No, we should not abolish OSHA
The statistical evidence is clear: workplace safety inspections work.
Table of Contents
Introduction
On January 3rd, Republican Representative Andy Biggs introduced the “Nullify Occupational Safety and Health Administration Act”. The objective of the bill was simple: abolish the Occupational Safety and Health Administration, otherwise known as OSHA. Yes, you read that right.
At the moment, policy experts believe the bill has little chance of being signed into law. It currently sits with zero co-sponsors and would ultimately need to receive 60 votes in the Senate to overcome the filibuster. But, if there is anything that we ought to have learned during the past three months, it is that the current administration is more than happy to bypass Congress to shutter federal agencies in the name of government efficiency. In fact, the Trump administration has already taken action to:
Fire hundreds of employees at the National Institute for Occupational Safety and Health (NIOSH), leading to the suspension of the Coal Workers’ Health Surveillance Program (CWHSP) - the government’s primary method of screening miners for black lung disease.
Terminate the leases for a dozen OSHA field offices including the sole office located in Louisiana, a state with elevated rates of workplace fatalities
Given that there is a non-zero chance that the Trump administration will aim its next killshot at OSHA then, I would like to make a robust case in defense of the agency. In particular, I would like to argue that OSHA actually delivers on its promise to protect workers from threats to their safety and health. Let’s get into it!
OSHA 101
To start, what is OSHA and what do they actually do? OSHA, or the Occupational Safety and Health Administration, is tasked with ensuring that “America's workers have safe and healthful working conditions free from unlawful retaliation”. They accomplish this objective in three primary ways:
Establishing Standards & Regulations: OSHA establishes standards and regulations designed to protect worker health in a variety of situations. Examples include protecting agricultural workers from heat stroke, mandating proper labeling of chemical hazards in the workplace, and issuing standards governing lead exposure in the workplace.
Conducting Inspections: Unfortunately, standards and regulations alone cannot protect workers. They must be enforced to have any real-world impact. To that end, OSHA conducts approximately 30,000 workplace safety inspections each year to verify whether employers in high-risk industries are compliant. In cases of non-compliance, OSHA levies penalties of up to $16,550 per violation.1
Providing Resources: Lastly, OSHA provides many resources to workers, companies, and state agencies to promote workplace health. For example, OSHA provides:
Whistleblower protection services to workers who wish to report safety violations without fear of retaliation from their employer.
Educational materials and certification programs to employers who wish to remain compliant with existing standards and regulations (e.g. the Voluntary Protection Program)
Funding to state agencies who wish to target workplace safety needs specific to their region
Budget-wise, these programs led OSHA to request $655 million in funding for the 2025 fiscal year.
How Can We Estimate OSHA’s Impact?
Given that OSHA has historically been allocated hundreds of millions of dollars each year to support workplace health and safety, one might reasonably ask: how do we know whether all of this money is actually doing anything?
One initial way to tackle this question is to simply estimate whether injury & illness rates decrease at a workplace after it has been inspected by OSHA.
While this strategy has an obvious intuitive appeal, it is hopelessly contaminated by a statistical phenomenon known as regression to the mean (or as some folks like to say, “lightning doesn’t strike twice!”)
As an example, let’s suppose that a workplace happens to have a high injury rate in a given year due to sheer rotten luck and is inspected by OSHA as a result. Because the initial high injury rate was due to bad luck, we should expect the injury rate at the same workplace to go down in the next year not because of the inspection, but simply due to the absence of the “bad luck” factor. Thus, we cannot rely on simple changes in injury rates over time to understand the efficacy of OSHA inspections.
An alternative strategy might be to compare injury rates at workplaces inspected by OSHA to injury rates at workplaces not inspected by OSHA. However, this strategy is hopelessly contaminated by selection bias. Because OSHA selectively targets workplaces with high historical injury rates for inspection, any naive comparison between inspected and uninspected workplaces will be biased.
Given that neither a simple pre-post comparison nor a cross-sectional comparison can tell us the causal impact of OSHA inspections, one might readily wonder: what empirical strategies can? There are two promising candidates: the randomized controlled trial (RCT) and the regression discontinuity design (RDD).
Randomized Controlled Trials (RCT)
The first type of statistical strategy we can use to estimate the causal impact of OSHA inspections is called a randomized controlled trial, or RCT for short. The basic idea behind an RCT is to randomly assign individuals to an experiment arm (where they receive the intervention in question) or a control arm (where they do not). By ensuring that the experiment arm and control arm are comparable to one another, random assignment allows us to cleanly estimate the causal impact of the intervention in question on the outcome(s) of interest.
What relevance does this have to OSHA? Well historically speaking, OSHA has not had enough inspectors to inspect all workplaces which are deemed high risk to workers. Consequently, they’ve used randomized algorithms to determine which high-risk workplaces should be inspected during a given year.2 Drawing on the logic of an RCT then, we can estimate the causal impact of those inspections by comparing injury & illness rates between randomly inspected high-risk workplaces and randomly uninspected high-risk workplaces. If OSHA inspections truly improve workplace health and safety, we should expect to see the former workplaces have lower injury & illness rates than the latter.
Regression Discontinuity Designs (RDD)
The second type of statistical strategy we can use to estimate the causal impact of OSHA inspections is called a regression discontinuity design, or RDD for short. The basic idea behind an RDD is to compare individuals right above and right below an arbitrarily-set threshold which determines whether they receive the intervention in question. By creating two groups which are generally comparable to one another, the arbitrarily-set threshold allows us to cleanly estimate the causal impact of the intervention in question on the outcome(s) of interest.
What relevance does this have to OSHA? Well each year, OSHA administers the site-specific targeting program, or SST for short. The SST Program targets workplaces with a high historical injury rate for inspections. In particular, workplaces with a DART rate - the rate of injuries causing days away from work, restricted job duties, or transfer to a new job - above a preset threshold are substantially more likely to be inspected via the program.
To estimate the causal impact of workplace inspections then, we can simply compute the change in the rate of workplace injuries & illnesses between workplaces just above and below the DART rate threshold divided by the change in the probability of receiving an OSHA inspection at the same threshold.
The RDD strategy thus provides an alternative method of estimating the causal impact of OSHA inspections which does not require deliberate random assignment.
Poor Measurement
While the RCT and RDD strategies undeniably represent an improvement over more naive empirical strategies, they are both susceptible to the issue of poor measurement.
You see, in an ideal world, OSHA would have perfect knowledge of all injuries and illnesses that occur in every workplace in the United States. Sadly, we do not live in this world. We instead live in a world where OSHA must ask companies/workers to document injuries and illnesses and report them back to OSHA via Form 300A.
This reliance on self-reporting has led OSHA to systematically under-count non-fatal workplace injuries and illnesses.
The 2009 GAO Report
In 2009, the Government Accountability Office (GAO) issued a scathing evaluation of OSHA auditing practices, noting that OSHA inspectors often did not ask workers to corroborate whether employers were accurately keeping records of workplace injuries in illnesses. The report also noted that occupational health practitioners (e.g. on-site nurses or other trained medical staff) were sometimes pressured to downplay the severity of workplace injuries and administer treatments which would not result in a recorded safety violation:
From our survey, we found that more than one-third of health practitioners were asked by company officials or workers to provide treatment that resulted in an injury or illness not being recorded, but also was not sufficient to properly treat the injury or illness.
The National Emphasis Program
In response to the GAO report, OSHA initiated a National Emphasis Program (basically just a fancy name for a government initiative) dedicated towards improving data quality on workplace injuries and illnesses. Inspecting a set of 350 workplaces in high-risk industries, OSHA found that roughly 50% failed to accurately record at least one workplace injury and/or illness. Interviews with workers indicated numerous causes of mis-reporting including but not limited to fear of retaliation, difficulty navigating the process of reporting a workplace safety violation, and gatekeeping by on-site occupational health practitioners.
The Multi-Source Enumeration Project
Between 2009 and 2012, OSHA conducted the multi-source enumeration project, an initiative which was inspired by the previous success of the Census of Fatal Occupational Injuries (CFOI). For context, the CFOI provides a complete census of all fatal workplace injuries by leveraging multiple data sources - death certificates, coroner’s reports, local news reports, and more - to corroborate the existence of a workplace fatality. The hope with the multi-source enumeration project was that the same approach could be applied to construct a complete census of non-fatal workplace injuries: leverage OSHA recordkeeping logs, workers’ compensation databases, and hospital records to accurately tally the # of non-fatal workplace injuries.
Sadly however, this task simply proved to be too much. First and foremost, the # of non-fatal workplace injuries is multiple orders of magnitude higher than the # of fatal injuries. Without sufficient manpower to pore over the details of each case, OSHA was unable to construct a complete census. But even if the manpower issue was resolved, OSHA encountered more intractable logistical challenges.
Take the issue of tallying up the # of workplace-related amputations as an example. You would think that this exercise would be easy because amputations seem like a pretty binary medical procedure: either the worker had to amputate a limb because they were injured at work or they didn’t. But alas, numerous challenges arose even in this theoretically optimal setting. Wiatrowski (2014) notes:
In some instances, amputations are not properly identified. Many cases identified in [the Survey of Occupational Injuries and Illnesses (SOII)], workers’ compensation, and elsewhere involved loss of tissue but not loss of bone, such as with the loss of a fingertip. In some of the research, it was not possible to separate such an injury, known as an avulsion, from bone-loss amputations.
Amputation is both a type of injury and a surgical procedure. Various coding schemes used by the data sources handled this distinction in different ways.
The timing of an amputation can lead to differential recording of the same event. For example, a worker whose finger is crushed at work may have the injury recorded on the OSHA log (and therefore in SOII) as a crushing; days or months later, infection or other complications may result in the finger being amputated, leading to a workers’ compensation claim for an amputation. OSHA does not require that case details be updated, meaning cases with delayed amputations would be reported in the SOII and workers’ compensation systems differently.
Ultimately, the multi-source enumeration project ended with the rather sobering conclusion that (1) OSHA was almost certainly under-counting the prevalence of non-fatal workplace injuries and illnesses (2) it would be very difficult to mitigate this issue without more fundamental changes to staffing and data collection practices.
In response to the above information, a curious reader might ask:
It’s definitely a shame that OSHA systematically under-counts workplace injuries and illnesses. But does this under-reporting actually matter in the context of a randomized controlled trial (RCT) or regression discontinuity design (RDD)? In both types of studies, the two sets of workplaces being compared should be very similar to one another across multiple dimensions including their rates of under-reporting. So does the under-reporting issue truly matter?
This is a great question, and the answer is unfortunately, yes. The crux of the issue is that, while the two sets of workplaces being compared should have similar rates of under-reporting before an inspection takes place, OSHA inspections may very well impact the probability of reporting injuries thereafter. More specifically:
OSHA inspections might be spuriously associated with increased injuries. During the course of an inspection, an inspector might discover that a company has failed to appropriately document workplace injuries and illnesses and issue that company a recordkeeping violation. To avoid the possibility of additional penalties, the company might then increase reporting of such issues to OSHA, leading us to falsely believe that OSHA inspections increase workplace injuries.
OSHA inspections might be spuriously associated with decreased injuries. While some companies may choose to comply with OSHA guidelines following an inspection, other companies might choose to decrease reporting of injuries and illnesses in the hopes of evading additional scrutiny, leading us to falsely believe that OSHA inspections decrease workplace injuries.
Researchers deal with these challenges in two ways:
Leverage multiple data sources. Rather than solely relying on data collected by OSHA to track workplace injuries and illnesses, one might choose to rely on data derived from worker’s compensation claims. Relying on this latter data source has the advantage that it consists of worker-reported data rather than company-reported data. There are two disadvantages to this approach, however:
Workers’ compensation databases serve an administratively different purpose than OSHA data sources. Violations that entitle a worker to workers’ compensation may not meet the criteria to be considered OSHA violations and vice versa. And in the case of contracted workers, the violation may be linked to different workplaces in each database: OSHA links the violation to the company whose worksite the contractor was actually located at whereas workers’ compensation is paid out by the agency with whom the contractor is formally employed.
There is ample evidence to indicate that workers’ compensation databases also under-count injuries and illnesses. As the multi-source enumeration project demonstrated, there is no singular data source which fully captures the # of workplace-related non-fatal injuries and illnesses.
Look at workplace fatalities. Just as criminologists consider homicide to be the most accurately tracked category of crime because it is easy to corroborate using multiple sources of documentation (e.g. death certificates, coroner’s reports, and law enforcement documentation), occupational health experts consider workplace fatalities to be the most accurately tracked category of workplace injuries for largely the same reasons. Thus, one way to deal with poor measurement is to simply assess whether the results hold up when analyzing workplace fatalities. The disadvantage of this approach, however, is that it greatly reduces statistical power. Fatal workplace injuries are much rarer than non-fatal ones, so it is considerably harder to detect a statistically significant decrease in said injuries.
Do OSHA Inspections Work?
Now that we understand what OSHA does and how to estimate its impact, we can finally answer the key question of interest: do OSHA inspections meaningfully improve workplace health and safety? The answer across multiple empirical strategies and types of workplace injuries is yes.
We can begin with randomized controlled trials (RCTs) which estimate the impact of OSHA inspections on workplace injuries in relative terms (i.e. percent reductions). These studies indicate that planned OSHA inspections reduce workplace injuries by approximately 9%.
This result is corroborated by randomized controlled trials (RCTs) and regression discontinuity designs (RDDs) which estimate the impact of OSHA inspections in absolute terms. These studies indicate that OSHA inspections reduce both non-fatal and fatal workplace injuries. This latter result is particularly important as it provides strong evidence that the estimated improvements in workplace safety are unlikely to be an artifact of poor measurement.
A necessary caveat, however, is that this particular study only analyzed OSHA inspections which took place during 1987 to 1997 and thus, may not generalize to present-day conditions. Nonetheless, the consistency of results across different time periods, outcomes, and methodologies is highly promising, reinforcing the common sense view that workplace safety inspections do, in fact, improve workplace safety.
Follow-Up Questions
While the above studies provide strong evidence that OSHA inspections meaningfully improve workers’ safety and health, a curious reader might have a number of additional questions about OSHA’s efficacy. This section is intended to answer these such questions.
Hidden Costs
Thus far, you’ve assumed that OSHA reducing the rate of workplace injuries is necessarily good for workers. However, this isn’t guaranteed to be the case. OSHA regulations increase compliance costs for businesses, and burdensome costs might actually hurt workers by reducing their wages. So what reason do we have to believe that OSHA improves the welfare of workers on net?
This is a great question and, as it turns out, one that is addressed by two studies estimating the impact of OSHA inspections through randomized controlled trials.
Lee and Taylor (2019) estimate the causal impact of OSHA inspections conducted between 1987 and 1997 on two outcomes: workplace fatalities and workers’ wages. Using these impact estimates, they then construct an estimate of the value of a statistical life (VSL). In plain language, they estimate how much each worker values their own life in dollars based on the pay cut they are willing to take to reduce their risk of a workplace fatality. The authors ultimately estimate that manufacturing workers value their own lives at roughly $8 -10 million dollars (in 2016 terms). Thus, workers are more than willing to take a pay cut to avoid the risk of workplace fatality.
But let us suppose that one does not trust this VSL estimate because it is based on OSHA inspections that took place in the 1990s. Well, Levine, Johnson, and Toffel (2012) estimated the impact of random OSHA inspections in the 2000s on company payroll and found little evidence of a harmful impact.3
Thus, regardless of which randomized experiment one trusts, I am quite skeptical of the idea that OSHA inspections ultimately do more harm to workers than good.
Inadequate Efficiency
The above statistical analyses merely show that OSHA is “better than literally nothing”. In statistical terms, the impact of OSHA inspections is significantly different from zero. But shouldn’t we expect even greater efficiency from an organization which spends hundreds of millions of dollars in taxpayer money to promote workplace health and safety? “Better than literally nothing” just doesn’t seem to cut it.
One way to assess OSHA’s efficiency is by comparing its budget to the dollar value of the lives saved by OSHA inspections. Should the latter value exceed the former, we would have strong evidence that OSHA does not merely “do better than nothing” but instead is highly efficient in its mission to promote worker health and safety.
To carry out this comparison, we can revisit the VSL estimate generated by Lee and Taylor (2019): namely, that manufacturing workers value their own lives at roughly $8 - $10 million in 2016 dollars. Note that OSHA’s budget request during fiscal year 2016 was $590 million. Thus, to generate a net improvement in worker welfare, OSHA would have needed to avert $590 million / $8 million = 74 fatalities in that year to completely recoup the cost of its budget. Is this plausible? I think so.
Using 1987-1997 data, Lee and Taylor (2019) estimate that each programmed OSHA inspection at a manufacturing plant averted 1.4 fatalities per 10,000 employees.
Using 2001-2010 data, the average randomly inspected OSHA establishment via the Site-Specific Targeting (SST) Program had roughly 150 employees.
Thus, the average OSHA inspection averts roughly 3 fatalities / 10,000 employees * 150 employees = 0.021 fatalities.
During 2001-2010, there were roughly 5,000 programmed inspections of manufacturing plants that took place each year via the Site-Specific Targeting (SST) Program.
Thus, OSHA averted at least 5,000 inspections * 0.021 fatalities / inspection = 105 fatalities that year, which is above the 74 fatalities threshold.
Obviously, this napkin math rests on a number of assumptions (e.g. the estimated impact of OSHA inspections in the 1990s generalizes to the estimated impact of OSHA inspections today) and it does not take into account any negative productivity impacts of OSHA inspections so I would not take my numbers here as gospel. Rather, I think they provide a useful sanity test of the idea that OSHA’s economic impact exceeds the size of its budget.
That being said, I certainly acknowledge that OSHA can do more to increase efficiency. For example, Johnson, Levine, and Toffel (2023) find that the impact of OSHA inspections on subsequent workplace injuries varies considerably across inspected workplaces. In plain language, inspections are much more effective in some workplaces than others.
They go on to state:
Establishments with the top 20 percent estimated treatment effects have substantially more historical injuries, … more predicted injuries, … more employees, are more likely to be in the manufacturing sector, are less likely to be nursing homes, experienced more fatal injuries two years prior to randomization, and had more prior inspections triggered by a worker complaint.
Thus, a targeting scheme which prioritizes these establishments may have even greater returns to workplace safety than the current procedure.
Existing research also indicates that public shaming is a likely a more cost-effective supplement to inspections in promoting workplace safety. Johnson (2020) provides an illustrative example. The author begins by noting that the probability that OSHA publishes a negative press release about a company rises sharply when OSHA recently issued that company a penalty above a particular threshold.
Using a regression discontinuity design, he then provides suggestive evidence that these press releases deter bad behavior among uninspected workplaces which happened to be in the same industry and nearby the workplace had been subjected to the press release.
Because press releases are both cost-effective and can achieve deterrence effects in uninspected workplaces, OSHA should continue to empirically explore what the right balance is between using press releases vs inspections to improve workplace safety.
Lastly (and more speculatively), it’s possible that Large Language Models (LLMs) can assist OSHA in increasing the quality of workplace injury datasets. Because LLMs can process vast volumes of text in a span of minutes, they could sift through local news articles to identify workplace safety issues that would have otherwise gone unreported to OSHA.
Burdensome Regulation
Even if OSHA regulations don’t have downstream negative impacts on the welfare of workers, they almost certainly negatively impact businesses. This is because, as stated before, such regulations increase compliance costs, and businesses have to eat that cost somehow whether it be through lowering wage/non-wage benefits to workers or simply producing fewer goods. Thus, isn’t there a very real risk that OSHA is over-regulating businesses?
I think there are two points worth noting here. First, I am not someone who is pro-regulation by default. Regulations are merely tools which can be used to achieve both positive and negative aims. Thus, I do not automatically believe that every regulation adopted by OSHA as justifiable, and this view is entirely consistent with the view that OSHA should not be abolished.
Second, while I agree that over-regulation of workplace safety comes at the expense of businesses, under-regulation of workplace safety often comes at the expense of taxpayers. To elaborate, note that, in most states, employers are expected to pay for workers’ compensation insurance which pays out medical and wage benefits to workers in the event that they are injured on the job. In other words, the government expects the employer to take on the costs associated with workplace safety hazards.
However, this expectation is not always met in the real world. For one thing, not all workplace injuries end up resulting in a successful workers’ compensation claim being filed. But even in the event that a workers’ compensation claim is successfully filed, injured workers often rely on welfare programs to supplement these benefits. For example, Ladd and Neumark (2024) estimate that around 50% of permanently disabled workers receiving workers’ compensation also receive welfare benefits through Social Security Disability Insurance (SSDI).
Thus, when OSHA is deciding to establish a new regulation, it is not merely choosing between a world where businesses face greater compliance costs vs a world in which they don’t. It is often choosing between a world where businesses face greater compliance costs vs a world where taxpayers subsidize the consequences of unsafe workplace environments. When presented in this way, I imagine most taxpayers would prefer that businesses fully internalize the costs of workplace safety.
Why Can’t We Just Send OSHA Money Back To The States?
I can certainly understand why an organization like OSHA needs to exist. But is there any particular reason why it has to exist at the federal level? What would be wrong if the current administration simply chose to re-distribute OSHA funding back to the states to support their own workplace safety initiatives?
Two points are worth noting at the outset:
OSHA already allocates a substantial portion of its budget to state-level occupational safety and health agencies. This is included in the “State Programs” component of the budget chart I showed previously.
There are currently 22 states which have established their own equivalents of OSHA which operate independently of the federal agency.
So to some extent, OSHA is already doing what this question proposes.
“But why not distribute all of OSHA’s budget back to the states?” one might ask. There are two challenges with this proposal, one of which I expect to resonate with workers and one of which I expect to resonate with business owners:
The Worker Answer: Currently, states can only operate independently of OSHA if OSHA has certified that their standards are at least as effective as the federal agency. Completely redistributing OSHA’s funding back to the states would likely eliminate this requirement, leading to large variation in workplace safety regulations that could easily come at the expense of workers in states with poor track records around labor rights.
The Business Owner Answer: Uniform workplace safety standards across states ultimately allow businesses to scale because business owners do not have to spend as much time worrying about whether workplace safety regulations at a particular worksite will generalize to worksites in other states.
Thus, I would argue that there is good reason for OSHA to remain a federal agency.
Conclusion
To recap, the best available evidence indicates that OSHA inspections meaningfully improve workers’ health and safety in high-risk occupations. Consequently, systematically shuttering OSHA outposts would be, as former OSHA Chief David Michaels put it, “penny-wise and pound-foolish”.
However, a curious reader might naturally wonder:
Given that the Trump administration has publicly stated that it wishes to “build things in America again”, why would the same administration be more than happy to slash OSHA’s funding? Wouldn’t such actions only harm the very manufacturing workers who’d be needed to actually build things domestically?
The answer to the second question is a clear yes. Even under the Biden administration, OSHA experienced considerable challenges in its mission to uphold workplace safety. For example:
The average time needed to establish a new safety standard was more than 7 years.
The average whistleblower investigation submitted to OSHA sat in limbo for 200 days.
Continued cuts to the agency would only end up exacerbating such issues.
As to why the Trump administration would carry out such actions, that is for the pundits to debate. It is worth noting, however, that the OSHA issue is emblematic of the type of double-think which has permeated the current administration more generally. For example, Trump administration officials have:
Claimed to revitalize the coal industry, while suspending black lung screening for coal miners.
Claimed to invest in nuclear energy, while planning to gut a key agency which distributes loans to build nuclear power plants.
Claimed to prioritize research into obesity, while pushing out one of the world’s leading obesity researchers.
Claimed to defend women from “gender ideology”, while defunding clinical trials designed to improve women’s health.
… and so on and so forth.
Thus, while the exact motivations behind the Trump policy agenda are beyond the scope of this article, I think any reasonable observer can see that there is an obvious disconnect between the purported policy goals of this administration and the policy actions which it has chosen to take.
Appendix
I sought to identify all studies which met the following inclusion criteria:
Estimated the impact of OSHA inspections on measures of workplace injury and/or illness rates such as:
DART: Workplace incidents causing days away from work, restricted job duties, or transfer to a new job.
DAFW: Workplace incidents causing days away from work.
Workers’ Compensation Claims: Workplace incidents causing a worker to file a workers’ compensation claim.
Fatalities: Workplace incidents causing death.
The empirical strategy was either a randomized controlled trial (RCT) or a regression discontinuity design (RDD).
For RDD’s specifically, the threshold in question had to substantially increase the probability that a workplace was inspected. In technical terms, the threshold should not be a “weak” instrument (i.e. F-statistic below 10).
The study estimated the actual impact of being inspected as opposed to the impact of merely being assigned to inspection. In technical terms, the study estimated the treatment-on-treated (TOT) effect, not just the intent-to-treat (ITT) effect.
I identified 5 studies which met these criteria. Importantly, it was not possible to meta-analyze these studies because impacts were sometimes measured in relative terms and sometimes in absolute terms. While it is trivial to convert relative point estimates to absolute point estimates and vice versa, it is unfortunately not trivial to do the same for the corresponding standard errors (which is a necessary step to conduct meta-analysis).
Works Cited
Baram, M. (2025). OSHA Office Closure in ‘Cancer Alley’ Raises Fears About Risks to Workers. The American Prospect. https://prospect.org/labor/2025-04-18-osha-office-closure-cancer-alley-louisiana/
Bendix, A., Fattah, M., & Edwards, E. (2025). RFK Jr.'s insistence that the government ignores chronic disease is misguided. NBC News. https://www.nbcnews.com/health/health-news/rfk-jr-says-government-ignores-chronic-disease-misguided-rcna190042
Biggs, A. (2025). Nullify Occupational Safety and Health Administration Act or the NOSHA Act. https://www.congress.gov/bill/119th-congress/house-bill/86
Department of Energy (2025). Secretary Wright Acts to “Unleash Golden Era of American Energy Dominance”. https://www.energy.gov/articles/secretary-wright-acts-unleash-golden-era-american-energy-dominance
Department of Government Efficiency (2025). https://doge.gov/savings
Department of Labor (2016). Congressional Budget Justification - Occupational Safety and Health Administration. https://www.dol.gov/sites/dolgov/files/general/budget/2016/CBJ-2016-V2-12.pdf
Department of Labor (2022). US Department of Labor begins rulemaking process to revise standards for occupational exposure to lead. OSHA National News Release. https://www.osha.gov/news/newsreleases/national/06282022
Department of Labor (2023). Site-Specific Targeting. https://www.osha.gov/sites/default/files/enforcement/directives/cpl-02-01-064.pdf
Department of Labor (2024a). Department of Labor takes critical step in heat safety rulemaking, continues heightened enforcement efforts, focuses on dangers to agricultural workers. OSHA National News Release. https://www.osha.gov/news/newsreleases/national/05082024-0
Department of Labor (2024b). Department of Labor announces final rule updating the Hazard Communication Standard to better protect workers, first responders. OSHA National News Release. https://www.osha.gov/news/newsreleases/national/05202024
Department of Labor (2024c). Occupational Safety and Health Administration (OSHA) Enforcement. https://www.osha.gov/enforcement/current-enforcement-summary
Department of Labor (2024d). Congressional Budget Justification - Occupational Safety and Health Administration. https://www.dol.gov/sites/dolgov/files/general/budget/2025/CBJ-2025-V2-12.pdf
Department of Labor. Voluntary Protection Programs. https://www.osha.gov/vpp
Department of Labor. Injury & Illness Recordkeeping Forms - 300, 300A, 301. https://www.osha.gov/recordkeeping/forms
Fagan, K. M., & Hodgson, M. J. (2017). Under-recording of work-related injuries and illnesses: An OSHA priority. Journal of Safety Research, 60, 79–83. https://doi.org/10.1016/j.jsr.2016.12.002
Gilmore, E. (2025). 2025 Annual Adjustments to OSHA Civil Penalties. https://www.osha.gov/memos/2025-01-07/2025-annual-adjustments-osha-civil-penalties
Government Accountability Office (2009). Enhancing OSHA’s Records Audit Process Could Improve the Accuracy of Worker Injury and Illness Data. (2009). https://www.gao.gov/products/gao-10-10
Hall, K. (2025). https://x.com/KevinH_PhD/status/1913739360614166667
Hochman, T. (2025). The LPO is Already Efficient. Green Tape. https://www.greentape.pub/p/the-lpo-is-already-efficient
Johnson, M. S. (2020). Regulation by Shaming: Deterrence Effects of Publicizing Violations of Workplace Safety and Health Laws. American Economic Review, 110(6), 1866–1904. https://doi.org/10.1257/aer.20180501
Johnson, M. S., Levine, D. I., & Toffel, M. W. (2023). Improving Regulatory Effectiveness through Better Targeting: Evidence from OSHA. American Economic Journal: Applied Economics, 15(4), 30–67. https://doi.org/10.1257/app.20200659
Ladd, D., & Neumark, D. (2024). Workplace Injuries and Receipt of Benefits from Workers Compensation and SSDI. Michigan Retirement and Disability Research Center. https://mrdrc.isr.umich.edu/wp-content/uploads/2024/04/rb424.pdf
Landsman, G. et al (2025). Letter to President Trump and Secretary Kennedy Regarding The National Institute for Occupational Safety and Health. https://deluzio.house.gov/sites/evo-subsites/deluzio.house.gov/files/evo-media-document/2025.04.04_national%2Binstitute%2Bof%2Boccupational%2Bsafety%2Band%2Bhealth-compressed.pdf
Lee, J. M., & Taylor, L. O. (2019). Randomized Safety Inspections and Risk Exposure on the Job: Quasi-experimental Estimates of the Value of a Statistical Life. American Economic Journal: Economic Policy, 11(4), 350–374. https://doi.org/10.1257/pol.20150024
Levine, D. I., Toffel, M. W., & Johnson, M. S. (2012). Randomized Government Safety Inspections Reduce Worker Injuries with No Detectable Job Loss. Science, 336(6083), 907–911. https://doi.org/10.1126/science.121519
Li, L., & Singleton, P. (2019). The Effect of Workplace Inspections on Worker Safety. ILR Review, 72(3), 718–748. https://doi.org/10.1177/0019793918801575
Michaels, D. (2025). Former OSHA Chief Dr. David Michaels: Closing OSHA Offices Will Increase Worker Injuries, Illnesses, Deaths. Industrial Safety and Hygiene News. https://www.ishn.com/articles/114675-former-osha-chief-dr-david-michaels-closing-osha-offices-will-increase-worker-injuries-illnesses-deaths
Peto, B., Hoesly, L., Cave, G., Kretch, D., & Dieterle, E. (2016). Evaluation of the Occupational Safety and Health Administration’s Site-Specific Targeting Program Final Report. Summit Consulting. https://www.dol.gov/sites/dolgov/files/OASP/legacy/files/SST_Evaluation_Final_Report.pdf
Phillips, Z. (2025). Bill to abolish OSHA has ‘zero chance’ of becoming law, attorney says. Facilities Dive. https://www.facilitiesdive.com/news/osha-abolish-zero-chance-trump-labor-changes/739684/
van Breen, J., Devarakonda, K., & Liem, M. (2023). Can Homicide Serve as an Indicator of Non-lethal Crime? A Systematic Literature Review. International Criminology. 3, 1-17. https://doi.org/10.1007/s43576-023-00086-1.
Volcovici, V. (2025). As Trump eyes coal revival, his job cuts hobble black lung protections for miners. Reuters. https://www.reuters.com/business/world-at-work/trump-eyes-coal-revival-his-job-cuts-hobble-black-lung-protections-miners-2025-04-21/
Wadman, M., Kaiser, J., & Reardon, S. (2025). NIH guts its first and largest study centered on women. Science Insider. https://www.science.org/content/article/nih-cancels-its-first-and-largest-study-centered-women
White House. (2025a). Reinvigorating America’s Beautiful Clean Coal Industry and Amending Executive Order 14241. https://www.whitehouse.gov/presidential-actions/2025/04/reinvigorating-americas-beautiful-clean-coal-industry-and-amending-executive-order-14241/
White House. (2025b). Defending Women From Gender Ideology Extremism and Restoring Biological Truth To The Federal Government. https://www.whitehouse.gov/presidential-actions/2025/01/defending-women-from-gender-ideology-extremism-and-restoring-biological-truth-to-the-federal-government/
Wiatrowski, W. (2014). Examining the completeness of occupational injury and illness data: An update on current research. Monthly Labor Review. https://doi.org/10.21916/mlr.2014.24
Willful/repeated violators may pay up to $165,514 per violation.
In practice, the actual process of selecting workplaces for inspection is a bit more sophisticated and would more accurately be described as conditionally random (i.e. conditional on industry, region, year, historical injury rate, etc., inspections are randomly determined).
I realize that this result contradicts Lee and Taylor (2019) because they do estimate a negative impact on workers’ wages. I can’t explain the discrepancy in results between the two papers, but what I can say is this: regardless of which randomized experiment one trusts, OSHA inspections are unlikely to harm worker welfare on net.
I appreciate the careful review of the empirical literature; I certainly learned much about the difficulties of estimating the causal effects of OSHA inspections despite believing that the final estimates are somewhat unconnected to the claim that "OSHA actually delivers on its promise to protect workers from threats to their safety and health." This is because the relevant counterfactual in the context of NOSHA is not between safety outcomes in the presence/absence of OSHA inspections where OSHA requirements are already common knowledge. Rather, it is between those attained under OSHA versus under the private occupational safety regimes that firms would adopt in its absence. Even this may be misleading insofar as current OSHA standards may constrain workers' and firms' flexibility in trading off occupational safety for higher wages; this possibility is surfaced in the response to the 'Hidden Costs' objection, but the results cited there again don't address the relevant counterfactual.
My takeaway then is to increase my appreciation for the difficulty of attaining ecological validity for the noteworthy proposals of the day. That is, the degree to which a bill is noteworthy depends on the breadth of effects it would have on society, in turn lowering the likelihood that the relevant variation already exists for estimating the causal effects of interest.
Interesting detail on data quality around occupational injuries... this potentially applies to work-related ill health more broadly (work-related mental ill health, for example). Organisational data quality can be poor, regardless of what is reported. Better data quality would likely benefit organisations directly as well as feed in to broader workplace health and safety initiatives.