IRS SOI ZIP Code Data 2013 (with AGI) - inferred
- Rows
- 166,740
- Columns
- 114
- Data range
- 0 to 99999
ZIP Code data for tax year 2013 with AGI classes. Federal IRS SOI, all states CSV (pattern from available years).
AI analysis
IRS SOI ZIP Code Data 2013 — Analyst Report
1. Dataset Overview & Research Value
This dataset originates from the IRS Statistics of Income (SOI) division, sourced directly from https://www.irs.gov/pub/irs-soi/13zpallagi.csv, making it an authoritative federal publication reflecting tax year 2013 filing activity. It contains 166,740 rows across 114 columns, structured as a ZIP code × AGI class panel — meaning each ZIP code appears up to 6 times, once per agistub income bracket (1=under $25K through 6=over $200K). This design makes it exceptionally powerful for studying income stratification within and across geographies, capturing counts of returns, filing statuses, and dollar amounts for dozens of income and deduction categories simultaneously. The 2013 vintage is now roughly a decade old, so temporal caution applies when pairing with contemporary data, but it remains a gold standard for cross-sectional geographic-income analysis given the near-universal coverage of the U.S. tax-filing population.
2. Data Quality Observations
The data quality is remarkably clean for a government administrative dataset. Total null cells across all 114 columns is exactly 0 — an unusual and highly favorable characteristic that eliminates imputation concerns entirely. No type mismatches are reported, and all numeric columns parse cleanly into their declared Integer or Decimal types. The duplicate row count is flagged as pending Phase D recomputation, which warrants follow-up before any aggregation work, since undetected duplicates on the (zipcode, agistub) composite key could silently inflate totals. One structural quality note: zipcode=0 exists (min=0), which is a known IRS convention for state-level rollup records rather than a genuine ZIP code — these should be filtered or handled separately before any geographic join. The presence of 27,690 distinct ZIP codes against a theoretical U.S. maximum of ~42,000 active ZIPs is consistent with coverage of filing-active locations only, which is expected and not a data defect.
3. Key Column Distributions & Notable Patterns
The extreme skewness present across virtually every financial column is the defining distributional characteristic of this dataset and deserves careful attention:
- N1 (number of returns): mean=1,724 vs. median=230, σ=36,918, skew=86.3, with 18,905 outliers. This multi-order-of-magnitude gap between mean and median reflects the dominance of large urban ZIP codes — the top tail (e.g., aggregated state rows or dense metros) dwarfs typical suburban/rural ZIPs. - A00100 (Adjusted Gross Income, in $000s): mean=$109,624 vs. median=$14,294, σ=$2.33M, skew=93.6. At the ZIP×AGIstub level, this is expected — high-income brackets in wealthy ZIPs create extreme right tails. - A01000 (Net capital gains/losses): mean=$4,793 vs. median=$111, skew=189, min=-$1,396. This is among the most skewed columns, with a max of $60.9M, reflecting hyper-concentration of investment income in a small number of high-wealth ZIP codes. - A09600 (Alternative Minimum Tax): skew=250 — the single highest skewness in the dataset — and median=0, confirming AMT's extreme concentration among top earners. Q1 and Q3 are both 0, meaning over 75% of ZIP×AGI cells report zero AMT. - MARS1 (single filers): skew=107 with Q1=0, contrasting with MARS2 (joint filers): skew=61.4 with Q1=40, suggesting joint filing is more broadly distributed across ZIP codes while single filers are more concentrated — likely reflecting urban density patterns. - Several deduction columns (N03300/A03300 — student loan interest; N03240/A03240 — domestic production deduction) show Q1=Q3=0, meaning the majority of records have zero claimants, making them sparse features suitable only for high-income or targeted analyses.
4. Recommended Join Key Columns
The natural composite join key is (zipcode, agistub), which uniquely identifies each row and enables precise income-stratified geographic matching. For broader geographic joins, zipcode alone (after filtering out the zipcode=0 rollup rows) connects to any ZIP-level reference data, while STATEFIPS and STATE enable state-level aggregation joins. Important caveats: ZIP codes are postal constructs that change over time — a 2013 ZIP-to-ZCTA or ZIP-to-county crosswalk (e.g., from HUD or Census) will be necessary for most spatial joins, since raw ZIP codes don't map cleanly to Census geographies. For income-bracket analysis, agistub (1–6, perfectly uniform with 0 nulls and exactly 6 distinct values) is a reliable stratification key.
5. Recommended Pairing Datasets for Correlation Discovery
Several dataset categories would pair powerfully with this data:
- Census ACS 5-Year Estimates (2013 vintage) by ZIP/ZCTA: Correlate AGI distributions with educational attainment, housing values, commute patterns, and demographic composition — excellent for testing whether income predictors from tax data align with survey-based measures. - IRS SOI ZIP Code Data from other years (2011, 2012, 2014–2019): Panel analysis across years would reveal income growth trajectories, the effect of tax law changes (e.g., ATRA 2012, TCJA 2017), and ZIP-level gentrification signals via rising agistub 5–6 shares. - HUD Fair Market Rents or Zillow ZHVI by ZIP: The relationship between median AGI (A00100/N1) and housing costs is a classic affordability correlation worth quantifying at scale. - SBA Small Business Loan data or BLS Quarterly Census of Employment: Cross-referencing A00900 (Schedule C self-employment income) and N00900 with business formation or employment counts would illuminate entrepreneurship geography. - CDC PLACES or County Health Rankings: Linking income stratification (especially the agistub=1 low-income segment) with health outcomes such as obesity rates, diabetes prevalence, or life expectancy would support social determinants of health research, with ZIP-level granularity enabling neighborhood-scale analysis.
Columns
- zipcode (integer)
- agi_stub (integer)
- STATEFIPS (integer)
- STATE (string)
- N1 (decimal)
- MARS1 (decimal)
- MARS2 (decimal)
- MARS4 (decimal)
- PREP (decimal)
- N2 (decimal)
- NUMDEP (decimal)
- A00100 (decimal)
- N02650 (decimal)
- A02650 (decimal)
- N00200 (decimal)
- A00200 (decimal)
- N00300 (decimal)
- A00300 (decimal)
- N00600 (decimal)
- A00600 (decimal)
- N00650 (decimal)
- A00650 (decimal)
- N00700 (decimal)
- A00700 (decimal)
- N00900 (decimal)
- A00900 (decimal)
- N01000 (decimal)
- A01000 (decimal)
- N01400 (decimal)
- A01400 (decimal)
- N01700 (decimal)
- A01700 (decimal)
- SCHF (decimal)
- N02300 (decimal)
- A02300 (decimal)
- N02500 (decimal)
- A02500 (decimal)
- N26270 (decimal)
- A26270 (decimal)
- N02900 (decimal)
- A02900 (decimal)
- N03220 (decimal)
- A03220 (decimal)
- N03300 (decimal)
- A03300 (decimal)
- N03270 (decimal)
- A03270 (decimal)
- N03150 (decimal)
- A03150 (decimal)
- N03210 (decimal)
- A03210 (decimal)
- N03230 (decimal)
- A03230 (decimal)
- N03240 (decimal)
- A03240 (decimal)
- N04470 (decimal)
- A04470 (decimal)
- A00101 (decimal)
- N18425 (decimal)
- A18425 (decimal)
- N18450 (decimal)
- A18450 (decimal)
- N18500 (decimal)
- A18500 (decimal)
- N18300 (decimal)
- A18300 (decimal)
- N19300 (decimal)
- A19300 (decimal)
- N19700 (decimal)
- A19700 (decimal)
- N04800 (decimal)
- A04800 (decimal)
- N05800 (decimal)
- A05800 (decimal)
- N09600 (decimal)
- A09600 (decimal)
- N07100 (decimal)
- A07100 (decimal)
- N07300 (decimal)
- A07300 (decimal)
- N07180 (decimal)
- A07180 (decimal)
- N07230 (decimal)
- A07230 (decimal)
- N07240 (decimal)
- A07240 (decimal)
- N07220 (decimal)
- A07220 (decimal)
- N07260 (decimal)
- A07260 (decimal)
- N09400 (decimal)
- A09400 (decimal)
- N10600 (decimal)
- A10600 (decimal)
- N59660 (decimal)
- A59660 (decimal)
- N59720 (decimal)
- A59720 (decimal)
- N11070 (decimal)
- A11070 (decimal)
- N10960 (decimal)
- A10960 (decimal)
- N06500 (decimal)
- A06500 (decimal)
- N10300 (decimal)
- A10300 (decimal)
- N85330 (decimal)
- A85330 (decimal)
- N85300 (decimal)
- A85300 (decimal)
- N11901 (decimal)
- A11901 (decimal)
- N11902 (decimal)
- A11902 (decimal)