• Prior Models
• Model Overview
  • Differences Compared to the Residential Model
    • Features Used
    • Valuation
    • Multi-PIN Sales
  • Condo Strata
• Ongoing Issues
  • Unit Heterogeneity
  • Buildings With Few Sales
  • Buildings Without Sales
• FAQs
• Usage
  • Getting Data
• License
• Contributing

⚠️ NOTE ⚠️

The condominium model (this repo) is nearly identical to the residential (single/multi-family) model, with a few key differences. Please read the documentation for the residential model first.

This repository contains code, data, and documentation for the Cook County Assessor’s condominium reassessment model. Information about prior year models can be found at the following links:

The duty of the Cook County Assessor’s Office is to value property in a fair, accurate, and transparent way. The Assessor is committed to transparency throughout the assessment process. As such, this document contains:

• A description of the differences between the residential model and this (condominium) model
• An outline of ongoing issues specific to condominium assessments

The repository itself contains the code for the Automated Valuation Model (AVM) used to generate initial assessed values for all condominium properties in Cook County. This system is effectively an advanced machine learning model (hereafter referred to as “the model”). It uses previous sales to generate estimated sale values (assessments) for all properties.

The Cook County Assessor’s Office has started to track a limited number of characteristics (building-level square footage, unit-level square footage, bedrooms, and bathrooms) for condominiums, but the data we have varies in both the characteristics available and their completeness between triads. Staffing limitations have forced the office to prioritize smaller condo buildings less likely to have recent unit sales in certain parts of the county.

Like most assessors nationwide, our office staff cannot enter buildings to observe property characteristics. For condos, this means we cannot observe amenities, quality, or any other interior characteristics which must instead be gathered from listings and a number of additional third-party sources.

The only complete information our office currently has about individual condominium units is their age, location, sale date/price, and percentage of ownership. This makes modeling condos particularly challenging, as the number of usable features is quite small. Fortunately, condos have two qualities which make modeling a bit easier:

1. Condos are more homogeneous than single/multi-family properties, i.e. the range of potential condo sale prices is much narrower.
2. Condo are pre-grouped into clusters of like units (buildings), and units within the same building usually have similar sale prices.

We leverage these qualities to produce what we call strata, a feature unique to the condo model. See Condo Strata for more information about how strata is used and calculated.

Because our individual condo unit characteristics are sparse and incomplete, we primarily must rely on aggregate geospatial features, economic features, strata, and time of sale to determine condo assessed values. The features in the table below are the ones used in the 2024 assessment model.

For the most part, condos are valued the same way as single- and multi-family residential property. We train a model using individual condo unit sales, predict the value of all units, and then apply any post-modeling adjustment.

However, because the CCAO has so little information about individual units, we must rely on the condominium percentage of ownership to differentiate between units in a building. This feature is effectively the proportion of the building’s overall value held by a unit. It is created when a condominium declaration is filed with the County (usually by the developer of the building). The critical assumption underlying the condo valuation process is that percentage of ownership correlates with the relative market value differences between units.

Percentage of ownership is used in two ways:

1. It is used directly as a predictor/feature in the regression model to estimate differing unit values within the same building.
2. It is used to reapportion unit values directly i.e. the value of a unit is ultimately equal to % of ownership * total building value.

Visually, this looks like:

For what the office terms “nonlivable” spaces — parking spaces, storage space, and common area — the breakout of value works differently. See this excel sheet for an interactive example of how nonlivable spaces are valued based on the total value of a building’s livable space.

Percentage of ownership is the single most important feature in the condo model. It determines almost all intra-building differences in unit values.

The condo model is trained on a select number of “multi-PIN sales” (or “multi-sales”) in addition to single-parcel sales. Multi-sales are sales that include more than one parcel. In the case of condominiums, many units are sold bundled with deeded parking spaces that are separate parcels. These two-parcel sales are highly reflective of the unit’s actual market price. We split the total value of these two-parcel sales according to their relative percent of ownership before using them for training. For example, for a $100,000 sale of a unit (4% ownership) and a parking space (1% ownership), the sale would be adjusted to $80,000:

$$\frac{0.04}{0.04 + 0.01} * $100,000 = $80,000$$

The condo model uses an engineered feature called strata to deliver much of its predictive power. Strata is the binned, time-weighted, 5-year average sale price of the building. There are two strata features used in the model, one with 10 bins and one with 100 bins. Buildings are binned across each triad using either quantiles or 1-dimensional k-means. A visual representation of quantile-based strata binning looks like:

To put strata in more concrete terms, the table below shows a sample 5-level strata. Each condominium unit would be assigned a strata from this table (Strata 1, Strata 2, etc.) based on the 5-year weighted average sale price of its building. All units in a building will have the same strata.

Some additional notes on strata:

• Strata is calculated in the ingest stage of this repository.
• Calculating the 5-year average sale price of a building requires at least 1 sale. Buildings with no sales have their strata imputed via KNN (using year built, number of units, and location as features).
• Number of bins (10 and 100) was chosen based on model performance. These numbers yielded the lowest root mean-squared error (RMSE).

The CCAO faces a number of ongoing issues specific to condominium modeling. We are currently working on processes to fix these issues. We list the issues here for the sake of transparency and to provide a sense of the challenges we face.

The current modeling methodology for condominiums makes two assumptions:

1. Condos units within the same building are similar and will sell for similar amounts.
2. If units are not similar, the percentage of ownership will accurately reflect and be proportional to any difference in value between units.

The model process works even in heterogeneous buildings as long as assumption 2 is met. For example, imagine a building with 8 identical units and 1 penthouse unit. This building violates assumption 1 because the penthouse unit is likely larger and worth more than the other 10. However, if the percentage of ownership of each unit is roughly proportional to its value, then each unit will still receive a fair assessment.

However, the model can produce poor results when both of these assumptions are violated. For example, if a building has an extreme mix of different units, each with the same percentage of ownership, then smaller, less expensive units will be overvalued and larger, more expensive units will be undervalued.

This problem is rare, but does occur in certain buildings with many heterogeneous units. Such buildings typically go through a process of secondary review to ensure the accuracy of the individual unit values.

The condo model relies on sales within the same building to calculate strata. This method works well for large buildings with many sales, but can break down when there are only 1 or 2 sales in a building. The primary danger here is unrepresentative sales, i.e. sales that deviate significantly from the real average value of a building’s units. When this happens, buildings can have their average unit sale value pegged too high or low.

Fortunately, buildings without any recent sales are relatively rare, as condos have a higher turnover rate than single and multi-family property. Smaller buildings with low turnover are the most likely to not have recent sales.

When no sales have occurred in a building in the 5 years prior to assessment, the building’s strata features are imputed. The model will look at nearby buildings that have similar unit counts/age and then try to assign an appropriate strata to the target building.

Most of the time, this technique produces reasonable results. However, buildings without sales still go through an additional round of review to ensure the accuracy of individual unit values.

Note: The FAQs listed here are for condo-specific questions. See the residential model documentation for more general FAQs.

Q: What are the most important features in the condo model?

As with the residential model, the importance of individual features varies by location and time. However, generally speaking, the most important features are:

• Location, location, location. Location is the largest driver of county-wide variation in condo value. We account for location using geospatial features like neighborhood.
• Condo percentage of ownership, which determines the intra-building variation in unit price.
• Condo building strata. Strata provides us with a good estimate of the average sale price of a building’s units.

Q: How do I see my condo building’s strata?

Individual building strata are not included with assessment notices or shown on the CCAO’s website. However, strata are stored in the sample data included in this repository. You can load the data (input/condo_strata_data.parquet) using R and the read_parquet() function from the arrow library.

Q: How do I see the assessed value of other units in my building?

You can use the CCAO’s Address Search to see all the PINs and values associated with a specific condominium building, simply leave the Unit Number field blank when submitting a search.

Q: How do I view my unit’s percentage of ownership?

The percentage of ownership for individual units is printed on assessment notices. You may also be able to find it via your building’s board or condo declaration.

Installation and usage of this model is identical to the installation and usage of the residential model. Please follow the instructions listed there.

The data required to run these scripts is produced by the ingest stage, which uses SQL pulls from the CCAO’s Athena database as a primary data source. CCAO employees can run the ingest stage or pull the latest version of the input data from our internal DVC store using:

dvc pull

Public users can download data for each assessment year using the links below. Each file should be placed in the input/ directory prior to running the model pipeline.

• assmntdata.parquet
• modeldata.parquet

• assessment_data.parquet
• condo_strata_data.parquet
• land_nbhd_rate_data.parquet
• training_data.parquet

• assessment_data.parquet
• condo_strata_data.parquet
• land_nbhd_rate_data.parquet
• training_data.parquet

Due to a data issue with the initial 2024 model run, there are actually two final 2024 models. The run 2024-02-16-silly-billy was used for Rogers Park only, while the run 2024-03-11-pensive-manasi was used for all subsequent City of Chicago townships.

The data issue caused some sales to be omitted from the 2024-02-16-silly-billy training set, however the actual impact on predicted values was extremely minimal. We chose to update the data and create a second final model out of an abundance of caution, and, given low transaction volume in 2023, to include as many arms-length transactions in the training set as possible.

• assessment_data.parquet
• char_data.parquet
• condo_strata_data.parquet
• land_nbhd_rate_data.parquet
• training_data.parquet

• assessment_data.parquet
• char_data.parquet
• condo_strata_data.parquet
• land_nbhd_rate_data.parquet
• training_data.parquet

For other data from the CCAO, please visit the Cook County Data Portal.

Distributed under the AGPL-3 License. See LICENSE for more information.

We welcome pull requests, comments, and other feedback via GitHub. For more involved collaboration or projects, please see the Developer Engagement Program documentation on our group wiki.