Making Data-Driven Decisions for Community Development

A Collaborative Deep Dive into FAR Analysis with Jeff Marx

I'm thrilled to collaborate on this post with a familiar face from my past at DBRS Morningstar, Jeff Marx. Jeff and I worked together at DBRS, where he led the cashflow modeling team while I served as the CIO. Today, he's making waves in his community of Rollingwood, TX, bringing his data expertise to the fore as a member of the Rollingwood Comprehensive Residential Code Review Committee.

Stepping into the role at the CRCRC was an honor, but I quickly realized it came with its fair share of challenges. While I had been designated the resident data expert, the irony was, there wasn't any data at hand. In my professional capacity, I'm accustomed to leading a global team where brainstorming is the norm and tasks can be delegated. However, in this scenario, there was no such support system in place. Having managed to obtain the necessary data, I was confronted with the constraint of time for the required analysis. Thankfully, being an avid reader of this newsletter, I knew exactly who to turn to. To my delight, within an hour of sending the data to Sean, I not only had the answer I sought but was also presented with an insightful geo-heat map.

And this marked the genesis of the collaboration and the creation of the article you're reading now!

Residential Code Considerations: Balancing Space, Design, and Community

When diving into the topic of residential construction codes, one quickly realizes that it's a balancing act. A key concern expressed in resident emails centers around the potential for speculative builders to fully exploit the square footage of new constructions. Theoretically, an enterprising developer could build a property that extends from one setback to another, while also reaching the maximum height limit.

This possibility stirs resistance, particularly from those who argue against limiting architectural creativity, especially on unique lots. Under current codes, a height of 35’ is permitted, with an additional 10’ allowed for sloped lots. This has sparked a debate: Should this height allowance be adjusted?

On the one side, there are those who point out that some lots are so steeply sloped that height restrictions could drastically affect the design of new homes. On the flip side, critics argue that the bonus 10’ is being used simply to maximize square footage.

What's missing from the current code, however, is a measure that takes a step back to evaluate the overall mass of a home in relation to its lot size.

This brings us to the essence of Floor to Area Ratio (FAR) analysis – a tool that can add a much-needed layer of insight into residential construction codes and planning.

The Importance of FAR Analysis in Residential Real Estate Code

The Floor to Area Ratio (FAR) plays a crucial role in residential real estate code. It acts as a measure of a building's mass relative to the lot size, revealing the built intensity of a property. For zoning and development purposes, FAR can effectively illustrate the scale of development in a neighborhood, influencing aspects like livability and compatibility with the existing built environment. As such, towns and cities should consider FAR analysis for better community planning, managing the pace and nature of development, and maintaining a balance between progress and preservation.

Unleashing Data Analysis Power with ChatGPT

Data is the lifeblood of any insightful analysis. In his new role, Jeff was able to access county data that highlighted the square footage of homes and lots. This crucial information served as our jumping-off point to better understand the complexities of the zoning situation in Rollingwood.

The best data source that I could find is provided by the Travis Central Appraisal District (TCAD). TCAD has data on hundreds of fields that can be mined. I was able to obtain historical information on both the square footage of the home and the square footage of the lot across the entire neighborhood. TCAD obtains data that’s provided by the cities through the permitting process. In Texas, we don’t have a state income tax so the majority of the funding for the schools, county and city are administered through property taxes. Those annual tax rates can go as high as 2.23% of the home’s appraised value! Given the recent increase in Texas property values (and appraised values), there’s a lot of interest from homeowners in having an accurate value reflected. Some consider protesting appraised values as the second most popular sport in Texas behind football. The level of interest in this protest process has a nice benefit for data quality. The protest form provides the protestor with instructions for how to update the county if there are any data discrepancies. This leads to an incentive structure for homeowners to clarify if their home or lot size has been overstated, or if their neighbor’s values have been understated.

With volumes of data at our disposal, parsing it to generate actionable insights became our next challenge. Enter ChatGPT – a cutting-edge AI language model that is not only a source of rich information but also a tool to analyze and visualize data in a way that fosters comprehension and stimulates discussion.

Let's put ChatGPT to work and see how it can assist us in understanding FAR analysis better. Our journey begins with our first prompt:

You are a member of the zoning board in a rapidly growing city, Rollingwood Texas. The Floor to Area Ratio (FAR) has been rapidly rising over recent years and you would like to put together a plan to introduce a FAR cap for the city. What would you need access to, to highlight trends in this space and identify this as an issue?

To highlight trends and identify issues related to the Floor Area Ratio (FAR) in Rollingwood, Texas, you would need access to several resources:

1. Historical and Current Building Data: You would need comprehensive data about the city's existing buildings and any under construction. This would include the total square footage of each building and the size of the lot upon which it stands. This information will help you calculate the FAR for each structure.

2. Zoning Records: These records would provide information about how the city's land use and zoning regulations have changed over time. They would also offer insights into how these changes have impacted the city's FAR trends.

3. Population and Housing Data: Information about population growth rates, household sizes, and housing demand could help you understand the factors driving the increase in FAR.

4. Urban Planning and Growth Projections: Future projections for city growth, housing needs, business needs, and transportation plans can help predict how FAR trends might evolve if no action is taken.

5. Stakeholder Input: Feedback from community members, local business owners, and developers can provide additional context for the data trends. They can also help identify potential impacts of a FAR cap.

6. Case Studies: Looking at how other similar cities have handled rapid growth and FAR issues can provide valuable insights and potential solutions. It's important to consider the local context, as what works in one city might not work in another.

7. Environmental Impact Reports: These reports could show how higher FAR affects local ecosystems, stormwater management, heat islands, and other environmental factors.

8. Infrastructure and Public Services Data: Information about the strain on existing infrastructure (like roads, water and sewer systems) and public services (like schools and emergency services) can help you determine whether the increased FAR is sustainable.

Once you have access to these resources, you can analyze the data to identify trends and issues. This would involve calculating the FAR for individual properties and for the city as a whole, and then comparing these figures over time. You would also need to consider how these trends align with the city's growth projections and the capacity of its infrastructure and public services.

For #1 : I have historical building data back to 2000, including the square footage and lot size. I also have current building permits, including the lot size and square footage of new structure. If we made a streamlit dashboard to display this data, what kind of charts would we want to show?

Creating a Streamlit dashboard to visualize your building data is a great idea. Here are some charts you could include to help tell the story of your data:

1. Time Series Line Chart of Average FAR: This chart could show the average FAR in Rollingwood, Texas from 2000 to the present. Each data point would represent the average FAR for that year. This chart would clearly show how the average FAR has changed over time.

2. Bar Chart of New Construction FAR: This chart could show the FAR for each new construction project, sorted by year. This would provide a clear visualization of how the size of new buildings relative to their lots has changed over time.

3. Scatter Plot of Building Size vs. Lot Size: This chart could show each building as a point, with the building's total square footage on the y-axis and the lot size on the x-axis. This would provide a visual representation of the relationship between building size and lot size.

4. Heatmap of FAR by Neighborhood: If your data includes geographic information, you could create a heatmap showing the average FAR in different neighborhoods. This could help identify areas where high FAR is particularly common.

5. Histogram of FAR Values: This chart could show the distribution of FAR values across all buildings. This would provide a sense of how common different FAR values are.

6. Pie Chart of Building Types: If your data includes information about the type of each building (e.g., residential, commercial, industrial), you could create a pie chart showing the proportion of each type. This could provide context for your other charts by showing what types of buildings are most common.

Each of these charts would provide a different perspective on your data, helping you and your stakeholders understand the trends and patterns in FAR in your city.

So, these were all interesting, though, given the growth of FAR % within Rollingwood, I thought it would be nice to build out a stacked bar chart in a Streamlit dashboard. If you are not familiar with the use of Streamlit and ChatGPT, please have a look at some of my earlier posts, such as Navigating Occupancy Rates in CMBS.

Can we add a stacked bar chart to our streamlit dashboard of FAR concentrations over time?

by year
with 4 series
% of homes >35% FAR
% of homes 20-35% FAR
% of homes 10-20% FAR
% of homes < 10% FAR

ChatGPT came back with some Python code, and with a little back and forth on some bugs, I was able to get it to produce the following, which, I think is a great visual example of the issue at hand in the town.

Next, I wanted to tackle #4 that ChatGPT had suggested: Heatmap of FAR by Neighborhood.

okay, now if i have street addresses for all of these properties, how could i:

- look up the long/lat of those properties
- based on the selected year, plot a bubble on each property that year and relative sized bubble for the FAR for that property on that year

I've used a similar type of visualization in a previous post - Supercharging Learning and Personal Growth with ChatGPT. The approach leverages the geopy library to obtain latitude and longitude coordinates, and folium to render these data points as bubbles on a map. For brevity's sake, I won't go into the Python code here, but if you're interested in the technicalities, feel free to reach out or drop a comment below.

This process results in an interactive visual, where a slider adjusts the year and the map below reflects the corresponding data. For instance, here's how the FAR concentration in Rollingwood looked in 2002:

The scale goes from blue → green → yellow → red in terms of density. In 2023, there appear to be a lot more red areas in certain parts of the town as seen below:

Insights from the Residential Code Review Committee Meeting

Jeff had the opportunity to present these findings at the Residential Code Review Committee Meeting on May 23rd, 2023.

The presentation I gave (at the 8:52 mark) during the 5.23.23 meeting was well received. Its purpose was to introduce a public discussion on the subject, and it accomplished that goal. A few attendees expressed skepticism regarding extreme application of FAR to our code, which I consider a valid concern. My hope is for us to deliberate on any proposed solutions with thoroughness and from diverse perspectives. I advocate that FAR could be applied alongside other tools to help shape the code to achieve whatever result is broadly desired by the residents. For example, we could limit the application of FAR for lots below a certain size, or apply different height rules for for homes that exceed a certain FAR. FAR doesn’t have to be applied as a blunt instrument. Interestingly, one participant provided a historical context, sharing that FAR was a significant consideration back in 2004. This procedure has a considerable way to go before completion, which includes carrying out a public survey. It will be exciting to witness how AI can assist us throughout this process.

Conclusion

This collaborative project with Jeff exemplifies the power of data-driven decision-making in community development. It's been a fascinating journey, delving into the specifics of FAR analysis, pulling and crunching numbers, and ultimately revealing insights that might inform policy and planning decisions. AI tools like ChatGPT can significantly streamline this process, making it more accessible and efficient for community members to engage with and understand the data. Here's to the potential of more such collaborations in the future!

Happy Prompting!