Pre-IC3 Workstream A floor references — v1 promotion (E3, E4/E5, E8/E9)

runs/sipp_e8_e9_floors_v1.env.json

@@ -0,0 +1,16 @@
+{
+  "artifact": "sipp_e8_e9_floors_v1.json",
+  "status": "MEASUREMENT_ENVIRONMENT",
+  "environment": {
+    "python": "3.10.13",
+    "numpy": "2.1.3",
+    "pandas": "2.3.3",
+    "sklearn": "1.5.2",
+    "scipy": "1.13.1",
+    "platform": "macOS-26.5.2-arm64-arm-64bit",
+    "fitting_stack": {
+      "populace_fit": "absent",
+      "populace_frame": "absent"
+    }
+  }
+}

runs/sipp_e8_e9_floors_v1.inputs.json

@@ -0,0 +1,16 @@
+{
+  "artifact": "sipp_e8_e9_floors_v1.json",
+  "status": "SOURCE_INPUT_DIGESTS",
+  "official_source": {
+    "url": "https://www2.census.gov/programs-surveys/sipp/data/datasets/2023/pu2023_csv.zip",
+    "archive_sha256": "9c5363d56aca2041db20d46d17b81e9be931eb5b18bd5f5238b367d2dd7fb74b",
+    "archive_bytes": 109036604,
+    "archive_member": "pu2023.csv"
+  },
+  "staged_input": {
+    "path": "pu2023.csv.gz",
+    "sha256": "1e49df7e013970ea60443e4e15ded3e8fda07643038471bf6b3429a7383fbf69",
+    "bytes": 109032744,
+    "transport_note": "gzip -n recompression of the sole CSV member from the verified official Census ZIP"
+  }
+}

runs/sipp_e8_e9_floors_v1.json

@@ -0,0 +1,152 @@
+{
+  "artifact": "sipp_e8_e9_floors",
+  "version": "v1",
+  "status": "PRE-LOCK REFERENCE - NOT RATIFIED; IC3 not locked; no thresholds. v1 is a pinning event, not a ratification",
+  "issue": "192",
+  "deployment_scale_note": "RECORDED GAP (review of #212): these floors are half-vs-half, i.e. the sampling noise of ~50%-of-source estimates, while IC3 proposes scoring on a 0.20 person holdout - there is no candidate-context floor (gate-1 ctx20 analog). Under root-n scaling, a 20% scoring frame has ~sqrt(0.5/0.2)=1.58x the sampling noise of the half-split basis, so these floors are mildly ANTI-conservative (too tight), not conservative. RECORDED_NOT_SATISFIED: IC3 must accept a registered analytic scale adjustment or require matching-context floors before candidate runs.",
+  "source": "pu2023 (reference year 2022), persons observed all 12 reference months (censoring-free draft restriction, recorded; ESTIMAND NOTE per review: candidate runs scored against these cells must apply the identical full-year-persons restriction)",
+  "method": "person-disjoint sha256 half-splits, seeds 0-19 (raised from 5 per review: E8 cells where floor sd exceeds the mean need a stable across-seed sd); rates floored on |log rate ratio|, earnings-change medians/IQRs on absolute gaps in log-points; weighted by WPFINWGT",
+  "seam_caveat": "identical to the E4/E5 floors: half-splits share SIPP's seam structure; #214 carries the seam measurement",
+  "stay_median_heaping_caveat": "within-job SIPP monthly earnings are mostly constant across a wave (dependent-interview reporting), so the stay-transition median log-change heaps at exactly 0 and its half-vs-half floor is degenerate (0.0) - the same failure class as the tenure quantile heaping; E9-stay thresholds should be stated on the IQR or a distributional distance, not the median",
+  "thin_flag_units": {
+    "e8_nonemployment_by_age": "rows per half, equal to persons (one row per person in the E8 frame) vs THIN_CELL_PERSONS=200",
+    "e9_transitions.earnings_change": "distinct persons per half (person_id.nunique(); rows are consecutive-month transition pairs and a person can contribute up to 11) vs THIN_CELL_PERSONS=200"
+  },
+  "sipp_jobs_reader_commit": "a059193e4fad80ceb1c2e1f4177aa5c69abb1048",
+  "source_input": {
+    "path": "pu2023.csv.gz",
+    "sha256": "1e49df7e013970ea60443e4e15ded3e8fda07643038471bf6b3429a7383fbf69"
+  },
+  "e8_nonemployment_by_age": {
+    "16_24": {
+      "any_nonemp_share": 0.4145,
+      "long_nonemp_share": 0.3422,
+      "persons_unweighted": 1781,
+      "any_nonemp": {
+        "floor_abs_log_ratio_mean": 0.05149,
+        "floor_abs_log_ratio_sd": 0.0356,
+        "thin": false
+      },
+      "long_nonemp": {
+        "floor_abs_log_ratio_mean": 0.08586,
+        "floor_abs_log_ratio_sd": 0.04431,
+        "thin": false
+      }
+    },
+    "25_34": {
+      "any_nonemp_share": 0.1401,
+      "long_nonemp_share": 0.0973,
+      "persons_unweighted": 3240,
+      "any_nonemp": {
+        "floor_abs_log_ratio_mean": 0.05885,
+        "floor_abs_log_ratio_sd": 0.04819,
+        "thin": false
+      },
+      "long_nonemp": {
+        "floor_abs_log_ratio_mean": 0.10659,
+        "floor_abs_log_ratio_sd": 0.06993,
+        "thin": false
+      }
+    },
+    "35_44": {
+      "any_nonemp_share": 0.1017,
+      "long_nonemp_share": 0.0663,
+      "persons_unweighted": 3543,
+      "any_nonemp": {
+        "floor_abs_log_ratio_mean": 0.09917,
+        "floor_abs_log_ratio_sd": 0.07372,
+        "thin": false
+      },
+      "long_nonemp": {
+        "floor_abs_log_ratio_mean": 0.13369,
+        "floor_abs_log_ratio_sd": 0.0899,
+        "thin": false
+      }
+    },
+    "45_54": {
+      "any_nonemp_share": 0.0852,
+      "long_nonemp_share": 0.0555,
+      "persons_unweighted": 3241,
+      "any_nonemp": {
+        "floor_abs_log_ratio_mean": 0.10871,
+        "floor_abs_log_ratio_sd": 0.078,
+        "thin": false
+      },
+      "long_nonemp": {
+        "floor_abs_log_ratio_mean": 0.09797,
+        "floor_abs_log_ratio_sd": 0.07416,
+        "thin": false
+      }
+    },
+    "55_64": {
+      "any_nonemp_share": 0.1045,
+      "long_nonemp_share": 0.083,
+      "persons_unweighted": 3449,
+      "any_nonemp": {
+        "floor_abs_log_ratio_mean": 0.09033,
+        "floor_abs_log_ratio_sd": 0.05384,
+        "thin": false
+      },
+      "long_nonemp": {
+        "floor_abs_log_ratio_mean": 0.11965,
+        "floor_abs_log_ratio_sd": 0.05523,
+        "thin": false
+      }
+    },
+    "65_99": {
+      "any_nonemp_share": 0.1887,
+      "long_nonemp_share": 0.1546,
+      "persons_unweighted": 2286,
+      "any_nonemp": {
+        "floor_abs_log_ratio_mean": 0.05524,
+        "floor_abs_log_ratio_sd": 0.04697,
+        "thin": false
+      },
+      "long_nonemp": {
+        "floor_abs_log_ratio_mean": 0.08085,
+        "floor_abs_log_ratio_sd": 0.04997,
+        "thin": false
+      }
+    }
+  },
+  "e9_transitions": {
+    "transition_rates": {
+      "entry": 0.0101,
+      "exit": 0.0094,
+      "j2j": 0.0035,
+      "stay": 0.977
+    },
+    "earnings_change": {
+      "stay": {
+        "median_log_change": 0.0,
+        "iqr_log_change": 0.0656,
+        "pairs_unweighted": 163552,
+        "persons_unweighted": 16286,
+        "floor_abs_median_gap": {
+          "mean": 0.0,
+          "sd": 0.0
+        },
+        "floor_abs_iqr_gap": {
+          "mean": 0.0,
+          "sd": 0.0
+        },
+        "thin": false
+      },
+      "j2j": {
+        "median_log_change": 0.2264,
+        "iqr_log_change": 1.069,
+        "pairs_unweighted": 545,
+        "persons_unweighted": 524,
+        "floor_abs_median_gap": {
+          "mean": 0.06659,
+          "sd": 0.05025
+        },
+        "floor_abs_iqr_gap": {
+          "mean": 0.12848,
+          "sd": 0.07392
+        },
+        "thin": false
+      }
+    }
+  }
+}

runs/sipp_spell_floors_v1.env.json

@@ -0,0 +1,16 @@
+{
+  "artifact": "sipp_spell_floors_v1.json",
+  "status": "MEASUREMENT_ENVIRONMENT",
+  "environment": {
+    "python": "3.10.13",
+    "numpy": "2.1.3",
+    "pandas": "2.3.3",
+    "sklearn": "1.5.2",
+    "scipy": "1.13.1",
+    "platform": "macOS-26.5.2-arm64-arm-64bit",
+    "fitting_stack": {
+      "populace_fit": "absent",
+      "populace_frame": "absent"
+    }
+  }
+}

runs/sipp_spell_floors_v1.inputs.json

@@ -0,0 +1,16 @@
+{
+  "artifact": "sipp_spell_floors_v1.json",
+  "status": "SOURCE_INPUT_DIGESTS",
+  "official_source": {
+    "url": "https://www2.census.gov/programs-surveys/sipp/data/datasets/2023/pu2023_csv.zip",
+    "archive_sha256": "9c5363d56aca2041db20d46d17b81e9be931eb5b18bd5f5238b367d2dd7fb74b",
+    "archive_bytes": 109036604,
+    "archive_member": "pu2023.csv"
+  },
+  "staged_input": {
+    "path": "pu2023.csv.gz",
+    "sha256": "1e49df7e013970ea60443e4e15ded3e8fda07643038471bf6b3429a7383fbf69",
+    "bytes": 109032744,
+    "transport_note": "gzip -n recompression of the sole CSV member from the verified official Census ZIP"
+  }
+}

runs/sipp_spell_floors_v1.json

@@ -0,0 +1,149 @@
+{
+  "artifact": "sipp_spell_floors",
+  "version": "v1",
+  "status": "PRE-LOCK REFERENCE - NOT RATIFIED; IC3 not locked; no thresholds. v1 is a pinning event, not a ratification",
+  "issue": "192",
+  "deployment_scale_note": "RECORDED GAP (review of #212): these floors are half-vs-half, i.e. the sampling noise of ~50%-of-source estimates, while IC3 proposes scoring on a 0.20 person holdout - there is no candidate-context floor (gate-1 ctx20 analog). Under root-n scaling, a 20% scoring frame has ~sqrt(0.5/0.2)=1.58x the sampling noise of the half-split basis, so these floors are mildly ANTI-conservative (too tight), not conservative. RECORDED_NOT_SATISFIED: IC3 must accept a registered analytic scale adjustment or require matching-context floors before candidate runs.",
+  "source": "pu2023 (reference year 2022)",
+  "method": "person-disjoint sha256 half-splits, seeds 0-4; per-cell |log(rate_a/rate_b)| mean/sd across seeds; weighted by WPFINWGT",
+  "seam_caveat": "both halves share SIPP seam structure; the seam-vs-J2J reconciliation run is a separate required artifact before IC3 thresholds lock",
+  "thin_flag_units": {
+    "e4_retention_by_age_sex": "distinct persons per half (person_id.nunique(); rows are person-month retention pairs) vs THIN_CELL_PERSONS=200",
+    "e5_runs_by_age": "rows per half, equal to persons (one row per person in the run-length frame) vs THIN_CELL_PERSONS=200"
+  },
+  "sipp_jobs_reader_commit": "a059193e4fad80ceb1c2e1f4177aa5c69abb1048",
+  "source_input": {
+    "path": "pu2023.csv.gz",
+    "sha256": "1e49df7e013970ea60443e4e15ded3e8fda07643038471bf6b3429a7383fbf69"
+  },
+  "e4_retention_by_age_sex": {
+    "16_24|sex1": {
+      "rate": 0.9897,
+      "pairs_unweighted": 8055,
+      "floor_abs_log_ratio_mean": 0.00182,
+      "floor_abs_log_ratio_sd": 0.00107,
+      "thin": false
+    },
+    "16_24|sex2": {
+      "rate": 0.9901,
+      "pairs_unweighted": 7093,
+      "floor_abs_log_ratio_mean": 0.00198,
+      "floor_abs_log_ratio_sd": 0.00232,
+      "thin": false
+    },
+    "25_34|sex1": {
+      "rate": 0.9948,
+      "pairs_unweighted": 17797,
+      "floor_abs_log_ratio_mean": 0.0008,
+      "floor_abs_log_ratio_sd": 0.00036,
+      "thin": false
+    },
+    "25_34|sex2": {
+      "rate": 0.9953,
+      "pairs_unweighted": 15555,
+      "floor_abs_log_ratio_mean": 0.00143,
+      "floor_abs_log_ratio_sd": 0.00114,
+      "thin": false
+    },
+    "35_44|sex1": {
+      "rate": 0.9968,
+      "pairs_unweighted": 19920,
+      "floor_abs_log_ratio_mean": 0.00097,
+      "floor_abs_log_ratio_sd": 0.00058,
+      "thin": false
+    },
+    "35_44|sex2": {
+      "rate": 0.9964,
+      "pairs_unweighted": 17203,
+      "floor_abs_log_ratio_mean": 0.00089,
+      "floor_abs_log_ratio_sd": 0.00065,
+      "thin": false
+    },
+    "45_54|sex1": {
+      "rate": 0.9978,
+      "pairs_unweighted": 17779,
+      "floor_abs_log_ratio_mean": 0.0003,
+      "floor_abs_log_ratio_sd": 0.00028,
+      "thin": false
+    },
+    "45_54|sex2": {
+      "rate": 0.998,
+      "pairs_unweighted": 16516,
+      "floor_abs_log_ratio_mean": 0.00051,
+      "floor_abs_log_ratio_sd": 0.00045,
+      "thin": false
+    },
+    "55_64|sex1": {
+      "rate": 0.9984,
+      "pairs_unweighted": 18342,
+      "floor_abs_log_ratio_mean": 0.00067,
+      "floor_abs_log_ratio_sd": 0.00058,
+      "thin": false
+    },
+    "55_64|sex2": {
+      "rate": 0.999,
+      "pairs_unweighted": 17410,
+      "floor_abs_log_ratio_mean": 0.0003,
+      "floor_abs_log_ratio_sd": 0.00014,
+      "thin": false
+    },
+    "65_99|sex1": {
+      "rate": 0.9992,
+      "pairs_unweighted": 12113,
+      "floor_abs_log_ratio_mean": 0.00058,
+      "floor_abs_log_ratio_sd": 0.00039,
+      "thin": false
+    },
+    "65_99|sex2": {
+      "rate": 0.9996,
+      "pairs_unweighted": 10437,
+      "floor_abs_log_ratio_mean": 0.00039,
+      "floor_abs_log_ratio_sd": 0.00032,
+      "thin": false
+    }
+  },
+  "e5_runs_by_age": {
+    "16_24": {
+      "full_year_run_share": 0.4669,
+      "persons_unweighted": 1830,
+      "floor_abs_log_ratio_mean": 0.04937,
+      "floor_abs_log_ratio_sd": 0.04904,
+      "thin": false
+    },
+    "25_34": {
+      "full_year_run_share": 0.7499,
+      "persons_unweighted": 3299,
+      "floor_abs_log_ratio_mean": 0.01944,
+      "floor_abs_log_ratio_sd": 0.01104,
+      "thin": false
+    },
+    "35_44": {
+      "full_year_run_share": 0.8245,
+      "persons_unweighted": 3555,
+      "floor_abs_log_ratio_mean": 0.01764,
+      "floor_abs_log_ratio_sd": 0.01639,
+      "thin": false
+    },
+    "45_54": {
+      "full_year_run_share": 0.8671,
+      "persons_unweighted": 3256,
+      "floor_abs_log_ratio_mean": 0.01096,
+      "floor_abs_log_ratio_sd": 0.00768,
+      "thin": false
+    },
+    "55_64": {
+      "full_year_run_share": 0.8673,
+      "persons_unweighted": 3458,
+      "floor_abs_log_ratio_mean": 0.01742,
+      "floor_abs_log_ratio_sd": 0.00998,
+      "thin": false
+    },
+    "65_99": {
+      "full_year_run_share": 0.7941,
+      "persons_unweighted": 2294,
+      "floor_abs_log_ratio_mean": 0.01327,
+      "floor_abs_log_ratio_sd": 0.01036,
+      "thin": false
+    }
+  }
+}

runs/tenure_floors_v1.env.json

@@ -0,0 +1,16 @@
+{
+  "artifact": "tenure_floors_v1.json",
+  "status": "MEASUREMENT_ENVIRONMENT",
+  "environment": {
+    "python": "3.10.13",
+    "numpy": "2.1.3",
+    "pandas": "2.3.3",
+    "sklearn": "1.5.2",
+    "scipy": "1.13.1",
+    "platform": "macOS-26.5.2-arm64-arm-64bit",
+    "fitting_stack": {
+      "populace_fit": "absent",
+      "populace_frame": "absent"
+    }
+  }
+}