Does increased safety really reduce worker productivity?

I just came across the linked paper [https://doi.org/10.1177/0019793920931495] that analyzed the impact of MSHA flagrant violations levied on US mines on worker productivity in those mines. The author, Ling Li, found that productivity was reduced after flagrant violations were issued and that these costs together with the associated fines outweighed the benefits from reduced safety incident rates that resulted after these fines were levied.

The MINER Act that authorized the MSHA flagrant violations rule (fines of more than $220,000) was passed in 2006. Li’s paper confirms that mines experiencing a flagrant violation meaningfully improved safety and reduced injuries after being cited but also finds that the tons of coal per hour produced at those mines also decreased resulting in a total cost impact 30% greater from decreased production than would have been paid to treat injured workers.

I think there are a few weaknesses with this conclusion that could be addressed in future analyses.

• First, all mines are potentially affected by publicized flagrant violations, not just the mine that received it, since these are made public, and other management teams are likely to pay attention and update procedures where necessary to avoid a similar fine themselves. An analysis that considered the industry as a whole rather than the effects on individual mines would be more relevant to the effectiveness of the regulation.
• Second, there are potential explanations for reduced coal output beyond safety-related changes in response to a fine from MSHA including mine age resulting in depleting coal seams, or reduced investment by the owner to focus on other sites, or a loss of more experienced workers. These and other alternative explanations for the decline in productivity need to be investigated rather than assuming it is the result of the response to the flagrant violation.
• Third, the costs of safety incidents were estimated at a flat rate of $69,363 each, and there weren’t enough fatalities occurring (thankfully) to be able to assess whether these were reduced as well. This is important since the policy analysis of the rule was heavily dependent on reductions in fatalities which have values around $10 million each. However, fatalities usually result from the same sorts of events that cause less severe injuries, so it should be possible to assess the expected number of reduced fatalities based on the observed reductions in more common injuries.
• Lastly, the comparison groups of mines with and without flagrant violations are matched via another mine at the shortest distance rather than matching mine characteristics such seam height, surface or underground, equipment employed, mine size in terms of overall production or employment, and so on. These attributes can dramatically change the baseline of risks to workers. Our group has some previous research on this topic: https://doi.org/10.1016/j.ssci.2018.10.005. While most mines nearby are likely to share some of these characteristics, they need to be explicitly included in the analysis as predictors, otherwise conclusions could be misleading.