{
 "version":89,
 "status":"fixed_spec_external_source_mechanism_check_not_exact_SIPP_replication",
 "years":[2024,2025],
 "link":"One-to-one exact shedid across annual files; use 2025 panel_weight > 0 only. Never substitute cross-sectional weights.",
 "domain":"Respondents age 25-60 in 2024, reporting paid/profit work (D1A=1) and no layoff/job loss in preceding 12 months (D44_f=0). Includes self-employment; not an any-member household exposure domain.",
 "exposure":"2025 D44_f=1: respondent laid off or lost a job in preceding 12 months, including temporary layoff. A control requires 2025 D44_f=0 and D44_e=0 (no voluntary quit). Exposed respondents may also have quit a different job.",
 "resource":"2024 EF1=0 is low buffer: no designated emergency fund covering three months of expenses. This self-reported expense-based measure is not bank deposits below one month of income.",
 "outcome":"I(2025 FD3 in 3,4) - I(2024 FD3 in 3,4), each requiring valid codes 1-4. Food insufficiency in past month, not USDA six-item annual food insecurity or actual consumption.",
 "temporal_limit":"Annual fieldwork roughly one year apart but no dated layoff variable is used. The retrospective 12-month window may overlap the baseline interview by a short boundary interval. Exact pre-event resource timing is not certified.",
 "primary":"Complete valid core responses with all required SHED iflags=0. Baseline D1A,D44_f,EF1,FD3 and follow-up D44_f,FD3 always required; follow-up D44_e reporting required for controls only. Core reported does not certify that profile covariates are unimputed.",
 "sensitivity":"All valid edited core responses, including imputed ones; same years, domain and model. Always report even when less favorable. Never promote this to primary after seeing results.",
 "model":"2025 panel_weight WLS: change_food_insufficiency ~ intercept + job_loss + low_buffer + job_loss*low_buffer + baseline_food_insufficiency + (baseline_age-40)/10 + baseline_household_size + baseline_female + baseline_income_7cat_dummies(2..7,ref1) + baseline_education_4cat_dummies(2..4,ref1)",
 "inference":"HC3 sandwich around probability-weighted WLS, with independent respondent pairs as a working model. No public replicate/stratum/PSU design was identified in these CSV headers. HC3 is model-robust, not full complex-survey design variance, and not a causal interval.",
 "availability":{"each_raw_cell_n":30,"each_cell_kish_n":20,"full_rank":true,"finite_hc3":true},
 "precision_target_halfwidth_pp":5,
 "decision":"No new calibration or confirmation on nominal significance. Report direction, interval, precision and proxy differences. This separately fielded survey is external to SIPP; individual-level non-overlap between surveys is not demonstrated. Do not combine observations or coefficients with SIPP.",
 "audit_before_estimate":"Commit spec and analysis code, run outcome-value-blinded membership/weight audit, commit audit before estimating contrasts. Questionnaire definitions and public codebook marginal tables were already inspected; no external preregistration.",
 "model_calibration":false,
 "causal":false
}