Following is an in-depth review of the Houston Public Works 3/16/2024 draft report on the 11th Street Safety Improvement Project cited in Axios. The author of the review was Jason Starr, a member of the Houston Heights East community, who does financial modeling, forecasting and data analysis for a living. He wanted to see and understand for himself what the report and available data showed. Jason came to the conclusion that “In the 24 months following completion, total crashes in the TEST area (same as HPW Study Area)…are up 15x compared to the control group...." Also, "...[C]rashes are...up since the project was completed. Not only are they up, but they are up a higher rate than the control group, and the entire COH."
For more information re how the review was done and comments, see his nextdoor.com post here. Because we are limited on the number of photos we can upload and the number of times we can post (every 24 hours), more will come later!
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
I. HPW Draft Report Background
The Houston Public Works (HPW) released a draft report on 3/16/2024 to report on the results of the 11th Street Safety Improvement Project. This report stated that, “Study Period crashes throughout the Study Area decreased from 45 in 2019 to 40 crashes in 2023, an 11% decrease, and injury–causing crashes decreased from four in 2019 to two in 2023, a 50% decrease.” This statement largely aligns with my findings when using the same time samples. (In the newly updated research, total crashes may be up or down one or two as great effort was made to mirror the polygons used in the original HPW study)
The HPW study also provides data on pedestrian and bicycle volumes at key crossing locations. This would seem to be good information, but the before and after only consisted of 1 day (before) and 1 day (after). Such a limited sample size effectively renders findings meaningless. Similarly, the HPW study did not capture “pre’ measurements for travel time delays. To gauge the impact on travel times, a simulation model was used, and the HPW study determined that delays were minimal.
The HPW study compared a 6-month window (April through May) in 2019 to the same 6-month window in 2023. The HPW study compared crashes that occurred in a defined Study Area that captured 10th, 11th, 12th streets, from N. Shepherd to Studewood, and a small section of White Oak Dr. from Norhill Blvd to Usener St.
HPW Study used the year 2019 as the base period and noted, “pandemic data should not be used for comparison as it is based on significantly different travel volumes and patterns”. The year 2020 was significantly affected by the pandemic, and one could make the case 2021 was partially impacted. As you will see, 2022 was not affected by the pandemic.
Had the study period (April through May) 2023 results been compared to the same period in 2022, it would have shown a 35% increase in total crashes. It appears that 2019 was used due to having an unnaturally high base period and therefore yielding a more favorable reduction in crashes in the study. For this exact reason, a study of this nature demands the use of a comparison group (control). Comparing only one sample, even in a before / after scenario, is not recognized as a valuable means of actionable conclusions.
The author of the HPW study notes, “six months of crash data is not typically considered sufficient to assess long-term crash patterns”. (HPW Study, Sec. C, p. 11.)
II. The Analysis
We now have 24 full months of data since significant completion. Using the same Study Area outlined in the original HPW study, a comparison was made between the 24-month period before construction (pre), and the 24-month period after significant completion (post).
Table A below shows data for the TEST area, for the 24-month period before and after. Table B below shows the same time frames, shows total crashes for the TEST vs. CONTROL area. This is a larger area south of I-10, filtered identically.
HPW Study Area (TEST)_24 Months Pre_24 Months Post
Table A
______________________________________pre____post____+(-)_____%____
all crashes……………………………….........…….......115 148 33 28.7%
pedestrian / pedalcyclist-involved crashes….. 4 4 0 0%
injury-involved crashes……………………….......... 35 35 0 0%
death-involved crashes…………………………....... 0 1 1 100.0%
____________________________________________________________________
(Pre) November 2020 - October 2022; (Post) April 2023 - March 2025
* * * * * * * * * *
(TEST vs. CONTROL_all crashes)_24 Month Pre_Post
Table B
____________________________________pre____post____+(-)_____%___
all crashes (TEST GROUP)…………………..... 115 148 33 28.7%
all crashes (TEST GROUP)……………….....5,255 5,354 99 1.9%
________________________________________________________________
(Pre) November 2020 - October 2022; (Post) April 2023 - March 2025
III. Conclusion
In the 24 months following completion, total crashes in the TEST area (same as HPW Study Area) are up 28.7%. Total crashes in the CONTROL area are up only 1.9% (total crashes in the Study Area used by HPW study, are up 15x compared to the control group.
There have been articles published in multiple publications that cite the HPW draft study while making a case the project has been successful in reducing the number of crashes. Quotes from these articles are heavily touted by proponents of the project. However, the data now show that the number of crashes in the Study Area have increased significantly when compared to a larger sample size (Control Group).
The data presented in this file have been filtered to preserve the same conditions that were used in the HPW Study. As some in the community have noted, there is a misattribution issue stemming from crashes occurring on Interstate-10 (or the I-10 feeder). This means that there have been several crashes occurring on or near I-10, but the database is reporting that they are occurring on 10th St. Fortunately, each of these instances are also tagged with an “interstate” attribute making it simple to filter out of the data.
All data appearing in this report are publicly available on the TxDot Crash Query Tool. All data used in HPW Report were also retrieved with the same tool.
I encourage anyone to use the same tool to research for themselves.