Header_White-14.png
Particle_Gradient_TopRight-39.png

Automate Property Data Maintenance

Location, Location, Location! – But Also Details
by Matt Amann, TrueRoll Data Engineer

Property Assessors are tasked with determining the relative value of all real property in their jurisdictions to levy proportional taxes on the owners of that property, and so they are no strangers to the three rules of real estate. While location is an excellent indicator of property value, determining the difference in value between neighboring homes comes down to characteristics: construction material, upkeep, number of bedrooms and bathrooms, and the size and finish of basements can swing home prices by tens of thousands of dollars on the same block. 


Unfortunately, these details are notoriously difficult to track and even more difficult to verify – how does a single government agency investigate thousands (or hundreds of thousands) of homes with any regularity? This problem has long seemed irresolvable, and most jurisdictions rely on surveys, permit applications, and occasional “anonymous” tips from neighbors with an axe to grind.


Assessors faces two challenges in their characteristics data: 

  • Keeping them up to date in order to capture changes to a property over time

  • Tracking the differentiating features that are the best determiners of value

Case in Point

The Cook County Assessor’s Office (CCAO), responsible for the appraisal of over a million residential parcels, discussed these challenges in a recent presentation. Senior Data Scientist Dan Snow demonstrated improvements the office has made to their valuation model in recent years and addressed the ongoing challenge of data integrity at the parcel level. He highlighted several homes that have undergone complete renovations which are not reflected in their data. In one instance, their valuation model accurately predicted the pre-renovation sale price but post-renovation it sold for nearly five times that amount.

“Models are only as good as their data,” said Snow.  “We just don’t have property characteristics for interior features, pools, condition, etc., and the data we do have is outdated and messy.” 

Pools and many other features that may indicate a “deluxe” residence are actually outside the scope of the current CCAO valuation model, making it nearly impossible to differentiate between similarly sized homes on the same block. One might be maintained as originally constructed, the other updated with hardwood floors, a chef’s kitchen and a hot tub out back. 

beforeafter.png

Before & After: 708 East 44th St

A jurisdiction wishing to tackle the problem should consider a two-stage process: 

  1. Verify, correct or fill in the missing data from the current characteristic domain

  2. Expand the valuation model to encompass a larger domain tailored to the jurisdiction

Same Data, Different Application

While working to deploy TrueRoll™ for our clients across the country, we’ve collected hundreds of millions of real estate listings from thousands of sources to ensure compliance with property tax exemption statutes. Most residential tax exemptions require the owner to claim the exempt property as their primary residence, and we’ve found that a rental listing is a good indicator of an unqualified exemption. A sale is a potential indicator of a qualified resident who doesn’t know they are eligible for a benefit, and a proactive Assessor or Tax Collector can send the new owner information about eligibility requirements and how to apply.


We’ve been gathering data to serve one purpose, but Cook County’s situation helped us realize that we had a natural capacity to handle and track characteristics for assessments as well. Our vision of better taxation through data integrity and accuracy certainly fits with helping to ensure fair and equitable assessments. The more sources of data you use, the more confident you can be.
                                                                                    Tyler Masterson, TrueRoll CEO


While TrueRoll only needs to know if a property has been listed for sale or to rent, these listings contain a wealth of information that can inform the characteristics model. To operationalize the data, we use an algorithm to map individual sources’ encoding to that used by our clients’ valuation models and provide comparisons that can be easily aggregated. If an assessor’s internal data has parcel A’s interior square footage listed as 1275, but we find several recent sources that list 1875, we can be fairly confident that the latter is closer to the actual value. On the other hand, if parcel B is in the assessor’s dataset at 1550 square feet, and we observe multiple instances of 1550, we can mark that characteristic as confirmed.

Many jurisdictions also track improvement condition, or how well the building on the parcel has been maintained. To tease this out of the listings, we’ve developed a dictionary of indicator buzzwords (e.g., “repair” for below average condition, “grand” for above average) and used an advanced text search function in PostgreSQL to count occurrences of these indicators. Using this technique will enable us to determine other features correlated with higher and lower value properties and recommend new characteristics for our clients to expand their valuation model, or alternatively to create a composite index that can act as a proxy for “how nice” a home is.

attributes.png

Confirmed, Flagged, & New Characteristics for 708 East 44th St

Improved Equity

This is particularly important as the tax burden falls more heavily on those who own less expensive property. Incorrect characteristics push valuations towards the mean for the neighborhood, overvaluing properties that have fallen into disrepair due to financial hardship and undervaluing uniquely appointed or significantly upgraded homes. When taxes are levied against these values, the owners of the more expensive homes pay less than their fair share, and the owners of the less expensive homes pay more.

image (8).png

Solving for Equality: Effect of Unknown & Inaccurate Property Characteristics

Using different data sets to surface and correct inaccuracies can effectively shift the tax burden in a more equitable way, something that holds significant promise for counties of all sizes across the country. Accurate data also allows counties to easily defend appeals with specific, objective, and documented evidence – oftentimes provided by the homeowners themselves to third-party websites like Redfin.

TrueRoll and the CCAO Data Science Department have partnered to deploy an alpha version of this promising new technique. The results are being independently evaluated by a team from the Department of Public Administration at the University of Illinois at Chicago, with a full report forthcoming in the spring of 2021.

The solution is scalable to jurisdictions of all sizes, and we’re actively looking for a diverse group of assessors to partner with for beta development. Please reach out if you think your county could benefit from more accurate data!

matt_amann_profile_bw.png
Matt Amann

Data Engineer, Innovation Specialist

Matt Amann develops data-driven solutions to public sector problems that enable governments to provide efficient service. He has extensive experience working with issues of taxation equity at the state and local level, and is focused on supporting government agencies to deploy Big Data responsibly. He holds graduate and undergraduate degrees in Public Policy from the University of Illinois at Chicago.