How Did We Identify Black and Hispanic Real Estate Developers?
Most public and proprietary business databases classify businesses based on at least one of two common industry classification systems, the North American Industry Classification System (NAICS) and the Standard Industrial Classification (SIC) system. However, NAICS and SIC do not define real estate development as an industry at any level of aggregation. Each NAICS and SIC industry that contains real estate developers also includes companies that are not real estate developers. Therefore, researchers cannot use these standard industry categories to reliably identify and study real estate developers.
To address this challenge, we created a custom list of Black and Hispanic real estate developers. We began our search for real estate developers with Dun & Bradstreet’s (D&B) Hoovers database. We developed a list of industry codes in real estate and construction industries that were likely to contain real estate developers. We also used D&B’s built-in search tool to scan company websites for key phrases such as “real estate development.” Our website search included an even broader list of industries. We downloaded all companies that D&B classified as “minority-owned” and that matched one of our two search criteria.
We then used a two-step search process to determine whether each company was a real estate developer and whether it was Black- or Hispanic-owned. We broadly defined a real estate developer as a company that, as a core part of its business model, takes property that it owns (either undeveloped land or existing building(s)), increases the market value of that property by building a new building or substantially upgrading the existing structure(s), and sells or rents the property to an individual or another company to generate a profit. We used a standard definition of “Black-owned” or “Hispanic-owned” that requires a company to be at least 51 percent owned by one or more people who are either Black or Hispanic, respectively.
For our analyses and the map and directory, we used a company’s headquarters or primary location as the sole physical address for that company. While developers may operate and have offices in multiple cities, we did not have a way to identify branch locations for all companies in our dataset and therefore used each company’s headquarters.
If a company had a website, we were usually able to determine whether a company was a real estate developer based on that company’s website. Most real estate developers with websites clearly described the business’s activities or clearly identified the company as a real estate developer. For companies that did not have their own website, we used publicly available information on websites such as LinkedIn, Buzzfile, and Manta to determine whether the company was a real estate developer.
If we determined that a company was a real estate developer, we searched for information about the company’s owner(s) to find out whether they identified as Black or Hispanic. We counted a company as Black- or Hispanic-owned only if we could find that information stated openly in a publicly available source. We found this information through sources such as company websites, social media pages, public versions of government contracting and lending databases, news articles, and interviews. We did not discover any companies that were both Black- and Hispanic-owned.
We conducted this two-stage research process for each company and created an initial list of Black and Hispanic real estate developers. This list included companies from 57 different NAICS codes. This large number of NAICS codes demonstrates that standard industry codes do not line up with real estate development as an industry.
To supplement this list, we also downloaded a list of “minority-owned” real estate firms and funds from Preqin, a proprietary data provider that focuses on alternative assets. We completed the same two-stage research process for each company on the Preqin list and added any new Black and Hispanic developers to our list. The Preqin data gave us unique insight into deal-level information for a small subset of Black and Hispanic developers. We could not directly match the Preqin data to D&B data because many of the companies in that dataset either are not in D&B or do not have revenue estimates. The Preqin data also do not have information about revenue, so we could only infer information about the size of the developers based on the size of their deals. The typical deal size in Preqin is large (more than $30 million), which suggests that the coverage is primarily for large real estate developers.
We used only the lists from D&B and Preqin in our quantitative analysis. We downloaded 2022 data from D&B and Preqin for our quantitative variables. We found a relatively small number of additional Black and Hispanic real estate developers through other sources such as online articles, local and national chambers of commerce, and real estate professionals’ associations; we added these developers to our directory but did not include them in our analysis. We excluded these developers from our analysis because we wanted to construct a comparison group of white-owned companies and could not replicate the additional searches for the comparison group.
How Did We Create a List of Companies to Which We Could Compare Black and Hispanic Real Estate Developers?
We were not only interested in identifying Black and Hispanic real estate developers and analyzing their businesses’ characteristics and performance; we also wanted to compare Black and Hispanic developers to their white peers. We could contextualize Black and Hispanic developers’ performance by comparing their businesses to white developers’ businesses.
As with Black and Hispanic developers, we used D&B and Preqin to identify white real estate developers. For each dataset, we assumed that privately held, for-profit companies that the data provider did not mark as “minority-owned” were “white-owned.” We did not attempt to verify that each company was white-owned because this information is generally unavailable.
We could not search for information about all of the white-owned businesses that met our search criteria because we found many times more white-owned businesses than “minority-owned” businesses. Based on our search criteria, D&B included 252,000 white-owned businesses that could potentially be real estate developers; Preqin included 5,455 white-owned potential real estate developers. We therefore took a random sample of 1,527 white-owned businesses from D&B and 1,871 white-owned businesses from Preqin, using the same search criteria that we used for the initial lists of Black and Hispanic developers. We verified that each company was a real estate developer and removed non-developers.
What Are the Limitations of Our Methodology?
Our data are subject to self-selection bias because we relied on companies to publicly identify themselves both as real estate developers and as either Black- or Hispanic-owned. We drew on sources such as company websites and government procurement databases, both of which may be biased towards larger companies. For similar reasons, we suspect that both D&B and Preqin may undercount the number of small minority-owned businesses.
Second, both D&B and Preqin are themselves samples of businesses. Both providers aim to be comprehensive but we were not able to test whether or to what extent their datasets represent the broader industry. Finally, D&B estimates both revenue and employment for some companies. Representatives from D&B described a valid approach for developing these estimates but even the best estimation procedures are subject to more uncertainty than reported or observed data. We have no reason to believe that this estimation process introduced systematic error that would bias an analysis by race and ethnicity.
Finally, although we aimed to be comprehensive the list of developers in our analysis and in the map and directory is not a complete list. There may be Black or Hispanic developers who are not registered in any database and have not been named in any list or article available on the internet.
How Did We Estimate the Impact of Removing the Constraints on Black and Hispanic Developers?
To estimate the economic impacts of removing the constraints on Black and Hispanic developers, we developed a simulation that accounted for each of the key constraints our research identified: the representation crisis, the revenue gap for medium-sized Black developers, and the revenue cliffs for large Black and Hispanic developers. To address the representation crisis, we modeled what the industry would look like if it grew so that the percent of developers who are Black or Hispanic was equal to the percent of the U.S. population who are Black or Hispanic, respectively. To address the remaining two constraints, we assume that Black and Hispanic developers would have the same overall average revenue and employment as all of the developers in our sample if they were not constrained ($1.97 million in revenue and 31 employees).
We assumed that all new Black and Hispanic businesses are net additions and that they do not displace any existing businesses. We also assumed that the jobs and revenue that these new developers create are net additions to the economy and do not cause other businesses to lose employees or revenue. These are standard assumptions for this type of analysis based on widely accepted economic theory about the long-run dynamics of a growing industry.
Are the Comparisons We Make Statistically Significant?
The following comparisons are statistically significant at the .10 level:
- Average revenue for small Black developers compared to small white developers.
- Average revenue for small Hispanic developers compared to small white developers.
- Average revenue for mid-sized Black developers compared to mid-sized white developers.
- Average revenue for mid-sized Black developers compared to mid-sized Hispanic developers.
- Median deal size for Black and Hispanic developers (combined) compared to deal size for white developers.
We encourage readers not to make additional comparisons because they have not been tested for statistical significance.