It's Not Just Murder That Is Falling Rapidly
Death rates from several causes in the US are falling fast.
It has been a dark year in a lot of ways. My Saints are 3-10, my Pelicans are 3-1920, my Nationals fired everybody after going 66-96, and my Longhorns went 9-3 and will miss the playoff after starting the year as the preseason #1 team in the greatest injustice in modern American history.
But I don’t want to focus on those depressing sports outcomes, I want to look at areas that are unquestionably improving.
Murder is one area that I harp on quite a bit. Murder rose a ton in 2020 and stayed that high for most of the next two years. It has been in freefall since then with 2025 on track to have the largest one-year drop in murder ever recorded potentially to the lowest rate ever recorded (FBI counts starting in 1960). This year may feature the lowest murder rate ever recorded (more on that next week).
The drop in murder nationally is clearly seen in the below graph of murders rolling over 12 months nationally from the Real-Time Crime Index.
It’s not just police departments showing this. The same trend is clearly visible in Gun Violence Archive data showing fatal shooting victims rolling over time.
And then there’s homicide data from the CDC pointing to the same downward trend though with more of a lag than the RTCI and GVA have. Note that the lag has been made worse by the government shutdown so I stopped my count of CDC data in January. The trend in homicides is identical to what we see elsewhere and a good sign that the official data will eventually match the unofficial data.
But it’s not just homicides that are falling fast in the United States. Deaths from traffic accidents, drug overdoses, and alcohol-related causes are all falling fast in a similar manner to homicides based on data from the CDC. Here is the graph of each of those types of deaths rolling over 12 months:
And here’s an estimate of the rate of deaths per 100k for each issue rolling over 12 months.
Last Week On The Podcast:
Ganesha Martin was a really cool guest to talk to because she has worked to improve policing at a number of different places. Ganesha started doing police reform in Baltimore and is now running reform in the Minneapolis Police Department. We had a great conversation about data, police reform, and how to improve policing in America.
Give it a listen on Spotify, Apple, Amazon, Substack, or wherever you get your podcasts! Also, all episodes of the podcast are now being uploaded to Substack so you can grab them straight off the page here.
The homicide, drug overdose, motor vehicle accident, and alcohol death trends share a number of similarities.
All of these problems were generally worsening during the 2010s though the increase was much steeper for drug and alcohol-related deaths than homicides and traffic accidents in that span. Each type of death got far worse starting in 2020, peaked sometime between 2021 and 2023, and has been plunging ever since per the CDC’s data.
It’s hard to say exactly what is happening past the beginning of this year because of the government shutdown plus the inherent 6-ish month lag in the CDC’s data. It’s clear that homicides have continued the downward trend into 2025 based on RTCI and GVA data.
But what about traffic accident, drug overdose, and alcohol-related deaths?
The National Highway Transit Safety Administration (NHTSA) reported in September that the number of fatalities in the first half of this year was down 8 percent. The drop expands slightly to nearly 9 percent drop in the rate of fatalities per 100 million miles traveled.
Two factors stand out to me as likely contributing causes in falling traffic deaths: COVID ending and surging infrastructure spending. The theory goes that fewer cars on the streets and less traffic enforcement during COVID led to riskier driving behaviors and more fatalities. COVID-driven stress probably created worse drivers too, though I’ve fortunately forgotten just how stressful it was spending several weeks cooped up with 4 kids under 5.
Jumps in traffic enforcement over the last few years (see NOLA’s traffic stops dashboard here, for example) and a huge surge in infrastructure spending can be argued as prime drivers in traffic accident deaths peaking and falling. The US Department of Transportation has excellent open data, and the large surge in state and local government spending on highways/streets and lighting is clearly evident in that data.
It’s also plausible — though I’m just spitballing — that falling motor vehicle thefts is contributing to fewer traffic accident deaths. There’s something of a correlation between the two metrics — even accounting for seasonality — and it would make sense that fewer people stealing cars would make for fewer reckless drivers (and wreck less!). Still, the relationship isn’t incredibly strong and I’d call the linkage a somewhat wild guess at best.
Next, drug overdose deaths were preliminarily down 25 percent in the 12 months ending in March 2025 compared to the April 2023 to March 2024 period according to a paper released in November. Why are deaths declining? I won’t pretend to be an expert on this issue, but that paper points to a number of potential factors:
Two crucial drivers may be less widespread drug use and changes in the fentanyl supply.
…
Additional influences likely contribute, such as naloxone distribution among laypersons,17 adulteration of the fentanyl supply with less lethal substances (for example, with xylazine),18 more people avoiding fentanyl injection through other, possibly less lethal,19 routes like smoking,20 and easing of pandemic-related disruptions (to trusted drug suppliers, treatment availability, and use of drugs with people instead of alone).15
Finally, alcohol-related deaths are also falling fast. This seems to be pretty clearly tied to changing reduced alcohol consumption patterns over the last few years.
A Gallup poll from July 2025 found only 54 percent of Americans saying they drink alcohol, down from 67 percent in 2022. The question has been asked by Gallup on and off since the 1930s (it wasn’t asked in 2020), and the 54 percent in 2025 is the lowest share ever recorded.
People drank a lot of alcohol during the pandemic and are drinking less now producing better health outcomes. Seems simple enough though there are undoubtedly other factors at play in this trend (again, not an expert on the issue).
The biggest overarching takeaway for me from these trends is that the pandemic was a helluva thing that damaged society in a myriad of horrible ways. Though these problems are improving — some rapidly — they remain intractable issues that are by no means solved. Drug overdose deaths, for example, fell dramatically in 2024 but were still more common than in 2019 — and way more common than in 2010.
Finally, strong policy responses to enduring problems can lead to better outcomes. I’ve previously argued that investment in communities in the 2021/2022 timeframe helped bring murder down over the last three years. Similarly, making highways & streets safer through higher spending on infrastructure has helped bring down deaths from traffic accidents. And making naloxone more easily available is contributing to fewer drug overdose deaths.
These problems persist as serious issues, but sometimes it’s simply nice to remember that even intractable problems can improve.
Maybe there’s hope for the Pelicans after all...oh, wait, I forgot to update my opening paragraph. They’re 3-21 now.
Exceptional work connecting these disparate trends into a coherent narrative. The synchronized timing across all four causes is the real story here - the 2020 spike followed by post-2022 declines suggests something deeper than category-specific interventions. What stands out most is how policy responses emerged at different scales: federal infrastructure bills for traffic, community-level naloxone distribution for overdoses, localized policing shifts for crime. This decentralized approach might actually explain the sustained improvments better than any single national policy could. The fact that we're seeing broad gains without a unified top-down strategy hints that addressing societal shocks requiers matching solutions to the specific mechanisms of harm rather than searching for one-size-fits-all answers.
All positive trends, hopefully to continue. Drug overdose data suffers from issues that would be familiar to people working with crime data. In surprisingly large sections of the country, cause of death may be determined by an elected coroner, who may not be a medical professional. Where death is ruled as caused by an overdose and lab work is performed, it is either incomplete as to all drugs present or so complete as to required informed judgement as to the primary cause of death. However, also like crime data, much work and investment since 2010 has improved the timeliness, accuracy and completeness of data to make trending reliable enough. Credit here to partnerships with CDC augmented by support from ONDCP.
The period of the charts from 2010 to 2025 roughly maps the transition of opioid overdoses and deaths from prescription opioids though heroin, and to fentanyl and its derivatives. It also marks a key transition in the response to overdoses from overwhelmingly criminal justice, to growing partnership with public health, and attention to harm reduction. The wide distribution from naloxone has gone from rejected as encouraging moral hazard to standard issue. Distribution of fentanyl testing strips, medicine for opioid use disorder (MOUD) beginning during incarceration and other practices are reducing fatalities. Folks are recovering.
Darkly, the acceleration of fatalities since fentanyl has dominated the marketplace may also play a role. You can see it in the chart. Waking up to the problem with the diversion/overprescribing of medical opioids has helped curtail what was estimated to be the production of 1,000,000 new dependent persons a year, but you can do the math of how many people have died since 2010, not as a rate but as a staggering total of users.
Things change and I am a couple of years out of this now, but I do wonder about a conclusion that fentanyl has been diluted by other substances and drugs to a less lethal level. What has been more prevalent in my experience is that other drugs have been mixed with fentanyl to yield all but heroin, methamphetamine or cocaine flavored fentanyl. Similarly, fentanyl has been pressed into counterfeit medication pills such as Xanax, perhaps as concealment, but also resulting in unwitting exposure and overdose. Still, like the other charts, we need not to so much take comfort as to find and encourage practices that are contributing.