent variable (Supplementary Section).

• population_estimate_by_age_group_2010_2022.R
  Data pipeline for generating age-group-specific population estimates for denominator alignment.