← All Chicago LVT reports

Chicago, IL — Land Value Tax model (CLASSIFICATION-PRESERVED VARIANT)

Variant of model.ipynb. The split-rate and abatement are run on raw assessed values (which keep Cook County's 10%/25% class ratios) instead of market value, isolating the land-shift effect from the classification-removal effect. Outputs use *_classpreserved slugs and do not overwrite the canonical exports.

Models a revenue-neutral shift of the City of Chicago's own property tax levy from the current value-based system toward land, under two reforms:

1. 2.5:1 split-rate — land taxed at 2.5× the improvement rate (canonical cross-city export).
2. Full building abatement — improvements 100% exempt (pure land value tax), shown as a comparison.

The 2.5:1 ratio is set by the Illinois Constitution's classification limit (Art. IX §4(b)): in counties over 200,000, the ratio between the highest- and lowest-assessed classes may not exceed 2.5:1 — the same cap that gives Cook County its 10% residential / 25% commercial structure. Applying it to the land-vs-improvement split keeps the reform within that legal ceiling.

Policy decisions (confirmed before modeling)

Decision	Choice
Government body	City of Chicago levy only (not the composite of all taxing agencies)
Reform	Both — 2.5:1 split-rate and 100% building abatement, compared
Exemptions	Preserve intent, but modeled on gross EAV — see limitations
Validation target	City of Chicago Corporate extension, TY2024

Data sources (public, reproducible)

Need	Cook County dataset	Fields
Land/building assessed values	Assessor – Assessed Values (uzyt-m557)	board_land, board_bldg, board_tot (+ certified/mailed fallback), class, township_code
City filter + census geometry	Assessor – Parcel Universe (nj4t-kc8j)	lat, lon, tax_code, chicago_community_area_name

The eight Cook County assessment townships 70–77 are coterminous with the City of Chicago; the filter uses those township codes (≈ 883k parcels, matching the city's parcel count).

Cook County assessment mechanics (TY2024)

• Assessed value → market value: Cook County assesses by class — residential/vacant/multi-family (major classes 1/2/3/9) at 10% of market value, commercial/industrial (class 5) at 25%, not-for-profit (class 4) at 20%, incentive classes (6/7/8) at a reduced 10% in their active phase. The published values are assessed values; market value = assessed ÷ level-of-assessment.
• Equalized Assessed Value (EAV) = assessed × the IDOR Cook County state equalizer = 3.0355 (TY2024 final).
• The City of Chicago Corporate levy (TY2024) is $1,642,587,611 at an agency rate of 1.495780% (Cook County Clerk 2024 Tax Rate Report). Total Chicago EAV that year was $109.8B.

Why the reform is computed on market value

The current system taxes EAV, which embeds Cook County's non-uniform class ratios (a $ of commercial land is taxed 2.5× a $ of residential land). A land value tax taxes a dollar of land value uniformly, so the split-rate / abatement are applied to market-value land and improvement (assessed ÷ class level of assessment). The modeled change therefore reflects both the move to a uniform land base and the land/building shift.

Limitations

• Exemptions not applied. The public Assessed Values dataset carries no per-parcel homeowner/senior exemptions, so the model runs on gross EAV. The effective city rate derived below is therefore lower than the published 1.4958% (the same levy spread over a broader, pre-exemption base). Because the model is revenue-neutral and current tax is proportional to EAV, the distribution of winners and losers is robust to this; absolute rates are not exemption-exact.
• TIF increment not removed from the base (same reasoning).
• Incentive classes (6/7/8) are assumed to be at their 10% active-phase level; a minority in years 11–12 (15%/20%) or expired would have market value slightly overstated. Small share of parcels.
• TY2024 is the City of Chicago triennial reassessment year, so Board of Review certification is still rolling out township-by-township. Each parcel uses the most final value available — Board where present, else certified, else mailed (coalesce below) — so a minority of parcels reflect a pre-Board stage. Values are final for the majority; this will settle as the reassessment completes.

Section 1 — Imports and setup

import sys
import time
from pathlib import Path

import geopandas as gpd
import matplotlib
matplotlib.use('Agg')  # headless
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import requests
from shapely.geometry import Point
from dotenv import load_dotenv

sys.path.insert(0, '../..')
REPO_ROOT = Path('../..').resolve()
load_dotenv(REPO_ROOT / '.env')

from lvt.lvt_utils import (
    model_split_rate_tax,
    model_full_building_abatement,
    calculate_current_tax,
    calculate_category_tax_summary,
    print_category_tax_summary,
    save_standard_export,
)
from lvt.census_utils import get_census_data_with_boundaries, match_to_census_blockgroups

# ---- Parameters: single source of truth, shared with the report pipeline ----
# (edit cities/chicago/chicago_params.py to change any of these — the notebook and
#  every downstream report import from the same module, so they cannot drift.)
sys.path.insert(0, '.')
from chicago_params import (
    CITY_NAME, STATE_FIPS, COUNTY_FIPS, MODEL_TYPE, LAND_IMPROVEMENT_RATIO, TAX_YEAR,
    EQUALIZER, CITY_LEVY, PUBLISHED_CITY_RATE, CHICAGO_TOWNSHIPS, class_to_loa, classify,
)

DATA_DIR = Path('data')
DATA_DIR.mkdir(exist_ok=True)
print('setup OK')

setup OK

Section 2 — Fetch / load parcel data

Downloads from the Cook County open-data Socrata API with pagination, filtered to the eight Chicago townships and TAX_YEAR. Assessed values (uzyt-m557) are joined to the Parcel Universe (nj4t-kc8j) for latitude/longitude (used to build the census geometry) and locational fields. Cached to data/parcels.gpq.

SOCRATA = 'https://datacatalog.cookcountyil.gov/resource'

def fetch_socrata(resource, select, where, page=25000, max_pages=400, retries=5):
    # Page through a Socrata resource into a DataFrame, with retry/backoff
    # (the API can be slow/flaky for these large, wide tables).
    rows, offset = [], 0
    for _ in range(max_pages):
        params = {
            '$select': select,
            '$where': where,
            '$limit': page,
            '$offset': offset,
            '$order': ':id',
        }
        batch = None
        for attempt in range(retries):
            try:
                r = requests.get(f'{SOCRATA}/{resource}.json', params=params, timeout=120)
                r.raise_for_status()
                batch = r.json()
                break
            except Exception as e:
                if attempt == retries - 1:
                    raise
                time.sleep(2 ** attempt)   # exponential backoff
        if not batch:
            break
        rows.extend(batch)
        offset += page
        if len(batch) < page:
            break
        time.sleep(0.2)
    return pd.DataFrame(rows)

PARCEL_PATH = DATA_DIR / 'parcels.gpq'

if PARCEL_PATH.exists():
    gdf = gpd.read_parquet(PARCEL_PATH)
    print(f"Loaded {len(gdf):,} parcels from cache")
else:
    town_list = ','.join(f"'{t}'" for t in CHICAGO_TOWNSHIPS)

    # Assessed values (land / building / total), all three assessment stages
    av = fetch_socrata(
        'uzyt-m557',
        select=('pin,class,township_code,'
                'board_land,board_bldg,board_tot,'
                'certified_land,certified_bldg,certified_tot,'
                'mailed_land,mailed_bldg,mailed_tot'),
        where=f"year={TAX_YEAR} AND township_code in({town_list})",
    )
    print(f"Assessed values: {len(av):,} rows")

    # Parcel Universe: location + census geometry source
    pu = fetch_socrata(
        'nj4t-kc8j',
        select='pin,lat,lon,tax_code,chicago_community_area_name',
        where=f"year={TAX_YEAR} AND township_code in({town_list})",
    )
    print(f"Parcel universe: {len(pu):,} rows")

    df = av.merge(pu, on='pin', how='left')

    num_cols = ['board_land','board_bldg','board_tot','certified_land','certified_bldg',
                'certified_tot','mailed_land','mailed_bldg','mailed_tot','lat','lon']
    for c in num_cols:
        df[c] = pd.to_numeric(df[c], errors='coerce')

    # Point geometry from parcel centroid (used by the census block-group join)
    df = df[df['lat'].notna() & df['lon'].notna()].copy()
    gdf = gpd.GeoDataFrame(
        df, geometry=[Point(xy) for xy in zip(df['lon'], df['lat'])], crs='EPSG:4326'
    )
    gdf.to_parquet(PARCEL_PATH)
    print(f"Downloaded and cached {len(gdf):,} parcels")

print(gdf[['pin','class','board_land','board_bldg','board_tot']].head())

Loaded 882,904 parcels from cache
              pin class  board_land  board_bldg  board_tot
0  19151120251019   299       915.0     16834.0    17749.0
1  19151190381006   299       868.0     11132.0    12000.0
2  19151250350000   234      3661.0     24339.0    28000.0
3  19153260150000   203      3780.0     19220.0    23000.0
4  19154300260000   590     12263.0       896.0    13159.0

Section 3 — Classify, normalize to market value, flag exempt

Column mapping

Concept	Column	Notes
Land assessed value	assessed_land	coalesce board → certified → mailed
Improvement assessed value	assessed_bldg	coalesce board → certified → mailed
Total assessed value	assessed_tot	coalesce board → certified → mailed
Class / use code	class	Cook County 3-char class (e.g. 299, EX)
Exempt flag	full_exmp	1 for class EX/RR or zero/blank assessment
Level of assessment	loa	0.10 / 0.20 / 0.25 by major class
Market land value	market_land	assessed_land / loa
Market improvement value	market_bldg	assessed_bldg / loa

Levels of assessment and category labels follow the Cook County Assessor Classifications of Real Property (revised 12/16/2024). A Transportation - Parking category captures automotive improvements (5-22, 7-22, 8-22, 7-52), gasoline stations (classes ending in -23), and minor improvements (classes ending in -90, including 1-90 and surface parking lots); these keep their major-class level of assessment and are held out of the "$0 improvement → Vacant Land" override. Class 2-41 (vacant land under common ownership with an adjacent residence) is categorized as Vacant Land but, as a Class 2 parcel, is taxed at the 10% residential level.

# --- Coalesce assessment stages: board (final) -> certified -> mailed ---
def coalesce(*cols):
    out = gdf[cols[0]].copy()
    for c in cols[1:]:
        out = out.where(out > 0, gdf[c])
    return out.fillna(0).clip(lower=0)

gdf['assessed_land'] = coalesce('board_land', 'certified_land', 'mailed_land')
gdf['assessed_bldg'] = coalesce('board_bldg', 'certified_bldg', 'mailed_bldg')
gdf['assessed_tot']  = coalesce('board_tot',  'certified_tot',  'mailed_tot')

# Level of assessment + property category come from chicago_params (imported in §1),
# so the notebook and the report pipeline share one definition.
gdf['loa'] = gdf['class'].map(class_to_loa)
gdf['PROPERTY_CATEGORY'] = gdf['class'].map(classify)

# --- Exempt flag: class EX/RR, or no taxable assessment ---
gdf['full_exmp'] = (
    gdf['class'].astype(str).str.upper().isin(['EX', 'RR'])
    | (gdf['assessed_tot'] <= 0)
    | gdf['loa'].isna()
).astype(int)

# --- Market values (assessed / level of assessment) ---
loa = gdf['loa'].fillna(0.10)
gdf['market_land'] = (gdf['assessed_land'] / loa).clip(lower=0)
gdf['market_bldg'] = (gdf['assessed_bldg'] / loa).clip(lower=0)

# Override: zero improvement -> Vacant Land (regardless of class), EXCEPT parking
# (surface lots often have $0 building value but should stay Transportation - Parking).
gdf.loc[(gdf['assessed_bldg'] <= 0) & (gdf['PROPERTY_CATEGORY'] != 'Transportation - Parking'),
        'PROPERTY_CATEGORY'] = 'Vacant Land'

print(f"Parcels: {len(gdf):,}   Fully-exempt (EX/RR/zero): {(gdf['full_exmp']==1).sum():,}")
print("\nCategory counts:")
print(gdf['PROPERTY_CATEGORY'].value_counts())
print("\nLevel-of-assessment distribution (taxable):")
print(gdf.loc[gdf['full_exmp']==0, 'loa'].value_counts(dropna=False))

Parcels: 882,904   Fully-exempt (EX/RR/zero): 52,735

Category counts:
Condominium                       287166
Single Family Residential         278520
Small Multi-Family (2-4 units)    120202
Vacant Land                        92560
Townhome / Rowhouse                23020
Mixed Use                          15962
Transportation - Parking           15957
Retail / General Commercial        13232
Industrial                         11404
Large Multi-Family (5+ units)      11258
Other Commercial                    4359
Other Residential                   4309
Office / Commercial Condo           4245
Hotel                                436
Other                                172
Exempt                               100
Agricultural                           2
Name: PROPERTY_CATEGORY, dtype: int64

Level-of-assessment distribution (taxable):
0.10    781295
0.25     48664
0.20       210
Name: loa, dtype: int64

Section 4 — Current tax (City of Chicago levy only)

The City of Chicago Corporate levy is spread across all taxable equalized assessed value at a uniform citywide agency rate. We compute gross EAV per parcel, then derive the effective city rate that reproduces the official levy on this (pre-exemption) base. Revenue match to the official levy is exact by construction; the meaningful check is that the derived effective rate is in the neighborhood of the published 1.4958% — lower, because exemptions/TIF are not removed from the base.

# Gross EAV
gdf['eav'] = gdf['assessed_tot'] * EQUALIZER

taxable = gdf['full_exmp'] == 0
total_eav_taxable = float(gdf.loc[taxable, 'eav'].sum())

# Derive the effective city rate that reproduces the official City Corporate levy
city_rate = CITY_LEVY / total_eav_taxable
gdf['millage_rate'] = city_rate * 1000.0   # calculate_current_tax divides by 1000

current_revenue, _, gdf = calculate_current_tax(
    df=gdf,
    tax_value_col='eav',
    millage_rate_col='millage_rate',
    exemption_flag_col='full_exmp',
)

print(f"Taxable gross EAV:        ${total_eav_taxable:,.0f}")
print(f"Derived effective rate:    {city_rate*100:.4f}%   (published City Corporate: {PUBLISHED_CITY_RATE*100:.4f}%)")
print(f"Modeled current revenue:  ${current_revenue:,.0f}")
print(f"Official City levy:       ${CITY_LEVY:,.0f}")
print(f"Gap: {(current_revenue/CITY_LEVY - 1)*100:+.3f}%   (≈0 by construction)")

rate_ratio = city_rate / PUBLISHED_CITY_RATE
print(f"\nDerived/published rate ratio: {rate_ratio:.3f} "
      f"(<1 expected: gross EAV exceeds the net taxable base by exemptions + TIF)")

Taxable gross EAV:        $139,992,554,872
Derived effective rate:    1.1733%   (published City Corporate: 1.4958%)
Modeled current revenue:  $1,642,587,611
Official City levy:       $1,642,587,611
Gap: +0.000%   (≈0 by construction)

Derived/published rate ratio: 0.784 (<1 expected: gross EAV exceeds the net taxable base by exemptions + TIF)

Section 5 — Split-rate model (2.5:1) on market value

Fully-exempt parcels are held out of the revenue-neutral solver and excluded from the export and report charts (they carry no signal). The reform is solved on market-value land and improvement so a dollar of land is taxed uniformly across classes.

# Hold out fully-exempt parcels (excluded from model, export, and charts)
n_exempt = int((gdf['full_exmp'] == 1).sum())
gdf = gdf[gdf['full_exmp'] == 0].copy()

# Taxable value columns expected by save_standard_export = the reform (market) base
gdf['taxable_land_value'] = gdf['assessed_land']        # classification-preserved (assessed, embeds 10%/25% class ratios)
gdf['taxable_improvement_value'] = gdf['assessed_bldg']  # classification-preserved

land_millage, improvement_millage, new_revenue, gdf = model_split_rate_tax(
    df=gdf,
    land_value_col='taxable_land_value',
    improvement_value_col='taxable_improvement_value',
    current_revenue=gdf['current_tax'].sum(),
    land_improvement_ratio=LAND_IMPROVEMENT_RATIO,
)

print(f"Held out {n_exempt:,} fully-exempt parcels (excluded from model, export, and charts).")
print(f"Land millage: {land_millage:.5f}   Improvement millage: {improvement_millage:.5f}   "
      f"(ratio {land_millage/improvement_millage:.1f}:1)")
print(f"Revenue check: ${new_revenue:,.0f}  vs current ${gdf['current_tax'].sum():,.0f}")

category_summary = calculate_category_tax_summary(
    df=gdf, category_col='PROPERTY_CATEGORY',
    current_tax_col='current_tax', new_tax_col='new_tax',
)
print_category_tax_summary(category_summary, title=f"{CITY_NAME} — {LAND_IMPROVEMENT_RATIO}:1 split-rate tax impact")

Held out 52,735 fully-exempt parcels (excluded from model, export, and charts).
Land millage: 65.64175   Improvement millage: 26.25670   (ratio 2.5:1)
Revenue check: $1,642,587,611  vs current $1,642,587,611

chicago — 2.5:1 split-rate tax impact
================================================================================
                      Category  Count Total Tax Δ ($) Total Δ (%) Mean Δ ($) Median Δ ($) Avg % Δ Median % Δ % Parcels > +10% % Parcels < -10%
                   Condominium 287166    $-32,235,577      -10.8%      $-112         $-71   -9.8%     -14.3%             7.5%            64.5%
     Single Family Residential 278520     $15,870,770        4.9%        $57          $10    3.8%       1.6%            29.9%            16.7%
Small Multi-Family (2-4 units) 120202      $7,609,022        4.0%        $63         $-24    0.4%      -2.4%            21.4%            24.4%
                   Vacant Land  39925     $10,896,860       84.7%       $273          $84  257.2%      84.3%           100.0%             0.0%
           Townhome / Rowhouse  23020       $-585,131       -1.7%       $-25         $-14    0.1%      -1.8%            22.6%            27.4%
                     Mixed Use  15962      $2,794,447        4.1%       $175         $-43    0.8%      -3.4%            22.3%            27.0%
      Transportation - Parking  15957     $17,511,011       50.4%     $1,097         $360  339.9%      78.2%            91.9%             2.2%
   Retail / General Commercial  13232     $17,251,406       25.7%     $1,304         $289   21.0%      15.8%            58.8%             8.7%
                    Industrial  11404     $24,828,619       34.4%     $2,177         $483   47.0%      53.7%            79.7%             2.5%
 Large Multi-Family (5+ units)  11258    $-12,104,551       -8.7%    $-1,075        $-153    3.0%      -5.3%            25.9%            37.4%
              Other Commercial   4359     $-5,424,195       -6.3%    $-1,244         $-91    2.8%      -8.1%            32.0%            47.1%
             Other Residential   4309        $342,754        4.6%        $80         $109   67.4%      75.4%            95.5%             1.4%
     Office / Commercial Condo   4245    $-40,249,389      -15.8%    $-9,482        $-154   -6.2%      -9.9%            11.0%            49.5%
                         Hotel    436     $-6,847,685      -13.3%   $-15,706      $-2,326   -5.4%     -12.5%            16.1%            56.2%
                         Other    172        $343,297       12.4%     $1,996         $156   16.5%       9.0%            49.4%            14.5%
                  Agricultural      2         $-1,658      -15.6%      $-829        $-829   19.6%      19.6%            50.0%            50.0%

OVERALL SUMMARY:
Total Properties: 830,169
Total Tax Δ: $0
Net Revenue Δ: $0
Average Percent Δ (mean of means): 47.38%
Median Percent Δ (median of medians): -0.12%
Percent of ALL parcels with tax increase > +10%: 26.46%
Percent of ALL parcels with tax decrease < -10%: 33.98%

Section 5b — Gate 5 artifact scan

Read the category table against economic priors before trusting it. Watch for: distinct categories piled at one extreme value (ceiling clustering), implausible building shares, or large placeholder/zero building counts. A condo unit with apportioned land is normal in Cook County; a built commercial category showing ~0% building share is not.

d = gdf.copy()
d['bldg_share'] = d['market_bldg'] / (d['market_land'] + d['market_bldg']).replace(0, np.nan)
scan = d.groupby('PROPERTY_CATEGORY').agg(
    n=('tax_change_pct', 'size'),
    median_pct=('tax_change_pct', 'median'),
    median_bldg_share=('bldg_share', 'median'),
    pct_zero_bldg=('market_bldg', lambda s: (s <= 1).mean()),
).sort_values('median_pct', ascending=False)
pd.set_option('display.width', 160)
print(scan.round(3))

                                     n  median_pct  median_bldg_share  pct_zero_bldg
PROPERTY_CATEGORY                                                                   
Vacant Land                      39925      84.300              0.000          0.998
Transportation - Parking         15957      78.226              0.055          0.004
Other Residential                 4309      75.356              0.081          0.000
Industrial                       11404      53.712              0.277          0.000
Agricultural                         2      19.627              0.585          0.000
Retail / General Commercial      13232      15.764              0.620          0.000
Other                              172       9.043              0.681          0.000
Single Family Residential       278520       1.589              0.748          0.000
Townhome / Rowhouse              23020      -1.832              0.779          0.000
Small Multi-Family (2-4 units)  120202      -2.366              0.784          0.000
Mixed Use                        15962      -3.389              0.793          0.000
Large Multi-Family (5+ units)    11258      -5.282              0.810          0.000
Other Commercial                  4359      -8.102              0.836          0.000
Office / Commercial Condo         4245      -9.873              0.852          0.000
Hotel                              436     -12.507              0.875          0.000
Condominium                     287166     -14.277              0.891          0.000

Section 6 — Building abatement scenario (100% improvement exemption)

A full building abatement (pure land value tax), revenue-neutral to the same City levy, on the same modeled (non-exempt) parcels. Shown for comparison; the canonical cross-city export remains the 2.5:1 split-rate.

gdf_abate = gdf.copy()
abate_millage, abate_revenue, gdf_abate = model_full_building_abatement(
    df=gdf_abate,
    land_value_col='taxable_land_value',
    improvement_value_col='taxable_improvement_value',
    current_revenue=gdf_abate['current_tax'].sum(),
    abatement_percentage=1.0,
)
print(f"Abatement land millage: {abate_millage:.5f}")
print(f"Revenue check: ${abate_revenue:,.0f}")

abate_summary = calculate_category_tax_summary(
    df=gdf_abate, category_col='PROPERTY_CATEGORY',
    current_tax_col='current_tax', new_tax_col='new_tax',
)
print_category_tax_summary(abate_summary, title=f"{CITY_NAME} — 100% building abatement (pure LVT)")

# Side-by-side median % change by category
cmp = pd.DataFrame({
    'split_4to1_median_pct': gdf.groupby('PROPERTY_CATEGORY')['tax_change_pct'].median(),
    'abatement_median_pct':  gdf_abate.groupby('PROPERTY_CATEGORY')['tax_change_pct'].median(),
}).round(1)
print("\nMedian % change by category — split-rate vs abatement:")
print(cmp.sort_values('split_4to1_median_pct'))

Building abatement model (100.0% abatement)
Millage rate: 149.8960
Total tax revenue: $1,642,587,611.00
Target revenue: $1,642,587,611.00
Revenue difference: $0.00 (0.0000%)

Abatement land millage: 149.89601
Revenue check: $1,642,587,611

chicago — 100% building abatement (pure LVT)
================================================================================
                      Category  Count Total Tax Δ ($) Total Δ (%) Mean Δ ($) Median Δ ($) Avg % Δ Median % Δ % Parcels > +10% % Parcels < -10%
                   Condominium 287166   $-122,687,582      -41.2%      $-427        $-271  -37.6%     -54.3%            12.9%            79.8%
     Single Family Residential 278520     $60,463,261       18.6%       $217          $38   14.4%       6.1%            47.1%            37.7%
Small Multi-Family (2-4 units) 120202     $28,989,356       15.4%       $241         $-91    1.4%      -9.0%            36.4%            48.9%
                   Vacant Land  39925     $41,258,268      320.9%     $1,033         $321  320.9%     320.9%           100.0%             0.0%
           Townhome / Rowhouse  23020     $-2,220,434       -6.4%       $-96         $-55    0.4%      -7.0%            37.9%            47.9%
                     Mixed Use  15962     $10,646,357       15.6%       $667        $-163    2.9%     -12.9%            35.2%            52.0%
      Transportation - Parking  15957     $66,637,999      191.9%     $4,176       $1,370  875.9%     297.7%            94.0%             4.2%
   Retail / General Commercial  13232     $65,669,855       97.9%     $4,963       $1,102   80.1%      60.0%            70.1%            20.8%
                    Industrial  11404     $94,497,085      130.9%     $8,286       $1,837  179.0%     204.4%            88.0%             7.0%
 Large Multi-Family (5+ units)  11258    $-46,042,832      -32.9%    $-4,090        $-580   11.3%     -20.1%            35.4%            55.7%
              Other Commercial   4359    $-20,627,870      -23.9%    $-4,732        $-346   10.2%     -30.8%            39.0%            56.3%
             Other Residential   4309      $1,305,875       17.5%       $303         $415  256.6%     286.8%            96.8%             2.5%
     Office / Commercial Condo   4245   $-153,138,038      -60.2%   $-36,075        $-585  -23.4%     -37.6%            15.6%            65.2%
                         Hotel    436    $-26,052,066      -50.7%   $-59,752      $-8,849  -20.6%     -47.6%            19.7%            72.2%
                         Other    172      $1,307,075       47.2%     $7,599         $594   62.8%      34.4%            58.1%            29.7%
                  Agricultural      2         $-6,308      -59.2%    $-3,154      $-3,154   74.7%      74.7%            50.0%            50.0%

OVERALL SUMMARY:
Total Properties: 830,169
Total Tax Δ: $0
Net Revenue Δ: $0
Average Percent Δ (mean of means): 113.06%
Median Percent Δ (median of medians): -0.44%
Percent of ALL parcels with tax increase > +10%: 37.48%
Percent of ALL parcels with tax decrease < -10%: 51.63%

Median % change by category — split-rate vs abatement:
                                split_4to1_median_pct  abatement_median_pct
PROPERTY_CATEGORY                                                          
Condominium                                     -14.3                 -54.3
Hotel                                           -12.5                 -47.6
Office / Commercial Condo                        -9.9                 -37.6
Other Commercial                                 -8.1                 -30.8
Large Multi-Family (5+ units)                    -5.3                 -20.1
Mixed Use                                        -3.4                 -12.9
Small Multi-Family (2-4 units)                   -2.4                  -9.0
Townhome / Rowhouse                              -1.8                  -7.0
Single Family Residential                         1.6                   6.1
Other                                             9.0                  34.4
Retail / General Commercial                      15.8                  60.0
Agricultural                                     19.6                  74.7
Industrial                                       53.7                 204.4
Other Residential                                75.4                 286.8
Transportation - Parking                         78.2                 297.7
Vacant Land                                      84.3                 320.9

Section 7 — Census join + export

Census join must run before export. The canonical join uses parcel-centroid geometry to attach block-group demographics. The split-rate result is the canonical cross-city export (analysis/data/chicago.csv + analysis/reports/chicago/); the abatement scenario gets its own report folder (analysis/reports/chicago_abatement/) but is not added to the cross-city dataset.

# Census join — must happen before export.
# Cook County has 4,002 census tracts; the default TIGERweb block-group fetch
# (get_census_data_with_boundaries) queries tract-by-tract and takes ~40 min here.
# Instead pull the whole Illinois block-group file once from the Census FTP (seconds)
# and filter to the county — same std_geoid + boundaries the canonical path produces.
from lvt.census_utils import (
    get_census_blockgroups_from_ftp, get_census_data, match_to_census_blockgroups,
)

_fips = STATE_FIPS + COUNTY_FIPS
try:
    census_gdf = get_census_blockgroups_from_ftp(_fips, 2022)   # one state zip, fast
    census_data = get_census_data(_fips, 2022)                  # ACS 5-year demographics
    census_gdf = census_gdf.merge(
        census_data, on='std_geoid', how='left', suffixes=('', '_census'))
    census_gdf = census_gdf.loc[:, ~census_gdf.columns.str.endswith('_census')]
    gdf = match_to_census_blockgroups(gdf, census_gdf)
    if 'minority_pct' not in gdf.columns and 'total_pop' in gdf.columns and 'white_pop' in gdf.columns:
        gdf['minority_pct'] = ((gdf['total_pop'] - gdf['white_pop']) / gdf['total_pop'] * 100).round(2)
    if 'black_pct' not in gdf.columns and 'total_pop' in gdf.columns and 'black_pop' in gdf.columns:
        gdf['black_pct'] = (gdf['black_pop'] / gdf['total_pop'] * 100).round(2)
    print(f"Census join: {gdf['std_geoid'].notna().mean()*100:.1f}% matched")
except Exception as e:
    print(f"Census join failed: {e}")
    for _col in ['std_geoid', 'median_income', 'minority_pct', 'black_pct']:
        gdf[_col] = float('nan')

📥 Downloading block groups for state 17 from Census FTP...

✅ Downloaded 18,258,643 bytes
📊 Reading shapefile: tl_2022_17_bg.shp

🎯 Filtered to 4002 block groups in county 17031

Census join: 100.0% matched

# Propagate census columns onto the abatement frame (same index)
for _col in ['std_geoid', 'median_income', 'minority_pct', 'black_pct']:
    if _col in gdf.columns:
        gdf_abate[_col] = gdf[_col]

# Export (canonical) — split-rate 2.5:1
from lvt.lvt_utils import save_standard_export
out_df = save_standard_export(
    df=gdf,
    city=CITY_NAME + '_classpreserved',
    output_path=f'data/{CITY_NAME}_classpreserved.csv',
    model_type=MODEL_TYPE,
    land_millage=land_millage,
    improvement_millage=improvement_millage,
    property_category_col='PROPERTY_CATEGORY',
    current_tax_col='current_tax',
    new_tax_col='new_tax',
    tax_change_col='tax_change',
    tax_change_pct_col='tax_change_pct',
    taxable_land_col='taxable_land_value',
    taxable_improvement_col='taxable_improvement_value',
)

# Standard report — 7 PNGs in analysis/reports/chicago/
from lvt.viz import create_city_report
create_city_report(out_df, CITY_NAME + '_classpreserved', show=False)

# Building-abatement scenario — separate export (not in cross-city) + own report folder
abate_out = save_standard_export(
    df=gdf_abate,
    city=f'{CITY_NAME}_abatement_classpreserved',
    output_path=f'data/{CITY_NAME}_abatement_classpreserved.csv',
    model_type='abatement:100pct',
    land_millage=abate_millage,
    improvement_millage=0.0,
    property_category_col='PROPERTY_CATEGORY',
    current_tax_col='current_tax',
    new_tax_col='new_tax',
    tax_change_col='tax_change',
    tax_change_pct_col='tax_change_pct',
    taxable_land_col='taxable_land_value',
    taxable_improvement_col='taxable_improvement_value',
)
create_city_report(abate_out, f'{CITY_NAME}_abatement_classpreserved', show=False)
print("Done.")

  ✓ chicago_classpreserved: 830,169 rows → data/chicago_classpreserved.csv  [model: split_rate:2.5]

/opt/anaconda3/lib/python3.9/site-packages/scipy/__init__.py:155: UserWarning: A NumPy version >=1.18.5 and <1.25.0 is required for this version of SciPy (detected version 1.26.4
  warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"

  ✓ chicago_abatement_classpreserved: 830,169 rows → data/chicago_abatement_classpreserved.csv  [model: abatement:100pct]

Done.

Validation summary

Check	Result
Revenue match	Current revenue set to the official City of Chicago Corporate levy $1,642,587,611 (TY2024, Cook County Clerk 2024 Tax Rate Report) — exact by construction
Effective rate	Derived effective city rate (see Section 4) vs published 1.4958%; difference reflects gross-EAV modeling (exemptions/TIF not removed)
Parcel count	≈883k Chicago parcels; fully-exempt held out & excluded from charts
Census coverage	See Section 7 output
PNGs generated	analysis/reports/chicago/ (split-rate) and analysis/reports/chicago_abatement/
Artifact scan (Gate 5)	See Section 5b

Figures populated on execution.