Lowering the Legal BAC to 0.05?

A recent peer-reviewed study claims that Utah's 0.05 law works, so we can expect to see this study cited in various news articles and proposed legislation.

In my practice I defend people charged with operating while intoxicated, and I am frequently asked to testify as an expert on breath testing and standardized field sobriety testing. So when a study makes national news claiming that lowering the legal limit to 0.05 saved lives, clients, colleagues, and occasionally judges ask me what to make of it. I think the honest answer matters, because the number that separates a lawful driver from a criminal defendant should rest on sound evidence rather than on a well-timed headline.

The study was published this year in the American Journal of Preventive Medicine (Li, Fell, Hosseinichimeh, Hu & Vaca). Its message has been confident: Utah's move from a 0.08 to a 0.05 blood alcohol limit, still the strictest in the nation and the only one of its kind enforced in this country, drove down alcohol-related crash deaths more than in Utah's neighboring states, and the rest of the country should follow. The press coverage went further, telling readers that drunk-driving deaths "decreased across the board from 2018 to 2023." When I set that claim beside Utah's actual year-by-year fatality counts through 2023, the headline does not survive contact with the data.

What the study set out to measure

It is worth stating the study's design fairly. Utah's 0.05 law passed in 2017 and took effect on December 30, 2018. To measure its impact, the investigators analyzed fatal-crash data from the federal Fatality Analysis Reporting System (FARS) and compared Utah with its six contiguous states: Arizona, Colorado, Idaho, Nevada, New Mexico, and Wyoming. The central method was a difference-in-differences comparison, a technique that asks whether Utah's change differed from the change in similar nearby states over the same period. For that comparison the authors used a subsample restricted to crashes occurring in two years, 2016 before the law and 2019 after it. That single before-and-after pair produced the study's statistically significant results, including its marquee finding of a large drop in deaths involving an alcohol-positive driver.

The difference-in-differences design is a genuine strength, because it attempts to control for regional trends that a single state's raw counts cannot capture. But the design carries a vulnerability that a two-year snapshot cannot reveal: it depends entirely on the two years chosen. In Utah, 2019 was among the lowest years on record for alcohol-related deaths, and the years on either side of it were considerably worse. When the post-policy year sits near the bottom of the range, almost any intervening policy will appear successful.

Utah's full record, 2016 through 2023

The table below sets out Utah's statewide fatality counts, broken out by the driver's blood alcohol level, for the full period the study examined. The thresholds are cumulative: the column for 0.08 and above, for example, includes every driver at 0.08 or higher.

The 2019 row tells the story. In every alcohol category it sits at or near the bottom of the eight-year range. The any-alcohol count of 49 is the second lowest of the eight years; the count of 39 for drivers at 0.08 and above is also second lowest; and the 26 deaths involving drivers at 0.15 and above is near the floor as well. What the difference-in-differences comparison captured, in other words, was largely the distance from a high year, 2016, to a low one, 2019.

The reduction did not persist

If a lower limit genuinely changed how Utah drivers behaved, the improvement should have held. It did not. Within two years, every alcohol category had climbed back above its 2016 starting point, not merely above the 2019 low, but above the very baseline the study used to claim a reduction. Deaths involving a driver at any alcohol level rose to 89 in 2021, compared with 66 in 2016, twenty-three more than before the law. Deaths involving a driver at 0.08 and above reached 79 in 2021, again twenty-three more than the 56 recorded in 2016. And deaths involving the highest-BAC drivers, those at 0.15 and above, rose to 52 in 2021 against 34 in 2016, eighteen more. An effect that reverses this completely within twenty-four months is difficult to attribute to a durable change in driver behavior; it is more consistent with the ordinary year-to-year variation that a small state's fatality counts always show.

"Decreased across the board," measured against the data

The press summary's specific claim was that deaths fell "across the board from 2018 to 2023." Running those exact bookends against Utah's own numbers does not support it. Total traffic deaths rose over that span, from 260 to 280, an increase of about 7.7 percent. Non-alcohol deaths rose more sharply, from 190 to 211, about 11.1 percent. The alcohol-involved categories barely moved, down two or three deaths apiece, changes far smaller than the ordinary swings in a series that fell from 89 to 43 in the single span from 2021 to 2022. Nothing decreased across the board. The two categories that moved the most, total deaths and non-alcohol deaths, moved up.

Alcohol's share of fatalities did not improve

There is a cleaner way to test whether a law aimed at drinking drivers actually changed their behavior. If it did, alcohol's share of all traffic deaths should fall, even in years when the overall total bounces around. In Utah, the share moved the wrong way. Drivers with any alcohol present accounted for about 22.1 percent of all traffic deaths in 2016 and about 24.3 percent in 2023. Drivers at 0.08 and above went from about 18.7 percent to about 21.1 percent. The highest-BAC drivers went from about 11.4 percent to about 14.6 percent. Every alcohol share ended the period higher than it began. Whatever was happening on Utah's roads, drinking drivers did not become a smaller part of the toll.

The study's own longer analysis

The most important point is not my arithmetic on Utah's raw counts, which cannot by itself prove what would have happened without the law. It is that the study's own data undercut its headline. The published appendix contains a second difference-in-differences analysis using a longer, more representative window, several years before the law compared against several years after it. In that version, the dramatic results do not hold. For drivers at 0.05 and above, at 0.08 and above, and at 0.15 and above, the reductions are no longer statistically significant; only the broadest any-alcohol category retains significance. The authors say as much in their own discussion:

Sensitivity analyses using the longer observation period ... showed a similar pattern of reductions in alcohol-related fatalities in Utah relative to those in contiguous states, although the estimates were not statistically significant.

That concession deserves emphasis. The strong conclusion being promoted to the public depends on comparing 2016 to 2019 and stopping there. Extend the observation window, as the authors themselves did, and the statistical support for the threshold-specific reductions falls away. Utah's full record through 2023 shows why.

What can fairly be said

I want to be fair to the data and to the researchers. Two things are true. Utah's alcohol-related deaths did fall from 2016 to 2019; the study did not invent that decline. And a single state's year-to-year counts cannot, standing alone, prove what would have happened in the absence of the law, which is precisely why the authors compared Utah to its neighbors in the first place. Those are real limits, and an honest critic should name them. The national backdrop the study invokes is also real: alcohol-impaired driving remains a serious problem, and the federal data the authors cite show thousands of alcohol-involved fatalities each year and an increase over the past decade.

None of that, however, rescues the claim being offered to the public. A reduction that appears in one favorably chosen year, reverses within two, leaves alcohol's share of deaths higher than before, and loses statistical significance under the authors' own longer analysis is not the robust evidence that a nationwide rewrite of impaired-driving law would require. It is a single good year asked to carry a great deal of weight.

Why this matters in Michigan

Michigan currently sets its per se threshold at 0.08 under MCL 257.625, and proposals to lower that threshold to 0.05 are constantly under discussion here, with Utah held up as the proof that the change works. I have no quarrel with the goal of fewer deaths on Michigan roads; in my practice I see the human cost of impaired driving from both sides of the courtroom. But lowering a criminal threshold is a serious legislative act. It redraws the line between lawful conduct and a misdemeanor, and it does so for every driver in the state. A change of that consequence should rest on the full eight-year picture, not on a snapshot taken at the bottom of a dip. On the complete record that Utah itself reported, the 0.05 experiment did not deliver the lasting reduction its champions describe. When I evaluate evidence, whether it is a breath test in a single case or a study offered to justify a statewide policy, I look for results that hold up when the window is widened and the favorable assumptions are removed. By that ordinary standard, this study does not yet make the case.