This study is the first to evaluate the short-term impact of cannabis legalisation in Germany on both cannabis use and DUIC. Employing a DiD study design with Austria as control, the findings suggest that cannabis use further increased in Germany, however, this trend is comparable to what can be observed in Austria, where cannabis possession and cultivation remained illegal. Moreover, DUIC among at least monthly cannabis users appears not to have increased in Germany in the first months after the liberalisation of cannabis laws for personal use and for traffic participation.

Commercial cannabis legalisation in North America has been linked to small increases in use prevalence (recent overview:⁴), particularly in jurisdictions with more liberal policies,²⁹ driven by, for instance, increased availability of legal cannabis.³⁰ In Germany, however, commercial sales of recreational cannabis remain prohibited and only a few dozen ‘cannabis clubs' had been authorised, covering less than 1% of estimated users at the time of the survey. Meanwhile, the rapidly expanding medical market—imports of medical cannabis flower nearly quadrupled in 2024³¹—may partially offset restricted availability. It remains to be investigated how the availability of medical and recreational cannabis in Germany affects cannabis use prevalence. So far, the non-significant increase in past-year use prevalence observed in this and another survey³² suggest that prior trends⁹ remain unchanged by the Cannabis Act.

Official crash statistics show a continued rise in MVCs under the influence of psychoactive substances (including but not limited to cannabis), with no abrupt change following legalisation.³³ Our findings are consistent with this trend: population-level DUIC is believed to increase with rising cannabis use prevalence at stable DUIC rates among users. As this trend was also observed in Austria, no short-term changes appear to have been caused by cannabis legalisation. While both MVC statistics and survey data suggest that pre-existing trends have continued, two caveats in the data should be considered. First, MVC statistics do not allow a quantification of the cannabis-specific contribution. Second, the population-level DUIC estimate (Table S9.3) relies on the assumption that DUIC among less than monthly users remained unchanged; violations of this assumption would bias the estimate. Embedding this finding in previous research,⁴ we can summarise that cannabis legalisation is not necessarily causing a short-term increase in DUIC. Nevertheless, long-term monitoring is necessary to determine whether the legalisation of cannabis will gradually impact DUIC and traffic safety. Importantly, the implications of exemptions for cannabis patients to drive after using medical cannabis also warrant further study.

Previous findings of increased cannabis involvement in fatal crashes after legalisation⁵ or in regions with more liberal cannabis policies⁶ may be explained by higher prevalence of use in the population rather than changes in driving behaviour among users in response to legalisation. In other words, population-level increases in cannabis use may result in an increase in the number of MVCs—despite stable DUIC patterns among users. This aligns with Canadian survey data, where DUIC increased among all respondents but decreased moderately among cannabis users five years after legalisation.¹⁴ Like Germany, Canada applies a per-se THC limit (2 ng/ml whole blood, ∼2·8 ng/ml blood serum), with exceedance constituting a summary conviction offence (Blood Drug Concentration Regulations, SOR/2018-148). Among users, our data did not indicate short-term behavioural adaptation in DUIC following the introduction of the new THC-limit for most adult drivers (from 1 ng/ml to 3·5 ng/ml blood serum, the 1 ng/ml limit for novice drivers, those under 21, and combined use with alcohol remained unchanged). Such behavioural adaptation may be moderated by enhanced deterrence¹² or prevention measures. In Germany, the legalisation coincided with prevention campaigns (e.g., “Don't Drive High”) and announcements of intensified roadside controls by police authorities in several federal states; however, these factors were not examined in the present study. Similar to previous studies, we cannot ascertain if self-reported DUIC was consistent with the THC-limit, as no objective blood measurements were available. Subtler behavioural adjustments—such as shorter waiting times before driving—would not be captured by our DUIC measure, but may be detectable with THC blood concentrations. A regionally limited study from southwest Saxony using administrative police data from roadside drug testing observed higher THC blood concentrations following legalisation³⁴; however, the absence of information on testing strategies limits the interpretability of these findings. Therefore, more robust evidence is needed from studies examining changes in THC levels in relation to cannabis legalisation.

Findings of our explorative analysis suggest that the majority of DUIC episodes involved only cannabis and no alcohol or other drugs. Most studies on this subject have not addressed co-use of alcohol or other drugs in the context of DUIC, despite elevated risks and differential fines for mixing cannabis and other substances (e.g., Canada: Blood Drug Concentration Regulations, SOR/2018-148). DUIC(−) episodes were concentrated among daily users, whereas DUIC(+) episodes were concentrated among weekly users. While our study does not provide evidence on impairment by use frequency, previous research suggests that daily users may experience less impairment from the same dose of THC due to tolerance.¹⁶ Therefore, the greatest potential traffic safety risk among cannabis users may occur in those using cannabis weekly and combine it with other substances before driving. Further research is required to assess the implications for traffic safety. Identifying a key research gap, we propose that future studies account for the co-use of alcohol or other drugs in the context of DUIC. This differentiation may be crucial to deepen our understanding on the impact of cannabis legalisation on traffic safety.

Certain limitations should be acknowledged. We relied on quota samples from ISO-certified online access panels, that may not be representative of the general populations of Germany and Austria. While representativeness is not necessarily required for testing of the hypotheses, generalisability of the findings is limited to the recruited samples. Notably, the sample is skewed towards individuals with higher socioeconomic status while hard-to-reach populations such as people experiencing homelessness, people in institutions, or without internet access are not sufficiently covered and may exhibit other trends in cannabis use and DUIC. Comparisons with national probability-based surveys showed substantial overlap, though addictive behaviours were higher in our samples (Table S3.2). True population prevalence may not be captured; however, higher prevalence in our samples may reflect reduced underreporting, supporting the suitability for studying sensitive behaviours like cannabis use and DUIC. Gender- or ethnicity-stratified analyses were not conducted, as the primary focus of this study was on population-level effects of legalisation. Future research is warranted to assess potential differential effects.

DUIC was assessed via self-report, and no objective verifications (e.g., THC levels or tests of driving impairment) were available. Therefore, DUIC may be affected by recall and social desirability biases. However, social desirability regarding DUIC in our sample remained stable over time, making substantial time-varying bias unlikely.

As a quasi-experiment, unmeasured confounding cannot be fully excluded. Further, spillover to Austria or anticipation effects are possible; although at baseline (Nov–Dec 2023), the implementation of legalisation was still uncertain (parliament passed the bill only in Feb 2024), likely limiting behaviour changes driven by expectations of legalisation.

The t₁-survey assessed 12-month DUIC only 8 months after legalisation, so the reference period post-legalisation overlaps with the pre-period, which may lead to an underestimation of the true legalisation effect. However, a sensitivity analysis using past-30-day DUIC confirmed the absence of a significant change, suggesting the overlapping periods do not explain the null finding. While cannabis policy changes can have both immediate and lagged effects,³⁵ the present study, given its short follow-up period since legalisation, captures only immediate effects on cannabis use and DUIC.

Testing the parallel trends assumption underlying the DiD was limited by the absence of earlier survey waves. Using pre-legalisation data from national surveys and official statistics, we found no evidence of divergent pre-trends in cannabis use or drug-impaired MVCs between Germany and Austria. Although constrained by the lack of Austrian cannabis use data after 2020 and of cannabis-specific MVC data, these findings, together with the absence of major policy or enforcement changes and the overall structural comparability between both countries, support the plausibility of the parallel trends assumption.

The sample sizes for at least monthly cannabis users (Sample 2, n = 86–92 in Austria and n = 393–589 in Germany per timepoint) were sufficient to detect main effects reflecting temporal changes in DUIC within Germany. However, the smaller Austrian sample limited the power to detect interaction effects (minimum detectable DiD-effect: OR = 3·68), meaning small to medium between-country differences may go undetected. Nonetheless, the observed decline in DUIC in Germany and increase in Austria support the hypothesis of non-increasing DUIC prevalence among at least monthly users, despite limited power.

Overall, this study found no short-term impact on cannabis use or DUIC after legalisation of cannabis in Germany. Exploratory analyses indicate that daily users reported the most DUIC episodes involving cannabis only, whereas weekly users reported the most DUIC episodes involving cannabis in combination with other substances. Further studies with longer observation periods and other data, most importantly MVC involving substance use and toxicological information, are required to monitor the long-term implications of the law.