Author(s)
This methodology explains how “roadsTaken,” which documents the history of highway planning, construction, and displacement in Houston, Texas, was designed and displayed.
Mapping Houston’s Highway Displacements
In recent years, the impact of the United States’ first round of freeway construction has become a topic of extensive study. However, no project has yet attempted a systematic accounting of the residents displaced by these freeways. It is understandable why, as even studies of more recent events have struggled to accurately count displaced residents.[1] Understanding exactly who was displaced and the reasons behind these displacements is essential as our society continues to grapple with the legacy of systemic racism and urban renewal in our cities. This is especially true for policymakers, who are now being asked to address and correct perceived past injustices while ensuring future urban highway projects are equitable and well-planned.
Unfortunately, records on right-of-way clearance and displacement due to freeway construction are often incomplete or missing from archives. The federal government's statistics and records on these displacements only date back to 1991, and even these statistics are often inadequately reported by state governments.[2] As a result, studying the post-war decades of highway displacement in Houston requires reconstructing the historical record using a variety of archives and sources.
To construct as accurate a historical record as possible, roadsTaken has built two interlinked databases: one composed of every building in the highways’ paths and the other composed of people who lived in these buildings. In Houston, neither the city nor the state has a complete record of the structures — homes, garages, stores, and apartments — that were claimed through eminent domain for highway right-of-way. While some maps of freeway right-of-way were found in the archives of the Texas Department of Transportation, much of the information appears to have been lost or destroyed. As a result, many displaced buildings and people had to be identified through a variety of other historical sources.
First, an outline of the research process was created, mapping out how to achieve the goal of answering who was displaced by freeway construction (Figure 1). Because of the incomplete historical record, our research team began searching through the archives for two types of data: maps and buildings, and people and demographic data. By merging these two streams of information into a searchable map, this project aims uncover some of the historical realities of highway displacement.
Figure 1 — A Schematic of the Research Architecture Developed at the Start of the Project
Figure 2 — Paths of the 209 Linear Miles of Freeways and Adjacent Right-of-Way Clearance in Harris County (1946–74)
Finding Cleared Buildings
To identify buildings that were in the path of highway construction, our team utilized various historical geographic sources, including fire insurance adjustor maps from the Sanborn Fire Insurance company, right-of-way surveys conducted by the Texas Highway Department, Tax Assessor maps from Harris County, and aerial photography from the United States Department of Agriculture. No single set of maps provided a complete picture: Highway department maps only existed for later periods of construction, while fire insurance maps were restricted to older neighborhoods within the city’s core.
To complicate matters further, the Sanborn fire insurance maps — which were by far the most valuable historical source for identifying structures claimed by right-of-way — existed in several different editions, each revised and changed up to a particular date. In several cases, our researchers had to use magnification and bright lights to see through paper corrections pasted over buildings that had been cleared (Figure 3). Despite these challenges, by combining and cross-referencing this diverse set of historical sources, the project’s geospatial database successfully documents almost every identifiable structure cleared for highway construction from 1946 through 1974.
Figure 3 — Example of Updated Map, With Corrections Pasted Over Buildings of Interest
All of these maps were then georeferenced, i.e., stretched and reshaped so that they accurately overlaid a digital projection of the city (Figure 4). Our team of researchers then individually traced each of the thousands of structures using geographic information system (GIS) software. These digitized building shapes were then linked to a tabular database filled with historical information from the map sources. For each structure, researchers recorded details such as the address, size, number of stories, and any other specific information provided by the source. This included whether the building was a dwelling or a store, any known names of the building, and, in some cases, the owner of the building (Figure 5). Each structure was also classified into several broad categories: residences (houses, apartments, and garage apartments), nonhousing residential buildings (garages and sheds), and commercial, industrial, religious, educational, agricultural, and civic buildings.
Figure 4 — Examples of Fire Insurance Maps (Left) and Texas Highway Department Survey Maps (Right)
Note: These maps were modified for readability and georeferenced on a digital projection of the city.
Figure 5 — Example of a Partially Traced (in Blue) City Block of Cleared Buildings
Note: Each building is now a digital shape to which data can be attached.
As each structure cleared by the highways was traced from historical maps, it was assigned a unique identification number, connecting the information in the database to a digital representation of the building as it stood before it was cleared. In total, this project has uncovered 11,007 such structures, ranging from tiny outhouses, sheds, and garages to massive apartment buildings and large factories, with all sizes of buildings in between. For a large portion of the buildings, multiple historical sources had to be cross-referenced to uncover these details. For example, on many maps, the historical addresses were missing. To address this, our researchers used microfiche archives of Houston city street directories or “white pages,” to find the corresponding address of each building. Once all the cleared structures were georeferenced, described, and categorized, our researchers focused on identifying residential buildings and other housing units.
Finding Displaced Residents
The project’s primary goals were to determine who was displaced by highway construction and the reasons behind their displacement. While the U.S. government has tracked data on displacement over the last three decades, there are no official statistics for the period covered by this study.[5] Almost all of Houston’s urban displacements from highway construction took place between the late 1940s and early 1970s. As detailed in the section on the history of highway planning, this early planning and construction was directed by City Hall. Our research team obtained a temporary internship at the City Planning and Development Department and, with the help of city employees, reviewed the department’s entire archives, along with archived materials at the Houston History Research Center. However, detailed records from the City Planning Department and Houston Planning Commission for our research period were destroyed.[6] Even if these records still existed, it is unlikely that they would contain the details of each displaced person. Therefore, to create as accurate a count as possible of displaced Houstonians, our research team had to assemble the database using a variety of other sources, each contributing pieces to the overall picture.
Previous studies have examined whole census tract data to analyze the impact of highway construction, but our research team chose not to rely solely on this data. Many census tracts include a mix of different income levels and demographics, while the residences cleared by highways were typically concentrated in one part of a tract. A survey of the historical maps, which were collected for the map of cleared buildings, revealed an unequal density of displacement throughout the city. In some areas, only a few buildings were cleared per mile, while in others, over a thousand buildings were removed in a single highway mile. Additionally, our researchers found that along certain stretches of highways, apartment complexes, duplexes, or large houses often contained dozens of residents. Although the database and map of structures allowed our researchers to make some use of this data, the overall goal was to connect actual residents — as well as summary statistics — to each building. To identify who was displaced by highway construction, our project needed to determine who resided in each cleared building when it was cleared for highway construction.
Fortunately, the complete records of the 1950 U.S. census population schedules were released to the public in April 2022.[7] Census schedules include a range of information on residents such as addresses, names, ages, race, gender, occupations, and more. Combined with the previously released 1940 census schedules, this data allowed our researchers to gather the most detailed decennial information for over a quarter of the households displaced in Harris County, as most major right-of-way alignments were completed during the 1940s and 1950s. To transform the thousands of pages of census schedules into usable data, the project used optical character recognition to convert the handwritten census schedules of every census enumeration district that overlapped with highway right-of-way.[8] The result was the creation of a dataset containing information on over 720,000 Harris County residents across the two censuses, encompassing a total of 45,168,703 cells of data.
House numbers and street names were used as keys to join the census-based dataset with our dataset on buildings, identifying individuals who lived in the cleared structures. By using a structured query language “join” function to match our tables based on common street address information, our research team was able to extract the respondents in the census who lived in the houses and buildings within our cleared structures database (Figure 6).
Figure 6 — Graphic Representation of How Cleared Buildings are Linked With Displaced People
Overall, this methodology is highly accurate but not comprehensive. The resulting database and map show the names and details of real people and where they lived in 1940 or 1950 within areas that would later become highway right-of-way. The project’s website provides links to scans of the original documents from which this information was sourced. This linked dataset not only shows the exact names of people, but it also includes the exact number of people living in each building and the relationships among the occupants (i.e., spouse, child, parent, lodger, etc.). This methodology is also useful for identifying residences in unexpected buildings, as several matches reveal people who lived in stores, garages, or other buildings identified on our mapping sources as nonresidential.
One gap in our data is mapping Hispanic populations in Houston. In 1930, “Mexican” was included as a response category to the census questionnaire, but it was removed before the 1940 and subsequent counts after a successful lobbying effort by the Texas-based League of the United Latin American Citizens.[9] Hispanic as a distinct category of ethnicity was not added to the full short-form census until 1980, after the roadsTaken study ends.[10] As Hispanic Houstonians were already a sizable community in the city when freeway construction began and continued to grow throughout the period of study, this presents a challenge that official data cannot adequately address. The methodologies discussed here present the census data and rolls as they are, based off the often-flawed judgements of census enumerators. According to Dan Bouk, and borne out by the roadsTaken database, white census takers occasionally still used the “Mexican” category for Hispanic residents, despite clear instructions to categorize these residents as “White.” This is especially true for likely Hispanic Houstonians who lived in predominately Black neighborhoods.[11]
Our database reflects the census as it was recorded, so most people who today might identify as Hispanic are listed solely as “White.” However, highway displacement significantly impacted Hispanic neighborhoods, such as the Near Northside, Frost Town, and Harrisburg. To study the magnitude of these impacts, we have matched residents who have common Spanish surnames that on later censuses would identify as Hispanic, residents who list their native language as Spanish, residents who identify a Latin American nation as their place of birth, their father’s place of birth, or their mother’s place of birth.[12] While none these indicators will wholly or comprehensively account for people who today would likely identify as Hispanic, this provides a starting point for future research.
Determining who was displaced by a highway requires careful consideration of the timing of the displacement. Our research team decided the most important date to consider was the date when the final alignment decision was made. For example, if a highway’s alignment was finalized in the 1940s, as with the Gulf Freeway in 1943, then the 1940 census is included in the dataset for that section of the city’s urban freeway network. Subsequent sections will detail the results and uses of each of these datasets. The results have been mapped on the project’s website under the “Census Rolls” button (Figure 7).
Figure 7 — Example Portion of a Map Showing the Southwest Freeway at Montrose Boulevard
Note: Dots represent people matched with buildings differentiated by race (blue represents white residents, and green represents Black residents). Empty residences are unmatched.
While the resulting database is highly detailed, it is important to note that it does not capture the residents living in right-of-way areas at the actual date of clearance by the city or the state. For example, if clearances began around 1957, as was the case for the Southwest Freeway, then the database only displays those who lived in these buildings in 1950. Unfortunately, the detailed records of the 1960 and 1970 censuses will not be released until 2032 and 2042, respectively. As a result, the further we move forward in time from our two available census years — 1940 and 1950 — the less likely it is that the matched census rolls accurately reflect the residents at the time of alignment or clearance. The South Freeway, South & Southwest concurrency near downtown, the Harrisburg Freeway spur, Interstate 10 north of downtown, and the North Loop East were all cleared around two decades after the last available data. As a result, demographic or income summaries based on these datasets for blocks or neighborhoods that underwent demographic or economic changes during this period are likely to be inaccurate or misleading.
Although this method of linking addresses to census records has accurately identified over 9,000 people from the 1940 census and over 7,000 people from the 1950 census, there are still omissions and gaps. The methodology requires an accurate street address to establish a match, so if a street address is missing or incorrectly recorded, the building remains unpopulated in the dataset. A majority of the residential buildings remain unmatched with any residents. This could be because these buildings were not residences, were unoccupied, or were unfinished during the census. However, many omissions are caused by either errors in the address data in the structures database or errors in the census roll dataset.
These errors often stem from the optical character recognition process, which can misinterpret handwritten house numbers and street names. In some cases, the handwritten information is simply illegible. Other omissions occur when the historical maps lack addresses or when street names or house numbers change. Mistakes may have also occurred during the data entry process or when census takers incorrectly or illegibly recorded street names or numbers. The greatest number of errors occurred in areas with unpaved roads or informal street names.
By linking the data and viewing the results on a map, our research team was able to easily identify structures for which no residents were linked, which was useful for describing the geographic distribution of these errors. The project was able to identify and correct some errors by cross-referencing original sources on the interactive map, but fixing all the omissions caused by errors would require thousands of person-hours. Because of these omissions, the totals generated represent an undercount of the actual number of displaced people. Additionally, because of the large time lag between the data and actual displacement, this methodology should not be used for overall estimates of displacement.
The limitations of this methodology, particularly the lack of data after 1950, prompted the development of a second approach to identifying the people displaced by highways. Through the creation of a detailed database of structures, each coded with housing information, this data can now be compared with census tract summary data available for every census — 1940, 1950, 1960, and 1970 — during the target time period. Census tracts in Harris County between 1940 and 1980 typically contained populations ranging from a few thousand to 20,000 people. Summary statistics — such as total population by race, age, income, and tenure — provide less precise but more contemporary information about residents living in highway right-of-way and help validate the results of our previous census roll matching method.
For each structure in our geodatabase, a coding process identified whether it served primarily as a residence, including dwellings, garage apartments, duplexes, and apartment buildings. Then, using the spatial join function in our GIS software, each structure was attached to census tract data for each decennial census up to the date of the building’s clearance. This allowed each residential building — excluding unoccupied structures like sheds — to carry the corresponding census tract information. Of particular importance are the number of persons per housing unit, as well as their race, age, income, place of birth, and tenure.
To estimate the total population residing in structures from the database for each census decennial, the following equation was applied: The number of housing unit structures was multiplied by the persons-per-housing-unit statistic from the linked census tract data for the corresponding year. This calculation accounts for persons living in group housing and vacancy rates, and applies a multiplier to structures that contain more than one housing unit, such as duplexes and apartment buildings. The results of this equation treat each housing unit in the structures database as representative of the average persons-per-housing-unit statistic for its census tract year. This approach assumes that each building is perfectly representative of the averages of the census tract, which, of course, is not the case. For example, many mixed-income tracts were quite economically segregated, and the highway right-of-way often affected only one part of a tract.
As with the first methodology from this database, five total datasets were generated: four from each of our censuses — 1940, 1950, 1960, 1970 — and a final dataset based on the closest date to freeway right-of-way clearance. As with the first approach, subsequent sections will detail the results of these datasets. The final dataset is mapped on the project’s website under the “Estimate” button (Figure 8).
Figure 8 — Map With Demographic Overlay by Race
Note: This map uses the date of the nearest clearance dataset and is differentiated by race (blue represents white residents, and green represents Black residents).
Despite its limitations, this census tract data approach has two key advantages. First, it provides an overall estimate of the total number of people displaced by highway construction. Second, it uses data that are temporally closer to the construction dates of later highways, making it especially useful for analyzing highway displacement in the 1960s and 1970s. Additionally, this approach does not require exact street addresses or house numbers, ensuring no buildings go unmatched. While this methodology yields a more comprehensive dataset, it is less specific and may be less accurate. As with the previous approach, the certainty of the data decreases as the clearance and alignment dates move further from a decennial census year.
While neither methodology is without limitations, presenting both allows for as much historical information as possible to be presented, with the hope that researchers can continue to build on the analysis started here.
Creating an Interactive Map for the Public and Scholars
Both linked databases — cleared buildings and displaced persons — were incorporated into an interactive map using ESRI’s ArcGIS Experience Builder. The map features a moveable timeline slider, allowing users to track when alignments were finalized, buildings were cleared, and highways were completed. Other historical maps, geospatial data, and documents gathered in the research phase of the project were added to this temporal framework. Users can replace the default background map with U.S. Department of Agriculture aerial photographic collages, city planning and street maps, and historical freeway plans to observe the evolving built landscape of the city and how highway alignments shifted over time.
Historical photographs of highway construction can be overlayed onto “view cones” on the interactive website, showing the visible landscape depicted in the photographs. Demographic summaries can also be displayed, along with other historical geographic information including pre-expressway highway designations, racial population distribution maps, Home Owners’ Loan Corporation (HOLC) “redlining” maps, routes of historical urban railways, and the city of Houston’s historical industrial districts. Additionally, important historical news articles detailing planning decisions, alignment changes, and public protests are available to view.
By archiving these diverse types of historical sources with temporal and geospatial metadata, this project has made a vast amount of information available to scholars and the public in an organized and user-friendly format.
The project in its current form is not the end of our research, nor is it the end of our engagement with the public. In the future, several other features — such as allowing the public to upload historical photographs and personal histories of highway displacement — will be implemented. After giving permission and agreeing to our terms of services, users will be able to add their own histories to the website. These submissions, including personal photographs and testimonies, will be reviewed by project members before being made public. The hope is that photographs of lost houses and buildings and stories of a city transformed by freeways will help illuminate this history. The intention is for this historical information to continue as a living research project, continually updated as new archival information is uncovered or new methods of analysis are developed.
Both scholars and the public are encouraged to make use of these datasets, and the hope is that the scholarly conversation will continue on the project’s website. Additionally, our research team will continue its work on the historical impacts of freeway construction, as plans for in-depth discussions of key neighborhoods — such as the Fourth Ward, Riverside Terrace, the Fifth Ward, Wesleyan Place, Frenchtown, and Independence Heights — are already underway. Important primary sources, such as flyers and maps of aborted plans for the Harrisburg Freeway and a map documenting business owners from Denver Harbor and the Fifth Ward who protested a highway-induced road closure, are also slated for inclusion and discussion. Additionally, our research team has outlined plans to investigate where displaced residents and businesses relocated to as Houstonians adapted to freeway displacement.
As this work continues, our team has come to appreciate how cataloging and mapping the history of Houston’s highways reveals the incredible vibrancy and durability of its communities.
Click here for the full-view version of the “roadsTaken” interactive map and database of Houston’s highway displacement.
Previous Section | Next section
Notes
[1] Ben Poston and Liam Dillon, “How We Reported the Story on Highway Displacements,” Los Angeles Times, November 11, 2021, https://www.latimes.com/california/story/2021-11-11/how-we-reported-the-story-on-freeway-displacements.
[2] Poston and Dillon.
[3] Matt Drwenski, Uilvim Ettore Gardin Franco, and Bruno Sousa, “roadsTaken: The History of Highway Displacement in Houston,” Rice University’s Baker Institute for Public Policy, https://www.bakerinstitute.org/roads-taken.
[4] Houston Public Library, “Houston and Texas,” https://houstonlibrary.org/sb.php?subject_id=209042; Texas Department of Transportation, “Real Property Asset Map,” https://maps.dot.state.tx.us/AGO_Template/TxDOT_Viewer/.
[5] “Annual Right-of-Way Statistics,” Federal Highway Administration, updated October 13, 2022, https://www.fhwa.dot.gov/real_estate/uniform_act/stats/.
[6] Ralph Ellifrit, interviewed by Louis Marchiafava, Houston History Research Center (HHRC) Oral History Collection, Houston Public Library Special Collections, September 26, 1979, https://cdm17006.contentdm.oclc.org/digital/collection/p17006coll119/id/602/rec/2.
[7] “1950 Census,” 1950census.archives.gov., accessed November 20, 2023, https://1950census.archives.gov/.
[8] “Questions Asked on the 1950 Census,” National Archives, October 29, 2020, https://www.archives.gov/research/census/1950/questions-asked.
[9] Gene Denby, “On The Census, Who Checks ‘Hispanic,’ Who Checks ‘White,’ And Why,” Code Switch, NPR, https://www.npr.org/sections/codeswitch/2014/06/16/321819185/on-the-census-who-checks-hispanic-who-checks-white-and-why.
[10] Beverly M. Pratt et al., “Infographic: Measuring Race and Ethnicity Across the Decades, 1790–2010,” Population Division, U.S. Census Bureau, https://bit.ly/3Yjj3sh.
[11] Dan Bouk, Democracy’s Data: The Hidden Stories in the U.S. Census and How to Read Them, (MCD, 2022), 135–44.
[12] David L. Word and R. Colby Perkins Jr., “Building a Spanish Surname List for the 1990’s—A New Approach to an Old Problem,” Technical Working Paper No. 13, Population Division, U.S. Census Bureau, (March 1996), https://www2.census.gov/library/working-papers/1996/demographics/POP-twps0013.pdf.
This material may be quoted or reproduced without prior permission, provided appropriate credit is given to the author and Rice University’s Baker Institute for Public Policy. The views expressed herein are those of the individual author(s), and do not necessarily represent the views of Rice University’s Baker Institute for Public Policy.
