Abstract
In a letter to Time & Society, José Maria Martín-Olalla (2023b) offers a critique of my team’s study on traffic fatalities in misaligned time zones in the United States (Gentry et al., 2022). Unlike research that is met only with the sound of crickets, it is positively refreshing when other scholars take notice. This essay is reminiscent of his response to Malow (2022) on the subject of Daylight Saving Time (Martín-Olalla 2023a), and other constructive critiques.
Our study identified a 21.8% elevated risk of traffic fatalities in eccentric time localities (ETLs) compared to solar zones in the United States. Dr. Martín-Olalla’s essay takes issue with what he calls this “staggering large” result (p. 1). His subtitle is “Not that elevated” (p. 1).
In framing my reply it helps to consider Martín-Olalla’s (2023b) early statement, “In their analysis, the authors implicitly assume that F [fatalities] scales with P [population] through different geographic localities” (p. 1). Far from assuming, we tested a falsifiable hypothesis using data from the Fatality Analysis Reporting System (FARS) annual census. This carefully managed resource, curated by the National Highway Traffic Safety Administration (NHTSA), enumerates each U.S. highway death by county. Whole-population data carry the advantage of no standard error and no need to eliminate data (Alexander, 2015), a concept we explained rather than assumed. Whole-population data are appropriately analyzed via descriptive rather than inferential statistics.
In contrast, Dr. Martín-Olalla’s (2023b) extensive meta-statistical analyses seem to assume we collected only a sample of the population rather than the whole population. Note his concern about our “scarcity of observations” (p. 3). In fact, we accounted for each road fatality over a 12-year period, not a mere sample (n = 417,399). We cited Alexander (2015) and Wood (2016), who explain why inferential-statistical tests are not only unnecessary for whole-population data, but counterproductive. Wood (2016) notes, “Obviously if you’ve got the time and energy and resources to study the whole population you should do so because then you get the whole answer and don’t need to mess around with p values or confidence intervals etc. This must be more rigorous. There is obviously no good reason to ignore data you can get!”
Naturally, factors such as speeding and drunk driving influence road fatalities, as we acknowledged. But a 21.8% higher death-rate was the empirical difference between solar zones and ETLs over 12 years. These are not estimates, not extrapolations, but actual death counts. The pattern held across all four principal U.S. time zones. Eastern Time’s ETLs were 23.8% worse than its solar zones. ETLs in Central Time were 17.7% worse, and Mountain Time’s ETLs were 26.5% worse. The combined vehicle-fatality rate in these ETLs was 50.4% higher than in Pacific Time, which contains no ETLs in the United States.
Again making probabilistic assumptions, Martín-Olalla (2023b) calls for “More observations” to produce a “precise value” (p. 3). Our data-set included each road fatality over 12 years. How many more observations can there be above complete enumeration? He concludes: “Very likely the impact of the eccentric time zones is not an increase of 20%” as we claimed (p. 3, emphasis mine). Such questions of likelihood entail sample data, with concerns about standard error, p-values, and the like.
Dr. Martín-Olalla’s (2023b) comment that “Disaggregated numbers by county may provide some insight” (p. 3) is puzzling because this is precisely what we did. Our narrative and tables plainly report comprehensive data from 3094 counties, each of which we carefully placed in its respective solar zone or ETL according to the sun’s natural movement. His Table 2 (p. 3) would even suggest that ETLs in the Central zone were safer than solar counties (vehicle-fatality rate minus 2.3%). Empirically, however, ETLs in Central Time indicated a death rate 17.7% higher. Superimposing point-estimation here (with no unknown parameter) masked reality, another case of forcing sample-version assumptions on whole-population data.
Our findings came as no surprise to chronobiologists and neurologists. As our literature-review noted, chronic circadian deficits are associated with higher risks of stroke, heart disease, obesity, dementia, cancer, and now, highway deaths. We then provided measures of practical significance, such as the total number of excess deaths, years-of-life lost, and associated economic costs. We never used the word “causality,” but I am convinced by established theory and research in chronobiology that natural circadian alignment is beneficial for the health and safety of human beings.
Our Time and Society article even addressed the problem of “scientific ritualism” (Cowger, 1984), in which the same statistical tests are applied whether they are appropriate to the data or not. Dr. Martín-Olalla’s (2023b) essay overlooks this point, and how we expressly justified our methodological and statistical choices. Alexander (2015) notes that scholars continue to apply inferential-statistical tests regardless of their appropriateness. Even taken at face value, Dr. Martín-Olalla’s strenuous manipulations of data produced a 3.9% worse outcome in ETLs. This result would be publishable, but it would also be inaccurate.
In Dr. Martín-Olalla’s defense, it is understandable when scholars are unfamiliar with whole-population (a.k.a. census) data due to their relative scarcity. These data are not unheard of, however (Alexander, 2015), considering the exhaustive effort that national governments expend enumerating statistics on health and well-being in their decennial or 5-year censuses. We provided an example of how they can be analyzed without discarding salient data over misplaced concerns resulting from inferential-statistical assumptions.
Finally, our study provides an improved framework for research on the “east-west gradient” (Martín-Olalla, 2023b: p. 1). Prior studies focused on time-zone boundaries or their western edges. But some time zones and western edges are solar-appropriate, such as Pacific Time in the U.S. Distinguishing eccentric time localities from solar zones addresses the problem more scientifically. We also weighed the policy implications, noting how artificial time zones are rhetorical constructs and calling for U.S. time-zone boundaries to move east. Last fall in Barcelona, international scholars with the Working Group on Natural Time (2022) called on European nations to adjust their erratic time-zone boundaries to better align with solar time, as well.
To summarize our study’s contribution, we (1) provided a natural experiment that found an elevated risk of death on U.S. roads in eccentric time localities, (2) identified statistics for the analysis of whole-population data, (3) raised awareness of the problem of scientific ritualism in academic research, and (4) introduced a parsimonious framework to distinguish eccentric time localities in chronobiology.
In conclusion, the empirical death rate in ETLs was 21.8% higher than solar zones over a 12-year period in the U.S. Although Dr. Martín-Olalla’s introduction said he would offer an alternative explanation for our findings, no alternate causality is noted aside from “Not that elevated.” The most parsimonious inference is that our results support chronobiology theory. Modern-day insults to natural circadian processes are associated with negative health and safety outcomes, including traffic fatalities. However, I applaud Dr. Martín-Olalla’s continued scrutiny of methodological and statistical choices in academic research. Discussion and debate fuel creative energy and innovation. In contrast, science without healthy debate quickly becomes moribund.
